LX CAS 522 Syntax I
Week 3a. Categories, features, natural classes, and morphology.
Previously, in LX522...
n n
n
So, here's where we were. We're trying to characterize our knowledge of syntax, using English speaker's knowledge of English as a window to the kinds of things we need to describe language. Words seem to come in categories (N, V, A, P, C, I, PRN, D, ...). English treats these differently, so to describe English, our theory needs to treat them differently.
Previously, in LX522...
n
The surface
n
n n n
We've collected a number of features that seem to make a difference in various ways. We have the category, e.g., [+N], [+V]. But we seem to have subcategories too, e.g., [+Pl], [+Common], [+Count]. These features matter to how we can combine words. Language cares, and so, therefore, do we.
n
n n
Before going on, let's take a detour, because if we don't, things are just going to get confusing. So far, we have looked at a word, and tried to determine what its relevant features are. Books [+N, +Count, +Common +Pl]. Written [+V, +Participle, +Perfect]
Where's the feature?
n
Why is this so confusing?
n
This is useful in that we get a hint as to what features are required. But consider:
n n n n n n
So, do seems to be:
n n n
Bill ate lunch. Bill will eat lunch. Bill did not eat lunch. Bill does not eat lunch. You do not eat lunch. We do not eat lunch.
[+Pres], and not [+3, +Sg], and only shows up in the negative.
n n n
n
What are the features of ate? eat? did? does? do?
n
Ouch. That's rather inelegant. Here's the problem: I sat on the bank. I saw the candidate with the binoculars. Visiting relatives can be tedious. See?
1
We're going the wrong way
n
Generative grammar
n n n
n n
There are two different intents underlying Visiting relatives can be tedious (so I do it as little as possible), and Visiting relatives can be tedious (so I avoid them as often as possible). They happen to sound the same, but they have a different underlying structure. In general, what's unique is the underlying intent/structure, not the pronounced form.
The syntactic system we are going to build is a generative grammar. It builds up an underlying structure, which is then pronounced. The two versions of Visiting relatives can be tedious are different sentences. ...But wasn't our goal to explain how people could tell if sentences they hear are part of their language or not?
n
Judging sentences
n
A very, very little bit of French
n
The view of sentence judgment we'll adopt here is basically one of asking oneself: Could I say that sentence? When listening to somebody, you of course need to decode what that person meant, but it is a process of recovering the underlying form of their utterance.
If you've tried to learn any French at all, you've come across this phenomenon:
n n n
n
n n n n n
de `of' le `the (masculine)' `at' la `the (feminine)' la biblioteque `to the library (fem)' * le cinma `to the movies (masc)' au cinema `to the movies (masc)' de la mayonnaise `of mayonnaise (fem)' de le lait of milk (masc) du lait `of milk' (masc)
A very, very little bit of French
n
What does this have to do with eating lunch?
n
This is usually taught as:
au = + le n du = de + le
n
And now we can return to the point:
n n n n
n
If your underlying intent is `at' + le `the', you pronounce it like au.
n
Bill ate lunch. Bill eats lunch. Bill does not eat lunch. Bill will (not) eat lunch.
n
So is au a preposition or an article?
What generalizations can we come up with here? How are the features organized in these simple sentences?
n
Why did I juxtapose +le=au from French with Bill ate lunch? Where is tense? Where is the verb? What is ate? What does do mean?
2
Falling into place
n
Falling into place
n
If we suppose that these sentences all have the same form... Subject Tense/Agreement (Not) Verb Object ...things start to look a lot more regular, describable. This is the structure of a sentence.
n
Moving one step closer to syntactic structure: [NP Subject] I (not) V [NP Object]
That Tense+Verb comes out as "Tensed Verb" is a matter of pronunciation. If you separate Tense from the Verb with not, they no longer can combine. In order to pronounce Tense, you insert do.
n n n
So [Past, 1, 2, Pl] are features of I. [1, 2, Pl, Common, Count, ...] are features of N. [+V] is a feature of V.
Those pesky participles
n
Crosscategorial features
n
Bill will have been eating lunch [NP Bill] [I will] [V have] [V been] [V eating] [NP lunch] will [+Fut] have [+Vaux] been [+Vaux, +Participle +Perf] eating [+V, +Participle, -Perf]
Consider what un can attach to.
untie, unfold, unwrap, unpack unhappy, unfriendly, undead n *uncity, *uncola, *unconvention n *unupon, *unalongside, *unat
n n
n
n n n n
n
Basically, it applies to reversible verbs and adjectives, but not to nouns or prepositions. How can we state that?
Crosscategorial features
n
Crosscategorial features
n
Suppose that nouns and verbs are the most basic categories. A noun is a noun and not a verb, and verb is a verb and not a noun.
n n
Noun: [+N, -V]. Verb: [-N, +V].
Looked at this way, adjectives are kind of "verby" in that they are also attributing properties.
n
n
A possible conceptual reason to separate nouns and verbs is that verbs are basically predicates-- they attribute some property to the noun. Nouns are basically arguments, to be assigned properties by verbs.
It's hard to make that really precise, but we have a more concrete syntactic similarity between verbs and adjectives too: both can take un-, while nouns and prepositions cannot.
3
Supercategories
n
Supercategories
n
Chomsky (1970) proposed that we explain this by supposing that [N] and [V] are the two basic features that determine the four lexical categories (N, V, A, P).
n n
So, un attaches to a [+V] category. It doesn't care about [N]. [+V] defines a natural class that language refers to. Why is A [+V, +N] and P [-V, -N]?
Suppose we had a morpheme that attaches just to V and P, how could we state that? n Do V and P form a natural class too?
n
N: [+N, -V] P: [-N, -V]
V: [-N, +V] A: [+N, +V]
n
n
Given that, what does un attach to?
Russian Case
n
Functional and lexical
n n
Other languages can give us evidence of natural classes as well. E.g., Russian nouns (all nouns) are marked for Case (like English pronouns are: me vs. I), but when they are modified by an adjective, the adjective is also marked for case. What gets marked for Case in Russian?
koshku v cat pustuyu korobku box
That takes care of N, V, A, P, but what about our functional categories? In fact, the functional categories (C, I, D, PRN) each seem a little like a lexical category.
n
n
n n n
Krasivaya dyevushka vsunula chornuyu beautiful girl put black
in empty
`The beautiful girl put the black cat in the empty box'
Auxiliaries seem a lot like (have, verbs be, do), and inflect like verbs do. Could and can, Would and will might be supposed to differ in tense. Complementizers and infinitival to seem a bit like prepositions (e.g., for, to). Pronouns are kind of nouny. Determiners are a bit adjectivey.
Auxiliary verbs and verbs
n
Auxiliary verbs and I
n
Verbs and auxiliary verbs are subject to some of the same processes.
n
They inflect for tense, they inflect for subject agreement.
n n n n
Suggests: They form a natural class. Suppose there's a feature (say [+F] for "functional") that differentiates them. Both are [+V, -N], but be is [+F], and eat is [-F]. Thus: [+V, -N] inflects for tense and subject agreement.
n n
On the other hand, auxiliary verbs act like elements of category I, appearing in that spot between the Subject and (Not) V. Auxiliary verbs are [+V, -N, +F]. Other elements of category I might be like prepositions, e.g., to. If these are "functional prepositions", then they are [-V, -N, +F]. Do Auxiliary verbs and to form a natural class?
n
4
"Grammatical category"
n
Sentences are made of words?
n n
So what, then is a grammatical category? A grammatical category is a set of elements which have the same value(s) for a given set of grammatical features. It's really a natural class. Category labels like "N", or "Aux" are really just shorthand for feature matrices like [+N, -V, -F], or [-N, +V, +F]. Notationally convenient, but only respected as such by some parts of the grammar.
n
n n
Bill kicked the pail. Bill mailed the letter. Bill did not kick the pail. Bill did not mail the letter. Remembering that +le=au in French, we said that underlyingly, these are: Bill Tense (Not) Verb Object Is Tense a word?
n
n
n
Tense is not a word
n n
Bill will kick the pail
n
In English, past tense is not really a word. It's a morpheme. It's a suffix. Regularly, -ed.
n
Actually, tense can be a word, if it's the future tense will. Note: If would is the past tense of will, then it is probably not correct to think of will as being simply a future marker. Rather, it's one of the modals, an "unrealized" marker, which makes sense as long as time goes invariably forward, as it seems to. Many people nevertheless consider will to be the same category of thing as -ed, so we will for now ignore this complication, since it matters little to what we're going to do.
Of course there are lots of special cases (wrote, fed, drew), but these at least all seem to be modifications of the end of the word.
n
n n
Suppose then, that we have: Bill -ed (not) kick the pail. Of course, you can't pronounce an affix. An definitional property of an affix is that it attaches to a word (of a particular category: past tense -ed is a verbal affix). If forced to pronounce -ed, you insert a meaningless verb to attach it to (do).
Sentences are made of morphemes
n
Possessive 's
n
We will have more success if we assume that sentences are made of things that can be smaller than words. Here's another example: Bill's pail. What is that `s there? What does it mean?
Is 's a suffix?
n n
The man from Australia's hat. The man who left's hat.
n n n n
n
What does 's attach to? The possessive 's is not really an affix, but it's not really a word either. These things usually go by the name clitic. They're like a little word that leans on a nearby word.
5
Clitics
n
Moral: Underlying Surface
n
Plenty of languages have clitics. English has a few.
Isn't that Bill's hat? Yes, that's Bill's hat. n Wouldn't you like a hat like Bill's?
n n
The larger point here is that sentences have two forms: the underlying form (which is our primary concern) and the surface form (which is really where our data comes from).
n
n n
French (again) has them: je pars, *je, moi. Essentially, we want to treat clitics, affixes, and words on a par in the underlying structure-- they differ in pronunciation.
We can deduce things about the structure of the underlying form from the surface form, and by positing abstract elements like affixes, clitics, features, we can describe in a concise (and...