The Ascent of Babel: An Exploration of Language, Mind, and Understanding
Page 13
You're wrong-Sam didn't buy MARY a pizza [he bought it for someone else]
You're wrong-Sam didn't BUY Mary a pizza [he made her one]
You're wrong-Sam DIDN'T buy Mary a pizza [he did something else, or nothing at all]
You're wrong-SAM didn't buy Mary a pizza [someone else did]
In each one of these cases, the speaker can use emphatic stress to convey some extra meaning in addition to the literal meaning of the sentence itself. But whether we would say that the sentence `Sam didn't buy Mary a pizza', in its written form, is ambiguous, because we cannot figure out which of the many alternative meanings is intended, is unclear. Does the stress disambiguate the alternative meanings, or add alternative meanings? Probably, in some cases it does one, and in other cases it does the other. Either way, there is a lot more to figure out when listening to sentences than one might expect on the basis of our experience of reading them.
So the role of prosody in all of this is still unclear. Although there are extremely good accounts of what kinds of thing prosody can and cannot signal, there is still considerable controversy about when the signals are used. Are they used after all the other things are worked out, like ambiguity, or literal meaning? Or can they influence the ways in which these other things are worked out in the first place? Even though these issues have been around for a long time, it is only very recently that researchers have begun to explore these issues. At present, they are far from being resolved.
On minds, meanings, and grammar
There is no doubt that in order to recognize meaning, we have to recognize the grammatical conventions (including the prosodic ones) that signal that meaning. But although we must construct mental representations which capture the meanings conveyed by the sentences we hear and read, do we also construct mental representations that do nothing more than characterize the grammatical conventions themselves? Do we construct mental representations that capture the prosodic characteristics of the spoken sentence without capturing the meaning that these characteristics convey?
The descriptions that linguists use to characterize the grammatical conventions employed in a language are expressed in terms of syntactic categories such as noun and verb. These descriptions encompass the fact that `The girl likes the language' and `The boy ate the pizza' share certain common properties; the language is the object of the liking just as the pizza is the object of the eating, and the boy is the eater just as the girl is the liker. A description of this similarity requires no appreciation, in fact, of what the individual words in the two sentences actually mean. We could just as well say of each sentence that `There is an X being Y'd, and a Z that is doing the Y'ing'. But although this description makes no reference to the meanings of X, Y, and Z, it does nonetheless capture an important element of that meaning. To understand these sentences, it is not enough to simply know the meanings of the words `girl', `language', `likes', and so on. We need also to know that other element. But does this mean that some aspect of our own mental representation of these sentences corresponds to that element? Is there some aspect of our own mental representation of these sentences which is also independent of, and expressed without reference to, the meanings of these sentences?
Another way of asking the same question is to ask whether the kinds of mental representations that are constructed when listening to, or reading, a sentence somehow mirror the descriptions and linguistic representations that linguists use. The question is an important one, because of the claim we encountered earlier that certain kinds of ambiguity may be resolved on the basis of which grammatical conventions (which linguistic representations) are least complex, or are most frequently encountered in the language at large. To resolve ambiguities on such bases would presumably require some mental version of those conventions-a version of those conventions that was divorced from the individual meanings they would normally convey.
Whether or not such mental representations are in fact derived as we hear and read is at the heart of many a psycholinguistic debate-what could a psycholinguist possibly want except to discover the nature of the internal representations that are constructed to depict the who-didwhat-to-whom of sentence understanding? It is unlikely that we shall ever discover the nature of these representations. We cannot open up the brain and have a look. We must instead look to our theories of what meaning and understanding are, and how they may be contained within the neural structures of the brain. In Chapters 9 and 13 we shall do just this.
To a certain extent, the kinds of ambiguity we have encountered in this chapter have been the battleground on which opposing theories have been paraded and fought over. But ambiguities are just a part of the problem that faces us each time we have to understand a sentence. Sentence processing is not simply a matter of resolving the multitude of ambiguities that can occur in the language. It is about figuring out who or what the individual participants are in the real-world situation being described. It is about figuring out what roles are being played out, and consequently which roles should be assigned to which participants. The kinds of ambiguity we have dealt with so far complicate this whole process because they make it hard to figure these things out. But they do need to be figured out, and it is to how we might do this that we now turn.
Who did what, and to whom?
As children, we work hard to learn about words and the different ways of combining them. As adults, we rarely stop to think about the meanings of the individual words and about how we unscramble their combination. We do not notice that, despite a lifetime spent learning about meaning and grammar, we often ignore both. We use words with no meaning and we ignore grammatical convention.
Words have meanings in much the same way that objects have meanings. We use the objects around us in different ways according to how they work and what they are for. A car, for example, has various things that enable us to use it in a particular way; things like its accelerator, clutch, brakes, gears, and so on. These things contribute to whatever it is that makes a car a car-they contribute to its meaning. But when we drive that car, we need more than just its accelerator, clutch, brakes and gears. We need also a set of conventions to tell us which side of the road to drive on, what to do at a red light, where not to park, and so on. These conventions are the equivalent of a grammar. So to now say, returning to language, that we ignore meaning and we ignore grammar is to say that we do the linguistic equivalent of driving on the wrong side of the road, through red, with no brakes!
Perhaps that is a slight exaggeration. We do not use words that have no meaning. We just use words whose meanings can change-as if the car had pedals, but what the pedals did could change. And we do not ignore grammar. We just anticipate grammatical conventions before knowing for sure which ones are actually being used-we go through the lights without checking they are green. It all sounds rather dangerous. How do we manage? Which are these words that can take on different meanings? And why the grammatical recklessness?
What we do when we interpret a sentence can be described very simply: we establish who is doing what to whom. If the sentence is a theatre stage, the who and the whom are the actors, and the what determines their roles. If we identify the actors, and identify which particular one is doing what to which other, then we have established the whodid-what-to-whom. Within the sentence, the nouns describe the actors, and the verb conveys the roles being acted out. The grammar, of course, tells us which roles should be assigned to which actors. The problems start to arise when we use words which, on their own, do not describe anything.
In the previous chapter we encountered the ambiguity introduced by the word `that'. But if ambiguity means that something has more than one meaning, what are the different meanings of `that'? Its meanings are, in fact, pretty hard to tie down. As we saw in Chapter 7, it is a word whose function, sometimes, is to signal how to interpret subsequent words, either as a relative clause ('the car that he drove'), or as part of a message ('he told her that it was superb'). In these cases, it really has no meaning of its own.
But what about when it is used as an alternative to `this'? Although it still has no really tangible meaning here, what it signals is very different from what it signals in the earlier cases.
Words like `this' and `that' are used to direct attention to somewhere, and they are sometimes referred to as pointing words. As such, they highlight one of the most important aspects of language-the fact that many of the words in the language refer to things. Pointing is one way to convey what thing is being referred to, but when we speak, we rarely point (except when giving directions), and when we write, pointing is out of the question. Generally, this is not a problem, because words have meaning and the meaning of a word determines what it refers to. In fact, using words can be a lot better than pointing-saying `Sam bit his tongue' is a lot more specific than saying `Sam bit' and pointing in the vague direction of his head. So words like `tongue' are great for referring to things out there in the world. If there happens to be more than one thing out there that has that name, then we can use additional descriptions, like relative clauses, to indicate which specific thing we intend. So we could say `The baby who's wearing the pretty frock' to indicate which baby (`who', unlike `that', unambiguously signals a relative clause). But sometimes we need to be able to refer to someone or something that we have just been talking about. In these cases, what could we possibly point at even if we wanted to point? Instead, we use pronouns such as `he', `she', `it', `they', and so on. Sounds simple enough. But, as usual, it is not.
Why `he' is not `himself'
In a language like American Sign Language, and other sign languages of the world, the equivalent to using a pronoun is to do something that really is very much like pointing. When one introduces someone into the topic of conversation, the sign for them is made in a particular location of the space in front of the signer. This individual can then be referred to by pointing to that location. But another way of referring to them, for instance as having done something to someone else, is to start the sign for the appropriate verb in the location at which that first individual was introduced, and to finish the sign for that verb in the location at which the second individual (to whom the something was done) was introduced. In this respect, sign languages have the advantage over spoken and written languages. And a part of this advantage is the ability to explicitly represent in space the individuals being referred to. This avoids the kinds of ambiguity that we encounter in sentences like `The president discussed his resignation with him'. Who did the resigning?
But so what? Surely we normally speak and write in such a way as to avoid these kinds of ambiguity? How often do we really say things like The president discussed his resignation with him' without first being absolutely clear about who had resigned and who the president was having his discussion with? How often do parents give each of their children the same name? Not often, because no matter how much they like a name, they probably have an inkling of the confusion they would cause. But does this reflect some startling insight into human nature? Probably not. But the way in which we use pronouns does offer some interesting insights into, if not human nature, human language.
A pronoun like `him' can be used either to refer to someone physically present, or to refer to someone already mentioned in the conversation. Why a pronoun like `him', or even `he', is used instead of an expression like `the president' (if that is who the pronoun is referring to) will be discussed in more detail in the next chapter. Here, we shall be concerned more with what pronouns cannot refer to. A pronoun cannot refer to just anyone already mentioned. In `The president wrote him a letter', the pronoun `him' cannot refer to the president. They have to be different people. In a sentence like `The president wrote himself a letter', the reflexive pronoun `himself must refer to the president. But in `A friend of the president wrote himself a letter', the reflexive pronoun must not refer to the president. The facts surrounding what a reflexive pronoun can and cannot refer to are not too surprising given what it means to use such a pronoun. Generally, it indicates that an action was self-oriented-done to oneself-which means that the pronoun should only refer to whoever did the doing. And in a sentence like `The attorney-general said the president's letter to himself was a fake', the pronoun refers to the president simply because what has been done is a letter, and who did the letter is the president.
So much for reflexive pronouns. On the face of it, they look quite easy. Nonreflexive pronouns like `him' or `her' can refer to just about anything, but are other-oriented; `him' cannot refer to the president in `The president wrote him a letter'. It turns out that if you replace a reflexive pronoun ('himself) by an ordinary nonreflexive one ('him'), the `him' version can never refer to the same person that the `himself version would have referred to. This is not just a fact about English. It is apparently true of every language that has both kinds of pronouns.
Intermission: innateness revisited
For years, linguists have searched for properties of language that were universal to all languages. If such properties existed, they would most likely reflect something that was common to the minds of the speakers of all those different languages. The argument has been taken further. Many linguists (and psycholinguists) believe that these universal properties of language must reflect universal and innate knowledge about language. Only in this way could each different language have some piece of itself that was the same as an equivalent piece in every other language. The him/himself facts are a good example of this kind of reasoning. Many linguists believe that these facts have to be explained in terms of the child having innate knowledge that reflexive pronouns (like `himself) must do something different from what nonreflexive pronouns (like `him') do. Many go further and believe that the child has innate knowledge of what each kind of pronoun does. That is, of what each kind of pronoun can and cannot refer to (the child does not know that `him' is nonreflexive, or `himself reflexive, but has innate knowledge of these types and of what these types should or should not refer to). Some linguists go so far as to suggest that even though children get the facts wrong for quite a while, they still `know' the facts-there just happen to be various good reasons why they might ignore them.
Children take a long time to learn the correct facts about what pronouns can and cannot refer to. Errors on both kinds of pronoun can be found up until around 10 years of age, although children tend to get pretty good at them by between four and six years (and they learn about what the reflexive `himself'-like pronouns can refer to sooner than they learn about what the nonreflexive `him'-like pronouns cannot refer to). When children do make mistakes, they might interpret `The monkey is tickling him' to mean the monkey is tickling himself.
We shall come back to children in a moment. For now, what counts is that, although pronouns may seem like things of little consequence, they present to linguists and psycholinguists alike an enduring puzzle whose solution is the source of considerable controversy and debate.
Ambiguity, pronouns, and children's instincts
So `him' and `himself cannot refer to the same things if they are in the same position within the sentence. In all likelihood there is an innate cause for this, as we shall see.
For every rule there is an exception. In Frisian, a language from the North Netherlands, the equivalent of `him' can, on occasion, be used like 'himself. However, Frisian allows this kind of thing to happen only in those cases where no ambiguity would ensue. Might this be a hint as to why `him' and 'himself' annot be swapped around with impunity?
Nonreflexive pronouns like `him', as we have seen, can refer to just about anything, and it is often unclear which person or thing they are supposed to refer to. Languages tolerate this kind of uncertainty (that is, their users do). But if we tolerate the ambiguity in `The president says he voted for him', why do we not also tolerate the ambiguity, and let `him' refer to the president, in `The president voted for him'? That is the real puzzle-we tolerate all kinds of ambiguity except this one. And what is surprising is that `The president says he voted for him' contains many more possibili
ties regarding who the `he' and `him' could refer to than the sentence `The president voted for him' would contain if `him' could refer to the president. So by excluding this kind of ambiguity from the language we save absolutely nothing. We are still left with having to figure out where this constraint on what these pronouns can and cannot refer to comes from. Again, why does it appear to operate across different languages?
We can really only speculate at this point. Perhaps the linguists who claim that it is all innate are right. But perhaps children are the key to a slightly different solution. Perhaps the facts concerning these pronouns stem from the instinct described in Chapter 4 which generally prevents children from associating two different words with the same thing. When learning new words, children assume that if they already have a word for something, a new word should not refer to that same something. Pronouns are just the same. `Him' cannot refer to the same thing as 'himself'. Perhaps children already suspect this. In this sense, perhaps the him/himself facts do after all stem from innate factors, whose origins can be traced, perhaps, to properties of associative learning. And perhaps children take a long time to get the facts straight because the meaning of a pronoun, what it can refer to, is changing the whole time. They are chasing a moving target.