In many ways, Steven Pinker's 1999 book Words and Rules picks up where the author left off in 1994's The Language Instinct. Once again, the topic is language and the human mind, only this time the MIT cognitive psychologist and linguist focuses his analysis on syntax (grammar) with only occasional forays into more general discussions of linguistics and cognition.
The book's title refers to a theory that the mental machinery devoted to language is made up of two interlinked and overlapping systems: one devoted to lexical items and their semantic meanings, and one devoted to syntactic, phonetic, phonological, and morphological transformations that shift words into different forms in regular ways. Morphosyntactic rules like "add '-s' to the end of a word to pluralize it" are governed by the latter system; grammatical exceptions like "the plural of 'child' is 'children'" are learned one at a time and stored in more of a mental dictionary. This theory combines the two extreme ends of syntactic theory.
At the "rules" extreme is a theory called generative morphology, which was given its definitive treatment in Noam Chomsky and Morris Halle's 1968 The Sound Pattern of English (revised in 1991 with linguist K.P. Mohanan). I've tried to read the book in question, and highly recommend Words and Rules's short summary, which presents all the major points of the Chomsky-Halle-Mohanan theory without any dumbing down or excessive reverence. As Pinker writes, "Any theory that can tame the quintessentially unruly English irregular past-tense system with only three rules, each delicately adjusting a single feature, is undeniably brilliant. But is it true? Not necessarily." (100) This example may sound familiar to Pinker afficionados, perhaps because Chapter 4 of The Language Instinct provided a similarly priceless translation of Chomsky's more well-known related theory of transformational generative grammar, as support for the idea that humans have an innate capacity for language use. But I digress.
At the other end of the syntactic theory spectrum are approaches based on artificial neural network models of the brain, and language in particular. The groundbreaking work in this area was published in 1968 by psychologists David Rumelhart and James McClelland. Their basic hypothesis was that language learning is a function of associative memory, which is to say that what Chomsky, Halle, and Mohanan would have us believe are "rules" are actually a product of our human capacity for finding patterns in data, and not hardwired into the brain in any way at all.
Pinker's compromise between these two extremes is a classic case of having your cake and eating it too: he posits that regular words are governed by rules, but irregulars are learned one at a time, pattern-associator-style. He provides evidence from other languages and the history of English in a chapter called "The Horrors of the German Language" which suggests that once upon a time the words we know as irregular verbs of English (strong verbs) were regular, but as new words were invented by borrowing, verbing nouns, and other word formation processes, the regular patterns became more prevalent. See weak verb and past tense of new verbs for more on this.
I liked Pinker's linguistic analysis a lot, especially the stuff about historical linguistics, which was how I justified buying this book in the first place ("It's related to thesis", I told myself, "really it is."). I was less enamored of his attempts to relate words and rules theory to general theories of cognition. Part of the problem was that I was reading as quickly as I could at that point (more on this in the next paragraph) so I probably missed a lot. The basic idea was that regular and irregular words are a lot like classical and family-resemblance categories, respectively. I was glad to see him incorporate George Lakoff's work in category theory, even if it means I have to really read Women, Fire, and Dangerous Things instead of just scanning the chapters assigned in my semantics class. The very end of Words and Rules is a little pat and dated, but that's outweighed by all the good stuff that precedes it.
As I mentioned in my Steven Pinker writeup, I own an autographed copy of this book: a souvenir from the author's appearance at the Claremont McKenna College Athenaeum on 19 February 2001. This review is based on that presentation and my recent reading of this book in an attempt to have it ready to lend to Igloowhite at the Portland neo-imperialist nodermeet of October 2002. Because both he and I are lazy slackers, the planned book exchange never took place: I didn't finish reading until the last day of the party, which he did not attend. Oops. Anyway, this writeup is dedicated to Igloowhite, as a consolation prize of sorts.
Pinker, Steven. Words and Rules. Perennial (an imprint of HarperCollins Publishers), 2000. ISBN 0-06-095840-5
Pinker, Steven. Lecture at the Claremont McKenna College Athenaeum, 19 February 2001.