Читать книгу Syntax - Andrew Carnie - Страница 19

Consider sentences such as (14). Native speakers will have to read this sentence a couple of times to figure out what it means.

14) # Cotton shirts are made from comes from India.

This kind of sentence (called a garden path sentence) is very hard to understand and process. In this example, the problem is that the intended reading has a noun, cotton, that is modified by a reduced relative clause: (that) shirts are made from. The linear sequence of cotton followed by shirt is ambiguous with the noun phrase cotton shirts. Note that this kind of relative structure is okay in other contexts; compare: That material is the cotton shirts are made from. Sentences like (14) get much easier to understand with really clear pauses (where … is meant to indicate a pause): Cotton … shirts are made from … comes from India. Or by insertion of a that which breaks up the potentially ambiguous cotton shirts sequence: The cotton that shirts are made from comes from India. What is critical about these garden path sentences is that, once one figures out what the intended meaning is, native speakers can identify them as acceptable sentences or at the very least as sentences that have structures that would otherwise be acceptable in them. The problem for us as linguists is that native speakers have a really hard time figuring out what the intended meaning for these sentences is on those first few passes!

A similar situation arises when we have really long sentences with complex syntactic relations. Look at (15). A first reading of this sentence will boggle your average speaker of English. But if you read it a couple of times, it becomes obvious what is intended. In fact, the sentence seems to be structured grammatically.

15) Who did Keisha say Monique claimed that Suzanne seems to have been likely to have kissed?

The reason this sentence is hard to understand is that the question word who is very far away from where it gets interpreted (as the object of kiss), and what lies in between those two points is quite a lot of sophisticated embeddings and structure. But once you get a chance to think about it, it gets better and better as a sentence. The most famous example of this kind of effect is called center embedding. English speakers tolerate a small amount of stacking of relative clauses between subjects and verbs, so (16) – while a little clumsy – is still a good sentence for most speakers of English. We have some cheese, the kind that mice love, and it stinks. If you have trouble with this sentence put a big pause after cheese and before stinks.

16) Cheese mice love stinks.

But no pauses will fix a sentence in which we put another reduced relative right after mice, with the intended meaning that cheese which is loved by mice who are caught by cats is stinky:

17) #Cheese mice cats catch love stinks

This sentence is essentially uninterpretable for English speakers. Chomsky (1965) argued that the problem here is not one of the grammar (as English grammar allows reduced relative clauses after subjects and before verbs), but instead either a constraint on short- term memory⁷ or a constraint on our mental ability to break apart sentences as we hear them. The English parsing system – that is the system that breaks down sentences into their bits for comprehension – has certain limits, and these limits are distinct from the limits on what it means to be “grammatical”. Sentences (14), (15), and (16) are unacceptable to native speakers in a qualitatively different way than the ones in (13).

The distinction we’ve been looking at here is often known as the competence/performance distinction. When we speak or listen, we are performing the act of creating a piece of language output. This performance can be interrupted by all sorts of extraneous factors: we can be distracted or bored; we can cough or mumble our words; we can forget what we had previously heard; the noise of the bus driving past can blot out a crucial word. Performance refers to the kinds of language that are actually produced and heard. Competence, by contrast, refers to what we know about our language; it is unimpeded by factors that might muddy the waters of performance. So, think about the really long complicated sentence in (15). The first time you read it, things like your memory and how complicated it was interfered with your ability to understand it. So the initial unacceptability of the sentence was due to a performance problem. But once you thought about it and stared at it a bit, you saw that it was actually a fairly standard grammatical sentence of English – just a really complicated one. When you did this, you were accessing your competence in (or knowledge of) English grammar. If syntax is part of cognitive science which is about what we know, then we should probably be most interested in competence.

This takes us to a new point. Listen carefully to someone speak (not lecture or read aloud, but someone really speaking in a conversation). You’ll notice that they don’t speak in grammatical sentences. They leave stuff off and they speak in fragments. They start and they stop the same sentence a couple of times. Everyone does this, even the most eloquent among us. So much of what you hear (or see in spoken language corpora) consists of actually “ungrammatical” forms. Nevertheless, if you’re a native English speaker, you have the ability to judge if a sentence is acceptable or not. These two tasks, understanding spoken conversational language and being able to judge the well- formedness of a sentence, seem to actually be different skills corresponding roughly to performance and competence.

An analogy that might clarify these distinctions: imagine that you’re a software engineer and you’re writing a piece of computer code. First you run it on your own beautiful up-to-date computer and it behaves beautifully. The output of the computer code is one kind of performance of the underlying competence. Then you run it on your little sister’s ancient PC. The program doesn’t perform as you expect. It’s really slow. It crashes. It causes the fan to run continuously and the processor to overheat. Now you go back and look at the code. There are no errors in the code. It meets all the requirements of the computer language. So, from the perspective of competence, your program is okay. The real problem here is not with your code, but with the machine you’re running it on. The processor is too old, there isn’t enough memory and you have a computer that tends to overheat. These are all performance problems.

What does this mean for the linguist using acceptability judgments as a tool for investigating syntax? It means that when using a judgment, you have to be really clear about what is causing the acceptability or unacceptability of the sentence. Is the sentence acceptable just because you have gleaned enough information from the conversational context (in which case we might consider it a performance effect)? If you hear a sentence that you judge as unacceptable, is it because someone was speaking too quickly and left out a word, or is it because the sentence really doesn’t work as an English sentence at all? This distinction is very subtle, but it is one that syntacticians have to pay careful attention to as they do their work.