If we assumed the existence of a language processing mechanism as described in the previous post and grammar were reduced to nothing but a collection of residual records of how this computational system goes about combining words, the question remains as to how adult language learners should go about acquiring a new language/building a new mechanism. Are adults even able to reverse engineer a brand new language processing unit (LPU)? Are all adult language learners instead building a translation mechanism? Is translation a scaffold towards building the new mechanism?
Now, we know that words have partcular properties. Do we try to discover these properties by ourselves, or should we look at descriptions (grammar)? A description would contribute towards the speed of language acquisition. It would also reinforce translation. Memory decay and constant refreshment through native examples should greatly attenuate this effect. There is no guarantee that trying to soak up the language through exposure would not result in subconscious (or even conscious) translation.
So, the use of aids should be deemed beneficial. How should we use these aids? From the above, I would assume that the best approach would be to dive into the text and simply look up the meanings. The properties of words should be inferred. I don't see why a grammar could not also be consulted, albeit sparingly, as a reference book. So, consult it, don't study it. Do this until you can get the gist of things and later whenever the meaning is absolutely crucial to understanding the gist. Or don't. My main reason for not using the dictionary is that it's six feet away, under the table. And it's heavy.
Breaking into a new language language, with or without the use of aids is the first step. Even high levels of comprehension however do not guarantee adequate production. Language use may guarantee automatization but not necessarily native-like fluency. This disconnect seems to confirm the existence of a very complex mechanism.