The Articulate Mammal (4 page)

‘It’s all right, sir, Really it is. We’ve got a connection wrong somewhere, that’s all. You must remember, Mr Bohlen, there’s over a million feet of wiring in this room.’

‘It’ll never work,’ Mr Bohlen said.

Roald Dahl,
The Great Automatic Grammatizator

Every normal human being can talk. So the average person tends to think that there is little or nothing mysterious about language. As the linguist Noam Chomsky has pointed out:

We lose sight of the need for explanation when phenomena are too familiar and ‘obvious’. We tend too easily to assume that explanations must be transparent and close to the surface … As native speakers, we have a vast amount of data available to us. For just this reason it is easy to fall into the trap of believing that there is nothing to be explained. Nothing could be further from the truth …

(Chomsky 1972a: 25–6)

But the mysterious nature of human language becomes more apparent when one realizes that no one has yet managed to simulate the language ability of a human being. Computers can play chess, sort bank statements, and even talk about limited topics such as cubes, squares and cones. But we are far from producing a ‘great automatic grammatizator’ which could unaided hold conversations on any topic. Why is this? Perhaps we should think about language more carefully.

NATURE OR NURTURE?

When people start thinking about language, the first question which often occurs to them is this: is language
natural
to humans? – in the same way that grunting is natural to pigs, and barking comes naturally to dogs. Or is it just something we happen to have
learned
? – in the same way that dogs may learn to beg, or elephants may learn to waltz, or humans may learn to play the guitar.

Clearly, in one sense, children ‘learn’ whatever language they are exposed to, be it Chinese, Nootka or English. So no one would deny that ‘learning’ is very important. But the crucial question is whether children are born with ‘blank sheets’ in their head as far as language is concerned – or whether humans are ‘programmed’ with an outline knowledge of the structure of languages in general.

This question of whether language is partly due to
nature
or wholly due to learning or
nurture
is often referred to as the
nature–nurture
controversy, and has been discussed for centuries. For example, it was the topic of one of Plato’s dialogues, the
Cratylus.
Controversies which have been going on for literally ages tend to behave in a characteristic fashion. They lie dormant for a while, then break out fiercely. This particular issue resurfaced in linguistics in 1959 when the linguist Noam Chomsky wrote a devastating and witty review of
Verbal Behavior
, a book by the Harvard psychologist B.F. Skinner (Skinner 1957; Chomsky 1959). This book claimed to ‘explain’ language as a set of habits gradually built up over the years. According to Skinner, no complicated innate or mental mechanisms are needed. All that is necessary is the systematic observation of the events in the external world which prompt the speaker to utter sounds.

Skinner’s claim to understand language was based on his work with rats and pigeons. He had proved that, given time, rats and pigeons could be trained to perform an amazing variety of seemingly complex tasks, provided two
basic principles were followed. First, the tasks must be broken down into a number of carefully graduated steps. Second, the animals must be repeatedly rewarded.

In a typical experiment, a rat was put in a box containing a bar. If it pressed the bar, it was rewarded with a pellet of food. Nothing forced it to press the bar. The first time it possibly did so accidentally. When the rat found that food arrived, it pressed the bar again. Eventually it learned that if it was hungry, it could obtain food by pressing the bar. Then the task was made more difficult. The rat only got rewarded if it pressed the bar while a light was flashing. At first the rat was puzzled. Eventually it learned the trick. Then the task was made more difficult again. This time the rat only received food if it pressed the bar a certain number of times. After initial confusion, it learned to do this also. And so on, and so on.

This type of ‘trial-and-error’ learning was called
operant conditioning
by Skinner, which can be translated as ‘training by means of voluntary responses’ (the word ‘operant’ means a voluntary response rather than an automatic one). Skinner suggested that it is by means of this mechanism that the vast majority of human learning takes place, including language learning:

The basic processes and relations which give verbal behaviour its special characteristics are now fairly well understood. Much of the experimental work responsible for this advance has been carried out on other species, but the results have proved to be surprisingly free of species restrictions. Recent work has shown that the methods can be extended to human behaviour without serious modification.

(Skinner 1957: 3)

All one needed to do in order to understand language, he said, was to identify the ‘controlling variables’, which would enable us to predict specific utterances. For example, in the same way as it was possible to say that a rat’s bar-pressing behaviour was partly ‘under the control’ of a flashing light, so a feeling of hunger might ‘control’ or predict a human utterance such as ‘Please pass the bread and butter.’ Or the presence of a beautiful painting might call forth the exclamation, ‘Oh how beautiful.’ Or a bad smell might cause one to exclaim ‘Oh what a terrible smell.’ A French notice, such as ‘
Ne touchez pas
’, might result in one saying, ‘That means “Don’t touch”.’ And if a child said ‘Hickory dickory dock’, you are likely to continue ‘The mouse ran up the clock.’ In theory, Skinner saw no difficulty in linking up any particular set of words which a human might wish to produce with an identifiable external happening.

In practice, the matter is far from simple, as Chomsky pointed out. Chomsky made two major criticisms of Skinner’s work. First, the behaviour
of rats in boxes is irrelevant to human language. Second, Skinner fundamentally misunderstood the nature of language.

THE IRRELEVANCE OF RATS

Chomsky pointed out that the simple and well-defined sequence of events observed in the boxes of rats is just not applicable to language. And the terminology used in the rat experiments cannot be re-applied to human language without becoming hopelessly vague.

For example, how do you know that someone is likely to say ‘Oh what a beautiful picture’ when looking at a beautiful painting? They might say instead, ‘It clashes with the wallpaper’, ‘It’s hanging too low’, ‘It’s hideous.’ Skinner would say that instead of the utterance being ‘controlled’ by the beauty of the picture, it was ‘controlled’ by its clash with the wallpaper, its hanging too low, its hideousness. But this reduces the idea of ‘control’ to being meaningless, because you have to wait until you hear the utterance before you know what controlled it. This is quite unlike the predictable behaviour of rats which could be relied upon to respond to certain stimuli such as a flashing light with a fixed response.

Another problem was that the rats were repeatedly rewarded. It is quite clear that children do not receive pellets of food when they make a correct utterance. However, the idea of reward or
reinforcement
(since it reinforces the behaviour that is being learned) can in humans be naturally extended to approval or disapproval. One might suppose that a parent smiles and says ‘Yes dear, that’s right’ when a child makes a correct utterance. Even if this were so, what happens to this idea of approval when there is nobody around, since children are frequently observed to talk to themselves? Skinner suggested that in these cases children automatically ‘reinforce’
themselves
because they know they are producing sounds which they have heard in the speech of others. Similarly, Skinner assumed that someone like a poet who is uttering words aloud in an empty room will be ‘reinforced’ by the knowledge that others will be influenced by the poetry in the future. So reinforcement seems a very woolly notion, since an actual reward need not exist, it need only be imagined or hoped for. Such a notion is certainly not comparable to the food pellets given to rats when they make a correct response.

Studies by Roger Brown and his associates provided even more problems for Skinner’s notion of reinforcement. After observing mother–child interactions they pointed out that parents tend to approve statements which are
true
rather than those which are grammatically correct. So a boy who said ‘Teddy sock on’ and showed his mother a teddy bear wearing a sock would probably meet with approval. But if the child said the grammatically correct utterance
‘Look, Teddy is wearing a sock’, and showed his mother a bear
without
a sock, he would meet with disapproval. In other words, if approval and disapproval worked in the way Skinner suggested, you would expect children to grow up telling the truth, but speaking ungrammatically. In fact the opposite seems to happen (Brown
et al.
1968).

Another example of a problem which crops up in trying to match rat and human behaviour is that of defining the notion of
response strength.
When a rat has learned to respond to a particular external happening, the extent to which it has learned the lesson can be measured in terms of the speed, force and frequency of the bar-pressing. Skinner suggested that similar measures of response strength might be found in some human responses. For example, a person who was shown a prized work of art might, much to the gratification of the owner, instantly exclaim ‘Beautiful!’ in a loud voice. Chomsky pointed out:

It does not appear totally obvious that in this case the way to impress the owner is to shriek ‘Beautiful’ in a loud, high-pitched voice, repeatedly, and with no delay (high response strength). It may be equally effective to look at the picture silently (long delay), and then to murmur ‘Beautiful’ in a soft lowpitched voice (by definition, very low response strength).

(Chomsky 1959: 35)

Chomsky used these and similar arguments to show the irrelevance of Skinner’s experiments to the problem of understanding language. Perhaps ‘irrelevance’ is too strong a word, since there are areas of language where habit forming works. For example, some people invariably say ‘Damn’ if they drop a raw egg, or ‘Good night’ when they are going to bed, or ‘London transport gets worse every day’ when standing at a bus-stop. And there is one sad character in a Beatles’ song who only ever says ‘Good morning’:

I’ve got nothing to say but it’s OK

Good morning, good morning, good morning.

But apart from trivial exceptions such as these, language is infinitely more complex and less predictable than Skinner’s theory would suggest.

Of course, just because Skinner’s ideas were over-simple does not automatically mean that Chomksy’s ideas were right. Maybe both Skinner’s and Chomsky’s views are outdated. Now, in the twenty-first century, we know a lot more about language and its special qualities, partly because Chomsky in particular inspired so many to take language seriously as a key to understanding the human mind, and to work on it further.

THE NATURE OF LANGUAGES

What is there about language that makes it so special? There are a large number of human activities such as learning to drive or learning to knit which seem to be learnt in the same way as bar pressing by rats. Why not language also?

Chomsky pointed out some of the special properties of language in his review of Skinner’s book, where he suggested that Skinner was not in a position to talk about the causation of verbal behaviour, since he knew little about the character of such behaviour:

There is little point in speculating about the process of acquisition without a much better understanding of what is acquired.

(Chomsky 1959: 55)

Chomsky has since discussed the nature of language in a number of places (e.g. Chomsky 1972a, 1986, 1995b 2000, 2002). One point which he stressed is that language makes use of
structure-dependent operations.
By this he means that the composition and production of utterances is not merely a question of stringing together sequences of words. Every sentence has an inaudible internal structure which must be understood by the hearer.

In order to see more clearly what is meant by a
structure-dependent
operation, it is useful to look at
structure-independent
operations.

Suppose a Martian had landed on earth, and was trying to learn English. She might hear the sentence:

AUNT JEMIMA HAS DROPPED HER FALSE TEETH DOWN THE DRAIN

as well as the related question:

HAS AUNT JEMIMA DROPPED HER FALSE TEETH DOWN THE DRAIN?

If she was an intelligent Martian, she would immediately start trying to guess the rules for the formation of questions in English. Her first guess might be that English has a rule which says, ‘In order to form a question, scan the sentence for the word
has
and bring it to the front.’ Superficially, this strategy might occasionally work. For example, a sentence such as:

PETRONELLA HAS HURT HERSELF

would quite correctly become:

HAS PETRONELLA HURT HERSELF?

But it is clearly a wrong strategy, because it would also mean that the Martian would turn a statement such as: