Just How Sane Are We?

“Lord, what fools these mortals be!” Shakespeare put that line in A Midsummer’s Night Dream, first staged in the 1590s, and I doubt that anyone at the play’s opening thought the idea was new. Thus, a whole book asserting that we mortals do indeed have our foolish side might seem unnecessary. But a series of doubters have appeared since Puck had his laugh, people who insist that, to paraphrase Dr. Pangloss, we mortals are the most rational of all possible species. Now there is an idea that could give both Puck and his audience a hearty laugh. Who, you might demand, argues such nonsense?

Many economists do. Market enthusiasts insist that people behave rationally, and books with catchy titles like Freakonomics turn out to be expositions of how rational our seemingly strange behavior is.

Evolutionists also do so. A year ago on this blog I extensively examined Jean Louis Dessalles book, Why We Talk. One of its claims was that evolution always optimizes its products, so that if, say, we speak at about 120 words per minute, that is the optimal production rate.

Noam Chomsky has been arguing for about 20 years now that the syntactic tree structure is nigh-well perfect for generating sentences.

Anti-evolutionists (intelligent design proponents) also argue for the well made mind. One of the jokes of science history is that in Darwin’s day the challenge for biologists was to explain where all the order and design in nature came from, and now the challenge for creationists is to explain away nature’s clutter and confusion.

These are the main targets of Gary Marcus’s new book Kluge: The Haphazard Construction of the Human Mind. He knocks them all down in well under 200, easy-to-read pages.

Marcus’s thesis is not just that humans can be foolish. More specifically, he argues that evolution has not given us a well-engineered brain. We have evolved a Swiss-Army knife of gadgets and thingamajigs that was originally born to serve apes and even more distant ancestors. Thus, many of our problems arise from the way we inherited a trait that performed one thing which we jerry-rigged to do something else. The work-around nature of the solution (the kluge) limits performance. Engineers, starting afresh, could have, and would have, done a better job.

Put that generally, I have no trouble with the thesis. But there is a danger, when we get down to cases, of blaming an organ for not doing something well when it evolved to do something else. Marcus devotes a chapter to language and it seemed to me largely oblivious to the risk that he might be getting the function of language wrong. He writes:

Computer languages don’t suffer from these problems [of instability, vagueness, and ambiguity]; in Pascal, C, Fortran, or LISP, one finds neither rampant irregularity nor pervasive ambiguity—proof in principle that languages don’t have to be ambiguous. [p. 103]

Does that really settle the matter, in principle?

Well, it sure shows one problem inherent in languages—the tendency to confuse categories by lumping separate things under the same name. Pascal is NOT a language like English in the way English is like French. This post that you are reading right now can be translated into French; it cannot be translated into Pascal or any other computer “language.” Why not? Simply because computer languages are entirely procedural. Their one function is to control the actions of a machine. It is true that advertisers, publicists, and propagandists all wish language could control us so exactly, but their ambitions are often frustrated. George Orwell’s 1984 satirizes the propagandist’s ambition by introducing Newspeak, a language that is so restricted in meanings that Big Brother’s subjects cannot stray from the official line. But Orwell’s point was that language is one of those things that let humans form a human society, and if you want to control society you must control its speech. Pascal and the other computer languages are tools for controlling machines; they are models for a desirable human language only to the extent that controlling other people is desirable.

So I am quite suspicious of the book’s anti-humanism and its doctrine that sound engineering standards are the measure of all things. But Kluge does have the strength of debunking other doctrines whose anti-humanism rests on a faith in human rationality that is nowhere supported by the evidence. Marcus is particularly strong where he is not arguing directly with evolution’s mysterious goals, but critiquing theorists who insist that humans, when left on their own, will perform in the best and most rational way possible. For language students, debunking such optimism means critiquing Noam Chomsky’s claim that we create and parse syntax the way a soundly-made machine does.

Marcus summarized his objections to Chomsky’s position at the Evolang conference in Barcelona last month. (see: The Practicality of Studying Language Origins) His main criticism of generative syntax is that humans cannot follow long, complicated embeddings. Even short ones give us trouble if they are sufficiently absurd: e.g., people people left left. You have to be enamored of a very strict theory to think there is something wrong with us because we have a rough time figuring out what this grotesquery means, but Chomsky does have a strict theory and it does accept “sentences” of this sort. (What does the sentence mean? [The] people[, whom some other] people left[,] left.]) The problem, as understood on this blog, is that speech directs attention. Embeddings always get in the way of that easy direction, and even when more felicitously handled, causes problems: The man who broke the bank at Monte Carlo has just come into the room. We can follow this sentence because the word who warns us that our attention is being detoured for a moment. A great deal of well-designed speech keeps listeners alert to where their attention should shift. Commands written in Pascal or C++ don’t do that, but perhaps that absence reflects a difference in function.

One area where Marcus and I are in full accord is the problem of writing. Speech is, at an absolute minimum, 100 thousand years old; writing is one tenth that age and most of the people who lived during the past 10 thousand years of written language have been illiterate. Our whole linguistic biology, therefore, is oral, and written language has all the limitations of something tacked on. Most writing seems comparable to the stunt of a circus animal performing a trick outside its genetic lineage. Generally, we perform better than bears on bicycles, but perhaps not as well as seals balancing balls on their noses. Meter and style are only partial substitutes for the shades of emotion and attitude carried by tone of voice. The complete absence of audience feedback would seem to guarantee confusion as sentences mount. The solution, however, is to master literary craftmanship, not to write more sentences like those in C++. And since Marcus himself is a very smooth writer, I am sure he knows just what I mean.

Comments

[yair shimron]

Writing is but 5 thousands years old, as far as Sumer and Egypt are concerned.

[David]

I"ve been a lurker here for awhile and thanks for all the thought provoking stuff.....

I really love your analogy between computer and human language and being involved as I am in this area, you hit the keyboard in the right places....

I'm gonna get me this book, just so I can be a better doubter.

David

[Dave]

I'm glad you mentioned written language. I think that much of linguistics is somewhat unfortunately biased towards theories of language that are steeped in the idea that writing is natural and fundamental to language. This bias, I think, is present from the lowest levels of phonetics and phonology (where vaguely defined and often ambiguous categories of sounds are represented both in writing and, supposedly, in the mind, using discrete graphemes or letters) to the highest levels of morphology and syntax (where understanding spoken language is equivalent to parsing discrete written tokens).

There's a lot of interesting research and theorizing waiting to be done concerning the nature of physical, biological, social and historical constraints on language. Or, perhaps it's already been done and I just haven't had a good enough linguistics education...