How so? Can you back this up?

Check out Rules and Representations by Noam Chomsky. One of his premises, for which he makes a fairly strong argument, is that at least some basic linguistic structures probably relate to human brain structure. He uses language acquisition in children as part of his backing.. the information set they're exposed to is unusually sparse for the amount of structure they pick up. IIRC, simulations with neural nets tend to support that premise. A raw neural net doesn't pick up language patterns nearly as quickly as a child does, but you can speed things up by 'predisposing' a net to certain types of patterns.

It was a bit cavalier of me to include such a statement without going into detail, but at least I didn't pull it straight out of the air.