The crucial end product of much teaching is that students should ‘know’ language in an unconscious sense.
Reciting rules simply does not cut it. The ability to consciously refer to a grammatical rule does not necessarily help a learner produce language because when producing language, there is simply too much to have to consciously think about. Working memory can hold maybe 7 items. Learning a language has to involve converting knowledge from conscious to automatic processing.
If learners are expected to synthesize everything they know about language when speaking or writing then it’s an awful lot to remember.
Working memory is thought to be limited to seven chunks of information. (Miller’s Law: number of objects an average human can hold in working memory is 7 (+ or – 2), and this may only be 5 for words, and is even less for young children who can maybe only hold three items at once). Now when speaking or writing, learners have also got to think about meaning, so this will use up some of the available slots in working memory. There won’t be much left for grammar!
We often see how learners will get everything correct when doing a focused task on a particular language item, but then make errors on this same language item a few minutes later in speech and writing. This is essentially because production of the language item has not become automatic yet – they have to consciously think about the language item to get it right. If you draw their attention to the error they made, they are usually able to correct themselves, but only because you’ve forced the language item back into working memory.
Therefore, learning a second language is essentially a process of moving from conscious processing to automatic processing.