The Articulate Mammal (5 page)

THE MAN WHO HAS RUN AWAY SHOUTING WAS ATTACKED BY A WASP

Into:

*HAS THE MAN WHO RUN AWAY SHOUTING WAS ATTACKED BY A WASP?

which is not English. (An asterisk denotes an impossible sentence.)

Looking at the Aunt Jemima sentence again, the Martian might make a second guess, ‘In order to form a question, bring the third word to the front.’ Once again, this might superficially appear to work because a sentence such as:

THE ALLIGATOR HAS ESCAPED

would correctly become:

HAS THE ALLIGATOR ESCAPED?

But it is obviously accidental that this type of rule gets the right result, because it also produces a number of non-sentences:

SLUGS ARE SLIMY

would become:

*SLIMY SLUGS ARE?

And:

MARY HAS SWALLOWED A SAFETY PIN

turns into:

*SWALLOWED MARY HAS A SAFETY PIN?

The Martian went wrong in her guesses because she was trying out structure-independent operations – manoeuvres which relied solely on mechanical counting or simple recognition procedures without looking at
the
internal
structure of the sentences concerned. In order to grasp the principles of question formation, the Martian must first realize that:

AUNT JEMIMA, THE MAN WHO HAS RUN AWAY SHOUTING, SLUGS, MARY

each behaves as a unit of structure. The number of words within each unit is irrelevant, so no amount of counting will produce the right result for question formation. In these sentences (though not in all English sentences) the solution is to take the word which follows the first unit and bring it to the front:

This may seem an obvious solution to people who already know English – but it is not at all clear
why
language should behave in this way. As Chomsky pointed out:

The result is … surprising from a certain point of view. Notice that the structure-dependent operation has no advantages from the point of view of communicative efficiency or ‘simplicity’. If we were, let us say, designing a language for formal manipulations by a computer, we would certainly prefer structureindependent operations. These are far simpler to carry out, since it is only necessary to scan the words of the sentence, paying no attention to the structures which they enter, structures that are not marked physically in the sentence at all.

(Chomsky 1972b: 30)

Yet, amazingly, all children learning language seem to know automatically that language involves structure-dependent operations. On the face of it, one might expect them to go through a prolonged phase of testing out Martianlike solutions – but they do not. This leads Chomsky to suggest that humans may have an innate knowledge of this phenomenon:

Given such facts, it is natural to postulate that the idea of ‘structuredependent operations’ is part of the innate schematism applied by the mind to the data of experience.

(Chomsky 1972b: 30)

This knowledge, he argued (somewhat controversially), ‘is part of the child’s biological endowment, part of the structure of the language faculty’ (Chomsky 1988: 45).

The structure-dependent nature of the operations used in language is all the more remarkable because there are often no overt clues to the structure. Experiments carried out by psycholinguists have made it clear that listeners do not have to rely on auditory clues for interpreting the main structural divisions. For example, Garrett
et al.
(1966) constructed two sentences which each contained the words:

GEORGE DROVE FURIOUSLY TO THE STATION:

1 IN ORDER TO CATCH HIS TRAIN GEORGE DROVE FURIOUSLY TO THE STATION.

2 THE REPORTERS ASSIGNED TO GEORGE DROVE FURIOUSLY TO THE STATION.

In the first sentence, it is GEORGE who is driving furiously. In the second, it is the REPORTERS. In order to understand the sentence, the listener must (mentally) put the structural break in the correct place:

Just to check that the listeners were
not
using auditory clues, the experimenters recorded both these sentences on to tapes. Then they cut the words GEORGE DROVE FURIOUSLY TO THE STATION off each tape, and spliced them to the
other
sentence:

They then played the newly spliced tapes to students – but into one ear only. In the other ear the students heard a click, which was placed in the middle of a word, for example, GEORGE. The students were then asked whereabouts in the sentence the click had occurred. The interesting result was that in their reports students tended to move the location of the click in the direction of the structural break:

This indicates clearly that listeners impose a structure on what they hear for which there is often
no
physical evidence.

Another point made by Chomsky (1959) and others is that simple slot-filling operations are inadequate as explanations of language. It has sometimes been suggested that anyone learning language allocates to each sentence a number of ‘slots’ and then fits units of structure into each hole, for example:

No one would deny the existence of such substitutions and their value in language learning. But the problem is that there is a lot more going on besides, which cannot be accounted for by the ‘slot’ idea: ‘It is evident that more is involved in sentence structure than insertion of lexical items in grammatical frames’ (Chomsky 1959: 54). For example, look at the following sentences:

As soon as we try to find other words to fit into the slot occupied by
can be
, we run into problems.
Are
fits in with the first sentence but not the second, whereas
is
fits in with the second but not the first:

If slot-filling was the sole principle on which language worked, one would not expect this result. In fact, slot-filling makes it quite impossible to explain how the listener knows, in the sentences where the centre slot is filled by
can be
, that it is the fleas who are performing, but that it is not the tiddlywinks who are playing. But examples of ‘constructional homonymity’ (as Chomsky calls such superficially similar utterances) are by no means rare.

Even more inexplicable from a slot-filling point of view are sentences which can be interpreted in two different ways:

CLEANING LADIES CAN BE DELIGHTFUL:

1 LADIES WHO CLEAN CAN BE DELIGHTFUL.

2 TO CLEAN LADIES CAN BE DELIGHTFUL.

THE MISSIONARY WAS READY TO EAT:

1 THE MISSIONARY WAS ABOUT TO EAT.

2 THE MISSIONARY WAS ABOUT TO BE EATEN.

Sentences such as these indicate that merely filling a grammatical frame may be only part of what is happening when we speak. Such examples led Chomsky in the 1960s to suggest that language might be organized on two levels: a
surface
level, in which words are in the place where they actually occur, and a
deep
level, in which words are located in their ‘proper’ place in the slot structure.

Chomsky’s arguments that a ‘deeper’ level of syntax underlay the surface level were interesting, but not necessarily right. Other explanations are possible, as he himself later stressed (Chomsky 1995b). The important point is that the differing interpretations of the ambiguous sentences described above can
not
be explained by means of the bar-pressing antics of rats, nor by means of simple slot-filling operations. Some more complex procedure is involved.