The New Republic

The Limits of Language

By THOMAS NAGEL
Issue date: 01.24.00
Post date: 01.13.00

Words and Rules:
The Ingredients of Language

by Stephen Pinker
Basic Books, 348 pp.


What makes language such a happy object of study is that it exists in all of us, and each of us can use himself as a source of data to confirm or disconfirm general theories about it. In this excellent work of popular science, Steven Pinker introduces contemporary psycholinguistics through the special topic of the English past tense, and the distinction between regular and irregular verbs.

"He lied as he lay down beside her" is the past tense of "He lies as he lies down beside her." The irregulars, apart from the highly individual be and go, tend to fall into patterns. Shrink, shrank, [have] shrunk goes with drink, sink, stink; sleep, slept, slept with keep, creep, weep. There is blow, blew, blown; find, found, found; wear, wore, worn; hit, hit, hit; and so on. But they have to be remembered individually. The regulars follow the rule "add -ed"--resulting in three different pronunciations, as in lived, licked, and listed, also governed by strict rules of phonology. The -ed rule takes over automatically when not blocked by an entrenched alternative, so it works for most verbs in the language, including new coinages, as in Borked and faxed, or verbs formed from nouns, as in "Shouts rang out from the crowd that ringed the speaker," or "The batter flied out to left field."

The irregular verbs are among those with the highest frequency in the language, which explains why they are remembered. Once an irregular form falls into infrequent use, it tends to be replaced by the regular, as crew has given way to crowed. The balance between the two mechanisms--individual memorization and application of the rule--is intriguingly illustrated by the so-called U-shaped learning curve followed by children. At first they acquire verb forms individually by memory, regulars and irregulars alike. Then, as they get the hang of the regular rule in their third year, they tend to overapply it, producing errors such as goed and sended that they didn't make earlier. Only after a few more years are these errors subdued by the force of irregular forms on memory. As Pinker says, "Those who cannot remember the past are condemned to compute it."

Pinker also has some discussion of plurals and related matters (such as why we say still lifes rather than still lives). And he tells us a lot about how the history of the language accounts for many of its properties, including the descent of many irregulars from earlier rules that are now defunct. For someone like myself, who becomes apoplectic upon encountering criteria and data treated as singulars, or kudos as the plural of kudo, this book is strangely soothing. Pinker is not immune to such feelings--he says he was barely able to contain himself while listening to a speaker refer repeatedly to the "alumnis" of his university--but he takes the long view. He quotes Jonathan Swift fulminating against the barbarians who in his day were not treating -ed as a separate syllable in words such as disturbed and rebuked. The fact that language joins ferocious normativity with radical contingency should encourage a certain relativism, and even tolerance, in all of us, even as we stick to our guns. Here, as in so many respects, civilization is in perpetual decline but is at the same time being perpetually reborn.

The problem with Pinker's book is that it takes itself too seriously. Much of it revolves around a controversy among psychologists of language that outsiders will find boring--between a radical associationist view of the learning of grammar and the view, favored by Pinker, that the identification of formal rules with deductive consequences plays a large part in the process. Associationists try to explain everything in terms of phonetic or sometimes semantic similarities and dissimilarities among words. I have no reason to believe that Pinker has not presented the opposition in its best light, but it certainly seems like a pretty hopeless position in advance; he seems to be spending more ammunition than needed to persuade us that regular verbs provide a counterexample. Associationism is an old and bad idea about how the mind works in general, but it has recently been revived in the "connectionist" model of David Rumelhart and James McClelland, complete with computer simulations, and Pinker spends endless pages on its refutation. He offers the charming but dispiriting excuse that "as an experimental psychologist I have been trained not to believe anything unless it can be demonstrated in the laboratory on rats or sophomores."

It requires only a little reflection, however, to see that even though the mind operates partly by perceptions of similarity and difference, this cannot account for its capacity to apply precise concepts that subsume radically different cases, and to derive conclusions from them that mere similarity would never reveal. Mathematical and geometrical concepts are particularly clear examples, but many forms of mental operation in humans depend on a deductive logical structure. REliance on the more primitive mechanism of association was one of the things that made B.F. Skinner's behavioristic reinforcement theory so limited as a way of understanding human beings, especially their use of language.

Pinker certainly seems to establish that the regular English past tense, and various other grammatical rules, are learned not simply by association but through identification of a logical structure. Yet he greatly exaggerates the significance of this dispute by linking it to the venerable conflict between rationalism and empiricism: Descartes and Leibniz versus Locke and Hume. It is true that purely psychological hypotheses were part of what divided those figures of the past, but their more fundamental difference was epistemological: on what, if anything, can we base our claims to knowledge about the world? If, as seems clearly true of modern science, those claims go beyond our experiences, in what way are they nevertheless supported by those experiences? What additional contribution from a priori reason or understanding is needed to get beyond experience, and is such a thing available?

The dispute over the relative importance of rule-detection and memorization in the learning of grammar has nothing to do with this issue. The reason is that it has nothing to do with epistemology--that is, with the question of how our beliefs about the world can be justified. The truth about language depends on how people speak, but the truth about the natural world does not depend on what people believe. So there is a problem about the validity of our methods of belief formation that does not exist with regard to our methods of language learning: the problem is not just whether those methods are associative or logical, but whether they lead us to the truth.

The nature of the inborn, unlearned contribution to the process by which linguistic skills are acquired is of great psychological interest in itself, but it has no more implications for the justification of our knowledge of physics or biochemistry than do the nest-building instincts of the robin. The impulse to present empirical findings in psychology as the solutions to basic philosophical problems may be an occupational hazard of the field, but it is no more excusable than the urge of sociobiologists to offer moral insight.

Learning a grammatical system appears to depend on a distinct capacity, rather than on the application of a more general learning ability, associative or logical, to the special case of language. Evidence of this comes not only from the speed with which very small children accomplish the task, automatically and without any conscious formulation of the complex rules that they learn to obey. It comes also from the separability of this skill from other skills, through genetic deficit. There is a rare genetic defect called Williams Syndrome, whose victims stay in most respects at a mental age of around six or seven, but whose speech is extremely fluent, grammatically sophisticated, and rich in adult vocabulary. And there is a different genetic defect that results in great difficulty in learning to speak grammatically, but without comparable effects on other aspects of intelligence: such people have to memorize verb forms one by one. These data supplement the long-standing evidence for the complexity and the specificity of the language faculty from the different forms of aphasia caused by injury to different regions of the brain.

All work in this area stems from or responds to the great contributions of Noam Chomsky, who posed the basic question of the size and the specificity of the innate contribution to language learning more than forty years ago. Since any normal human infant raised in the appropriate environment can, remarkably, learn any language whatever with ease, this innate contribution, whatever it is, must be universal. Chomsky proposed that there should be deeply buried universal features common to all languages, features that are in some sense correlative with this universal faculty, permitting any child to find the key when exposed to an example.

Pinker makes a stab at identifying a linguistic universal toward the end of his book, but it is only the phenomenon of a "default rule," governing all cases not governed by more specific, individually memorized formations. The regular past tense is such a default rule in English. In Chinese, which is not inflected, Pinker identifies a default rule that supplies an all-purpose classifier, or a measure word, to go with nouns that are not assigned a specific classifier that has to be memorized. In English, we say "a piece of fruit," and "piece" is a classifier, but we do not need one with every noun. In Chinese, by contrast, a classifier is obligatory, and the default, Pinker reports, is ge. Fine, but it is hard to see this as the kind of deep similarity between the languages that would help explain why little children can pick them both up with ease.

Steven Pinker is a masterful explainer, a great collector of amusing examples, and smart. He ought to recognize that his subject is interesting enough in its own right, and that it does not need to be bolstered with inflated claims of deeper significance.


THOMAS NAGEL is the author, most recently, of The Last Word (Oxford University Press).



(Copyright 1999, The New Republic)