Eng Syntax
Eng Syntax
Det N V Det N
. . . . . . . . . . . . . . .
We assume a lexicon, a list of categorized words, to be part of the grammar along with the rule
in (32):
(34) a. Det: a, that, the, this, . . .
b. N: ball, man, piano, string, student, . . .
c. V: kicked, hit, played, sang, threw, . . .
d. A: handsome, happy, kind, long, tall, . . .
By inserting lexical items into the appropriate pre-terminal nodes in the structure, where the
labels above . . . are, we can generate grammatical examples like those (31) as well as those like
the following, not all of which describe a possible real-world situation:
(35) a. That ball hit a student.
b. The piano played a song.
c. The piano kicked a student.
d. That ball sang a student.
Such examples are all syntactically well-formed, even if semantically in some cases, implying
that syntax is rather autonomous from semantics. Note that any anomalous example can be
preceded by the statement Now, heres something hard to imagine:....
12
Notice that even this simple grammar rule can easily extend to generate an innite number
of English sentences by allowing iteration of the A:
13
(36) S Det A
N V Det A
N
12
See Exercise 9 of this chapter and the discussion of selectional restrictions in Chapter 4.
13
This iteration operator
is called the Kleene Star Operator, and is a notation meaning zero to innitely many
occurrences. It should not be confused with the * prexed to a linguistic example, indicating ungrammaticality.
17
The operator allows us to repeat any number of As, thereby generating sentences like (37). Note
that the parentheses around A in (33) are no longer necessary in this instance, for the Kleene
Star operator means any number including zero.
(37) a. The tall man kicked the ball.
b. The tall, handsome man kicked the ball.
c. The tall, kind, handsome man kicked the ball.
One could even generate a sentence like (38):
(38) The happy, happy, happy, happy, happy, happy man sang a song.
A grammar using only lexical categories can be specied to generate an innite number of
well-formed English sentences, but it nevertheless misses a great deal of basic properties that
we can observe. For example, this simple grammar cannot capture the agreement facts seen in
examples like the following:
(39) a. The mother of the boy and the girl is arriving soon.
b. The mother of the boy and the girl are arriving soon.
Why do the verbs in these two sentences have different agreement patterns? Our intuitions tell
us that the answer lies in two different possibilities for grouping the words:
(40) a. [The mother of [the boy and the girl]] is arriving soon.
b. [The mother of the boy] and [the girl] are arriving soon.
The different groupings shown by the brackets indicate who is arriving: in (40a), the mother,
while in (40b) it is both the mother and the girl. The grouping of words into larger phrasal units
which we call constituents provides the rst step in understanding the agreement facts in (40).
Now, consider the following examples:
(41) a. John saw the man with a telescope.
b. I like chocolate cakes and pies.
c. We need more intelligent leaders.
These sentences have different meanings depending on how we group the words. For example,
(41a) will have the following two different constituent structures:
(42) a. John saw [the man with a telescope].
(the man had the telescope)
b. John [[saw the man] with a telescope].
(John used the telescope)
Even these very cursory observations indicate that a grammar with only lexical categories is
not adequate for describing syntax. In addition, we need a notion of constituent, and need to
consider how phrases may be formed, grouping certain words together.
18
2.4 Phrasal Categories
In addition to the agreement and ambiguity facts, our intuitions may also lead us to hypothesize
constituency. If you were asked to group the words in (43) into phrases, what constituents would
you come up with?
(43) The student enjoyed his English syntax class last semester.
Perhaps most of us would intuitively assign the structure given in (44a), but not those in (44b)
or (44c):
(44) a. [The student] [enjoyed [his English syntax class last semester]].
b. [The] [student enjoyed] [his English syntax class] [last semester].
c. [The student] [[enjoyed his English] [syntax class last semester]].
What kind of knowledge, in addition to semantic coherence, forms the basis for our intuitions of
constituency? Are there clear syntactic or distributional tests which demonstrate the appropriate
grouping of words or specic constituencies? There are certain salient syntactic phenomena
which refer directly to constituents or phrases.
Cleft: The cleft construction, which places an emphasized or focused element in the X posi-
tion in the pattern It is/was X that . . . , can provide us with simple evidence for the existence
of phrasal units. For instance, think about how many different cleft sentences we can form from
(45).
(45) The policeman met several young students in the park last night.
With no difculty, we can cleft almost all the constituents we can get from the above sentence:
(46) a. It was [the policeman] that met several young students in the park last night.
b. It was [several young students] that the policeman met in the park last night.
c. It was [in the park] that the policeman met several young students last night.
d. It was [last night] that the policeman met several young students in the park.
However, we cannot cleft sequences that not form constituents:
14
(47) a. *It was [the policeman met] that several young students in the park last night.
b. *It was [several young students in] that the policeman met the park last night.
c. *It was [in the park last night] that the policeman met several young students.
Constituent Questions and Stand-Alone Test: Further support for the existence of phrasal
categories can be found in the answers to constituent questions, which involve a wh-word such
as who, where, when, how. For any given wh-question, the answer can either be a full sentence
or a fragment. This stand-alone fragment is a constituent:
(48) A: Where did the policeman meet several young students?
14
The verb phrase constituent met . . . night here cannot be clefted for independent reasons (see Chapter 12).
19
B: In the park.
(49) A: Who(m) did the policeman meet in the park?
B: Several young students.
This kind of test can be of use in determining constituents; we will illustrate with example (50):
(50) John put old books in the box.
Are either old books in the box or put old books in the box a constituent? Are there smaller
constituents? The wh-question tests can provide some answers:
(51) A: What did you put in your box?
B: Old books.
B: *Old books in the box.
(52) A: Where did you put the book?
B: In the box.
B:*Old books in the box.
(53) A: What did you do?
B: *Put old books.
B: *Put in the box.
B: Put old books in the box.
Overall, the tests here will show that old books and in the box are constituents, and that put old
books in the box is also a (larger) constituent.
The test is also sensitive to the difference between particles and prepositions. Consider the
similar-looking examples in (54), including looked and up:
(54) a. John looked up the inside of the chimney.
b. John looked up the meaning of chanson.
The examples differ, however, as to whether up forms a constituent with the following material
or not. We can again apply the wh-question test:
(55) A: What did he look up?
B: The inside of the chimney.
B: The meaning of chanson.
(56) A: Where did he look?
B: Up the inside of the chimney.
B:*Up the meaning of chanson.
(57) A: Up what did he look?
B: The inside of the chimney.
B: *The meaning of chanson.
20
What the contrasts here show is that up forms a constituent with the inside of the chimney in
(54a) whereas it does not with the meaning of chanson in (54b).
Substitution by a Pronoun: English, like most languages, has a system for referring back to
individuals or entities mentioned by the use of pronouns. For instance, the man who is standing
by the door in (58a) can be substituted by the pronoun he in (58b).
(58) a. What do you think the man who is standing by the door is doing now?
b. What do you think he is doing now?
There are other pronouns such as there, so, as, and which, which also refer back to other con-
stituents.
(59) a. Have you been [to Seoul]? I have never been there.
b. John might [go home], so might Bill.
c. John might [pass the exam], and as might Bill.
d. If John can [speak French uently] which we all know he can we will have
no problems.
A pronoun cannot be used to refer back to something that is not a constituent:
(60) a. John asked me to put the clothes in the cupboard, and to annoy him I really
stuffed them there [there=in the cupboard].
b. John asked me to put the clothes in the cupboard, and to annoy him I stuffed
them there [them=the clothes].
c. *John asked me to put the clothes in the cupboard, but I did so [=put the clothes]
in the suitcase.
Both the pronoun there and them refer to a constituent. However, so in (60c), referring to a VP,
refers only part of a constituent put the clothes, making it unacceptable.
Coordination: Another commonly-used test is coordination. Words and phrases can be co-
ordinated by conjunctions, and each conjunct is typically the same kind of constituent as the
other conjuncts:
(61) a. The girls [played in the water] and [swam under the bridge].
b. The children were neither [in their rooms] nor [on the porch].
c. She was [poor] but [quite happy].
d. Many people drink [beer] or [wine].
If we try to coordinate unlike constituents, the results are typically ungrammatical.
(62) a. *Mary waited [for the bus] and [to go home].
b. *Lee went [to the store] and [crazy].
Even though such syntactic constituent tests are limited in certain cases, they are often
adopted in determining the constituent of given expressions.
21
2.5 Phrase Structure Rules
We have seen evidence for the existence of phrasal categories. We say that phrases are projected
from lexical categories, and hence we have phrases such as NP, VP, PP, and so on. As before,
we use distributional evidence to classify each type, and then specify rules to account for the
distributions we have observed.
2.5.1 NP: Noun Phrase
Consider (63):
(63) [liked ice cream].
The expressions that can occur in the blank position here are once again limited. The kinds of
expression that do appear here include:
(64) Mary, I, you, students, the students, the tall students, the students from Seoul, the
students who came from Seoul, etc.
If we look into the sub-constituents of these expressions, we can see that each includes at least
an N and forms an NP (noun phrase). This leads us to posit the following rule:
15
(65) NP (Det) A* N (PP/S)
This rule characterizes a phrase, and is one instance of a phrase structure rule (PS rule). The rule
indicates that an NP can consist of an optional Det, any number of optional A, an obligatory N,
and then an optional PP or a modifying S.
16
The slash indicates different options for the same
place in the linear order. These options in the NP rule can be represented in a tree structure:
(66)
NP
.
.
.
.
.
.
.
.
.
.
.
.
.
(Det) A* N (PP/S)
. . . . . . . . . . . .
Once we insert appropriate expressions into the pre-terminal nodes, we will have well-formed
NPs; and the rule will not generate the following NPs:
(67) *the whistle tune, *the easily student, *the my dog, . . .
One important point is that as only N is obligatory in NP, a single noun such as Mary, you, or
students can constitute an NP by itself. Hence the subject of the sentence She sings will be an
NP, even though that NP consists only of a pronoun.
15
The relative clause who came from Seoul is kind of modifying sentence (S). See Chapter 11.
16
To license an example like the very tall man, we need to make A* as AP*. For simplicity, we just use the former in
the rule.
22
2.5.2 VP: Verb Phrase
Just as N projects an NP, V projects a VP. A simple test environment for VP is given in (68).
(68) The student .
(69) lists just a few of the possible phrases that can occur in the underlined position.
(69) snored, ran, sang, loved music, walked the dog through the park, lifted 50 pounds,
thought Tom is honest, warned us that storms were coming, etc.
These phrases all have a V as their head as projections of V, they form VP. VP can be charac-
terized by the rule in (70), to a rst level of analysis:
(70) VP V (NP) (PP/S)
This simple VP rule says that a VP can consist of an obligatory V followed by an optional NP
and then any number of PPs or an S. The rule thus does not generate ill-formed VPs such as
these:
(71) *leave the meeting sing, *the leave meeting, *leave on time the meeting, . . .
We can also observe that the presence of a VP is essential in forming a grammatical S, and the
VP must be nite (present or past tense). Consider the following examples:
(72) a. The monkey wants to leave the meeting.
b. *The monkey eager to leave the meeting.
(73) a. The monkeys approved of their leader.
b. *The monkeys proud of their leader.
(74) a. The men practice medicine.
b. *The men doctors of medicine.
These examples show us that an English well-formed sentence consists of an NP and a (nite)
VP, which can be represented as a PS rule:
(75) S NP VP
We thus have the rule that English sentences are composed of an NP and a VP, the precise
structural counterpart of the traditional ideas of a sentence being a subject and predicate or a
noun and a verb.
One more aspect to the structure of VP involves the presence of auxiliary verbs. Think of
continuations for the fragments in (76):
(76) a. The students .
b. The students want .
For example, the phrases in (77a) and (77b) can occur in (76a) whereas those in (77c) can appear
in (76b).
(77) a. run, feel happy, study English syntax, . . .
b. can run, will feel happy, must study English syntax, . . .
23
c. to run, to feel happy, to study English syntax, . . .
We have seen that the expressions in (77a) all form VPs, but how about those in (77b) and
(77c)? These are also VPs, which happen to contain more than one V. In fact, the parts after
the auxiliary verbs in (77b) and (77c) are themselves regular VPs. In the full grammar we
will consider to and can and so on as auxiliary verbs, with a feature specication [AUX +] to
distinguish them from regular verbs. Then all auxiliary verbs are simply introduced by a second
VP rule:
17
(78) VP V[AUX +] VP
One more important VP structure involves the VP modied by an adverb or a PP:
(79) a. John [[read the book] loudly].
b. The teacher [[met his students] in the class].
In such examples, the adverb loudly and the PP in the class are modifying the preceding VP. To
form such VPs, we need the PS rule in (80):
(80) VP VP Adv/PP
This rule, together with (75) will allow the following structure for (79b):
18
(81)
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
The teacher VP
PP
VP
.
.
.
.
.
.
.
.
.
.
.
.
Det A N V NP
PP
. . . . . . . . . . . . Det N P NP
. . . . . . . . . Det N
. . . . . .
With the structural possibilities shown here, let us assume that we have the following lexical
entries:
(99) a. Det: a, an, this, that, any, some, which, his, her, no, etc.
b. A: handsome, tall, fat, large, dirty, big, yellow, etc.
c. N: book, ball, hat, friend, dog, cat, man, woman, John, etc.
d. V: kicked, chased, sang, met, believed, thinks, imagines, assumes etc.
Inserting these elements in the appropriate pre-terminal nodes (the places with dots) in (98), we
are able to generate various sentences like those in (100):
21
(100) a. This handsome man chased a dog.
b. A man kicked that ball.
c. That tall woman chased a cat.
d. His friend kicked a ball.
There are several ways to generate an innite number of sentences with this kind of grammar.
As we have seen before, one simple way is to repeat a category like A innitely. There are also
other ways of generating an innite number of grammatical sentences. Look at the following
two PS rules from (97) again:
(101) a. S NP VP
b. VP V S
As we show in the following tree structure, we can recursively apply the two rules, in the sense
that one can feed the other, and then vice versa:
21
The grammar still generates semantically anomalous examples like # The desk believed a man or # A man sang
her hat. For such semantically distorted examples, we need to refer to the notion of selectional restrictions (see Chapter
7).
27
(102)
S
7654 0123
NP VP
?>=< 89:;
N V S
7654 0123
John believes NP VP
?>=< 89:;
N V S
NP VP
7654 0123
N V[AUX +] VP
7654 0123
They will V NP
study English syntax
This means that we will also have a recursive structure like the following:
22
22
Due to the limited number of auxiliary verbs, and restrictions on their cooccurrence, the maximum number of
auxiliaries in a single English clause is 3.
28
(105)
S
.
.
.
.
.
.
.
.
.
NP VP
.
.
.
.
.
.
.
.
.
?>=< 89:;
N V [AUX +] VP
.
.
.
.
.
.
.
.
.
7654 0123
They will V[AUX +] VP
.
.
.
.
.
.
.
.
.
?>=< 89:;
have V[AUX +] VP
.
.
.
.
.
.
.
.
.
.
.
PP
V NP
Conj PP
Conj AP
Part
put the customers off
As represented here, the particle does not form a constituent with the following or preceding
NP whereas the preposition does form a constituent with it.
In summary, we have seen that a grammar with lexical categories can not only generate
an innite number of grammatical English sentences, it does account for some fundamental
properties, such as agreement and constituency.
25
This motivates the introduction of phrases
into the grammar.
25
In this chapter, we have not discussed the treatment of agreement with PS rules. Chapter 6 discusses the subject-
verb agreement in detail.
31
2.7 Exercises
1. Discuss the categorial status of each of the words in the following sentences, giving
detailed reasons (based on meaning, form, or function) in support of your analysis.
(i) a. Oil companies will have to pass on all of the benets of tax reform to the
consumer.
b. That young child in the corner probably will fall off his bed onto the cold,
hard oor early in the morning.
2. Consider the lexical category status of italicised nonsense words in the following sen-
tences and provide arguments in support of your analysis.
26
(i) a. John blonks on Sundays.
b. John likes to blonk in the afternoons.
c. John was feeling murgy, but happy.
d. Hes murgier than anyone I know.
e. John is a garoon, and so is Fred.
f. In fact, theyre both typical garoons.
g. She put the car ong the garage.
3. Determine the lexical category of for and that in the following examples and decide what
kind of PS rule(s) we need to account for them.
(i) a. It is important for us to spend time with children.
b. He was arrested for being drunk.
c. I think that person we met last week is insane.
d. We believe that he is quite reasonable.
e. I forgot to return the book that I borrowed from the teacher.
4. Consider the following data carefully and describe the similarities and differences among
that, for, if and whether. In so doing, rst compare that and for and then see how these
two are different from if and whether. Finally, see how if and whether are similar or
different from wh-words like what and where.
(i) a. I am anxious that you should arrive on time.
b. *I am anxious that you to arrive on time.
(ii) a. I am anxious for you to arrive on time.
b. *I am anxious for you should arrive on time.
(iii) a. I dont know whether/if I should agree.
b. I wonder whether/if youd be kind enough to give us information.
(vi) a. If students study hard, teachers will be happy.
26
This exercise is adopted from Radford (1988).
32
b. Whether they say it or not, most teachers expect their students to study
hard.
(iv) a. I dont know whether/*if to agree.
b. I dont know *that to agree.
c. I dont know what to do.
d. I dont know where to go.
(v) a. I am not certain about when he will come.
b. I am not certain about whether he will go or not.
c. I am not certain about *if he will go or not.
5. Check if the underlined parts form a constituent or not with at least two constituenthood
tests (e.g., cleft, pronoun substitution, stand-alone, etc). Also provide tree structures for
each of the following sentences.
(i) a. John bought a book on the table.
b. John put a book on the table.
(ii) a. She turned down the side street
b. She turned down his offer.
(iii) a. He looked at a book about swimming.
b. He talked to a girl about swimming.
c. He talked with a girl about swimming.
(iv) a. I dont know the people present.
b. John called the president a fool.
6. Explain why the examples in (i) are ungrammatical. Do this by drawing trees for each
sentence while referring to the PS rules related to particles and coordination.
(i) a. *Could you turn off the re and on the light?
b. *A nuclear explosion would wipe out plant life and out animal life.
c. *He ran down the road and down the President.
d. *I know the truth and that you are innocent.
e. *Lee went to the store and crazy.
7. Provide a tree structure for each of the following sentences and suggest what kind of
rules for VP will be necessary. In doing so, pay attention to the position of modiers like
proudly, by the park, and so forth.
(i) a. John refused the offer proudly.
b. I consider Telma the best candidate.
c. I saw him leaving the main building.
d. He took Masako to the school by the park.
e. John sang a song and danced to the music.
f. John wants to study linguistics in near future.
g. They told Angelica to arrive early for the award.
h. That Louise had abandoned the project surprised everyone.
33
8. Each of the following sentences is structurally ambiguous. Represent their structural
ambiguities by providing different tree structures for each string of words.
(i) a. I know you like the back of my hand.
b. I forgot how good beer tastes.
c. I saw that gas can explode.
d. I saw her duck.
e. Time ies like an arrow.
f. I need to have that report on my desk by tomorrow.
9. Provide tree structures for the bracketed clauses or sentences. See if there are any PS
rules that we need to introduce.
(i) [Different languages may have different lexical categories], or they might
associate different properties to the same one. For example, Spanish uses ad-
jectives almost interchangeably as nouns while English cannot. [Japanese has
two classes of adjectives whereas English has one]; Korean, Japanese, and
Chinese have measure words while European languages have nothing resem-
bling them; [many languages dont have a distinction between adjectives and
adverbs], or adjectives and nouns, etc. [Many linguists argue that the formal
distinctions between parts of speech must be made within the framework of a
specic language or language family], and should not be carried over to other
languages or language families.
27
27
Adapted from http://en.wikipedia.org/wiki/Part of speech
34
3
Syntactic Forms, Grammatical Functions, and
Semantic Roles
3.1 Introduction
In the previous chapter, we analyzed English sentences with PS rules. For example, the PS
rule S NP VP represents the basic rule for forming well-formed English sentences. As we
have seen, such PS rules allow us to represent the constituent structure of a given sentence in
terms of lexical and phrasal syntactic categories. There are other dimensions of the analysis
of sentences; one such way is using the notion of grammatical functions such as subject and
object:
(1) a. Syntactic categories: N, A, V, P, NP, VP, AP, . . .
b. Grammatical functions: SUBJ (Subject), OBJ (Object), MOD (Modier), PRED
(Predicate), . . .
The notions such as SUBJ, OBJ and PRED represent the grammatical function each constituent
plays in the given sentence. For example, consider one simple sentence:
(2) The monkey kicked a boy on Monday.
This sentence can be structurally represented in terms of either syntactic categories or gram-
matical functions as in the following:
(3) a. [
S
[
NP
The monkey] [
VP
kicked [
NP
a boy] [
PP
on Monday]]].
b. [
S
[
SUBJ
The monkey] [
PRED
kicked [
OBJ
a boy] [
MOD
on Monday]]].
As shown here, the monkey is an NP in terms of its syntactic form, but is the SUBJ (subject)
in terms of its grammatical function. The NP a boy is the OBJ (object) while the verb kicked
functions as a predicator. More importantly, we consider the entire VP to be a PRED (predicate)
which describes a property of the subject. On Monday is a PP in terms of its syntactic category,
but serves as a MOD (modier) here.
We also can represent sentence structure in terms of semantic roles. Constituents can be
considered in terms of conceptual notions of semantic roles such as agent, patient, location,
35
instrument, and the like. A semantic role denotes the underlying relationship that a participant
has with the relation of the clause, expressed by the main verb. Consider the semantic roles of
the NPs in the following two sentences:
28
(4) a. John tagged the monkey in the forest.
b. The monkey was tagged in the forest by John.
Both of these sentences describe a situation in which someone named John tagged a particular
monkey. In this situation, John is the agent and the monkey is the patient of the tagging event.
This in turn means that in both cases, John has the semantic role of agent (agt), whereas the
monkey has the semantic role of patient (pat), even though their grammatical functions are
different. We thus can assign the following semantic roles to each constituent of the examples:
(5) a. [[
agt
John] [
pred
tagged [
pat
the monkey] [
loc
in the forest]]].
b. [
S
[
pat
The monkey] [
pred
was tagged [
loc
in the wood] [
agt
by John]]].
As noted here, in addition to agt (agent) and pat (patient), we have pred (predicate) and loc
(locative) semantic roles that express the semantic role that each expression performs in the
described situation.
Throughout this book we will see that English grammar refers to these three different levels
of information (syntactic category, grammatical function, and semantic role), and they interact
with each other. For now, it may appear that they are equivalent classications: for example, an
agent is a subject and an NP, and a patient is an object and an NP. However, as we get further
into the details of the grammar, we will see many ways in which the three levels are not simply
co-extensive.
3.2 Grammatical Functions
How can we identify the grammatical function of a given constituent? Several tests can be used
to determine grammatical function, as we show here.
3.2.1 Subjects
Consider the following pair of examples:
(6) a. [The cat] [devoured [the rat]].
b. [The rat] [devoured [the cat]].
These two sentences have exactly the same words and have the same predicator devoured. Yet
they are signicantly different in meaning, and the main difference comes from what serves as
subject or object with respect to the predicator. In (6a), the subject is the cat, whereas in (6b)
it is the rat, and the object is the rat in (6a) but the cat in (6b).
The most common structure for a sentence seems to be one in which the NP subject is the one
who performs the action denoted by the verb (thus having the semantic role of agent). However,
28
Semantic roles are also often called thematic roles or -roles (theta roles) in generative grammar (Chomsky
1982, 1986).
36
this is not always so:
(7) a. My brother wears a green overcoat.
b. This car stinks.
c. It rains.
d. The committee disliked her proposal.
Wearing a green overcoat, stinking, raining, or disliking ones proposal are not agentive activi-
ties; they indicate stative descriptions or situations. Such facts show that we cannot rely on the
semantic roles of agent for determining subjecthood.
More reliable tests for subjecthood come from syntactic tests such as agreement, tag ques-
tions, and subject-auxiliary inversion.
Agreement: The main verb of a sentence agrees with the subject in English:
(8) a. She never writes/*write home.
b. These books *saddens/sadden me.
c. Our neighbor takes/*take his children to school in his car.
As we noted in Chapter 1, simply being closer to the main verb does not entail subjecthood:
(9) a. The book, including all the chapters in the rst section, is/*are very interesting.
b. The effectiveness of teaching and learning *depend/depends on several factors.
c. The tornadoes that tear through this county every spring *is/are more than just a
nuisance.
The subject in each example is book, effectiveness, and tornadoes respectively, even though
there are nouns closer to the main verb. This indicates that it is not simply the linear position of
the NP that determines agreement; rather, agreement shows us what the subject of the sentence
is.
Tag questions: A tag question, a short question tagged onto the end of an utterance, is also a
reliable subjecthood test:
(10) a. The lady singing with a boy is a genius, isnt she/*isnt he?
b. With their teacher, the kids have arrived safely, havent they/ *hasnt he?
The pronoun in the tag question agrees with the subject in person, number, and gender it refers
back to the subject, but not necessarily to the closest NP, nor to the most topical one. The she in
(10a) shows us that lady is the head of the subject NP in that example, and they in (10b) leads
us to assign the same property to kids. The generalization is that a tag question must contain a
pronoun which identies the subject of the clause to which the tag is attached.
Subject-auxiliary inversion: In forming questions and other sentence-types, English has
subject-auxiliary inversion, which applies only to the subject.
37
(11) a. This teacher is a genius.
b. The kids have arrived safely.
c. It could be more detrimental.
(12) a. Is this teacher a genius?
b. Have the kids arrived safely?
c. Could it be more detrimental?
As seen here, the formation of Yes/No questions such as these involves the rst tensed auxil-
iary verb moving across the subject: more formally, the auxiliary verb is inverted with respect to
the subject, hence the term subject-auxiliary inversion. This is not possible with a non-subject:
(13) a. The kids in our class have arrived safely.
b. *Have in our class the kids arrived safely?
Subject-auxiliary inversion provides another reliable subjecthood test.
3.2.2 Direct and Indirect Objects
A direct object (DO) is canonically an NP, undergoing the process denoted by the verb:
(14) a. His girlfriend bought this computer.
b. That silly fool broke the teapot.
However, this is not a solid generalization. The objects in (15a) and (15b) are not really affected
by the action. In (15a) the dog is experiencing something, and in (15b) the thunder is somehow
causing some feeling in the dog:
(15) a. Thunder frightens [the dog].
b. The dog fears [thunder].
Once again, the data show us that we cannot identify the object based on semantic roles. A
much more rm criterion is the syntactic construction of passivization, in which a notional
direct object appears as subject. The sentences in (16) can be turned into passive sentences in
(17):
(16) a. His girlfriend bought this computer for him.
b. The child broke the teapot by accident.
(17) a. This computer was bought for him by his girlfriend.
b. The teapot was broken by the child by accident.
What we can notice here is that the objects in (16) are promoted to subject in the passive
sentences. The test comes from the fact that non-object NPs cannot be promoted to the subject:
(18) a. This item belongs to the student.
b. *The student is belonged to by this item.
(19) a. He remained a good friend to me.
b. *A good friend is remained to me (by him).
38
The objects that undergo passivization are direct objects, distinct from indirect objects.
An indirect object (IO) is one which precedes a direct object (DO), as in (20); IOs are NPs
and have the semantic roles of goal, recipient, or benefactive:
(20) a. I threw [the puppy] [the ball]. (IO = goal)
b. John gave [the boys] [the CDs]. (IO = recipient)
c. My mother baked [me] [a birthday cake]. (IO = benefactive)
In examples like (20), passive has the property of making the IO into the subject.
(21) a. The boys were given the CDs (by John).
b. She was sent a review copy of the book (by the publisher).
Note that sentences with the IO-DO order are different from those where the semantic role
of the IO is expressed as an oblique PP, following the DO (such a process is sometimes called
dative alteration):
(22) a. John gave the CDs to the boys.
b. The publisher sent a review copy of the book to her.
c. My mother baked a cake for me.
In this kind of example, it is once again the DO which can be passivized, giving examples like
the following:
(23) a. The CDs were given to the boys by John.
b. A review copy of the book was sent to her by the publisher.
c. This nice cake was baked for me by my mother.
3.2.3 Predicative Complements
There also are NPs which follow a verb but which do not behave as DOs or IOs. Consider the
following sentences:
(24) a. This is my ultimate goal.
b. Michelle became an architect.
(25) a. They elected Graham chairman.
b. I consider Andrew the best writer
The italicized elements here are traditionally called predicative complements in the sense that
they function as the predicate of the subject or the object. However, even though they are NPs,
they do not passivize:
(26) a. *Chairman was elected Graham.
b. *The best writer was considered Andrew.
The difference between objects and predicative complements can also be seen in the following
contrast:
(27) a. John made Kim a great doll.
39
b. John made Kim a great doctor.
Even though the italicized expressions here are both NPs, they function differently. The NP a
great doll in (27a) is the direct object, as in John made a great doll for Kim, whereas the NP
a great doctor in (27b) cannot be an object: it serves as the predicate of the object Kim. If we
think of part of the meaning informally, only in the second example would we say that the nal
NP describes the NP Kim.
(28) a. (27)a: Kim = a great doll
b. (27)b: Kim = a great doctor
In addition, phrases other than NPs can serve as predicative complements:
(29) a. The situation became terrible.
b. This map is what he wants.
c. The message was that you should come on time.
(30) a. I made Kim angry.
b. I consider him immoral.
c. I regard Andrew as the best writer.
d. They spoil their kids rotten.
The italicized complements function to predicate a property of the subject in (29) and of the
object in (30).
3.2.4 Oblique Complements
Consider now the italicized expressions in (31):
(31) a. John put books in the box.
b. John talked to Bill about the exam.
c. She reminded him of the last time they met.
d. They would inform Mary of any success they have made.
These italicized expressions are neither objects nor predicative complements. Since their pres-
ence is obligatory, for syntactic well-formedness, they are called oblique complements. Roughly
speaking, oblique contrasts with the direct functions of subject and object, and oblique
phrases are typically expressed as PPs in English.
As we have seen before, most ditransitive verbs can also take oblique complements:
(32) a. John gave a book to the student.
b. John bought a book for the student.
The PPs here, which cannot be objects since they are not NPs, also do not serve as predicate of
the subject or object they relate directly to the verb, as oblique complements.
40
3.2.5 Modiers
The functions of DO, IO, predicative complement, and oblique complement all have one com-
mon property: they are all selected by the verb, and we view them as being present to comple-
ment the verb to form a legitimate VP. Hence, these are called complements (COMPS), and
typically they cannot be omitted.
Unlike these COMPS, there are expressions which do not complement the predicate in the
same way, and which are truly optional:
(33) a. The bus stopped suddenly.
b. Shakespeare wrote his plays a long time ago.
c. They went to the theater in London.
d. He failed chemistry because he cant understand it.
The italicized expressions here are all optional and function as modiers (also called adjuncts
or adverbial expressions). These modiers specify the manner, location, time, or reason,
among many other properties, of the situations expressed by the given sentences informally,
they are the (how, when, where, and why) phrases.
One additional characteristic of modiers is that they can be stacked up, whereas comple-
ments cannot.
(34) a. *John gave Tom [a book] [a record].
b. I saw this lm [several times] [last year] [during the summer].
As shown here, temporal adjuncts like several times and last year can be repeated, whereas the
two complements a book and a record in (34a) cannot. Of course, temporal adjuncts do not
become the subject of a passive sentence, suggesting that they cannot serve as objects.
(35) a. My uncle visited today.
b. *Today was visited by my uncle.
3.3 Form and Function Together
We now can analyse each sentence in terms of grammatical functions as well as the structural
constituents. Let us see how we can analyze a simple sentence along these two dimensions:
(36)
S
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP: SUBJ
VP: PRED
N V CP: OBJ
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
John believes C S
.
.
.
.
.
.
.
.
.
.
.
.
VP: PRED
VP: PRED
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Det A N VP: PRED
NP: MOD
last night
devoured a mouse
Here the expression last night is an adverbial NP in the sense that it is categorically an NP but
functions as a modier (adjunct) to the VP.
As we go through this book, we will see that the distinction between grammatical functions
and categorical types is crucial in the understanding of English syntax.
3.4 Semantic Roles
As noted before, semantic roles were introduced as a way of classifying the arguments of pred-
icators (mostly verbs and adjectives) into a closed set of participant types. Even though we
43
cannot make any absolute generalizations about the relationship between grammatical func-
tions and semantic roles, the properties of semantic roles do interact in regular ways with cer-
tain grammatical constructions. A list of the most relevant thematic roles and their associated
properties is given below.
32
Agent: A participant which the meaning of the verb species as doing or causing some-
thing, possibly intentionally. Examples: subject of eat, kick, hit, hammer, etc.
(44) a. John ate his noodle quietly.
b. A boy hit the ball.
c. A smith hammered the metal.
Patient: A participant which the verb characterizes as having something happen to it, and
as being affected by what happens to it. Examples: object of kick, hit, hammer, etc.
33
(45) a. A boy hit the ball.
b. A smith hammered the metal.
Experiencer: A participant who is characterized as aware of something. Examples: sub-
ject of perception verbs like feel, smell, hear, see, etc.
(46) a. The students felt comfortable in the class.
b. The student heard a strange sound.
Theme: A participant which is characterized as changing its position or condition, or
as being in a state or position. Examples: direct object of give, hand, subject of come,
happen, die, etc.
(47) a. John gave a book to the students.
b. John died last night.
Benefactive: The entity that benets from the action or event denoted by the predicator.
Examples: oblique complement of make, buy, etc.
(48) a. John made a doll for his son.
b. John bought a lot of books for his sons.
Source: The one from which motion proceeds. Examples: subject of promise, object of
deprive, free, cure, etc.
(49) a. John promised Bill to leave tomorrow morning.
b. John deprived his sons of game cards.
Goal: The one to which motion proceeds. Examples: subject of receive, buy, indirect
object of tell, give, etc.
(50) a. Mary received an award from the department.
b. John told the rumor to his friend.
32
The denition of semantic roles given here is adopted from Dowty (1989).
33
Patient and theme are often unied into undergoer in the sense that both the patient and theme individual undergo
the action in question.
44
Location: The thematic role associated with the NP expressing the location in a sentence
with a verb of location. Examples: subject of keep, own, retain, locative PPs, etc.
(51) a. John put his books in the attic.
b. The government kept all the money.
Instrument: The medium by which the action or event denoted by the predicator is carried
out. Examples: oblique complement of hit, wipe, hammer, etc.
(52) a. John hit the ball with a bat.
b. John wiped the window with a towel.
An important advantage of having such semantic roles available to us is that it allows us to
capture the relationship between two related sentences, as we have already seen. As another
example, consider the following pair:
(53) a. [
agt
The cat] chased [
pat
the mouse].
b. [
pat
The mouse] was chased by [
agt
the cat].
Even though the above two sentences have different syntactic structures, they have essentially
identical interpretations. The reason is that the same semantic roles assigned to the NPs: in both
examples, the cat is the agent, and the mouse is the patient. Different grammatical uses of verbs
may express the same semantic roles in different arrays.
The semantic roles also allow us to classify verbs into more ne-grained groups. For exam-
ple, consider the following examples:
(54) a. There still remains an issue to be solved.
b. There lived a man with his grandson.
c. At the same time there arrived a lone guest, a tall, red-haired and incredibly well
dressed man.
(55) a. *There sang a man with a pipe.
b. *There dances a man with an umbrella.
All the verbs are intransitive, but not all are acceptable in the there-construction. The dif-
ference can come from the semantic role of the postverbal NP, as assigned by the main verb.
Verbs like arrive, remain, live are taken to assign the semantic role of theme (see the list of
roles above), whereas verbs like sing, dance assign an agent role. We thus can conjecture that
there-constructions do not accept the verb whose subject carries an agent semantic role.
While semantic roles provide very useful ways of describing properties across different con-
structions, we should point out that the theoretical status of semantic roles is still unresolved.
34
For example, there is no agreement about exactly which and how many semantic roles are
needed. The problem is illustrated by the following simple examples:
(56) a. John resembles his mother.
b. A is similar to B.
34
See Levin and Rappaport Hovav (2005) for further discussion of this issue.
45
What kind of semantic roles do the arguments here have? Both participants seem to be playing
the same role in these examples they both cannot be either agent or patient or theme. They are
also cases where we might not be able to pin down the exact semantic role:
(57) a. John runs into the house.
b. Mary looked at the sky.
The subject John in (57a) is both agent and theme: it is agent since it initiates and sustains the
movement but also theme since it is the object that moves.
35
Also, the subject Mary in (57b)
can either be an experiencer or an agent depending on her intention one can just look at the
sky with no purpose at all.
36
Even though there are theoretical issues involved in adopting semantic roles in the grammar,
there are also many advantages of using them. We can make generalizations about the grammar
of the language: typically the agent takes the subject position, while an NP following the
word from is serving as the source. As we will see in the next chapter, semantic roles are
also recognized as the standard concepts used for organizing predicate-argument structures for
predicates within the lexicon. In the subsequent chapters, we will refer to semantic roles in
various places.
3.5 Exercises
1. Draw tree structures for the following sentences and then assign the appropriate gram-
matical function to each phrase.
(i) a. They parted the best of friends.
b. You will need comprehensive travel insurance.
c. In the summer we always go to France.
d. Benny worked in a shoe factory when he was a student.
e. Last year I saw this lm several times.
f. He baked Tom the bread last night.
g. That they have completed the course is amazing.
h. This poem was written by my uncle.
i. Everyone hoped that she would sing.
j. The gang robbed her of her necklace.
k. She passed through sheer hard work.
l. They helped us edit the script.
35
Jackendoff (1987) develops an account of thematic roles in which agency and motion are two separate dimensions,
so, in fact, a single NP can be agent and theme.
36
To overcome the problem of assigning a right semantic role to an argument, one can assume that each predicator
has its own (individual) semantic roles. For example, the verb kick, instead of having an agent and a patient, has two
individualized semantic roles kicker and kicked. See Pollard and Sag (1987).
46
m. We reminded him of the agreement.
2. Consider the following examples:
(i) a. There is/*are only one chemical substance involved in nerve transmission.
b. There *is/are more chemical substances involved in nerve transmission.
With respect to the grammatical function of there, what can we infer from these data?
Try out more subjecthood tests, to determine the grammatical function of there in these
examples. In addition, think of the following locative inversion examples and decide
the subject of the sentence.
(ii) a. In the garden stands/*stand a statue.
b. Among the guests was/*were sitting my friend.
3. Construct sentences containing the following grammatical functions:
(i) a. subject, predicator, direct object
b. subject, predicator, indirect object, direct object
c. subject, predicator, adjunct
d. adjunct, subject, predicator
e. adjunct, subject, predicator, direct object
f. subject, predicator, direct object, oblique complement
g. subject, predicator, predicative complement
h. subject, predicator, direct object, predicative complement
i. subject, predicator, predicative complement, adjunct
j. subject, predicator, direct object, predicative complement, adjunct
4. Give the grammatical function of the italicized phrases in the following examples:
(i) a. All of his conversation was reported to me.
b. Sandy removed her ballet shoes.
c. The school awarded a few of the girls in Miss Kims class scholarships.
d. She was the nicest teacher in the Senior School.
e. They elected him Americas 31st President.
f. The next morning we set out for Seoul.
g. Doing syntax is not easy.
h. Why dont you promise to see him later?
i. This is the place to go to.
j. He saw the man with the stick.
k. This week will be a difcult one for us.
l. We need to nish this project this week.
5. Decide the grammatical function of the underlined phrase, providing at least one syntac-
tic test we have discussed in the chapter.
(i) a. This proved a decisive factor.
b. This proved my hypothesis.
47
c. He felt a fool standing there alone on the platform.
d. He felt a fool pushing in front of me on the platform.
e. The students all enjoyed that summer.
f. The students all worked that summer.
g. The scientist made her a robot.
h. The students called me a teacher.
6. Assign a semantic role to each argument in the following sentences.
(i) a. A big green insect ew into the soup.
b. Johns mother sent a letter to Mary.
c. John smelled the freshly baked bread.
d. We placed the cheese in the refrigerator.
e. Frank threw himself into the sofa.
f. The crocodile devoured the doughnut.
g. John came from Seoul.
h. John is afraid of Bill.
i. The ice melted.
j. The vacuum cleaner terries the child.
7. After reading the following texts, circle the agent, underline the patient, and box the
theme expression.
(i) Scientists found that the birds sang well in the evenings, but performed
badly in the mornings. After being awake several hours, however, the
young males regained their mastery of the material and then improved on
the previous days accomplishments. To see whether this dip in learning
was caused by the same kind of pre-coffee fog that many people feel in
the morning, the researchers prevented the birds from practicing rst thing
in the morning. They also tried keeping the birds from singing during the
day, and they used a chemical called melatonin to make the birds nap at
odd times. The researchers concluded that their study supports the idea
that sleep helps birds learn. Studies of other animals have also suggested
that sleep improves learning.
37
37
From Science News Online, Feb 2, 2007
48
4
Head, Complements, and Modiers
4.1 Projections from Lexical Heads to Phrases
4.1.1 Internal vs. External Syntax
As we have seen in the previous chapters, both syntactic categories (NP, AP, VP, PP, etc.) and
grammatical functions (subject, complement, and modier) play important roles in the analysis
of English sentences. We have also observed that the grammatical function and form of each
constituent depend on where it occurs or what it combines with.
The combinatory properties of words and phrases involve two aspects of syntax: internal
and external syntax.
38
Internal syntax deals with how a given phrase itself is constructed in a
well-formed manner whereas external syntax is concerned with how a phrase can be used in a
larger construction. Observe the following examples:
(1) a. *John [put his gold].
b. *John [put under the bathtub].
c. *John [put his gold safe].
d. *John [put his gold to be under the bathtub].
e. John [put his gold under the bathtub].
Why is only (1e) acceptable? Simply, because only it satises the condition that the verb put
selects an NP and a PP as its complements, and it combines with them in the VP. In the other
examples, this condition is not fullled. This combinatory requirement starts from the internal
(or lexical) properties of the verb put, and is not related to any external properties of the VP.
In contrast, the external syntax is concerned with the external environment in which a phrase
occurs. Some of the unacceptable examples in (1) can be legitimate expressions if they occur in
the proper (syntactic) context.
(2) a. This is the box in which John [put his gold]. (cf. (1a))
b. This is the gold that John [put under the bathtub]. (cf. (1b))
38
The terms internal and external syntax are from Baker (1995).
49
Meanwhile, the well-formed VP in (1e) can be unacceptable, depending on external contexts.
For example, consider frame induced by the governing verb kept in (3):
(3) a. *The king kept [put his gold under the bathtub].
b. The king kept [putting his gold under the bathtub].
The VP put his gold under the bathtub is a well-formed phrase, but cannot occur in (3a) since
this is not the environment where such a nite VP occurs. That is, the verb kept requires the
presence of a gerundive VP like putting his gold under the bathtub, and therefore imposes an
external constraint on VPs.
4.1.2 Notion of Head, Complements, and Modiers
One important property we observe in English phrase-internal syntax is that in building up any
phrase, there is one obligatory element in each phrase. That is, each phrase has one essential
element as represented in the diagrams in (4):
(4) a.
NP
... N
7654 0123
b.
VP
V
7654 0123
...
c.
AP
A
7654 0123
...
The circled element here is the essential, obligatory element within the given phrase. We call
this essential element the head of the phrase.
39
The head of each phrase thus determines its
projection into a larger phrasal constituent. The head of an NP is thus N, the head of a VP is
V, and the head of an AP is A.
The notion of headedness plays an important role in the grammar. For example, the verb
put, functioning as the head of a VP, dictates what it must combine with two complements,
NP and PP. Consider other examples:
(5) a. The defendant denied the accusation.
b. *The defendant denied.
(6) a. The teacher handed the student a book.
b. *The teacher handed the student.
The verb denied here requires an NP object whereas handed requires two NP complements, in
this use. The properties of the head verb itself determine what kind of elements it will combine
with. As noted in the previous chapter, the elements which a head verb should combine with
are called complements. The complements include direct object, indirect object, predicative
complement, and oblique complement since these are all potentially required by some verb or
other.
The properties of the head become properties of the whole phrase. Why are the examples in
(7b) and (8b) ungrammatical?
39
See section 1.3 in Chapter 1 also.
50
(7) a. They [want to leave the meeting].
b. *They [eager to leave the meeting].
(8) a. The senators [know that the president is telling a lie].
b. *The senators [certain that the president is telling a lie].
The examples in (7b) and (8b) are unacceptable because of the absence of the required head.
The unacceptable examples lack a nite (tensed) VP as the bracketed part, but we know that
English sentences require a nite VP as their immediate constituent, as informally represented
as in (9):
(9) English Declarative Sentence Rule:
Each declarative sentence must contain a nite VP as its head.
Each nite VP is headed by a nite verb. If we amend the ungrammatical examples above to
include a verb but not a nite one, they are still ungrammatical:
(10) a. *They [(to) be eager to leave the meeting].
b. *The senators [(to) be certain that the president is telling a lie].
The VP is considered to be the (immediate) head of the sentence, with the verb itself as the head
of the VP. In this way, we can talk about a nite or non-nite sentence, one which is ultimately
headed by a nite or nonnite verb, respectively.
40
In addition to the complements of a head, a phrase may also contain modiers:
(11) a. Tom [
VP
[
VP
offered advice to his students] in his ofce].
b. Tom [
VP
[
VP
offered advice to his students] with love].
The PPs in his ofce or with love here provide further information about the action described
by the verb, but are not required as such by the verb. These phrases are optional and func-
tion as modiers, and they function to augment the minimal phrase projected from the head
verb offered. The VP which includes this kind of modier forms a maximal phrase. We might
say that the inner VP here forms a minimal VP which includes all the minimally required
complements, and the outer VP is the maximal VP which includes optional modiers.
What we have seen can be summarized as follows:
(12) a. Head: A lexical or phrasal element that is essential in forming a phrase.
b. Complement: A phrasal element that a head must combine with or a head select.
These include direct object, indirect object, predicative complement, and oblique
complement.
c. Modier: A phrasal element not selected by the verb functions as a modier to
the head phrase.
d. Minimal Phrase: A minimal phrase is the phrase including this head and all of
its complements.
40
See section ?? of Chapter 5 for the detailed discussion of English verb form (VFORM) values including nite and
nonnite.
51
e. Maximal Phrase: A maximal phrase is the phrase that includes complements as
well as modiers.
4.2 Differences between Complements and Modiers
Given these notions of complements and modiers, the question that then follows is how we can
distinguish between complements and modiers. There are several tests to determine whether a
phrase is a complement or a modier.
41
Obligatoriness: As hinted at already, complements are strictly-required phrases whereas mod-
iers are not. The examples in (13)(15) show that the verb placed requires an NP and a PP as
its complements, kept an NP and a PP or an AP, and stayed a PP.
(13) a. John placed Kim behind the garage.
b. John kept him behind the garage.
c. *John stayed Kim behind the garage.
(14) a. *John placed him busy.
b. John kept him busy.
c. *John stayed him busy.
(15) a. *John placed behind the counter.
b. *John kept behind the counter.
c. John stayed behind the counter.
In contrast, modiers are optional. Their presence is not required by the grammar:
(16) a. John deposited some money in the bank.
b. John deposited some money in the bank on Friday.
In (16b), the PP on Friday is optional here, serving as a modier.
Iterability: The possibility of iterating identical types of phrase can also distinguish between
complements and modiers. In general two or more instances of the same modier type can
occur with the same head, but this is impossible for complements.
(17) a. *The UN blamed global warming [on humans] [on natural causes].
b. Kim and Sandy met [in Seoul] [in the lobby of the Lotte Hotel] in March.
In (17a) on humans is a complement and thus the same type of PP on natural causes cannot
co-occur. Yet in Seoul is a modier and we can repeatedly have the same type of PP.
41
Most of the criteria we discuss here are adopted from Pollard and Sag (1987).
52
Do-so Test: Another reliable test often used to distinguish complements frommodiers is the do
so or do the same thing test. As shown in (18), we can use do the same thing to avoid repetition
of an identical VP expression:
(18) a. John deposited some money in the checking account and Mary did the same
thing (too).
b. John deposited some money in the checking account on Friday and Mary did the
same thing (too).
What we can observe in (18b) is that the VP did the same thing can replace either the minimal
phrase deposited some money in the checking account or the maximal phrase including the
modier on Friday. Notice that this VP can replace only the minimal phrase, leaving out the
modier.
(19) John deposited some money in the checking account on Friday and Mary did the
same thing on Monday.
From these observations, we can draw the conclusion that if something can be replaced by do
the same thing, then it is either a minimal or a maximal phrase. This in turn means that this
replacement VP cannot be understood to leave out any complement(s). This can be veried
with more data:
(20) a. *John [deposited some money in the checking account] and Mary did the same
thing in the savings account.
b. *John [gave a present to the student] and Mary did the same thing to the teacher.
Here the PPs in the checking account and to the student are both complements, and thus they
should be included in the do the same thing phrase. This gives us the following informal gener-
alization:
(21) Do-so Replacement Condition:
The phrase do so or do the same thing can replace a verb phrase which includes
at least any complements of the verb.
This condition explains why we cannot have another locative complement phrase in the
savings account or to the teacher in (20). The unacceptability of the examples in (22) also
supports this generalization about English grammar:
(22) a. *John locked Fido in the garage and Mary did so in the room.
b. *John ate a carrot and Mary did so a radish.
Constancy of semantic contribution: An adjunct can cooccur with a relatively broad range of
heads whereas a complement is typically limited in its distribution. Note the following contrast:
(23) a. Kim camps/jogs/mediates on the hill.
53
b. Kim jogs on the hill/under the hill/over the hill.
(24) a. Kim depends/relies on Sandy.
b. Kim depends on Sandy/*at Sandy/*for Sandy.
The semantic contribution of the adjunct on the hill in (23a) is independent of the head whereas
that of the complement on Sandy is idiosyncratically dependent upon the head.
Structural Difference: We could distinguish complements and modiers by tree structures,
too: complements combine with a lexical head (not a phrase) to form a minimal phrase whereas
modiers combine with a phrase to form a maximal phrase. This means that we have structures
of the following forms:
(25)
XP
XP
Modier
X Complement(s)
As represented in the tree structures, complements are sisters of the lexical head X, whereas
modiers are sisters of a phrasal head. This structural difference between complements and
modiers provides a clean explanation for the contrast in do-so test. Given that the verb ate
takes only an NP complement whereas put takes an NP and a PP complement, we will have the
difference in the two structures shown in (26):
(26) a.
VP
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
VP
.
.
.
.
.
.
.
PP
V NP
in the room
ate some food
b.
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
V NP
PP
N
c. VP V (NP) (VP)
d. VP V NP AP
e. VP V NP NP
f. VP V S
g. AP A VP
h. PP P NP
i. VP Adv VP
One common property of all these rules is, as we have discussed, that every phrase has its own
head. In this sense, each phrase is the projection of a head, and thereby has the endocentricity.
However, we can ask the theoretical question of whether or not we can have rules like the
following, in which the phrase has no head at all:
(30) a. VP P NP
42
See (49) for the structural differences between with long hair and of linguistics.
43
The discussion of this section is based on Sag et al. (2003).
55
b. NP PP S
Nothing in the grammar makes such PS rules unusual, or different in any way from the set in
(29). Yet, if we allow such non-endocentric PS rules in which a phrase does not have a lexical
head, the grammar would then be too powerful to generate only the grammatical sentences of
the language.
Another limit that we can nd from the simple PS rules concerns an issue of redundancy.
Observe the following:
(31) a. *The problem disappeared the accusation.
b. The problem disappeared.
(32) a. *The defendant denied.
b. The defendant denied the accusation.
(33) a. *The boy gave the book.
b. The boy gave the baby the book.
What these examples show is that each verb has its own requirement for its complement(s).
For example, deny requires an NP, whereas disappear does not, and gave requires two NPs as
its complements. The different patterns of complementation are said to dene different subcate-
gories of the type verb. The specic pattern of complements is known as the subcategorization
requirement of each verb, which can be represented as following (IV: intransitive, TV: transitive,
DTV: ditransitive):
(34) a. disappear: IV,
b. deny: TV, NP
c. give: DTV, NP NP
In addition, in order to license the grammatical sentences in (31)(33), we need to have the
following three VP rules:
(35) a. VP IV
b. VP TV NP
c. VP DTV NP NP
We can see here that in each VP rule, only the appropriate verb can occur. That is, a DTV cannot
form a VP with the rules in (35a) or (35b): It forms a VP only according to the last PS rule.
Each VP rule thus also needs to specify the kind of verb that can serve as its head.
Taking these all together, we see that a grammar of the type just suggested must redundantly
encode the subcategorization information both in the lexical type of each verb (e.g., DTV) and
in the PS rule for that type of verb.
A similar issue of redundancy arises in accounting for subject-verb agreement:
(36) a. The bird devours the worm.
b. The birds devour the worm.
56
To capture the fact that the subject NP agrees with the predicate VP, we need to differentiate the
S rule into the following two:
(37) a. S NP
singular
VP
singular
(for (36)a)
b. S NP
plural
VP
plural
(for (36)b)
Descriptively, there is no problem with a grammar with many specic parts. From a theoretical
perspective, though, we have a concern about the the endocentricity and redundancy issues. A
more particular related question is that of how many PS rules English has. For example, how
many PS rules do we need to characterize English VPs?Presumably there are as many rules
as there are subcategories of verb.
We need to investigate the abstract content of PS rules, in order to develop a theoretical view
of them. For example, it seems to be the case that each PS rule must have a head. This will
disallow many possible PS rules which we can write using the rule format, from being actual
rules of any language.
In order to understand more about the structures that rules describe, we need two more no-
tions, intermediate category and specier. We motivate the idea of the intermediate category,
and then specier is a counterpart of it. Consider the examples in (38):
(38) a. Every photo of Max and sketch by his students appeared in the magazine.
b. No photo of Max and sketch by his students appeared in the magazine.
What are the structures of these two sentences? Do the phrases every photo of Max and sketch
by his students form NPs? It is not difcult to see sketch by his students is not a full NP by itself,
for if it was, it should be able to appear as subject by itself:
(39) *Sketch by his students appeared in the magazine.
In terms of the semantic units, we can assign the following structures to the above sentences, in
which every and no operate over the meaning of the rest of the phrase:
(40) a. [Every [[photo of Max] and [sketch by his students]]] appeared in the magazine.
b. [No [[photo of Max] and [sketch by his students]]] appeared in the magazine.
The expression photo of Max and sketch by his students are phrasal elements but not full NPs
so what are they? We call these intermediate phrases, notationally represented as N-bar or
N
. The phrase N
is thus intuitively bigger than a noun, but smaller than a full NP, in the sense
that it still requires a determiner from the class the, every, no, some, and the like.
The complementary notion that we introduce at this point is specier (SPR), which can
include the words just mentioned as well as phrases, as we illustrate in (41):
(41) a. [the enemys] [
N
destruction of the city]
b. [The enemy] [
VP
destroyed the city].
The phrase the enemys in (41a) and the subject the enemy in (41b) are semantically similar
in the sense that they complete the specication of the event denoted by the predicate. These
phrases are treated as the speciers of N
is a DP.
44
Now let us compare the syntactic structures of (41a) and (41b):
(43)
NP
DP
the enemys N PP
NP
VP
The enemy V
NP
which in turn combines with the specier DP the enemys. In the S structure, the head
V combines with its complement the city and forms a VP. This resulting VP then combines with
44
Some analyses take the whole expression in (43) to be a DP (e.g., a little dog, my little dogs) in which expressions
like little dog is not an N
but an NP.
58
the subject the enemy, which is also a specier. In a sense, the VP is an intermediate phrase that
requires a subject in order to be a full and complete S.
Given these similarities between NP and S structures, we can generalize over them as in (45),
where X is a variable over categories such as N, V, P, and other grammatical categories:
45
(45)
XP
Specier X
X Complement(s)
This structure in turn means the grammar now includes the following two rules:
46
(46) a. XP Specier, X
(Head-Specier Rule)
b. XP X, YP
(Head-Complement Rule)
These Head-Specier and Head-Complement Rules, which form the central part of X
-theory,
account for the core structure of NP as well as that of S. In fact, these two general rules can also
represent most of the PS rules we have seen so far. In addition to these two, we just need one
more rule:
47
(47) XP Modier, X
(Head-Modier Rule)
This Head-Modier Rule allows a modier to combine with its head as in the PS rule VP
VP Adv/PP, as presented in the following.
48
(48)
XP
Specier X
Modier
X Complement(s)
One thing to notice in the Head-Complement Rule is that the head must be a lexical el-
ement. This in turn means that we cannot apply the Head-Modier Rule rst and then the
45
We can assume that the head of S is VP and that VP is an intermediate phrase in the sense that it still requires a
subject as its specier.
46
Unlike the PS rules we have seen so far, the rules here are further abstracted, indicated by the comma notation
between daughters on the right-hand side. We assume that the relative linear order of a head and complements etc. is
determined by a combination of general and language-specic ordering principles, while the hierarchical X
-structures
themselves are universal.
47
The comma indicates that the modier can appear either before the head or after the head.
48
The modier can also precede the head X
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
the
N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
PP
N
PP
with a hat
king of Rock and Roll
b.
NP
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
DP *N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
the
N
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
PP
N
PP
but not an N or an NP. This will also account for the following contrast, too:
(52) A: Which student were you talking about?
B: The one with long hair.
60
B: *The one of linguistics with long hair.
The phrase of linguistics is the complement of student. This means the N-bar pronoun one
should include this. However, the modier with long hair cannot be within the N
.
There exist several more welcoming consequences that the three X
rules directly solve the endocentricity issue, for they refer to Head.
Assume that X is N, then we will have N, N
Specier X
[POS 1 ]
[POS 1 ]
Modier
X[POS 1 ] Complement(s)
The notation 1 shows that whatever value the feature has in one place in the structure, it has
the same value somewhere else. This is a representational tag, in which the number 1 has no
signicance: it could as easily be 7 or 437 . We provide more details of the formal feature system
in the following section.
So (53) indicates that the phrases POS value is identical to its head daughter, capturing
the headedness of each phrase: the grammar just does not allow any phrase without a head.
The redundancy issue mentioned above for agreement is now a matter of introducing another
feature, NUMBER. That is, with the new feature NUMBER, with values singular and plural,
we can add a detail to the Head-Specier Rule as following:
(54) XP Specier[NUMBER 1 ], X
[NUMBER 1 ]
The rule states that the subjects NUMBER value is identical with that of the predicate VPs
NUMBER value. The two rules in (37) are both represented in (54).
4.4 Lexicon and Feature Structures
In the previous section, we have seen that the properties of a lexical head determine the compo-
nents of the minimal phrase, in terms of complements, and that other properties of the head are
61
directly properties of the phrase. This information is encoded in a lexical entry, for each word
in the lexicon.
Every lexical entry at least includes phonological (but in practice, orthographic), morpho-
logical, syntactic, and semantic information. For example, the word puts will have at least the
following information:
(55) Minimal Lexical Information for puts:
a. phonological information: <puts>
b. syntactic information: verb, nite, 3rd singular
c. argument information: <agent
i
, theme
j
, location
k
>
d. semantic information: put
(i,j,k)
The phonological information is the information about how the word is pronounced; the syn-
tactic information indicates that this particular word is a verb and is in the 3rd singular present
(nite) form. The argument structure represents the number of arguments which the verb se-
lects, to indicate the participants that are minimally involved in the event expressed by the verb.
This argument information is linked to its more precise meaning as indicated by the indexes i,
j and k. These indexes refer to the participants denoted by the arguments. Finally, the semantic
structure represents that the verbs meaning relates three participants someone i who is doing
the action of putting, something j being put in a place, and someplace k it is put in. These lexi-
cal entries can be represented in a more systematic and precise way with the system of feature
structures, which we now introduce.
4.4.1 Feature Structures and Basic Operations
Most modern grammars rely on a representation of lexical information in terms of features
and their values.
49
We present here a formal and explicit way of representing it with feature
structures. Each feature structure is an attribute-value matrix (AVM):
(56)
_
_
Attribute1 value1
Attribute2 value2
Attribute3 value3
. . . . . .
_
_
The value of each attribute can be an atomic element, a list, a set, or a feature structure:
(57)
_
_
Attribute1 atomic
Attribute2
Attribute3
_ _
Attribute4
_
. . .
_
_
_
49
In particular, grammars such as Head-driven Phrase Structure Grammar (HPSG) and Lexical Functional Grammar
(LFG) are couched upon mathematically well-dened feature structure systems. The theory developed in this textbook
heavily relies upon the feature structure system of HPSG. See Sag et al. (2003).
62
One important property of every feature structure is that it is typed.
50
That is, each feature
structure is relevant only for a given type. A simple illustration can show why feature structure
needs to be typed. The upper left declaration in italics is the type of the feature structure:
(58) a.
_
_
university
NAME kyunghee univ.
LOCATION seoul
_
_
b.
*
_
_
university
NAME kyunghee univ.
MAJOR linguistics
_
_
The type university may have many properties, including its name and location, but having a
MAJOR, as a subject of study, is inappropriate. In the linguistic realm, we might declare that
TENSE is appropriate only for verb, for example.
Now consider the following example of a typed feature structure, information about one
author of this book:
(59)
_
_
author
NAME kim
SONS Edward, Richard
HOBBIES
_
swimming, jogging, reading, . . .
_
ADVANCED-DEGREE
_
_
FIELD linguistics
AREA syntax-semantics
YEAR 1996
_
_
_
_
This illustrates the different types of values that attributes (feature names) may have. Here, the
value of the attribute NAME is atomic, whereas the value of SONS is a list which represents
something relative about the two values, in this case that one is older than the other. So, for
example youngest son would be the right-most element in the list value of SONS. Meanwhile,
the value of HOBBIES is a set, showing that there is no signicance in the relative ordering.
Finally, the value of the feature ADVANCED-DEGREE is a feature structure which in turn has
three attributes.
One useful notion in the feature structure is structure-sharing, which we have already seen
above in terms of the 1 notation (see (??)). This is to represent cases where two features (or
attributes) have an identical value:
50
Even though every feature structure is typed in the present grammar, we will not specify the type of each feature
structure unless it is necessary for the discussion.
63
(60)
_
_
individual
NAME kim
TEL 1
SONS
_
_
_
individual
NAME richard
TEL 1
_
_
,
_
_
individual
NAME edward
TEL 1
_
_
_
_
_
For the type individual, attributes such as NAME and TEL and SONS are appropriate. (60)
represents a situation in which the particular individual kimhas two sons, and their TEL attribute
has a value which is the same as the value of his TEL attribute, whatever the value actually is.
In addition to this, the notion of subsumption is also important in the theoretical use of
feature structures; the symbol represents subsumption. The subsumption relation concerns
the relationship between a feature structure with general information and one with more specic
information. In such a case, the general one subsumes the specic one. Put differently, a feature
structure A subsumes another feature structure B if A is not more informative than B.
(61)
A:
_
individual
NAME kim
_
B:
_
_
individual
NAME kim
TEL 961-0892
_
_
In (62), A represents more general information than B. This kind of subsumption relation is
used to represent partial information, for in fact we cannot represent the total information
describing all possible worlds or states of affairs. In describing a given phenomenon, it will be
more than enough just to represent the particular or general aspects of the facts concerned. Each
small component of feature structure will provide partial information, and as the structure is
built up, the different pieces of information are put together.
The most crucial operation in feature structures is unication, represented by
. Feature
unication means that two compatible feature structures are unied, conveying more coherent
and rich information. Consider the feature structures in (62); the rst two may unify to give the
third:
(62)
_
individual
NAME kim
_
_
individual
TEL 961-0892
_
_
_
individual
NAME kim
TEL 961-0892
_
_
The two feature structures are unied, resulting in a feature structure with both NAME and TEL
information. However, if two feature structures have incompatible feature values, they cannot
be unied:
(63)
_
individual
NAME edward
_
_
individual
NAME richard
_
64
*
_
_
individual
NAME edward
NAME richard
_
_
Since the two smaller feature structures have different NAME values, they cannot be unied.
Unication will make sure that information is consistent as it is built up in the analysis of a
phrase or sentence.
4.4.2 Feature Structures for Linguistic Entities
Any individual or entity including a linguistic expression can be represented by a feature struc-
ture. For example, the word puts, whose general type is verb, can have a feature structure like
the following:
51
(64)
_
_
verb
PHON puts
SYN
_
_
POS verb
VFORM n
_
_
ARG-ST
_
[agt]
i
, [th]
j
, [loc]
k
_
SEM
_
_
PRED put-relation
AGENT i
THEME j
LOCATION k
_
_
_
_
This feature structure has roughly the same information as the informal representation in (55).
The verb puts, like any verb, has its own phonological (PHON) value, syntactic (SYN), ar-
gument structure (ARG-ST), and semantic (SEM) information. The SYN attribute indicates
that the POS (part of speech) value is verb, that it has a nite verbal inectional form value
(VFORM).The ARG-ST attribute indicates that the verb selects for three arguments (with the-
matic roles agent, theme, location), which will be realized as the subject and two complements
in the full analysis. The SEM feature represents the information this verb denotes the predicate
relation put-relation, whose three participants are linked to the elements in the ARG-ST via the
indexing values i, j, and k.
One thing to note here is that since there are some cases where we have difculties in assign-
ing a specic named semantic role to a selected argument discussed in Chapter 3, we typically
just indicate the number of arguments each predicate is selecting in ARG-ST: we underspecify
the information unless it is necessary to show more details. So, for example, verbs like smile,
devour and give will have the following ARG-ST representations, respectively:
(65) a.
_
ARG-ST [ ]
_
51
Later on, we will not represent the PHON and SEM values unless relevant to the discussion at hand.
65
b.
_
ARG-ST [ ], [ ]
_
c.
_
ARG-ST [ ], [ ], [ ]
_
One-place predicates like smile select for just one argument, two-place predicates like devour
take two arguments, and three-place predicates take three arguments. Eventually, the arguments
selected by each predicate are linked to grammatical functions, to the core semantic properties,
and to other parts of the representation of the grammatical properties.
4.4.3 Argument Realization
Each element on the ARG-ST list is realized as SPR (specier) or COMPS (complements),
through one of the rules in (46).
52
In general, the basic pattern is that the rst element on the list
is realized as subject and the rest as complements:
(66) Argument Realization Constraint (ARC):
The rst element on the ARG-ST list is realized as SPR, the rest as COMPS in
syntax.
This realization is obligatory in English; for example, the three arguments of put are realized as
subject and complements, with the putter (agent) as subject:
53
(67) a. John put the book in the box.
b. *John put in the box.
c. *In the box put John the book.
d. #The book put John in the box.
We see that the arguments selected by a lexical head should be all realized as SPR and COMPS,
which are combined in the notion of valence (VAL) features.
54
More formally, we can represent
this constraint as applied to put as the following feature structure:
(68)
_
_
VAL
_
_
SPR 1 NP
COMPS 2 NP, 3 PP
_
_
ARG-ST 1 , 2 , 3
_
_
The boxed tags show the different identities in the overall structure. For example, the rst el-
ement of ARG-ST and of SPR have the boxed tag 1 , ensuring that the two are identical. The
general ARC constraint blocks examples like (67c) in which the location argument is realized
as the subject, as shown in (69):
52
Once again, remember that the term SPR includes subject as well as the nouns specier.
53
The notation # indicates that the structure is technically well-formed from a syntactic perspective, but semantically
anomalous.
54
The term valence refers to the number of arguments that a lexical item can combine with, to make a syntactically
well-formed sentence, often along with a description of the categories of those constituents. It is inspired by the notion
of valence as used in atomic theory in chemistry.
66
(69) *
_
_
VAL
_
_
SPR 3 PP
COMPS 1 NP, 2 NP
_
_
ARG-ST 1 , 2 , 3
_
_
This violates the ARC, which requires the rst element of ARG-ST be realized as the SPR (the
subject of a verb or the specier of a noun).
Notice that the arguments can be realized into different categories, depending on the proper-
ties of the given verb:
(70) a. The election results surprised everybody.
b. That he won the election surprised everybody.
The data indicate that verbs like surprise will take two arguments, but the rst argument can be
realized either as an NP subject as in (70a) or a CP subject as in (70b). This difference in the
argument realization can be represented as the following, respectively:
(71) a.
_
_
VAL
_
_
SPR 1 NP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
b.
_
_
VAL
_
_
SPR 1 CP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
Though there is no difference in terms of the number of arguments that surprise select, the
arguments can be realized in a different phrase. As the book goes on, we will see how the
argument realization is further constrained by the lexical properties of the verb in question or
by other grammatical components.
4.4.4 Verb Types and Argument Structure
As mentioned earlier, lexical elements in the classes V, A, N, and P, select one or more comple-
ment(s) to form a minimal phrase. With the construct of ARG-ST, we know that every lexical
element has ARG-ST information which will be realized in surface form through the SPR and
COMPS values. Verb types can be differentiated by looking only at the COMPS value since
every verb will have one SPR (subject) element. This is exactly the way that verbs are differen-
tiated using the traditional notion of subcategorization.
Intransitive: This is a type of verb that does not have any COMPS:
(72) a. John disappeared.
b. *John disappeared Bill.
(73) a. John sneezed.
b. *John sneezed the money.
67
These verbs have no COMPS elementthe list is necessarily empty. Such a verb will have just
one argument that is realized as subject:
55
(74)
_
_
disappear
SPR 1 NP
COMPS
ARG-ST 1
_
_
Linking verbs: Verbs such as look, seem, remain, and feel require different complements that
are typically of category AP:
(75) a. The president looked [weary].
b. The teacher became [tired of the students].
c. The lasagna tasted [scrumptious].
d. John remained [somewhat calm].
e. The jury seemed [ready to leave].
These verbs also can select other phrases (here, NP):
(76) a. John became a success.
b. John seemed a fool.
c. John remained a student.
Though each verb may select different types of phrases, they all at least select a predicative
complement, where a property is ascribed to the subject. (Compare John remained a student
and John revived a student.) This subcategorization requirement can be represented as follows:
(77)
_
_
become
SPR 1 NP
COMPS 2 XP[PRD +]
ARG-ST 1 , 2
_
_
This kind of verb selects two arguments: one is canonically an NP to be realized as the subject
and the other is any phrase (XP) that can function as a predicate (PRD +) (see also the examples
in (85)). Of course, this presupposes an accurate characterization of which phrases can be [PRD
+], which we simply assume here.
Transitive verbs: Unlike linking verbs, pure transitive verbs select a referential, non-predicative
NP as their complement, functioning as direct object:
(78) a. John saw Fred.
b. Alice typed the letter.
c. Clinton supported the health care bill.
55
For convenience reason, we adopt a shorthand system in representing feature structures, suppressing unrelated
features. For example, the fully specied feature structure in (74) will include VAL as well as PHON, SYN, SEM, etc.
68
d. Raccoons destroyed the garden.
Such verbs will have the following lexical information:
(79)
_
_
destroy
SPR 1 NP
COMPS 2 NP
ARG-ST 1 , 2
_
_
Ditransitive: There are also ditransitive verbs that require IO and DO:
(80) a.
b. The school board leader asked the students a question.
c. The parents bought the children non-ction novels.
d. John taught new students English Syntax.
Such verbs have three arguments: one subject and two complement NPs functioning as indi-
rect and direct object, respectively:
(81)
_
_
teach
SPR 1 NP
COMPS 2 NP, 3 NP
ARG-ST 1 , 2 [goal], 3 [theme]
_
_
These verbs typically have a related realization with the goal argument is realized as a dative
PP complement:
(82) a. The school board leader asked a question of the students.
b. The parents bought non-ction novels for the children.
c. John taught English Syntax to new students.
direct object and an prepositional complement:
(83)
_
_
teach
SPR 1 NP
COMPS 2 NP, 3 PP
ARG-ST 1 , 2 [theme], 3 [goal]
_
_
In this realization, the second argument has the theme role while the third one has the goal role.
(84)
_
_
teach
SPR 1 NP
COMPS 3 NP, 2 NP
ARG-ST 1 , 2 [theme], 3 [goal]
_
_
The second argument, the theme, is realized as the nal complement in the COMPS whereas the
goal argument becomes the direct object that can be promoted to the subject of a passive verb.
69
This argument realization will project sentences like the following in which the goal is realized
as the indirect object:
Complex Transitive: There is another type of transitive verb which selects two complements,
one functioning as a direct object and the other as a predicative phrase (NP, AP, or VP), describ-
ing the object:
(85) a. John regards Bill as a good friend.
b. The sexual revolution makes some people uncomfortable.
c. Ad agencies call young people Generation X-ers.
d. Historians believe FDR to be our most effective president.
In (85a), the predicative PP as a good friend follows the object Bill; in (85b), the AP uncom-
fortable serves as a predicate phrase of the preceding object some people. In (85c), the NP
Generation X-ers is the predicative phrase. In (85d), the predicative phrase is an innitive VP.
Just like linking verbs, these verbs require a predicative ([PRD +]) XP as complement:
(86)
_
_
call
SPR 1 NP
COMPS 2 NP, 3 XP
ARG-ST 1 , 2 , 3 [PRD +]
_
_
This means that the verbs in (85) all select an object NP and an XP phrase that function as a
predicate.
Even though these ve types of verb that we have seen so far represent many English verb
types, there are other verbs that do not t into these classes; for instance, the use of the verb
carry in (87).
(87) a. *John carried to the door.
b. *John carried her.
c. John carried her on his back.
The examples in (87) illustrate that carried requires an NP and a PP, as represented in the feature
structure:
(88)
_
_
carry
SPR 1 NP
COMPS 2 NP, 3 PP
ARG-ST 1 [agt], 2 [th], 3 [loc]
_
_
The PP here cannot be said to be predicate of the object her; it denotes the location to which
John carries her.
Of course, there exist various other verb types that we have not described here, in terms of
complementation patterns. As the book goes on, we will see yet more different types of verbs.
70
4.5 Exercises
1. Provide tree structures for the following two sentences while checking the grammatical
function of each phrase with valid distributional tests. What differences do we need to
represent?
(i) a. Tom locked Fido in the garage.
b. Tom bathed Fido in the garage.
2. For each example below, draw its structure and give the lexical entry (SPR and COMPS
value) of the main verb.
(i) a. Tom hid the manuscript in the cupboard.
b. Fred hired Sharon to change the oil.
c. They pushed the prisoners into the truck.
d. Frank hopes to persuade Harry to make the cook wash the dishes.
e. George mailed the attorney his photograph of the accident.
f. Gordon tried to open the jar.
g. Tom keeps asking Karens sister to buy the car.
h. Jane left the book on the table.
i. We have not conrmed whether the ight had been booked.
j. We saw him beaten by the champion.
k. They conned his remarks to the matter under discussion.
3. The verbs in the following sentences are used incorrectly. Correct the errors or replace
the verb with another one. In addition, provide the COMPS value for each verb.
(i) a. *Oliver ascribed his longevity there.
b. *Oliver mentioned Charles the problem.
c. *Oliver ned ten pounds to the prisoner.
d. *Oliver drove me a lunatic.
e. *Oliver addressed the king the letter.
f. *Oliver absented his brother from the meeting.
4. Draw tree structures for the following two sentences. In particular, provide detailed NP
structures using the intermediate phrase N
:
(i) a. The love of my life and father of my children would never do such a thing.
b. The museum displayed no painting by Miro or drawing by Klee.
5. Consider the following English sentences and provide the lexical entries for wrote and
worded:
(i) a. Kim wrote the beginning paragraph carelessly.
b. Kim worded the beginning paragraph carelessly.
(ii) a. Kim wrote that letter.
b. *Kim worded that letter.
As we have seen in this chapter, do so can replace a minimal or maximal VP phrase.
71
Based on the difference between wrote and worded and the property of do so test, can
you see that in (iiia) so did can have two readings whereas in (iiib) it can have only one
reading? If you can, explain why?
(iii) a. Kim wrote the preface carelessly, and so did Lee.
b. Kim worded the preface carelessly, and so did Lee.
6. Read the following texts and provide the ARG-ST of the italicized words and their SPR
and COMPS values:
(i) Learning to use a language freely and fully is a lengthy and effortful pro-
cess. Teachers cannot learn the language for their students. They can set
their students on the road, helping them to develop condence in their own
learning powers. Then they must wait on the sidelines, ready to encourage
and assist.
(ii) Deep ecologists put a reign on human exploitation of natural resources
except to satisfy vital needs. Thus, the use of a eld by an African tribe to
growgrain for survival is an example of a vital need whereas the conversion
of a swamp to an exclusive golf course would not. Rest assured that much
of the mining, harvesting, and development of our technological age would
not meet the requirement of this principle. Rather than being concerned
about how to raise automobile production, this ethic would be interested
in solving the problem of human mobility in a way that would not require
the disruption of highways, roads, and parking lots. It rebels against an
industrialist world view: Before it is possessed and used, every plant is a
weed and every mineral is just another rock.
56
56
From http://www.unitedearth.com.au/deepecology.html
72
5
More on Subjects and Complements
5.1 Grammar Rules and Principles
As we have seen in the previous chapter, the arguments in ARG-ST are realized as the syntactic
elements SPR (subject of a verb and determiner of a noun) and COMPS. The X
rules control
their combination with a relevant head:
(1) a. XP Specier, Head
b. XP Head, Complement(s)
c. XP Head, Modier
The rule (1a) represents the case where a head combines with its specier (e.g., a VP with
its subject and an N
with its determiner), whereas (1b) says that a head combines with its
complement(s) and forms a phrase. (1c) allows a combination of a head with its modier. As
noted earlier, in order to guarantee that the heads POS (part of speech) value is identical with
its mother phrase, we need to introduce the category variable X and the feature POS:
(2) a. Head-Specier Rule:
XP
_
POS 1
_
Specier, X
[POS 1 ]
b. Head-Complement Rule:
XP
_
POS 1
_
X[POS 1 ], Complement(s)
c. Head-Modier Rule:
XP
_
POS 1
_
Modier, XP
_
POS 1
_
The POS feature is thus a head feature which passes up to a mother phrase from its head
daughter, as shown in (3):
(3)
VP[ POS 1 verb ]
V[ POS 1 verb ]
PP
73
This percolation from a head to its mother is ensured by the following Head Feature Principle:
(4) The Head Feature Principle (HFP):
A phrases head feature (e.g., POS, VFORM, etc.) is identical with that of its
head.
The HFP thus ensures that every phrase has its own lexical head with the identical POS value.
The HFP will apply to any features that we declare to be head features, VFORMbeing another.
The grammar does not allow hypothetical phrases like the following:
(5)
*VP[POS verb]
.
.
.
.
.
.
.
.
.
.
.
A[POS adj] PP
We have not yet spelled out clearly what ensures a lexical head to combine not just with one
of its complements but with all of its COMPS elements. Consider the following examples:
(6) a. Kim put the book in the box.
b. *Kim put the book.
c. *Kim put in the box.
As seen from the contrast here and as noted in the lexical entry in (7), the verb put selects two
complements and must combine with all of its complements.
(7)
_
_
HEAD| POS verb
SPR NP
COMPS NP, PP
_
_
We can also see that a nite verb must combine with its subject:
(8) a. *Is putting the book in the box.
b. *Talked with Bill about the exam.
Such combinatorial requirements can be formally stated in the revised grammar rules as given
in (9):
(9) a. Head-Specier Rule:
XP
_
SPR
_
1 , H[SPR 1 ]
b. Head-Complement Rule:
XP
_
COMPS
_
H[COMPS 1 ,. . . , n ], 1 ,. . . , n
c. Head-Modier Rule:
XP [MOD 1 ], 1 H
The grammar rules here are well-formedness conditions on possible phrases in English, indi-
cating what each head combines with and then what happens as the result of the combination.
74
For example, in (9a) when a head, requiring a SPR, combines with it, we have a well-formed
head-specier phrase with the SPR value discharged; in (9b), a head combines with all of its
COMPS value, it forms a Head-Complement phrase with no further COMPS value; in (9c),
when a modier combines with the head it modies, the resulting phrase forms a well-formed
head-modier phrase.
57
These three grammar rules, interacting with the general principles such as the HFP, license
grammatical sentences in English. Let us consider one example in a little more detail:
(10)
S
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Subj
H
1 NP
VP
_
_
HEAD 4 | POS verb
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
H
C
C
Kim
V
_
_
HEAD 4 | POS verb
SPR 1
COMPS 2 , 3
_
_
2 NP
3 PP
nonn
_
knows
HEAD
_
_
POS verb
VFORM pres
_
_
VAL
_
_
SPR
_
1 NP
_
COMPS
_
2 NP
_
_
_
ARG-ST
_
1 , 2
_
_
_
This [VFORM pres] value will be projected to the S in accordance with the HFP, as represented
in the following:
78
(24)
S
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
NP
VP
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
The student
V
_
VFORM pres
_
NP
_
keep
HEAD| POS verb
COMPS
_
VP[prp]
_
_
_
b.
_
_
make
HEAD| POS verb
COMPS
_
NP, VP[bse]
_
_
_
Such lexical specications on the VPs VFORM value will make sure that these verbs only
combine with a VP complement with the appropriate VFORM value. The following structure
represents one example:
79
(27)
S
_
_
HEAD 4
_
_
POS verb
VFORM past
_
_
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD4
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
N
V
_
_
HEAD 4
SPR 1
COMPS 2 VP
_
_
2 VP
_
_
VFORM 6 prp
SPR NP
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
John kept
V
_
_
VFORM 6 prp
SPR NP
COMPS 5
_
_
5 NP
_
eager
HEAD| POS adj
SPR NP
COMPS VP[VFORM inf ]
_
_
b.
_
_
fond
HEAD| POS adj
SPR NP
COMPS PP[PFORM of ]
_
_
Such lexical entries will project sentences like the following:
81
(31)
S
_
VFORM pres
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
V
_
VFORM pres
_
.
.
.
.
.
.
.
.
.
.
.
.
Monkeys
V
_
_
VFORM pres
COMPS 2
_
_
2 AP
are
A
_
COMPS 1
_
1 VP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
.
eager
V
_
VFORM inf
_
VP
_
proximity
HEAD| POS noun
SPR
_
1 DP
_
COMPS
_
2 PP[PFORM to]
_
_
_
b.
_
_
faith
HEAD| POS noun
SPR
_
1 DP
_
COMPS
_
2 PP[PFORM in]
_
_
_
Though many more details remain to be covered for the various complement types of lexical cat-
egories, the discussion so far has given an idea of what kinds of complement lexical categories
select for.
5.3 Feature Specications for the Subject
In general, most verbs select a regular NP as subject:
(34) a. John/Some books/The spy disappeared.
b. The teacher/The monkey/He fooled the students.
However, as noted in the previous chapter, certain English verbs select only it or there as subject:
(35) a. It/*John/*There rains.
b. There/*The spy lies a man in the park.
The pronouns it and there are often called expletives, indicating that they do not have or con-
tribute any meaning. The use of these expletives is restricted to particular contexts or verbs,
though both forms have regular pronoun uses as well. One way to specify such lexical speci-
cations for subjects is to make use of a form value specication for nouns: all regular nouns
[NFORM norm(al)] as default specication; overall we classify nouns as having three different
NFORM values: normal, it, and there. Given the NFORM feature, we can have the following
lexical entries for the verbs above:
(36) a.
_
_
rained
SPR NP[NFORM it]
COMPS
_
_
58
DP covers not only simple determiners like a, the, and that, but also includes a possessive phrase like Johns. In
Chapter 6 we cover NP structures in detail.
83
b.
_
_
fooled
SPR NP[NFORM norm
COMPS NP
_
_
One thing to note here is that weather verbs like rain and existential verbs like be do not assign
any semantic role to their subject. This in turn means that in terms of the ARG-ST, weather verbs
at least do not have any semantic arguments at all. However, English syntax has a rule that every
clause must have a subject,
59
We need to ensure that the Argument Realization Constraint forces
all verbs to have a SPR value even if there is no corresponding value in ARG-ST:
60
(37)
_
_
rain
SPR NP[NFORM it]
COMPS
ARG-ST
_
_
We can also observe that only a limited set of verbs require their subject to be [NFORMthere]:
61
(38) a. There exists only one truly amphibian mammal.
b. There arose a great storm.
(39) a. There exist few solutions which are cost-effective.
b. There is a riot in the park.
c. There remained just a few problems to be solved.
The majority of verbs do not allow there as subject:
(40) a. *There runs a man in the park.
b. *There sings a man loudly.
For the sentences with there subjects, we rst consider the verb forms which have regular sub-
jects. A verb like exist in (38a) takes one argument when used in an example like Only one truly
amphibian mammal exists, and the argument will be realized as the SPR, as dictated by the
entry in (41a). In addition, such verbs can introduce there as the subject, through the Argument
Realization option given in (41b), which is the form that occurs in the structure of (38a).
(41) a.
_
_
exists
SPR 1 NP
COMPS
ARG-ST 1 NP
_
_
59
This constraint can be found instantiated as the Final 1 Law of Relational Grammar (Perlmutter (1983), Perlmutter
and Rosen (1984)), or the Extended Projection Principle in Government-Binding Theory (Chomsky (1981)), or the
subject condition in Lexical-Functional Grammar (Bresnan (1982)).
60
This in turn means that we need to have a ner-grained approach to the Argument Realization Constraints.
61
Some verbs such as arise or remain sound a little archaic in these constructions.
84
b.
_
_
exists
SPR NP[NFORM there]
COMPS 1 NP
ARG-ST 1 NP
_
_
5.4 Clausal Complement or Subject
5.4.1 Verbs Selecting a Clausal Complement
We have seen that the COMPS list includes predominantly phrasal elements. However, there are
verbs selecting not just a phrase but a whole clause as a complement, either nite or nonnite.
For example, consider the complements of think or believe:
(42) a. I think (that) the press has a check-and-balance function.
b. They believe (that) Charles Darwins theory of evolution is just a scientic the-
ory.
The C (complementizer) that here is optional, implying that this kind of verb selects for a nite
complement clause of some type, which we will notate as a [VFORM n] clause. That is, these
verbs will have one of the following two COMPS values:
(43) a.
_
COMPS
_
S[VFORM n]
_
_
b.
_
COMPS
_
CP[VFORM n]
_
_
If the COMPS value only species a VFORM value, the complement can be either S or CP. This
means that we can subsume these two uses into the following single lexical entry, suppressing
the category information of the sentential complement:
62
(44)
_
_
believe
HEAD| POS verb
COMPS
_
[VFORM n]
_
_
_
We can also nd somewhat similar verbs like demand and require:
(45) a. John demanded [that she stop phoning him].
b. The rules require [that the executives be polite].
Unlike think or believe, these verbs which introduce a subjunctive clause typically only take a
CP[VFORM bse] as their complement: the nite verb itself is actually in the bse form. Observe
the structure of (45b):
62
Although the categories V or VP are also potentially specied as [VFORM n], such words or phrases cannot
be complements of verbs like think or believe. This is because complements are typically saturated phrases, with no
unsatised requirements for complements or speciers. While S and CP are saturated categories projected from V, VP
and V are not saturated.
85
(46)
S
NP
VP
_
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
.
The rules
V
_
COMPS 2
_
2 CP
_
VFORM bse
_
.
.
.
.
.
.
.
.
.
.
.
.
require
C
_
VFORM bse
_
S
_
VFORM bse
_
.
.
.
.
.
.
.
.
.
.
.
.
_
that
HEAD
_
_
POS comp
VFORM 1
_
_
SPR
COMPS S[VFORM 1 ]
_
_
This lexical information will then allow us to pass on the VFORM value of S to the head C
and then be percolated up to the CP according to the HFP. This encodes the intuition that a
complementizer agrees with regard to VFORM value with its complement sentence.
There are also verbs which select a sequence of an NP followed by a CP as complements. NP
and CP are abbreviations for feature structure descriptions that include the information [HEAD
noun] and [HEAD comp], respectively:
(48) a. Joe warned the class that the exam would be difcult.
b. We told Tom that he should consult an accountant.
c. Mary convinced me that the argument was sound.
The COMPS value of such verbs will be as in (49):
86
(49)
_
COMPS NP, CP[VFORM n]
_
In addition to the that-type of CP, there is an innitive type of CP, headed by the comple-
mentizer for. Some verbs select this nonnite CP as the complement:
(50) a. Tom intends for Sam to review that book.
b. John would prefer for the children to nish the oatmeal.
The data show that verbs like intend and prefer select an innitival CP clause. The structure of
(50a) is familiar, but now has a nonnite VFORM value within it:
(51)
S
NP
VP
_
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
Tom
V
_
COMPS 2
_
2 CP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
intends
C
_
VFORM inf
_
S
_
VFORM inf
_
for NP
VP
_
VFORM inf
_
.
.
.
.
.
.
.
.
.
.
.
Sam
V
_
VFORM inf
_
VP
_
intends
HEAD
_
_
POS verb
VFORM pres
_
_
COMPS CP[VFORM inf ]
_
_
87
To ll out the analysis, we need explicit lexical entries for the complementizer for and for to,
which we treat as an (innitive) auxiliary verb. In fact, to has a distribution very similar to the
nite modal auxiliaries such as will or must, differing only in the VFORM value.
(53) a.
_
_
for
HEAD
_
_
POS comp
VFORM inf
_
_
COMPS S[VFORM inf ]
_
_
b.
_
_
to
HEAD
_
_
POS verb
VFORM inf
_
_
COMPS VP[VFORM bse]
_
_
Just like the complementizer that, the complementizer for selects an innitival S as its comple-
ment, inheriting its VFORM value too. The evidence that the complementizer for requires an
innitival S can be found from coordination data:
(54) a. For John to either [make up such a story] or [repeat it] is outrageous.
(coordination of bse VPs)
b. For John either [to make up such a story] or [to repeat it] is outrageous.
(coordination of inf VPs)
c. For [John to tell Bill such a lie] and [Bill to believe it] is outrageous.
(coordination of inf Ss)
Given that only like categories (constituents with the same label) can be coordinated, we can
see that base VPs, innitival VPs, and innitival Ss are all constituents.
One thing to note here is that the verbs which select a CP[VFORM inf ] complement can also
take a VP[VFORM inf ] complement:
(55) a. John intends to review the book.
b. John would prefer to nish the oatmeal.
By underspecifying the category information of complements, we can generalize this subcate-
gorization information:
(56)
_
_
intend
HEAD| POS verb
COMPS
_
[VFORM inf ]
_
_
_
Since the specication [VFORM inf ] is quite general, it can be realized either as CP[VFORM
inf ] or VP[VFORM inf ].
However, this does not mean that all verbs behave alike: not all verbs can take variable
complement types such as an innitival VP or S. For examples, try, tend, hope, and others
88
select only a VP[inf ] as attested by the data:
(57) a. Tom tried to ask a question.
b. *Tom tried for Bill to ask a question.
(58) a. Tom tends to avoid confrontations.
b. *Tom tends for Mary to avoid confrontations.
(59) a. Joe hoped to nd a solution.
b. *Joe hoped for Beth to nd a solution.
Such subcategorization differences are hard to predict just from the meaning of verbs: they are
simple lexical specications which language users need to learn.
There is another generalization that we need to consider with respect to the property of verbs
that select a CP: most verbs that select a CP can at rst glance select an NP, too:
(60) a. John believed it/that he is honest.
b. John mentioned the issue to me/mentioned to me that the question is an issue.
Should we have two lexical entries for such verbs or can we have a simple way of representing
such a pattern? To reect such lexical patterns, we will assume that parts of speech come in
families and can protably be analyzed in terms of typed feature structures. The part-of-speech
types we will assume form the hierarchy illustrated in (61):
63
(61)
part-of-speech
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
nominal
_
SPR 1 [POS nominal]
COMPS 2 NP
ARG-ST 1 , 2
_
_
b.
_
_
SPR 1 NP
COMPS 2 NP
ARG-ST 1 , 2
_
_
These different realizations all hinge on the lexical properties of the given verb, and only some
verbs allow the dual realization described by (68)a.
A clausal subject is not limited to a nite that-headed CP, but there are other clausal types:
(69) a. [That John sold the ostrich] surprised Bill.
(that-clause CP subject)
b. [(For John) to train his horse] would be desirable.
(innitival CP or VP subject)
c. [That the king or queen be present] is a requirement on all Royal weddings.
(subjunctive that-clause CP subject)
90
d. [Which otter you should adopt rst] is unclear.
(wh-question CP subject)
Naturally, each particular predicate dictates which kinds of subjects are possible, as in (69),
and which are not, as in (70):
(70) a. *That Fred was unpopular nominated Bill.
b. *That Tom missed the lecture was enjoyable.
c. *For John to remove the mother is undeniable.
d. *How much money Gordon spent is true.
For example, the difference between the two verbs nominate and surprise can be seen in these
partial lexical entries:
(71) a.
_
_
nominate
VAL
_
_
SPR 1 NP
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
b.
_
_
surprise
VAL
_
_
SPR 1 [POS nominal]
COMPS 2 NP
_
_
ARG-ST 1 , 2
_
_
Unlike nominate, the rst argument of surprised can be a nominal. This means that its subject
can be either an NP or a CP.
5.4.3 Adjectives Selecting a Clausal Complement
Like verbs, certain adjectives can also select CPs as their complements. For example, condent
and insistent select a CP, whereas eager selects an innitival CP:
(72) a. Tom is condent [that the elephants respect him].
b. Tom is insistent [that the witnesses be truthful].
(73) a. Tom seems eager [for her brother to catch a cold].
b. Tom seems eager [to catch a cold].
We can easily nd more adjectives which select a CP complement:
(74) a. I am ashamed that I neglected you.
b. I am delighted that Mary nished his thesis.
c. We are content for the cleaners to return the drapes next week.
The lexical entries for some adjectives are given in (75):
91
(75) a.
_
_
condent
HEAD| POS adj
COMPS CP[VFORM n]
_
_
b.
_
_
insistent
HEAD| POS adj
COMPS CP[VFORM bse]
_
_
c.
_
_
eager
HEAD| POS adj
COMPS [VFORM inf ]
_
_
Such lexical entries, interacting with the Head-Complement Rule, the Head-Specier Rule, and
the HFP, can license analyses such as (76), for (73b):
(76)
S
NP
VP
Tom
V
_
COMPS 2
_ 2 AP
seems
A
_
COMPS 3
_
3 VP
_
VFORM inf
_
eager V
VP
_
VFORM bse
_
to V NP
catch a cold
When the head adjective eager combines with its complement, VP[inf ], it satises the Head-
Complement Rule. The same rule allows the combination of the verb seem with its AP comple-
ment.
5.4.4 Nouns Selecting a Clausal Complement
Nouns can also select a CP complement, for example, eagerness:
92
(77) a. (Johns) eagerness [for Harry to win the election]
b. (Johns) eagerness [to win the election]
These examples imply that eagerness will have the following lexical information:
(78)
_
_
eagerness
HEAD| POS noun
COMPS
_
[VFORM inf ]
_
_
_
This lexical entry will allow a structure like the following:
(79)
NP
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
DP
.
.
.
.
.
.
.
.
.
Johns
N
_
COMPS 2
_
2 VP
_
VFORM inf
_
_
his
HEAD| POS det
SPR
COMPS
_
_
b.
_
_
John
HEAD| POS noun
SPR
COMPS
_
_
c.
_
_
s
HEAD| POS det
SPR NP
COMPS
_
_
d.
_
_
friend
HEAD| POS noun
SPR DP
COMPS
_
_
These lexical entries will project NP structures like the following:
(7) a.
NP
1 DP
N
_
_
SPR 1 DP
COMPS
_
_
Det
N
_
_
SPR 1 DP
COMPS
_
_
his friend
99
b.
NP
1 DP
_
_
SPR 1 DP
COMPS
_
_
2 NP
Det
_
_
SPR 2 NP
COMPS
_
_
N
_
_
SPR 1 DP
COMPS
_
_
John s friend
Keen readers may have noticed that no rules that we have covered so far license the projection
of N to N
: this projection has only the head element. To be more precise and allow this kind of
vacuous projection, the grammar needs the following Head-Only Rule:
(8) Head-Only Rule:
XP
_
VAL 1
_
X
_
VAL 1
_
This rule will also license a lexical element to project into a phrase, either X
or XP, with no
change in the VAL (SPR and COMPS) values, as illustrated in the following:
(9)
NP
_
_
VAL 1
_
_
SPR
COMPS
_
_
_
_
VP
_
_
VAL 1
_
_
SPR 1 NP
COMPS
_
_
_
_
N
_
_
VAL 1
_
_
SPR
COMPS
_
_
_
_
V
_
_
VAL 1
_
_
SPR 1 NP
COMPS
_
_
_
_
cookies ran
Applied to nouns, the Head-Only Rule will predict examples like the following in which bare
plural or mass nouns are projected into a fully-formed NPs:
(10) a. Students ran.
b. John wants to buy cookies.
c. Advice is cheap.
100
(11) a. *Student ran.
b. *John wants to buy cookie.
c. *An advice is cheap.
In fact, singular countable nouns as well as uncountable nouns can be used without a specier
when they are used as a predicative expression, as denoting a generic individual, or as repre-
senting a habitual activity:
(12) a. He is president of the university. (predicative)
b. Pepper makes people sneeze. (generic)
c. Sally brews beer. (habitual)
(13) a. Your friends are Europeans. (predicative)
b. Ostriches are large ightless birds. (generic)
c. John sells shoes. (habitual)
In these usages, the nouns will directly project to a full NP without combining with any com-
plement, modier, or specier.
6.2.2 Projection of Pronouns
The core class of pronouns in English includes at least three main subgroups:
(14) a. Personal pronouns: I, you, he, she, it, they, we
b. Reexive pronouns: myself, yourself, himself, herself, itself
c. Reciprocal pronoun: each other
Personal pronouns refer to specic persons or things and take different forms to indicate person,
number, gender, and case. They participate in agreement relations with their antecedent, the
phrase which they are understood to be referring to (indicated by the underlined parts of the
examples in (15)).
(15) a. After reading the pamphlet, Judy threw it/*them into the garbage can.
b. I got worried when the neighbors let their/*his dogs out.
Reexive pronouns are special forms which typically are used to indicate a reexive activity
or action, which can include mental activities.
(16) a. After the party, I asked myself why I had faxed invitations to everyone in my
ofce building.
b. Edward usually remembered to send a copy of his e-mail to himself.
As noted earlier, these personal or reexive pronouns neither take a determiner nor combine
with an adjective except in very restricted constructions.
68
.
68
These restricted constructions involve some indenite pronouns (e.g., a little something, a certain someone)
101
6.2.3 Projection of Proper Nouns
Since proper nouns usually refer to something or someone unique, they do not normally take a
plural form and cannot occur with a determiner:
(17) a. John, Bill, Seoul, January, . . .
b. *a John, *a Bill, *a Seoul, *a January, . . .
However, proper nouns can be converted into countable nouns when they refer to a particular
individual or type of individual:
(18) a. No John Smiths attended the meeting.
b. This John Smith lives in Seoul.
c. There are three Davids in my class.
d. Its nothing like the America I remember.
e. My brother is an Einstein at maths.
In such cases, proper nouns are converted into common nouns, may select a specier, and take
other nominal modiers.
6.3 Agreement Types and Morpho-syntactic Features
6.3.1 Noun-Determiner Agreement
Common nouns in English participate in three types of agreement. First, they are involved in
determiner-noun agreement. All countable nouns are used either as singular or plural. When
they combine with a determiner, there must be an agreement relationship between the two:
(19) a. this book/that book
b. *this books/*that books/these books/those books
c. *few dog/few dogs
The data in turn means that the head nouns number value should be identical to that of its spec-
ier, implying that determiners and nouns have NUM (number) information as their syntactic
AGR (agreement) value:
(20)
a.
_
_
a
HEAD
_
_
POS det
AGR| NUM sing
_
_
SPR
COMPS
_
_
b.
_
_
book
HEAD
_
_
POS noun
AGR | NUM sing
_
_
SPR DP[NUM sing]
COMPS
_
_
Common nouns thus impose a specic NUM value on the specier:
102
(21)
NP
_
AGR| NUM sing
_
1 DP[NUM sing]
N
_
AGR| NUM sing
SPR 1 DP
COMPS
_
_
a
N
_
_
AGR| NUM sing
SPR 1 DP
COMPS
_
_
book
The singular noun book selects a singular determiner like a. Notice that the AGR value on the
head noun book is passed up to the whole NP, marking the whole NP as singular, so that it can
combine with a singular VP, if it is the subject.
In addition, there is nothing preventing a singular noun from combining with a determiner
which is not specied at all for a NUM value:
(22) a. *those book, *these book, . . .
b. no book, the book, my book, . . .
Determiners like the, no and my are not specied for a NUM value. Formally, their NUM value
is underspecied as num(ber). That is, the grammar of English has the underspecied value num
for the feature NUM, with two subtypes sing(ular) and pl(ural):
(23)
num
sing pl
Given this hierarchy, nouns like book requiring a singular Det can combine with determiners
like the whose AGR value is num. This is in accord with the grammar since the value num is a
supertype of sing.
103
6.3.2 Pronoun-Antecedent Agreement
As noted earlier, a second type of agreement is pronoun-antecedent agreement, as indicated in
(24).
(24) a. If John wants to succeed in corporate life, he/*she has to know the rules of the
game.
b. The critique of Platos Republic was written from a contemporary point of view.
It was an in-depth analysis of Platos opinions about possible governmental
forms.
The pronoun he or it here needs to agree with its antecedent not only with respect to the number
value but also with respect to person (1st, 2nd, 3rd) and gender (masculine, feminine, or neuter)
values too. This shows us that nouns have also information about person and gender as well as
number in the AGR values:
(25)
a.
_
_
book
HEAD
_
_
POS noun
AGR
_
_
NUM sing
GEN neut
PER 3rd
_
_
_
_
SPR DP[NUM sing]
COMPS
_
_
b.
_
_
he
HEAD
_
_
POS noun
AGR
_
_
NUM sing
PER 3rd
GEN masc
_
_
_
_
SPR
COMPS
_
_
As we have briey shown, nouns have NUM, PER(SON), and GEN(DER) for their AGRvalues.
The PER value can be 1st, 2nd or 3rd; the GEN value can be masc(uline), fem(inine) or neut(er).
The NUM values are shown in (23) above.
6.3.3 Subject-Verb Agreement
The third type of agreement is subject-verb agreement, which is one of the most important
phenomena in English syntax. Let us look at some slightly complex examples:
(26) a. The characters in Shakespeares Twelfth Night *lives/live in a world that has been
turned upside-down.
b. Students studying English read/*reads Conrads Heart of Darkness while at
university.
As we can see here, the subject and the verb need to have an identical number value; and the
person value is also involved in agreement relations, in particular when the subject is a personal
pronoun:
(27) a. You are/*is the only person that I can rely on.
b. He is/*are the only person that I can rely on.
104
These facts show us that a verb lexically species the information about the number as well as
person values of the subject that it selects for.
To show how the agreement system works, we will use some simpler examples:
(28) a. The boy swims/*swim.
b. The boys swim/*swims.
English verbs will have at least the following selectional information:
(29)
_
_
swims
HEAD
_
_
POS verb
VFORM pres
_
_
SPR
_
NP
_
_
PER 3rd
NUM sing
_
_
_
_
_
The present-tense verb swims thus species that its subject (SPRs value) carries a 3rd singular
AGR information. This lexical information will license a structure like the following:
(30)
S
2 NP
_
_
PER 3rd
NUM sing
_
_
VP
_
_
SPR
_
2 NP
_
_
PER 3rd
NUM sing
_
_
_
_
_
The boy
V
_
SPR 2 NP
_
swims
Only when the verb combines with a subject satisfying its AGR requirement, will we have a
well-formed head-subject phrase. In other words, if this verb were to combine with a subject
with an incompatible agreement value, we would generate an ungrammatical example like *The
boys swims in (28b). In this system, what subject-verb agreement is is structure-sharing between
the AGR value of the subject (SPR value of the verb) and that of the NP that the verb combines
with.
The acute reader may have noticed that there are similarities between noun-determiner agree-
ment and subject-verb agreement, that is, in the way that agreement works inside NP and inside
105
S. Both NP and S require agreement between the head and the specier. Reecting this obser-
vation, we can modify the Head-Specier Rule as following:
(31) Head-Specier Rule:
XP SPR[AGR 1 ], H
_
AGR 1
_
This revised rule guarantees that English head-specier phrases require their head and specier
to share agreement features.
6.4 Semantic Agreement Features
6.4.1 Morpho-syntactic and Index Agreement
What we have seen so far is that the morphosyntactic AGR values of noun or verb can be spec-
ied, and may be inherited by phrases built out of them. However, consider now the following
examples:
(32) a. [The hash browns at table nine] are/*is getting cold.
b. [The hash browns at table nine] is/*are getting angry.
When (32b) is spoken by a waiter to another waiter, the subject refers to a person who ordered
hash browns. A somewhat similar case is found in (33):
(33) King prawns cooked in chili salt and pepper was very much better, a simple dish
succulently executed.
Here the verb form was is singular to agree with the dish being referred to, rather than with a
plurality of prawns. If we simply assume that the subject phrase inherits the morphosyntactic
agreement features of the head noun (hash) browns in (32b) and (King) prawns in (33), and
requires that these features match those of the verb, we would not expect the singular verb form
to be possible at all in these examples.
In the interpretation of a nominal expression, it must be anchored to an individual in the
situation described. We call this anchoring value the noun phrases index value. The index of
hash browns in (32a) must be anchored to the plural entities on the plate, whereas that of hash
browns in (32b) is anchored to a customer who ordered the food.
English agreement is not purely morpho-syntactic as described in the sections above, but
context-dependent in various ways, via the notion of index that we have just introduced. Often
what a given nominal refers to in the real world is important for agreement index agreement.
Index agreement involves sharing of referential indexes, closely related to the semantics of a
nominal, and somewhat separate from the syntactic agreement feature AGR. This then requires
us to distinguish the morphological AGR value and semantic (SEM(ANTIC)) IND(EX) value.
So, in addition to the morphological AGR value introduced above, each noun will also have a
semantic IND value representing what the noun refers to in the actual world.
106
(34) a.
_
_
boy
SYN| HEAD
_
_
POS noun
AGR| NUM sing
_
_
SEM| IND| NUM sing
_
_
b.
_
_
boys
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND| NUM pl
_
_
The lexical entry for boy indicates that it is syntactically a singular noun (through the feature
AGR) and semantically also denotes a singular entity (through the feature IND). And the verb
will place a restriction on its subjects IND value rather than its morphological AGR value:
69
(35)
_
_
swims
SYN
_
_
HEAD
_
_
POS verb
AGR| NUM sing
_
_
VAL| SPR
_
NP[IND| NUM sing]
_
_
_
SEM| IND s0
_
_
The lexical entry for swims in (35b) indicates that it is morphologically marked as singular (the
AGR feature) whereas it selects a subject to be linked to a singular entity in the context (by the
feature IND). Different from nouns, the verbs own IND value is a situation index (s0) in which
the individual referred to through the SPR value is performing the action of swimming. If the
referent of this subject (its IND value) does not match, we would generate an ungrammatical
example like *The boys swims:
69
The IND value of a noun will be an individual index (i, j, k, etc) whereas that of a verb or predicative adjective will
be a situation index such as s0, s1, s2, etc.
107
(36)
*S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
2 NP
_
_
IND i
_
_
PER 3rd
NUM pl
_
_
_
_
VP
_
_
SPR
_
2 NP
i
_
_
PER 3rd
NUM sing
_
_
_
_
_
The boys
V
_
SPR 2 NP
i
_
swims
In the most usual cases, the AGR and IND value are identical, but they can be different, as in
examples like (32b). This means that depending on the context, hash browns can have different
IND values:
70
(37)
a.
_
_
hash browns
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND 1 | NUM pl
_
_
(when referring to the food itself)
b.
_
_
hash browns
SYN| HEAD
_
_
POS noun
AGR| NUM pl
_
_
SEM| IND 1 | NUM sing
_
_
(when referring to a customer, or to a dish)
In the lexical entry (37b), the AGRs NUM value is plural but its INDs NUM value is singu-
lar. As shown by (32), the reference hash browns can be transferred from cooked potatoes to
the customer who ordered it. This means that given an appropriate context, there could be a
mismatch between the morphological form of a noun and the index value that the noun refers
to.
6.4.2 More on Semantic Aspects of Agreement
Here we introduce one more complex aspect of English agreement patterns. Consider the ex-
amples in (38):
70
As indicated here, the lexical expression now has two features SYN (SYNTAX) and SEM (SEMANTICS). The
feature SYN includes HEAD and SPR and COMPS. The feature SEM is for semantic information. As our discussion
goes on, we will add more to this part.
108
(38) a. [Five pounds] is/*are a lot of money.
b. [Two drops] deodorizes/*deodorize anything in your house.
c. [Fifteen dollars] in a week is/*are not much.
d. [Fifteen years] represents/*represent a long period of his life.
e. [Two miles] is/*are as far as they can walk.
In all of these examples with measure nouns, the plural subject combines with a singular verb.
An apparent conict arises from the agreement features of the head noun. For proper agreement
inside the noun phrase, the head noun has to be plural, but for subject-verb agreement the noun
has to be singular.
A similar mismatch is also found in cases with terms for social organizations or collections,
as in (39) and (40):
71
(39) a. [This/*these government] has/*have broken its promises.
b. [This/*these government] have/*has broken their promises.
(40) a. [This/*these England team] have/*has put themselves in a good position to win
the championship.
b. [This/*these England team] *have/has put itself in a good position to win the
championship.
The head noun has to be singular so that it can combine with a singular determiner. But the
conicting fact is that the singular noun phrase can combine even with a plural verb as well
as a singular verb. Since the only possible number value of the determiner is singular for the
head noun, the head noun cannot be anchored to plural entities unless we allow the mode of
individuation to be changeable even within the same sentence domain.
What this indicates is that subject-verb agreement and noun-specier agreement are differ-
ent. In fact, English determiner-noun agreement is only a reection of morpho-syntactic agree-
ment features between determiner and noun, whereas both subject-verb agreement and pronoun-
antecedent agreement are index-based agreement. This is represented in (41), and shown by the
example in (42), where the underlined parts have singular agreement with four pounds, which
is internally plural.
(41) Morpho-syntactic agreement (AGR)
Det
head-noun
verb . . .
Index agreement (IND)
(42) [Four pounds] was quite a bit of money in 1950 and it was not easy to come by.
Given the separation of the morphological AGR value and the semantic IND value, nothing
blocks mismatches between the two (AGR and IND) as long as all the other constraints are
71
The accepability of some of these examples varies in different varieties of English; British English typically allows
more of the mismatching type of example.
109
satised. Consider the example in (38). The nouns pounds and drops here are morphologically
plural and thus must select a plural determiner, as argued so far. But when these nouns are
anchored to the group as a whole that is, conceptualized as referring to a single measure, the
index value has to be singular, as represented in (43).
(43)
_
_
pounds
SYN
_
_
HEAD
_
_
POS noun
AGR 1 | NUM pl
_
_
SPR
_
DP
_
AGR 1
_
_
_
_
SEM| IND | NUM sing
_
_
As indicated in the lexical entry (43), the morpho-syntactic number value of pounds is plural
whereas the index value is singular. In the present analysis, this would mean that pounds will
combine with a plural determiner but with a singular verb. This is possible, as noted earlier in
section 2, since the index value is anchored to a singular individual in the context of utterance.
The present analysis thus generates the following structure for the sentence (38a):
(44)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
3 NP
_
_
AGR 1
IND i
_
_
VP
_
SPR 3 NP
i
Det
_
AGR 1 | NUM pl
_
N
_
_
AGR 1
IND i
_
_
V
_
SPR 3 NP
i
_
NP
_
this
HEAD
_
_
POS det
AGR| NUM sing
_
_
_
_
b.
_
_
government
SYN
_
_
HEAD
_
_
POS noun
AGR| NUM sing
_
_
_
_
SEM| IND | NUM pl
_
_
As represented in (47a) and (47b), this and government agree each other in terms of the morpho-
syntactic agreement number value whereas the index value of government is what matters for
subject-verb agreement. This in turn means that when government refers to the individuals in
this given group, the whole NP this government carries a plural index value. This then allows
the NP this government to combine even with a plural VP dislike changes.
Now consider the examples (40) with pronoun-antecedent agreement.
(48) a. [This England team] now puts [itself/*themselves] in a good position to win the
championship.
b. [This England team] now put [themselves/*itself] in a good position to win the
championship.
The point here is that the number value of the verb matches that of the reexive pronoun itself
and themselves. Pronoun-antecedent agreement is also index-based, rather than morpho-syntax-
based. Given a simple Binding Condition specifying that a reexive pronoun such as himself,
itself, themselves has to be bound by a preceding argument of the same verb in the argument
structure (ARG-ST), the grammar can predict the contrast here. Consider the ARG-ST of the
main verb put:
72
(49)
_
_
put
ARG-ST NP
i
, NP[anaphor]
i/j
, PP
_
_
The verb put selects for three arguments. If the second argument is an anaphor, it must be bound
by a preceding argument with respect to its IND value in accordance with the assumed Binding
72
These forms such as himself, myself, itself are sometimes called anaphors, or sometimes simply reexives. See
Exercise 7 for further discussion.
111
Condition. This means that in (48a) the head noun team must have a singular index value for
subject agreement since the verb is singular: any reexive noun in the same argument structure
would have to have a singular index value too. Meanwhile in (48b), the verb is plural, implying
that the subject is anchored to individuals constituting a group. This mode of individuation
cannot be changed, thus requiring a 3rd-person plural reexive pronoun.
6.5 Partitive NPs and Agreement
6.5.1 Basic Properties
With regard to the NP-internal elements between which we may nd instances of agreement,
there are two main types of NP in English: simple NPs and partitive NPs, shown in (50) and
(51) respectively.
(50) a. some objections
b. most students
c. all students
d. much worry
e. many students
f. neither cars
(51) a. some of the objections
b. most of the students
c. all of the students
d. much of her worry
e. many of the students
f. neither of the cars
As in (51), the partitive phrases have a quantier followed by an of -phrase, designating a set
with respect to which certain individuals are quantied. In terms of semantics, these partitive
NPs are different from simple NPs in several respects.
First, the lower NP in partitive phrases must be denite; and in the of -phrase, no quantica-
tional NP is allowed, as shown in (52):
(52) a. Each student vs.*each of students
b. Some problems vs.*some of many problems
Second, not all determiners with quanticational force can appear in partitive constructions.
As shown in (53), determiners such as the, every and no cannot occupy the rst position:
(53) a. *the of the students vs. the students
b. *every of his ideas vs. every idea
c. *no of your books vs. no book(s)
Third, simple NPs and partitive NPs have different restrictions relative to the semantic head.
Observe the contrast between (54) and (55):
112
(54) a. She doesnt believe much of that story.
b. We listened to as little of his speech as possible.
c. How much of the fresco did the ood damage?
d. I read some of the book.
(55) a. *She doesnt believe much story.
b. *We listened to as little speech as possible.
c. *How much fresco did the ood damage?
d. *I read some book.
The partitive constructions in (54) allow a mass (non-count) quantier such as much, little and
some to cooccur with a lower of -NP containing a singular count noun. But as we can see in
(55), the same elements serving as determiners cannot directly precede such nouns.
Another difference concerns lexical idiosyncrasies.
(56) a. One of the people was dying of thirst.
b. Many of the people were dying of thirst.
(57) a. *One people was dying of thirst.
b. Many people were dying of thirst.
The partitives can be headed by quantiers like one and many, as shown in (56) and (57) but
unlike many, one cannot serve as a determiner when the head noun is collective as in (57a).
What the observations we have seen so far suggest is that cannot simply derive partitive con-
structions from simple noun phrases. The two constructions induce quite different lexical and
syntactic properties that no independently motivated transformation mechanisms can capture.
6.5.2 Two Types of Partitive NPs
We classify partitive NPs into two types based on the agreement facts, and call them Type I and
Type II. In Type I, the number value of the partitive phrase is always singular:
(58) Type I:
a. Each of the suggestions is acceptable.
b. Neither of the cars has air conditioning.
c. None of these men wants to be president.
In Type II, the number value depends on the head noun inside the of -NP phrase.
(59) Type II:
a. Most of the fruit is rotten.
b. Most of the children are here.
c. Some of the soup needs more salt.
d. Some of the diners need menus.
e. All of the land belongs to the government.
f. All of these cars belong to me.
113
As shown in (59), when the NP following the preposition of is singular or uncountable, the main
verb is singular. When the NP is plural, the verb is also plural. From a semantic perspective, we
see that the class of quanticational indenite pronouns including some, half, most and all may
combine either singular or plural verbs, depending upon the reference of the of -NP phrase. If
the meaning of these phrases is about how much of something is meant, the verb is singular; but
if the meaning is about how many of something is meant, the verb is plural. The expressions in
(60) also exhibit similar behavior in terms of agreement.
(60) half of, part of, the majority of, the rest of, two-thirds of, a number of (but not the
number of )
An effective way of capturing the relations between Type I and Type II constructions involves
the lexical properties of the quantiers. First, Type I and Type II involve pronominal forms
serving as the head of the constructions, which select an of -NP inside which the NP is denite:
(61) a. *neither of students, *some of water
b. neither of the two linguists/some of the water
However, we know that the two types are different in terms of agreement: the pronouns in the
Type I construction are lexically specied to be singular whereas the number value for Type II
comes from inside the selected PP.
A slight digression is in order. It is easy to see that there are prepositions whose functions
are just grammatical markers.
(62) a. John is in the room.
b. I am fond of him.
The predicative preposition in here selects two arguments John and the room. Meanwhile, the
preposition of has no predicative meaning, but just functions as a marker to the argument of
fond. As for the PPs headed by these markers, as in the partitive construction, their semantic
features are identical with the prepositional object NP. There is no semantic difference (such as
deniteness effect represented as the feature DEF in the present system) between the PP of him
and the NP him.
Within the complement PP, the agreement or index features of an NP such as of him and
those of the internal him are identical. We show this in (63) by sharing the SEM value of the NP
with that of the PP; this will also share any deniteness information from the NP to the PP:
114
(63)
PP
_
_
HEAD 2
SEM 3
_
_
P
_
_
HEAD2
_
_
AGR 1
PFORM of
_
_
SEM 3
_
_
NP
_
_
HEAD| AGR 1
SEM 3
_
_
of him
Given this basic part of the analysis, we can lexically encode the similarities and differences
between Type I and Type II in a simple manner:
(64) a.
_
_
neither
HEAD
_
_
POS noun
AGR| NUM sing
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
_
_
_
_
_
b.
_
_
some
HEAD
_
_
POS noun
AGR| NUM 1
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
AGR| NUM 1
_
_
_
_
_
(64) shows that both Type I neither and Type II some are lexically specied to require a PP
complement whose semantic value includes the denite (DEF) feature (with the value +). This
will account for the contrast in (61). However, the two types are different in terms of their
AGRs NUM value. The NUM value of Type I neither is singular, whereas that of Type II is
identied with the PPs NUM value which is actually coming from its prepositional object NP.
Showing these differences in the syntactic structures, we have the alternatives in (65):
115
(65) a.
NP[ NUM sing ]
N[ NUM sing ]
PP
neither P
NP
of the students
b.
NP[ NUM 1 ]
some P
NP[ NUM 1 ]
of the students
As shown in (65a), for Type I, it is neither which determines the NUM value of the whole
NP phrase. However, for Type II, it is the NP the students which determines the NUM value of
the whole NP.
We can check a few of the consequences of these different specications in the two Types.
Consider the contrast in (66):
(66) a. many of the/those/her apples
b. *many of some/all/no apples
(66)b is ungrammatical since many requires an of -PP phrase whose DEF value is positive.
This system also offers a simple way of dealing with the fact that quantiers like each affect
the NUM value as well as the countability of the of -NP phrase. One difference between Type I
and Type II is that Type I selects a plural of -NP phrase whereas Type II has no such restriction.
This is illustrated in (67) and (68).
(67) Type I:
a. one of the suggestions/*the suggestion/*his advice
b. each of the suggestions/*the suggestion/*his advice
c. neither of the students/*the student/*his advice
(68) Type II:
a. some of his advice/students
116
b. most of his advice/students
c. all of his advice/students
The only additional specication we need for Type I pronouns relates to the NUM value on the
PPs complement as given in (69):
(69)
_
_
each
HEAD
_
_
POS noun
AGR| NUM sing
_
_
COMPS
_
PP
_
_
PFORM of
DEF +
NUM pl
_
_
_
_
_
We see that quantiers like each select a PP complement whose NUM value is plural.
Type II pronouns do not have such a requirement on the PP complement note that all the
examples in (70) are acceptable, in contrast to those in (71) (cf. Baker 1995):
(70) a. Most of Johns boat has been repainted.
b. Some of the record contains evidence of wrongdoing.
c. Much of that theory is unfounded.
(71) a. *Each of Johns boat has been repainted.
b. *Many of the record contained evidence of wrongdoing.
c. *One of the story has appeared in your newspaper.
The contrast here indicates that Type II pronouns can combine with a PP whose inner NP is
singular. This is simply predicted since our analysis places no restriction on the NUM value of
the inner NP.
We are also in a position now to understand some differences between simple NPs and par-
titive NPs. Consider the following examples:
(72) a. many dogs/*much dog/the dogs
b. much furniture/*many furniture/the furniture
(73) a. few dogs/*few dog/*little dogs/*little dog
b. little furniture/*little furnitures/*few furniture/*few furnitures
The data here indicate that in addition to the agreement features we have seen so far, common
nouns also place a restriction on the countability value of the selected specier. Specically, a
countable noun selects a countable determiner as its specier. To capture this agreement restric-
tion, we introduce a new feature COUNT (COUNTABLE):
117
(74)
a.
_
_
dogs
HEAD| POS noun
SPR DP[COUNT +]
_
_
b.
_
_
furniture
HEAD| POS noun
SPR DP[COUNT ]
_
_
The lexical specication on a countable noun like dogs requires its specier to be [COUNT +],
to prevent formations like *much dogs. This in turn means that determiners must also carry the
feature COUNT:
(75)
a.
_
_
many
HEAD
_
_
POS det
COUNT +
_
_
_
_
b.
_
_
the
HEAD
_
_
POS det
COUNT bool
_
_
_
_
c.
_
_
little
HEAD
_
_
POS det
COUNT
_
_
_
_
Notice here that some determiners such as the are not specied for a value for COUNT. Ef-
fectively, the value can be either + or , licensing combination with either a countable or an
uncountable noun (the book or the furniture).
Now consider the following contrast:
(76) a. much advice vs. *many advice
b. *much story vs. many stories
(77) a. much of the advice vs. *many of the advice
b. much of the story vs. many of the stories
Due to the feature COUNT, we understand now the contrast between much advice and *many
advice or the contrast between *much story and many stories. The facts in partitive structures
are slightly different, as (77) shows, but the patterns in the data directly follow from these lexical
entries:
(78)
a.
_
_
many
HEAD| POS noun
COMPS
_
PP
_
_
NUM pl
DEF +
_
_
_
_
_
b.
_
_
much
HEAD| POS noun
COMPS
_
PP
_
_
NUM sing
DEF +
_
_
_
_
_
The pronoun many requires a PP complement whose inner NP is plural, whereas much does not.
118
6.5.3 Measure Noun Phrases
There are also so-called measure noun phrase constructions, which are similar to partitive
constructions. Consider the following contrast:
(79) a. one pound of those beans
b. three feet of that wire
c. a quart of Bobs cider
(80) a. one pound of beans
b. three feet of wire
c. a quart of cider
Notice here that (79) is a kind of partitive construction whereas (80) just measures the amount
of the NP after of . As the examples show, measure noun phrases do not require a denite article,
which is not an option for the true partitive constructions, repeated here:
(81) *many of beans, *some of wire, *much of cider, *none of yogurt, *one of straw-
berries
In addition, there are several more differences between partitive and measure noun phrases.
For example, measure nouns cannot occur in simple noun phrases. They obligatorily require an
of -NP phrase:
(82) a. *one pound beans vs. one pound of beans
b. *three feet wire vs. three feet of wire
c. *a quart cider vs. a quart of cider
Further, unlike partitive constructions, measure noun phrases require a numeral as their spec-
ier:
(83) a. *one many of the books, *several much of the beer
b. one pound of beans, three feet of wire
Further complications arise due to the existence of defective measure noun phrases. Consider
the following examples:
(84) a. *a can tomatoes/a can of tomatoes/one can of tomatoes
b. a few suggestions/*a few of suggestions/*one few of suggestions
c. *a lot suggestions/a lot of suggestions/*one lot of suggestions
Expressions like few and lot actually behave quite differently. With respect to few, it appears
that a few acts like a complex word. However, lot acts more like a noun, but unlike can, it does
not allow its specier to be a numeral.
In terms of agreement, measure noun phrases behave like Type I partitive constructions:
(85) a. A can of tomatoes is/*are added.
b. Two cans of tomatoes are/*is added.
119
We can see here that it is the head noun can or cans which determines the NUM value of the
whole NP. The inner NP in the PP does not affect the NUM value at all. These observations lead
us to posit the following lexical entry for a measure noun:
(86)
_
_
pound
HEAD
_
_
POS noun
NUM sing
_
_
SPR DP
COMPS PP
_
PFORM of
_
_
That is, a measure noun like pound requires one obligatory SPR and a PP complement. Unlike
partitive constructions, there is no deniteness restriction on the PP complement.
Finally, there is one set of words whose behavior leaves them somewhere between quantity
words and measure nouns. These are words such as dozen, hundred, and thousand:
(87) a. three hundred of your friends
b. *three hundreds of your friends
c. *three hundreds of friends
d. three hundred friends
e. hundreds of friends/*hundreds friends
Consider the behavior of hundred and hundreds here. The singular hundred, when used as
noun, obligatorily requires a PP[of ] complement as well as a numeral specier, as in (87a). The
plural hundreds requires no specier although it also selects a PP complement. Not surprisingly,
similar behavior can be observed with thousand and thousands:
(88) a. several thousand of Bills supporters
b. *several thousands of Bills supporters
c. *several thousands of supporters
d. several thousand supporters
e. thousands of supporters/*thousands supporters
One way to capture these properties is to assign the following lexical specications to hundred
and hundreds:
(89)
a.
_
_
hundred
HEAD noun
SPR
_
[ ]
_
COMPS <PP[PFORM of]>
_
_
b.
_
_
hundreds
HEAD| POS noun
SPR
COMPS <PP[PFOR of]>
_
_
120
Even though there may be some semantic reasons for all these different kinds of lexical spec-
ications, for now, stating it all directly in the lexical entries will account at least for the data
given here.
6.6 Modifying an NP
6.6.1 Prenominal Modiers
Adjectives are expressions commonly used to modify a noun. However, not all adjectives can
modify nouns. Even though most adjectives can be used either as in a modifying (attributive)
function or as a predicate (as in She is tall), certain adjectives are restricted to their usages. Ad-
jectives such alive, asleep, awake, afraid, ashamed, aware, utter, can be used only predicatively,
whereas others such as wooden, drunken, golden, main, mere are only used attributively:
(90) a. He is alive.
b. He is afraid of foxes.
(91) a. It is a wooden desk.
b. It is a golden hair.
c. It is the main street.
(92) a. *It is an alive sh. (cf. living sh)
b. *They are afraid people. (cf. nervous people)
(93) a. *This objection is main. (cf. the main objection)
b. *This fact is key. (cf. a key fact)
The predicately-used adjectives are specied with the feature PRD, and with a MOD value
being empty as default, as shown here:
(94)
_
_
alive
HEAD
_
_
POS adj
PRD +
MOD
_
_
_
_
This says that alive is used predicatively, and does not have a specication for a MOD value
(the value is empty). This lexical information will prevent predicative adjectives from also func-
tioning as noun modiers.
73
In contrast to the predicative adjective, a modifying adjective will have the following lexical
entry:
(95)
_
_
brave
HEAD
_
_
POS adj
MOD N
_
_
_
_
73
In addition, all predicative expressions select one argument, their subject (SPR). This information is not shown
here.
121
This species an adjective which modies any value whose POS is noun. This will license a
structure like the following:
(96)
NP
_
_
_
_
_
_
_
_
_
2 DP
N
_
SPR 2 DP
_
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
the
AdjP
_
MOD 1
_
1 N
_
SPR 2 DP
_
brave child
6.6.2 Postnominal Modiers
Postnominal modiers are basically the same as prenominal modiers with respect to what they
are modifying. The only difference is that they come after what they modify. Various phrases
can function as such postnominal modiers:
(97) a. [The boy [in the doorway]] waved to his father.
b. [The man [eager to start the meeting]] is Johns sister.
c. [The man [holding the bottle]] disappeared.
d. [The papers [removed from the safe]] have not been found.
e. [The money [that you gave me]] disappeared last night.
All these postnominal elements bear the feature MOD. Leaving aside detailed discussion of the
relative clause(-like) modiers in be until Chapter 12, we can say that example (97)a will have
the following structure:
(98)
NP
_
_
_
_
_
_
_
_
_
2 DP
N
_
SPR 2 DP
_
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
the
1 N
_
SPR 2 DP
_
PP
_
MOD 1
_
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
V S
.
.
.
.
.
.
.
.
.
.
.
seems NP VP[inf ]
Stephen
to be irritating
The movement of the subject Stephen to the higher subject position will correctly generate
(26b). This kind of movement to the subject position can be triggered by the requirement that
each English declarative have a surface subject (cf. Chomsky 1981). A similar movement pro-
cess can be applied to the object raising cases:
(28) a. Deep structure: Tom believes [Stephen to be irritating].
b. Surface structure: Tom believes Stephen to be irritating.
Here the embedded subject Stephen moves not to the matrix subject but to the matrix object
position:
132
(29)
S
NP
VP
John
V
NP
S
believes NP VP[inf ]
Mary
to be irritating
Control constructions are different: there is no movement operation involved. Instead, it is
the lower subject position which has special properties. Consider the examples in (30):
(30) a. John tried to please Stephen.
b. John persuaded Stephen to be more careful.
Since try and persuade assign a semantic roles to their subject, and objects, an unlled position
of the kind designated above by cannot be allowed. Instead, it is posited that there is an
unexpressed subject of the innitival VP to please Stephen and to be more careful. This is
traditionally represented as the element called PRO (a silent pronoun),, and the examples
will have the following deep structures:
(31) a. John tried [PRO to please Stephen].
b. John persuaded Stephen [PRO to be more careful].
The nal tree representations of these are as follows:
(32) a.
S
NP
VP
John
i
V S
tried NP VP[inf ]
PRO
i
to please Stephen
133
b.
S
NP
VP
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
John
V
NP
S
persuaded Stephen
i
NP VP[inf ]
PRO
i
to be more careful
An independent part of the theory of control links PRO in each case to its antecedent, marked
by coindexing. In (32a), PRO is coindexed with John whereas in (32b), it is coindexed with
Stephen.
These analyses which involve derivations on tree structures are driven by the assumption
that the mapping between semantics and syntax is very direct. For example, in (29), the verb
believe semantically selects an experiencer and a proposition, and this is reected in the initial
structure. In some syntactic respects, though, believe acts like it has an NP object (separate from
the inntival complement), and the raising operation creates this object. In contrast, persuade
semantically selects an agent, a patient, and a proposition, and hence the structure in (32)b
reects this: the object position is there all along, so to speak.
The classical transformational approach provides a useful graphical approach to understand-
ing the difference between raising and control. However, it requires assumptions about the na-
ture of grammar rather different from what we have made throughout this book. In the rest of
this chapter, we present a non-transformational account of control and raising.
7.4 A Nontransformational Approach
7.4.1 Identical Syntactic Structures
Instead of the movement approach in which movement operations and various kinds of empty
elements or positions play crucial roles, we simply focus here directly on the surface structures
of raising and control constructions. Going back to seem and try, we can observe that both select
an innitival VP, as in (33), giving the structures in (34):
(33) a.
_
_
seems
SPR NP
COMPS VP
_
VFORM inf
_
_
_
134
b.
_
_
tries
SPR NP
COMPS VP
_
VFORM inf
_
_
_
(34) a.
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
John
V
VP[inf ]
.
.
.
.
.
.
.
.
.
.
.
seems V VP[bse]
to be irritating
b.
S
.
.
.
.
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
.
.
.
.
John
V
VP[inf ]
.
.
.
.
.
.
.
.
.
.
.
tries V VP[bse]
to please Stephen
As shown here, seems and tries actually have identical structures.
The object raising verb expect and the control verb persuade also have identical valence
(SPR and COMPS) information:
(35) a.
_
_
expects
SPR NP
COMPS NP, VP
_
VFORM inf
_
_
b.
_
_
persuaded
SPR NP
COMPS NP, VP
_
VFORM inf
_
_
These two lexical entries will license the following structures:
135
(36) a.
S
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
Kim V
NP
VP
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
_
Kim V
NP
VP
_
seemed
SPR 1
COMPS
_
VP
_
_
VFORM inf
SPR 1
_
_
_
_
_
b.
_
_
tried
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
SPR NP
i
_
_
_
_
_
These two lexical entries represent the difference between seem and try: for seemed, the subject
of the VP complement is identical with its own subject (notated by 1 ) whereas for tried, only the
index value of its VP complement is identical to that of its subject. That is, the VP complements
understood subject refers to the same individual as the subject of tried. This index identity in
control constructions is clear when we consider examples like the following:
(40) Someone
i
tried NP
i
to leave the town.
The example here means that whoever someone might refer to, that same person left town.
Object raising and control predicates are no different. Raising verbs select a VP complement
whose subject is fully identical with the object. Control verbs select a VP complement whose
subjects index value is identical with that of its subject. The following lexical entries show
these properties:
(41) a.
_
_
expect
SPR 1 NP
i
COMPS
_
2 , VP
_
_
VFORM inf
SPR 2 NP
_
_
_
_
_
b.
_
_
persuade
SPR NP
COMPS
_
NP
i
, VP
_
_
VFORM inf
SPR NP
i
_
_
_
_
_
Let us look at the structures these lexical entries eventually project:
137
(42)
S
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
VP
_
_
HEAD 4
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
Kim
V
_
_
HEAD 4
SPR 1
COMPS 2 , 3
_
_
2 NP
VP
_
SPR
_
2
_
_
_
HEAD 4 | POS verb
SPR
COMPS
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
1 NP
VP
_
_
HEAD 4
SPR 1
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Kim
V
_
_
HEAD 4
SPR 1
COMPS 2 NP, 3 VP
_
_
2 NP
i
3 VP
_
SPR NP
i
_
(j, b)
This shows that the verb hit takes two arguments with the predicate relation hit, with the
notation to indicate the semantic value. The relevant semantic properties can be represented in
a feature structure system as follows:
(45)
_
_
hit
SYN| VAL
_
_
SPR NP
i
COMPS NP
j
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED hit
AGENT i
PATIENT j
_
_
_
_
_
_
_
In terms of syntax, hit is a verb selecting a subject and a complement, as shown in the value
of the feature SYN(TAX). The semantic information of the verb is represented with the fea-
ture SEM(ANTICS). It rst has the attribute IND(EX), representing what this expression refers
to; as a verb, hit refers to a situation s0 in which an individual i hits and individual j. The se-
mantic relation of hitting is represented using the feature for semantic relations (RELS). The
feature RELS has as its value a list of one feature structure, here with three further features,
PRED(ICATE), AGENT, and PATIENT. The predicate (PRED) relation is whatever the verb
denotes: in this case, hit takes two arguments.
The AGENT argument in the SEM value is coindexed with the SPR in the SYN value,
while the the PATIENT is coindexed with COMPS. This coindexing links the subcategorization
information of hit with the arguments in its semantic relation. Simply put, the lexical entry in
(45) is the formal representation of the fact that in X hits Y, X is the hitter and Y is the one hit.
Now we can use these extra parts of the representation for the semantic differences in raising
and control verbs. The subject of a raising verb like seem is not assigned any semantic role,
while that of a control verb like try is denitely linked to a semantic role. Assuming that s0
139
or s1 stand for situations denoted by an innitival VP, seem and try will have the following
simplied meaning representations:
(46) a. seem
_
seem
SYN| VAL
_
_
SPR 1
COMPS
_
VP
_
_
VFORM inf
SPR
_
1
_
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
__
PRED seem
SIT s1
__
_
_
_
_
b.
_
_
try
SYN| VAL
_
_
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
SPR
_
NP
i
_
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED try
AGENT i
SIT s1
_
_
_
_
_
_
_
We can see here that even though the verb seem selects two syntactic arguments, its meaning
relation has only one argument: note that the subject (SPR) is not coindexed with any argument
in the semantic relation. This means that the subject does not receive a semantic role (from
seem). Meanwhile, try is different. Its SPR is coindexed with the AGENT role in the semantics,
and the SPR is also coindexed with the VP complements SPR.
Now we look at object-related verbs like expect and persuade. Just like the contrast between
seem and try, the key difference lies in whether the object (y) receives a semantic role or not:
(48) a. expect
(x, s0)
140
b. persuade
(x, y, s1)
What one believes, as an experiencer, is a proposition denoted by the VP complement, whereas
what a person x persuades is not a proposition but rather, one persuades an individual y denoted
by the object to perform the proposition denoted by the VP. Once again, these differences are
more clearly represented in feature structures:
(49) a.
_
_
expect
SYN| VAL
_
_
SPR NP
i
COMPS
_
2 , VP
_
_
VFORM inf
SPR 2 NP
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED expect
EXPERIENCER i
SIT s1
_
_
_
_
_
_
_
b.
_
_
persuade
SYN| VAL
_
_
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
j
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED persuade
AGENT i
THEME j
SIT s1
_
_
_
_
_
_
_
As seen in the lexical entries, expect has two semantic arguments, EXPERIENCER and SIT: the
object is not linked to a semantic argument of expect. In contrast, persuade has three semantic
arguments: AGENT, THEME, and SIT. We can thus conclude that raising predicates assign one
less semantic role in their argument structures than the number of syntactic dependents, while
with control predicates, there is a one-to-one correlation.
141
7.5 Explaining the Differences
7.5.1 Expletive Subject and Object
Recall that for raising verbs like seem and believe, the subject and object respectively is depen-
dent for its semantic properties solely upon the type of VP complement. This fact is borne out
by the examples in (50):
(50) a. There/*It/*John seems [to be a fountain in the park].
b. We believed there/*it/*John [to be a fountain in the park].
Control verbs are different, directly assigning a semantic role to the subject or object. Hence
expletives cannot appear (illustrated here for the subject of try):
(51) a. *There/*It/John tries to leave the country.
b. We believed *there/*it/John to try to leave the country.
7.5.2 Meaning Preservation
We noted above that in a raising example such as (52a), the idiomatic reading can be preserved,
but not in a control example like (52b):
(52) a. The cat seems to be out of the bag.
b. The cat tries to be out of the bag.
This is once again because the subject of seems does not have any semantic role: its subject is
identical with the subject of its VP complement to be out of the bag, whereas the subject of tries
has its own agent role.
Exactly the same explanation applies to the following contrast:
(53) a. The dentist is likely to examine Pat.
b. Pat is likely to be examined by the dentist.
Since likely is a raising predicate, as long as the expressions The dentist examines Pat and Pat
is examined by the dentist have roughly the same meaning, the two raising examples will also
have roughly the same meaning.
However, control examples are different:
(54) a. The dentist is eager to examine Pat.
b. Pat is eager to be examined by the dentist.
The control adjective eager assigns a semantic role to its subject independent of the VP com-
plement, as given in the following lexical entry:
142
(55)
_
_
eager
SYN| VAL
_
_
SPR NP
i
COMPS
_
VP
_
_
VFORM inf
IND s1
_
_
_
_
_
SEM
_
_
IND s0
RELS
_
_
_
PRED eager
EXPERIENCER i
SIT s1
_
_
_
_
_
_
_
This then means that (54a) and (54b) must differ in that in the former, it is the dentist who is
eager to perform the action denoted by the VP complement, whereas in the latter, it is Pat who
is eager.
7.5.3 Subject vs. Object Control Verbs
Consider nally the following two examples:
(56) a. They persuaded me to leave.
b. They promised me to leave.
Both persuaded and promised are control verbs since their object is assigned a semantic role
(and so is their subject). This in turn means that their object cannot be an expletive:
(57) a. *They persuaded it to rain.
b. *They promised it to rain.
However, the two are different with respect to the controller of the innitival VP. Consider who
is understood as the unexpressed subject of the innitival verb here. In (56a), it is the object
me which semantically functions as the subject of the innitival VP. Yet, in (56b), it is the
subject they who will do the action of leaving. Due to this fact, verbs like promise are known as
subject control verbs, whereas those like persuade are object control verbs. This difference
is straighforwardly represented in their lexical entries:
(58)
_
_
persuade
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
j
_
_
_
_
_
143
_
_
promise
SPR NP
i
COMPS
_
NP
j
, VP
_
_
VFORM inf
SPR NP
i
IND s1
_
_
_
_
_
Based on world knowledge, we know that when one promises someone to do something, this
means that the person who makes the promise will do the action. Meanwhile, when one per-
suades someone to do something, the person who is persuaded will do the action. The lexical
entries here reect this knowledge of the relations in the world.
In sum, the properties of rasing and control verbs presented here can be summarized as
follows:
.
Unlike control predicates, raising predicates are unusual in that they do not assign a semantic
role to their subject or object. The absence of a semantic role accounts for the possibility of
expletives it or there or parts of idioms as subject or object with raising predicates, and not
with control predicates.
.
With control predicates, the VP complements unexpressed subject is coindexed with one
of the syntactic dependents. With raising predicates, the entire syntactic-semantic value of
the subject of the innitival VP is structure-shared with that of one of the dependents of
the predicate. This ensures that whatever category is required by the raising predicates VP
complement is the raising predicates subject (or object). Notice that even non-NPs can be
subject in certain kinds of example (see (59)).
(59) a. Under the bed is a fun place to hide.
b. Under the bed seems to be a fun place to hide.
c. *Under the bed wants to be a fun place to hide. (want is a control verb)
144
7.6 Exercises
1. Draw trees for the following sentences and provide the lexical entries for the italicized
verb.
(i) a. Kim may have admitted to let Mary mow the lawn.
b. Gregory appears to have wanted to be loyal to the company.
c. Jones would prefer for it to be clear to Barry that the city plans to sue him.
d. John continues to avoid the conict.
e. The captain ordered the troops to proceed.
f. He coaxed his brother to give him the candy.
g. Frank hopes to persuade Harry to make the cook wash the dishes.
2. Explain why the following sentences are ungrammatical, based on the lexical entries of
the predicates in the following sentences.
(i) a. *John seems to rain.
b. *John is likely to appear that he will win the game.
c. *Beth tried for Bill to ask a question.
d. *He believed there to be likely that he won the game.
e. *It is likely to seem to be arrogant.
f. *Sandy appears that Kim is happy.
g. *Dana would be unlikely for Pat to be called upon.
h. *Robin is nothing in the box.
i. *It said that Kim was happy.
j. *There preferred for Sandy to get the job.
3. Decide whether the following lexical elements are raising or control verbs. In particular,
use it, there, and an idiom expression to decide whether an expression is a raising or
control predicate:
(i) certain, anxious, lucky, sure, apt, liable, bound, careful, reluctant
(ii) tend, decide, manage, fail, happen, begin, hope, intend, refuse
4. Discuss the similarities and differences among the following three sentences. In so doing,
please use it, there, and an idiomatic expression. Also see what is the controller of the
innitival VP in each case.
(i) a. Pat expected Leslie to be aggressive.
b. Pat persuaded Leslie to be aggressive.
c. Pat promised Leslie to be aggressive.
5. Consider the following data and discuss what can be the antecedent of her and herself .
(i) a. Kevin urged Anne to be loyal to her.
b. Kevin urged Anne to be loyal to herself.
Also observe the following data and discuss the binding conditions of ourselves and us
here. In particular, see if the value of the ARG-ST can tell us anything about this relation.
145
(ii) a. We
i
expect the dentist to examine us
i
.
b. *We
i
expect the dentist to examine ourselves
i
.
c. We expect them to examine themselves.
d. *We expect them
i
to examine them
i
.
(iii) a. We
i
persuaded the dentist to examine us
i
.
b. *We
i
persuaded the dentist to examine ourselves
i
.
c. We persuaded them
i
to examine themselves
i
.
d. *We persuaded them
i
to examine them
i
.
6. Read the following passage and provide tree structures for the bracketed sentences and
lexical entries for the italicized words.
(i) I vividly recall a story I read in a sales book about a salesperson who were
so persistent that the customer nally, physically threw him out of the house.
As the salesperson lays there on the ground, hurting all over from the wounds
hes gotten after being thrown out, with great difculty he asks the angry
person hes tried to sell to: Will you now buy my product? The customer
is so surprised by the salespersons persistence that he exhausted and out of
pure compassion nally says, Okay then! But only on the condition that [you
promise never to bother me again].
[Some people still seem to have the same idea]. A tough job! If [youve ever
tried to persuade other people to buy your product or service], you also know
that this can be one of the most discouraging and difcult things to try to do
as a business owner. In fact, this way of trying to get business by trying to
persuade other people, is one of the factors that causes most business owners
to dislike, yes even hate, the process of marketing and selling. [Its very tough
to try to convince other people to buy fromyou] - especially if its against their
will. After all, if [you try to persuade someone to buy from you], you try to
cause that person to do something. And usually theres always some kind of
pressure involved in this process.
146
8
Auxiliary Constructions
8.1 Basic Issues
The English auxiliary system involves a relatively small number of elements interacting with
each other in complicated and intriguing ways. This has been one of the main reasons for mak-
ing the system the most extensively analyzed empirical domains in the literature on generative
syntax.
Ontological Issues: One of the main issues in the study of English auxiliary system concerns
ontological issues: is it necessary to posit auxiliary as an independent part of speech or not?
Auxiliary verbs can be generally classied as follows:
.
modal auxiliary verbs such as will, shall, may, etc.: have only nite forms
.
have/be: have both nite & nonnite forms
.
do: has a nite form only with vacuous semantic meaning
.
to: has a nonnite form only with vacuous semantic meaning
Such auxiliary verbs behave differently from main verbs in various respects. There have been
arguments to treat these auxiliary verbs just as carrying the lexical category V, though they
are different in terms of syntactic distribution and semantic contribution. For example, both
auxiliary and main verbs behave alike in carrying tense information and undergoing the same
syntactic operations such as gapping, as shown in (1):
(1) a. John drank water and Bill wine.
b. John may drink water, but Bill drink beer.
Such phenomena provide apparent stumbling blocks to assign a different lexical category to the
English auxiliary verbs from the main verbs.
Distinction between auxiliary and main verbs: Another important issue that raises in the
study of the English auxiliary system is the question of which words function as auxiliary verbs
and how we can differentiate the two. Most reliable criteria for auxiliaryhood seems to lie in
syntactic phenomena such as negation, inversion, contraction, and ellipsis (henceforth, NICE):
147
1. Negation: Only auxiliary verbs can be followed by not as a sentential negation (have and
be too).
(2) a. Tom will not leave.
b. *Tom kicked not a ball.
2. Inversion: Only auxiliary verbs can undergo the subject-auxiliary inversion.
(3) a. Will Tom leave the party now?
b. *Left Tom the party already?
3. Contraction: Only auxiliary verbs can have contracted forms with the sufx nt.
(4) a. John couldnt leave the party.
b. *John leftnt the party early.
4. Ellipsis: The complement of an auxiliary verb, but not of a main verb can be elided.
(5) a. If anybody is spoiling the children, John is .
b. *If anybody keeps spoiling the children, John keeps .
In addition to these NICE properties, tag questions can be another criterion: an auxiliary verb
can appear in the tag of tag questions, but not a main verb:
(6) a. You should leave, shouldnt you?
b. *You didnt leave, left you?
The position of adverbs or oating quantiers can also be adopted in differentiating auxiliary
verbs from main verbs. The difference can be easily observed from the following contrast:
(7) a. She would never believe that story.
b. *She believed never his story.
(8) a. The boys will all be there.
b. *Our team played all well.
Adverbs such as never and oating quantiers such as all can follow an auxiliary verb, but not
a main verb.
Ordering Restrictions: The third main issue centers on how to capture the ordering re-
strictions among auxiliary elements. Auxiliary verbs are subject to restrictions that limit the
sequences in which they can occur and the forms with which they can combine. Observe the
following contrast:
(9) a. The children will have been being seen.
b. He must have been being interrogated by the police at that very moment.
(10) a. *The house is been remodelling.
b. *Margret has had already left.
c. *He has will seeing his children.
148
d. *He has been must being interrogated by the police at that very moment.
As can be observed here, when we have more than two auxiliary verbs, they must come in a
certain order. In addition, each auxiliary verb requires that the immediately following one be in
a particular morphological form.
In the study of the English auxiliary system, we thus need to address the following issues at
least:
.
Should we posit an auxiliary category?
.
How can we distinguish main verbs from auxiliary verbs?
.
How can we account for phenomena (such as NICE) that are sensitive to the presence of an
auxiliary verb?
.
How can we capture the ordering and co-occurrence restrictions among auxiliary verbs?
This chapter providers answers to these questions.
8.2 Transformational Analyses
The seminal work on these three issues is that of Chomsky (1957). His analysis, introducing the
rule in (11), directly stipulates the ordering relations among auxiliary verbs:
(11) Aux Tense (Modal) (have + en) (be + ing)
The PS rule in (11) would generate sentences with or without auxiliary verbs as in (12):
(12) a. Mary solved the problem.
b. Mary would solve the problem.
c. Mary was solving the problem.
For example, see the following structure generating sentences like (12a):
(13)
S
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
NP
AUX
.
.
.
.
.
.
.
.
.
.
.
.
.
VP
Adv VP
Past will V NP
_
must
HEAD
_
_
POS verb
VFORM plain
AUX +
_
_
VAL
_
_
SPR 1 NP
COMPS
_
VP
_
_
VFORM bse
SPR 1 NP
_
_
_
_
_
_
_
In the lexical information given here, we need to notice at least three things: rst, the auxiliary
verbs have the head feature AUX, different from the main verb. This feature thus will thus
distinguish auxiliary verbs from main verbs. In addition, it tells us that the modal verb selects a
base VP as its complement. This subcategorization information will rule out examples like the
following:
(22) a. *Kim must
V P[fin]
[bakes a cake].
b. *Kim must
V P[fin]
[baked a cake].
c. *Kim must
V P[fin]
[will bake a cake].
The possible and impossible structures can be more clearly represented in the tree format:
151
(23)
*VP
V[AUX +]
VP[n]
_
HEAD
_
_
POS verb
AUX +
_
_
VAL
_
_
SPR 1 NP
COMPS
_
XP
_
_
PRD +
SPR 1
_
_
_
_
_
_
_
All the three bes thus bear the feature AUX and select a predicative phrase whose subject is
identical with its own subject. This in turn means that like modals, it is also a raising verb. The
main difference just lies in the XPs VFORM value:
77
(32) a. copula be:
_
COMPS XP
_
b. passive be:
_
COMPS VP[VFORM pass]
_
c. progressive be:
_
COMPS VP[VFORM prp]
_
As given here, the copula be needs no further specication: any phrase that can function as a
predicate can be its COMPS. The passive be requires its complement to be a VP[pass], and the
progressive be asks its complement to be a VP[prp]. Given these common property as well as
differences, we then can easily account for the following:
76
XP is a variable encompassing NP, VP, AP, and PP.
77
The VFORM value pass(ive) is derived from psp (past participle). See Chapter 9.
153
(33) a. John is [
AP
happy about the outcome].
b. The children are [
V P[pass]
seen in the yard].
c. John was [
V P[prp]
seeing his children].
The grammar will not allow examples like the following:
(34) a. *John was knowing the answer.
b. *This is belonging to John.
c. *John is having sung a song.
The complement of be needs to be at least one of the three: XP[PRD +], VP[pass], or VP[prp].
However, knowing, belonging, and having here cannot be progressive, they cannot occur here.
78
Like the auxiliary be, have also behaves just like an auxiliary verb:
(35) a. John has not sung a song.
b. Has John sung a song?
c. John hasnt been singing a song.
d. John has sung a song and Mary has , too.
Given these observations, we can posit the following information as the lexical entry for the
aspectual have:
(36)
_
_
have
HEAD
_
_
POS verb
AUX +
_
_
SPR 1
COMPS
_
VP
_
_
VFORM psp
SPR 1
_
_
_
_
_
The interaction of subcategorization and morphosyntactic information is enough to predict the
ordering restrictions among modals:
(37) a. He has [seen his children].
b. He will [have [been[seeing his children]]].
c. He must [have [been [being interrogated by the police at that very moment]]].
(38) a. *Americans have [paying income tax ever since 1913]
b. *George has [went to America].
(38a) is ungrammatical since have requires a perfect participle VP. (38b) is out since the fol-
lowing VP is nite.
79
78
This does not mean that there is no progressive having at all. As in John is having lunch now, the progressive form
is determined by meaning.
79
One thing to note here is the copula be and the auxiliary have used as a main verb in British English:
154
8.3.3 Periphrastic do
The so-called dummy do has several similar as well as different properties compared with other
auxiliaries:
First of all, the periphrastic do also exhibits the NICE properties like other auxiliaries:
(39) a. John does not leave the town.
b. In no other circumstances does John drink alcohol.
c. They dont leave the town.
d. Jane likes the apples, but Mary doesnt .
Like other modals, do does not appear in innitive clauses.
(40) a. *They expected us to do leave him.
b. *They expected us to can leave him.
There are also some properties that distinguish do from other auxiliaries. First, unlike other
auxiliaries, do appears neither before nor after any other auxiliary:
(41) a. *He does be leaving.
b. *He does have been eating.
c. *They will do come.
Second, the verb do has no obvious intrinsic meaning to speak of. Except for the grammatical
information such as tense and agreement, it does not carry any semantic value. Third, if do
itself is positive, then do needs to be emphatic (stressed). But in negative sentences, no such
requirement exists.
(42) a. *John does leave.
b. John DOES leave.
(43) a. John did not come.
b. John DID not come.
The most economic way of representing these lexical properties seems to assume that the
periphrastic do has the lexical entry given in (44).
(i) a. John is a student.
b. John has not enough money.
(ii) a. Is John a student?
b. Has John enough money? (British English)
As shown in the subject-aux inversion, they both have the AUX feature. One can questions where is the main verb in
both cases? Are there any main verbs here? The analysis presented here implies that these are all verbs with the feature
AUX: Nothing in the grammar requires the presence of a so-called main verb.
155
(44)
_
_
do
HEAD
_
_
POS verb
AUX +
VFORM n
_
_
VAL
_
_
SPR 1 NP
COMPS
_
2 VP
_
_
AUX
VFORM bse
SPR 1
_
_
_
_
_
_
_
Like other auxiliaries including modals, do is specied to be [+AUX]. The feature specication
[+AUX] ensures that like other auxiliary elements, do is also sensitive to negation, inversion,
contraction, and ellipsis (NICE properties). Further, like other auxiliaries, do selects a subject
NP and a VP complement whose unrealized subject is structure-shared with its subject ( 1 ).
Treating do as a raising verb like other English auxiliaries is based on typical properties of
raising verbs, one of which is that raising verbs allow expletives as their subject:
(45) a. John may leave.
b. It may rain.
c. *John may rain.
(46) a. John did not leave.
b. It did not rain.
c. *John did not rain.
The [AUX +] specication and raising-verb treatment of do enable us to capture its similar-
ities with other auxiliaries and modals. But its differences stem from the lexical specications
on feature values for HEAD|POS and its complement VP.
Unlike auxiliaries have and be, do is specied to be n(ite). This property then accounts for
why no auxiliary element can precede do.
80
(47) a. He might [have left].
b. *He might [do leave].
The rst requirement on the complement VP of the auxiliary do is [VFORM bse]. This fea-
ture specication blocks modals from heading the VP following do. Since modals are specied
to be [n], the ungrammaticality of (48) is a natural expectation.
(48) a. *He does can leave here.
b. *He does may leave here.
The lexical entry further species that its complement VP be [AUX ]. This requirement will
correctly predict the ungrammaticality of examples in (49) and (50).
80
Like do, modals also do not have non-nite forms.
156
(49) a.*Jim [DOES [have supported the theory]].
b.*The proposal [DID [be endorsed by Clinton]].
(50) a. *I [do [not [have sung]]].
b. *I [do [not [be happy]]].
In (49) and (50), the VPs following the auxiliary do, stressed or not, bear the feature [+AUX]
inherited from the auxiliaries have and be. This explains their ungrammaticality.
81
8.3.4 Innitival Clause Marker to
The auxiliary verbs to and do, in addition to differing by one phonological feature, voicing,
differ in one small way: do appears only in nite contexts, and to only in non-nite contexts.
(51) a. *John believed Kim to do leave here.
b. John believes Kim to leave here.
Other than that, they share the property that they obligatorily take bare verbal complements
(hence not modals):
(52) a. *John believed Kim to leaving here.
b. *John did not leaving here.
In terms of NICE properties, to observes the VP ellipsis criterion:
(53) a. Tom wanted to go home, but Peter didnt want to .
b. Lee voted for Bill because his father told him to .
These properties mean that to would have a lexical entry like the following:
(54)
_
_
HEAD
_
_
POS verb
AUX +
VFORM inf
_
_
SPR
_
1 NP
_
COMPS
_
VP
_
_
VFORM bse
SPR 1
_
_
_
_
_
The lexical entry of to is thus similar to that of do, in that they both are raising verbs.
81
But note that there are differences between do and dont in imperatives and in non-imperatives. One telling differ-
ence is that do in imperatives can occur before another auxiliary like be and have.
(i) a. Do be honest!
b. Dont be silly!
do and dont in imperatives also have one distinct property: only dont allows the subject you to follow. Their properties
indicate that they have different lexical information from those in non-imperatives.
157
8.4 Explaining the NICE Properties
Let us here then see how we can account for the NICE properties that are highly sensitive to the
presence of auxiliary verbs.
8.4.1 Auxiliaries with Negation
The English negator not leads a double life: one as a nonnite VP modier when it is constituent
negation and the other as a complement of a nite auxiliary verb when it is sentential negation.
Constituent Negation: The properties of not as a nonnite VP modier can be supported
from its similarities with adverbs such as never in nonnite clauses as given in (55):
(55) a. Kim regrets [never/not [having seen the movie]].
b. We asked him [never/not [to try to call us again]].
c. Duty made them [never/not [miss the weekly meeting]].
If we assume that not modies a nonnite VP, we can predict its various positional possibilities
in nonnite clauses, as represented the following:
(56)
_
_
not
HEAD
_
_
POS adv
NEG +
MOD
_
VP[VFORM nonn]
_
_
_
_
_
This means that the adverb not modies any nonnite VP, as represented in the following:
(57) Constituent Negation:
VP[nonn]
.
.
.
.
.
.
.
.
.
.
.
.
.
Adv
_
MOD 1 VP[nonn]
_
1 VP[nonn]
not
For example, in all the good examples in (58) and (59), not simply modies a nonnite VP. But
in the bad examples, this nonnite VP modifying lexical constraint is violated.
(58) a. [Not [speaking English]] is a disadvantage.
b. *[Speaking not English] is a disadvantage.
c. *Lee likes not Kim.
(59) a. Lee is believed [not
V P[inf]
[to like Kim]].
b. Lee is believed to [not
V P[inf]
[like Kim].
158
c. *Lee is believed [to
V P[inf]
[like not Kim]].
Sentential Negation: But in nite clauses, it is well-known that not has restricted distribu-
tions:
(60) a. Lee never/ *not left.
b. Lee will not leave.
This is one clear difference between never and not. not can modify a nonnite VP, but not a
nite VP, as further shown by the following contrast:
(61) a. John could [not [leave the town]].
b. John wants not to leave the town.
(62) a. *John not left the town.
b. *John not could leave the town.
Another difference between never and not comes from VP ellipsis. Observe the following:
(63) a. Mary sang a song, but Lee never did .
b. *Mary sang a song, but Lee did never .
c. Mary sang a song, but Lee did not .
The data here indicate that not behaves differently form adverbs like never and always in the
nite clauses even though it behaves alike in nonnite clauses.
One possible piece of evidence to differentiate two types of not may come from scope pos-
sibilities in an example like (64) (cf. Warner 2000).
(64) The president could not approve the bill.
The negation here could have the two different scope readings as given in (65).
(65) a. It would not be possible for the president to approve the bill.
b. It would be possible for the president not to approve the bill.
The most economical way to differentiate sentential negation from constituent negation
seems to assume that the sentential negation is a syntactic complement of a nite auxiliary
verb (cf. Kim and Sag 1995, 2002). That is, we can assume that when not is used as a sentential
negation, it can be selected by the nite auxiliary verb through the lexical rule:
(66) Negative Auxiliary Verb Lexical Rule:
_
_
HEAD
_
_
AUX +
VFORM n
_
_
COMPS 1
_
_
_
_
HEAD
_
_
AUX +
VFORM n
NEG +
_
_
COMPS Adv[NEG +], 1
_
_
This lexical rule means that an auxiliary verb basically selects a complement ( 1 ), but can also
be realized as a sentential negation (with [NEG +]) that selects an [NEG +] element (not) as
159
an additional complement. This view of negation will then generate the following structure for
sentential negation.
(67)
VP
_
_
VFORM n
AUX +
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
VFORM n
AUX +
COMPS 2 [NEG +], 3
_
_
2 Adv
3 VP[bse]
could
Adv
_
MOD 1 VP
_ 1 VP[VFORM bse]
_
VFORM n
NEG +
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
VFORM n
AUX +
NEG +
COMPS 2 [NEG +], 3
_
_
2 Adv 3 VP[VFORM bse]
V S
.
.
.
.
.
.
.
Are NP
VP
.
.
.
.
.
.
.
you V VP
However, there are certain exceptions that present problems for the analysis of inverted inter-
rogatives via movement transformation. Observe the following contrast:
(72) a. I shall go downtown.
b. Shall I go downtown?
Here there is a semantic difference between the auxiliary verb shall in (72)a and the one in
(72)b: the former conveys futurity whereas the latter has a deontic sense.
Further, there are inected forms that only occur in inversion constructions, e.g. the rst
person singular negative contracted form of the copula illustrated in (73):
(73) a. *I arent going.
b. Arent I going?
To state such lexical idiosyncrasies, movement approaches run into difculties.
Notice that English has various Subject-Aux inversion constructions:
83
(74) a. Wish: May she live forever!
b. Matrix Polar Interrogative: Boy, was I stupid!
c. Negative Imperative: Dont you even touch that!
d. Subjunctive: Had they been here now, we wouldnt have this problem.
e. Exclamative: Am I tired!
Each of these constructions has its own constraints that can hardly be predicted from other
constructions. For example, in wish constructions, only the modal auxiliary may is possible.
In negative imperative, only dont allows the subject to follow. These idiosyncratic properties
support a non-movement approach. One effective way is to assume the following Subject-Aux
Inversion (SAI) Rule:
84
(75) Subject-Aux Inversion Rule:
83
See Fillmore (1999) for detailed discussion.
84
Another option is to assume a lexical rule that turns a nite auxiliary verb into an inverted nite one selecting the
subject as its complement as well. See Borsley (1989a,b).
162
S
_
SPR
_
H
_
_
HEAD
_
_
INV +
AUX +
_
_
SPR A
COMPS B
_
_
, A, B
This rule licenses an an inverted, nite, auxiliary verb to combine with its subject (the SPR
value A) and complements (the COMPS value B ), forming a well-formed subject-auxiliary
inverted phrase. An inverted nite auxiliary verb will have the following lexical information:
(76)
_
_
will
HEAD
_
_
AUX +
INV +
_
_
SPR 1 NP
COMPS 2 VP[bse]
ARG-ST 1 , 2
_
_
This lexical entry then can license the structure like the following:
(77)
S
_
_
HEAD
_
_
AUX +
INV +
_
_
SRR
COMPS
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
HEAD
_
_
AUX +
INV +
_
_
SRR 1 NP
COMPS 2 VP
_
_
1 NP 2 VP[VFORM bse]
_
PHON 1
HEAD
_
_
POS verb
VFORM n
AUX +
_
_
_
_
PHON 1 + nt
HEAD
_
_
VFORM n
AUX +
NEG +
_
_
_
_
This means that words like can will turn into cant with the addition of the NEG feature. As we
have seen earlier, the head feature NEG will play an important role in forming tag questions:
(82) a. They can do it, cant they?
b. They cant do it, can they?
c. *They cant do it, cant they?
The tagged part needs to have the oppositive value of the NEG feature of the main sentence.
8.4.4 Auxiliaries with Ellipsis
The standard generalization of VPE is that it is possible only after an auxiliary verb as shown
in the contrast (83) and (84).
(83) a. Kim can dance, and Sandy can , too.
b. Kim has danced, and Sandy has , too.
c. Kim was dancing, and Sandy was , too.
164
(84) a. *Kim considered joining the navy, but I never considered .
b. *Kim got arrested by the CIA, and Sandy got , also.
c. *Kim wanted to go and Sandy wanted , too.
The data means that the VP complement of an auxiliary can undergo VP ellipsis as long as the
context provides its interpretation.
This generalization can be succinctly stated in the form of lexical rule:
(85) VP Ellipsis Rule:
_
_
HEAD| AUX +
COMPS XP
_
_
_
_
HEAD| AUX +
COMPS
_
_
Given this lexical rule, the canonical auxiliary verb can will be changed into a VPE can as
following:
(86)
_
_
can
SPR 1 NP
COMPS 2 VP[bse]
ARG-ST 1 , 2
_
_
can
SPR 1
COMPS
ARG-ST 1 , 2
_
_
Notice here that even though the VP complement in the output is elided in the output, the ARG-
ST is intact. This simple lexical rule can explain all the following data:
(87) Kim must have been dancing and
_
_
a. Sandy must have been , too.
b. Sandy must have , too.
c. Sandy must , too.
_
_
The elided VP is all the complement of the auxiliary verbs been, have, and must.
The analysis also immediately predicts the behavior VPE after the innitival marker to that
has been taken to be an auxiliary verb, too:
(88) a. Tom wanted to go home, but Peter didnt want to .
b. Lee voted for Bill because his father told him to .
(89) a. Because John persuaded Sally to , he didnt have to talk to the reporters.
b. Mary likes to tour art galleries, but Bill hates to .
As we have seen earlier, to is a type of auxiliary verb. This means that its complement can be
freely elided.
The present system also can account for the contrast we have seen earlier in (63). A similar
contrast can be found in the following:
(90) a. *Mary sang a song, but Lee could never .
b. Mary sang a song, but Lee could not .
The negator not in (90b) is the complement of the nite auxiliary verb could. This means we
can apply the VPE lexical rule to could as shown in the following:
165
(91)
_
_
could
SPR 1 NP
COMPS 3 VP[bse]
ARG-ST 1 , 2 [NEG +], 3
_
_
can
SPR 1
COMPS 2 ,
ARG-ST 1 , 2 , 3
_
_
As seen from the output, the VP complement of the auxiliary verb could is not realized as a
COMP element. The lexical information (91b) would then project the syntactic structure in
(92):
(92)
VP
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
AUX +
SPR 1
COMPS 2 Adv[NEG +]
ARG-ST 1 , 2 , 3 VP[bse]
_
_
2 Adv
_
NEG +
_
could not
As represented here, the auxiliary verb could forms a well-formed head-complement structure
with not. But as for never, consider the following structure:
(93)
VP
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
V[AUX +] *VP
has Adv[MOD VP]
never
The adverb never modies a VP through the feature MOD. The head feature MOD guarantees
the fact that the adverb selects the head VP it modies. The absence of this VP then means that
there is no VP the adverb can modify. And this results in an ill-formed structure: no well-formed
phrasal conditions in our system renders such a structure acceptable.
166
8.5 Exercises
1. Each of the following sentences contains something (in the parenthesis) that we might
want to call an auxiliary. In each case, construct relevant examples that will clarify
whether it actually is one.
(i) a. John got sent to prison. (got)
b. He ought to leave his luggage here. (ought)
c. They neednt sit this exam. (need)
d. You better not leave it here. (better)
e. He dared not argue against his parents.
2. Provide an analysis of the grammaticality/ungrammaticality of the following examples.
Hint: give a tree structure and then lexical entries.
(i) a. Ann may spend/*spending/*spends/*spent her vacation in Italy.
b. It has rained/*raining/*rains/*rain every day for the last week.
c. Togalog is spoken/*speak/* speaks/*spoke in Philiphines.
d. The roof is leaking/*leaked/*leaks/*leak.
(ii) a. *Americans have musted pay income tax ever since 1913.
b. *George is having lived in Toledo for thirty years.
c. *The house is been remodeling.
d. *Margaret has had already left.
e. *A medal was been given to the mayor by the sewer commissioner.
(iii) a. Sam may have been being interrogated by the FBI.
b. *Sam may have been being interrogating by the FBI.
c. *Sam may be had been interrogating by the FBI.
3. Provide tree structures for the following sentences. In addition, mark the SPR and
COMPS value on each node and see what kind of grammar rules licenses each phrase:
(i) a. John is believed to have left the town.
b. There may have occurred a disaster.
c. John has sold his car and bought a bicycle.
4. Observe the following data, and explain how we can generate the questions in the fol-
lowing. In so doing, please ensure what kind of lexical rules have been used.
(i) a. I dont think we can trust voting computers. [Can we trust them?]
b. An ape can do this. [Can we not do this?]
In addition, also consider the following examples where we have elliptical parts. Can our
system sketched in this chapter also account for such examples? If it can, explain how. If
it cannot, state why not.
(ii) a. I dont think we can trust voting computers. [Can we?]
b. An ape can do this. [Can we not?]
5. English allows so called negative inversions as seen from the following contrast:
167
(i) a. There was hardly any rain falling.
b. I did little know that more trouble was just around the corner.
c. I have never been spoken to so rudely!
(ii) a. [ Hardly] was there any rain falling.
b. [Little] did I know that more trouble was just around the corner.
c. [Never] have I been spoken to so rudely!
Draw tree structures for the sentences (ii) and provide the lexical entries for hardly, little
and never. In addition, think of how your analysis can account for the unacceptable
examples in the following:
(iii) a. As a statesman, he scarcely could do anything worth mentioning.
b. As a statesman, scarcely could he do anything worth mentioning.
c. *As a statesman, scarcely he could do anything worth mentioning.
d. *As a statesman, he scarcely couldnt do anything worth mentioning.
6. Read the following passages and anlayze the bracketed parts or sentences as far as you
can.
(i) [Evolution by natural selection is not by any means the only example of a
spontaneous natural order]. Even in human affairs, where people might think
that everything that happens is planned and intended by someone, there are
many examples of spontaneous order emerging without either plan, inten-
tion, or purpose. [One excellent case is in the structure of human languages].
Grammar, although it can be consciously described and is often the object
of deliberate attempts at control, usually escapes such control and emerges
in new forms that no one ever anticipated and that no one, if [they had been
aware of it], probably would have allowed. A nice example is the auxiliary
verb system in English.
Although French and German both have verb systems that extensively use
words like have, be, will, should, can, etc. to modify main verbs, [English
takes this to extremes]. An English sentence like, I am walking home, ex-
presses a present tense and a progressive aspect (where the action is on-going)
by using the verb be as an auxiliary, while an English sentence like, I have
walked home, expresses a perfect aspect (where the action has just been com-
pleted), which is also a present tense, by using the verb have as an auxiliary.
[French and German use have to express a simple past tense and do not ex-
press true progressive or perfect aspects at all].
[This expanded role for auxiliaries in English has resulted in some curious
rules]. One is that when a sentence is to be negated, the word not must fol-
low, not the main verb (as used to be the case), but the auxiliary. This rule
creates an awkward dilemma in the occasional instance when [the sentence
168
to be negated actually doesnt have an auxiliary verb]. Thus, if I wish to deny
the sentence, I walked home, I must add an entirely meaningless auxiliary
from the verb do just to stand as the prop for the word not. The result is
the sentence, I didnt walk home. Now, do and did are often added to show
emphasis, but in those cases they are spoken with emphasis. Thus there is
a difference between saying I didnt walk home and saying I DIDNT walk
home. The latter sentence expresses emphasis, but in the former sentence the
verb did expresses nothing at all; it is merely there to hang the not on. If we
tried to say, I walked not home, [this would have an unacceptably odd sound
to it]. It would, indeed, sound archaic. English literature is full of such ar-
chaisms, since putting not after the main verb was still good usage in the time
of Shakespeare and a century or more later.
[I have elaborated this example because most native speakers of English are
not even aware of the rules of the auxiliary verb system]. They just use the
rules constantly without thinking about them. Indeed, [the rules came into
existence without anyone thinking about them]. Nothing could be more ob-
vious than that there never was anyone who planned or intended to produce
the forms of the auxiliary verb system that we see in English today. [Neither
King James nor Samuel Johnson, much as they contributed to the xing of
English usage by the one commissioning a translation of the Bible and by the
other compiling a dictionary of English, ever sat down] and thought, Oh!
Lets have an elaborate auxiliary verb system in English, expressing things
like progressive and perfect aspects! Indeed, no languages they would have
looked to for precedents, German, French, Latin, or Greek, even possessed a
distinct progressive aspect as English does possess it; and no other language
with a developed aspect system, like Russian, Arabic, or Hebrew, combines
this with a complete and distinct tense system as does English. So the En-
glish language, regardless of anyones beliefs, desires, or purposes, actually
evolved structures, in a spontaneous and unintended way, that mostly do not
exist in languages related to it either genetically, by cultural association, or
by artice. [The grammar of all natural languages evolves in a spontaneous
and self-organizing way], often in the very teeth of attempts to stop change
and freeze grammatical forms.
85
85
From Creationism & Darwinism, Politics & Economics by Kelley L. Ross.
169
9
Passive Constructions
9.1 Introduction
One important aspect in doing syntax is capturing systematic relations between related con-
structions. For example, the following two sentences are similar in their meanings:
(1) a. One of Koreas most famous poets wrote these lines.
b. These lines were written by one of Koreas most famous poets.
We recognize (1b) as the passive counterpart of the active sentence (1a). These two sentences
are truth-conditionally no different: they both describe the event of writing the lines by one
Korean poet. In terms of semantic roles, the one who wrote the lines and the things that he or
she wrote are all identical. The only difference is the grammatical functions: In the active voice
(1a), one of Koreas most famous poets is the subject, whereas in the passive voice (1b), these
lines is the subject.
Observing these relationships, the question that follows is why do we use different voices in
expressing or describing the same situation or proposition? It is generally accepted that passive
sentences are used for certain discourse reasons. For example, when it is more important to
draw our attention to the person or thing acted upon, we use passive. Compare the following:
(2) a. Somebody apparently struck the unidentied victim during the early morning
hours.
b. The unidentied victim was apparently struck during the early morning hours.
We can easily notice here that the passive in (2b) assigns more attention to the victim than the
active in (2a). In addition, when the actor in the situation is not important, it is recommended to
use the passive voice:
(3) The aurora borealis can be observed in the early morning hours.
Similarly, we use the passive voice in formal, scientic, or technical writing or reports to place
an emphasis or an objective view in the process or principle being described. For example,
compare the following pair:
(4) a. I poured 20cc of acid into the beaker.
171
b. 20cc of acid was poured into the beaker.
It is clear that unlike the active sentence (4a), the passive sentence (4b) assigns a more objective
perspective to the meaning of the sentence.
In this chapter, leaving aside these discourse properties of passive constructions, we will look
into the syntactic and semantic relationships between active and passive and the properties of
different passive constructions.
9.2 Relationships between Active and Passive
Consider the two canonical active and passive counterpart sentences:
(5) a. The executive committee approved the new policy.
b. The new policy was approved by the executive committee.
Grammatical Functions and Subcategorization: As observe, one of the main differences
between the active and the passive, we rst is that the passive sentence promotes the active
object into the passive subject while it demotes the active subject into an optional PP (headed
by by). We can observe that the complement of the main verb is missing, and the subject of the
sentence has the main properties of this missing element. For example, the active transitive verb
taken or chosen must have its object:
(6) a. John has taken Bill to the library.
b. John has chosen Bill for the position.
(7) a.*John has taken to the library.
b.*John has chosen for the position.
However, in the passive, the object NP must be absent. That is, it must not appear right after
the passive verb:
(8) a.*John has been [taken Bill to the library].
b.*John has been [chosen Bill for the position].
(9) a. John has been [taken to the library].
b. John has been [chosen for the position].
The absence of the object in the passive is due to the fact that the object of the verb is promoted
to the subject in the passive.
The other subcategorization requirement stays unchanged. For example, the active handed
requires an NP and a PP[to] as its complements, and the passive handed still requires the PP as
its complement:
(10) a. Pat handed Chris a book.
b.*Pat handed Chris.
c.*Pat handed a book.
172
(11) a. Chris was handed a book (by Pat).
b.*Chris was handed (by Pat).
Morpho-syntactic changes: In addition to such changes in the grammatical functions, the
passive also adds the auxiliary verb be and changes the verb form of the active main verb into
the passive. Such changes happen even when we have more than one auxiliary verb:
(12) a. John drove the car. The car was driven.
b. John was driving the car. The car was being driven.
c. John will drive the car. The car will be driven.
d. John has driven the car. The car has been driven.
e. John has been driving the car. The car has been being driven.
f. John will have been driving the car.
The car will have been being driven.
Raising Properties: The third important property, following from the fact that the active
object is promoted to the passive subject, is that the passive construction is similar to the raising
construction whose properties we have seen in Chapter 7. Whatever the active transitive verb
requires, it must be the subject of a passive. For example, if the postverbal constituent must be
an expletive, so must be the subject. Compare the following:
(13) a. They believe it/*Stephen to be easy to annoy Ben.
b. They believe there/Stephen to be a dragon in the wood.
(14) a. It/*Stephen is believed to be easy to annoy Fred
b. There/Stephen is believed to be a dragon in the wood.
Or if the postverbal constituent must be a clause, so must be the subject of the passive verb:
(15) a. Everyone believes/*kicks that he is a fool.
b. [That he is a fool] is believed/*kicked by everyone.
Finally, if the postverbal constituent can be understood as part of an idiom, so can be the subject:
(16) a. They believe the cat to be out of the bag.
b. The cat is believed to be out of the bag.
We thus can conclude that the subject of the passive corresponds to the object of the active.
Semantics: In terms of meaning, as noted before there is no change in the semantic role
assigned to the arguments. The difference is that the agent denoted by the active subject is
expressed as an optional oblique argument of the PP headed by the preposition by in the passive:
(17) a. Pat handed Chris a note.
b. Chris was handed a note (by Pat)
173
(18) a. TV puts ideas in childrens heads.
b. Ideas are put in childrens heads (by TV).
The observations we have made so far tell us that any grammar needs to capture the following
basic facts in passive:
.
Passive turns the active object into the passive subject;
.
Passive optionally allows the subject to turn into the object of a PP headed by the preposition
by;
.
Passive leaves the COMPS value unchanged (except the object promoted to the subject;
.
Passive makes the appropriate morphological change in the form of the main verb
.
Passive leaves the semantics unchanged
9.3 Three Approaches
There can be several ways to capture the systematic, syntactic and semantic relationships be-
tween active and passive. Given our discussion so far, one can rely on grammatical forms (NP,
VP, S, etc), or grammatical functions (SPR and COMPS), or semantic roles (agent, patient), and
so forth. In what follows, we will see that we need to refer to all these three notions for a proper
treatment of English passive constructions.
9.3.1 From Structural Description to Structural Change
The classical transformational grammar assumes the so-called Passive Formation Rule in terms
of structural descriptions (SD) and structural change (SC):
(19) Passive Formation Rule:
SD: X NP Y V NP Z
1
4
.
5
.
6
SC: 1 5 3 be 4+en 6 (by 2)
This rule means that if there is anything that ts the SD in (19), it will be changed into the given
SC: that is, if we have any string in the order ofX NP Y V NP Z (in which X, Y,
and Z are variables), the order can be changed into X NP Y be V+en Z by NP. For
example, consider one example:
174
(20)
Yesterday, the child really kicked a monkey in the street.
X NP Y V NP Z
1
4
.
5
6
1 5 3 be 4+en 6 (by 2)
Yesterday, a monkey really was kicked in the street (by the child)
As noted here, the main change that occurs in the SC is that the rst NP became an optional PP
whereas the second NP became the rst NP. The change also accompanies the addition of be
and the change of the verb form into the passive. Even though this SD-SC style rule can handle
basic facts for passivization, it leaves many questions unanswered. For example, this does not
answer questions like why we have such a SD-SC operation, whether this rule can cover all the
facts related to English passivization, and what triggers such a movement.
9.3.2 A Transformational Approach
A more elaborated approach is a transformational approach assuming a passive movement op-
eration represented in the following:
(21)
S
.
.
.
.
.
.
.
NP
VP
.
.
.
.
.
.
.
e V
VP
.
.
.
.
.
.
.
was V NP
deceived Hill
The operation basically moves the object Hill to the subject position. This kind of movement
analysis is based on the following three basic assumptions:
.
Move : Move anything anywhere (as long as the movement observes general principles)
.
Case Theory: Every NP needs to get Case (nominative (NOM) or accusative (ACC) ). The
subject gets NOM case from tense, and the object gets ACC from an active transitive verb.
86
.
A passive participle does not license ACC case.
These basic assumptions are taken to trigger the movement of Hill. That is, the NP Hill in
(21) does not have ACC case since the passive participle deceived cannot assign any case. This
86
In English, case appears only on pronouns: he is nominative, whereas him is accusative.
175
would violate the Case Theory requiring every NP to have a case. To salvage this case violation,
the NP must be moved to the subject position whose case is assigned by the tensed verb was.
This kind of movement or derivational analysis is an improvement, compared to the SD-SC
analysis. However, it still suffers from problems. For example, this kind of analysis relies on the
congurational structure only, it is hard to explain cases where we need to refer to grammatical
functions or semantic roles, which we will discuss in what follows.
9.3.3 A Lexicalist Approach
Unlike the movement (or transformational) approach that relies on the tree structure, we can
resort to the lexical properties of passive verbs. We can observe that there are many exceptions
to passives. For example, transitive verbs like resemble do not have any passive counterpart:
(22) a. John resembles his father.
b. *His father is resembled by John.
These transitive verbs meet the SD requirement or the tree structure in (21), but cannot be
passivized.
There are also verbs like rumor used only passive forms, as observed from the following
contrast:
87
(23) a. It is rumored that he is on his way out.
b. John is said to be rich.
c. He is reputed to be a good scholar.
(24) a.*Someone rumors that he is on his way out.
b.*They said him to be a good teacher.
c.*They reputed him to be a good scholar.
Neither the SD-SC approach nor the movement approach can predict such lexical idiosyn-
crasies.
One easiest way of capturing the fact that not all but a limited set of transitive verbs can
undergo a passive lexical rule is to adopt a lexical rule like the following:
(25) Simple Passive Lexical Rule (to be revised):
_
_
tran-v
HEAD| POS verb
SPR 1
COMPS 2 ,...
_
_
pass-tran-v
HEAD
_
_
POS verb
VFORM pass
_
_
SPR 2
COMPS
_
...,
_
PP
_
by 1
_
_
_
_
_
87
There exist more verbs that are typically used in the passive as in be born, be deemed, be stranded, be taken aback,
etc.
176
This rule simply says that if there is a transitive verb (tran(sitive)-v) lexeme selecting one SPR
and at least one COMPS element,
88
there is another verb (pass-tran-v) which selects the rst
COMPS in the input as its SPR (subject) and an optional PP[by] with the the remaining COMPS
value unchanged.
89
Notice that this lexical rule is not precise enough. For example, consider the following:
(26) a. He kicked the ball. vs. The ball was kicked by him.
b. John kicked him. vs. He was kicked by John.
We can observe that the active subject is changed from he to him, indicating that we need to
change the case value from nominative to accusative, or just refer to the index value of the active
subject like the following reformulated one:
(27) Passive Lexical Rule (Revised):
_
_
HEAD| POS verb
SPR NP
i
COMPS NP
j
,...
_
_
_
_
HEAD
_
_
POS verb
VFORM pass
_
_
SPR NP
j
COMPS ...,
_
PP
i
_
by
_
_
_
The revised Passive Lexical Rule now refers to the index value of the SPR and COMPS value,
avoiding the issue of changing the case values of the arguments.
Let us consider one simple example rst:
(28) a. John sent her to Seoul.
b. She was sent to Seoul.
The active verb sent is turned into the passive verb sent by the Passive Lexical Rule in (27):
(29) _
_
send
HEAD|POS verb
SPR NP
i
COMPS NP
j
, 3 PP[to]
_
_
sent
HEAD
_
_
POS verb
VFORM pass
_
_
SPR NP
j
COMPS 3 , (PP
i
[by])
_
_
As seen from the output, the passive sent takes a SPR whose index value is identical to that
of the object in the input. The passive sent also inherits the PP[to] complement and selects an
optional PP whose index value is identical to the SPR (subject) of the input.
90
This output lexical entry will then assign the following structure to (28b):
88
A lexeme is an abstract unit of morphological analysis. For example, drive, drives, driving, drove, drive are forms
of the same lexeme DRIVE. In this sense, we can take a lexicon to consist of lexemes as headwords.
89
Verbs like resemble will thus not undergo such a lexical rule whereas rumor exists only as a type of pass-tran-v.
90
As noted in 6.5.2, the preposition functioning as a marker rather than a predicate does not contribute to the meaning
of the head PP. This makes its index value identical to that of its object NP.
177
(30)
S
_
_
_
_
_
_
_
_
_
_
_
_
2 NP
VP
_
SPR 2
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
N
V
_
SPR 2
COMPS 5
_
5 VP
_
SPR 2
COMPS
_
.
.
.
.
.
.
.
.
.
.
.
.
She was
V
_
SPR 2
COMPS 3
_
3 PP
sent to Seoul
As given in the structure, the passive sent combines with its PP[to] complement, forming a VP
that still requires a SPR. This VP also functions as the complement of the auxiliary be. Since
be is a raising verb as given here again in (31), its subject (SPR value) is identical to its VP
complements subject.
91
(31)
_
_
be
HEAD| POS verb
SPR 2
COMPS
_
VP
_
_
VFORM pass
SPR 2
_
_
_
_
_
The raising property of be thus ensures that its SPR value is identical with the VP complements
SPR value. This SPR requirement on be will be passed upto the highest VP in accordance with
the VALP. When this VP combines with the subject Stephen in accordance with the Head-
Specier Rule, we will have a well-formed passive sentence.
The Passive Rule can be applied to sentences including a CP complement too. Consider the
following examples:
(32) a. They widely believed that John was ill.
b. That John was ill was widely believed.
The application of the Passive Lexical Rule to the active believed will generate the passive
output:
91
Chapter 7 for the basic properties of raising verbs and adjectives.
178
(33) _
_
believed
HEAD|POS verb
SPR NP
i
COMPS CP
j
_
believed
HEAD
_
_
POS verb
VFORM pass
_
_
SPR CP
j
COMPS (PP
i
)
_
_
The output passive believed then can project the following structure:
(34)
S
2 CP
VP
_
SPR 2
_
.
.
.
.
.
.
.
.
.
.
That
John
was ill
V
_
SPR 2
COMPS 5
_
5 VP
_
SPR 2
_
was Adv
VP
_
SPR 2
_
.
.
.
.
.
.
.
.
.
.
widely
V
_
SPR 2
COMPS 3
_
3 PP
_
decide
HEAD|POS verb
SPR NP
i
COMPS S
j
[QUE +]
_
_
decided
HEAD
_
_
POS verb
VFORM pass
_
_
SPR S
j
[QUE +]
COMPS (PP
i
)
_
_
The output passive decided then will generate the following structure:
(37)
S
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
_
1 S[QUE +]
VP
_
SPR 1
_
.
.
.
.
.
.
.
.
.
.
.
.
Which attorney
will give the
closing
argument
V
_
SPR 1
COMPS 4
_
4 VP
has
V
_
SPR 1
COMPS 3
_
3 VP
been
V
_
_
VFORM pass
SPR 1
COMPS 2
_
_
decided
The passive verb decided selects the optional PP complement and an indirect question subject.
The auxiliary verb been combines with the lowest VP. Since been is a raising verb, its subject is
identical with the VP complements subject. The auxiliary raising verb has also combines with
its complement VP whose subject is identical with its own subject. In this way, the subject of
has is identical with that of the passive verb decided.
9.4 Prepositional Passive
In addition to the passivization of an active transitive verb, prepositional verbs can also have
passive:
180
(38) a. You can rely on Ben.
b. Ben can be relied on.
(39) a. They talked about the scandal for days.
b. The scandal was talked about for days.
This kind of passive examples is unexpected if we apply a passive operation only to transi-
tive verbs. Neither the SD-SC analysis nor the transformation approach can account for such
examples. The present analysis also needs a revision to deal with such cases.
One thing we can notice is that even though such verbs select a PP complement, the passive
verbs here form a coherent syntactic unity with the following preposition. Observe the following
contrast:
(40) a. They talked repeatedly about the scandal for days.
b. You can rely absolutely on Ben.
(41) a.*The scandal was talked repeatedly about for days.
b.*Ben can be relied absolutely on.
As shown in the contrast, unlike the active, the passive allows no adverb to intervene between
the passive verb and the preposition.
The difference between active and passive further can be attested by a constituent test:
(42) a. It was about the scandal that they talked for days.
b. You can rely on Ben and on Maja.
(43) a. *It was about for days that the scandal was talked repeatedly.
b. *Ben can be relied absolutely on and completely on.
As for the behavior of active examples here, we can simply assume that such prepositional
verbs select a PP complement:
(44)
S
NP VP
They V PP
V
.
.
.
.
.
.
.
.
.
.
.
NP
V P the scandal
talked about
b.
VP
.
.
.
.
.
.
.
.
.
.
.
V P NP
_
prepositional-v
SPR NP
i
COMPS PP
j
_
PFORM 4
_
_
_
_
VFORM pass
SPR NP
j
COMPS 4 , (PP
i
[PFORM by])
_
_
This rule ensures that a prepositional verb (prepositional-v) can have its counterpart passive
verb.
93
The output passive prepositional verb selects a SPR whose index value is identical to
the input verbs PP complement. The output can select two complements: a preposition and an
optional PP complement. This preposition is identical with the PFORM value of the input PP.
Consider the following example:
(47) a. The lawyer looked closely into the document.
b. The document was closely looked (*closely) into by the lawyer.
The prepositional verb look will undergo the rule in (46) and generates a passive output:
93
Verbs selecting a PP can be divided at least into two groups: prepositional verbs and non-prepositional verbs. For
example, verbs like look, come, live, recover select a PP complement but are not prepositional verbs since they do not
have passive counterparts.
182
(48)
_
_
look
SPR NP
i
COMPS PP
j
_
into
_
_
looked
VFORM pass
SPR NP
j
COMPS P[into], PP
i
[by]
_
_
The output passive now selects a preposition and an optional PP as its complement and a subject
NP. This will generate a structure like the following:
(49)
VP[n]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
VFORM pass
COMPS 2 P[into], ( 3 )
_
_
2 P 3 PP
_
HEAD|POS verb
SPR NP
i
COMPS NP
j
_
_
_
HEAD| POS adj
SPR NP
j
COMPS
_
PP
i
_
_
As given here, the output now functions as an adjective with no specic constraint on the
PFORM value. The output of this lexical rule can generate a structure like the following:
(65)
S
NP
VP
He V AP
was Adv AP
very A PP
_
get
HEAD|POS verb
VAL
_
_
SPR 1 NP
COMPS 2 VP[pass]
_
_
ARG-ST 1 [pat], 2
_
_
This will then generate a structure like the following:
(69)
S
NP
VP[n]
Rosie V VP[pass]
got V PP
struck by lightning
In terms of structure, get-passives are not different from be-passives.
9.7 Middle Voice
In addition to the active and passive voices, in English there exists another voice, often called
middle. Consider the following:
95
With a more elaborated feature structure for pragmatic information, we can formulate this in the feature structure
system.
186
(70) a. John opened the door.
b. John cooked the casserole in the oven.
(71) a. The door was opened by John.
b. The casserole was cooked in the oven by John.
(72) a. The door opened.
b. The casserole cooked in the oven.
The sentences in (71) are passive forms of those in (70). Then what about (72)? The subject here
is identical with that of the passive, but the verb is not in the passive but in the active form. As
such, an intransitive verb that appears to be active but expresses a passive action characterizes
the English middle voice. That is, we can say that English middle voices are syntactically active
but semantically passive:
(73) a. John rang the bell. The bell rang.
b. John broke the window. The window broke.
c. John smashed the vase. The vase smashed.
d. John melted the ice. The ice melted.
e. John sank the ship. The ship sank.
However, not all transitive verbs have middle voices:
(74) a. John kicked the bell. *The bell kicked.
b. John hit the window. *The window hit.
c. John bought the vase. *The vase bought.
Such middle voices in general describe permanent properties of the subject. This general se-
mantic condition makes middle voice incompatible with duration adverbs:
96
(75) a. ??This car drove smoothly last night.
b. ??This clothes washed well last night.
In addition, these middle verbs do not allow the by phrase:
(76) a. *The bell rang by John.
b. *The window broke by the child.
c. *The vase smashed by the baby.
Given these observations, we can introduce the following lexical rule for a limited set of
transitive verbs in English:
96
The double question marks mean that such a sentence is in general unacceptable, but can be used in certain context.
187
(77) Middle-Voice Lexical Rule:
_
_
HEAD| POS verb
VAL
_
_
SPR 1 NP
i
COMPS 2 NP
j
_
_
ARG-ST
_
1 [agt], 2 [th]
_
_
_
HEAD| POS verb
SPR NP
j
COMPS
_
_
The lexical rule means that a verb selecting an agent and a theme argument can be turned into a
verb selecting this theme as the subject. For example, consider the verb open:
(78)
_
_
open
VAL
_
_
SPR 1 NP
i
COMPS 2 NP
j
_
_
ARG-ST 1 [agt], 2 [th]
_
_
open
SPR NP
j
COMPS
_
_
There is no change in the ARG-ST. The output realizes the input verbs theme argument as its
subject.
(79)
S
NP
VP[pst]
The door V
opened
Though such a lexical process can add a special semantic and pragmatic meaning (such as
describing permanent properties), it is another way of expanding lexical usages.
188
9.8 Exercises
1. Draw the complete tree diagram for each of the following sentences and then provide the
lexical entry for the italicized passive verb.
(i) a. Peter has been asked to resign.
b. I assume the matter to have been led in the appropriate records.
c. Smith wants the picture to be removed from the ofce.
d. The events have been described well.
e. Over 120 different contaminants have been dumped into the river.
f. Heart disease is considered the leading cause of death in the United States.
g. The balloon is positioned in an area of blockage and is inated.
2. Consider the following examples and provide the putative active counterpart.
(i) a. That we should call the police was suggested by her son.
b. Whether this is feasible hasnt yet been determined.
c. Paying taxes cant be avoided.
Also see if there is any relationships between the above sentences with the following
passives:
(ii) a. It was suggested by her son that we should call the police.
b. It hasnt yet been determined whether this is feasible.
c. It cant be avoided paying taxes.
3. The verbs like get and have can be used in so-called pseudo passives:
(i) a. Frances has had the drapes cleaned.
b. Shirley seems to have forgotten Fred promoted.
(ii) a. Nina got Bill elected to the committee.
b. We got our car radio stolen twice on holiday.
Check if we can replace the italicized verbs with different verb forms (e.g., base or inni-
tive) and then discuss if such a replacement test can tell us anything about the properties
of such constructions. In addition, provide the tree structures for the sentences with the
lexical entries for have, get, and the italicized verbs.
4. The following sentences appear to have similar meanings. Discuss the relationships
among the following sentences if you can nd any.
(i) a. Joe rolled the barrel down the hill.
b. Joe had the barrel rolled down the hill.
c. The barrel was rolled down the hill.
d. The barrel rolled down the hill.
5. Consider the following passive examples.
(i) a. We cannot put up with the noise anymore.
b. He will keep up with their expectations.
(ii) a. This noise cannot be put up with.
189
b. Their expectations will be kept up with.
Do the Passive Lexical Rules proposed in this chapter explain such? Also observe the
following examples which have two different kinds of passive:
(iii) a. They paid a lot of attention to the matter.
b. The son took care of his parents.
(iv) a. The matter was paid a lot of attention to.
b. A lot of attention was paid to the matter.
Can you think of any way to account for such examples?
6. Check if the following verbs are middle verbs or not. In so doing, try to construct relevant
examples.
(i) ll, break, withdraw, move, march, jump, load
Also consider the following examples:
(ii) a. These shirts wash *(well).
b. The meat cuts *(easily).
c. The books sell *(well).
In these examples, the presence of an adverb is obligatory. Are they any other verbs that
behave like these? Also, provide a lexical rule that can capture such a middle formation
operation.
7. Read the following passage and identify the errors in the verbs verb form. In addition,
provide the lexical information for the corrected form.
(i) This survey aim at investigating the effectiveness of the appraisal system in
our company. The survey conduct last month. The data collect by means of a
questionnaire survey and three focus group interviews. In the questionnaire,
respondents ask ten questions regarding the current appraisal system. It nd
that the respondents generally quite satisfy with the system, but about half of
them state it should carry less frequently. In the focus group interviews, the
respondents give the opportunity to discuss the system openly. Some respon-
dents complain that the appraiser know too little about them to give detailed
and objective comments. The ndings indicate that the rationale for conduct-
ing the appraisal exercise should explain more clearly to our staff.
(ii) Syntax is the discipline that examining the rules of a language that dictate
how the various parts of sentences gone together. While morphology looks
at how individual sounds formed into complete words, syntax looks at how
those words are put together for complete sentences. One part of syntax, call-
ing inection, deals with how the end of a word might changed to tell a lis-
tener or reader something about the role that word is playing. Regular verbs
in English, for example, change their ending based for the tense the verb is
representing in a sentence, so that when we see Robert danced, we know the
sentence is in the past tense, and when we see Robert is dancing, we know
190
it is not. As another example, regular nouns in English become plural simply
by adding an s to the end. Cues like these play a large role for helping hearers
understanding sentences.
97
97
From http://www.wisegeek.com/what-is-syntax.htm
191
10
Wh-Questions
10.1 Clausal Types and Interrogatives
Like other languages, English also distinguishes a set of clause types that are characteristically
used to perform different kinds of speech acts:
(1) a. Declarative: John is clever.
b. Interrogative: Is John clever? Who is clever?
c. Exclamative: How clever you are!
d. Imperative: Be very clever.
Each clause type in general has its own functions to represent speech acts. For example,
declarative makes a statement, interrogative asks a question, exclamative makes an exclamatory
statement, and imperative issues a directive.
As for the interrogative, there are basically two types, yes-no question and wh-question:
(2) a. Yes-No questions: Can the child read the book?
b. Wh-questions: What can the child read?
Yes-no questions are different from their declarative counterparts by having subject and aux-
iliary verb in the reverse order. As we have seen in Chapter 8, such yes-no questions are gener-
ated from the combination of an inverted nite auxiliary verb with its subject and complement
in accordance with the SAI Rule:
193
(3)
S
.
.
.
.
.
.
.
.
.
.
.
.
V
_
_
INV +
AUX +
SPR 1
COMPS 2
_
_
1 NP
2 VP
NP C
who C S
will NP
VP
they V VP
V NP
recommend e
The wh-phrase who is originally in the object of recommend and then moved to the specier po-
sition of the intermediate phrase C
.
98
The auxiliary verb will is also moved from the V position
to the C. This kind of movement operation at rst glance can be appealing in capturing the link-
age between the two. However, moving an overt element to form a wh-question immediately
runs into a problem for examples like the following:
(9) a. Who did Kim work for and Sandy rely on ?
b. *Who did Kim work for and Sandy rely on Mary?
If we adopt a movement process for such examples, there must be an operation that the two NP
gaps are collapsed into one NP and become who. We cannot simply move one NP, because it
will generate an ill-formed sentence like (9b).
99
Unlike this kind of movement operation, we can assume that there is no movement process
at all to generate such wh-questions, but there exists just a mechanism of feature percolations.
In this system, the missing information of a gap is passed up to the tree until it meets the
corresponding ller:
98
In transformational analyses, the movement of a wh-phrase is often called A
NP S/NP
who V NP VP/NP
S/NP
and S/NP
NP VP/NP
NP VP/NP
S/NP
and S/PP
NP VP/NP
NP VP/PP
_
recommend
VAL
_
_
SPR 1
COMPS 2
GAP
_
_
_
_
_
_
recommend
ARG-ST 1 , 2
_
_
_
recommend
VAL
_
_
SPR 1
COMPS 2
GAP
_
_
_
_
This indicates that the verb recommend selecting two arguments seen from the ARG-ST can
be realized in two different ways. That is, the syntactic valence information of the verb can be
different with respect to how the two arguments are realized. In (15a), the two arguments are
realized as the SPR and COMPS value, generating examples like (14a). However, in (15b), the
second argument is realized not as a COMPS value but as a GAP value.
100
This lexical information in (15b) will eventually project a structure like the following:
(16)
S
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
NP S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
_
_
_
_
Who V NP
VP[GAP 1 NP]
will they V[GAP 1 NP]
recommend
In this structure the head verb recommend itself has the GAP information. This GAP value will
be passed up to the point where it will be discharged by the ller.
Passing up the GAP information upto the mother froma non-head is ensured by the following
principle:
(17) GAP Inheritance Principle (GIP):
The mothers GAP value is the sum of each daughters GAP value minus the bound
100
We thus introduce the feature GAP as a kind of VAL feature.
198
GAP value.
This GAP value will be percolated upto a higher tree until it meets an appropriate ller and then
is discharged. In English this dischargement takes place at the S level in accordance with the
following Head-Filler Rule:
(18) Head-Filler Rule:
S
_
GAP
_
1 , S
_
GAP 1
_
This grammar rule says when the head expression S containing a nonempty GAP value com-
bines with its GAP value, the resulting phrase will form a grammatical head-ller phrase, with
the discharge of the GAP value.
10.3.2 Non-subject Wh-questions
One thing to note here is that in English only complements can be realized as a GAP value.
Unlike languages like Korean or Japanese where both subject and object can be extracted, Indo-
European languages including English exhibit subject-object asymmetry in various phenomena.
For example, though the object can be easily realized as a gap element, the subject is not:
101
(19) a. *I saw the car that you think that John claimed that could hit the man
b. I saw the car that you think that John claimed that the man could hit
Also observe the following contrast:
(20) a. *What did [that John bought ] upset Jack?
b. What did Julia think [that John bought ]?
(21) a. *Who is [a book about ] being ready by the class?
b. Who is the class reading [a book about ]?
The data here indicate that an element from the subject is less extractable than one from the
complement.
102
Reecting this subject/object asymmetry, we can assume the Argument Realization Con-
straint as following:
103
(22) Argument Realization Constraint:
The non-initial argument can be realized either as a COMPS element or as a GAP
element.
This constraint tells us that the rst argument must be realized as the SPR and the remaining
arguments are realized as COMPS and GAP value. For example, lets see the verb put. This
verb will select three arguments:
101
See (36) also.
102
Such a constraint is often called subject island. See 11.4 for the discussion of the so called island constraints.
103
The notation is an minus operation on the two lists, and
A
is a variable over the list.
199
(23)
_
_
put
ARG-ST
_
1 , 2 , 3
_
_
_
Of these three, the rst one will be always realized as the SPR element whereas the non-initial
two arguments can be realized either as a COMPS or as a GAP element. This means we have at
least the following three realizations of put:
(24) a.
_
_
put
VAL
_
_
SPR 1
COMPS 2 , 3
GAP
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
b.
_
_
put
VAL
_
_
SPR 1
COMPS 3
GAP 2
_
_
ARG-ST
_
1 , 2 , 3
_
_
_
c.
_
_
put
VAL
_
_
SPR 1
COMPS 2
GAP 3
_
_
ARG-ST 1 , 2 , 3
_
_
Each of these three are meant to be examples like the following:
(25) a. John put the books in a box.
b. Which book did John put in the box?
c. Where did John put the book?
As we can see here, the verb put in (25a) has the canonical realization of arguments whereas
in (25b), the NP argument is realized as a gap and in (25c), the PP is realized as a gap. The
structure for (25b) is given in the following:
104
104
We assume that every wh-element in questions carries the feature [QUE +].
200
(26)
S[QUE +]
2 NP[QUE +]
S
_
GAP 2
_
Which book V
1 NP
VP
_
_
SPR 1
GAP 2
_
_
did John
V
_
_
SPR 1
COMPS 3
GAP 2
_
_
3 PP
NP S[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
_
saw
VAL
_
_
SPR 1
COMPS
GAP 2
_
_
ARG-ST 1 , 2
_
_
Since the verb has its COMPS be realized as its GAP value, it need not look for its complement
in its sister. This GAP information, in accordance with the GIP, will be passed upto the second
highest S where it is discharged by the ller who in accordance with the Head-Filler Rule in
(54). Once again, we can observe that the grammar rules are closely interacting with the general
principles such as the HFP and GIP.
It is also easy to verify that this system captures examples like (29) in which the gap is a
non-NP phrase:
(29) a. [In which box] did John put the book ?
b. [How happy] has John been ?
It is not difcult to observe that the categorial status of the ller is identical with that of the gap.
The structure of (29a) can be represented as following:
202
(30)
S
.
.
.
.
.
.
.
.
.
.
.
3 PP
S[GAP 3 PP]
.
.
.
.
.
.
.
.
.
.
.
did John
V
_
_
COMPS 2 NP
GAP 3 PP
_
_
2 NP
1 NP VP[GAP 1 NP]
.
.
.
.
.
.
.
.
.
.
Who V NP
PP
1 NP
S
_
GAP 1
_
Who
VP
_
GAP 1
_
.
.
.
.
.
.
.
.
.
.
V
_
GAP 1
_
NP
PP
NP[QUE +]
VP
.
.
.
.
.
.
.
.
.
.
Who V NP
PP
C S[GAP 2 NP]
Sara invited
b.
*CP[GAP 1 NP]
C VP[GAP 1 NP]
V S[GAP 2 NP]
Sara invited
Just like (36)a, this sentence also has the object of the embedded clause is gapped. The verb
believe combines with this incomplete S.
How can then we account for (39b) in which the subject is extracted? One thing to note
here is that such a subject-gapped example is possible only with so-called parenthetical verbs
likethink, imagine, assume and so forth:
(40) a. Who do you believe likes Mary?
b. Who do you imagine likes Mary?
(41) a. *Who do you know won the prize?
b. *Who do you recognize won the prize?
All these verbs select either an S or CP as their complement:
(42) a. I believe that Mary won the prize.
b. I know that Mary won the prize.
The data here indicate that the subject extraction is rather lexically controlled. That is, we can
assume that a limited set of verbs selecting either a CP or an S can undergo a lexical rule so that
they can select a nite VP with their subject lexically realized as a GAP in accordance with the
following rule:
106
(43) Subject Extraction Lexical Rule:
_
_
parenthetical-v
HEAD|POS verb
ARG-ST 1 , S
_
_
_
_
HEAD| POS verb
SPR 1
COMPS
_
VP
_
_
VFORM n
GAP 1
_
_
_
_
_
This lexical rule means that given a parenthetical verb that selects a nite sentence, we have
another counterpart that selects a nite VP with its subject as the GAP value. For example, the
verb think selects a nite sentence so that it can undergo the lexical rule in (43):
106
This rule is reminiscent of the meta rule assumed in Gazdar et el. (1987).
206
(44)
_
_
think
HEAD|POS verb
ARG-ST 1 , S[n]
_
_
_
_
think
HEAD| POS verb
SPR 1
COMPS
_
VP
_
_
VFORM n
GAP 1
_
_
_
_
_
The output of think now selects a nite VP with one GAP value. This can generate a structure
like the following:
(45)
S
1 NP S[GAP 1 NP]
do you V
VP
_
_
VFORM n
GAP 1 NP
_
_
think
V
NP
likes Robin
Every local structure here is well-formed and licensed by the grammar rules and principles.
The verb think combines with its complement VP likes Robin. This nite VP has a GAP value
which is identical with the VPs subject. This GAP value is passed up to the point where it is
discharged by the ller who.
107
The SELR will not apply to verbs selecting a CP. For example, verbs like wonder will select
a CP so that it cannot undergo this lexical rule:
(46) a. *The man who I wondered [whether chased Fido] returned.
b. *The man who I wondered [if chased Fido] returned.
c. *The man I wondered [chased Fido] returned.
In addition, verbs like complain subcategorize only for a CP in many varieties of English:
(47) a. Who did you complain that you hated?
107
One thing to note is that the SELR in (??) introduces the GAP value at the VP level, not at the VP level. This
means that the GAP value introduced by this rule is not sensitive to the ARC in (22) which applies only to word level
expressions.
207
b. *Who did you complain you hated?
In sum, by admitting that English has subject and object asymmetry and does not basically
allow the subject to be gapped, we can account for relevant facts rather easily. Of course, in this
process, we also need to introduce a lexical rule to account for the examples where the subject
seems to be gapped.
10.5 Indirect Questions
10.5.1 Basic Structure
Among the verbs selecting a sentential complement, there also exist verbs requiring an indirect
question:
(48) a. John asks [whose book [his son likes ]].
b. John asks [what [his son likes ]].
c. John has forgotten [which player [his son shouted at ]].
d. He told me [how many employees [Karen introduced to the visitors]].
Notice that not all verbs allow such indirection questions as their complement:
(49) a. Tom denied (that) he had been reading that article.
b. *Tom denied which book he had been reading.
(50) a. Tom claimed (that) he had spent ve thousand dollars.
b. *Tom claimed how much money she had spent.
Factive verbs like deny or claim cannot combine with an indirect question: only a nite CP can
function as their complement. Verbs selecting an indirect question as their complement can be
in general classied by the meaning:
108
(51) a. interrogative verbs: ask, wonder, inquire
b. verbs of knowledge: know, learn, forget
c. verbs of increased knowledge: teach, tell, inform
d. decision verbs/verbs of concern: decide, care
The complement CP of these verbs cannot be a canonical CP: it must be an indirect question:
(52) a. *John inquired that he should read it.
b. *John forgot that the drawer contained the money.
c. *Peter will decide that we should review the book.
(53) a. John inquired which book he should read.
b. John forgot which drawer contained the money.
c. Peter will decide which book we should review.
108
Unlike verbs such as wonder, those like tell can select either a declarative CP or an indirect question.
208
This means that we need to distinguish indirect questions from CPs headed by that or simple Ss.
This in turn means that verbs like inquire, forget, decide, wonder will be different from those
like deny in that the formers CP complement is specied with [QUE +].
(54)
a.
_
_
wonder
HEAD| POS verb
SPR 1
COMPS
_
CP
_
QUE +
_
_
_
_
b.
_
_
deny
HEAD| POS verb
SPR 1
COMPS
_
CP
_
QUE
_
_
_
_
As given in the following, the indirect question needs to be marked with the feature [QUE
+] so that the clause is distinguished from the canonical S:
(55)
VP
V S[QUE +]
asks 1 NP[QUE +]
S[GAP 1 NP]
whose book NP
VP[GAP 1 NP]
his son V[GAP 1 NP]
likes
The feature QUE starts from the wh-word which. This feature, similar to the feature GAP, will
pass up to the point where it is required by a verb or to the highest position to tell us the given
sentence is a question:
(56) a. [
S[QUE +]
In which box did he put the book ]?
b. [
S[QUE +]
Which book of his father did he read ]?
(57) a. John asks [
S[QUE +]
in which box he put the book].
b. John asks [
S[QUE +]
which book of his father he read].
The percolation of the feature QUE (including GAP) can be ensured by the following inheri-
tance principle:
(58) Nonlocal Feature Inheritance Principle (NIP): A phrases nonlocal feature such as
GAP and QUE is the sum of its daughters nonlocal feature values minus the bound
nonlocal features.
209
This principled constraint allows the QUE value to pass up to the mother even from a nonhead
as given in the following:
(59) a. Kim has wondered [[in which room] Gary stayed ].
b. Lee asked me [[how fond of chocolates] the monkeys
are ]].
The structure of (59a) will look like the following:
(60)
VP
V S[QUE +]
wondered
1 PP
_
QUE +
_
S[GAP 1 PP]
in which room NP
VP[GAP 1 PP]
Gary V[GAP 1 PP]
stayed
The verb forgotten selects an indirection question sentence, an S with the feature [QUE +]. This
system will not generate examples like the following in which such verbs combine with a [QUE
] S:
(61) a. *Kim has wondered that Gary stayed in the room.
b. *Kim asked me that the monkeys are very fond of chocolates.
Another important restriction here is that the missing phrase must correspond with the wh-
phrase in the initial position of the indirect phrase. For example, the following structure is not
licensed simply because there is no Head-Filler Rule that allows the combination of the ller
NP with a PP missing S:
210
(62)
VP
V *S[QUE +]
forgotten NP[QUE +]
S[GAP 1 PP]
which room NP
VP[GAP 1 PP]
Gary V[GAP 1 PP]
stayed
In a similar fashion, the present system also predicts the following contrast:
(63) a. John knows [whose book [Mary bought ] and [Tom
borrowed from her]].
b. *John knows [whose book [Mary bought ] and [Tom
talked ].
The partial structure of these can be represented as following:
(64) a.
S[QUE +]
1 NP[QUE +]
S[GAP 1 NP]
and S[ GAP 1 NP ]
and S[ GAP 2 PP ]
_
whether
SYN
_
_
HEAD| POS comp
VAL| COMPS S
QUE +
_
_
_
_
According to the lexical information, whether selects a nite S with the [QUE +] value, gener-
ating a structure like the following:
(67)
CP[QUE +]
.
.
.
.
.
.
.
.
.
.
.
.
.
C[QUE +]
S[n]
VP[SPR NP[PRO]]
V
NP
V
NP
1 NP[QUE +]
S
_
_
VFORM inf
GAP 1
_
_
which politician
VP
_
_
VFORM inf
SPR PRO
GAP 1
_
to support
Consider from the bottom. The verb support selects two arguments whose second argument is
realized as a GAP:
(79)
_
_
support
VAL
_
_
SPR PRO
COMPS
GAP 2 NP
_
_
ARG-ST 1 NP, 2 NP
_
_
The verb will then form a VP with the innitival marker to. Since this VPs subject is PRO, the
VP can be projected into an S with the GAP value. The S then forms a well-formed head-ller
phrase with the ller which politician. The QUE value on the phrase makes the whole innitival
clause as an indirect question that can be combined with the verb knows.
215
10.6 Exercises
1. Draw tree structures for the following and mark what kind of grammar rules (e.g., Head-
Specier, Head-Complement, Head-Modier, Head-Only, and Head-Filler Rule) licenses
each phrasal combination. In addition, Provide a tree structure for the following sen-
tences with the lexical entry for the italicized predicates:
(i) a. Joseph has forgotten how many matches he has won.
b. The committee knows whose efforts to achieve peace the world should
honor.
c. Fred will warn Martha that she should claim that her brother is patriotic.
d. That Bill tried to discover which drawer Alice put the money in made us
realize that we should have left him in Seoul.
e. Mary told me how brave he was.
f. Jasper wonders which book he should attempt to persuade his students to
buy.
g. What proof that he has implicated have you found?
2. Draw tree structures for the following sentences?
(i) a. Whose car is blocking the entrance to the store?
b. Which textbook was used in his class last summer?
c. Which textbook did the teacher use in the class last summer?
d. To whom did you send your job application?
3. We have seen wh-questions and indirect questions in which only arguments are gapped.
How then can the present system account for examples like the following in which wh-
phrases are not arguments but behave like an adjunct?
(i) a. How carefully have you considered your future career?
b. When can we register for graduation?
c. Where do we go to register for graduation?
d. Why have you borrowed my pencil?
Now consider the following examples:
(ii) a. Why did you say that she will invite me?
b. How long did he tell you that he waited?
c. I asked when they think that they will meet Mary.
Do these examples have more than one interpretation, depending on the scope of the
wh-elements? Compare these with the following?
(iii) a. Why do you wonder whether she will invite me?
b. How often did he ask when she will meet at the party?
Can you see the difference between examples in (ii) and those in (iii)? Also, think of how
to account for these examples within the present system.
216
4. Analyze the following sentences in the paragraph as far as you can. Use tree structures
and lexical entries.
(i) Within grammar lies the power of expression. Understand grammar, and you
will understand just how amazing a language is. You uncover the magicians
tricks, you nd the inner workings of not only your own language, but you can
also see how it is different from the language youre studying. You will nd
that different languages are better for expressing different ideas, and you will
be able to make conscious decisions about how you communicate. Grammar
gives you the formula, the canvas, or the blank notation sheet that you then
choose which variables, paints, or notes you want to put down. Once you
know how to use each part of speech, you will be able to expand outside of the
box and express yourself in ways that no one has ever expressed themselves
before. A solid understanding of the grammar of a language gives you the
skeleton, and your words bring it to life. That is why we study grammar.
111
111
From GRAMMAR (no, dont run, I want to be your friend!) by Colin Suess
217
11
Relative Clause Constructions
11.1 Introduction
English relative clauses, basically functioning as a modier, are also long distance dependency
constructions in that the gap in the relative clause is long-distance dependent upon the relative
pronoun ller:
(1) a. The video [which you recommended ] was really terric.
b. The video [which I thought you recommended ] was really terric.
Relative clauses can be classied according to several criteria. We can rst classify them by
the type of missing elements in the relative clause:
(2) a. the student who won the prize
b. the student who everyone likes
c. the baker from whom I bought these bagels
d. the person whom John gave the book to
e. the day when I met her
f. the place where we can relax
As given here, the missing element can be subject, object, or oblique argument, prepositional
object, or even temporal and place adjunct.
We also can divide relative clauses by the type of relative pronoun: wh-relatives, that-
relatives, and bare relatives.
112
(3) a. The president [who [Fred voted for]] has resigned.
b. The president [that Fred voted for] dislike his opponents.
c. The president [ Fred voted for] has resigned.
In (3c) we have no relative pronoun like who or that, but the clause Fred voted for modies the
president.
Depending on the tenseness of the relative clause, we have nite and innitival relative
clauses:
112
We take that as a relative pronoun too.
219
(4) a. He is the kind of person [with whom to consult].
b. There is not a whole lot [with which to disagree].
c. We will invite volunteers [on whom to work].
In addition, examples like (5) are often called reduced relative clauses in that these expres-
sions seem to omit the string wh-phrase + be as indicated in the parenthesis:
(5) a. the person (who is) standing on my foot
b. the prophet (which was) descended from heaven
c. the bills (which were) passed by the House yesterday
d. the people (who are) in Rome
e. the people (who are) happy with the proposal
11.2 Restrictive vs. Nonrestrictive Relative Clauses
11.2.1 Basic Differences
In addition to the types of relative clause we seen before, there is a canonical distinction between
restrictive and nonrestrictive. Consider the following:
(6) a. The person who John asked for help thinks John is a foolish.
b. Mary, who John asked for help, thinks John is a foolish.
The relative clause in (6a) semantically restricts the denotation of person whereas the one in
(6b) just adds extra information about Mary. Let us consider one more pair:
(7) a. John has two sisters who became lawyers.
b. John has two sisters, who became lawyers.
As pointed out, the rst difference between the two types of relative clauses comes from the
meaning, as represented in the following diagrams. The denotation of the restrictive relative
clause two sisters who became lawyers is the interaction between the set of two sisters and that
of the lawyers. There can be more than two sisters, but there are only two who became lawyers.
Meanwhile two sisters, who became lawyers mean that there are two sisters and they all became
lawyers: there is no interaction meaning here. Thus, there exist only two sisters.
(8) a. Meaning of the restrictive relative clause
sisters
lawyers
D E
F
A
B C
I
G H
220
b. Meaning of the non-restrictive relative clause
sisters
lawyers
A B A B
This meaning difference explains the difference in the types of possible antecedents. For
example, only the nonrestrictive relative clause can modify a proper noun:
(9) a. Regan, whom the Republicans nominated in 1980, now lives in California.
b. *Regan who began his career as a radio announcer came to hold the nations high-
est ofce.
Considering that a personal pronoun denotes one and the only one, nothing can restrict its
meaning.
The meaning difference is also related to the fact that only a restrictive clause can modify
indenite pronouns such as everyone and nothing, or indenite determiners like every and no.
(10) a. Every student who attended the party had a good time.
b. *Every student, who attended the party, had a good time.
(11) a. No student who scored 80 or more in the exam was ever failed.
b. *No student, who scored 80 or more in the exam, was ever failed.
The phrases with no, any, or every semantically have no referential interpretations. That is, if
we look at (11)b, who has its antecedent no student. However, since no student has no reference,
who has no referent.
The two types are also different with respect to stacking and ordering relations. For example,
restrictive clauses can be stacked, but nonrestrictive clauses cannot:
(12) a. The student who took the qualifying exam who failed it wants to retake it.
b. *Sam Bronowsky, who took the qualifying exam, who failed it, wants to retake it.
(13) a. I met the man that grows peaches that lives near your cousin.
b. I dont like the bills passed by the House yesterday that we objected to died in
the Senate.
c. Harold borrowed the book from Sally that he had been wanting to read.
In addition, the restrictive clause must precede the nonrestrictive clause:
(14) a. The contestant who won the rst prize, who is the judges brother-in-law, sang
dreadfully.
b. *The contestant, who is the judges brother-in-law, who won the rst prize sang
dreadfully.
221
11.2.2 Capturing the Differences
Let us consider some canonical restrictive relative clauses, rst:
(15) a. the senators [who] Fred met
b. the apple [that] John ate
c. the problem [ ] you told us about
Just like wh-questions, we can notice that relative clauses have bipartite structures: a wh-
element and a sentence with one missing element:
(16) a. wh-element S/XP
b. that S/XP
c. [ ] S/XP
We can represent the structure of (15a) as following:
(17)
N
S
_
REL +
_
senators 1 NP[REL +]
S[GAP 1 NP]
who NP
VP[GAP 1 NP]
Fred V[GAP 1 NP]
met
Notice that the complement of the verb met is realized as a GAP value which is percolated up
until it is discharged. The incomplete sentence Fred met combines with the relative pronoun
who, forming a head-ller phrase. The relative pronoun has the feature [REL +], passing up to
the S too. This REL feature on the sentence ensures that the whole relative clause functions as
a modier to the head senators.
222
(18)
N
3 N
S
_
REL +
_
senators 2 NP[REL +]
S
_
GAP 2
_
who NP
VP
_
GAP 2
_
Fred V
_
GAP 2
_
met
As given in the structure, the verb met has its object as a GAP value, and the ller who functions
as its ller. The combination of the ller whom and the gapped sentence Fred met forms a
well-formed head-ller phrase that can modify a nominal element. This ller has the nonlocal
REL feature that percolates upto the mother. This REL feature also observes the NIP (Nonlocal
Inheritance Principle) in the sense that its value is inherited to the mother phrase from a non-
head as illustrated in the following:
113
(19) a. The teacher set us a problem [the answer to which] we can nd in the textbook.
b. We just nished the nal exam [the result of which] we can nd next week.
c. I just met the friend [in whose apartment] I would be staying.
The REL value here is originated from a non-head and percolated upto the relative clause.
One thing to notice in the structure (19) is that the restrictive relative clause modies not
a fully saturated NP but an N
DP N
the N
NP
S
John, NP
whom we respect
If the these restrictive and non-restrictive types of relative clauses are structurally different,
how do they diff in terms of syntax and semantics? Lets compare the following two sentences:
These differences can be predicted from the structural differences: the restrictive relative
clause modies an N
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
NP
Det N
contestant
who won the
rst prize,
Since the combination of the restrictive relative clause who wont the rst prize and the head con-
testant forms an N
which in turn forms an NP with its specier the. Note that the nonrestrictive
clause who is the judges... can easily modify this nal NP. However, when the nonrestrictive
relative clause precedes the restrictive one, we will get a wrong structure:
(26)
*NP
NP
NP
who won...
the contestant,
who is the judges
brother-in-law,
As given in the structure, when the norestrictive relative clause who is the judges... modies
the full NP the contestant, the result will be a fully saturated NP. This in turn means that the
restrictive relative clause, requiring to modify an N
_
AP
PP
nonnite VP
S[REL +]
_
_
HEAD| MOD N
_
The head feature MOD will pass up the AP or PP or nonnite VP, enabling them to modify
the prenominal position. The MOD value on the S[REL +] can be added by a constructional
constraint.
116
116
This MOD value can be added by a morphological element in the verb of the relative clause, as in languages like
Korean. See Kim (2002).
226
(32)
N
3 N
S
_
_
MOD 3 N
REL +
_
_
senators 2 NP[REL +]
S
_
GAP 2
_
whom NP
VP
_
GAP 2
_
Fred V
_
GAP 2
_
met
As noted here, the [REL +] feature passed up on the S triggers the relative clause whom Fred met
to function as a modier to the N
3 N
VP
_
_
MOD 3 N
VFORM prp
_
_
.
.
.
.
.
.
.
.
.
.
.
.
.
PP
.
.
.
.
.
.
.
.
.
.
.
.
.
standing P
NP
on my foot
11.3 Subject Relative Clauses
Subject relative clauses behave like non-subject relatives clauses we have seen so far. However,
one main difference is that the presence of a wh-relative pronoun including that is obligatory:
(34) a. the senators [who] met Fred
227
b. the apple [that] fell down on the ground
Notice here that English does not allow a nite VP to function as the postnominal modier:
(35) a. *[The student [met John]] came.
b. *[The problem [intrigued us]] bothered me.]
As we have seen in the previous chapter, English does not allow the subject to be gapped, either.
As in subject wh-questions, subject relative clauses are generated when the subject of a matrix
verb is realized as a [REL +] expression. For example, the subject of meet can be realized as a
[REL +] expression:
(36)
_
_
meet
SPR [REL +]
COMPS
_
NP
_
_
_
This lexical entry will then be able to project a structure like the following:
(37)
N
3 N
i
S
_
_
REL +
MOD 3 N
i
_
_
senators
1 NP[REL +]
VP
_
SPR 1 NP[REL +]
_
who
V
_
_
SPR 1 NP[REL +]
COMPS 2
_
_
2 NP
met Fred
The VP met Fred combines with the subject relative pronoun who which carries the [REL +]
feature. The resulting S will still carry the REL feature, introducing the MOD feature. The
S[REL +] then modies the head senators by the Head-Modier Rule.
11.4 That-relative clauses
As noted earlier, that can be used as a relative pronoun:
117
117
Not all the analyses take that as a relative pronoun. One can develop an analysis in which that is taken to be a
complementizer.
228
(38) a. the people that voted in the election
b. the book that Sandy thought we had read
c. each argument that Sandy thought was unconvincing
A complexity arises when that is used as a complementizer as in (39):
(39) a. Mary knows that John was elected.
b. That John was elected surprised Frank.
c. It surprised Frank that John was elected.
d. The fact that John was elected surprised Frank.
e. Mary told Bill that John was elected.
How can we distinguish between the two usages of that? One clear difference is that the rela-
tive pronoun that requires its following clause be a sentence with one missing element whereas
the complementizer that combines with a complete S, as represented in the following:
(40)
a. CP
Comp S
that ...
b. S[REL +]
that ...
To see a clear difference between complementizer that and relative pronoun that, let us con-
sider some examples:
(41) a.
_
_
_
Mary told us
*Mary met the man
_
_
_
[that John disappeared].
b.
_
_
_
*Mary told us
Mary met the man
_
_
_
[that disappeared].
The clause John disappeared in (41a) is a complete clause whereas disappeared in (41b) is just
a VP clause still requiring a subject. This in turn means that John disappeared in (41a) is a CP
clause headed by the complementizer that whereas that disappeared in (41b) is a relative clause
with the relative pronoun that. A further contrast can be observed from the following too:
(42) a.
_
_
_
Mary told us
*Mary met the man
_
_
_
[that we will speak with John].
b.
_
_
_
*Mary told us
Mary met the man
_
_
_
[that we will speak with].
The verb told selects for an NP and a complete sentence as its complements whereas met just
for a simple NP. In (42a), the clause we will speak with John is a complete sentence and that
can be only a complementizer; the clause thus can function as the sentential complement of
229
told, but it is not required for met. Meanwhile, in (42b) that can be either a relativizer or a
complementizer. In either case, this cannot occur with told requiring an NP and a complete
sentence as its complements. However, as for met, when that is a relativizer, we have a well-
formed structure like the following:
(43)
N
3 N
S
_
_
MOD 3 N
REL +
_
_
men 2 NP[REL +]
S
_
GAP 2
_
that NP
VP
_
GAP 2
_
2 N
i
S
_
_
REL +
MOD 2 N
_
_
bench 1 PP[REL +]
S
_
_
MOD 2
GAP 1
_
_
on which VP
_
_
MOD 2
GAP 1
_
_
to sit
As given here, the VP to sit modies an N
2 N
i
*S
_
_
REL +
MOD 2 N
_
_
bench 1 PP[REL +]
CP
_
GAP 1
_
Det N
[GAP ]
the 1 N
S
_
_
GAP NP
i
MOD 1
i
_
_
bench
VP
_
_
GAP NP
i
MOD 1
i
_
_
to sit on
As given here, the VP clause to sit on will have the MOD feature together with one GAP
element. The issue here is how to deal with the GAP value here since there is no ller for the
GAP value. One thing we can notice here is that English allows an incomplete S allows to
function as a postnominal modier.
(53) a. the person [I met ].
b. the box [we put the books in ].
What this means is that English allows an S with missing an accusative NP can also function as
a postnominal modier with discharging the GAP value as the result:
(54) Head-REL Phrase Rule:
N
_
GAP
_
1 N
i
, S
_
MOD 1 GAP (NP
i
[acc])
_
This rule means the English relative phrase is the combination of a N
with a postnominal S
which can have either no GAP value or a GAP value whose case is accusative and index value
is identical with the head N
V CP[GAP NP]
V *NP[GAP NP]
claim Det N
[GAP NP]
V
*NP
_
_
HEAD| POS noun
SPR
COMPS
GAP NP
_
refute Det N
[GAP NP]
_
HEAD| POS noun
SPR
COMPS
_
_
GAP
_
This licensing condition means that an expression whose POS value is noun and whose SPR
and COMPS value are discharged must have no GAP value.
In a similar manner, the SSC (sentential subject constraint) can be interpreted as a constraint
on the Head-Specier Rule:
(71) Head-Specier Rule:
XP 1
_
GAP
_
, H[SPR
_
1
_
]
This constraint means that the specier of a head must have an empty GAP value. This assigns
the following structure to the unacceptable sentence (63b):
(72)
S
2 NP
S
_
_
SPR
COMPS
GAP 2
_
Who V
CP
_
_
SPR
COMPS
GAP 2
_
AP
_
SPR 1
_
extremely unlikely
Notice that this constraint can also account for the Left-Branching Constraint (LBC):
(73) a. You saw the presidents guard.
b. *Whose did you see guard?
119
This constraint is a soft constraint which can be overridden.
237
In (73b), the specier of the head guard is gapped, violating the licensing rule in (71).
The AC (Adjunct Constraint) is predicted since only an argument can be gapped. However,
notice that the following famous parasitic gap examples allow a gap in the adjunct when the
head has the identical gap:
(74) a. Which book did she review without reading it?
b. *Which book did she review it without reading ?
c. Which book did she review without reading ?
As seen in (74b), an element in the without adjunct clause cannot be wh-questioned. However,
as in (74c), this gap in the modier clause can be licensed when the head VP that the clause
modies has the identical gap. The second gap in the adjunct clause is thus considered to be
parasitic since this second gap (unlike the rst gap) cannot easily stand on its own. The one
simple way of implementing this constraint is to treat the GAP as a HEAD feature as well:
120
This means when the modier has a GAP value, the head it modies will also must have the
identical feature GAP value. This is due to the GAP value is a HEAD feature. This in turn means
whenever a modier has a GAP, its head will also have the identical GAP value, enforcing a
modier cannot have a GAP value by itself. This explains the difference between the following
two tree structures:
(75) a.
*VP
VP
_
GAP
_
PP
_
GAP 1 NP
_
VP
_
GAP 1 NP
_
PP
_
GAP 1 NP
_
_
HEAD| POS verb
COMPS
_
S
_
_
QUE +
GAP
_
_
_
_
_
This means that a verb selecting an indirect question as its complement requires its sentential
complement to have an empty GAP value. This is why the following structure is ill-formed:
(77)
*VP
V
_
_
COMPS
_
S
_
_
QUE +
GAP
_
_
_
_
_
S
_
_
QUE +
GAP NP
_
_
_
HEAD| POS adj
VAL
_
_
SPR NP
i
COMPS
_
_
_
VFORM inf
GAP 1 NP
i
_
_
_
_
TO-BIND| GAP 1
_
_
b. eager-type adjectives
_
_
HEAD| POS adj
VAL
_
_
SPR NP
COMPS VP
_
VFORM inf
_
_
_
_
_
The lexical entry in (11a) species that the innitival complement (VP or S) of easy contains
a GAP value which is coindexed with the subject. Meanwhile, that of eager in (11b) is just
a complete innitival VP with no missing element. These lexical differences will assign the
following structures to (4a) and (4b), respectively:
245
(12) a.
S
NP VP[GAP ]
.
.
.
.
.
.
.
.
.
.
Kim
V
AP[ GAP ]
is
A
_
B-GAP 1
_
VP[ GAP 1 ]
easy V VP[GAP 1 ]
to V[GAP 1 ]
please
b.
S
NP
VP
Kim
V AP
is A
VP
eager V
VP
to V
please
As noted in (12a), the VP complement of easy has a GAP value. Notice that the lexical entry
easy in (11a) also species that the VPs GAP value is identical with its own TO-BIND|GAP
(B-GAP) value. This ensures that the VPs GAP value is lexically discharged. Meanwhile, the
VP complement of eager is already saturated.
The present analysis then also licenses sentences like the following in which the subject of
the innitival VP complement appears:
(13) a. Kim is easy for us to please.
b. Kim is easy for us to make Lee accept.
246
In both cases, easy combines with an innitival CP with one GAP value.
The present analysis then can also predict examples in which the VP complement includes
more than one GAP elements. Compare the following pair of sentences:
(14) a. This sonata is easy to play on with the piano.
b. Which piano is this sonata easy to play on with ?
The present analysis assigns the following structure to (14a):
(15)
S
1 NP
i
VP
_
SPR 1
_
This sonata V
AP
_
_
SPR 1
GAP
_
_
.
.
.
.
.
.
.
.
is
A
_
_
SPR 1
COMPS 3
B-GAP 2
_
_
VP
_
_
SPR 1
GAP 2
_
_
easy V
3 VP
_
GAP 2
i
_
...........
to play on with the piano?
As given in (11a), the adjective easy selects a VP complement with one GAP value coindexed
with its subject. Its lexical information ensures this GAP value to be discharged. In the structure,
thus, the GAP value started from the object position of on is percolated up to the VP where it is
lexically bound by the feature B-GAP (TO-BIND|GAP) of easy.
Now consider the structure of (14b) in which the object of on is linked to the subject this
sonata and the object of with is wh-questioned:
247
(16)
S
.
.
.
.
.
.
.
.
.
.
.
2 NP
S
_
GAP 2
_
.
.
.
.
.
.
.
.
.
.
.
Which piano V
1 NP
AP
_
GAP 2
_
is
this
sonata
Adj
_
_
SPR 1
COMPS 3
B-GAP 4
_
_
3 VP
_
GAP 4
i
, 2
_
3 NP
VP
It [t]
VP
V NP
you came early
surprised me
This movement process also introduces the insertion of that. To generate nonextraposed sen-
tences like That you came early surprised me, the system posits a process of deleting it in (24a)
and then adding the complementizer that.
249
A slightly different analysis has also been suggested with the difference in the direction of
movement. That is, instead of extraposing the clause from the subject, we might assume that
the clause is already in the extraposed position as in (26a):
(26) a. [[ ]
V P
[surprised [me]
S
[that you came early]]].
b. [[It]
V P
[surprised me that you came early]].
As given in (26a), the extraposed S is base-generated within the VP. When we insert the exple-
tive it in the subject position here, we generate (26b). When this clause is moved to the subject,
we would have the nonextraposed sentence That you came early surprised me.
Most current movement approaches follow this second line of assumption. Though such
derivational analyses can capture certain aspects of English subject extraposition, they are not
enough to predict lexical idiosyncrasies as well as non-local properties of extraposition.
124
12.3.3 A Lexicalist Analysis
One obvious constraint in subject extraposition we can observe is that even though there exist
systematic relationships between extraposed and nonextraposed pairs, we also nd examples
with no non-extraposed counterparts:
(27) a. *[That Pat is innocent] proves.
b. It proves [that Pat is innocent].
(28) a. *[That Sandy had lied] suggested.
b. It suggested [that Sandy had lied].
As observed here, verbs like appear, happen, chance, intend, and fall do not allow non-
extraposed counterparts. This kind of lexical idiosyncracy implies that subject extraposition
is lexically controlled. As a way of formally representing the systematic relationship depending
on the lexical property of verbs, we can introduce the following lexical rule with the feature
EXTRA:
(29) Extraposition Lexical Rule:
_
ARG-ST ..., 1 [IND prop],...
_
_
_
ARG-ST ...,NP[NFORM it],...
EXTRA 1
_
_
What this rule means is that if a verbal element (adjective or verb) selects a clausal argument
(whose meaning is a proposition prop) its propositional argument can be realized as the value
of the feature EXTRA together with the introduction of the expletive argument it.
For example, consider the following data set:
(30) a. Fidos barking annoys me.
b. That Fido barks annoys me.
c. It annoys me that Fido barks.
124
See Kim and Sag (2005).
250
As given here, the verb annoys can take either a CP or an NP as its subject. This means that
when the verb annoys select a CP subject, it can undergo the Extraposition Lexical Rule in (29)
as following:
(31)
_
_
annoys
ARG-ST 1 CP[prop] , 2 NP
_
_
_
_
ARG-ST NP[NFORM it], 2 NP
EXTRA 1 CP
_
_
The output annoy now selects the expletive it as subject and an object NP together with the
original CP as an EXTRA element. This output verb will then allow us to generate a structure
like the following:
(32)
S
3 NP
VP
_
_
SPR 3
EXTRA
_
_
It
VP
_
_
SPR 3
EXTRA 1
_
_
1 CP
V
_
_
SPR 3
COMPS 2
ARG-ST 3 NP[it], 2
EXTRA 1
_
_
2 NP
that Fido barks.
annoys me
As noted in the tree, the two arguments of the verb annoys are realized as SPR and COMPS
respectively. When it combines with the NP me, it forms a VP with the nonempty EXTRA
value. This VP then combines with the extraposed clause CP in accordance with the following
Head-Extra Rule, forming a complete VP:
(33) Head-Extra Rule:
251
_
EXTRA
_
H
_
EXTRA 1
_
, 1
As noted, the rule also discharges the feature EXTRA passed up to the head position. This gram-
mar rule reects the fact that English independently allows a phrase in which a head element
combines with an extraposed element as represented in the following:
(34)
_
EXTRA
_
H
_
EXTRA 1
_
1
We can observe English freely employs this kind of well-formed phrase condition even in the
extraposition of an adjunct element.
(35) a. [[A man came into the room] [that no one knew]].
b. [[A man came into the room] [with blond hair]].
c. I [read a book during the vacation [which was written by Chomsky]].
All these examples are licensed by the Head-Extra Rule that allows the combination of a head
element with an extraposed element.
Object extraposition is no different. For example, consider the following examples:
(36) a. Ray found the outcome frustrating.
b. Ray found it frustrating [that his policies made little impact on poverty].
The data mean that the lexical entry for nd that selects three arguments can undergo the Extra-
position Lexical Rule:
(37)
_
_
nd
ARG-ST 1 NP, 2 XP, 3 AP
_
_
_
_
ARG-ST 1 , NP[it], 3 AP
EXTRA 2 [prop]
_
_
As observed, when the second argument XP is realized not as an NP but as a clause with the
semantic type of proposition, the verb yields an extraposed output. This output then can generate
sentences like (36) as represented in the following simplied structure:
252
(38)
VP
_
EXTRA
_
.
.
.
.
.
.
.
.
.
.
.
.
.
VP
_
EXTRA 2
_
.
.
.
.
.
.
.
.
.
.
.
.
.
2 CP
V
_
_
ARG-ST 1 , NP[it], 3 AP
EXTRA 2
_
_
NP 3 AP
that....
found it frustrating
The verb found requires an expletive object and an AP as its complement. It also has a clausal
element as its EXTRA element. The rst VP thus has a nonempty EXTRA value projected from
the verb. This VP then forms a well-formed phrase with the extraposed CP clause.
One additional language independent constraint relevant in the extraposition is that English
independently prohibits a CP from having any element to its right stated:
(39) a. *Would [that John came] surprise you?
b. Would it surprise you [that John came]?
(40) a. I believe strongly [that the world is round].
b. *I believe [that the world is round] strongly.
This constraint basically bars any argument from appearing after a sentential argument:
(41) Ban on the Non-sentence Final Clause (BNFC):
English prohibits a clausal element (CP or S) from having any element to its right
position.
In the present context this means that there exists no word whose COMPS list contains some-
thing after its CP complement.
This constraint, combined with the present analysis of extraposition, can explain the follow-
ing contrast:
(42) a. *I made [to settle the matter] my objective.
b. I made it [my objective] to settle the matter.
c. I made [the settlement of the matter] my objective.
(43) a. *I owe [that the jury acquitted me] to you.
253
b. I owe it [to you] that the jury acquitted me.
c. I owe [my acquittal] to you.
Verbs like made and owe here take an object and an obligatory predicative phrase (NP for made
and PP for owe) as their arguments. This means that when the object is realized as a CP, it must
be extraposed to the sentence nal position not to violate the BNFC constraint.
12.4 Cleft constructions
12.4.1 Basic Properties
The examples in (44) represent the canonical types of three kinds of cleft constructions: it-cleft,
wh-cleft, and inverted wh-cleft in English:
(44) a. It-cleft: In fact its their teaching material that were using.
b. Wh-cleft: What were using is their teaching material.
c. Inverted wh-cleft: Their teaching material is what we are using.
These three types of clefts all denote the same proposition as the following simple declarative
sentence:
(45) We are using their teaching material.
The immediate question that follow is then why are we using clefts instead of the simple
sentence (45)? It is commonly accepted that these three types of clefts share the identical
information-structure properties given in (46):
(46) a. Presupposition (Background): We are using X.
b. Highlighted (Foreground): their teaching material
c. Assertion: X is their teaching material.
In terms of the structures, these three types of cleft constructions all consist of a matrix clause
headed by a copula and a relative-like cleft clause whose relativized argument is coindexed
with the predicative argument of the copula. These structural properties can be represented in
the formula as given in the following table:
(47) Three main types of cleft constructions:
Types of cleft Formula
(a) It-cleft It + be + XP
i
+ Cleft clause
(b) Wh-cleft Cleft clause + be + XP
i
(c) Inverted wh-cleft XP
i
+ be + Cleft Clause
The choice of one rather than another of these three clefts is determined by various formal and
pragmatic factors, some of which we will look into here.
12.4.2 Distributional Properties of the Three It-clefts
It-cleft Constructions: As given in (47a), the it-cleft construction has the pronoun it as the
subject of the matrix verb be, the highlighted phrase XP, and a cleft clause. The pronoun it here
254
functions just as a place holder though it is similar to the referential pronoun it. For example, it
is hard to claim that it in the following dialogue has any referential property:
(48) A: I share your view but I just wonder why you think thats good.
B: Well I suppose its the writer that gets you so involved.
As for the type of foreground XP, we observe that only the limited phrases are used:
(49) a. It was [
NP
the gauge] that was the killer in the rst place .
b. It was [
AdvP
then] that he felt a sharp pain.
c. It was [
PP
to Stanford] that he gave his full loyalty.
d. It wasnt [
S
till I was perhaps twenty-ve or thirty] that I read and enjoyed them.
Phrases such as an innitival VP, AP, or CP cannot function as the XP:
(50) a. *It was [
VP
to nish the homework] that John tried.
b. *It is [
AP
fond of Bill] that John seems to be.
c. *It is [
CP
that Bill is honest] that John believes.
The wh-word that introduces the cleft-clause ranges from that to who and which:
(51) a. Its the second Monday [that] we get back from Easter holiday.
b. It was the girl [who] kicked the ball.
c. Its mainly the content [which] differs rather than the actual language itself.
WH-cleft Constructions: Unlike the it-cleft construction, the wh-cleft construction places
a cleft clause in the subject position followed by the highlighted XP in the postcopular position.
There exists a wide ranger of highlighted types. As given in (52), almost all the phrase types
can serve as the highlighted XP phrase in the wh-cleft:
(52) a. What you want is [
NP
a little greenhouse].
b. Whats actually happening in London at the moment is [
AP
immensely exciting].
c. So what is to come is [
PP
in this document].
d. What Ive always tended to do is [
VP
to do my own stretches at home].
e. What I meant was [
CP
that you have done it really well].
Different from it-clefts, the wh-cleft construction allows AP, base VP, and clauses (content
clause as in (51), pure S, and even an wh-clause) to serve as the highlighting XP:
(53) a. What you do is [
VP
wear it like that].
b. What happened is [
S
they caught her without a licence].
c. What the gentleman seemed to be asking is [
S
how policy would have differed].
Inverted wh-cleft constructions: Though the inverted wh-cleft construction is similar to the
wh-cleft, not many types can be highlighted:
(54) a. [
NP
That] is what theyre trying to do.
255
b. [
S
What one wonders] is what went on in his mind.
c. [
AP
Insensitive] is how I would describe him.
d. [
PP
In the early morning] is when I do my best research.
(55) a. *[
VP
Wear it like that] is what you do.
b. *[
S
They caught her without a license] is what happened.
c. [
CP
That you have done it really well] is what I meant.
The inverted wh-cleft also can be introduced with the head noun thing, all, or one:
(56) a. [The last thing I want to do] is to put you to any more trouble personally.
b. [All I had to do] was heat it up.
In terms of the cleft clause type, all the wh-words, except which, are possible:
(57) a. Thats when I read.
b. That was why she looked so nice.
c. Thats how they do it.
d. Thats who I played with over Christmas.
12.4.3 Syntactic Structures of the Three Clefts
As noted before, the three types of clefts all provide unique options for presenting salient
discourse information in a particular serial order. Each of these three types have different syn-
tactic properties which make it hard to drive one from the other. For example, one noticeable
difference lies in that only wh-clefts allow bare innitives as the highlighted XP phrase:
(58) a. What you should do is [
VP
order one rst].
b. *It was [
VP
order one rst] that you should do rst.
c. *[
VP
Order one rst] is what you should do.
The three are different with respect to the occurrence of an adverbial subordinate clause, too:
(59) a. It wasnt till I was perhaps twenty-ve or thirty that I read them and enjoyed
them.
b. *When I read them and enjoyed them was not until I was perhaps twenty-ve.
c. *Not until I was perhaps twenty-ve was when I read them and enjoyed them.
As noted here, the not until adverbial clause appears only in it-clefts.
It is not difcult to nd no isomorphic relationships among the three clefts. For example,
neither wh-clefts nor inverted wh-clefts allow the cleft clause headed by that:
(60) a. Its the writer [that gets you so involved].
b. *[That gets you so involved] is the writer.
c. *The writer is [that gets you so involved].
In addition, only the cleft clause of if-clefts can have the PP wh-head:
256
(61) a. And it was this matter [on which I consulted with the chairman of the Select
Committee].
b. *[On which I consulted with the chairman of the Select Committee] was this
matter.
c. *This matter was [on which I consulted with the chairman of the Select Com-
mittee].
The lack of such isomorphic relations among the three clefts indicates that the three clefts
have no strong syntactic closeness. These do not mean that there exist no commonalities. In
terms of its argument structure, it is obvious that the cleft copula be selects two arguments
which refer to the identical individual:
125
(62)
_
_
be
ARG-ST
_
NP
i
, XP
i
_
_
_
These two arguments will canonically be realized as SPR and COMPS in syntax:
(63) Canonical Argument Realization of be:
_
ARG-ST
_
1 NP
i
, 2 XP
i
_
_
_
SPR
_
1 NP
i
_
COMPS
_
2 XP
i
_
_
_
Such an argument realization will generate canonical specicational sentences like the follow-
ing:
(64) a. The recipient of this years award is President Kim.
b. The one who broke the window was Mr. Kim.
However, there are various different ways of argument realization, depending on how the
information structure (IS) is realized. That is, the three types of clefts reect how the arguments
are realized differently with respect to the information structure of the sentence in question.
Two common information structure sensitive features are TOPIC and FOCUS which are usually
linked to given and newinformation, respectively. In addition to these two features, we introduce
the feature HIGHLIGHT. The feature HIGHLIGHT is similar to the notion of salient: the
information that is most salient in the given context bears this feature. Consider the following
simple question and answer dialogue:
(65) A: What did John drink?
B: John drank beer.
It is clear that in the expressions John and drank here are both given information (topic),
whereas beer is new information (focus). The difference between John and drank is just
125
The copula in the cleft construction is specicational, not predicational. In sentences like John is happy, the
copula is used as predicational, whereas in sentences like The culprit is John, the copula is specicational. One main
difference is that in the former the postcopular element denotes the property of the subject whereas in the latter it
denotes an individual. See Heycock and Kroch (1999).
257
that John is more salient than drank since it is what the sentence is about. This kind of
comparison also holds between completive (pure) focus and contrastive focus:
(66) A: Did John drink beer or coke?
B: John drank beer.
Unlike the NP beer in (65), the NP beer here is focus, but has a contrastive meaning com-
pared to coke. In this sense, we call beer is contrastive focus, the most salient information in
this given discourse. The feature HIGHLIGHT is thus given to the topic and contrastive focus.
The feature thus can be assigned either to a TOPIC or to a FOCUS expression. These three
features are called information-structure (INFO-ST), distinguished from phonology (PHON),
syntax (SYN), and semantics (SEM) information. Given these, the contrastive focus phrase beer
in (66)B will have the following information:
126
(67)
_
_
PHON
_
beer
_
SYN| HEAD| POS noun
SEM
_
_
IND i
RELS
_
_
_
PRED beer-rel
ARG1 i
_
_
_
_
_
INFO-ST
_
_
HIGHLIGHT +
FOCUS +
_
_
_
_
The feature structure means that the nominal element beer refers to an individual i with a beer-
relation. This expression is also used as a highlighted focus.
Equipped with this system, we can assume that depending on the realization of these three
IS features, TOPIC, FOCUS, and HIGHLIGHT, we have different cleft constructions.
Lets start with wh-clefts. We assume that wh-clefts reect the following argument realiza-
tion of the specicational be:
(68) Argument Realization for the Wh-cleft Formation:
_
ARG-ST
_
1 , 2
_
_
_
SPR
_
1 NP
i
_
_
FREL +
HIGHLIGHT +
TOPIC +
_
_
_
COMPS
_
2 XP
i
[FOCUS +]
_
_
_
The two arguments of be are realized as SPR and COMPS in order. The subject here also is
TOPIC as well as HIGHLIGHT, functioning as the salient element in the discourse. The coin-
dexing relation between the two arguments ensures that the COMPS element species the prop-
erty of the subject. In addition, the highlighted subject carries the feature FREL (free relative)
126
See Engdahl and Vallduv (1996) for the arguments introducing the INFO-ST level.
258
which is assigned to wh elements like what, when, and where, but not to why or how since they
cannot serve as the head of a free relative clause as seen from the following contrast:
(69) a. He got what he wanted.
b. He put the money where Lee told him to put it.
c. The concert started when the bell rang.
(70) a. *Lee wants to meet who Kim hired.
b. *Lee bought which car Kim wanted to sell to him.
c. *Lee solved the puzzle how Kim solved it.
In the examples in (69), what, where and when can head the free relative clause in the sense that
they are interpreted as the thing that, the place where, and the time when. However, this kind
of interpretation is not possible with who, which or how:
127
(71) a. *Who achieved the best result was Angela.
b. *Which book he read was this.
Given the output in (68), we then can generate a structure like the following:
(72)
S
NP
i
_
_
FREL +
TOP +
HIGHLIGHT +
_
VP
NP
i
[FREL +] S/NP
i
V NP
i
[FOC +]
.
.
.
.
.
.
.
.
.
.
_
SPR
_
2 XP
i
_
_
HIGHLIGHT +
TOPIC +
_
_
_
COMPS
_
1 NP
i
_
_
_
As noted here, unlike the wh-clefts, the discourse salient and highlight information is the second
argument. This second argument is also functioning as a topic, as attested by the unnaturalness
of an indenite NP in the position:
(78) a. #A question is what we have been trying to answer.
b. #A book is what I recommended to you.
Since topic comes rst in natural discourse, it will be realized as SPR, generating the following
structure:
260
(79)
S
NP
i
_
_
TOPIC +
HIGHLIGHT +
_
_
VP
is
NP
_
FREL +
_
S/NP
i
_
be
SPR NP[it]
COMPS 2 YP
_
HIGHLIGHT +
_
EXTRA CP[GAP 2 ]
_
_
This lexical realization introduces the expletive it as the subject, the contrastive focus as the
HIGHLIGHT element together with placing the rst argument in the extraposition. This work is
done through the feature EXTRA, adopting the treatment of it-clefts as an extraposition process
(cf. Akamajian 1970, Emonds 1976, Gundel 1978, among others).
128
Notice that unlike wh-
clefts, the extraposing cleft clause has no restriction on the feature FREL. This ensures that
even a content clause can function as a cleft clause:
(81) a. *That you heard was an explosion.
b. It was an explosion that you heard.
The output in (80) will then generate a structure like the following:
128
See Kim and Sag (2005) for the detailed discussion of English extraposition constructions.
261
(82)
S
NP
i
VP
_
GAP
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
It VP[EXTRA
_
1
_
]
1 S
_
GAP NP
i
V
NP
i
_
HIGHLIGHT +
_
.
.
.
.
.
.
.
.
.
.
that we used
is their teaching material
This structure is different from wh-clefts in that the HIGHLIGHTED expression is a contrastive
focus. In addition, the value of the feature EXTRA is discharged by the grammar rule in (83):
(83) Head-Extra Rule:
_
EXTRA
_
H
_
EXTRA 1
_
, 1
There are several facts that support such a structure in which the cleft-clause is not a comple-
ment of the copula and but is extraposed to the sentential nal position. For example, consider
the following:
(84) a. It was the boy, I believe, who brought the letter.
b. It was in the church, presumably, where he married her.
As given here, a parenthetical or an adverb can intervene between the highlighted XP and the
cleft clause. If the XP and the cleft clause are both complements of be, such data are not ex-
pected. In addition, consider the following coordination data:
(85) a. *It was [beer that Kim drank] and [tango that Lee danced].
b. It [was beer that Kim drank] and that Mary tasted.
As observed here, the XP and the cleft clause do not form a constituent. This is what the present
analysis predicts.
In addition, the present analysis, in which the cleft clause is not a complement of the copula
verb but but a modier to the VP, can predict the difference between canonical sentential com-
plement and cleft-clause. Observe that unlike the sentential complement, the cleft clause does
not allow its element to be extracted:
(86) a. Which book do you think John put in the box?
262
b. *Which book is it John that put in the box?
This kind of ungrammatical sentence cannot be blocked if we simply assume that be selects a
CP as its complement.
Notice that it-clefts have two different subtypes. The examples we have discussed so far
contain a syntactic gap in the cleft clause. However, there are it-clefts that do not have any
syntactic gap in the cleft clause. Compare the following:
(87) a. It is Bill [
CP/NP
that John relies on ].
b. It is Bill [
S
[on whom] [John relies]].
Even though (87a) the cleft clause contains a gap, the cleft clause is a complete sentence when
John relies combines with on whom. More clear examples will be those with an adjunct element
being highlighted:
(88) a. It was then when we all went to bed.
b. It was only gradually that I came to realize how stupid I was.
As a way of explaining this second type of it-clefts, we assume that there is another realization
of the copula for it-clefts:
(89)
_
_
be
SPR NP[it]
COMPS 2 YP
_
HIGHLIGHT +
_
EXTRA
_
S
_
_
MOD
_
2
_
GAP
_
_
_
_
_
This means that the EXTRA sentence is modifying the hightlighted YP element. Other than this,
there is no tight dependency between the cleft clause and the highlighted YP. This realization
will then project a structure like the following:
263
(90)
S
NP
i
VP
_
GAP
_
It
2 VP[EXTRA 1 ]
1 S
_
MOD 2
_
V
NP
i
_
HIGHLIGHT +
_
on whom we rely
is Tom
(91)
S
NP
i
VP
_
GAP
_
It
2 VP[EXTRA 1 ]
1 S
_
MOD 2
_
V
NP
i
_
HIGHLIGHT +
_
-movement, 195
accusative, 175, 177, 233, 243
adjective, 3, 12, 13, 24, 25, 43, 81, 91, 92, 101,
121, 245
attributive, 121
control, 128
predicative, 121
raising, 128
tough, 244
adjunct, 41, 43, 54, 235, 252
Adjunct Clause Constraint (ACC), 235
adverb, 12, 2426, 41, 42, 148, 158, 166, 181,
190, 262
adverbial, 41, 181, 256
affectedness, 183, 184
Afx Hopping Rule, 150
agent, 36, 37, 4446, 48, 65, 139, 141, 173,
184, 188
AGR (agreement), 65, 102104, 106, 107
agreement, 18, 31, 37, 101, 155
index, 106
mismatch, 108
morphosyntactic, 106, 111
noun-determiner, 102
pronoun-antecedent, 104
subject-verb, 7, 43, 56, 104, 109
ambiguity
structural, 8, 18
anaphor, 111
antecedent, 101
ARG-ST (argument-structure), 6567, 73, 90,
145
Argument Realization Constraint (ARC), 66,
67, 84, 199, 257
article, 5, 14, 57, 119
atomic, 62
attribute, 6264, 77
attribute-value matrix (AVM), 62
autonomous, 17
auxiliary verb, 15, 23, 28, 37, 180, 193, 203
benefactive, 39, 44
BNFC Constraint, 253
British English, 154
Chomsky, Noam, 1, 3, 36, 84, 149, 243
clausal
argument, 250
complement, 15, 85, 92
subject, 90, 235
clause, 11
cleft, 19, 243, 244
inverted wh-cleft, 254
it-cleft, 254
wh-cleft, 254
combinatory
properties, 49
requirement, 49
rules, 2
competence, 24
complement, 15, 41, 51
complement clause, 14
281
complementation pattern, 56, 70
complementizer, 14, 15, 30, 85, 87, 88, 205,
229, 249
Complex Noun Phrase Constraint (CNPC), 235
COMPS (complements), 66
conjunction, 14, 21
coordinating, 14
subordinating, 14
constituent, 11, 1821, 24, 35, 49, 88, 173, 248,
262
constituent question, 19
context dependent, 106
context free grammar, 26
contraction, 147, 148, 164
Coordinate Structure Constraint (CSC), 235
coordination, 21, 88, 150, 236, 262
Coordination Rule, 30, 236
copula, 82, 153, 154, 254, 257, 260
COUNT (countable), 117
count nouns, 4, 6, 7, 9, 98
creativity, 2
criteria
distribution, 12
meaning, 12
morphological, 12
syntactic function, 12
declarative, 77, 79, 193
deep structure, 127
DEF (denite), 115
denite, 114
demonstrative, 14, 57
descriptive rules, 4
determiner, 14, 57, 73, 97, 98, 102, 105, 110,
112, 118
discourse, 171, 172, 244, 260, 265
distributional criteria, 13
Do-so Replacement Condition, 53
do-so test, 53
Do-support, 150
DP: Determiner Phrase, 58
ellipsis, 148
endocentricity, 55, 57, 61
English Declarative Sentence Rule, 51
exclamative, 162, 193
experiencer, 44, 46, 141, 142
expletive, 83, 128, 131, 143, 156, 250, 251,
261, 266
external syntax, 49
EXTRA, 250
extraction, 203205
extraposition, 243
Extraposition Lexical Rule, 250
feature, 6163
name, 63
specication, 24
structure, 6264
system, 61
unication, 64
feature percolation, 195
feature structures, 62
Ferdinand de Saussure, 1
ller, 194, 195, 197, 201, 202, 219
nite, 2, 14, 23, 74, 76, 77, 90, 151, 159, 160,
165, 207
nite clause, 14
oating quantier, 148, 153
FOCUS, 264
fragment, 19, 23
FREL (free-relative), 259
GAP, 198, 206
gap, 194, 195
GAP Inheritance Principle, 198
gapping, 147
GEN (gender), 104
generative grammar, 4
generative linguistics, 1
generative syntax, 147
generativity, 27
generic, 101
goal, 39, 44, 69, 70
grammatical category, 11
grammatical competence, 3
grammatical function, 35, 42, 49
habitual, 101
282
head, 7, 15, 50, 51
head feature, 73, 74, 78, 166, 238
Head Feature Principle (HFP), 74, 75, 82, 179,
202, 231
Head-Complement Rule, 59, 7375, 80, 82
Head-Extra Rule, 251, 252
Head-Filler Rule, 75, 199, 202, 210, 232, 248
Head-Modier Rule, 59, 60, 73, 74, 179
Head-Only Rule, 100, 214, 231
Head-Specier Rule, 59, 61, 73, 74, 76, 92,
178, 179, 201, 216, 237
headedness, 50, 55, 61
hierarchical structure, 29, 59
HIGHLIGHT, 261
HPSG, 62
hypothesis, 6
imperative, 157, 162, 193, 213
IND (index), 106
indirect question, 208, 212, 213, 215, 235, 239
Indirect Wh-question Constraint (IWC), 235
innitival
clause, 14, 15, 230
CP, 91, 232, 247
marker, 15, 82
relative clause, 219, 232
S, 88
VP, 88, 127, 129, 130, 132, 136, 143, 214,
234
wh-relative, 231
INFO-ST (information-structure), 258
information
argument, 62
semantic information, 62
syntactic , 62
information: phonological, 62
instrument, 36, 45
interability, 52
intermediate category, 57, 59, 71, 195
internal syntax, 49
Kleene Star Operator, 17
Left-Branching Constraint (LBC), 235
lexical category, 11, 16
lexical idiosyncrasy, 113
lexicalist, 176
lexicon, 17, 46, 62
LFG, 62
location, 41, 45
locative, 36
long distance, 219
long distance dependency, 194
manner, 41
maximal phrase, 51
meaning preservation, 130
middle voice, 187
minimal phrase, 51
modals, 88
modier, 41, 51, 55, 61, 73, 121, 122, 219, 226
morphological criteria, 13
negation, 15, 147, 148, 156, 158, 159, 164
constituent, 158
sentential, 159
NFORM, 83
NICE properties, 148
nominal, 89
nominative, 175, 177, 243
non-count nouns, 4
nonnite, 76
nontransformation, 142, 176
noun, 12
collective, 109
common, 97
countable, 97
measure, 119
partitive, 112
pronoun, 97, 101
proper, 97, 102
uncountable, 97
NUM (number), 61, 106
object, 35
obligatoriness, 52
oblique complement, 40
ontological issue, 147
ordering restriction, 148
283
particle, 16, 20, 182
parts of speech, 11
passive
adjectival, 184
get, 185
prepositional, 180
semi, 185
Passive Lexical Rule, 177
passivization, 38, 39, 131, 180
past, 77
past participle, 77
patient, 36, 44, 46, 139
PER (person), 108
performance, 3
PFORM, 81, 96, 182, 185
PHON (phonology), 65
phrasal category, 19
phrasal verb, 30
Phrase Structure Rule (PS) Rule, 22
POS (part of speech), 61
position
of adverb, 153
possessive, 14, 58, 98, 99
PRD (predicate), 68, 121, 154
predicate, 23, 35, 36, 40, 41, 57, 66, 70, 121,
129, 131, 244
tough, 244
predicational, 257
predicative complement, 39
preposition, 8, 13, 20, 81, 94, 114, 115,
180182
grammatical, 114
predicative, 114
prepositional verb, 30, 180, 182
prescriptive rules, 4
present, 77
present participle, 77
PRO (pronoun), 133
projection, 50
pronoun, 21
proposition, 141, 171
PS rules, 24, 27, 29, 30, 35, 55, 56, 59, 61, 149
quantier, 14
QUE (question), 179
question, 193
raising properties, 173
reason, 41
recipient, 39
recursive application, 27
redundancy, 56
reexive, 101, 111
REL, 222, 223, 228
relative
nonrestrictive, 220
relativizer, 230
restrictive, 224
Sag, Ivan A., 3, 46, 52, 55, 62, 78, 159, 243,
250, 261
selectional restriction, 130
SEM (semantics), 65, 106
semantic criteria, 12
semantic restriction, 53
semantic role, 36, 128
Sentential Subject Constraint (SSC), 235
situation, 171
sounds and meanings, 1
source, 44
specicational, 257
specier (SPR), 57
SPR (specier), 66
stand-alone test, 19
structural
change, 174
description, 174
difference, 54
position, 14
structure sharing, 63, 105
subcategorization, 68, 129
subject, 35
Subject Extraction Lexical Rule, 206
subject-auxiliary inversion, 37, 148
subject-object asymmetry, 199
subjecthood tests, 37
substitution, 21, 60
subsumption, 64
284
SYN (syntax), 65
syntactic
category, 35, 49
tag question, 148, 161, 164
tag questions, 37
theme, 4446, 48, 65, 69, 141, 188
TO-BIND, 247
TOPIC, 264
topicalization, 1
tough, 243
transformation, 113, 149, 162, 174176, 181,
249
underspecication, 88, 103
unication, 64, 65, 196
VAL (valence), 66, 67, 130
Valence Principle (VALP), 75, 178, 231
verb, 12
complex transitive, 70
diransitive, 40
ditransitive, 56, 69
equi, 127
indirect question, 239
intransitive, 45, 56, 67, 187, 244
linking, 68, 70
transitive, 56, 68, 90, 175177, 180, 187
VFORM, 51, 65, 74, 77
VP
nite, 51
innitival, 43
nonnite, 158, 226
VP ellipsis, 157, 159, 160, 165
VP Ellipsis Rule, 165
wh-question, 20, 195, 203, 204, 222, 231, 235
wh-relative pronoun, 227
wh-subject, 204
word order, 1
X
theory, 59
285
This new textbook, focusing on the descriptive facts of English, provides a systematic introduc-
tion to English syntax for the students with no prior knowledge of English grammar or syntactic
analysis. The textbook aims to help students to appreciate the various sentence patterns avail-
able in English, understand insights into core data of English syntax, develop analytic abilities
to further explore the patterns of English, and learn precise ways of formalizing syntactic analy-
sis for a variety of English data and major English constructions such as agreement, raising and
control, the auxiliary system, passives, wh-questions, relative clauses, extraposition, and clefts.
Jong-Bok Kim is Associate Professor of School of English at Kyung Hee University, Seoul,
Korea. Peter Sells is Professor of Linguistics and Asian Languages at Stanford University.
287