The way children learn a language seems to be largely independent of what their parents teach them. For example, children use grammatical forms that they never heard from their parents. The ability of children to learn a language depends on their brains to be biased towards some kinds of phrases (e.g., noun phrases) and to be equipped with some kinds of procedures.

Pinker then attacks "linguistic determinism", Sapir's and Whorf's theory that language determines thought (that we learn an entire culture when we learn to speak a language). Pinker tries to "debunk" the myth that Eskimos have a lot more words for snow than other cultures, a fact that, in Whorf's eyes, proved linguistic determinism. Whether Pinker succeeds or not in doing it (there are as many versions of this controversy as there are supposedly words for snow in Eskimo), he does not convincingly demolish the argument. We know for a fact that the Italians are Italian and the Chinese are Chinese and Pinker is obviously an Anglosaxon by the way he writes. In many ways the French Canadians are more similar to their linguistic neighbors of France than to their geographical neighbors of the USA. There are clearly attitudes that obey linguistic borders. Pinker ignores the fact and declares that the fact does not exist because his theory does not provide for it. It's like an astronomer deciding that the Sun does not exist because his theory does not contemplate one. It would be more scientific to start with the assumption that the Sun exists and then provide a theory for this fact. Sapir and Whorf offered their theory for it, "linguistic determinism". Pinker does not offer any theory, implying that it must just be our imagination. Obviously, he has never had to deal with Italian postal officers and French waiters.

Pinker thinks that linguistic determinism is wrong because he thinks that language is merely a (very limited) medium to pack our thoughts and broadcast them to our fellow humans. In the process, we lose most of the reasoning that went on in our brains. Pinker thinks that the mental representation of what we are saying is way more complex and subtle, and just cannot be expressed in words. Words are an effective way to deliver the essential part of the meaning in a reasonable amount of time. Our thoughts are not verbal: they are mental representations of the kind of Jerry Fodor's "mentalese". And this mentalese is a genetic fact: we inherit it when we inherit human genes. It is universal. We all reason the same way. We think in mentalese, not in English or Chinese. It is only when we have to pack information for another human being that we use the language of our community (e.g., English or Chinese), and in doing so we have to limit our message to what can be said in that language.

(It does not occur to Pinker that this act of constraining our messages may actually influence "what" we think. Even if "how" we think is the same for everybody and is determined only by the genetic "mentalese", the "what" may very much depend on "what" we are communicated and we communicate, and the "what" is really our thought).

Pinker therefore sides with the school of the "physical symbolic processor": a mind is a purely syntactic processor of symbols, and the "intelligence" of that mind arises from processing the symbols, just like the Turing machine is capable of solving (almost) every problem by moving symbols around, without actually "knowing" what problem it is solving (or even that it is solving a problem).

Pinker's point that mental representation is a powerful invention is easier to understand. That was Turing's fundamental discovery. Once we equip something with the ability to represent the world through symbols that can be processed in some logical way, that something is suddenly endowed with the power to solve even the most complex of problems. Mental representation sounds indeed miraculous. The symbols are not intelligent. The algorithm to process them is not intelligent. But the result is intelligent. In fact, the result "is" intelligence itself.

Language and thought have little in common, according to Pinker: "knowing a language is knowing how to translate mentalese into strings of words, and viceversa". People without a language are still thinking in mentalese.

Pinker notes that mentalese might actually be simpler than the languages we speak, because it doesn't have to deal with the oddities of spoken language (such sa pronouns and indexicals) or with pronunciation. Presumably, Pinker thinks that the complexity of spoken languages evolved because it helps communicate the essential very quickly.

After a summary of how syntax and morphology work, Pinker gives a much more interesting summary of phonology. We often forget that language as originally used to speak, not to write, and that it is, ultimately, sounds. Sentences are made of words, words of morphemes, and morphemes of phonemes. And phonemes are fundamentally different from the other components of language, because they do not combine according to a grammar: they represent the analogic to digital interface.

At some point in evolution, the mouth and the ear developed additional functions: to help utter sounds and to help to "understand" sounds. Pinker claims that we can hear "words" where there are just sounds because phonetic perception is a sixth sense, another piece in the puzzle of the language instinct. Our brains are hardwired to recognize "meaningful" words out of a stream of "meaningless" sounds (there are actually no meaningless sounds, and, according to Pinker's own theory, words are not really meaningful, but Pinker uses "meaningful" as in "useful for the purpose of reacting to a sentence"). Pinker shows how we use an assembly of organs to create the sounds of sentences. Recognizing a phoneme is much more difficult than dealing with grammar, so much so that no machine has been built yet that can recognize speech the way the human brain does. First of all, there is hardly any separation between words when we speak: there is a continuous flow of sounds. Secondly, different speakers pronounce the same words in different ways. Thirdly, the same speaker can pronounce the same word in different ways (depending on whether she is sleepy or not, angry or not, in a hurry or not). Machines that try to recognize speech have to be "trained" to the voice of a particular speaker, and can generally recognize only a small subset of the vocabulary (usually, only a dozen of words, instead of the tens of thousands that the human brain recognizes effortlessly). Clearly, the wiring of the human brain is the secret to recognizing speech.

There are other features of the human brain that are difficult to replicate with a machine, and they all have to do with "analogic" versus "digital" reasoning. Uttering or listening to speech is a function very similar to grasping an object, another difficult act for a machine. The brain does not simply calculate distance and angle to program the movement of the arm. The movement of the arm and of many other organs, from the joints of the fingers to the muscles of the eyes are continuously refined as the movement itself takes place. A complex process of feedback makes sure that the movement achieves its goal. Pinker shows that something similar happens when we speak: a number of organs cooperate in making the sounds of words, and the sound wave is refined as it is being uttered; and viceversa when the soundwave is being heard. Thus speech belongs to the general class of "motor control". Just like with other kinds of motor control, the "expectation" contributes to the success of the operation. If one "expects" to hear something, then she is more likely to hear it. Since we expect a speaker to make sounds that are words in our language, we are more likely to detect them, even though the words are not coded in the exact same soundwave each and every time. Speech recognition is greatly facilitated if the hearer continuously guesses what the speaker is trying to say. The secret of speech is not in its "digital" workings (how the brain dissects and processes its constituents) but in its "analogic" workings (how the brain continuously readjusts the process to sculpt an output that matches the input).

If all humans are equipped with the same universal grammar, a legitimate question is why are there so many languages instead of just one. Pinker's answer is similar to the answer to the question of why are there so many species of animals if all animals are equipped with the same genetic code: it's the way evolution works, namely variation is an inherent element of evolution. Linguistic variation boosts cultural evolution the same way that genetic variation boosts biological evolution. New languages are born the same way that new species are born, through a process of variation, heredity and isolation. (This similarity had originally been pointed out by Darwin himself). Pinker does not elaborate on the linguistic equivalent of "natural selection", i.e. the role played by the "environment" (which, in the case of language, is the society of other speakers), but language too is subject to environmental pressure. If a child utters a meaningless sentence that brings no benefit (or is even harmful in achieving the goal), that sentence will die out. On the other hand, novel sentences or grammatical constructs or idiomatic expressions that turn out to be very effective are inherited by other speakers and spread throughout the population of speakers. This is the equivalent of what natural selection does to organs of bodies.

Pinker calls it "language instinct". Like all instincts, it must be implemented somewhere in the brain, and that implementation must be dictated by some genes of the human genome (and neurologists proved that language is implemented in the left hemisphere). That is why Pinker thinks that animals cannot speak language, "real" language: they lack the genes, and therefore they lack the brains. They can certainly be trained to recognize and react to certain sounds (although never with the same dexterity of a child) but they lack the "discrete combinatorial system" that would enable them to understand "other" sentences besides the ones they have been trained to react to. Children do not simply repeat the sentences that they have been taught: children come up with their own sentences. What children have learned is "language", not just a few words or a few sentences. That is what animals cannot learn. Animals can learn to react to the sentence "Kiss me" and "Dog", but they cannot understand the sentence "Kiss the dog". Even less likely is that they can understand the sentence "Kiss her". And even less likely to reply to a simple question such as "Why?". Children, instead, rapidly learn to deal with sentences such as "Why aren't you nice to her?" even if they never heard those words before in that specific sequence, as long as they have learned what "nice" means, and what pronouns stand for, and what is expected by a "why".

Neurally speaking, Pinker thinks that human language is ultimately controlled by the neocortex, whereas animal "language" is controlled by the evolutionary older structures in the brain stem and in the limbic system. Humans too have this primitive form of language controlled by the same ancient brain structures, but those are the kind of sounds that cannot be "combinatorially combined", for example a scream of terror or a burst of laughter. HUman language is not a combination of this primitive sounds, but a different process of syntax, morphology and phonology that takes place in a different region of the brain.

Pinker's speculations on how language evolved sound a lot less plausible. Pinker thinks that language evolved through a "revamping of primate brain circuits that originally had no role in vocal communication, and by the addition of some new ones."

A great deal of the book is devoted to linguistic details of Chomsky's theory applied beyond the original realm of syntax.

TM, ®, Copyright © 2005 Piero Scaruffi