I think going out into the world and talking to people (or even just chatting/posting on a forum, in writing) definitely helps. I don't like doing it, myself, early on (it makes me uncomfortable to suck at something in front of people, outside a classroom situation), but if you can handle it, it definitely helps. If you have access to a personal tutor, or a class, or even just a study-group or study partner, to converse with...that's brilliant. Having a competent tutor, to be your conversation partner, correct your errors, and provide you with custom made, up to date input (in the form of a personalized Anki deck, made just for you, for instance), would be the fastest way to learn a language...and, in that scenario, production would play a huge part in your learning process. But that's also prohibitively expensive for everyone not at least a millionaire.
What I don't think helps is doing production by trying to learn and apply grammar rules...in other words, doing production by composing your own sentences. That's not gonna work, because language isn't a structure built on grammar. It's the other way around: grammar is built on language. You can't construct a natural language based on grammar rules, you can only construct grammar rules based on a language (and even then, the rules are gonna be approximate at best).
So, sure, do production, but stick to saying things you've heard said before (producing collocations you know). Just to clarify, a collocation can be different things: it can be an exact cluster of words, like "to make an exception", or it can be the kind where some of the words are interchangeable, like "to enjoy a good book", "to enjoy a nice meal", etc. Once it becomes obvious that you can enjoy a wide variety of things, you don't need to have heard that verb used with every single thing. So it's not entirely as simple as "only say what you have heard said before". But you should stick pretty close to that, don't waste your time trying to infer your own version of Japanese from grammar rules you've studied, or patterns you've "noticed". Even in the examples above, you can enjoy a good book or a nice meal, but you can't enjoy a nice book, or a good meal, without sounding like a weirdo. Just let the patterns become internalized subconsciously, don't try to help the process along by learning or "noticing" them (I'm putting "notice" in quotes because you're not really noticing as much as guessing that there's a pattern there...when you're a beginner, you might think there's a pattern somewhere, only to run into half a dozen "exceptions" to your "pattern" in the next two days). So, when it comes to production, stick with what you know...you can guess from time to time (and, inevitably, make mistakes), that's fine, but once in a while, not every other word. You want to produce Japanese, not gibberish. Producing Japanese helps you get better at Japanese. Producing gibberish doesn't help you get better at anything.
As for doing exercises/Anki drills, I wouldn't do them for production too much. Maybe at the beginning, for really basic stuff, just to get you started with having basic conversations. But after that, I wouldn't bother. They help, sure, but the question is, how much do they help, compared to using Anki for input. I think that when you're using Anki for input (audio, or audio AND Kanji both in the question), you get through far more material. And that's the most important thing: to get through as much material as you can, as fast as you can, with the help of Anki (or textbooks, or whatever, I'm not saying SRS is the only valid language learning method), to then be able to let immersion solidify that knowledge.
There's a long road to listening comprehension, but there's an even longer road to fluency (the ability to fluently say the stuff you know). You want to use boring drills and exercises to get to listening comprehension, and then let immersion help you master the language. Because, if you want to drill your way to mastery, it's going to take thousands of hours of sitting in front of Anki. It's virtually impossible to do. On the other hand, spending those thousands of hours on immersion: very doable, and fun.
[disclaimer: a lot of the stuff I wrote about the importance of collocations, in the other thread, is the result of scientific research...most of the stuff I wrote in this post, on the other hand, is just my opinion, based on my own experiences...so it could definitely be wrong, personal experiences can be distorted].
What I don't think helps is doing production by trying to learn and apply grammar rules...in other words, doing production by composing your own sentences. That's not gonna work, because language isn't a structure built on grammar. It's the other way around: grammar is built on language. You can't construct a natural language based on grammar rules, you can only construct grammar rules based on a language (and even then, the rules are gonna be approximate at best).
So, sure, do production, but stick to saying things you've heard said before (producing collocations you know). Just to clarify, a collocation can be different things: it can be an exact cluster of words, like "to make an exception", or it can be the kind where some of the words are interchangeable, like "to enjoy a good book", "to enjoy a nice meal", etc. Once it becomes obvious that you can enjoy a wide variety of things, you don't need to have heard that verb used with every single thing. So it's not entirely as simple as "only say what you have heard said before". But you should stick pretty close to that, don't waste your time trying to infer your own version of Japanese from grammar rules you've studied, or patterns you've "noticed". Even in the examples above, you can enjoy a good book or a nice meal, but you can't enjoy a nice book, or a good meal, without sounding like a weirdo. Just let the patterns become internalized subconsciously, don't try to help the process along by learning or "noticing" them (I'm putting "notice" in quotes because you're not really noticing as much as guessing that there's a pattern there...when you're a beginner, you might think there's a pattern somewhere, only to run into half a dozen "exceptions" to your "pattern" in the next two days). So, when it comes to production, stick with what you know...you can guess from time to time (and, inevitably, make mistakes), that's fine, but once in a while, not every other word. You want to produce Japanese, not gibberish. Producing Japanese helps you get better at Japanese. Producing gibberish doesn't help you get better at anything.
As for doing exercises/Anki drills, I wouldn't do them for production too much. Maybe at the beginning, for really basic stuff, just to get you started with having basic conversations. But after that, I wouldn't bother. They help, sure, but the question is, how much do they help, compared to using Anki for input. I think that when you're using Anki for input (audio, or audio AND Kanji both in the question), you get through far more material. And that's the most important thing: to get through as much material as you can, as fast as you can, with the help of Anki (or textbooks, or whatever, I'm not saying SRS is the only valid language learning method), to then be able to let immersion solidify that knowledge.
There's a long road to listening comprehension, but there's an even longer road to fluency (the ability to fluently say the stuff you know). You want to use boring drills and exercises to get to listening comprehension, and then let immersion help you master the language. Because, if you want to drill your way to mastery, it's going to take thousands of hours of sitting in front of Anki. It's virtually impossible to do. On the other hand, spending those thousands of hours on immersion: very doable, and fun.
[disclaimer: a lot of the stuff I wrote about the importance of collocations, in the other thread, is the result of scientific research...most of the stuff I wrote in this post, on the other hand, is just my opinion, based on my own experiences...so it could definitely be wrong, personal experiences can be distorted].
Edited: 2016-04-11, 12:20 pm