This fascinating book delves into the structure of language, and the corresponding mental structures we presumably possess in order to design, use, and understand language. "Words and Rules," like Pinker's earlier book, "The Language Instinct" (also recommended), is easily readable by a non-expert, but nonetheless gives a rigorous explanation. I'll give "Words and Rules" a "+".
At issue is how language is organized in our minds-- do we use language by applying a series of rules (such as, "add -ed for most past tense verbs" in English), or do we store language in the form of many word lists (such as, "the past tense of 'drink' is 'drank'; the past tense of 'blink' is 'blinked', etc)? [Or, to use Pinker's definitions: "Words in the sense of memorized links between sound and meaning; rules in the sense of operations that assemble the words into combinations whose meaning can be computed from the meanings of the words and the way they are arranged"]. The answer seems to be that language is a complicated combination of both words and rules.
Much of the book is devoted to justifying this claim, which Pinker notes is not supported by all linguists. He gives many examples of other theories of language structure, then describes their shortcomings, for each fails in some cases. His hybrid words-and-rules theory holds up quite well. It fits the way children learn langauge (and make early mistakes) and the way patients with certain types of brain damage misuse language. For instance, people with Alzheimer's disease lose access to their words lists, while those with Parkinson's disease have difficulty applying rules.
Here's a simplified version of what seems to happen in a normal adult brain (Pinker gives more detail in the book): Some linguistic forms are stored in memorized lists-- these tend to be irregular verb and noun forms, such as dig-dug, strike-struck, and child-children. When you go to form a sentence, your brain looks for forms on these lists. If it comes up empty, then a default rule is applied-- the past tense is made by adding -ed, or the plural is made by adding -s. All of this occurs in a split second, of course.
The beauty of the combination words-and-rules system is that it can optimize brain function better than a system relying only on words or only on rules. If you had to remember every single form separately, that would take a lot of space-- obviously it is more efficient to have some general rules for how to relate the forms to each other (such as "add -ed for the past tense"). So why not use only rules? This turns out to be inefficient also. Not only does this require more computation time, but it's actually difficult to make a consistent set of rules that can capture all the words we need to express our complex thoughts. We know because people have tried to devise "logical" languages that rely on precise rules. They are a disaster. They quickly become unwieldy with literally hundreds of rules. And this creates a new problem: you have to memorize all those rules! Might as well go back to memorizing lists of words, right?
So the best compromise for our brains seems to be retaining some irregular forms as words lists, and using some rules (perhaps less than ten) for regular inflections. That's how we ended up with such a mishmash language.
But it's even more interesting than that. There is some rhyme and reason to how we inflect even the irregular forms. It's not just a random choice; the way the word is pronounced, the regular forms it sounds like, and the historical development of the word all influence the particular irregular inflection we give it. With some careful study, it is possible to explain why we say drink-drank, but blink-blinked.
And furthermore, regular and irregular forms are not set in stone. There are mechanisms in language evolution that allow one to transform itself into the other. Most irregular forms are relics of history-- they used to come from applying a rule, but because of shifts in pronunciation or vocabulary, the rule became less useful, and was eventually not recognized as a rule by a new generation of children. Hence, the forms were simply memorized. By this mechanism, regular forms can be irregularized over time. There are other methods of winding up with irregular forms, too, such as combining several verbs into one. This happened with go-went, which used to be two separate verbs. We just kept the present form of one and the past form of the other when we combined them.
Regular forms don't have much history, precisely because they are generated on the fly (you don't have a list in your head with the word "blinked" on it; you literally create this form each time you use it, then forget it again). But if an irregular form sounds like a regular one, the regular form may "co-opt" it, thus regularizing it. Also, unusual words are most likely to be regularized because we forget our lists if we don't use them enough. This happened with the past tense of "chide," which used to be "chid." Now we use "chided." Rememeber, when the irregular form fails to pop up from the list, the regular rule is applied. So if a word happens to be used less and less frequently, it will become regularized over time.
(This neatly explains, by the way, why irregular forms tend to be the most common words. Go-went, am-is-are, do-did, have-had, say-said: these are all very common verbs, and they are all irregular. Uncommon verbs like avow and truncate are unfailingly regular. And newly invented words, which we can't possibly have on our lists, tend to be made into regular forms by default).
Pinker also applies this nifty theory to other languages. There is a fascinating chapter on German, which I appreciated since I also speak German (but even non-speakers will get something out of it). It happens that German is a language where the "regular" form is not the most common! In English, it's easy to say that the past-tense "-ed" ending is "regular" simply because it is used so often. But that is not Pinker's definition of "regular." His idea is that the regular form is the one we use as the rule, while all other forms are memorized as words. By testing German speakers with nonsense words, he teased out the actual rule that they apply to generate regular forms, and it is not the most common inflection. This provides evidence for the mental organization he is talking about-- it exists even beyond our conscious knowledge of which words are used most frequently.
However, I was left with a few questions about foreign languages. First, I also study Japanese. This amazing language has just two truly irregular verbs. All other verbs fall into two perfectly inflected categories, for which knowing the verb stem tells you all the forms. (OK, OK, there are a few exceptions even here, but nothing as pronounced as strike-struck or ride-rode). Pinker touches briefly on Asian languages with a single mention of Chinese. But I was left hungering for a way to understand Japanese verbs. I guess that was beyond the scope of the book.
But second, the more I thought about foreign languages, the more I wondered whether Pinker's model is really an optimization issue like I described above. After all, languages range from Japanese-- with just two irregular verbs-- all the way to German-- with far more exceptions than rules. If the combination of "words and rules" is really acting to optimize brain function, why is there such a wide range seen among the languages of the world? It seems that the brain works just fine for a huge variety of linguistic structures.
I was convinced by Pinker's thorough research that he's onto something with the idea of having some words stored on lists, and then applying default rules for cases not covered by the lists. All his results support this idea. But I'm not sure he's figured out exactly why it works this way.
Anyway, Pinker also relates some of these language ideas to brain organization in an intriguing speculative section at the end. He talks about the way we "categorize" the world around us into "family groups" (like "birds", where we have a clear idea of what isn't included, but no exact definition of what is) and "Aristotelian categories" that are more clearly defined, such as "grandmother." He points out that much tension arises from the fuzzy borders around family groups conflicting with the sharp edges of clear categories. This is likened to the tension between words and rules, from which arises much of the richness of language.
As noted above, this book is a fabulous journey through modern linguistics theory for the non-specialist. Pinker doesn't oversimplify or talk down, but gives a fully comprehensible analysis of one of the most interesting parts of human cognition. This is a fairly young field, and it's fun to catch it near the beginning. There will be a chance to watch it evolve over the coming decades.
Copyright © Kim Allen 2001
