Review: The Language Instinct

I was referred to The Language Instinct by John McWhorter's highly recommended The Story of Human Language. The language instinct is a popular account of Noam Chomsky's theory of language, which is that we have a built in understanding of language grammars at birth, causing human languages to be built a certain way, and why certain new phrases "sound right", even when pedantic newspaper commentators and grammar textbooks indicate otherwise.

The argument comes in several forms, including one that was new to me, which is that children generalized from a paucity of exposure. If you come from machine learning/computer science, you're undoubtedly aware of the huge amount of data required to train a machine in statistical language comprehension. Yet even children in cultures where parents do not regularly speak to their children in childhood (a "no-no" in American/Western culture) pick up language from their peers and are able to generalize a grammar with no formal instruction.

There's a huge long section in the book full of parse trees and exploration of when the human's innate natural language parser breaks down, and how a computer wouldn't have any trouble with the stacking problem, yet cannot disambiguate sentences that humans have no problems with. The section is long, yet also serves to illustrate the point that humans don't think in any particular language, but have to translate thought into human languages through a mechanism. It's intriguing for me to think that there's a "machine language" of the brain, but obviously there's much work that's needed here.

There's a great defense of the idea that that language is born into humans. Pinker debunks all the famous stories about chimpanzees being taught sign language, and provides a plausible mechanism by which the language instinct could have developed evolutionary on a branch unique to humans.

This is a long book, but worth your time. Recommended.