Handout Syntax
Handout Syntax
Fall 2008
Lexical Categories
We have three tests to determine whether a word is a noun, verb, or an adjective: meaning objects, abstract concepts, individuals actions, sensations, states of being descriptions of nouns morphology plural, possessive past, progressive, 3 SG present comparative, superlative syntactic distribution She can The (it).
is/are here. .
Looking at meaning is a quick way to estimate the lexical category of a word, but its not always sufcient. For example, see, perceive, behold, look, view, and glance have very similar meanings, but see, perceive, and behold can only be used as verbs, while the other three can be used as verbs and as nouns. This is where morphology comes in handy, but morphology can sometimes be irregular. For example, some nouns have irregular plurals (men, oxen ) and some dont seem to even have plurals at all (*knowledges, *furnitures ). Similarly, some verbs have irregular past tenses (had, taught ), some of which are not distinct from their bare form (put, set ). Syntactic distribution can help conrm the lexical category of a word, but this has to be done carefully. Some of the resulting sentences used to test the lexical category may have odd meanings that make it difcult to tell if the problem is in the syntax or the semantics, so test sentences need to be constructed carefully. In most cases, a combination of all three tests will usually successfully reveal the lexical category of a word, but of course, if a word cant be classied by the tests above, we need some other way of knowing what lexical category it is. For example, there arent tests exactly like the ones above for prepositions.
Phrase Structure
Once we know what lexical categories the words of a sentence belong to, we can compile the possible ways these lexical categories can be combined with each other to form grammatical sentences of the language. In class, we discovered that English sentences could have a variety of patterns, including: NV Det N V ANV Det A N V NVN N V Det N NVAN N V Det A N etc. Boars ate. The boars ate. Angry boars ate. The angry boars ate. Boars ate humans. Boars ate the humans. Boars ate dead humans. Boars ate the dead humans.
More generally, we can write S (Det) (A) N V ((Det) (A) N), meaning that a sentence can be composed of an optional determiner (words like the, a, and some ), an optional adjective, a noun, a verb, and an optional sequence of an optional determiner, optional adjective, and noun.
There is a repeated pattern of (Det) (A) N occurring in two separate places in the sentence, with the three elements in exactly the same order, with exactly the same pattern of optionality. This is unlikely to be a coincidence, so it we want to have a rule that encodes this pattern. These three elements together constitute what is known as a noun phrase, and extracting this predictable NP pattern out into its own rule, we get: S NP V (NP) NP (Det) (A) N Further exploration of English sentences reveals repeated patterns for adjective phrases and for prepositional phrases, too, giving us a fuller set of phrase structure rules for English: S NP AP PP NP (PP) V (NP) (AP) (PP) (Det) (AP) N (Deg) A P NP
Constituents
So far, we have only come up with phrase structure rules to describe repeated patterns of words. It would be nice to nd some independent support for these rules, some other evidence that shows that a noun phrase really is a separate unit, a constituent. There are at least three tests that show how constituents behave as a unit, while other strings of consecutive words do not. Not only do these tests conrm the existence of NPs, APs, and PPs, but also suggest the existence of VPs. Coordination Test: Constituents of the same type can be joined together with conjunctions like and and or, while non-constituents cannot be joined in such a way. Clark lied to [a bald billionaire] and [a blonde reporter]. Clark [lied to a bald billionaire] and [revealed his secret to a blonde reporter]. *Clark lied to [a bald billionaire] and [protect his secret]. NPs VPs NP = VP
Substitution Test: Instead of repeating a complete constituent in a separate location, it may be replaced in its entirety by an appropriate substitute word (e.g., pronouns for NPs, do so for VPs, and there for locative PPs), while non-constituents (or incomplete constituents) cannot be substituted. A painting of an explosion covered the oor, and it spurred everyone to action. *A painting of an explosion covered the oor, and it of the cop was on a canvas. The older son saved the world if the younger son did so. *The older son saved the world if the younger son did so the cheerleader. NP not complete VP not complete
Movement Test: Under certain circumstance, a complete constituent can be moved to other parts of the sentence, sometimes with slightly different syntax, while non-constituents cannot be moved. There are a variety of movement tests (fronting and passives, for examples). Demonstrated here is pseudo-clefting, where the constituent tobe tested appears at the beginning of the sentence, followed by a statement of the form is/are who/what/where, and the rest of the sentence, with the constituent no longer in its usual position. (Note that pseudo-clefting of VPs requires do -insertion.) The phone is what Pam and Ryan answer *Phone is what Pam and Ryan answer the . NP not complete VP not complete . NP not complete
Answer the phone is what Pam and Ryan do . *Answer is what Pam and Ryan do the phone. A book of tasteless jokes is what Michael has *A book is what Michael has of jokes.
We can visualize the phrase structure of a sentence in a tree format, which is often easier to interpret than a purely linear bracketed, especially as the sentences get more complex. S NP N boars V ate Det the AP A dead [S [NP boarsN ] [VP ateV [NP theDet [AP deadA ] humansN ] ] ] N humans VP NP
Universal Patterns
We nd that this basic phrase structure seems to be the same across languages, except that the order is different. For example, NPs always seem to contain Det, AP, N, and PP, but not necessarily in the same order as in English. The phrase structure tree diagram can then be thought of as a mobile, with all of the parts strictly connected at the nodes in the tree, but with the individual branches able to freely swing around each other, but staying conned to their constituents. For example, while English has SOV word order, some languages like Hixkaryana have OVS word order, with the main subcomponents of S of VP completely reversed, though the components themselves appear in the same constituents (the subject NOP and the VP are both in directly under the S node, and the verb and object NP are both directly under the VP node). English SVO S NP Det the N jaguar V ate Det the VP NP N man NP N toto
man
NP N kamara
jaguar
The jaguar ate the man. *The man ate the jaguar.