Heh... our daughter was talking to herself in Japanese in her sleep last night.
http://www.kinokuniya.com/au/index.p...=9784906224500
Nothing beats just having conversations with people in Japanese (or whatever target language you're trying to learn) though. You learn tonnes more from speaking and listening than you would from reading piles of books. Speaking and listening are natural language skills -- when babies are born, they're instantly listening (some say they start listening while in utero; we were talking to our daughter and letting her listen to Mozart while she was in utero*) and later they will naturally try to talk. Simple utterances at first ("dada," "mama" etc.), then proper words, attempts to string sentences etc. But reading and writing are artificial skills. Less than 5% of civilisations invented writing. Literacy is a skill that must be instructed -- people will not naturally learn it (which is why you can have illiterate native speakers of languages. You need look no further than the internet to see grown adults who can't write properly in their native (and often only) tongue! ).
Having said that, literacy does massively improve one's ability to learn a language, so I wouldn't completely dismiss it either. When it comes to that, forget using Romaji -- just jump straight into Hiragana, Katakana and Kanji (in that order). I find learning Japanese with Romaji to be not as effective because learners often become too dependent on it, then they have to "ween" themselves off it when attempting to learn how to read/write Japanese script. I knew an American fella who lived in Japan for 8+ years -- spoke the language fluently, but was completely illiterate in it. Couldn't even read Hiragana... and all because he allowed himself to learn Japanese through Romaji first. I'm glad that I never allowed myself to become dependent on Romaji, and after meeting that guy, it convinced me that it certainly is a poor way to learn Japanese script.
----------------------------------------------------------
*While there's no definitive/conclusive evidence that this stuff actually works, we figured that it certainly wouldn't hurt.