Rabu, 09 Oktober 2019
Bab 2 (2) Syntax: The Sentence Patterns of Language
------------------------------------ HAL 78 ---------------------------------------
In each case the speaker could continue creating sentences by adding another adjective, prepositional phrase, or relative clause. In principle, this could go on forever. All languages have mechanisms of this sort that make the number of sentences limitless. Given this fact, the sentences of a language cannot be stored in a dictionary format in our heads. Rather, sentences are composed of discrete units that are combined by rules. This system of rules explains how speakers can store infinite knowledge in a finite space—our brains.
The part of grammar that represents a speaker’s knowledge of sentences and their structures is called syntax. The aim of this chapter is to show you what syntactic structures look like and to familiarize you with some of the rules that determine them. Most of the examples will be from the syntax of English, but the principles that account for syntactic structures are universal.
What the Syntax Rules Do
“Then you should say what you mean,” the March Hare went on.
“I do,” Alice hastily replied, “at least—I mean what I say—that’s the same thing, you know.”
“Not the same thing a bit!” said the Hatter. “You might just as well say that ‘I see what I eat’ is the same thing as ‘I eat what I see’!”
“You might just as well say,” added the March Hare, “that ‘I like what I get’ is the same thing as ‘I get what I like’!”
“You might just as well say,” added the Dormouse . . . “that ‘I breathe when I sleep’ is the same thing as ‘I sleep when I breathe’!”
“It is the same thing with you,” said the Hatter.
LEWIS CARROLL, Alice’s Adventures in Wonderland, 1865
The rules of syntax combine words into phrases and phrases into sentences. Among other things, the rules specify the correct word order for a language. For example, English is a Subject–Verb–Object (SVO) language. The English sentence in (1) is grammatical because the words occur in the right order; the sentence in (2) is ungrammatical because the word order is incorrect for English. (Recall that the asterisk or star preceding a sentence is the linguistic convention for indicating that the sentence is ungrammatical or ill-formed according to the rules of the grammar.)
1. The President nominated a new Supreme Court justice.
2. *President the new Supreme justice Court a nominated.
A second important role of the syntax is to describe the relationship between the meaning of a particular group of words and the arrangement of those words. For example, Alice’s companions show us that the word order of a sentence contributes crucially to its meaning. The sentences in (3) and (4) contain the same words, but the meanings are quite different, as the Mad Hatter points out.
3. I mean what I say.
4. I say what I mean.
The rules of the syntax also specify the grammatical relations of a sentence, such as subject and direct object. In other words, they provide the information about who is doing what to whom. This information is crucial to understanding the meaning of a sentence. For example, the grammatical relations in (5) and (6) are reversed, so the otherwise identical sentences have very different meanings.
5. Your dog chased my cat.
6. My cat chased your dog.
Syntactic rules also specify other constraints that sentences must adhere to. Consider, for example, the sentences in (7). As an exercise you can first read through them and place a star before those sentences that you consider to be ungrammatical.
7. (a) The boy found.
(b) The boy found quickly.
(c) The boy found in the house.
(d) The boy found the ball.
We predict that you will find the sentence in (7d) grammatical and the ones in (7a–c) ungrammatical. This is because the syntax rules specify that a verb like found must be followed by something, and that something cannot be an expression like quickly or in the house but must be like the ball. Similarly, we expect you will find the sentence in (8b) grammatical while the sentence in (8a) is not.
8. (a) Disa slept the baby.
(b) Disa slept soundly.
The verb sleep patterns differently than find in that it may be followed solely by a word like soundly but not by other kinds of phrases such as the baby. We also predict that you’ll find that the sentences in (9a, d, e, f) are grammatical and that (9b, c) are not. The examples in (9) show that specific verbs, such as believe, try, and want, behave differently with respect to the patterns of words that may follow them.
9. (a) Zack believes Robert to be a gentleman.
(b) Zack believes to be a gentleman.
(c) Zack tries Robert to be a gentleman.
(d) Zack tries to be a gentleman.
(e) Zack wants to be a gentleman.
(f) Zack wants Robert to be a gentleman.
The fact that all native speakers have the same judgments about the sentences in (7) to (9) tells us that grammatical judgments are neither idiosyncratic nor capricious, but are determined by rules that are shared by all speakers of a language.
In (10) we see that the phrase ran up the hill behaves differently from the phrase ran up the bill, even though the two phrases are superficially quite similar. For the expression ran up the hill, the rules of the syntax allow the word
orders in (10a) and (10c), but not (10b). In ran up the bill, in contrast, the rules allow the order in (10d) and (10e), but not (10f).
10. (a) Jack and Jill ran up the hill.
(b) Jack and Jill ran the hill up.
(c) Up the hill ran Jack and Jill.
(d) Jack and Jill ran up the bill.
(e) Jack and Jill ran the bill up.
(f) Up the bill ran Jack and Jill.
The pattern shown in (10) illustrates that sentences are not simply strings of words with no further organization. If they were, there would be no reason to expect ran up the hill to pattern differently from ran up the bill. These phrases act differently because they have different syntactic structures associated with them. In ran up the hill, the words up the hill form a unit, as follows:
He ran [up the hill]
The whole unit can be moved to the beginning of the sentence, as in (10c), but we cannot rearrange its subparts, as shown in (10b). On the other hand, in ran up the bill, the words up the bill do not form a natural unit, so they cannot be moved, and (10f) is ungrammatical.
Our syntactic knowledge crucially includes rules that tell us how words form groups in a sentence, or how they are hierarchically arranged with respect to one another. Consider the following sentence:
The captain ordered all old men and women off the sinking ship.
This phrase “old men and women” is ambiguous, referring either to old men and to women of any age or to old men and old women. The ambiguity arises because the words old men and women can be grouped in two ways. If the words are grouped as follows, old modifies only men and so the women can be any age.
[old men] and [women]
When we group them like this, the adjective old modifies both men and women. [old [men and women]]
The rules of syntax allow both of these groupings, which is why the expression is ambiguous. The following hierarchical diagrams illustrate the same point:
In the first structure old and men are under the same node and hence old modifies men. In the second structure old shares a node with the entire conjunction men and women, and so modifies both.
This is similar to what we find in morphology for ambiguous words such as unlockable, which have two structures, corresponding to two meanings, as discussed in chapter 1.
Many sentences exhibit such ambiguities, often leading to humorous results. Consider the following two sentences, which appeared in classified ads: For sale: an antique desk suitable for lady with thick legs and large drawrs. We will oil your sewing machine and adjust tension in your home for $10.00.
In the first ad, the humorous reading comes from the grouping [a desk] [for lady with thick legs and large drawers] as opposed to the intended [a desk for lady] [with thick legs and large drawers], where the legs and drawers belong to the desk. The second case is similar. Because these ambiguities are a result of different structures, they are instances of structural ambiguity.
Contrast these sentences with:
This will make you smart.
The two interpretations of this sentence are due to the two meanings of smart— “clever” or “burning sensation.” Such lexical or word-meaning ambiguities, as opposed to structural ambiguities, will be discussed in chapter 3.
Often a combination of differing structure and double word-meaning creates ambiguity (and humor) as in the cartoon:
Syntactic rules reveal the grammatical relations among the words of a sentence as well as their order and hierarchical organization. They also explain how the grouping of words relates to its meaning, such as when a sentence or phrase is ambiguous. In addition, the rules of the syntax permit speakers to produce and understand a limitless number of sentences never produced or heard before—the creative aspect of linguistic knowledge. A major goal of linguistics is to show clearly and explicitly how syntactic rules account for this knowledge. A theory of grammar must provide a complete characterization of what speakers implicitly know about their language.
What Grammaticality Is Not Based On
Colorless green ideas sleep furiously. This is a very interesting sentence, because it shows that syntax can be separated from semantics—that form can be separated from meaning. The sentence doesn’t seem to mean anything coherent, but it sounds like an English sentence.
HOWARD LASNIK, The Human Language: Part One, 1995
Importantly, a person’s ability to make grammaticality judgments does not depend on having heard the sentence before. You may never have heard or read the sentence
Enormous crickets in pink socks danced at the prom.
but your syntactic knowledge tells you that it is grammatical. As we showed at the beginning of this chapter, people are able to understand, produce, and make judgments about an infinite range of sentences, most of which they have never heard before. This ability illustrates that our knowledge of language is creative not creative in the sense that we are all poets, which we are not, but creative in that none of us is limited to a fixed repertoire of expressions. Rather, we can exploit the resources of our language and grammar to produce and understand a limitless number of sentences embodying a limitless range of ideas and emotions.
We showed that the structure of a sentence contributes to its meaning. However, grammaticality and meaningfulness are not the same thing, as shown by the following sentences:
Colorless green ideas sleep furiously.
A verb crumpled the milk.
Although these sentences do not make much sense, they are syntactically well formed. They sound funny, but their funniness is different from what we find in the following strings of words:
*Furiously sleep ideas green colorless.
*Milk the crumpled verb a.
There are also sentences that we understand even though they are not well formed according to the rules of the syntax. For example, most English speakers could interpret
*The boy quickly in the house the ball found.
although they know that the word order is incorrect. Similarly, we could probably assign a meaning to sentence (8a) (Disa slept the baby) in the previous section. If asked to fix it up, we would probably come up with something like “Disa put the baby to sleep,” but we also know that as it stands, (8a) is not a possible sentence of English. To be a sentence, words must conform to specific patterns determined by the syntactic rules of the language.
------------------------------------ HAL 83 ---------------------------------------
Some sentences are grammatical even though they are difficult to interpret because they include nonsense words, that is, words with no agreed on meaning. This is illustrated by the following lines from the poem “Jabberwocky” by Lewis Carroll:
’Twas brillig, and the slithy toves
Did gyre and gimble in the wabe
These lines are grammatical in the linguistic sense that they obey the word order and other constraints of English. Such nonsense poetry is amusing precisely because the sentences comply with syntactic rules and sound like good English. Ungrammatical strings of nonsense words are not entertaining:
*Toves slithy the and brillig ’twas wabe the in gimble and gyre did
Grammaticality also does not depend on the truth of sentences. If it did, lying would be impossible. Nor does it depend on whether real objects are being discussed or whether something is possible in the real world. Untrue sentences can be grammatical, sentences discussing unicorns can be grammatical, and sentences referring to pregnant fathers can be grammatical.
The syntactic rules that permit us to produce, understand, and make grammaticality judgments are unconscious rules. The grammar is a mental grammar, different from the prescriptive grammar rules that we are taught in school. We develop the mental rules of grammar long before we attend school, as we shall see in chapter 7.
Sentence Structure
I really do not know that anything has ever been more exciting than diagramming sentences.
GERTRUDE STEIN, “Poetry and Grammar,” 1935
Suppose we wanted to write a template that described the structure of an En glish sentence, and more specifically, a template that gave the correct word order for English. We might come up with something like the following:
Det—N—V—Det—N
This template says that a determiner (an article) is followed by a noun, which is followed by a verb, and so on. It would describe English sentences such as the following:
The child found a puppy.
The professor wrote a book.
That runner won the race.
------------------------------------ HAL 84 ---------------------------------------
The implication of such a template would be that sentences are strings of words belonging to particular grammatical categories (“parts of speech”) with no internal organization. We know, however, that such “flat” structures are incorrect. As noted earlier, sentences have a hierarchical organization; that is, the words are grouped into natural units. The words in the sentence
The child found a puppy.
may be grouped into [the child] and [found a puppy], corresponding to the subject and predicate of the sentence. A further division gives [the child] and then [[found] [a puppy]], and finally the individual words: [[the] [child]] [[found] [[a] [puppy]]]. It’s sometimes easier to see the parts and subparts of the sentence in a tree diagram:
The “tree” is upside down with its “root” encompassing the entire sentence, “The child found a puppy,” and its “leaves” being the individual words, the, child, found, a, puppy. The tree conveys the same information as the nested square brackets. The hierarchical organization of the tree reflects the groupings and subgroupings of the words of the sentence.
The tree diagram shows, among other things, that the phrase found a puppy divides naturally into two branches, one for the verb found and the other for the direct object a puppy. A different division, say, found a and puppy, is unnatural.
Constituents and Constituency Tests
The natural groupings or parts of a sentence are called constituents. Various linguistic tests reveal the constituents of a sentence. The first test is the “stand alone” test. If a group of words can stand alone, they form a constituent. For example, the set of words that can be used to answer a question is a constituent. So in answer to the question “What did you find?” a speaker might answer a puppy, but not found a. A puppy can stand alone while found a cannot.
The second test is “replacement by a pronoun.” Pronouns can substitute for natural groups. In answer to the question “Where did you find a puppy?” a speaker can say, “I found him in the park.” Words such as do can also take the place of the entire predicate found a puppy, as in “John found a puppy and Bill did too.” If a group of words can be replaced by a pronoun or a word like do, it forms a constituent.
------------------------------------ HAL 85 ---------------------------------------
A third test of constituency is the “move as a unit” test. If a group of words can be moved, they form a constituent. For example, if we compare the following sentences to the sentence “The child found a puppy,” we see that certain elements have moved:
It was a puppy that the child found.
A puppy was found by the child.
In the first example, the constituent a puppy has moved from its position following found; in the second example, the positions of a puppy and the child have been changed. In all such rearrangements the constituents a puppy and the child remain intact. Found a does not remain intact, because it is not a constituent.
In the sentence “The child found a puppy,” the natural groupings or constituents are the subject the child, the predicate found a puppy, and the direct object a puppy.
Some sentences have a prepositional phrase in the predicate. Consider
The puppy played in the garden.
We can use our tests to show that in the garden is also a constituent, as follows:
Where did the puppy play? In the garden (stand alone)
The puppy played there. (replacement by a pronoun-like word)
In the garden is where the puppy played. (move as a unit)
It was in the garden that the puppy played.
As before, our knowledge of the constituent structure of a sentence may be graphically represented by a tree diagram. The tree diagram for the sentence “The puppy played in the garden” is as follows:
In addition to the syntactic tests just described, experimental evidence has shown that speakers do not represent sentences as strings of words but rather in terms of constituents. In these experiments, subjects listen to sentences that have clicking noises inserted into them at random points. In somecases the click occurs at a constituent boundary, and in other sentences the click is inserted in the middle of a constituent. The subjects are then asked to report where the click occurred. There were two important results: (1) Subjects noticed the click and recalled its location best when it occurred at a major constituent boundary (e.g., between the subject and predicate); and (2) clicks that occurred inside the constituent were reported to have occurred between constituents. In other words, subjects displaced the clicks and put them at constituent boundaries. These results show that speakers perceive sentences in chunks corresponding to grammatical constituents.
------------------------------------ HAL 86 ---------------------------------------
Every sentence in a language is associated with one or more constituent structures. If a sentence has more than one constituent structure, it is ambiguous, and each tree will correspond to one of the possible meanings. For example, the sentence “I bought an antique desk suitable for a lady with thick legs and large drawers” has two phrase structure trees associated with it. In one structure the phrase [a lady with thick legs and large drawers] forms a constituent. For example, it could stand alone in answer to the question “Who did you buy an antique desk for?” In its second meaning, the phrase with thick legs and large drawers modifies the phrase a desk for a lady, and thus the structure is [[a desk for a lady][with thick legs and large drawers]].
------------------------------------ HAL 87 ---------------------------------------
Each grouping in the tree diagrams of “The child found a puppy” is a member of a large family of similar expressions. For example, the child belongs to afamily that includes the police officer, your neighbor, this yellow cat, he, John, and countless others. We can substitute any member of this family for the child without affecting the grammaticality of the sentence, although the meaning of course would change.
A police officer found a puppy.
Your neighbor found a puppy.
This yellow cat found a puppy.
A family of expressions that can substitute for one another without loss of grammaticality is called a syntactic category.
The child, a police officer, John, and so on belong to the syntactic category noun phrase (NP), one of several syntactic categories in English and every other language in the world. NPs may function as the subject or as an object in a sentence. NPs often contain a determiner (like a or the) and a noun, but they may also consist of a proper name, a pronoun, a noun without a determiner, or even a clause or a sentence. Even though a proper noun like John and pronouns such as he and him are single words, they are technically NPs, because they pattern like NPs in being able to fill a subject or object or other NP slots.
John found the puppy.
He found the puppy.
Boys love puppies.
The puppy loved him.
The puppy loved John.
NPs can be more complex as illustrated by the sentence:
The girl that Professor Snape loved married the man of her dreams.
The NP subject of this sentence is the girl that Professor Snape loved, and the NP object is the man of her dreams.
Syntactic categories are part of a speaker’s knowledge of syntax. That is, speakers of English know that only items (a), (b), (e), (f), and (g) in the following list are NPs even if they have never heard the term noun phrase before.
1. (a) a bird
(b) the red banjo
(c) have a nice day
(d) with a balloon
(e) the woman who was laughing
(f) it
(g) John
(h) went
You can test this claim by inserting each expression into three contexts: Who found _________, _________ was seen by everyone, and What/who I heard was _________. For example, *Who found with a balloon is ungrammatical, as is *Have a nice day was seen by everyone, as opposed to Who found it? or John was seen by everyone. Only NPs fit into these contexts because only NPs can function as subjects and objects.
------------------------------------ HAL 88 --------------------------------------
There are other syntactic categories. The expression found a puppy is a verb phrase (VP). A verb phrase always contains a verb (V), and it may contain other categories, such as a noun phrase or prepositional phrase (PP), which is a preposition followed by an NP, such as in the park, on the roof, with a balloon. In (2) the VPs are those phrases that can complete the sentence “The child __________ .”
2. (a) saw a clown
(b) a bird
(c) slept
(d) smart
(e) ate the cake
(f) found the cake in the cupboard
(g) realized that the earth was round
Inserting (a), (c), (e), (f), and (g) will produce grammatical sentences, whereas the insertion of (b) or (d) would result in an ungrammatical sentence. Thus, (a), (c), (e), (f), and (g) are verb phrases.
Lexical and Functional Categories
There are ten parts of speech, and they are all troublesome.
MARK TWAIN, “The Awful German Language,” in A Tramp Abroad, 1880
Syntactic categories include both phrasal categories such as NP, VP, AdjP (adjective phrase), PP (prepositional phrase), and AdvP (adverbial phrase), as well as lexical categories such as noun (N), verb (V), preposition (P), adjective (Adj), and adverb (Adv). Each lexical category has a corresponding phrasal category. Following is a list of lexical categories with some examples of each type:
Lexical categories
Noun (N) puppy, boy, soup, happiness, fork, kiss, pillow, cake, cupboard
Verb (V) find, run, sleep, throw, realize, see, try, want, believe
Preposition (P) up, down, across, into, from, by, with
Adjective (Adj) red, big, candid, hopeless, fair, idiotic, lucky
Adverb (Adv) again, carefully, luckily, never, very, fairly
Many of these categories may already be familiar to you. As mentioned earlier, some of them are traditionally referred to as parts of speech. Other categories may be less familiar, for example, the category determiner (Det), which includes the articles a and the, as well as demonstratives such as this, that, these, and those, and “counting words” such as each and every. Another less familiar category is auxiliary (Aux), which includes the verbs have, had, be, was, and were, and the modals may, might, can, could, must, shall, should, will, and would. Aux and Det are functional categories, so called because their members have a grammatical function rather than a descriptive meaning. For example, determiners specify whether a noun is indefinite or definite (a boy versus the boy), or the proximity of the person or object to the context (this boy versus that boy). Auxiliaries provide the verb with a time frame, whether ongoing (John is dancing), completed in the past (John has danced), or occurring in the future (John will dance). Auxiliaries may also express notions such as possibility (John may dance), necessity (John must dance), ability (John can dance), and so on.
------------------------------------- HAL 89 --------------------------------------
Lexical categories typically have particular kinds of meanings associated with them. For example, verbs usually refer to actions, events, and states (kick, marry, love); adjectives to qualities or properties (lucky, old); common nouns to general entities (dog, elephant, house); and proper nouns to particular individuals (Noam Chomsky) or places (Dodger Stadium) or other things that people give names to, such as commercial products (Coca-Cola, Viagra). But the relationship between grammatical categories and meaning is more complex than these few examples suggest. For example, some nouns refer to events (marriage and destruction) and others to states (happiness, loneliness). We can use abstract nouns such as honor and beauty, rather than adjectives, to refer to properties and qualities. In the sentence “Seeing is believing,” seeing and believing are nouns but are not entities. Prepositions are usually used to express relationships between two entities involving a location (e.g., the boy is in the room, the cat is under the bed), but this is not always the case; the prepositions of, by, about, and with are not locational. Because of the difficulties involved in specifying the precise meaning of lexical categories, we do not usually define categories in terms of their meanings, but rather on the basis of their syntactic distribution (where they occur in a sentence) and morphological characteristics. For example, we define a noun as a word that can occur with a determiner (the boy) and that can take a plural marker (boys), among other properties.
All languages have syntactic categories such as N, V, and NP. Speakers know thesyntactic categories of their language, even if they do not know the technical terms. Our knowledge of the syntactic classes is revealed when we substitute equivalent phrases, as we just did in examples (1) and (2), and when we use the various syntactic tests that we have discussed.
Phrase Structure Trees and Rules
Who climbs the Grammar-Tree distinctly knows
Where Noun and Verb and Participle grows.
JOHN DRYDEN, “The Sixth Satyr of Juvenal,” 1693
Now that you know something about constituent structure and grammatical categories, you are ready to learn how the sentences of a language are constructed. We will begin by building trees for simple sentences and then proceed to more complex structures. The trees that we will build here are more detailed than those we saw in the previous sections, because the branches of the tree will have category labels identifying each constituent. In this section we will also introduce the syntactic rules that generate (a technical term for describe or specify) the different kinds of structures.
------------------------------------- HAL 90 --------------------------------------
The following tree diagram provides labels for each of the constituents of the
sentence “The child found a puppy.” These labels show that the entire sentence belongs to the syntactic category of S (because the S-node encompasses all the words). It also reveals that the child and a puppy belong to the category NP, that is, they are noun phrases, and that found a puppy belongs to the category VP or is a verb phrase, consisting of a verb and an NP. It also reveals the syntactic category of each of the words in the sentence.
A tree diagram with syntactic category information is called a phrase structure tree or a constituent structure tree. This tree shows that a sentence is both a linear string of words and a hierarchical structure with phrases nested in phrases. Phrase structure trees (PS trees, for short) are explicit graphic representations of a speaker’s knowledge of the structure of the sentences of his language. PS trees represent three aspects of a speaker’s syntactic knowledge:
1. The linear order of the words in the sentence
2. The identification of the syntactic categories of words and groups of words
3. The hierarchical structure of the syntactic categories (e.g., an S is composed of an NP followed by a VP, a VP is composed of a V that may be followed by an NP, and so on)
In chapter 1 we discussed the fact that the syntactic category of each word is listed in our mental dictionaries. We now see how this information is used by
the syntax of the language. Words appear in trees under labels that correspond to their syntactic category. Nouns are under N, determiners under Det, verbs under V, and so on.
The larger syntactic categories, such as VP, consist of all the syntactic categories and words below that point, or node, in the tree. The VP in the PS tree above consists of syntactic category nodes V and NP and the words found, a, and puppy. Because a puppy can be traced up the tree to the node NP, this constituent is a noun phrase. Because found and a puppy can be traced up to the node VP, this constituent is a verb phrase. The PS tree reflects the speaker’s intuitions about the natural groupings of words in a sentence. In discussing trees, every higher node is said to dominate all the categories beneath it. S dominates every node. A node is said to immediately dominate the categories one level below it. VP immediately dominates V and NP, the categories of which it is composed. Categories that are immediately dominated by the same node are sisters. V and NP are sisters in the phrase structure tree of “the child found a puppy.”
------------------------------------- HAL 91 --------------------------------------
A PS tree is a formal device for representing the speaker’s knowledge of the structure of sentences in his language, as revealed by our linguistic intuitions. When we speak, we are not aware that we are producing sentences with such structures, but controlled experiments, such as the click experiments described earlier, show that we use them in speech production and comprehension. We will discuss these experiments further in chapter 8.
The information represented in a PS tree can also be represented by another formal device: phrase structure (PS) rules. PS rules capture the knowledge that speakers have about the possible structures of a language. Just as a speaker cannot have an infinite list of sentences in her head, so she cannot have an infinite set of PS trees in her head. Rather, a speaker’s knowledge of the permissible and impermissible structures must exist as a finite set of rules that generate a tree for any sentence in the language. To express the structure given above, we need the following PS rules:
1. S → NP VP
2. NP → Det N
3. VP → V NP
Phrase structure rules specify the well-formed structures of a language precisely and concisely. They express the regularities of the language and make explicit a speaker’s knowledge of the order of words and the grouping of words into syntactic categories. For example, in English an NP may contain a determiner followed by a noun. This is represented by rule 2. This rule conveys two facts:
A noun phrase can contain a determiner followed by a noun in that order.
A determiner followed by a noun is a noun phrase.
You can think of PS rules as templates that a tree must match to be grammatical. To the left of the arrow is the dominating category, in this case NP, and the categories that it immediately dominates—that comprise it—appear on the right side, in this case Det and N. The right side of the arrow also shows the linear order of these components. Thus, one subtree for the English NP looks like this:
Rule 1 says that a sentence (S) contains (immediately dominates) an NP and a VP in that order. Rule 3 says that a verb phrase consists of a verb (V) followed by an NP. These rules are general statements and do not refer to any specific VP, V, or NP. The subtrees represented by rules 1 and 3 are as follows:
------------------------------------- HAL 92 --------------------------------------
A VP need not contain an NP object, however. It may include a verb alone, as in
the following sentences:
The woman laughed.
The man danced.
The horse galloped.
These sentences have the structure:
Thus a tree must have a VP that immediately dominates V, as specified by rule 4, which is therefore added to the grammar:
4. VP → V
The following sentences contain prepositional phrases following the verb:
The puppy played in the garden.
The boat sailed up the river.
A girl laughed at the monkey.
The sheepdog rolled in the mud.
The PS tree for such sentences is
To permit structures of this type, we need two additional PS rules, as in 5 and 6.
5. VP → V PP
6. PP → P NP
------------------------------------- HAL 93 --------------------------------------
Another option open to the VP is to contain or embed a sentence. For example, the sentence “The professor said that the student passed the exam” contains the sentence “the student passed the exam.” Preceding the embedded sentence is the word that, which is a complementizer (C). C is a functional category, like Aux and Det. Here is the structure of such sentence types:
To allow such embedded sentences, we need to add these two new rules to our set of phrase structure rules.
7. VP → V CP
8. CP → C S
CP stands for complementizer phrase. Rule 8 says that CP contains a complementizer such as that followed by the embedded sentence. Other complementizers are if and whether in sentences like
I don’t know whether I should talk about this.
The teacher asked if the students understood the syntax lesson.
that have structures similar to the one above.
Here are the PS rules we have discussed so far. A few other rules will be considered later.
1. S → NP VP
2. NP → Det N
3. VP → V NP
4. VP → V
5. VP → V PP
6. PP → P NP
7. VP → V CP
8. CP → C S
------------------------------------- HAL 97 --------------------------------------
In this structure the VP rule 5 has applied three times and so there are three PPs: [down the street] [with a gun] [toward the bank]. It is easy to see that the rule could have applied four or more times, for example by adding a PP like for no good purpose.
NPs can also contain PPs recursively. An example of this is shown by the phrase the man with the telescope in a box.
To show that speakers permit recursive NP structures of this sort, we need to include the following PS rule, which is like the recursive VP rule 5.
9. NP → NP PP
The PS rules define the allowable structures of the language, and in so doing make predictions about structures that we may not have considered when formulating each rule individually. These predictions can be tested, and if they are not validated, the rules must be reformulated because they must generate all and only the allowable structures. For example, rule 7 (VP → V CP) in combination with rules 8 (CP → C S) and 1 (S → NP VP) form a recursive set. (The recursiveness comes from the fact that S and VP occur on both the left and right side of the rules.) Those rules allow S to contain VP, which in turn contains CP, which in turn contains S, which in turn again contains VP, and so on, potentially without end. These rules, formulated for different purposes, correctly predict the limitlessness of language in which sentences are embedded inside larger sentences, such as The children hope that the teacher knows that the principal said that the school closes for the day as illustrated on the following page.
------------------------------------- HAL 99 --------------------------------------
Recursive Adjectives and Possessives
Now we consider the case of multiple adjectives, illustrated at the beginning of the chapter with sentences such as “The kindhearted, intelligent, handsome boy had many girlfriends.” In English, adjectives occur before the noun. As a first approximation we might follow the system we have adopted thus far and introduce a recursive NP rule with a prenominal adjective:
NP → Adj NP
Repeated application of this rule would generate trees with multiple adjective positions, as desired.
But there is something wrong in this tree, which is made apparent when we expand the lowest NP. The adjective can appear before the determiner, and this is not a possible word order in English NPs.
------------------------------------- HAL 100 ------------------------------------The problem is that although determiners and adjectives are both modifiers of the noun, they have a different status. First, an NP will never have more than one determiner in it, while it may contain many adjectives. Also, an adjective directly modifies the noun, while a determiner modifies the whole adjective(s) + noun complex. The expression “the big dog” refers to some specific dog that is big, and not just some dog of any size. In general, modification occurs between sisters. If the adjective modifies the noun, then it is sister to the noun. If the determiner modifies the adjective + noun complex, then the determiner is sister to this complex. We can represent these two sisterhood relations by introducing an additional level of structure between NP and N. We refer to this level as N-bar (written as N').
This structure provides the desired sisterhood relations. The adjective handsome is sister to the noun boy, which it therefore modifies, and the determiner is sister to the N' handsome boy. We must revise our NP rules to reflect this new structure, and add two rules for N'. Not all NPs have adjectives, of course. This is reflected in the second N' rule in which N' dominates only N.
NP → Det N' (revised version of NP → Det N)
N' → Adj N'
N' → N
Let us now see how these revised rules generate NPs with multiple (potentially infinitely many) adjectives.
Thus far all the NPs we have looked at are common nouns with a simple definite or indefinite determiner (e.g., the cat, a boy), but NPs can consist of a simple pronoun (e.g., he, she, we, they) or a proper name (e.g., Robert, California, Prozac). To reflect determiner-less NP structures, we will need the rule
NP → N'
But that’s not all. We have possessive noun phrases such as Melissa’s garden, the girl’s shoes, and the man with the telescope’s hat. In these structures the possessor NP (e.g., Melissa’s, the girl’s, etc.) functions as a determiner in that it further specifies its sister noun. The ’s is the phonological realization of the abstract element poss. The structures are illustrated in each of the following trees.
------------------------------------- HAL 101 ------------------------------------
To accommodate the possessive structure we need an additional rule:
Det → NP poss
This rule forms a recursive set with the NP → Det N' rule. Together these rules allow an English speaker to have multiple possessives such as The student’s friend’s cousin’s book.
The embedding of categories within categories is common to all languages. Our brain capacity is finite, able to store only a finite number of categories and rules for their combination. Yet this finite system places an infinite set of sentences at our disposal.
This linguistic property also illustrates the difference between competence and performance, discussed in chapter 6. All speakers of English (and other languages) have as part of their linguistic competence—their mental grammar — the ability to embed phrases and sentences within each other ad infinitum. However, as the structures grow longer, they become increasingly more difficult to produce and understand. This could be due to short-term memory limitations, muscular fatigue, breathlessness, boredom, or any number of performance factors. (We will discuss performance factors more fully in chapter 8.) Nevertheless, these very long sentences would be well-formed according to the rules of the grammar.
------------------------------------- HAL 102 ------------------------------------
Heads and Complements
Phrase structure trees also show relationships among elements in a sentence. For example, the subject and direct object of the sentence can be structurally defined. The subject is the NP that is closest to, or immediately dominated by, the root S. The direct object is the NP that is closest to, or immediately dominated by, VP.
Another kind of relationship is that between the head of a phrase and its sisters. The head of a phrase is the word whose lexical category defines the type of phrase: the noun in a noun phrase, the verb in a verb phrase, and so on. Reviewing the PS rules in the previous section, we see that every VP contains a verb, which is its head. The VP may also contain other categories, such as an NP or CP. Those sister categories are complements; they complete the meaning of the phrase. Loosely speaking, the entire phrase refers to whatever the head verb refers to. For example, the VP find a puppy refers to an event of “finding.” The NP object in the VP that completes its meaning is a complement. The underscored CP (complementizer phrase) in the sentence “I thought that the child found the puppy” is also a complement. (Please do not confuse the terms complementizer and complement.)
Every phrasal category, then, has a head of its same syntactic type. NPs are headed by nouns, PPs are headed by prepositions, CPs by complementizers, and so on; and every phrasal head can have a complement, which provides further information about the head. In the sentence “The death of Lincoln shocked the nation,” the PP of Lincoln is the complement to the head noun death. Other examples of complements are illustrated in the following examples, with the head in italics and the complement underlined:
------------------------------------- HAL 103 ------------------------------------
an argument over jelly beans (PP complement to noun)
his belief that justice will prevail (CP complement to noun)
happy to be here (infinitive complement to adjective)
about the war in Iraq (NP complement to preposition)
wrote a long letter to his only sister (NP—PP complement to verb)
tell John that his mother is coming to dinner (NP CP complements to verb)
Each of these examples is a phrase (NP, AdjP, PP, VP) that contains a head (N, Adj, P, V), followed by a complement of varying composition such as CP in the case of belief, or NP PP in the case of wrote, and so on. The head complement relation is universal. All languages have phrases that are headed and that contain complements.
However, the order of the head and complement may differ in different languages. In English, for example, we see that the head comes first, followed by the complement. In Japanese, complements precede the head, as shown in the following examples:
Taro-ga inu-o mitsuketa
Taro-subject marker dog-object marker found (Taro found a dog)
Inu-ga niwa-de asonde iru
dog-subject marker garden-in playing is (The dog is playing in the garden)
In the first sentence, the direct object complement inu-o “dog” precedes the head verb mitsuketa “found.” In the second, the PP complement niwa-de “in the garden” also precedes the head verb phrase. English is a VO language, meaning that the verb ordinarily precedes its object. Japanese is an OV language, and this difference is also reflected in the head/complement word order.
Selection
Whether a verb takes a complement or not depends on the properties of the verb. For example, the verb find is a transitive verb. A transitive verb requires an NP complement (direct object), as in The boy found the ball, but not *The boy found, or *The boy found in the house. Some verbs like eat are optionally transitive. John ate and John ate a sandwich are both grammatical.
Verbs select different kinds of complements. For example, verbs like put and give take both an NP and a PP complement, but cannot occur with either alone:
Sam put the milk in the refrigerator.
*Sam put the milk.
Robert gave the film to his client.
*Robert gave to his client.
Sleep is an intransitive verb; it cannot take an NP complement.
Michael slept.
*Michael slept a fish.
Some verbs, such as think, select a sentence complement, as in “I think that
Sam won the race.” Other verbs, like tell, select an NP and a sentence, as in “I told Sam that Michael was on his bicycle”; yet other verbs like feel select either an AdjP or a sentence complement. (Complements are italicized.)
------------------------------------- HAL 104 ------------------------------------
Paul felt strong as an ox.
He feels that he can win.
As we will discuss later, sentences that are complements must often be preceded by a complementizer that.
Other categories besides verbs also select their complements. For example, the noun belief selects either a PP or a CP, while the noun sympathy selects a PP, but not a CP, as shown by the following examples:
the belief in freedom of speech
the belief that freedom of speech is a basic right
their sympathy for the victims
*their sympathy that the victims are so poor
Adjectives can also have complements. For example, the adjectives tired and proud select PPs:
tired of stale sandwiches
proud of her children
With noun selection, the complement is often optional. Thus sentences like “He respected their belief,” “We appreciated their sympathy,” “Elimelech was tired,” and “All the mothers were proud” are syntactically well-formed with a meaning that might be conveyed by an explicit complement understood from context. Verb selection is often not optional, however, so that *He put the milk is ungrammatical even if it is clear from context where the milk was put.
The information about the complement types selected by particular verbs and other lexical items is called C-selection or subcategorization, and is included in the lexical entry of the item in our mental lexicon. (Here C stands for “categorial” and is not to be confused with the C that stands for “complementizer”—we apologize for the “clash” of symbols, but that’s what it’s like in the linguistic literature.)
Verbs also include in their lexical entry a specification of certain intrinsic semantic properties of their subjects and complements, just as they select for syntactic categories. This kind of selection is called S-selection (S for semantic). For example, the verb murder requires its subject and object to be human, while the verb drink requires its subject to be animate and its object liquid. Verbs such as like, hate, and so on select animate subjects. The following sentences violate S-selection and can only be used in a metaphorical sense. (We will use the symbol “!” to indicate a semantic anomaly)
!The rock murdered the man.
!The beer drank the student.
!The tree liked the boy.
The famous sentence Colorless green ideas sleep furiously, discussed earlier in this chapter, is anomalous because (among other things) S-selection is violated (e.g., the verb sleep requires an animate subject). In chapter 3 we will discuss the semantic relationships between a verb and its subject and objects in far more detail.
------------------------------------- HAL 105 ------------------------------------
The well-formedness of a phrase depends then on at least two factors: whether the phrase conforms to the structural constraints of the language as expressed in the PS rules, and whether it obeys the selectional requirements of the head, both syntactic (C-selection) and semantic (S-selection).
What Heads the Sentence
Might, could, would—they are contemptible auxiliaries.
GEORGE ELIOT (MARY ANN EVANS), Middlemarch, 1872
We said earlier that all phrases have heads. One category that we have not yet discussed in this regard is sentence (S). For uniformity’s sake, we want all the categories to be headed, but what would the head of S be? To answer this question, let us consider sentences such as the following:
Sam will kick the soccer ball.
Sam has kicked the soccer ball.
Sam is kicking the soccer ball.
Sam may kick the soccer ball.
As noted earlier, words like will, has, is, and may are auxiliary verbs, belonging to the category Aux, which also includes modals such as might, could, would, can, and several others. They occur in structures such as the following one.
(From now on we will adopt the convention of using a triangle under a node when the content of a category is not crucial to the point under discussion.)
Auxiliary verbs specify a time frame for the event (or state) described by the verb, whether it will take place in the future, already took place in the past, or is taking place now. A modal such as may contains “possibility” as part of its meaning, and says it is possible that the event will occur at some future time. The category Aux is a natural category to head S. Just as the VP is about the situation described by the verb—eat ice cream is about “eating”—so a sentence is about a situation or state of affairs that occurs at some point in time.
------------------------------------- HAL 106 ------------------------------------
The parallel with other categories extends further. In the previous PS tree, VP is the complement to Aux. The selectional relationship between Aux and VP is demonstrated by the fact that particular auxiliaries go with particular kinds of VPs. For example, the auxiliary be takes a progressive (-ing) form of the verb,
The boy is dancing.
while the auxiliary have selects a past participle (-en) form of the verb,
The girl has eaten.
and the modals select the infinitival form of the verb (no affixes),
The child must sleep
The boy may eat.
To have a uniform notation, many linguists use the symbols T (= tense) and TP (= tense phrase) instead of Aux and S. Furthermore, just as the NP required the intermediate N-bar (N') category, the TP also has the intermediate T-bar (T') category, as in the phrase structure tree below.
Indeed, many linguists assume that all XPs, where XP stands for any of NP, PP, VP, TP, AdjP, or CP, have three levels of structure. This is referred to as X-bar theory. The basic three-level X-bar schema is as follows:
The first level is the XP itself. The second level consists of a specifier, which functions as a modifier (and which is generally an optional constituent), and an X' (i.e., “X-bar”). For example, an NP specifier is a determiner; a VP specifier is an adverb such as never or often; an AdjP specifier is a degree word such as very or quite. The third level is an expansion of X' and consists of a head X and a complement, which may itself be a phrasal category, thus giving rise to recursion. X-bar structure is thought to be universal, occurring in all the world’s languages, though the order of the elements inside XP and X' may be reversed, as we saw in Japanese
------------------------------------ HAL 107 -------------------------------------
We will not use X-bar conventions in our description of syntax except on the few occasions where the notation provides an insight into the syntax of the language. For sentences we will generally use the more intuitive symbols S and Aux instead of TP and T, but you should think of Aux and S as having the same relationship to each other as V and VP, N and NP, and so on. To achieve this more straightforward approach, we will also ignore the T' category until it is needed later on in the description of the syntax of the main verb be.
Without the use of TP, T', and T, we need an additional PS rule to characterize structures containing Aux:
VP → Aux VP
Like the other recursive VP rules, this rule will allow multiple Aux positions.
This is a desired consequence because English allows sentences with multiple auxiliaries such as:
The child may be sleeping. (modal, be)
The dog has been barking all night. (have, be)
The bird must have been flying home. (modal, have, be)
The introduction of Aux into the system raises a question. Not all sentences seem to have auxiliaries. For example, the sentence “Sam kicked the soccer ball” has no modal, have or be. There is, however, a time reference for this sentence, namely, the past tense on the verb kicked. In sentences without auxiliaries, the tense of the sentence is its head. Instead of having a word under the category Aux (or T), there is a tense specification, present or past, as in the following tree:
------------------------------------ HAL 108 -------------------------------------
The inflection on the verb must match the tense in Aux. For example, if the
tense of the sentence is past, then the verb must have an -ed affix (or must be an irregular past tense verb such as ate).
Thus, in English, and many other languages, the head of S may contain only an abstract tense specification and no actual word, as just illustrated. The actual morpheme, in this case -ed or an irregular past tense form such as went, is inserted into the tree after all the syntactic rules have applied. Most inflectional morphemes, which depend on elements of syntax, are represented in this way. Another example is the tense-bearing word do that is inserted into negative sentences such as John did not go and questions such as Where did John go? In these sentences did means “past tense.” Later in this chapter we will see how do-insertion works.
In addition to specifying the time reference of the sentence, Aux specifies the agreement features of the subject. For example, if the subject is we, Aux contains the features first-person and plural; if the subject is he or she, Aux contains the features third-person and singular. So, another function of the syntactic rules is to use Aux as a “matchmaker” between the subject and the verb. When the subject and the verb bear the same features, Aux makes a match; when they have incompatible features, Aux cannot make a match and the sentence is ungrammatical. This matchmaker function of syntactic rules is more obvious in languages such as Italian, which have many different agreement morphemes, as discussed in chapter 1. Consider the Italian sentence for “I buy books.”
The verb compro, “buy,” in the first sentence bears the first-person singular morpheme, -o, which matches the agreement feature in Aux, which in turn matches the subject Io, “I.” The sentence is therefore grammatical. In the second sentence, there is a mismatch between the first-person subject and the secondperson features in Aux (and on the verb), and so the sentence is ungrammatical.
------------------------------------ HAL 109 -------------------------------------
Structural Ambiguities
The structure of every sentence is a lesson in logic.
JOHN STUART MILL, Inaugural address at St. Andrews, 1867
As mentioned earlier, certain kinds of ambiguous sentences have more than one phrase structure tree, each corresponding to a different meaning. The sentence The boy saw the man with the telescope is structurally ambiguous. Its two meanings correspond to the following two phrase structure trees. (For simplicity we omit Aux in these structures and we return to the non-X-bar notation.)
------------------------------------ HAL 110 -------------------------------------
One meaning of this sentence is “the boy used a telescope to see the man.” The first phrase structure tree represents this meaning. The key element is the position of the PP directly under the VP. Notice that although the PP is under VP, it is not a complement because phrasal categories don’t take complements (only heads do), and because it is not selected by the verb. The verb see selects an NP. In this sentence, the PP has an adverbial function and modifies the verb.
In its other meaning, “the boy saw a man who had a telescope,” the PP with the telescope occurs under the direct object NP, where it modifies the noun man. In this second meaning, the complement of the verb see is the entire NP—the man with the telescope.
The PP in the first structure is generated by the rule
VP → VP PP
In the second structure the PP is generated by the rule
NP → NP PP
Two interpretations are possible because the rules of syntax permit different structures for the same linear order of words. Following is the set of PS rules that we have presented so far in the chapter. The rules have been renumbered.
1. S → NP VP
2. NP → Det N'
3. Det → NP poss
4. NP → N'
5. NP → NP PP
6. N' → Adj N'
7. N' → N
8. VP → V
9. VP → V NP
10. VP → V CP
11. VP → Aux VP
12. VP → VP PP
13. PP → P NP
14. CP → C S
This is not the complete set of PS rules for the language. Various structures in English cannot be generated with these rules, some of which we will talk about later. But even this mini phrase structure grammar generates an infinite set of possible sentences because the rules are recursive. These PS rules specify the word order for English (and other SVO languages, but not for Japanese, say, in which the object comes before the verb). Linear order aside, the hierarchical organization illustrated by these rules is largely true for all languages, as expressed by X-bar schema.
A PS tree is a formal device for representing the speaker’s knowledge of the structure of sentences in his language, as revealed by our linguistic intuitions. When we speak, we are not aware that we are producing sentences with such structures, but controlled experiments, such as the click experiments described earlier, show that we use them in speech production and comprehension. We will discuss these experiments further in chapter 8.
The information represented in a PS tree can also be represented by another formal device: phrase structure (PS) rules. PS rules capture the knowledge that speakers have about the possible structures of a language. Just as a speaker cannot have an infinite list of sentences in her head, so she cannot have an infinite set of PS trees in her head. Rather, a speaker’s knowledge of the permissible and impermissible structures must exist as a finite set of rules that generate a tree for any sentence in the language. To express the structure given above, we need the following PS rules:
1. S → NP VP
2. NP → Det N
3. VP → V NP
Phrase structure rules specify the well-formed structures of a language precisely and concisely. They express the regularities of the language and make explicit a speaker’s knowledge of the order of words and the grouping of words into syntactic categories. For example, in English an NP may contain a determiner followed by a noun. This is represented by rule 2. This rule conveys two facts:
A noun phrase can contain a determiner followed by a noun in that order.
A determiner followed by a noun is a noun phrase.
You can think of PS rules as templates that a tree must match to be grammatical. To the left of the arrow is the dominating category, in this case NP, and the categories that it immediately dominates—that comprise it—appear on the right side, in this case Det and N. The right side of the arrow also shows the linear order of these components. Thus, one subtree for the English NP looks like this:
Rule 1 says that a sentence (S) contains (immediately dominates) an NP and a VP in that order. Rule 3 says that a verb phrase consists of a verb (V) followed by an NP. These rules are general statements and do not refer to any specific VP, V, or NP. The subtrees represented by rules 1 and 3 are as follows:
------------------------------------- HAL 92 --------------------------------------
A VP need not contain an NP object, however. It may include a verb alone, as in
the following sentences:
The woman laughed.
The man danced.
The horse galloped.
These sentences have the structure:
Thus a tree must have a VP that immediately dominates V, as specified by rule 4, which is therefore added to the grammar:
4. VP → V
The following sentences contain prepositional phrases following the verb:
The puppy played in the garden.
The boat sailed up the river.
A girl laughed at the monkey.
The sheepdog rolled in the mud.
The PS tree for such sentences is
To permit structures of this type, we need two additional PS rules, as in 5 and 6.
5. VP → V PP
6. PP → P NP
------------------------------------- HAL 93 --------------------------------------
Another option open to the VP is to contain or embed a sentence. For example, the sentence “The professor said that the student passed the exam” contains the sentence “the student passed the exam.” Preceding the embedded sentence is the word that, which is a complementizer (C). C is a functional category, like Aux and Det. Here is the structure of such sentence types:
To allow such embedded sentences, we need to add these two new rules to our set of phrase structure rules.
7. VP → V CP
8. CP → C S
CP stands for complementizer phrase. Rule 8 says that CP contains a complementizer such as that followed by the embedded sentence. Other complementizers are if and whether in sentences like
I don’t know whether I should talk about this.
The teacher asked if the students understood the syntax lesson.
that have structures similar to the one above.
Here are the PS rules we have discussed so far. A few other rules will be considered later.
1. S → NP VP
2. NP → Det N
3. VP → V NP
4. VP → V
5. VP → V PP
6. PP → P NP
7. VP → V CP
8. CP → C S
------------------------------------- HAL 94 --------------------------------------
Some Conventions for Building Phrase Structure Trees
Everyone who is master of the language he speaks . . . may form new . . . phrases, provided they coincide with the genius of the language.
JOHANN DAVID MICHAELIS, Dissertation, 1769
One can use the phrase structure rules as a guide for building trees that follow the structural constraints of the language. In so doing, certain conventions are followed. The S occurs at the top or “root” of the tree (it’s upside down). Another convention specifies how the rules are applied: First, find the rule with S on the left side of the arrow, and put the categories on the right side below the S, as shown here:
Continue by matching any syntactic category at the bottom of the partially constructed tree to a category on the left side of a rule, then expand the tree with the categories on the right side. For example, we may expand the tree by applying the NP rule to produce:
The categories at the bottom are Det, N, and VP, but only VP occurs to the left of an arrow in the set of rules and so needs to be expanded using one of the VP rules. Any one of the VP rules will work. The order in which the rules appear in the list of rules is irrelevant. (We could have begun by expanding the VP rather than the NP.) Suppose we use rule 5 next. Then the tree has grown to look like this:
Convention dictates that we continue in this way until none of the categories at the bottom of the tree appears on the left side of any rule (i.e., no phrasal categories may remain unexpanded). The PP must expand into a P and an NP (rule 6), and the NP into a Det and an N. We can use a rule as many times as it can apply. In this tree, we used the NP rule twice. After we have applied all the rules that can apply, the tree looks like this:
Some Conventions for Building Phrase Structure Trees
Everyone who is master of the language he speaks . . . may form new . . . phrases, provided they coincide with the genius of the language.
JOHANN DAVID MICHAELIS, Dissertation, 1769
One can use the phrase structure rules as a guide for building trees that follow the structural constraints of the language. In so doing, certain conventions are followed. The S occurs at the top or “root” of the tree (it’s upside down). Another convention specifies how the rules are applied: First, find the rule with S on the left side of the arrow, and put the categories on the right side below the S, as shown here:
Continue by matching any syntactic category at the bottom of the partially constructed tree to a category on the left side of a rule, then expand the tree with the categories on the right side. For example, we may expand the tree by applying the NP rule to produce:
The categories at the bottom are Det, N, and VP, but only VP occurs to the left of an arrow in the set of rules and so needs to be expanded using one of the VP rules. Any one of the VP rules will work. The order in which the rules appear in the list of rules is irrelevant. (We could have begun by expanding the VP rather than the NP.) Suppose we use rule 5 next. Then the tree has grown to look like this:
Convention dictates that we continue in this way until none of the categories at the bottom of the tree appears on the left side of any rule (i.e., no phrasal categories may remain unexpanded). The PP must expand into a P and an NP (rule 6), and the NP into a Det and an N. We can use a rule as many times as it can apply. In this tree, we used the NP rule twice. After we have applied all the rules that can apply, the tree looks like this:
------------------------------------- HAL 95 --------------------------------------
By following these conventions, we generate only trees specified by the PS rules, and hence only trees that conform to the syntax of the language. By implication, any tree not so specified will be ungrammatical, that is, not permitted by the syntax. At any point during the construction of a tree, any rule may be used as long as its left-side category occurs somewhere at the bottom of the tree. By choosing different VP rules, we could specify different structures corresponding to sentences such as:
The boys left. (VP → V)
The wind blew the kite. (VP → V NP)
The senator hopes that the bill passes. (VP → V CP)
Because the number of possible sentences in every language is infinite, there are also an infinite number of trees. However, all trees are built out of the finite set of substructures allowed by the grammar of the language, and these substructures are specified by the finite set of phrase structure rules.
The Infinity of Language: Recursive Rules
So, naturalists observe, a flea
Hath smaller fleas that on him prey;
And these have smaller still to bite ’em,
And so proceed ad infinitum.
JONATHAN SWIFT, “On Poetry, a Rhapsody,” 1733
We noted at the beginning of the chapter that the number of sentences in a language is infinite and that languages have various means of creating longer and longer sentences, such as adding an adjective or a prepositional phrase. Even children know how to produce and understand very long sentences and know how to make them even longer, as illustrated by the children’s rhyme about the house that Jack built.
This is the farmer sowing the corn,
that kept the cock that crowed in the morn,
that waked the priest all shaven and shorn,
that married the man all tattered and torn,
that kissed the maiden all forlorn,
that milked the cow with the crumpled horn,
that tossed the dog,
By following these conventions, we generate only trees specified by the PS rules, and hence only trees that conform to the syntax of the language. By implication, any tree not so specified will be ungrammatical, that is, not permitted by the syntax. At any point during the construction of a tree, any rule may be used as long as its left-side category occurs somewhere at the bottom of the tree. By choosing different VP rules, we could specify different structures corresponding to sentences such as:
The boys left. (VP → V)
The wind blew the kite. (VP → V NP)
The senator hopes that the bill passes. (VP → V CP)
Because the number of possible sentences in every language is infinite, there are also an infinite number of trees. However, all trees are built out of the finite set of substructures allowed by the grammar of the language, and these substructures are specified by the finite set of phrase structure rules.
The Infinity of Language: Recursive Rules
So, naturalists observe, a flea
Hath smaller fleas that on him prey;
And these have smaller still to bite ’em,
And so proceed ad infinitum.
JONATHAN SWIFT, “On Poetry, a Rhapsody,” 1733
We noted at the beginning of the chapter that the number of sentences in a language is infinite and that languages have various means of creating longer and longer sentences, such as adding an adjective or a prepositional phrase. Even children know how to produce and understand very long sentences and know how to make them even longer, as illustrated by the children’s rhyme about the house that Jack built.
This is the farmer sowing the corn,
that kept the cock that crowed in the morn,
that waked the priest all shaven and shorn,
that married the man all tattered and torn,
that kissed the maiden all forlorn,
that milked the cow with the crumpled horn,
that tossed the dog,
------------------------------------- HAL 96 --------------------------------------
that worried the cat,
that killed the rat,
that ate the malt,
that lay in the house that Jack built.
The child begins the rhyme with This is the house that Jack built, continues by lengthening it to This is the malt that lay in the house that Jack built, and so on.
You can add any of the following to the beginning of the rhyme and still have a grammatical sentence:
I think that . . .
What is the name of the unicorn that noticed that . . .
Ask someone if . . .
Do you know whether . . .
Once we acknowledge the unboundedness of sentences, we need a formal device to capture that crucial aspect of speakers’ syntactic knowledge. It is no longer possible to specify each legal structure; there are infinitely many.
To see how this works, let us first look at the case of multiple prepositional phrases such as [The girl walked [down the street] [over the hill] [through the woods] . . .]. VP substructures currently allow only one PP per sentence (VP → V PP—rule 5). We can rectify this problem by revising rule 5:
5. VP → VP PP
Rule 5 is different from the previous rules because it repeats its own category (VP) inside itself. This is an instance of a recursive rule. Recursive rules are of critical importance because they allow the grammar to generate an infinite set of sentences. Reapplying rule 5 shows how the syntax permits structures with multiple PPs, such as in the sentence “The girl walked down the street with a gun toward the bank.”
that worried the cat,
that killed the rat,
that ate the malt,
that lay in the house that Jack built.
The child begins the rhyme with This is the house that Jack built, continues by lengthening it to This is the malt that lay in the house that Jack built, and so on.
You can add any of the following to the beginning of the rhyme and still have a grammatical sentence:
I think that . . .
What is the name of the unicorn that noticed that . . .
Ask someone if . . .
Do you know whether . . .
Once we acknowledge the unboundedness of sentences, we need a formal device to capture that crucial aspect of speakers’ syntactic knowledge. It is no longer possible to specify each legal structure; there are infinitely many.
To see how this works, let us first look at the case of multiple prepositional phrases such as [The girl walked [down the street] [over the hill] [through the woods] . . .]. VP substructures currently allow only one PP per sentence (VP → V PP—rule 5). We can rectify this problem by revising rule 5:
5. VP → VP PP
Rule 5 is different from the previous rules because it repeats its own category (VP) inside itself. This is an instance of a recursive rule. Recursive rules are of critical importance because they allow the grammar to generate an infinite set of sentences. Reapplying rule 5 shows how the syntax permits structures with multiple PPs, such as in the sentence “The girl walked down the street with a gun toward the bank.”
------------------------------------- HAL 97 --------------------------------------
In this structure the VP rule 5 has applied three times and so there are three PPs: [down the street] [with a gun] [toward the bank]. It is easy to see that the rule could have applied four or more times, for example by adding a PP like for no good purpose.
NPs can also contain PPs recursively. An example of this is shown by the phrase the man with the telescope in a box.
9. NP → NP PP
The PS rules define the allowable structures of the language, and in so doing make predictions about structures that we may not have considered when formulating each rule individually. These predictions can be tested, and if they are not validated, the rules must be reformulated because they must generate all and only the allowable structures. For example, rule 7 (VP → V CP) in combination with rules 8 (CP → C S) and 1 (S → NP VP) form a recursive set. (The recursiveness comes from the fact that S and VP occur on both the left and right side of the rules.) Those rules allow S to contain VP, which in turn contains CP, which in turn contains S, which in turn again contains VP, and so on, potentially without end. These rules, formulated for different purposes, correctly predict the limitlessness of language in which sentences are embedded inside larger sentences, such as The children hope that the teacher knows that the principal said that the school closes for the day as illustrated on the following page.
------------------------------------- HAL 98 --------------------------------------
------------------------------------- HAL 99 --------------------------------------
Recursive Adjectives and Possessives
Now we consider the case of multiple adjectives, illustrated at the beginning of the chapter with sentences such as “The kindhearted, intelligent, handsome boy had many girlfriends.” In English, adjectives occur before the noun. As a first approximation we might follow the system we have adopted thus far and introduce a recursive NP rule with a prenominal adjective:
NP → Adj NP
Repeated application of this rule would generate trees with multiple adjective positions, as desired.
But there is something wrong in this tree, which is made apparent when we expand the lowest NP. The adjective can appear before the determiner, and this is not a possible word order in English NPs.
------------------------------------- HAL 100 ------------------------------------The problem is that although determiners and adjectives are both modifiers of the noun, they have a different status. First, an NP will never have more than one determiner in it, while it may contain many adjectives. Also, an adjective directly modifies the noun, while a determiner modifies the whole adjective(s) + noun complex. The expression “the big dog” refers to some specific dog that is big, and not just some dog of any size. In general, modification occurs between sisters. If the adjective modifies the noun, then it is sister to the noun. If the determiner modifies the adjective + noun complex, then the determiner is sister to this complex. We can represent these two sisterhood relations by introducing an additional level of structure between NP and N. We refer to this level as N-bar (written as N').
This structure provides the desired sisterhood relations. The adjective handsome is sister to the noun boy, which it therefore modifies, and the determiner is sister to the N' handsome boy. We must revise our NP rules to reflect this new structure, and add two rules for N'. Not all NPs have adjectives, of course. This is reflected in the second N' rule in which N' dominates only N.
NP → Det N' (revised version of NP → Det N)
N' → Adj N'
N' → N
Let us now see how these revised rules generate NPs with multiple (potentially infinitely many) adjectives.
Thus far all the NPs we have looked at are common nouns with a simple definite or indefinite determiner (e.g., the cat, a boy), but NPs can consist of a simple pronoun (e.g., he, she, we, they) or a proper name (e.g., Robert, California, Prozac). To reflect determiner-less NP structures, we will need the rule
NP → N'
But that’s not all. We have possessive noun phrases such as Melissa’s garden, the girl’s shoes, and the man with the telescope’s hat. In these structures the possessor NP (e.g., Melissa’s, the girl’s, etc.) functions as a determiner in that it further specifies its sister noun. The ’s is the phonological realization of the abstract element poss. The structures are illustrated in each of the following trees.
------------------------------------- HAL 101 ------------------------------------
To accommodate the possessive structure we need an additional rule:
Det → NP poss
This rule forms a recursive set with the NP → Det N' rule. Together these rules allow an English speaker to have multiple possessives such as The student’s friend’s cousin’s book.
The embedding of categories within categories is common to all languages. Our brain capacity is finite, able to store only a finite number of categories and rules for their combination. Yet this finite system places an infinite set of sentences at our disposal.
This linguistic property also illustrates the difference between competence and performance, discussed in chapter 6. All speakers of English (and other languages) have as part of their linguistic competence—their mental grammar — the ability to embed phrases and sentences within each other ad infinitum. However, as the structures grow longer, they become increasingly more difficult to produce and understand. This could be due to short-term memory limitations, muscular fatigue, breathlessness, boredom, or any number of performance factors. (We will discuss performance factors more fully in chapter 8.) Nevertheless, these very long sentences would be well-formed according to the rules of the grammar.
------------------------------------- HAL 102 ------------------------------------
Heads and Complements
Phrase structure trees also show relationships among elements in a sentence. For example, the subject and direct object of the sentence can be structurally defined. The subject is the NP that is closest to, or immediately dominated by, the root S. The direct object is the NP that is closest to, or immediately dominated by, VP.
Another kind of relationship is that between the head of a phrase and its sisters. The head of a phrase is the word whose lexical category defines the type of phrase: the noun in a noun phrase, the verb in a verb phrase, and so on. Reviewing the PS rules in the previous section, we see that every VP contains a verb, which is its head. The VP may also contain other categories, such as an NP or CP. Those sister categories are complements; they complete the meaning of the phrase. Loosely speaking, the entire phrase refers to whatever the head verb refers to. For example, the VP find a puppy refers to an event of “finding.” The NP object in the VP that completes its meaning is a complement. The underscored CP (complementizer phrase) in the sentence “I thought that the child found the puppy” is also a complement. (Please do not confuse the terms complementizer and complement.)
Every phrasal category, then, has a head of its same syntactic type. NPs are headed by nouns, PPs are headed by prepositions, CPs by complementizers, and so on; and every phrasal head can have a complement, which provides further information about the head. In the sentence “The death of Lincoln shocked the nation,” the PP of Lincoln is the complement to the head noun death. Other examples of complements are illustrated in the following examples, with the head in italics and the complement underlined:
------------------------------------- HAL 103 ------------------------------------
an argument over jelly beans (PP complement to noun)
his belief that justice will prevail (CP complement to noun)
happy to be here (infinitive complement to adjective)
about the war in Iraq (NP complement to preposition)
wrote a long letter to his only sister (NP—PP complement to verb)
tell John that his mother is coming to dinner (NP CP complements to verb)
Each of these examples is a phrase (NP, AdjP, PP, VP) that contains a head (N, Adj, P, V), followed by a complement of varying composition such as CP in the case of belief, or NP PP in the case of wrote, and so on. The head complement relation is universal. All languages have phrases that are headed and that contain complements.
However, the order of the head and complement may differ in different languages. In English, for example, we see that the head comes first, followed by the complement. In Japanese, complements precede the head, as shown in the following examples:
Taro-ga inu-o mitsuketa
Taro-subject marker dog-object marker found (Taro found a dog)
Inu-ga niwa-de asonde iru
dog-subject marker garden-in playing is (The dog is playing in the garden)
In the first sentence, the direct object complement inu-o “dog” precedes the head verb mitsuketa “found.” In the second, the PP complement niwa-de “in the garden” also precedes the head verb phrase. English is a VO language, meaning that the verb ordinarily precedes its object. Japanese is an OV language, and this difference is also reflected in the head/complement word order.
Selection
Whether a verb takes a complement or not depends on the properties of the verb. For example, the verb find is a transitive verb. A transitive verb requires an NP complement (direct object), as in The boy found the ball, but not *The boy found, or *The boy found in the house. Some verbs like eat are optionally transitive. John ate and John ate a sandwich are both grammatical.
Verbs select different kinds of complements. For example, verbs like put and give take both an NP and a PP complement, but cannot occur with either alone:
Sam put the milk in the refrigerator.
*Sam put the milk.
Robert gave the film to his client.
*Robert gave to his client.
Sleep is an intransitive verb; it cannot take an NP complement.
Michael slept.
*Michael slept a fish.
Some verbs, such as think, select a sentence complement, as in “I think that
Sam won the race.” Other verbs, like tell, select an NP and a sentence, as in “I told Sam that Michael was on his bicycle”; yet other verbs like feel select either an AdjP or a sentence complement. (Complements are italicized.)
------------------------------------- HAL 104 ------------------------------------
Paul felt strong as an ox.
He feels that he can win.
As we will discuss later, sentences that are complements must often be preceded by a complementizer that.
Other categories besides verbs also select their complements. For example, the noun belief selects either a PP or a CP, while the noun sympathy selects a PP, but not a CP, as shown by the following examples:
the belief in freedom of speech
the belief that freedom of speech is a basic right
their sympathy for the victims
*their sympathy that the victims are so poor
Adjectives can also have complements. For example, the adjectives tired and proud select PPs:
tired of stale sandwiches
proud of her children
With noun selection, the complement is often optional. Thus sentences like “He respected their belief,” “We appreciated their sympathy,” “Elimelech was tired,” and “All the mothers were proud” are syntactically well-formed with a meaning that might be conveyed by an explicit complement understood from context. Verb selection is often not optional, however, so that *He put the milk is ungrammatical even if it is clear from context where the milk was put.
The information about the complement types selected by particular verbs and other lexical items is called C-selection or subcategorization, and is included in the lexical entry of the item in our mental lexicon. (Here C stands for “categorial” and is not to be confused with the C that stands for “complementizer”—we apologize for the “clash” of symbols, but that’s what it’s like in the linguistic literature.)
Verbs also include in their lexical entry a specification of certain intrinsic semantic properties of their subjects and complements, just as they select for syntactic categories. This kind of selection is called S-selection (S for semantic). For example, the verb murder requires its subject and object to be human, while the verb drink requires its subject to be animate and its object liquid. Verbs such as like, hate, and so on select animate subjects. The following sentences violate S-selection and can only be used in a metaphorical sense. (We will use the symbol “!” to indicate a semantic anomaly)
!The rock murdered the man.
!The beer drank the student.
!The tree liked the boy.
The famous sentence Colorless green ideas sleep furiously, discussed earlier in this chapter, is anomalous because (among other things) S-selection is violated (e.g., the verb sleep requires an animate subject). In chapter 3 we will discuss the semantic relationships between a verb and its subject and objects in far more detail.
------------------------------------- HAL 105 ------------------------------------
The well-formedness of a phrase depends then on at least two factors: whether the phrase conforms to the structural constraints of the language as expressed in the PS rules, and whether it obeys the selectional requirements of the head, both syntactic (C-selection) and semantic (S-selection).
What Heads the Sentence
Might, could, would—they are contemptible auxiliaries.
GEORGE ELIOT (MARY ANN EVANS), Middlemarch, 1872
We said earlier that all phrases have heads. One category that we have not yet discussed in this regard is sentence (S). For uniformity’s sake, we want all the categories to be headed, but what would the head of S be? To answer this question, let us consider sentences such as the following:
Sam will kick the soccer ball.
Sam has kicked the soccer ball.
Sam is kicking the soccer ball.
Sam may kick the soccer ball.
As noted earlier, words like will, has, is, and may are auxiliary verbs, belonging to the category Aux, which also includes modals such as might, could, would, can, and several others. They occur in structures such as the following one.
(From now on we will adopt the convention of using a triangle under a node when the content of a category is not crucial to the point under discussion.)
Auxiliary verbs specify a time frame for the event (or state) described by the verb, whether it will take place in the future, already took place in the past, or is taking place now. A modal such as may contains “possibility” as part of its meaning, and says it is possible that the event will occur at some future time. The category Aux is a natural category to head S. Just as the VP is about the situation described by the verb—eat ice cream is about “eating”—so a sentence is about a situation or state of affairs that occurs at some point in time.
------------------------------------- HAL 106 ------------------------------------
The parallel with other categories extends further. In the previous PS tree, VP is the complement to Aux. The selectional relationship between Aux and VP is demonstrated by the fact that particular auxiliaries go with particular kinds of VPs. For example, the auxiliary be takes a progressive (-ing) form of the verb,
The boy is dancing.
while the auxiliary have selects a past participle (-en) form of the verb,
The girl has eaten.
and the modals select the infinitival form of the verb (no affixes),
The child must sleep
The boy may eat.
To have a uniform notation, many linguists use the symbols T (= tense) and TP (= tense phrase) instead of Aux and S. Furthermore, just as the NP required the intermediate N-bar (N') category, the TP also has the intermediate T-bar (T') category, as in the phrase structure tree below.
Indeed, many linguists assume that all XPs, where XP stands for any of NP, PP, VP, TP, AdjP, or CP, have three levels of structure. This is referred to as X-bar theory. The basic three-level X-bar schema is as follows:
The first level is the XP itself. The second level consists of a specifier, which functions as a modifier (and which is generally an optional constituent), and an X' (i.e., “X-bar”). For example, an NP specifier is a determiner; a VP specifier is an adverb such as never or often; an AdjP specifier is a degree word such as very or quite. The third level is an expansion of X' and consists of a head X and a complement, which may itself be a phrasal category, thus giving rise to recursion. X-bar structure is thought to be universal, occurring in all the world’s languages, though the order of the elements inside XP and X' may be reversed, as we saw in Japanese
------------------------------------ HAL 107 -------------------------------------
We will not use X-bar conventions in our description of syntax except on the few occasions where the notation provides an insight into the syntax of the language. For sentences we will generally use the more intuitive symbols S and Aux instead of TP and T, but you should think of Aux and S as having the same relationship to each other as V and VP, N and NP, and so on. To achieve this more straightforward approach, we will also ignore the T' category until it is needed later on in the description of the syntax of the main verb be.
Without the use of TP, T', and T, we need an additional PS rule to characterize structures containing Aux:
VP → Aux VP
Like the other recursive VP rules, this rule will allow multiple Aux positions.
This is a desired consequence because English allows sentences with multiple auxiliaries such as:
The child may be sleeping. (modal, be)
The dog has been barking all night. (have, be)
The bird must have been flying home. (modal, have, be)
The introduction of Aux into the system raises a question. Not all sentences seem to have auxiliaries. For example, the sentence “Sam kicked the soccer ball” has no modal, have or be. There is, however, a time reference for this sentence, namely, the past tense on the verb kicked. In sentences without auxiliaries, the tense of the sentence is its head. Instead of having a word under the category Aux (or T), there is a tense specification, present or past, as in the following tree:
------------------------------------ HAL 108 -------------------------------------
The inflection on the verb must match the tense in Aux. For example, if the
tense of the sentence is past, then the verb must have an -ed affix (or must be an irregular past tense verb such as ate).
Thus, in English, and many other languages, the head of S may contain only an abstract tense specification and no actual word, as just illustrated. The actual morpheme, in this case -ed or an irregular past tense form such as went, is inserted into the tree after all the syntactic rules have applied. Most inflectional morphemes, which depend on elements of syntax, are represented in this way. Another example is the tense-bearing word do that is inserted into negative sentences such as John did not go and questions such as Where did John go? In these sentences did means “past tense.” Later in this chapter we will see how do-insertion works.
In addition to specifying the time reference of the sentence, Aux specifies the agreement features of the subject. For example, if the subject is we, Aux contains the features first-person and plural; if the subject is he or she, Aux contains the features third-person and singular. So, another function of the syntactic rules is to use Aux as a “matchmaker” between the subject and the verb. When the subject and the verb bear the same features, Aux makes a match; when they have incompatible features, Aux cannot make a match and the sentence is ungrammatical. This matchmaker function of syntactic rules is more obvious in languages such as Italian, which have many different agreement morphemes, as discussed in chapter 1. Consider the Italian sentence for “I buy books.”
The verb compro, “buy,” in the first sentence bears the first-person singular morpheme, -o, which matches the agreement feature in Aux, which in turn matches the subject Io, “I.” The sentence is therefore grammatical. In the second sentence, there is a mismatch between the first-person subject and the secondperson features in Aux (and on the verb), and so the sentence is ungrammatical.
------------------------------------ HAL 109 -------------------------------------
Structural Ambiguities
The structure of every sentence is a lesson in logic.
JOHN STUART MILL, Inaugural address at St. Andrews, 1867
As mentioned earlier, certain kinds of ambiguous sentences have more than one phrase structure tree, each corresponding to a different meaning. The sentence The boy saw the man with the telescope is structurally ambiguous. Its two meanings correspond to the following two phrase structure trees. (For simplicity we omit Aux in these structures and we return to the non-X-bar notation.)
------------------------------------ HAL 110 -------------------------------------
One meaning of this sentence is “the boy used a telescope to see the man.” The first phrase structure tree represents this meaning. The key element is the position of the PP directly under the VP. Notice that although the PP is under VP, it is not a complement because phrasal categories don’t take complements (only heads do), and because it is not selected by the verb. The verb see selects an NP. In this sentence, the PP has an adverbial function and modifies the verb.
In its other meaning, “the boy saw a man who had a telescope,” the PP with the telescope occurs under the direct object NP, where it modifies the noun man. In this second meaning, the complement of the verb see is the entire NP—the man with the telescope.
The PP in the first structure is generated by the rule
VP → VP PP
In the second structure the PP is generated by the rule
NP → NP PP
Two interpretations are possible because the rules of syntax permit different structures for the same linear order of words. Following is the set of PS rules that we have presented so far in the chapter. The rules have been renumbered.
1. S → NP VP
2. NP → Det N'
3. Det → NP poss
4. NP → N'
5. NP → NP PP
6. N' → Adj N'
7. N' → N
8. VP → V
9. VP → V NP
10. VP → V CP
11. VP → Aux VP
12. VP → VP PP
13. PP → P NP
14. CP → C S
This is not the complete set of PS rules for the language. Various structures in English cannot be generated with these rules, some of which we will talk about later. But even this mini phrase structure grammar generates an infinite set of possible sentences because the rules are recursive. These PS rules specify the word order for English (and other SVO languages, but not for Japanese, say, in which the object comes before the verb). Linear order aside, the hierarchical organization illustrated by these rules is largely true for all languages, as expressed by X-bar schema.
------------------------------------ HAL 111 -------------------------------------
More Structures
Many English sentence types are not accounted for by the phrase structure rules given so far, including:
1. The dog completely destroyed the house.
2. The cat and the dog were friends.
3. The cat is coy.
------------------------------------ HAL 112 -------------------------------------
The sentence in (1) contains the adverb (Adv) completely. Adverbs are modifiers that can specify how an event happens (quickly, slowly, completely) or when it happens (yesterday, tomorrow, often). As modifiers, adverbs are sisters to phrasal (XP) categories. In sentence (1) the adverb is a sister to VP, as illustrated in the following structure (we ignore Aux in this structure):
Temporal adverbs such as yesterday, today, last week, and manner adverbs such as quietly, violently, suddenly, carefully, also occur to the right of VP as follows:
Adverbs also occur as sisters to S (which, recall, is also a phrasal category, TP).
------------------------------------ HAL 113 -------------------------------------
At this point you should be able to write the three PS rules that will account for the position of these adverbs.1
The “Shoe” cartoon’s joke is based on the fact that curse may take an NP complement (“cursed at the day”) and/or be modified by a temporal adverbial phrase (AdvP) (“cursed on the day”), leading to the structural ambiguity:
Interestingly, I cursed the day I was born the day I was born, with both the NP and AdvP modifying the verb, is grammatical and meaningful. (See exercise 23b.) Sentence 2 contains a coordinate structure The cat and the dog. A coordinate structure results when two constituents of the same category (in this case, two NPs) are joined with a conjunction such as and or or. The coordinate NP has the following structure:
Though this may seem counterintuitive, in a coordinate structure the second member of the coordination (NP2) forms a constituent with the conjunction and. We can show this by means of the “move as a unit” constituency test. In sentence (5) the words and a CD move together to the end of the sentence, whereas in (6) the constituent is broken, resulting in ungrammaticality.
4. Caley bought a book and a CD yesterday.
5. Caley bought a book yesterday and a CD.
6. *Caley bought a book and yesterday a CD.
Once again, we encourage you to write the two PS rules that generate this structure.
------------------------------------ HAL 114 -------------------------------------
You can also construct trees for other kinds of coordinate structures, such as VP or PP coordination, which follow the same pattern.
Michael writes poetry and surfs. (VP and VP)
Sam rode his bicycle to school and to the pool. (PP and PP)
Sentence (3) contains the main verb be followed by an adjective. The structure of main verb be sentences is best illustrated using T' notation. The main verb be acts like the modals and the auxiliaries be and have. For example, it is moved to the beginning of the sentence in questions (Is the cat coy?). For this reason we assume that the main verb be occurs under T and takes an XP complement. The XP may be AdjP, as shown in the tree structure for (3):
or an NP or PP as would occur in The cat is a feline or The cat is in the tree.
As before we will leave it as an exercise for you to construct the PS rules for these sentence types and the tree structures they generate.3 (You might try drawing the tree structures; they should look very much like the one above.)
There are also embedded sentence types other than those that we have discussed, for example:
Hilary is waiting for you to sing. (Cf. You sing.)
The host wants the president to leave early. (Cf. The president leaves early.)
The host believes the president to be punctual. (Cf. The president is punctual.)
Although the detailed structure of these different embedded sentences is beyond the scope of this introduction, you should note that an embedded sentence may be an infinitive. An infinitive sentence does not have a tense. The embedded sentences for you to sing, the president to leave early, and the president to be punctual are infinitives. Such verbs as want and believe, among many others, can take infinitival complements. This information, like other selectional properties, belongs to the lexical entry of the selecting verb (the higher verb in the tree).
------------------------------------ HAL 115 -------------------------------------
Sentence Relatedness
I put the words down and push them a bit.
EVELYN WAUGH, quoted in The New York Times, April 11, 1966
Another aspect of our syntactic competence is the knowledge that certain sentences are related to one another, such as the following pair:
The boy is sleeping. Is the boy sleeping?
These sentences describe the same situation. The sentence in the first column asserts that a particular situation exists, a boy-sleeping situation. Such sentences are called declarative sentences. The sentence in the second column asks whether such a boy-sleeping situation holds. Sentences of the second sort are called yesno questions. The only actual difference in meaning between these sentences is that one asserts a situation and the other asks for confirmation of a situation. This element of meaning is indicated by the different word orders, which illustrates that two sentences may have a structural difference that corresponds in a systematic way to a meaning difference. The grammar of the language must account for this fact.
Transformational Rules
Method consists entirely in properly ordering and arranging the things to which we should pay attention.
RENÉ DESCARTES, Oeuvres, vol. X, c. 1637
Phrase structure rules account for much of our syntactic knowledge, but they do not account for the fact that certain sentence types in the language relate systematically to other sentence types. The standard way of describing these relationships is to say that the related sentences come from a common underlying structure. Yes-no questions are a case in point, and they bring us back to a discussion of auxiliaries. Auxiliaries are central to the formation of yes-no questions as well as certain other types of sentences in English. In yes no questions, the auxiliary appears in the position preceding the subject. Here are a few more examples:
The boy is sleeping. Is the boy sleeping?
The boy has slept. Has the boy slept?
The boy can sleep. Can the boy sleep?
The boy will sleep. Will the boy sleep?
A way to capture the relationship between a declarative and a yes-no question is to allow the PS rules to generate a structure corresponding to the declarative sentence. Another formal device, called a transformational rule, then moves the auxiliary before the subject. The rule “Move Aux” is formulated as follows:
------------------------------------ HAL 116 -------------------------------------
Move the highest Aux to adjoin to (the root) S.
That is, Move Aux applies to structures like:
The “__” shows the position from which the Aux is moved. For example:
The boy is sleeping. → Is the boy __ sleeping?
The rule takes the basic (NP-Aux) structure generated by the phrase structure rules and derives a second tree (the dash represents the position from which a constituent has been moved). The Aux is attached to the tree by adjunction. Adjunction is an operation that copies an existing node (in this case S) and creates a new level to which the moved category (in this case Aux) is appended.
------------------------------------ HAL 117 ------------------------------------------------------------------------- HAL 118 ------------------------------------------------------------------------- HAL 119 ------------------------------------------------------------------------- HAL 120--------------------------------------
More Structures
Many English sentence types are not accounted for by the phrase structure rules given so far, including:
1. The dog completely destroyed the house.
2. The cat and the dog were friends.
3. The cat is coy.
------------------------------------ HAL 112 -------------------------------------
The sentence in (1) contains the adverb (Adv) completely. Adverbs are modifiers that can specify how an event happens (quickly, slowly, completely) or when it happens (yesterday, tomorrow, often). As modifiers, adverbs are sisters to phrasal (XP) categories. In sentence (1) the adverb is a sister to VP, as illustrated in the following structure (we ignore Aux in this structure):
Temporal adverbs such as yesterday, today, last week, and manner adverbs such as quietly, violently, suddenly, carefully, also occur to the right of VP as follows:
Adverbs also occur as sisters to S (which, recall, is also a phrasal category, TP).
------------------------------------ HAL 113 -------------------------------------
At this point you should be able to write the three PS rules that will account for the position of these adverbs.1
The “Shoe” cartoon’s joke is based on the fact that curse may take an NP complement (“cursed at the day”) and/or be modified by a temporal adverbial phrase (AdvP) (“cursed on the day”), leading to the structural ambiguity:
Interestingly, I cursed the day I was born the day I was born, with both the NP and AdvP modifying the verb, is grammatical and meaningful. (See exercise 23b.) Sentence 2 contains a coordinate structure The cat and the dog. A coordinate structure results when two constituents of the same category (in this case, two NPs) are joined with a conjunction such as and or or. The coordinate NP has the following structure:
Though this may seem counterintuitive, in a coordinate structure the second member of the coordination (NP2) forms a constituent with the conjunction and. We can show this by means of the “move as a unit” constituency test. In sentence (5) the words and a CD move together to the end of the sentence, whereas in (6) the constituent is broken, resulting in ungrammaticality.
4. Caley bought a book and a CD yesterday.
5. Caley bought a book yesterday and a CD.
6. *Caley bought a book and yesterday a CD.
Once again, we encourage you to write the two PS rules that generate this structure.
------------------------------------ HAL 114 -------------------------------------
You can also construct trees for other kinds of coordinate structures, such as VP or PP coordination, which follow the same pattern.
Michael writes poetry and surfs. (VP and VP)
Sam rode his bicycle to school and to the pool. (PP and PP)
Sentence (3) contains the main verb be followed by an adjective. The structure of main verb be sentences is best illustrated using T' notation. The main verb be acts like the modals and the auxiliaries be and have. For example, it is moved to the beginning of the sentence in questions (Is the cat coy?). For this reason we assume that the main verb be occurs under T and takes an XP complement. The XP may be AdjP, as shown in the tree structure for (3):
or an NP or PP as would occur in The cat is a feline or The cat is in the tree.
As before we will leave it as an exercise for you to construct the PS rules for these sentence types and the tree structures they generate.3 (You might try drawing the tree structures; they should look very much like the one above.)
There are also embedded sentence types other than those that we have discussed, for example:
Hilary is waiting for you to sing. (Cf. You sing.)
The host wants the president to leave early. (Cf. The president leaves early.)
The host believes the president to be punctual. (Cf. The president is punctual.)
Although the detailed structure of these different embedded sentences is beyond the scope of this introduction, you should note that an embedded sentence may be an infinitive. An infinitive sentence does not have a tense. The embedded sentences for you to sing, the president to leave early, and the president to be punctual are infinitives. Such verbs as want and believe, among many others, can take infinitival complements. This information, like other selectional properties, belongs to the lexical entry of the selecting verb (the higher verb in the tree).
------------------------------------ HAL 115 -------------------------------------
Sentence Relatedness
I put the words down and push them a bit.
EVELYN WAUGH, quoted in The New York Times, April 11, 1966
Another aspect of our syntactic competence is the knowledge that certain sentences are related to one another, such as the following pair:
The boy is sleeping. Is the boy sleeping?
These sentences describe the same situation. The sentence in the first column asserts that a particular situation exists, a boy-sleeping situation. Such sentences are called declarative sentences. The sentence in the second column asks whether such a boy-sleeping situation holds. Sentences of the second sort are called yesno questions. The only actual difference in meaning between these sentences is that one asserts a situation and the other asks for confirmation of a situation. This element of meaning is indicated by the different word orders, which illustrates that two sentences may have a structural difference that corresponds in a systematic way to a meaning difference. The grammar of the language must account for this fact.
Transformational Rules
Method consists entirely in properly ordering and arranging the things to which we should pay attention.
RENÉ DESCARTES, Oeuvres, vol. X, c. 1637
Phrase structure rules account for much of our syntactic knowledge, but they do not account for the fact that certain sentence types in the language relate systematically to other sentence types. The standard way of describing these relationships is to say that the related sentences come from a common underlying structure. Yes-no questions are a case in point, and they bring us back to a discussion of auxiliaries. Auxiliaries are central to the formation of yes-no questions as well as certain other types of sentences in English. In yes no questions, the auxiliary appears in the position preceding the subject. Here are a few more examples:
The boy is sleeping. Is the boy sleeping?
The boy has slept. Has the boy slept?
The boy can sleep. Can the boy sleep?
The boy will sleep. Will the boy sleep?
A way to capture the relationship between a declarative and a yes-no question is to allow the PS rules to generate a structure corresponding to the declarative sentence. Another formal device, called a transformational rule, then moves the auxiliary before the subject. The rule “Move Aux” is formulated as follows:
------------------------------------ HAL 116 -------------------------------------
Move the highest Aux to adjoin to (the root) S.
That is, Move Aux applies to structures like:
The “__” shows the position from which the Aux is moved. For example:
The boy is sleeping. → Is the boy __ sleeping?
The rule takes the basic (NP-Aux) structure generated by the phrase structure rules and derives a second tree (the dash represents the position from which a constituent has been moved). The Aux is attached to the tree by adjunction. Adjunction is an operation that copies an existing node (in this case S) and creates a new level to which the moved category (in this case Aux) is appended.
------------------------------------ HAL 117 ------------------------------------------------------------------------- HAL 118 ------------------------------------------------------------------------- HAL 119 ------------------------------------------------------------------------- HAL 120--------------------------------------
Langganan:
Posting Komentar
(
Atom
)
Tidak ada komentar :
Posting Komentar