Lecture 1 Ling 4/503 Aug. 27, 2002

- language processing and production takes place without conscious thought.

- think of ideas, and lexical items which will encode them as precisely as possible, but don’t have to think about i) how to pronounce or understand them (how to move mouth, lips, tongue, etc. on the one hand, or how to match words and bits of words up with meaning, on the other) ii) how to create sentences out of lexical items which express what you want to say, or how to parse the incoming flow of words into a structured representation for interpretation. For example, you don’t have to wonder, how do I form a question? create a plural? create a relative clause? indicate that the object is the topic of the sentence?

- to approach it scientifically, think of language as a found system. Imagine that you ran across a species of prairie dogs that lived communally and appeared to coordinate and organize their behavior and pass on information by producing a series of barely-differentiated clicking noises. Just because you happen to be a prairie dog yourself doesn’t mean you can’t investigate this phenomenon impartially --- indeed, you’ve got an advantage; you can try to figure out things about your own clicking behavior, and ask your buddies what they think about their clicks.

- the part of the system that seems, in a way, the most mysterious and the least conducive to external study is the system that structures sounds into words and sentences. Thought itself has been investigated for centuries by philosophers (and now, from a different perspective, by psychologists). It seems, prima facie, obvious how to study the system that actually produces and hears linguistic signals: we can study the actual movements of the lips and tongue, and the neural reflexes of these actions, as well as the actual neural reflexes of linguistic stimulation of the auditory nervous system, because these are the parts of language production that are out there in the world, physical systems that we can conduct experiments on. (Where does the tip of the tongue contact the palate in English speakers? hook ‘em up to some equipment and find out. )

- in another way, though, the system for producing structures is available for experimental study in a purely behavioral sense. Such a system exists, and is demonstrably in the mind/brain of speakers. So we can examine some of the output of the system and develop a model of what operations had to occur to produce the particular output. When we have such a model, we can consider the set of possible outputs the model can generate, and attempt to see whether or not they match the set of possible outputs the system in a speaker’s head generates. (That is, we can make a hypothesis – if the mental system works like this model, then it should produce a sentence like this – and test it on a speaker – Hey, does this sentence sound like something you could theoretically say? If it does, what would it mean if you said it?).

Grammaticality judgement examples:

1. a. Who did you see Mary with?

b. *Who did you see Mary and?

2. a. If you don’t know the meaning of a word, look it up in a dictionary.

b. *If you don’t know the meaning of a word, look up it in a dictionary.

(but contrast:

c. I looked the word up in a dictionary.

d. I looked up the word in a dictionary.)

3. (ambiguity)

a. Sam loves you more than Jim.

(also:

b. Sam loves him. (him ‚ Sam)

c. Sam loves himself (himself must = Sam)

d. Sam thinks Bill loves him. (him can = Sam)

e. Sam thinks Bill loves himself (himself ‚ Sam)

- This is qualitatively identical to any other kind of scientific experiment. (A good analogy might be Mendel’s model of the genetic system of pea plants: he knew there was a system in there that dictated whether he’d get a tall or short plant, or wrinkled or smooth peas, and he could create a model of that system that predicted the result when he mated two plants of different types, as long as he knew the types of their parent plants. His actual model didn’t say anything sensible about the actual mechanisms, which were discovered buried within cells many many years later, and which are still being investigated, but he did discover a great deal about how those mechanisms, once discovered, would have to operate.) The syntactician’s problem, albeit more complex, is similar: we need to create a model of a system that can generate (predict) all and only the sentences of a human language (descriptive adequacy) – and, ideally, a system that with minimal variations in well-defined areas can generate all and only the sentences of all human languages. Not only that, the variations which create the differences between languages must be learnable in 2-4 years by human children. (A theory that can answer these and other "why" questions about the way language works is said to have explanatory adequacy). This will help some smart persons to eventually figure out how the system is neurologically instantiated.

- Back to the question of the data, and what the system is intended to model. Obviously if someone has a lisp, and can’t pronounce the s sound correctly, and yet can distinguish between the words ship and sip when pronounced by someone else, our model of their grammar will include the s/sh distinction, and we’ll ascribe their failure to pronounce it to the failure of a physical system outside the grammar. Similarly, if someone says, otherwise coherently scolding a student, "You have tasted the whole worm," we will not conclude that they’re crazy, but rather that there’s been a mixup in the phonological component, and that the sentence they intended to pronounce was You have wasted the whole term. All kinds of things can go wrong in production and comprehension, but we don’t want our grammar to produce all the sentences that all people might produce under any scenario, including misspeakings, self-interruptions, etc. Those types of sentences can clearly be ascribed to the operation of forces outside the grammatical system. Rather, we want to model the system as it would operate independently of other systems. (Again, the analogy with genetics is appropriate: if Mendel forgot to water a pea plant that he had every reason to expect would be a tall plant (e.g. it had tall parents which each had tall parents), and it turned out to be short, he wouldn’t throw out his model of the genetic system because of that plant. Similar examples can be multiplied from other scientific domains – for example, the observations that led to the theory of gravity as a constant force operating on all bodies equally is not contradicted by the difference between a falling feather and a falling stone; independent forces can be shown to create that difference). For some reason, however, in linguistics, this difference has its own name: the competence/performance distinction. (In genetics, it might be called the nature/nurture distinction).

The way Minimalism is sometimes presented, it’s as if linguists in the 1980s had forgotten how to use O’s razor. That might be true to some extent, but it’s far from the main point of the approach. It’s a much more radical program than that. The central question of minimalism has to do with the idea that language might be "perfect" in some sense. That is, let’s assume that the central system that generates linguistic structure is governed by a few, very simple principles, and that apparent complications or ‘imperfections’ are the result of the interaction of those principles with other systems of the mind/brain. In a way, the way Chomsky puts the question, it’s asking, what if language is an emergent system, like Fibonacci number sequences or the six-sided structure of snowflakes. This notion of perfection won’t exactly leap to the eye in the theoretical model that we will be learning to work with, but it’s worth keeping in mind, as when you go on to read Chomsky it’ll be ever-present.

If children learn language by "imitating", the things that they’re imitating are not specific examples of language use, but rather the rules which adults use to generate their own language use. We can see this because a) children produce sentences which are completely novel ("The moon is gigantic!") and b) they make mistakes which never occur in the input, but are exactly the kind of thing that misapplication of rules would produce:

4. *Daddy goed to the store

*She hitted me.

*Is I can do that?

*Is Ben did go?

*Is the apple juice won’t spill?

Now, think about the problem of ‘imitating’ rules. No adult speaker consciously knows them. ("Yes/No questions in English are formed by moving the auxiliary to the front of the sentence, dear!") Most adult English speakers don’t consciously know what auxiliaries are, or noun phrases or verb phrases, but children's brains obviously do, since they figure out rules that manipulate these things and not others. In fact, there’s certain kinds of rules that they never try.

5. Memories will fade away.

Will memories fade away?

A. (Construct a declarative, then) move the second word in the sentence to the front.

6. Memories of happiness will fade away.

*Of memories happiness will fade away?

(let’s say the child noticed that words like will, can, did, is, should etc. are a natural class, and behave similarly to each other. We’ll call the natural class auxiliaries):

B. Move the first auxiliary in the sentence to the front.

7. The dog that can walk on its hind legs was hit by a car today.

*Can the dog that walk on its hind legs was hit by a car today.

Was the dog that can walk on its hind legs hit by a car today?

C. Move the main auxiliary in the sentence to the front.

8. Down will come cradle, baby and all.

*Will down come cradle, baby and all.

Will cradle come down, baby and all?

9. Into the room will saunter a man holding a rose.

*Will into the room saunter a man holding a rose?

Will a man holding a rose saunter into the room?

10. Sue hates carrots. Beans she might like if they were prepared properly.

*Might beans she like if they were prepared properly?

Might she like beans if they were prepared properly?

D. Move the main auxiliary in the sentence in front of a preceding noun phrase which functions as the subject of the sentence.

In fact, children never seem to go through stages of trying A, B or C as possible rules. They do try treating auxiliaries as question markers (see examples in 4, e.g.), but once they figure out that English doesn’t use a question marker per se, but rather juggles other parts of the sentence around, they don’t bother trying any of our possible permutations above. Why not? The only possibility is that they already know enough about the way language works that they know such rules would be nonsense. That is, there’s an innate component of grammar that predisposes them to look for things like auxiliaries and noun phrases, and not to try rules that involve word-counting, for example.

So, let’s say that the component of grammar that predisposes them to look for auxiliaries and noun phrases (i.e. for cateogries) is invariant across all human children. It’s a principle of the system. Nonetheless, the syntax of different languages differ in other ways. (Most variation between languages is in simple signe/signifié correspondence, of course, but grammatical differences also exist. English is not just French with different morphemes, and Japanese is not just Latin with different morphemes). Children must look for the structures that vary across languages too. Again, the idea/hope is that there are only a few of these things, and they may vary only in a limited number of ways. These things are called parameters (hence Principles and Parameters model of language acquisition).

examples of parameters:

headedness

11. English vs. Japanese:

a) (John) ate pizza Verb-Object

(Taro) pizza-o tabe-ta Object-Verb

(Taro) pizza-ACC eat-PAST

b) to the party Preposition-Object

paatii-e …. Object-Postposition

party-to

c) …the college swhich the eldest son got into… Noun-Relative Clause

…gotyoonan-ga gookakusare-ta daigaku-o…. Relative Clause-Noun

eldest.son-NOM pass.into-PAST college-ACC

PP (English: P-first, Japanese, P-last)

VP (English: V-first, Japanese, V-last)

NP (English: N-first, Japanese, N-last)

null subjects

French vs. Italian

12. a) Maria parla francese

Maria speaks French

b) Marie parle français

Maria speaks French

c) Parla francese

(she) speaks French

d) *Parle français

(she/he) speaks French

wh-movement

English vs. Chinese

13. a) I think he will say that he is pleased.

b) What do you think he will say ____ ?

c) Ni xiangxin ta hui shuo shenme?

You think he will say what

polysynthesis

English vs. Navajo