(I typed a message similar to this yesterday, but I must have forgotten to actually post it.)
I haven't read "The Language Instinct" yet, but it's on my list. Meanwhile, the author (Steven Pinker) has another book that I have read called "How the Mind Works."
His view is that there are numerous autonomous agents that perform very specialized functions in the brain, very much the same as Marvin Minsky outlined in his much earlier "The Society of Mind." This view, predominant among cognitive scientists and AI experts, is crucially different than what many people intuitively believe about the brain. Previously people believed that the brain was a general purpose computer. (In fact, I've read a number of articles over the years that refered to the brain as a general purpose computer.) The new view is that it is actually a highly specially computer composed of a conglomeration of highly specialized, autonomous agents.
Heavily used neural pathways get stronger. In youth, new neural pathways can form (possibly in adulthood as well, but doubtfully so quickly).
We speak for the most part what we hear and what we read and what we have spoken and written previously. I imagine it's analogous to chess, but vastly more complicated. Players play speed chess to burn in the traps in the openings and to have a feel for the possibilities where certain paths will lead. Maybe they do a similar thing in, say, a creative writing course. I can imagine an exercise in such a class where a teacher presents a topic and the students have to think of as many ways as they can of describing the topic.
We can't even follow every path in advance, let alone memorize it. But we don't have to play chess or to write well. With a lot of practice we can gain an understanding of how certain words fit together and of where they might lead us. A good chess player, doesn't consider anywhere close to every possibility. A good writer doesn't consider every possible word. But some combinations (of movements or of words) are right at the tip.
Some people will have astonishing natural ability. They will play a relatively few games and become class B or A or maybe even expert players. It's doubtful that even the person with the greatest natural ability, though, could become a master, much less a grandmaster, without considerable study and practice. And some people will just have a knack for words, but will still have to practice to be really good.
At the other end of the spectrum, in either venue, some people will practice and practice and it will never come to them. They might become capable, but they won't become artists.
The vast majority of people will have a range of abilities that are neither very low nor very high. If they practice hard, they'll become highly proficient.
I think the different things that one does with words, the people with whom one converses, the topics that one discusses, the books one reads, the problems one addresses, the solutions one pursues, all of these things and probably much, much more will have some effect on one's diction, grammar, and overall style.
Things like repetition, reaching (stretching one's limits), failure (exceeding one's limits), reflection (hopefully understanding the points of failure) all increase the repertoire that one has at one's disposal.
I reckon there are agents that collect and classify, some that select and mask, some that organize, some that test, recognize, filter. I think as well there must be some kind of uber-agents that tell the other guys what to do by fiat or perhaps by voting. Agents and uber-agents and uber-uber-agents get stronger by getting used. And the roles of agent and uber-agent can be reversed and augmented, because agents are not arranged in a hierarchy, but in a tangled web.
Well, this is a vague, non-answer from someone reading at and beyond the periphery of his expertise. I'm guessing if we really knew how our own brains selected just the right words (or just as often, just the wrong words), we might expect that AI would have progressed a bit farther than what it has.
k
