‘Words and Rules’: An Exchange

John Searle can be an acute thinker, and has given us insightful analyses of such topics as the nature of speech acts and the construction of social reality. But when it comes to people who don’t share his convictions, he is apt to substitute condescension for analysis. As his fellow philosopher Dan Dennett once wrote, “Someone less self-confident might reason: ‘I must be missing something; these colleagues of mine are coming out too stupid for words!’ But if this occurs to Searle, it does not prompt any serious consideration by him.” Searle’s review of my book Words and Rules [NYR, March 14] continues in this vein, offering one misrepresentation after another.

Words and Rules explores the human ability to extract meaning from words and their combinations. We understand that Bush Defeats Gore conveys a different message from Gore Defeats Bush because we have neural circuitry that can distinguish the arrangements of the words. Searle reads this as saying that words and rules are “sufficient” for people to understand language, without using any background knowledge—that one can fully understand the sentence without knowing, say, what an electoral defeat is. Nowhere did I assert or let slip such a ridiculous belief, and elsewhere I have strenuously explained why it cannot be correct.¹ Searle has confused my claim that words and rules are necessary for understanding with the claim that they are sufficient.

I also note that many other linguistic forms cannot be interpreted via the arrangements of their parts but have to be memorized whole. They include the vocabulary of a language and its stock of irregular forms such as blow–blew and foot–feet. Searle claims I “neglect” historical evidence that “today’s irregular verbs are historical survivors of what were once systematic rules.” In fact an entire chapter (Chapter Three) was devoted to this very point: in the introduction, I wrote, “Because irregulars originated from rules they are not a random grab-bag but display patterns, fossils of the long-dead rules.” Even worse, Searle uses the existence of such rules to claim that I “mistakenly suppose that the irregular cases have to be memorized on a case-by-case basis.” Does he really believe that children don’t have to memorize feet as the plural of foot because they can deduce it from the umlaut rule of proto-Germanic, spoken two millennia ago? And how could they, even in principle, given that the rule was triggered by the plural suffix –i, which is no longer found on the word? One of the best-established causes of language change is that speakers sometimes cease to apply a rule productively and memorize its outputs instead. This, too, is a recurring theme of the book, which Searle apparently missed.

Regular and irregular forms have recently attracted much interest because they bear on the controversy, originating with rationalist and empiricist philosophers, on whether human intelligence works by generalizing from sensory associations, by manipulating symbolic expressions, or by some combination of the two. Searle accuses me of believing that their debate was “essentially” about how the mind works, whereas, he says, it was really about how knowledge claims are to be verified. But as I pointed out, the rationalist-empiricist debate embraced both an epistemological issue (“whether knowledge comes from making deductions using theories or gathering data from observation”) and a psychological issue; the latter, I noted, was simply “the issue that concerns us here.”²Many philosophers would dispute Searle’s contention that the debate was only “incidentally” about thought processes. Locke, Hume, Descartes, Leibniz, and Kant were avid theoretical psychologists who often connected their psychology to their epistemology in a day when the two were not separate fields. It is surely not a coincidence that the philosophers who attributed reliable knowledge to empirical observation also tended to believe that the mind works by generalizing from sensory input, whereas those who attributed knowledge to deducing truths tended to believe that the mind works by cranking through rules. Searle, moreover, is incorrect in suggesting that Descartes was uninterested in combinatorial grammar. Like Leibniz, he speculated about a perfectly grammatical language, free of irregularities and memorized names, and about a logic of thought by which an infinite set of ideas could be enumerated like numbers without committing them to memory.³

The most remarkable of Searle’s misattributions comes from a page on which I used a little icon as a mnemonic placeholder for a word’s meaning, so as to avoid an irrelevant digression on how best to handle this complex issue. Oblivious to my warning that the icon was being used “for convenience,” Searle tries to educate me on why word meanings are not pictures—reiterating arguments that I myself have made in previous books.⁴

How can we explain these cavalier misreadings? For ten years Searle has insisted that he has a philosophical argument which proves that cognitive psychology, cognitive neuroscience, and linguistics are based on “stunning mistakes.”⁵ The main mistake lies in analyzing mental phenomena as forms of information-processing. “There are brute, blind neurophysiological processes and there is consciousness,” Searle wrote, “but there is nothing else.”⁶ Few philosophers have been persuaded by the argument,⁷ and Searle’s eccentric decree has not kept thousands of cognitive scientists and neuroscientists from invoking signals, codes, rules, representations, neural computation, parallel distributed processing, and other information-theoretic constructs that are neither blind neurophysiology nor accessible to consciousness. Searle concedes that Words and Rules explains many puzzles of language. It must be hard for him to reconcile the richness of actual explanations in the sciences of mind with the certitude that he has proven them a priori to be “stunning mistakes.”

Steven Pinker

Peter de Florez Professor of Psychology
MacVicar Faculty Fellow
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
Cambridge, Massachusetts

John Searle replies:

Steven Pinker says that in my review of his book I offer “one misrepresentation after another.” I do not think I misrepresented his book, so I have no choice but to answer his letter, point by point.

1. Words and Rules. The full title of Pinker’s book is Words and Rules: The Ingredients of Language. He thinks I misrepresent the book when I say he claims that “words and rules” explain our understanding of language. Well, let’s look at the text. He begins, on page 1, with the following question: “What is the trick behind our ability to fill one another’s heads with so many different ideas?” And he answers that question as follows:

The premise of this book is that there are two tricks, words and rules. They work by different principles, are learned and used in different ways, and may even reside in different parts of the brain. Their border disputes shape and reshape languages over centuries, and make language not only a tool for communication but also a medium for word play and poetry and an heirloom of endless fascination.

Similar statements occur throughout the book. When I read such passages it certainly seemed to me that he is saying “there are two tricks, words and rules.” Now he tells us that this is a misrepresentation. He did not mean that the two tricks are sufficient to do the job, they are just part of the story. I considered that as a possible interpretation but found the weaker thesis—that languages contain meaningful units and rules for combining them—so obvious as to be “not worth writing a book about.” But now Pinker insists the weaker claim is what he was claiming all along. But where does he say that words and rules are not enough, that at least another “trick” is necessary? Well, he tells us that he said it in another book. Unfortunately, I happen to be reviewing this book and not some other book and this book certainly gives the impression, as this quotation and many others exhibit, that he thinks that two tricks do the job. (I looked at the other book and it contains some examples of how in particular cases, such as the understanding of the Watergate tapes, the understanding of literal meaning is not enough to understand the utterance in context, but even in that book I could not find him making the point I am trying to make: words and rules are never enough to determine interpretation, not even in the simplest cases.)

1. Rationalism and Empiricism. Pinker claims that there are two aspects to the debate between empiricism and rationalism, a psychological side (where the rationalists, unlike the empiricists, were “obsessed by combinatorial grammar”) and an epistemic side (where the rationalists thought “knowledge comes from making deductions using theories”). It is about the psychological side that he claims, “The past tense is the only case I know in which two great systems of Western thought (rationalism and empiricism) may be tested and compared on a single rich set of data, just like ordinary scientific hypotheses.”

I think his account is mistaken both historically and philosophically. To see this, let us look at the standard philosophical dictionary definition of rationalism and its difference from empiricism:

Rationalism, the position that reason has precedence over other ways of acquiring knowledge, or, more strongly, that it is the unique path to knowledge. It is most often encountered as a view in epistemology, where it is traditionally contrasted with empiricism, the view that the senses are primary with respect to knowledge.⁸

I have three basic objections to Pinker’s account.

First, an obsession with combinatorial grammar is not a defining feature of rationalism in its debate with empiricism. Some rationalists had such an obsession, some did not. Many empiricists had such an obsession, many did not. Another dictionary of philosophy lists ten “main tenets” of rationalism.⁹ An obsession with combinatorial systems is not one of them. An interest in combinatorial systems cuts across the rationalist–empiricist debate, and many of the great contributions to combinatorial systems were made by famous empiricists such as Bertrand Russell, Rudolf Carnap, and Willard Quine, just to mention three.

Second, when he does try to describe the essential difference between rationalism and empiricism he gets it wrong. The issue, he tells us, is about whether “knowledge comes from making deductions using theories (rationalism) or gathering data from observation (empiricism).” But that is just wrong about the distinction. Both sides think you have to make deductions using theories—indeed one of the most influential versions of the “theory-deduction” method in epistemology is by the famous empiricist Carl Gustav Hempel. The central thesis of rationalism is that pure reason is the foundation and source of knowledge, “the unique path to knowledge” in the definition quoted above.

Third, the debate about the past tense is not a case in which “two great systems of Western thought (rationalism and empiricism) may be tested and compared on a single rich set of data.” Precisely the features that make them “great systems of Western thought” are left untouched by the discussion of the past tense.

1. Meanings. Here is what Pinker says about meaning: “The meaning of a word is a link to an entry in the person’s mental encyclopedia, which captures the person’s concept of a rose. For convenience we can symbolize it with a picture…” (italics in the original). That certainly seems to me as if he is saying that meanings are some sort of mental entities. But now I gather that the view, like the picture, is just a matter of convenience. I agree that it may be convenient but my point is that it is false, and its falsity is nowhere recognized in the book under review.
2. History. When I said Pinker “neglects” the historical examples, he mistakenly interprets me as saying that his whole book neglects historical examples. But that is not at all what I said. Indeed, I praise the book by saying that one of the best things about it is its discussion of the history of linguistic examples. I made the comment in question in the context of the debate between Pinker and Charles Yang, professor of linguistics at Yale. Yang tries to show how a child could acquire both the regulars and the irregulars with a single mechanism, unlike Pinker’s postulation of two mechanisms, and Yang uses historical evidence to support this thesis. But the existence of the historically based rules that Yang postulates in no way implies that children have to know the historical origins of the rules. To know that “feet” is the plural of “foot,” they do not “have to deduce it from the umlaut rule of proto-Germanic, spoken two millennia ago.” I do not believe that Pinker has adequately answered Yang’s arguments. And there is a deeper point I raised in my review that he does not address: if we are going to appeal to rules in the explanation of language we need an answer to the questions “What exactly is a rule of language?” and “What role exactly does the rule play in linguistic behavior?”

3. Computation and information processing. I believe the real source of Pinker’s indignation here goes beyond the issues discussed in his book. He is outraged that I reject the computational theory of the mind. I do think it is a “stunning mistake” to think that all that is going on our heads that is relevant to cognition is the implementation of computer programs. I have refuted this view in a number of places and I continue to think it is a mistake. But in the last paragraph of his letter Pinker misrepresents my views. He thinks I am denying the possibilities of an information-processing cognitive science. Quite the contrary; and indeed, in the article he cites I provide a way of making sense of the notion of information processing in cognitive science. But I do insist that we need to distinguish between the sense of such notions as “information processing” and “computation” where they are dependent on outside observers, as when we say for example that my computer is processing information right now when I type into it, or the medulla processes information when it regulates my breathing, from the sense of these notions which is independent of observers, as when I am consciously computing my income tax or consciously processing information when I think about how to explain my views in a letter to The New York Review.

In the observer-relative sense just about any system can be described as computing and processing information. The stomach computes and processes information during digestion, and trees process information when they take in information about the cycle of the seasons and print it out in the form of rings in the trunk. Unless you keep clear about these two senses of these and related notions you are going to make some serious mistakes.

These are large and complex issues and I have to say that Pinker has seriously misrepresented my views in discussing these very points. Years ago, in these pages¹⁰ and elsewhere, I argued that the computer theory of the mind could not be right because if it were I would acquire a knowledge of Chinese just by following the steps in a computer program for answering questions in Chinese.

In responding to this “Chinese Room Argument,” Pinker writes as follows:

Searle’s tactic is to appeal over and over to our common sense. You can almost hear him saying, “Aw, c’mon! You mean to claim that the guy understands Chinese??!!! Geddadahere! He doesn’t understand a word!! He’s lived in Brooklyn all his life!!” and so on. But the history of science has not been kind to the simple intuitions of common sense, to put it mildly.¹¹

Perhaps Pinker can “almost” hear me saying this, but I cannot. The quoted passage is not in a dialect I speak, nor one I find congenial. The deeper misrepresentation is not in the quoted passage but in the substantive claim that my argument is just an appeal to common sense. The fact is that in over twenty years of debating these issues, I have never relied on common sense. I appeal to a logical distinction between the syntax of the implemented program and the semantics of actual human understanding, and the thought experiment in question is designed to illustrate the distinction between the syntax and the semantics. I am not sure I know what “common sense” is, but I doubt that it contains theories about the distinction between syntax and semantics.