Stephen Pinker is pushing an agenda, and a childish one at that. This may be why his claims here are such unadulterated nonsense. Note this: Pinker's own grammar is just fine, and rightly so: He's saying something more complicated than "gimme eat" or "lemme out". If he wrote incoherently he wouldn't be able to communicate his point at all.
". . . most of the 'ignorant errors' these rules are supposed to correct display an elegant logic and an acute sensitivity to the grammatical texture of the language, to which the mavens are oblivious."
Okay, there's the problem: He's living in a fantasy world populated by "noble savages". Here on E2 and elsewhere on planet Earth, most "bad grammar" is nothing like what Pinker describes. People with a crude grasp of the language don't communicate very much, nor very well. They throw words together at random and hope for the best. I doubt very much that Stephen Pinker wastes his time reading garbage written by fools. Instead, he reads prose written by people as well-educated as himself. He's not interested in condemning himself to this foolishness: He wants us to waste our lives deciphering ambiguous antecedents, malapropisms, and the like. He's got better things to do. I'm the guy who spends his days trying to pry coherent prose out of semi-literate kids on E2; Pinker has TA's for that.
Many of the "mavens" Pinker complains about are people like Vladimir Nabokov and William Safire -- people whose "acute sensitivity to . . . language" is real hard to miss if you're at all sensitive to the language yourself. "Oblivious"? Er . . . no.
Incidentally, Safire may well be a right wing maniac in his politics, but his writing about language is another matter: Safire is more often "descriptive" than "prescriptive", Pinker's disinformation to the contrary notwithstanding. Yes, I'm actually suggesting that you should read Safire's column (On Language, in the New York Times Magazine on Sundays) before deciding what he says in it. Scary, huh? It's good readin', though.
As for the "song of the humpback whale", that's nonsense too: If the humpback whale is attempting to communicate, it should damn well make noises that communicate effectively with other humpback whales. If if fails to communicate, it may as well have saved its breath. Language is not just random noise. It is a tool. For the "tool" to work, it has to be compatible with the "tools" everybody else is using. This is true for the humpback whale, it's true for Stephen Pinker, and it's true for the rest of us. You can express brutishly simple concepts by pointing and grunting, but that's all it's good for. When it comes time to say something complicated and interesting, a blunt intrument won't slice the tomato. You can call that the Wharfingerian Hypothesis, if you like.
Imagine a computer programmer who decides, with Stephen Pinker, that network protocols are arbitrary so it really doesn't matter whether his code is compatible with all the other computers out there. "Nobody defines correctness! TCP/IP is a tool of the oppressor! I'll make up my own protocol!" Okay, now imagine you paid good money for this crap, and now you want to visit E2: Sorry! Can't do that! Some idiot decided that network protocols are "a living language".
Language is not instinctive for humans. We're not chickadees building nests here: Read up about feral children (the forbidden experiment, wild children, Kaspar Hauser, wild boy of Aveyron): They never learn to express complicated concepts. They use arbitrary word order, contrary to Pinker's claims. Language is learned behavior, just like driving a car or cooking a meal. You can learn it well or learn it poorly.