Words and Rules, The Ingredients of Language is a truly great book about language and linguistics. Steven Pinker, as a professor of psychology and director of the Center for Cognitive Neuroscience at MIT, a courageous advocate of free speech, and husband of philosopher Rebecca Newberger Goldstein, happily can be taken to represent the best in contemporary linguistics and cognitive science. An earlier book, The Language Instinct, How the Mind Creates Language [William Morrow and Company, 1994], was already an excellent introduction to the best of recent thinking and discoveries about language, but Words and Rules is something a little different, both an accessible presentation of its results and a thorough argument for the thesis productive of those results. This is a genuine scientific monograph that is nevertheless discursive and intelligible to the non-specialist.
The thesis is a simple one, that grammar involves both rote learning ("words") and innate rules ("rules"). As a disciple of Noam Chomsky, it is not surprising that Pinker subscribes to the "universal grammar" of innate rules. But Pinker is a critical disciple of Chomsky and qualifies his support. Whereas Chomsky had simply announced that the Rationalists were right in the classical philosophical debate about knowledge, Pinker allows that both Rationalists and Empiricists were right: Grammar has an innate component and an empirical component. This sensible compromise would be no surprise to Kant, who himself was neither pure Rationalist nor pure Empiricist. Kant famously said, "Thoughts without content are empty; intuitions without concepts are blind" [Critique of Pure Reason, first edition p.51]. Pinker's thesis is thus of great interest here, as tending to support a Kantian construction of knowledge.
Now, Chomsky would certainly be enough of an empiricist to allow that words are learned by rote. But Pinker's argument goes well beyond that. Much of the book is taken up with an examination of the difference between regular and irregular verbs. While both kinds of verbs are learned by rote as words, most of the inflection and usage of irregular verbs must also be learned by rote. That is why they are called "irregular," i.e. "without rule." The Chomskian purist doesn't like this. There must be rules behind the irregular verbs, rules, as Pinker says, "all the way down." The strict empiricist, of course, wants the grammar of regular verbs to be learned by rote also. Pinker carefully deals with the arguments and evidence for both of these angles, disposing of them decisively. Chomsky's attempts to explain everything with rules are clever but inherently unbelievable. At the same time, the phenomenon of children learning irregular verb forms and then, for a while, switching to improper regular forms for them, which were never heard or taught [p.193], is a striking example of their emergent skill at applying rules across the board. Once the rule-applying faculty is active, special practice is required not to apply it universally.
The most intriguing question, however, is Why? Why do we have verb (or noun) systems with both regular and irregular forms? A clue comes from the fact that irregular verbs are often the most frequently occurring verbs, like "be" in languages like English and Greek. Frequently used words, like frequently used tools, you might want to have close to hand; and it turns out that the irregular forms can be produced faster than the regular forms. This appears to mean that they can simply be remembered, rather than generated by the processes of regular word formation, which take slightly longer and even, as some of Pinker's evidence suggests, in different parts of the brain. Remembering a lot of irregular parts of verbs is more taxing on memory, but it would be worth it if this makes it easier to call on them in frequent use. Verbs rarely used, however, will be stored more conveniently and efficiently if only one form need to be remembered and all the others can be generated by rules. If this is so, then irregular words that are passing out of common usage would tend to become regular. This seems to be the case. Thus, we have a classic cost-benefit analysis in which the cost of memory and the cost of processing are balanced against the benefit of access and the benefit of efficient storage. Where access is important, the cost of memory is paid; and where efficiencies of memory are important, the cost of processing is paid.
One of the great philosophical debates that intersects the argument of the book is Wittgenstein's thesis that concepts are not about crisply defined Aristotelian classes but are fuzzy groupings where "family resemblances" bring together things that no clear rule can include in the class. This view also appears in linguistics and cognitive science.
In his book Women, Fire, and Dangerous Things (a family resemblance category in an Australian aboriginal language), the linguist George Lukoff called attention to the fuzziness that lies at the heart of that traditional bastion of rules, grammar. He cited irregular verbs as the ultimate proof of the bankruptcy of the two-thousand-year-old Aristotelian tradition in Western thought that seeks precise definitions for everything in sight.
But Lakoff did not notice that right next door to the irregulars are the regular verbs, and they pass all the tests of classical categories. [p.277]
The human mind deals in both the fuzzy and the precise, which should not be surprising.
The facts about verbs and the facts about concepts converge to suggest that the human mind is a hybrid system, learning fuzzy associations and crisp rules in different subsystems... Some modelers even link the rule system to the frontal cortex and the exemplar-based system to the temporal and posterior cortex, much as we did for rules and words in the preceding chapter. [p.279]
What Pinker does not do is consider the larger philosophical implications of this. It is not just that Wittgenstein is an empiricist, he is a nominalist. To him, there are only fuzzy "family resemblances" because there is no real identity between abstract features in the world, and that is because there are no real universals to be thus identical. A German Shepherd may be more similar to a Dobberman than to a Siamese cat, but there is no real thing that they both actually share as dogs. We just lump them together for our own purposes on the basis of our own vague intuitions.
Wittgenstein cannot be entirely consistent about this. He decided that languages are games "because they are played." This is interesting because it implies that "being played" is the defining abstract feature, the Aristotelian "specific difference," of games. The problems with that, of course, are (1) Wittgenstein did not think there were "specific differences," only family resemblances, (2) some games, like the Olympic Games, are not in fact, to my recollection, "played" ("held" is more like it), and (3) languages are not "played" either, in ordinary language, they are "used" or "spoken." But an important part of Wittgenstein's nominalism is his notion of languages as non-representational "games" that we play. We are not knowing the world, just playing. This is important enough that it becomes essential, in a fully Aristotelian sense, that languages really be games and be played, despite the incongruity of this with the rest of the theory.
Pinker, and even Chomsky, in fact could still be nominalists despite their belief in an innate rule directed faculty. They could simply subscribe to psychologism, that the mind imposes its own structures that are unrelated to the world, or to conceptualism, that universals are such structures in the mind, rather than realism, where real universals and real abstact identities exist among things (Chomsky, indeed, appears to be a conceptualist). Because Pinker does not make the differences, or his own position, clear -- and really, as a linguist or scientist, he doesn't need to -- his discussion of Rationalism and Empiricism is a little odd. Thus, he says, "The idea that intelligence arises from the manipulation of symbols by rules is a major doctrine of the school of thought called rationalism, generally associated with Leibniz and Descartes" [p.88].
However, although this trait may be basic to some rationalism, it is not essential to all of it, nor is it exclusive to rationalism. Rationalism can also be intuitionistic, like Plato and Husserl, as well as formalistic -- so we would have, not "rules all the way down," but "rules down to intuitions." Indeed, Aristotle limits the scope of logical rules and leaves the truth of first principles to be apprehended by intuition. On the other hand, an empiricist and nominalist can say "intelligence arises from the manipulation of symbols by rules" by adopting logicism, the view that intelligence does consist of the manipulation of rules but that we have just made these up, treat them by convention, and elaborate them as we wish. Thus Leibniz, although certainly a Rationalist himself, has inspired generations of empiricists and nominalists, like the Logical Positivists, who can extract a logicism from its rationalistic and realistic context in Leibniz. The early Wittgenstein, like his English discoverer, Bertrand Russell, was more in this camp.
Similarly, Pinker doesn't get the Empricists quite right. Thus, he quotes Hume as saying that there are only "three principles of connection among ideas, namely, resemblance, contiguity in time or place, and cause or effect," and goes on to say:
In this passage from his 1748 Enquiry Concerning Human Understanding, David Hume summarizes the theory of associationism, a major tenet of the school of thought called empiricism. The mind connects things that are experienced together or that look alike (Hume later eliminated cause and effect as a separate principle) and generalized to new objects according to their resemblance to known ones. [p.89]
The statement of interest here is that "Hume later eliminated cause and effect as a separate principle." Pinker has made a mistake, for this is simply not true. We can tell why when it is noted that judgments based on resemblance and contiguity are about "relations of ideas," which are matters of certainty, while on the other hand, as Hume says, all judgments about "matters of fact" are based on cause and effect and are not matters of certainty. There is therefore more to cause and effect than resemblance and contiguity, and this is what introduces the uncertainty into judgments about matters of fact. It is resemblance that enables us to associate one cause with another cause, and one effect with another effect, but the connection between the cause and the effect is something else. Although cause and effect always occur in contiguity with each other, what makes them cause and effect is not contiguity but necessary connection. Since a necessary connection is never "discovered" in experience, it is supplied, psychologistically, by human custom and habit. Given cause and effect, Hume runs with it, ruling out chance, free will, and miracles as all violating the principle that everything has a cause.
Pinker, indeed, could easily assimilate his own thought to Hume, since Hume believed that our knowledge of cause and effect, in owing nothing to reason and only imperfectly based on experience, was actually due to a kind of instinct, with which we are wisely equipped by nature. Pinker's "language instinct" could as easily be something of the sort. Nothing in evidence would prevent Pinker from taking this up and adopted all the rest of Hume's nominalism and scepticism.
Thus, while Words and Rules is a splendid treatise of linguistics, and important and suggestive for a Kantian epistemology, Pinker's own epistemology and ontology are philosophically underdetermined, and confused when it comes to the characteristics of schools like Rationalism and Empricism. The real philosophical significance of his work, therefore, must be inferred rather than found in his book.
Philosophy of Science
A Letter to The Wall Street Journal, "Thus Fakes Zarathustra," about a column by Robert P. Crease against Steven Pinker