OK here’s a cool short example of of how we pick up grammar.
1) The word “bung” in the Blablabian language means “I make love.”
2) The word “bungbung” in Blablabian means “I made love.”
3) The word “nom” means “I eat”
4) How do we say “I ate” in Blablabian? That’s right, class– “nomnom.”
What’s the rule for past-tense first-person formation in Blablabian? That’s right– say the present-tense first-person word twice in a row.
Kids do this with whatever language they are learning because (a) their brains get 1000s of hours of input (literally– kids get 4,000-4,500 hours of input before even single words emerge) which is supported with pointing, gestures, repetition, physically being picked up and moved, etc etc. and (b) what Chomsky calls “the language organ” neural networks know (it’s genetically wired-in) to induce the rules from the data.
Comprehensible input language teaching relies on one simple trick: instead of focusing the conscious mind on the rule, we focus the conscious mind on the meaning. And if we give people enough repetitions of quality input where they (a) understand, (b) are paying attention and (c) feel comfortable, they will pick the rule up. Indeed, they can’t help but pick the rule up. The unconscious “language organ” will figure the rule out and eventually start applying it, initially to decode input and then later in speech and writing.
We also throw in explanations of the rule– pop-ups, or what my colleague Adriana Ramirez calls “grammar commercials”– because some people like those (and a few may even need them). Some researchers now think (in contradiction to Stephen Krashen’s contentions) that there is a bit of “leakage” from the conscious mind to the unconscious, so explicit instruction will in some cases be helpful (especially with writing, and with older learners).
But, basically, 95% of all good language teaching work is nothing more than delivering repeatable, interesting comprehensible input in a safe, fun environment.