1 Foundational issues

Prescriptive versus descriptive grammar

Rule formation and syntactic structure in language acquisition
o A thought experiment
o Rule-based word formation
o Question formation
More evidence for syntactic structure
o Intuitions about words belonging together
o Structural ambiguity
Universal Grammar
o Formal universals
o Recursion
o Parameters
Generative grammar
o Elementary trees and substitution
o Grammaticality
o Grammar versus language
Notes
Exercises and problems
Supplementary material
o Expletive elements in English
o Modals and auxiliary verbs in English
o Questions

This book is an introduction to generative grammar from a Chomskyan perspective. By the time
you finish this chapter, you will have a clearer understanding of what we mean by this sentence,
and by the time you finish the entire book, your understanding of it will be clearer and deeper
still. But for the moment, you have probably gained the impression that this book is about
grammar of some sort. And right there, we have a problem. The problem is that there is an
everyday sense of the term 'grammar' and a quite different sense in which the term is used in
linguistics.

Prescriptive versus descriptive grammar

In the everyday sense, 'grammar' refers to a collection of rules concerning what counts as
socially acceptable and unacceptable language use. Some of these rules, like the ones in (1),
make reference to particular words and apply to both spoken and written language.

(1) a. Don't use ain't.

 b. Don't use seen as the past tense of see (as in I seen him at the party last night).

c. Don't use contractions.

But mainly, the rules in question concern the proper composition of sentences in written
language. You may recall being taught rules at school like those in (2).

(2) a. Don't start a sentence with a conjunction.

b. Don't use sentence fragments.

c. Don't end a sentence with a linking verb.

d. Don't use dangling participles.

e. Don't end a sentence with a preposition.

f. Don't use an object pronoun for a subject pronoun in a conjoined subject.

g. Don't use a plural pronoun to refer back to a singular noun like everyone, no-one, someone, and
the like.

h. Don't split infinitives.

i. Use whom, not who, as the object of a verb or preposition.

Someone who composes sentences in accordance with rules like those in (2) is said to have good
grammar, whereas someone said to have bad grammar doesn't apply the rules when they ought to
be applied1 and so produces sentences like (3).

(3) a. Over there is the guy who I went to the party with. violates (2f), (2j)

b. Bill and me went to the store. violates (2g)

From the amount of attention that people devote to rules like those in (1) and (2), it is easy to get
the impression that they are the only linguistic rules there are. But it is also easy to see that that
can't be so. The reason is that even people who don't follow the rules in (1) and (2) don't produce
rampantly variable, confusing word salad. For instance, even people who invariably produce
sentences like (3) do not produce the likes of (4).

(4) a. Over there is guy the who I went to party the with.
 b. Over there is the who I went to the party with guy.

c. Bill and me the store to went.

The sentences in (3) may be instances of bad grammar in the everyday sense, but they are still
English sentences. By contrast, we don't need to rely on school rules to tell us that the examples
in (4) are not English sentences - even though they contain exactly the same English words as the
sentences in (3).

Since native speakers of English do not produce a variable mishmash of words of the sort in (4),
there must be another type of rules according to which sentences are composed. We can
determine what some of them are by taking a closer look at the sequences in (4). Why exactly is
it that they are word salad? In (4a), the article the is in the wrong order with respect to the nouns
that it belongs with, guy and party. In (4b), the relative clause (who I went to the party with) is in
the wrong order with respect to the noun that it modifies (guy). In (4c), the preposition to is in
the wrong order with respect to its object (the store). In other words, the sentences in (4) do not
follow the rules in (5).

(5) a. Articles precede the nouns that they belong with.

b. Relative clauses follow the noun that they modify.

c. Prepositions precede their objects.

(There's a further rule that's not followed in (4), which you are asked to formulate in the Exercise
1.1.)

Rules like those in (5) have a different intention than those in (2). The rules in (2) are
prescriptive; those in (5) are descriptive. Rules of prescriptive grammar have the same status as
rules of etiquette (like table manners or dress codes) or the laws of society, which divide the
spectrum of possible human behavior into socially acceptable or legal behavior, on the one hand,
and socially unacceptable or illegal behavior, on the other. Rules of prescriptive grammar make
statements about how people ought to use language. In contrast, rules of descriptive grammar
have the status of scientific observations, and they are intended as insightful generalizations
about the way that speakers use language in fact, rather than about they way that they ought to
use it. Descriptive rules are more general and more fundamental than prescriptive rules in the
sense that all sentences of a language are formed in accordance with them, not just a more or less
arbitrary subset of shibboleth sentences. A useful way to think about the descriptive rules of a
language (to which we return in more detail below) is that they produce, or generate, all the
sentences of a language. The prescriptive rules can then be thought of as filtering out some
(relatively minute) portion of the entire output of the descriptive rules as socially unacceptable.

In syntax, as in modern linguistics more generally, we adopt a resolutely descriptive perspective

concerning language. In particular, when linguists say that a sentence is grammatical, we don't
mean that it is correct from a prescriptive point of view, but rather that it conforms to descriptive
rules like those in (5). In order to indicate that a sequence of words or morphemes is
ungrammatical in this descriptive sense, we prefix it with an asterisk. Grammatical sentences are
usually not specially marked, but sometimes we prefix them with 'ok' for clarity. These
conventions are illustrated in (6) and (7).

(6) a. * Over there is guy the who I went to party the with. (= (4a))

b. * Over there is the who I went to the party with guy. (= (4b))

(7) a. ok Over there is the guy who I went to the party with. (= (3a))

b. ok Over there is the guy with whom I went to the party.

Prescriptive grammar is based on the idea that there is a single right way to do things. When
there is more than one way of saying something, prescriptive grammar is generally concerned
with declaring one (and only one) of the variants to be correct. The favored variant is usually
justified as being better (whether more logical, more euphonious, or more desirable on some
other grounds) than the deprecated variant. In the same situation of linguistic variability,
descriptive grammar is content simply to document the variants - without passing judgment on
them.

For instance, consider the variable subject-verb agreement pattern in (8).

(8) a. There 's some boxes left on the porch.

b. There are some boxes left on the porch.

In (8a), the singular verb is (contracted to 's) agrees in number with the preverbal expletive
subject there (in red), whereas in (8b), the plural verb are agrees with the postverbal logical
subject some boxes (in blue). The color of the verb indicates which of the two subjects it agrees
with.

The prescriptive and descriptive rules concerning this pattern are given in (9). The differences
between the two rules are emphasized by underlining.

(9) In a sentence containing both the singular expletive subject there and a plural logical subject ...

a. Prescriptive ... the verb should agree in number with the logical subject.
rule:

b. Descriptive rule: ... the verb can agree in number with either the expletive subject or the logical
 subject.

To take another example, let's consider the prescriptive rule that says, "Don't end a sentence with
a preposition."2 A prescriptivist might argue that keeping the preposition (in italics) together with
its object (in boldface), as in (10a), makes sentences easier to understand than does separating
the two, as in (10b).

(10) a. With which friend did you go to the party?

b. Which friend did you go to the party with?

But by that reasoning, (11a), where the verb and its object are adjacent, ought to be preferable to
(11b), where they are not. In fact, however, (11a) is completely ungrammatical in English.

(11) a. * Adopt which cat did your friend?

b. ok Which cat did your friend adopt?

It is important to understand that there is no conceptual or semantic reason that prepositions can
be separated from their objects in English, but that verbs can't. From a descriptive perspective,
the grammaticality contrast between (10a) and (11a) is simply a matter of fact, irreducible to
more basic considerations (at least given our present state of knowledge). (12) highlights the
difference between the relevant prescriptive and descriptive rule.

(12) When the object of a preposition appears in a position other than its ordinary one (as in a
question), ...

a. Prescriptive ... it should be preceded by the preposition.

rule:

b. Descriptive ... it can either be preceded by the preposition, or it may stand alone, with the
rule: preposition remaining in its ordinary position.

The contrasting attitude of prescriptive and descriptive grammar towards linguistic variation has
a quasi-paradoxical consequence: namely, that prescriptive rules are never descriptive rules. The
reason for this has to do with the way that social systems (not just language) work. If everyone in
a community consistently behaves in a way that is socially acceptable in some respect, then there
is no need for explicit prescriptive rules to ensure the behavior in question. It is only when
behavior that is perceived as socially unacceptable becomes common that prescriptive rules
come to be formulated to keep the unacceptable behavior in check. For example, if every
customer entering a store invariably wears both a shirt and shoes, there is no need for the store
owner to put up a sign that says "No shirt, no shoes, no service." Conversely, it is precisely at
illegal dump sites that we observe "No dumping" signs. In an analogous way, in the domain of
language use, rules of prescriptive grammar are only ever formulated in situations where
linguistic variation is common. But being prescriptive, they cannot treat all of the occurring
variants as equally acceptable - with the result that they can't ever be descriptive.

Rule formation and syntactic structure in language acquisition

As we have just seen, prescriptive and descriptive rules of grammar differ in intention. In addition, they
differ in how they come to be part of a speaker's knowledge. Prescriptive rules are taught at school, and
because they are taught, people tend to be conscious of them, even if they don't actually follow them.
By contrast, we follow the rules of descriptive grammar consistently3 and effortlessly, yet without
learning them at school. In fact, children have essentially mastered these rules on their own by first
grade. Ordinarily, we are completely unconscious of the descriptive rules of language. If we do become
conscious of them, it tends to be in connection with learning a foreign language whose descriptive
grammar differs from that of our native language. In order to emphasize the difference between the
unconscious way that we learn a native language (or several) in early childhood and the conscious way
that we learn a foreign language later on in life, the first process is often called language acquisition
rather than language learning.

As you consider descriptive rules like those in in (5), you might not find it all that surprising that
a child raised in an English-speaking community would acquire, say, the rule that articles
precede nouns. After all, you might say, all the child ever hears are articles and nouns in that
order.4 So why would it ever occur to such a child to put the article and the noun in the other
order? Isn't it just common sense that children learn their native language by imitating older
speakers around them?

Well, yes and no. It is true that children learn some aspects of their native language by imitation
and memorization. Children in English-speaking communities learn English words, children in
Navajo-speaking communities learn Navajo words, children in Swahili-speaking communities
learn Swahili words, and so on. But language acquisition isn't purely a process of memorization.
In fact, given current human life spans, it couldn't possibly be!

A thought experiment

To see this, let's consider a toy version of English that contains three-word sentences consisting
of a noun, a transitive verb, and another noun. The toy version contains sentences like (13) that
are sensible given the real world as well as sentences like (14) that aren't, but that might be
useful in fairy tale or science fiction contexts.

(13) a. Cats detest lemons. (14) a. Lemons detest cats.

("Secret life of citrus fruits")

b. Children eat tomatoes. b. Tomatoes eat children.

 ("Attack of the genetically modified tomatoes")

c. Cheetahs chase gazelles. c. Gazelles chase cheetahs.

("Avenger gazelle")

Again for the sake of argument, let's assume a (small) vocabulary of 1,000 nouns and 100 verbs.
This gives us a list of 1,000 x 100 x 1,000 (= 100 million) three-word sentences of the type in
(13) and (14). Numbers of this magnitude are difficult to put in human perspective, so let's
estimate how long it would take a child to learn all the sentences on the list. Again, for the sake
of argument, let's assume that children can memorize sentences quickly, at a rate of one sentence
a second. The entire list of three-word sentences could then be memorized in 100 million
seconds, which comes to 3.17 years. So far, so good. However, the minute we start adding
complexity to Toy English, the number of sentences and the time it would take to memorize
them quickly mushrooms. For instance, adding only 10 adjectives to the child's vocabulary
would cause the number of five-word sentences of the form in (15) to grow to 10 billion (100
million x 10 x 10).

(15) a. Black cats detest green peas.

b. Happy children eat ripe tomatoes.

c. Hungry cheetahs chase speedy gazelles.

Even at the quick rate of one sentence per second that we're assuming, the list of all such five-
word sentences would take a bit over 317 years to learn. Clearly, this is an absurd consequence.
For instance, how could our memorious child ever come to know, as every English speaker
plainly does, that the sentence in (16) is ungrammatical? If grammatical knowledge were based
purely on rote memorization, the only way to determine this would be to compare (16) to all of
the 10 billion five-word sentences and to find that it matches none of them.

(16) * Cats black detest peas green.

And even after performing the comparison, our fictitious language learner still wouldn't have the
faintest clue as to why (16) is ungrammatical!

In addition to this thought experiment with its comically absurd consequences, there is another
reason to think that language acquisition isn't entirely based on rote memorization - namely, that
children use what they hear of language as raw material to construct linguistic rules. How do we
know this? We know because children sometimes produce rule-based forms that they have never
heard before.

Rule-based word formation

One of the earliest demonstrations that children acquire linguistic rules, rather than simply
imitating the forms of adult language, was the well-known wug experiment (Berko 1958). In it,
the psycholinguist Jean Berko used invented words to examine (among other things) how
children between the ages of 4 and 7 form plurals in English. She showed the children cards with
simple line drawings of objects and animals and elicited plurals from them by reading them
accompanying texts like (17).

(17) This is a wug. Now there is another one. There are two of them. There are two ___.

More than 75% of the children pluralized the invented words cra, lun, tor, and wug in exactly the
same way that adults did in a control group: they added the sound -z to the word (Berko
1958:159-162).5 Since none of the children had encountered the invented words before the
experiment, their response clearly indicates that they had acquired a plural rule and were using it
to produce the novel forms.

Children are also observed to produce novel rule-based forms instead of existing irregular adult
forms (for instance, comed or goed instead of came or went). This process, which is known as
overregularization, is further illustrated in (18) (Marcus et al. 1992:148-149, based on Brown
1973).

(18) a. beated, blowed, catched, cutted, doed, drawed, drived, falled, feeled, growed, holded, maked,
sleeped, standed, sticked, taked, teached, throwed, waked, winned (Adam, between the ages of
2 and 5)

b. drinked, seed, weared (Eve, between the ages of 1 1/2 and 2)

Overregularized forms don't amount to a large fraction of the forms that children produce overall
(less than 5% in the case of past tense forms, according to Marcus et al. 1992:35), but they are
important because they clearly show that even the acquisition of words can't be completely
reduced to rote memorization.

Question formation

In addition to morphological rules (which concern the structure of words), children also acquire
syntactic rules (which concern the structure of sentences). Some of these rules are of particular
interest because they differ from the corresponding adult rules that the children eventually
acquire. At the same time, however, the children's novel rules don't differ from the rules of the
adult grammar in completely arbitrary ways. Rather, the children's rules share certain abstract
properties with the adult rules, even when they differ from them.

To see this, let's consider how young children form yes-no questions. Some 3- to 5-year-olds
form such questions from declarative sentences by copying the auxiliary element to the
beginning of the sentence, as in (19) (Crain and Nakayama 1987:536). (We use the term
'auxiliary element' as a convenient cover term for elements that invert with the subject in (adult)
English questions, like forms of the verb to be or modals like can. See Modals and auxiliary
verbs in English for more details.)

(19) a. The girl is tall. → Is the girl is tall?

b. The red pig can stand on the house. → Can the red pig can stand on the house?

In the course of language acquisition, the questions in (19) are eventually replaced by those in
(20), where we can think of the auxiliary element as having been moved rather than copied.

(20) a. Is the girl ___ tall?

b. Can the red pig ___ stand on the house?

But now notice a striking indeterminacy, first pointed out by Chomsky 1971:26-27. When
children produce questions like those in (20), there is no way of telling whether they are using
the adult rule for question formation in (21a) or the logically possible alternative rule in (21b).

(21) a. Adult question To form a question from a declarative sentence containing an auxiliary
formation rule: element, find the subject of the sentence, and invert the subject and the
auxiliary.

b. Logically possible To form a question from a declarative sentence containing an auxiliary

alternative: element, find the first auxiliary element, and move it to the beginning of
the sentence.

Don't confuse 'subject' with 'simple subject.'

Subjects, in contrast to simple subjects, are possible responses to questions like Who is
tall? and Who can stand on the house? The subjects in (20) are the noun phrases the girl
and the red pig.

If the subject consists of a single word or a clause, then the simple subject is identical to
the subject; otherwise, the simple subject of a sentence is obtained by stripping the
subject of any modifiers (yielding girl and pig as the simple subjects of (20)). The notion
of subject is basic to syntactic theory, but we will have no further use for the notion of
simple subject.

Both rules in (21) give the same result for simple sentences, which are likely to form most of the
data that young children attend to. Both rules also require children to identify auxiliary elements.
However, the adult rule additionally requires children to identify the subject of the sentence by
grouping together sequences of words like the girl or the red pig into a single abstract structural
unit. Because of this grouping requirement, the adult rule is called structure-dependent. By
contrast, the alternative rule in (21b) is not structure-dependent, since it requires the child only to
classify words according to their syntactic category (Is this word an auxiliary element?), but not
to group the words into structural units. The rule in (21b) is simpler in the sense that it relies on
fewer, as well as computationally less complex, cognitive operations, and children might
reasonably be expected to experiment with it in the course of acquiring question formation.
Nevertheless, Chomsky 1971 predicted that children would use only structure-dependent rules in
the course of acquisition.

As we mentioned, both rules give the same result for simple sentences. So how could we
possibly tell which of the two rules a child was actually using? Well, forming yes-no questions is
not restricted to simple sentences. So although we can't tell which rule a child is using in the case
of simple sentences like (19), the rules in (21) give different results for a complex sentence like
(22), which contains a relative clause (who was holding the plate).

(22) The boy who was holding the plate is crying.

In particular, the sentence in (22) contains two auxiliary elements - one for the relative clause
(was), and another one (is) for the entire sentence (the so-called matrix sentence, which contains
the relative clause). A child applying the structure-dependent question formation rule to (22)
would first identify the subject of the matrix sentence (the boy who was holding the plate) and
then invert the entire subject - including the relative clause and the auxiliary contained within it
(was) - with the matrix auxiliary (is). On the other hand, a child applying the structure-
independent rule would identify the first auxiliary (was) and move it to the beginning of the
sentence. As shown in (23), the two rules have very different results,

(23) a. Structure-dependent rule:

[ The boy who was holding the plate ] is → Is [the boy who was holding the plate] ___
crying. crying?

b. Structure-independent rule:

The boy who was holding the plate is crying. → Was the boy who ___ holding the plate is
crying?

Recall that Chomsky predicted that children would not use structure-independent rules, even
though they are simpler than structure-dependent ones. This prediction was tested in an
experiment with 3- to 5-year-old children by Crain and Nakayama 1987. In the experiment, the
experimenter had the children pose yes-no questions to a doll (Jabba the Hut from Star Wars).
For instance, the experimenter would say to each child Ask Jabba if the boy who was holding the
plate is crying. This task elicited various responses. Some children produced the adult question
in (23a), whereas others produced the copy question in (24a) or the restart question in (24b).
(24) a. Is [the boy who was holding the plate] is crying?

b. Is [the boy who was holding the plate], is he crying?

Although neither of the questions in (24) uses the adult rule in (21a), the rules that the children
used to produce them are structure-dependent in the same way that the adult rule is. This is
because children who produced (24a) or (24b) must have identified the subject of the sentence,
just like the children who produced (23a). Out of the 155 questions that the children produced,
none were of the structure-independent type in (23b). Moreover, no child produced the structure-
independent counterpart of (24a), shown in (25), which results from copying (rather than
moving) the first auxiliary element in the sentence.

(25) Was the boy who was holding the plate is crying?

In other words, regardless of whether a child succeeded in producing the adult question in (23a),
every child in the experiment treated the sequence the boy who was holding the plate as a
structural unit, thus confirming Chomsky's prediction.

More evidence for syntactic structure

We have seen that young children are capable of forming and applying both morphological and
syntactic rules. Moreover, as we have seen in connection with question formation, children do
not immediately acquire the rules of the adult grammar. Nevertheless, the syntactic rules that
children are observed to use in the course of acquisition are a subset of the logically possible
rules that they might postulate in principle. In particular, as we have just seen, children's
syntactic rules are structure-dependent. Another way of putting this is that the objects that
syntactic rules operate on (declarative sentences in the case of the question formation rule) are
not simply strings of words, but rather groups of words that belong together, so-called syntactic
constituents.

Intuitions about words belonging together

Evidence for syntactic structure isn't restricted to data from child language acquisition. Further
evidence comes from the intuitions that adults (and even children) have that certain words in a
sentence belong together, whereas others do not. For instance, in a sentence like (26), we have
the strong intuition that the first the belongs with dog, but not with did, even though the is
adjacent to both.

(26) Did the dog chase the cat?

Similarly, the second the in (26) belongs with cat and not with chase. But a word doesn't always
belong with the following word. For instance, in (27), dog belongs with the preceding the, not
with the following the.
(27) Did the dog the children like chase the cat?

Words that belong together can sometimes be replaced by placeholder elements such as
pronouns, as illustrated in (28).

(28) a. Did the dog chase the cat? → Did she chase him?

b. Did the dog the children like chase the cat? → Did the dog they like chase him?

The term 'pronoun' is potentially misleading since it suggests that pronouns substitute for nouns
regardless of syntactic context. In fact, pronouns substitute for noun phrases (as will be
discussed in more detail in Chapter 2). A less confusing term for them would be 'pro-noun
phrase,' but we'll continue to use the traditional term.

It's important to recognize that pronouns don't simply replace strings of words regardless of
context. Just because a string like the dog is a constituent in (28a) doesn't mean that it's always a
constituent. We can see this by replacing the dog by a pronoun in (28b), which leads to the
ungrammatical result in (29).

(29) Did the dog the children like chase the cat? → * Did she the children like chase the cat?

The ungrammaticality in (29) is evidence that the and dog belong together less closely in (28b)
than in (28a). In particular, in (28b), dog combines with the relative clause, and the combines
with the result of this combination, not with dog directly, as it does in (28a).

In some sentences, we have the intuition that words belong together even when they are not
adjacent. For instance, see and who in (30a) belong together in much the same way as see and
Bill do in (30b).

(30) a. Who will they see?

b. They will see Bill.

Finally, we can observe that there are various sorts of ways that words can belong together. For
instance, in a phrase like the big dog, big belongs with dog, and we have the intuition that big
modifies dog. On the other hand, the relation between see and Bill in (30b) isn't one of
modification. Rather, we have the intuition that Bill is a participant in a seeing event.
In the course of this book, we will introduce more precise ways of expressing and representing
intuitions like the ones just discussed. For the moment, what is important is that we have strong
intuitions that words belong together in ways that go beyond adjacency.

Structural ambiguity

A particularly striking piece of evidence for the existence of syntactic structure is the phenomenon of
structural ambiguity. The classified advertisement in (31) is a humorous illustration.

(31) Wanted: Man to take care of cow that does not smoke or drink.

World knowledge tells us that the intent of the advertiser is to hire a clean-living man to take
care of a cow. But because of the way the advertisement is formulated, it also has an
unintentionally comical interpretation - namely, that the advertiser has a cow that does not smoke
or drink and that the advertiser wants a man (possibly a chain-smoking alcoholic) to take care of
this clean-living cow. The intended and unintended interpretations describe sharply different
situations; that is why we say that (31) is ambiguous, rather than that it is vague. Moreover, the
ambiguity of the sentence can't be pinned on a particular word, as is possible in ambiguous
sentences like those in (32).

(32) a. As far as I'm concerned, any gender is a drag. (Patti Smith)

b. Our bikinis are exciting. They are simply the tops.

Sentences like those in (32) are examples of lexical ambiguity; their ambiguity is based on a
lexeme (= vocabulary item) with two distinct meanings. In (31), on the other hand, the words
themselves have the same meanings in each of the two interpretations, and the ambiguity derives
from the possibility of grouping the words in distinct ways. In the intended interpretation, the
relative clause that does not smoke or drink modifies man; in the unintended interpretation, it
modifies cow.

To avoid any confusion, we should emphasize that we are here considering structural ambiguity
from a purely descriptive perspective, focusing on what it tells us about the design features of
human language and disregarding the practical issue of effective communication. As writers of
advertisements ourselves, we would take care not to use (31), but to disambiguate it by means
of an appropriate paraphrase. For the ordinary interpretation of (31), where the relative clause
modifies man, we might move the relative clause next to the intended modifiee, as in (33a). The
comical interpretation of (31), on the other hand, cannot be expressed unambiguously by moving
the relative clause. If it were the desired interpretation, we would have to resort to a more drastic
reformulation, such as (33b).

(33) a. Wanted: Man that does not smoke or drink to take care of cow.
 b. Wanted: Man to take care of nonsmoking, nondrinking cow.

Universal Grammar

Formal universals

The structure-dependent character of syntactic rules is a general property of the human language
faculty (the part of the mind/brain that is devoted to language), often also referred to as
Universal Grammar, especially when considered in abstraction from any particular language.
There are two sources of evidence for this. First, as we have seen, the syntactic rules that
children acquire even when they are not the rules that adults use, are structure-dependent.
Second, even though structure-independent rules are logically possible and computationally
tractable, no known human language actually has rules that disregard syntactic structure as a
matter of course. For instance, no known human language has either of the computationally very
simple question formation rules in (34).

(34) a. To form a question, switch the order of The girl is tall. → Girl the is tall?
the first and second words in the
corresponding declarative sentence.

The blond girl is tall. → Blond the girl is tall?

b. To form a question, reverse the order of The girl is tall. → Tall is girl the?
the words in the corresponding
declarative sentence.

The blond girl is tall. → Tall is girl blond the?

The structure-dependent character of syntactic rules (often referred to more briefly as structure
dependence) is what is known as a formal universal of human language - a property common
to all human languages that is independent of the meanings of words. Formal universals are
distinguished from substantive universals, which concern the substance, or meaning, of
linguistic elements. An example of a substantive universal is the fact that all languages have
indexical elements such as I, here, and now. These words have the special property that their
meanings are predictable in the sense that they denote the speaker, the speaker's location, and the
time of speaking, but that what exactly they refer to depends on the identity of the speaker.

Recursion

Another formal universal is the property of recursion. A simple illustration of this property is
the fact that it is possible for one sentence to contain another. For instance, the simple sentence
in (35a) forms part of the complex sentence in (35b), and the resulting sentence can form part of
a still more complex sentence. Recursive embedding is illustrated in (35) up to a level of five
embeddings.

(35) a. She won.

b. The Times reported that

[she won].

c. John told me that

[the Times reported that
[she won]].

d. I remember distinctly that

[John told me that
[the Times reported that
[she won]]].

e. They don't believe that

[I remember distinctly that
[John told me that
[the Times reported that
[she won]]]].

f. I suspect that
[they don't believe that
[I remember distinctly that
[John told me that
[the Times reported that
[she won]]]]].

Parameters

Formal universals like structure dependence and recursion are of particular interest to linguistics in the
Chomskyan tradition. This is not to deny, however, that individual languages differ from one another,
and not just in the sense that their vocabularies differ. In other words, Universal Grammar is not
completely fixed, but allows some variation. The ways in which grammars can differ are called
parameters.

One simple parameter concerns the order of verbs and their objects. In principle, two orders are
possible: verb-object (VO) or object-verb (OV), and different human languages use either one or
the other. As illustrated in (36) and (37), English and French are languages of the VO type,
whereas Hindi, Japanese, and Korean are languages of the OV type.
(36) a. English Peter read the book.

b. French Pierre lisait le livre.

Pierre was.reading the book
'Pierre was reading the book.'
(37) a. Hindi Peter-ne kitaab parh-ii.

b. Japanese Peter-ga hon-o yon-da.

c. Korean Peter-ka chayk-ul ilk-ess-ta.

Peter book read
'Peter read the book.'

Another parameter of Universal Grammar concerns the possibility, mentioned earlier in

connection with prescriptive rules, of separating a preposition from its object, or preposition
stranding. (The idea behind the metaphor is that the movement of the object of the preposition
away from its ordinary position leaves the preposition stranded high and dry.) The parametric
alternative to preposition stranding goes by the name of pied piping,6 by analogy to the Pied
Piper of Hameln, who took revenge on the citizens of Hameln for mistreating him by luring the
town's children away with him. In pied piping of the syntactic sort, the object of the preposition
moves away from its usual position, just as in preposition stranding, but it takes the preposition
along with it. The two parametric options are illustrated in (38). (The alert reader will note that
one language can exhibit two parameter settings. We return to this issue later on in the chapter,
in the section on Grammar versus language.)

(38) a. Preposition stranding: ok Which house does your friend live in?

b. Pied piping: ok In which house does your friend live?

Just as in English, preposition stranding and pied piping are both grammatical in Swedish. (In
Swedish, it is preposition stranding that counts as prescriptively correct! Pied piping is frowned
upon, on the grounds that it sounds stiff and artificial.)

(39) a. Swedish ok Vilket hus bor din kompis i?

which house lives your friend in
'Which house does your friend live
in?'
b. ok I vilket hus bor din kompis?

In other languages, such as French and Italian, preposition stranding is ungrammatical. Speakers
of these languages reject examples like (40) as word salad, and accept only the corresponding
pied-piping examples in (41).
(40) a. French * Quelle maison est-ce que ton ami habite dans?
which house is it that your friend lives in
Intended meaning: 'Which house does your friend live in?'
b. Italian * Quale casa abita il tuo amico in?
which house lives the your friend in
Intended meaning: 'Which house does your friend live in?'
(41) a. French ok Dans quelle maison est-ce que ton ami habite?

b. Italian ok In quale casa abita il tuo amico?

Generative grammar

At the beginning of this chapter, we said that this book was an introduction to generative grammar from
a Chomskyan perspective. Until now, we have clarified our use of the term 'grammar,' and we have
indicated that a Chomskyan perspective on grammar is concerned with the formal principles that all
languages share as well as with the parameters that distinguish them. Let's now turn to the notion of a
generative grammar.

(42) A generative grammar is an algorithm for specifying, or generating, all and only the grammatical
sentences in a language.

What's an algorithm? It's simply any finite, explicit procedure for accomplishing some task,
beginning in some initial state and terminating in a defined end state. Computer programs are the
algorithms par excellence. More ordinary examples of algorithms include recipes, knitting
patterns, the instructions for assembling an Ikea bookcase, or the steps on the back of a bank
statement for balancing your checkbook.

An important point to keep in mind is that it is often difficult to construct an algorithm for even
trivial tasks. A quick way to gain an appreciation for this is to describe how to tie a bow. Like
language, tying a bow is a skill that most of us master around school age and that we perform
more or less unconsciously thereafter. But describing (not demonstrating) how to do it is not that
easy, especially if we're not familiar with the technical terminology of knot-tying. In an
analogous way, constructing a generative grammar of English is a completely different task than
speaking the language, and much more difficult (or at least difficult in a different way)!

Just like a cooking recipe, a generative grammar needs to specify the ingredients and procedures
that are necessary for generating grammatical sentences. We won't introduce all of these in this
first chapter, but in the remainder of the section, we'll introduce enough ingredients and
procedures to give a flavor of what's to come.

Elementary trees and substitution

The raw ingredients that sentences consist of are vocabulary items. These belong to various
syntactic categories, like noun, adjective, transitive verb, preposition, and so forth. Depending
on their syntactic category, vocabulary items combine with one another to form constituents,
which in turn belong to syntactic categories of their own. For instance, determiners (a category
that includes the articles a and the and the demonstratives this, that, these and those) can
combine with nouns to form noun phrases, but they can't combine with other syntactic categories
like adverbs, verbs, or prepositions.

(43) a. ok a house (44) a. * a slowly

b. ok the cats b. * the went

c. ok those books c. * those of

It's possible to represent the information contained in a constituent by using labeled bracketing.
Each vocabulary item is enclosed in brackets that are labeled with the appropriate syntactic
category. The constituent that results from combining vocabulary items is in turn enclosed in
brackets that are labeled with the constituent's syntactic category. The labeled bracketings for the
constituents in (43) are given in (45).

(45) a. [NounPhr [Deta ] [Noun house ] ]

b. [NounPhr [Detthe ] [Noun cats ] ]

c. [NounPhr [Detthose ] [Noun books ] ]

Noun phrases can combine with other syntactic categories, such as prepositions or transitive
verbs. Prepositions combine with a noun phrase to form prepositional phrases. A transitive verb
combines with one noun phrase to form a verb phrase, which in turn combines with a second
noun phrase to form a complete sentence.

(46) a. [PrepPhr [Prep on ] [NounPhr [Detthe ] [Noun table ] ] ]

b. [VerbPhr [TrVerb drafted ] [NounPhr [Deta ] [Noun letter ] ] ]

c. [Sentence [NounPhr [Detthe ] [Noun secretary ] ] [VerbPhr [TrVerb drafted ] [NounPhr [Deta ] [Noun letter ] ] ] ]

Again, however, noun phrases don't combine with any and all syntactic categories. For instance,
noun phrases can't combine with determiners (at least not in English).

(47) * the this letter

As constituent structure grows more complex, labeled bracketings very quickly grow difficult for
humans to process, and it's often more convenient to represent constituent structure with tree
diagrams. Tree diagrams, or trees for short, convey exactly the same information as labeled
bracketings, but the information is presented differently. Instead of enclosing an element in
brackets that are labeled with a syntactic category, the category is placed immediately above the
element and connected to it with a line or branch. The labeled bracketings that we have seen so
far translate into the trees in (48) and (49).7

(48) a. b. c.

(49) a. b. c.

Trees like those in (48) and (49) resemble dishes that are ready to serve; they don't provide a
record of how they were brought into being. We can provide such a record by representing
vocabulary items themselves in the form of trees that include combinatorial information. For
example, prepositions and transitive verbs can be represented as trees with empty slots for noun
phrases to fit into, as shown in (50).

(50) a. b.

We'll refer to trees for vocabulary items like those in (50) as elementary trees. The purpose of
elementary trees is to represent a vocabulary item's combinatorial possibilities, and so they
ordinarily contain unfilled nodes. Such nodes are called substitution nodes, and they are filled
by a substitution operation, as shown in (51).

(51) a. b. c.

Tree (a) has a The root (= topmost) node in Substitution occurs when the
substitution node of Tree (b) has the same root node of Tree (b) is
some syntactic syntactic category as the identified with the
 category. substitution node in Tree (a). substitution node in Tree (a).

Elementary trees don't necessarily contain substitution nodes, though; ones that invariably play
the role of Tree No. 2 in the substitution operation don't. The elementary tree for the noun in
(52b) is an example.

Notice, by the way, that there are two conceivable ways to arrive at trees for noun phrases like
those cats, depending on whether it is the noun that is taken as the substitution node, as in (52),
or the determiner, as in (53). At this point, there is no reason to prefer one way over the other,
but in Chapter 5, we will adopt a variant of (52).

(52) a. b.

(53) a. b.

In summary, a generative grammar as we've constructed it so far consists of a set of elementary

trees, which represent the vocabulary items in a language and the range of their combinatorial
possibilities, and a substitution operation, by means of which the elementary trees combine into
larger constituents and ultimately into grammatical sentences. In Chapter 4, we will introduce
two further formal operations. The first, adjunction, will enable the grammar to generate
sentences containing modifiers, such as adjectives or relative clauses modifying nouns (the big
dog, the dog that the children like). The second, movement, will enable the grammar to
represent, among other things, the similarities and the differences between declarative sentences
(They will see Bill) and questions corresponding to them (Will they see Bill?, Who(m) will they
see?).

Grammaticality

The aim of a generative grammar is to generate all and only the grammatical sentences of a language.
Since the notion of grammaticality is basic to syntactic theory, it is important to distinguish it from
notions with which it is easily confused.

First and foremost, 'is grammatical' is not the same thing as 'makes sense.' The sentences in (54)
all 'make sense' in the sense that it is easy to interpret them. Nevertheless, as indicated by the
asterisks, they are not grammatical.8

(54) a. * Is our children learning?

b. * Me wants fabric.
 c. * To where are we be taking thou, sir?

d. * The introduction explained that "the Genoese people, besides of hard worker, are good eater
too, and even 'gourmand,' of that honest gourmandise which will not drive a man to hell but
which is, after all, one of the few pleasures that mankind can enjoy in this often sorrowful
world."

Conversely, sentences can be grammatical, but not 'make sense.' The 'fairy tale' or 'science
fiction' sentences in (14) are of this type. Two further examples are given in (55). Since the
sentences are grammatical, they aren't preceded by an asterisk. However, a prefixed pound sign
can be used to indicate their semantic anomaly.

(55) a. # Colorless green ideas sleep furiously. (Chomsky cf. Revolutionary new ideas appear
1965:149) infrequently.

b. # I plan to travel there last year. cf. I plan to travel there next year.

Second, 'grammatical' must be distinguished from 'acceptable' or 'easily processable by human

beings.' This is because it turns out that certain well-motivated simple grammatical operations
can be applied in ways that result in sentences that are virtually impossible for human beings to
process. For instance, it is possible in English to modify a noun with a relative clause, and
sentences containing nouns that are modified in this way, like those in (56), are ordinarily
perfectly acceptable and easily understood. (Here and in the following examples, the relative
clauses are bracketed and the modified noun is underlined.)

(56) a. The mouse [that the cat chased] escaped.

b. The cat [that the dog scared] jumped out the window.

But now notice what happens when we modify the noun within the relative clause in (56a) with a
relative clause of its own.

(57) The mouse [that the cat [that the dog scared] chased] escaped.

Even though (57) differs from (56a) by only four additional words and a single additional level
of embedding, the result is virtually uninterpretable without pencil and paper. The reason is not
that relative clause modification can't apply more than once, since the variant of (56a) in (58),
which contains exactly the same words and is exactly as long, is perfectly fine (or at any rate
much more acceptable than (57)).
(58) The mouse escaped [that the cat chased ] [that the dog scared].

The reason that (57) is virtually uninterpretable is also not that it contains recursive structure (the
relative clause that modifies mouse contains the relative clause that modifies cat). After all, the
structures in (35) are recursive, yet they don't throw us for a loop the way that (57) does.

(57) is unacceptable not because it is ungrammatical, but because of certain limitations on human
short-term memory (Chomsky and Miller 1963:286, Miller and Chomsky 1963:471).
Specifically, notice that in the (relatively) acceptable (58), the subject of the main clause the
mouse doesn't have to "wait" (that is, be kept active in short-term memory) for its verb escaped
since the verb is immediately adjacent to the subject. The same is true for the subjects and verbs
of each of the relative clauses (the cat and chased, and the dog and scared). In (57), on the other
hand, the mouse must be kept active in memory, waiting for its verb escaped, for the length of
the entire sentence. What is even worse, however, is that the period during which the mouse is
waiting for its verb escaped overlaps the period during which the cat must be kept active,
waiting for its verb chased. What makes (57) so difficult, then, is not the mere fact of recursion,
but that two relations of exactly the same sort (the subject-verb relation) must be kept active in
memory at the same time. In none of the other relative clause sentences is such double activation
necessary. For instance, in (56a), the mouse must be kept active for the length of the relative
clause, but the subject of the relative clause (the cat) needn't be kept active since it immediately
precedes its verb chased.

Sentences like (56) and (57) are often referred to as center-embedding structures, and they
are said to contain nested dependencies.

The mouse that the cat chased escaped.

| |______| |
|____________________________|

The mouse that the cat that the dog scared chased escaped.
| | |______| | |
| |__________________________| |
|_________________________________________________|

A final important point to bear in mind is that any sentence is an expression that is paired with a
particular interpretation. Grammaticality is always determined with respect to a pairing of form
and meaning. This means that a particular string can be grammatical under one interpretation, but
not under another. For instance, (59) is ungrammatical under an subject-object-verb (SOV)
interpretation (that is, when the sentence is interpreted as Sue hired Tom).

(59) Sue Tom hired.

(59) is grammatical, however, under an object-subject-verb (OSV) interpretation (that is, when it
is interpreted as Tom hired Sue). On this interpretation, Sue receives a special intonation marking
contrast, which would ordinarily be indicated in writing by setting off Sue from the rest of the
sentence by a comma. In other words, the grammaticality of (59) depends on whether its
interpretation is analogous to (60a) or (60b).

(60) a. ok Her, he hired. (The other job candidates, he didn't even call back.)

b. * She him hired.

Grammar versus language

We conclude this chapter by considering the relationship between the concepts of grammar and
language. The notion of language seems straightforward because we are used to thinking and speaking
of "the English language," "the French language," "the Swahili language," and so forth. But these terms
are actually much vaguer than they seem at first glance because they cover a plethora of varieties,
including ones that differ enough to be mutually unintelligible. For instance, Ethnologue distinguishes 32
dialects of English in the United Kingdom alone. In addition, distinct dialects of English are spoken in
former British colonies, including Canada, the United States, Australia, New Zealand, India, and many
African, Asian, and Caribbean nations, and many of these dialects have subdialects of their own.
Similarly, Ethnologue distinguishes 11 dialects of French in France and 10 dialects of Swahili in Kenya,
and there are further dialects in other countries in which these languages are spoken. Moreover, we use
terms like "the English language" to refer to historical varieties that differ as profoundly as present-day
English does from Old English, which is about as intelligible to a speaker of modern English as German
(in other words, not very).

Although the most salient differences between dialects are often phonological (that is, speakers
of different dialects often have different accents), dialects of a so-called single language can
differ syntactically as well. For instance, in standard French, as in the Romance languages more
generally, adjectives ordinarily follow the noun that they modify. But that order is reversed in
Walloon, a variety of French spoken in Belgium. The two parametric options are illustrated in
(61) (Bernstein 1993:25-26).

(61) a. Standard French un chapeau noir

a hat black
b. Walloon on neûr tchapê
a black hat
'a black hat'

Another example of the same sort, though considerably more cathected for speakers of English,
concerns multiple negation in sentences like (62a).

(62) a. The kids didn't eat nothing.

 b. The kids didn't eat anything.

In present-day standard English, didn't and nothing each contribute their negative force to the
sentence, and the overall force of (62a) isn't negative; rather, the sentence means that the kids ate
something. In many nonstandard varieties of English, however, (62a) conveys exactly the same
meaning as standard English (62b); that is, the sentence as a whole has negative force. In these
dialects, the negation in nothing can be thought of as agreeing with (and reinforcing) the
negation in didn't rather than cancelling it; hence the term negative concord for this
phenomenon ('concord' is a variant term for 'agreement'). Negative concord is routinely
characterized as "illogical" by prescriptivists,9 and it is one of the most heavily stigmatized
features in present-day English.10 However, it was productive in earlier forms of English, and it
is attested in renowned masters of the language such as Chaucer and Shakespeare. Moreover,
negative concord is part of the standard forms of languages like French, Italian, Spanish, and
modern Greek. From a descriptive and generative point of view, negative concord is simply a
parametric option of Universal Grammar just like any other, and negative concord is no more
illogical than the noun-adjective order in (61a) or preposition stranding.

In both of the examples just discussed, we have dialects of "the same language" (English and
French, respectively) differing with respect to a parameter. The converse is also possible: two
"different languages" that are parametrically (all but) indistinguishable. For example, the same
linguistic variety spoken on the Dutch-German border may count as a dialect of Dutch or
German depending on which side of the political border it is spoken, and the same is true of
many other border dialects as well. According to Max Weinreich, "a language is a dialect with an
army and a navy." A striking (and sad) confirmation of this aphorism concerns the recent
terminological history of Serbo-Croatian. As long as Yugoslavia was a federal state, Serbo-
Croatian was considered a single language with a number of regional dialects. The 14th edition
of Ethnologue, published in 2000, still has a single entry for Serbo-Croatian. In the 15th edition,
published in 2005, the single entry is replaced by three new entries for Bosnian, Croatian, and
Serbian.

As the previous discussion has shown, the notion of language is based more on sociopolitical
considerations than on strictly linguistic ones. By contrast, the term 'grammar' refers to a
particular set of parametric options that a speaker acquires. For this reason, the distinction
between language and grammar that we have been drawing is also referred to as the distinction
between E-language and I-language (mnemomic for 'external' and 'internal' language)
(Chomsky 1986).

As we have seen, the same language label can be associated with more than one grammar (the
label "English" is associated with grammars both with and without negative concord), and a
single grammar can be associated with more than one language label (as in the case of border
dialects). It is important to distinguish the concept of shared grammar from mutual intelligibility.
To a large extent, standard English and many of its nonstandard varieties are mutually
intelligible even where their grammars differ with respect to one parameter or another. On the
other hand, it is perfectly possible for two or more varieties that are mutually unintelligible to
share a single grammar. For instance, in the Indian village of Kupwar (Gumperz and Wilson
1971), the three languages Marathi, Urdu, and Kannada, each spoken by a different ethnic group,
have been in contact for about 400 years, and most of the men in the village are bi- or trilingual.
Like the standard varieties of these languages, their Kupwar varieties have distinct vocabularies,
thus rendering them mutually unintelligible to monolingual speakers, but in Kupwar, the
considerable grammatical differences that exists among the languages as spoken in other parts of
India have been virtually eliminated. The difference between standard French and Walloon with
respect to prenominal adjectives is another instance of this same convergence phenomenon.
Here, too, the adjective-noun order in Walloon is due to language contact and bilingualism, in
this case between French and Flemish, the other language spoken in Belgium; in Flemish, as in
the Germanic languages more generally, adjectives ordinarily precede the nouns that they
modify.

It is worth noting that it is perfectly possible for a single speaker to acquire more than one
grammar. This is most strikingly evident in balanced bilinguals. Speakers can also acquire more
than one grammar in situations of syntactic change. For instance, in the course of its history,
English changed from an OV to a VO language, and individual speakers during the transition
period (which began in late Old English and continued into Middle English) acquired and used
both parametric options. Finally, speakers can acquire more than one grammar in situations of
diglossia or stable syntactic variation. For instance, English speakers whose vernacular grammar
has negative concord or (as for most of us) preposition stranding might acquire the parametric
variants without negative concord or with pied piping in the course of formal education.

Notes

1. It is also possible to overzealously apply rules like those in (2), even in cases where they
shouldn't be applied. This phenomenon is known as hypercorrection. Two common instances are
illustrated in (i).

Hypercorrect example Explanation

(i) a. Over there is the guy whom I Should be: the guy who I think took her to the party
think took her to the party. (the relative pronoun who is the subject of the relative clause,
not the object; cf. the guy { who, *whom } took her to the party)

b. This is strictly between you and I. Should be: between you and me
(the second pronoun is part of the conjoined object of the
preposition between, not part of a subject)

2. The prescriptive rule is actually better stated as "Don't separate a preposition from its object,"
since the traditional formulation invites exchanges like (i).
(i) A: Who are you going to the party with?

B: Didn't they teach you never to end a sentence with a preposition?

A: Sorry, let me rephrase that. Who are you going to the party with, Mr. Know-it-all?

3. As William Labov has often pointed out, everyday speech (apart from false starts and other
self-editing phenomena) hardly ever violates the rules of descriptive grammar.

4. Actually, that's an oversimplification. Not all the articles and nouns an English-speaking child
hears appear in the article-noun order. To see why, carefully consider the underlined sentence in
this footnote.

5. When children didn't respond this way, they either repeated the original invented word, or they
didn't respond at all. It's not clear what to make of these responses. Either response might
indicate that the children were stumped by the experimental task. Alternatively, repetition might
have been intended as an irregular plural (cf. deer and sheep), and silence might indicate that
some of the invented words (for instance, cra) struck the children as phonologically strange.

6. The term 'pied piping' was invented in the 1960's by John Robert Ross, a syntactician with a
penchant for metaphorical terminology.

7. Online corpora that are annotated with syntactic structure, such as the Penn Treebank, the
Penn-Helsinki Parsed Corpora of Middle English and of Early Modern English, and others like
them, tend to use labeled bracketing because the resulting files are computationally extremely
tractable. The readability of such corpora for humans can be improved by suitable formatting of
the labeled bracketing or by providing an interface that translates the bracketed structures into
tree diagrams.

8. (54a) is from a speech by George W. Bush

(http://politicalhumor.about.com/library/blbushisms2000.htm). (54b) was the subject line of an
email message in response to an offer of free fabric; the author is humorously attempting to
imitate the language of a child greedy for goodies. (54c) is from "Pardon my French" (Calvin
Trillin. 1990. Enough's enough (and other rules of life). 169). (54d) is from "Connoisseurs and
patriots" (Joseph Wechsberg. 1948. Blue trout and black truffles: The peregrinations of an
epicure. 127).

9. Two important references concerning the supposed illogicality of negative concord (and of
nonstandard English more generally) are Labov 1972a, 1972b.

Those who argue that negative concord is illogical often liken the rules of grammar to those of
formal logic or arithmetic, where one negation operator or subtraction operation cancels out
another; that is, (NOT (NOT A)) is identical to A, and (-(-5)) = +5. Such prescriptivists never
distinguish between sentences containing even and odd numbers of negative expressions. By
their own reasoning, (i.a) should have a completely different status than (i.b) - not illogical, but
at worst redundant.

(i) a. They never told nobody nothing.

b. They never told nobody.

10. Because of the social stigma associated with it, it is essentially impossible to study negative
concord in present-day English. This is because even for those speakers of negative concord
varieties who don't productively control standard English as a second dialect, the influence of
prescriptive grammar is so pervasive that if such speakers reject negative concord sentences as
unacceptable, we don't know whether they are rejecting them for grammatical or for social
reasons.

Exercises and problems

Exercise 1.1

The sentences in (4) violate several descriptive rules of English, three of which were given in (5).
As mentioned in the text, there is a fourth descriptive rule that is violated in (4). Formulate the
rule (you shouldn't need more than a sentence).

Exercise 1.2

(1)-(4) illustrate the facts of subject-verb agreement in the nonstandard variety of English spoken
in Belfast, Ireland (data from Henry 1995, chapter 2). Describe the data as clearly and briefly as
you can.

In order to avoid conflating morphological form with semantic content, you can refer to "is"
and "are" as "the i- form" and "the a- form", rather than as "singular" and "plural".

(1) a. ok The girl is late. (2) a. * The girl are late.

b. ok She is late. b. * She are late.

c. ok Is { the girl, she } late? c. * Are { the girl, she } late?

(3) a. ok The girls are late. (4) a. ok The girls is late.

b. ok They are late. b. * They is late.

 c. ok Are { the girls, they } late? c. * Is { the girls, they } late?

Exercise 1.3

Which of the newspaper headlines in (1) are lexically ambiguous, which are structurally
ambiguous, and which are a mixture of both types of ambiguity? Explain.

(1) a. Beating witness provides names

b. Child teaching expert to speak

c. Drunk gets nine months in violin case

d. Enraged cow injures farmer with ax

e. Prostitutes appeal to pope

f. Teacher strikes idle kids

g. Teller stuns man with stolen check

Exercise 1.4

In the text, we showed that sentences are recursive categories. In other words, one instance of the
syntactic category 'sentence' can contain another instance of the same category. Provide evidence
that noun phrases and prepositional phrases are recursive categories, too.

Be careful to give examples that are recursive, and not just ones in which the syntactic
category in question occurs more than once. For instance, (1) does not provide the evidence
required in this exercise, because the second prepositional phrase is not contained in the first.
This is clearly shown by the fact that the order of the prepositional phrases can be switched.

(1) The cat jumped [PP onto the table ] [PP without the slightest hesitation ].

Exercise 1.5

Which, if any, of the sentences in (1)-(5) are ungrammatical? Which, if any, are semantically or
otherwise anomalous? Briefly explain.

(1) a. They decided to go tomorrow yesterday.

 b. They decided to go yesterday tomorrow.

(2) a. They decided yesterday to go tomorrow.

b. They decided tomorrow to go yesterday.

(3) a. Yesterday, they decided to go tomorrow.

b. Tomorrow, they decided to go yesterday.

(4) They decided to go yesterday yesterday.

(5) How long didn't Tom wait?

Exercise 1.6

A. The following expressions are structurally ambiguous. For each reading (= interpretation),
provide a paraphrase that is itself unambiguous.

(1) a. chocolate cake icing

b. clever boys and girls

c. John will answer the question precisely at noon.

d. Watch the man from across the street.

e. They should decide if they will come tomorrow.

B. Provide a tree diagram for each reading. To do so, download the Trees program as well as the
tree-drawing grammar tool. In the Trees program, open the grammar tool with the file menu item
"Choose Grammar." Then select the file menu item "New." This will call up an empty workspace
on the right and a window containing syntactic categories on the upper left. Click on a syntactic
category, and a copy will appear in the window on the lower left. Click on this copy and drag it
into the workspace. You can build trees using any of the premade structures in the grammar tool,
adding or deleting nodes as needed. For the purposes of this exercise, all that is relevant is the
structure of the trees that you build (that is, the way the nodes are grouped, not the way they are
labeled). Therefore, you can simply label all nonterminal nodes (= nodes other than words) with
a dummy symbol like 'X'.

Exercise 1.7

A. How many elements does an expression need to contain to be three-ways ambiguous?

B. If an expression contains four elements, how many ways ambiguous can it be in principle?

Problem 1.1

Are syntactic structure and recursion equally basic properties of human language? Explain in a
brief paragraph.

Problem 1.2

Can you come up with a sentence (or other expression) that is structurally ambiguous more than two
ways? Paraphrase the distinct readings, and draw a tree for each reading. Feel free to use the tree-
drawing grammar tool (see Exercise 1.6 for instructions).

Problem 1.3

Download the grammar tool in which grammar. In the Trees program, open it with the file menu item
"Choose Grammar." Then select the file menu item "New." This will call up an empty workspace on the
right and a window containing a lexicon of one-letter expressions on the upper left. Click on one of the
expressions. A copy of the expression will appear in the lexical items window on the lower left. Click on
this copy and drag it into the workspace. You can build complex expressions out of simpler ones by
dragging them on top of each other or onto other nodes that appear in the course of a derivation.
Before beginning a derivation, you must select a grammar (G1 or G2) in the "choose-grammar" menu
above the workspace. The grammar tool requires you to produce the first combination by dragging one
Roman letter onto the Greek phi. Play with the tool to see what happens next. Once you are able to
construct complex expressions, briefly answer the following questions. There is no need to submit the
trees you construct.

What is the difference between G1 and G2?

If presented with substrings generated by G1 and G2 containing only Roman letters (i.e., if the
phi were somehow invisible), is it possible to tell which grammar has generated the string?

Problem 1.4

The grammars of Early Modern English (1500-1710) and present-day English differ enough for
certain Early Modern English sentences to be ungrammatical today. Find several such sentences,
and briefly describe the source of the ungrammaticality as best as you can. Early Modern English
texts that are easily accessible on the Web include Shakespeare's plays and the Authorized
Version of the Bible (also known as the King James Bible).

2 Syntactic constituenthood

Tests for determining syntactic constituenthood

o Substitution
o Movement
o Questions and short answers
 o It clefts
Some complications
o Mismatches between syntactic structure and other structure
o False negative results
 Phrasal versus lexical constituents
 Finiteness
 Islands
Representing syntactic constituenthood
Notes
Exercises and problems
Supplementary material
o Finiteness in English
o Node relations

At first glance, a sentence consists of a string of words arranged in a single dimension - that of
linear order. However, we presented evidence in Chapter 1 for a second dimension that is less
obvious (though no less real!) than linear order - the dimension of syntactic structure. Whether
a particular string of words is a syntactic constituent isn't always self-evident, and so several
diagnostic tests have been developed for constituenthood. In this chapter, we review these tests,
along with some of the complications that arise in applying them. We also discuss in more detail
how syntactic structure is represented in tree diagrams of the sort introduced in Chapter 1.

Tests for determining syntactic constituenthood

Substitution

The most basic test for syntactic constituenthood is the substitution test. The reasoning behind the test
is simple. A constituent is any syntactic unit, regardless of length or syntactic category. A single word is
the smallest possible constituent belonging to a particular syntactic category. So if a single word can
substitute for a string of several words, that's evidence that the word and the string are constituents of
the same category.

We mentioned in Chapter 1 that pronouns can substitute for noun phrases. Some examples are
given in (1).

(1) a. The little boy fed the cat. → He fed her.

b. Black cats detest green peas. → They detest them.

As we already said in Chapter 1, it's important to understand that a particular string of words can
be a noun phrase in one syntactic context, but not in another. For instance, the substitution test
tells us that the underlined strings are noun phrases in (1), but not in (2).

(2) a. The little boy from next door fed the cat without a → * He from next door fed her without a
tail. tail.

b. These black cats detest those green peas. → * These they detest those them.

Rather, in these sentences, the noun phrases are the longer strings in (3) that are underlined.

(3) a. The little boy from next door fed the cat without a tail. → He fed her.

b. These black cats detest those green peas. → They detest them.

Pronouns are not the only placeholder elements, or pro-forms. For instance, adverbs such as
here or there can substitute for constituents that refer to locations or directions. As in the case of
noun phrases, whether a particular string is a constituent depends on its syntactic context.

(4) a. Put it on the table. → Put it there.

b. Put it over on the table. → Put it over there.

c. Put it over on the table. → Put it there.

(5) a. Put it on the table that's by the door. → * Put it there that's by the door.

b. Put it over on the table that's by the door. → * Put it over there that's by the door.

c. Put it over on the table that's by the door. → * Put it there that's by the door.

The word so can substitute for adjective phrases (here, the most natural-sounding results are
obtained in contexts of comparison). As usual, the same string sometimes is a constituent and
sometimes isn't.

(6) a. I am very happy, and Linda is so, too.

b. I am very fond of Lukas, and Linda is so, too.

c. I am very fond of my nephew, * and Linda is so of her niece.

Finally, pronouns and sometimes the word so can substitute for subordinate clauses introduced
by that, as in (7).

(7) a. I { know, suspect } that they're invited. → I { know, suspect } it.

b. I { imagine, think } that they're invited. → I { imagine, think } so.

We conclude our discussion of substitution by noting a complicating factor. Nowhere is it

written that every constituent has a corresponding pro-form. For instance, although some
prepositional phrases can be replaced by the pro-forms here or there, other types of prepositional
phrases - for instance, ones referring to purposes or reasons - can't. As a result, it is perfectly
possible for the substitution test to give false negative results, and for the results of the
substitution to disagree with the results of the other tests presented in what follows. We return to
the issue of false negative results later on in the chapter.

Movement

Substitution by pro-forms is not the only diagnostic for whether a string is a constituent. If it is possible
to move a particular string from its ordinary position to another position - typically, the beginning of the
sentence - that, too, is evidence that the string is a constituent. In order to make the result of movement
completely acceptable, it's sometimes necessary to use a special intonation or to invoke a special
discourse context, especially in the case of noun phrases. In the examples that follow, "___" indicates
the ordinary position that a constituent has moved from, and appropriate discourse material (enclosed
in parentheses) may be added to make the examples more felicitous.

(8) a. I fed the cats. → The cats, I fed ___. (The dogs, I didn't.)

b. I fed the cats with long, fluffy → The cats with long, fluffy tails, I fed ___. (The other cats, I
tails. didn't.)

Movement of constituents other than noun phrases is illustrated in (9).

(9) a. Prepositional The cat strolled across the porch → With a confident air, the cat strolled
phrase: with a confident air. across the porch ___.

b. Adjective Ali Baba returned from his travels → Wiser than before, Ali Baba returned
phrase: wiser than before. from his travels ___.

c. Adverb phrase: They arrived at the concert hall more → More quickly than they had expected,
quickly than they had expected. they arrived at the concert hall ___.

As shown in (10), moving strings that aren't constituents yields ungrammatical results.
(10) a. I fed the cats with long, fluffy tails. → * The cats, I fed ___ with long, fluffy tails.1

b. The cat strolled across the porch with a → * With a, the cat strolled across the porch ___
confident air. confident air.

c. Ali Baba returned from his travels wiser → * Wiser than, Ali Baba returned from his travels
than before. ___ before.

d. They arrived at the concert hall more → * More quickly than they, they arrived at the
quickly than they had expected. concert hall ___ had expected.

Questions and short answers

Another diagnostic for whether a string is a constituent is whether the string can function as a
short answer to a question. The question itself also functions as a diagnostic test, since we can
think of it as being derived by substitution of a question word for a string, with subsequent
movement of the question word.2 (11) illustrates this pair of tests for a variety of constituent
types.

(11) a. Noun phrase: What do you see? The cats.

Cats with long, fluffy tails.
The cats with long, fluffy tails.

b. Prepositional How did the cat stroll across the With a confident air.
phrase: porch?

c. " Where did Ali Baba go? On a long journey.

To New York.

d. Adjective phrase: How did Ali Baba return? Wiser than before.
Fairly jet-lagged.

e. Adverb phrase: How did they do? Not badly.

Surprisingly well.
Much better than they had
expected.

Once again, attempting to question nonconstituents is ungrammatical.

(12) a. * What did you feed ___ long, fluffy tails? → * The cats with.
 b. * How did the cat stroll across the porch ___ confident air? → * With a.

c. * How did Ali Baba return from his travels ___ before? → * Wiser than.

d. * How did they arrive at the concert hall ___ had expected? → * More quickly than they.

Notice, incidentally, that so substitution for adjective phrases and subordinate clauses has a
variant that is reminiscent of questions. In addition to just substituting for the string of interest, as
illustrated earlier, so can move to the beginning of the sentence, triggering subject-aux inversion
- the same process that turns declarative sentences into yes-no questions. This variant of so
substitution is illustrated in (13) and (14).

(13) a. I am very happy, and so is Linda.

b. I am very fond of Lukas, and so is Linda.

c. I am very fond of my nephew, * and so is Linda of her niece.

(14) I { imagine, think } that they're invited, and so do they.

It clefts

The final constituent test that we'll consider is based on a special sentence type known as it
clefts. We begin by noting that ordinary sentences can often be divided into two parts: a part that
contains background information that is presupposed, the ground, and a part that is intended to
be particularly informative, the focus. In spoken language, this focus-ground partition (also
known as its information structure) is generally conveyed by intonation.3 In written language,
where intonation is difficult to represent, it is still possible to indicate a sentence's information
structure by fitting the focus and the ground into a syntactic frame consisting of it, a form of the
copula to be, and the subordinating conjunction that. In the examples in (15), the frame is in
black, the ground is in blue, and the focus is in red. Notice that a single sentence can be
partitioned into focus and ground in more than one way, giving rise to more than one it cleft.

(15) a. Ordinary cats detest the smell of citrus → It is ordinary cats that detest the smell of citrus
fruits. fruits.

b. Ordinary cats detest the smell of citrus → It is the smell of citrus fruits that ordinary cats
fruits. detest.

If a string can appear as the focus of an it cleft, then it is a constituent. Some examples for
various constituent types other than noun phrase are given in (16).
(16) a. Prepositional The cat strolled across the porch → It was with a confident air that the cat
phrase with a confident air. strolled across the porch ___.

b. Adjective Ali Baba returned from his travels → It was wiser than before that Ali Baba
phrase wiser than before. returned from his travels ___.

c. Adverb phrase They arrived at the concert hall → It was more quickly than they had
more quickly than they had expected that they arrived at the
expected. concert hall ___.

Because of their discourse function, it clefts don't always sound entirely natural out of the blue.
Nevertheless, it clefts where the focus is a constituent, as in (16), contrast sharply with the word
salad that results from attempting to focus a string that isn't a constituent, as in (17).

(17) a. Ordinary cats detest the smell of citrus → * It is the smell of that ordinary cats detest ___
fruits. citrus fruits.

b. The cat strolled across the porch with a → * It was with a confident that the cat strolled
confident air. across the porch ___ air.

c. Ali Baba returned from his travels wiser → * It was wiser than that Ali Baba returned from his
than before. travels ___ before.

d. They arrived at the concert hall more → * It was quickly than they had expected that they
quickly than they had expected. arrived at the concert hall more ___.

Some complications

Mismatches between syntactic structure and other structure

We mentioned earlier that it is not always self-evident whether a particular sequence of words is a
syntactic constituent. For instance, in reading a sentence like (18) out loud, we can perceive an
intonation break between cat and that (indicated by the slash).

(18) This is the cat / that chased the rat.

Because the intonation break is clearly audible, it is very tempting to equate the sentence's
abstract syntactic structure with its relatively concrete prosodic structure. Specifically, because
the and cat belong to the same prosodic constituent, it is tempting to treat the cat as a syntactic
constituent.
There are two pieces of evidence against doing so. First, as we have already seen in similar
examples, substituting a pronoun for the string the cat is ungrammatical in the context of (18)
(though not in other contexts).

(19) a. This is the cat that chased the rat. → * This is it that chased the rat.

b. I petted the cat. → ok I petted it.

Second, the string cat that chased the rat is shown to be a constituent by the grammaticality of
substituting the pro-form one. (One substitution is discussed in more detail in Chapter 5.)

(20) This is the cat that chased the rat. → This is the one.

The facts in (19a) and (20) converge to tell us that the word cat first combines with the relative
clause, not with the. Thus, (18) exhibits a mismatch between two types of linguistic structure:
syntactic and prosodic.

It is worth noting that the syntactic structure just described corresponds to the way that the
interpretation of the entire expression the cat that chased the rat is composed from the
interpretation of smaller expressions. In a simple semantics, the term cat denotes the set of all
cats. Combining cat with the relative clause yields cat that chased the rat, which denotes a
subset of all cats - namely, those with the property of having chased the rat. Further combining
cat that chased the rat with the definite article the yields as a denotation some unique individual
within the rat-chasing subset of cats (exactly which individual this is depends on the discourse
context).4

This correspondence of syntactic structure and semantic structure (the step-by-step composition
of the expression's meaning, as just illustrated for (18)) holds up as a first approximation, and it
is consistent with the correspondence between noun phrases and individuals, between adjective
phrases and properties, between prepositional phrases and locations, directions, etc., between
verb phrases and events, states, etc., and so on. Nevertheless, mismatches between syntactic
structure and semantic structure are possible. For instance, the sentence in (21) has two distinct
meanings, which can be paraphrased as in (22).

(21) Every student knows two languages.

(22) a. For every student, it is the case that he or she knows two languages.
(Abigail knows Arabic and Basque, Chris knows Chinese and Danish, Eric knows English and
French, ...)

b. There are two languages that every student knows.

(Arabic and Basque are known by Abigail, Chris, Eric, ...)
In the interpretation in (22a), the universal quantifier every is said to take scope over the number
two (EVERY > TWO). In the interpretation in (22b), the number takes scope over the universal
quantifier (TWO > EVERY). In either case, though, the ambiguous sentence itself (not the
paraphrases!) has a single syntactic structure. This is evident from the syntactic constituenthood
tests in (23), where the question and short answer pair are compatible with either scope
interpretation.

(23) a. Every student knows two languages. → Who knows two languages? Every student.

b. Every student knows two languages. → What does every student know? Two languages.

Other mismatches are also possible. Recall from the section on it clefts that one and the same
sentence can be associated with more than one information structure. Finally, mismatches
between syntactic and morphological structure are common; we discuss some examples in more
detail in Chapters 6 and 7.

False negative results

In a perfect world for syntacticians, the constituenthood tests would have perfect validity. That is, a
biconditional relation would hold between a string's being a constituent and passing the
constituenthood tests, as indicated in the penultimate column in (24). The examples presented so far
have been consistent with such a biconditional relation. However, the world is not made to order for
syntacticians, and it turns out to be possible for constituents to fail one or more constituenthood tests.
The actual state of affairs is thus as indicated in the rightmost column of (24).5

Ideal
Real world
world

(24) a. If a string passes the constituenthood tests, then it is a

TRUE TRUE
constituent.

b. If a string is a constituent, then it passes the constituenthood sometimes

TRUE
tests. FALSE

In other words, the failure of a string to pass a constituenthood test can be a false negative result.
In what follows, we present three such cases - constituents that fail at least some of
constituenthood tests

because they are words rather than phrases,

because they are finite, or
because they are contained within so-called syntactic islands.
Phrasal versus lexical constituents. Since single words are indivisible units, they are
constituents by definition.6 Nevertheless, they don't necessarily behave on a par with multiword
constituents. For instance, cats passes the constituenthood tests reviewed earlier in (25), but not
in (26).

(25) Cats are not social animals. → They are not social animals.

(26) a. The cats are hungry. → * The they are hungry.

b. Tabby cats are quite common. → * Tabby they are quite common.

c. Cats without tails are relatively rare. → * They without tails are relatively rare.

d. Those cats that have no tails are Manx cats. → * Those they that have no tails are Manx cats.

The reason for the grammaticality contrast in (25) and (26) is a systematic difference between
the syntactic contexts in these examples. In (26), cats is accompanied by a determiner or a
modifier of some sort, indicated by italics. In such contexts, cats combines with these other
words to form a noun phrase, but it isn't a noun phrase in its own right. In (25), on the other
hand, cats is a bare (= unmodified) noun. As such, it functions as a noun and as a noun phrase at
the same time. In other words, there are two levels of constituenthood: the lexical level, where
single words are constituents by definition, and the phrasal level, where single words don't
necessarily behave on a par with multiword constituents.

The constituenthood tests reviewed earlier turn out to be diagnostic only for phrasal constituents.
The ungrammatical results of attempting to move, question, and focus lexical constituents, rather
than phrasal ones, are illustrated in (27)-(29). The relevant lexical constituent is underlined, and
any words belonging with it to the same phrasal constituent are in italics.

(27) a. Attempt to I fed the cats. → * Cats, I fed the ___.

move:

b. The cat strolled across the porch → * With, the cat strolled across the porch
with a confident air. ___ a confident air.

c. Ali Baba returned from his travels → * Wiser, Ali Baba returned from his
wiser than before. travels ___ than before.

d. They arrived at the concert hall → * Quickly, they arrived at the concert hall
more quickly than they had more ___ than they had expected.
expected.

(28) a. Attempt to * What did you see the ___? → * Cats.

 question:

b. * How did the cat stroll across the → * With.

porch ___ a confident air?

c. * How did Ali Baba return from his → * Wiser.

travels ___ than before?

d. * How did they arrive at the concert → * Quickly.

hall more ___ than they had
expected?

(29) a. Attempt to Ordinary cats detest the smell of → * It is smell that ordinary cats detest the
focus: citrus fruits. ___ of citrus fruits.

b. The cat strolled across the porch → * It was with that the cat strolled across
with a confident air. the porch ___ a confident air.

c. Ali Baba returned from his travels → * It was wiser that Ali Baba returned
wiser than before. from his travels ___ than before.

d. They arrived at the concert hall → * It was quickly that they arrived at the
more quickly than they had concert hall more ___ than they had
expected. expected.

The examples in (30-(32), on the other hand, illustrate the grammatical results of moving,
questioning, and focusing phrasal constituents that happen to consist of a single word. (Notice
the absence of italicized material in this case.)

(30) a. Movement: I like cats. → ok Cats, I like ___.

b. Ali Baba returned from his → ok Wiser, Ali Baba returned from his
travels wiser. travels ___.

c. They arrived at the concert → ok Quickly, they arrived at the concert

hall quickly. hall ___.

(31) a. Question/short ok What do you like ___? → ok Cats.

answer:

b. ok How did Ali Baba return from → ok Wiser.

his travels ___?
 c. ok How did they arrive at the → ok Quickly.
concert hall ___?

(32) a. It cleft: Ordinary cats detest citrus. → ok It is citrus that ordinary cats detest.

b. Ali Baba returned from his → ok It was wiser that Ali Baba returned
travels wiser. from his travels ___.

c. They arrived at the concert → ok It was quickly that they arrived at

hall quickly. the concert hall ___.

Finiteness. Testing for the constituenthood of verb phrases is more complicated than is testing
for the constituenthood of other syntactic categories. First, there are no simple pro-forms for verb
phrases. The best we can do is to use the periphrastic forms do so for substitution and do what
for questions.7 (Notice that it's only what, rather than the entire pro-form do what, that moves to
the beginning of a question.)

(33) a. Substitution: She will write a book. → ok She will do so.

b. The two boys could order tuna salad → ok The two boys could do
sandwiches. so.

(34) a. Question/short What will she do? → ok Write a book.

answer:

b. What could the two boys do? → ok Order tuna salad

sandwiches.

Second and more importantly given our present focus on false negative results, verbs and the
verb phrases that contain them come in two varieties, finite and nonfinite (see Finiteness in
English for details). Now, two of the constituenthood tests - substitution and the question/short
answer test - yield grammatical results regardless of a verb phrase's finiteness, as shown in (35)
and (36).

(35) a. Substitution, nonfinite verb phrase: She will write a book. → ok She will do so.

b. finite verb phrase: She wrote a book. → ok She did so.

(36) a. Question/short answer, nonfinite verb phrase: What will she do? → ok Write a book.

b. finite verb phrase: What did she do? → ok Wrote a book.

But the results from the other two tests are more complex. Movement of nonfinite verb phrases is
grammatical,8 whereas movement of finite ones is not.

(37) a. Movement, nonfinite verb (She says that) she will write → ok (and) write a book, she will
phrase: a book, ___.

b. though she may write a book → ok write a book though she

may ___

(38) a. finite verb phrase: (She said that) she wrote a → * (and) wrote a book, she
book, ___.

b. though she wrote a book → * wrote a book though she

___

In it clefts, nonfinite verb phrases are marginally acceptable in focus, whereas finite verb phrases
are again clearly ruled out.

(39) a. It cleft, nonfinite verb phrase: She will write a book. → ? It is write a book that she will ___.

b. finite verb phrase: She wrote a book. → * It is wrote a book that she ___.

To summarize: we have good evidence that nonfinite verb phrases are constituents. In the case of
finite verb phrases, we have evidence for constituenthood from two of the four constituenthood
tests. Given this slightly complex state of affairs, we will proceed as follows. We will make the
simplifying assumption that the ungrammaticality of moving or focusing finite verb phrases has
nothing to do with their constituenthood, but that it is due to some other reason, yet to be
determined. (In fact, we will give you a crack at solving the problem in a later chapter.) Having
made this assumption, we are free to treat finite verb phrases as constituents on a par with their
nonfinite counterparts even though the syntactic behavior of the two types of verb phrases is not
identical in all respects.

Chances are that you are a bit leery of the simplifying assumption just described. If so, think of it
as comparable to taking out a loan. True, taking out a loan is risky, and taking out loans in a
careless or irresponsible way can lead to financial disaster. Nevertheless, the credit market is a
necessary and productive part of any modern economy. In a similar way, making simplifying
assumptions in science can help us to make progress where we would otherwise be stumped by
the complexity of the phenomena that we are investigating. Of course, we have to be careful
about what simplifying assumptions we make. Otherwise, we end up fooling ourselves into
believing that we are making progress, when in fact we are working on such a distorted model of
reality that our work is worthless.
Apart from this wrinkle concerning finiteness, verb phrases behave just as we have come to
expect from other constituent types. The tests yield grammatical results for verb phrases, but not
for verbs.

(40) a. Substitution: She will write a book. → * She will do so a book.

b. Movement: (She says that) she will write a → * and write, she will ___ a book.
book,

c. " though she may write a book → * write though she may ___ a
book

d. Question/short * What will she do a book? → * Write.

answer:

e. It cleft: She will write a book. → * It is write that she will ___ a
book.

And once again, particular strings can be phrasal constituents in one syntactic context, but not in
another. For instance, write isn't a phrasal constituent when it combines with a direct object, but
it is when used on its own. This is the source of the grammaticality contrast between (40) and
(41).

(41) a. Substitution: She will write. → ok She will do so.

b. Movement: (She says that) she will write, → ok and write, she will ___.

c. " though she may write → ok write though she may ___

d. Question/short answer: What will she do? → ok Write.

e. It cleft: She will write. → ? It is write that she will ___.

Islands. In (42a), the doctors is a constituent, as is evident from the possibility of substituting a
pronoun for the string, as in (42b).

(42) a. We should invite the lawyers and the doctors.

b. We should invite the lawyers and them. (pointing to the doctors)

But although the doctors passes the substitution test, the other three tests yield ungrammatical
results.
(43) a. * The doctors, we should invite the lawyers and ___.

b. * Who should we invite the lawyers and ___? The doctors.

c. * It is the doctors that we should invite the lawyers and ___.

Earlier, we pointed out that question formation can be thought of as a combination of

substitution and movement. The parallel between movement and question formation in
(43a,b) and the it cleft in (43c) suggests that the latter, too, involves movement - specifically,
movement of the focus to the position preceding that. In the remainder of this section, we
will therefore use the term 'movement' in a broad sense to include all three constituenthood
tests instantiated in (43).

Notice that there is nothing semantically wrong with (43), since it is possible to express the
intended meaning grammatically, as shown in (44).

(44) a. ok The doctors, we should invite ___ together with the lawyers.

b. ok Who should we invite ___ together with the lawyers? The doctors.

c. ok It is the doctors that we should invite ___ together with the lawyers.

Taken together with the grammaticality of (42b), the contrast between (43) and (44) shows that
movement of the noun phrase the doctors is somehow prevented by the syntactic configuration in
(43). Ross 1967 introduced the metaphorical term island for configurations that block movement
where one would otherwise expect it to be possible. The idea underlying the term is that the
constituents that might be expected to move, but can't, are stranded on an island like castaways.

Ross identified several types of islands, including coordinate structures, the type under
discussion here,9 and his discovery has given rise to a huge body of literature on the topic. Our
purpose here is neither to catalog the types of islands nor to pursue the proper linguistic analysis
of them (this is done in Chapters 11 and 12), but simply to draw attention to the fact that
constituenthood tests based on movement will yield false negative results for phrasal constituents
if they happen to be contained in islands.

Representing syntactic constituenthood

In Chapter 1, we introduced tree diagrams as a convenient way of representing syntactic

structure. For a mathematician working in the field of graph theory, the formal properties of tree
diagrams are interesting in their own right, but for a syntactician, the interest of trees lies in the
fact that they are representations, or models, of constituent structure. In other words, the
graphic structure of a tree on the page is intended as a statement about the way that speakers
group together syntactic elements in their minds. In a good model, we want the properties of the
model to correspond straightforwardly to the properties of the domain of inquiry. Such a close
correspondence allows us to state observations and generalizations about the domain of inquiry
without undue complication. Moreover, if we're lucky, we might even be able to use our
understanding of the model's formal properties as a sort of conceptual lever to generate
hypotheses and to discover facts and generalizations about the domain of inquiry that would
otherwise escape notice.

In light of these considerations, let's consider the sentence in (45), focusing particularly on the
constituenthood of the underlined string.

(45) The secretary drafted the letter.

According to the two tests that apply to finite verb phrases, the string drafted the letter is a
constituent.

(46) a. Substitution: The secretary drafted the letter. → The secretary did so.

b. Question/short answer: What did the secretary do? → Drafted the letter.

Having established this fact, let's now consider two alternative representations of the sentence.
We've already encountered (47a) in Chapter 1. (47b) is an alternative, 'flatter' tree.

(47) a. b.

At first glance, the flatter tree might seem preferable on the grounds that it is simpler in the sense
of containing fewer nodes. But let's focus on the question of which tree is a better representation
of the sentence. Another way of putting this question is to ask whether either of the trees in (47)
has some graphic property that corresponds to the results of the constituenthood tests in (46). In
(47a), the answer is 'yes,' since there is a single node (the one labeled VerbPhr) that
exhaustively dominates the string drafted the letter (see the section on exhaustive dominance in
Node relations for a definition). The tree in (47b), on the other hand, lacks such a node and has
no other graphic property that corresponds to the string's constituenthood. Clearly, then, (47a) is
a better representation of the sentence, because it follows the natural convention in (48).

(48) Syntactic constituents are represented graphically as nodes in a tree.

We will conclude this discussion of the model character of syntactic representations by
emphasizing that models are just that - models, and not the actual domain of inquiry itself. The
purpose of any model is to help us understand some part of reality that is too complex to
understand in all of its detail, at least all at once. This means that models are partial in two
respects. First, models often leave out many properties of a phenomenon that aren't relevant from
a particular point of view. This fact is often stated in the form of the maxim "Don't mistake the
map for the territory." For instance, a mountaineer's map might show topographical information
in great detail, but completely ignore political boundaries, whereas a diplomat's map might do
the reverse. Analogously, in linguistics, syntactic models leave out many important properties of
language, such as real-world plausibility, pragmatic felicity, the location of intonation breaks,
and so on. These are the focus of other subdisciplines of linguistics.

A second way that models are partial is that they are subject to revision as our understanding of a
particular domain improves and deepens. Another way of saying this - which perhaps sounds
more impressive - is to say that scientific progress is possible.

Notes

1. As we mentioned in Chapter 1, the grammaticality of a sentence depends on its interpretation.

Specifically, (10a), repeated here as (i), is ungrammatical under the ordinary interpretation where the
prepositional phrase with long, fluffy tails modifies cats.

(i) The cats, I fed with long, fluffy tails.

(i) also has an outlandish interpretation that can be paraphrased as I fed long, fluffy tails to the
cats. Under this interpretation, (i) is grammatical. In other words, in the pre-movement version of
(i) given in (ii), the string the cats is a constituent in the outlandish interpretation, though not in
the ordinary one.

(ii) I fed the cats with long, fluffy tails.

Conversely, the string the cats with long, fluffy tails is a constituent in the ordinary interpretation
of (ii), but not in the outlandish one. This is evident from the fact that (iii) has only the ordinary
interpretation.

(iii) The cats with long, fluffy tails, I fed. (The other cats, I didn't.)

2. Under certain discourse conditions, English allows the question word to remain in the place
where it substitutes (in situ), as illustrated in (i). See Information versus echo questions for more
discussion.
(i) a. Who did you see? (information question or echo question)

b. You saw who? (only echo question)

3. It is worth pointing out that focus-ground partitioning is relevant not just for it clefts, but also
for questions and (short) answers. The focus in a question is the unknown information expressed
by the question word. A short answer to a question consists only of a focus. Repeating the
ground of the question yields a full answer.

(i) a. Question: What animals detest the smell of citrus fruits?

b. Short answer: Ordinary cats.

c. Full answer: Ordinary cats detest the smell of citrus fruits.

(ii) a. Question: What do ordinary cats detest?

b. Short answer: The smell of citrus fruits.

c. Full answer: Ordinary cats detest the smell of citrus fruits.

4. Notice that the resultant interpretation is distinct from the one that would result from first
combining cat and the and then combining the cat with the relative clause. The denotation of the
cat is a unique member of the set of cats. Combining the cat with the relative clause would
attribute to this unique entity the property of having chased the rat. Given that the cat already
denotes a unique entity, the property of having chased the rat wouldn't be a defining property of
the cat in question; it would simply be an additional, more or less accidental one. The
interpretation in question is possible semantically, and it can be expressed by using a non-
restrictive relative clause, as in (i).

(i) This is the cat, which chased the rat.

But (i) is not synonymous with (18), where the rat-chasing property is restrictive (= defining).

5. In contrast to (24b), the statement in (i), derived from (24a) by the modus tollens rule of
propositional logic, is true.

(i) If a string isn't a constituent, then it doesn't pass the constituenthood tests.

6. For present purposes, we disregard syntax-morphology mismatches of the sort alluded to

earlier and discussed in Chapters 6 and 7.
7. For completeness, we should mention that do so substitution and the question test for verb
phrases are subject to a semantic restriction. Specifically, do so and do what cannot substitute for
verb phrases with so-called stative verbs like know or want.

(i) a. They know her parents; they want the cookies. → * They do so.

b. What do they do? → * Know her parents; want the cookies.

As their name implies, stative verbs refer to states (rather than to activities or accomplishments),
and a reasonably reliable diagnostic for them is their inability to appear in the progressive
construction.

(ii) a. Stative verb * They are knowing her parents; they are wanting the cookies.

b. Nonstative verb ok They are meeting her parents; they are eating the cookies.

Since do is the prototypical activity verb, it is not surprising that expressions containing it, like
do so and do what, give rise to a semantic clash when they substitute for verb phrases containing
stative verbs.

8. Though movement of nonfinite verb phrases in out-of-the-blue contexts, as in (i), is not very
felicitous, it is clearly grammatical given appropriate discourse contexts, as the examples in the
text show.

(i) Write a book, she will ___.

9. The island metaphor is not perfect. Although constituents can't move out of an island, islands
as a whole are able to move, as shown in (i).

(i) a. ok The doctors and the lawyers, we should invite ___.

b. ok Who should we invite ___? The lawyers and the doctors.

c. ok It is the lawyers and the doctors that we should invite ___.

Exercises and problems

Exercise 2.1
Using the constituenthood tests reviewed in this chapter, determine whether the underlined
strings in the following sentences are (phrasal) constituents. Be sure to provide the evidence on
which your conclusions are based.

(1) a. I put the car in the garage.

b. I put the car in the garage.

c. I put the car in the garage.

(2) a. I know the guy with the fedora.

b. I know the guy with the fedora.

(3) a. They threw in the towel.

b. They threw the towel in the closet.

Exercise 2.2

How well does each of the trees in (1) and (2) represent the syntactic structure of the sentence it
is intended to represent? Your discussion should be concise, but detailed enough to answer the
following questions:

Are any strings represented as constituents that shouldn't be?

Are any strings not represented as constituents that should be?
Are any of the trees misleading in other respects?

State the linguistic evidence on which your conclusions are based. (If you have completed
Exercise 2.1, you can simply refer to the evidence there rather than repeating it.)

(1) a. b.
 c.

(2) a. b.

c.

Problem 2.1

The substitution test introduced in Chapter 1 and discussed in further detail in this chapter and
the substitution operation introduced in Chapter 1 are not identical, but they are related. In a few
sentences, explain how.

Problem 2.2

It is sometimes argued that coordination (also known as conjunction) is a test for syntactic
constituenthood. Evaluate the argument, using the facts in (i) and others of your own devising.

(i) a. We saw my youngest sister and her husband.

b. Jim saw the yarn and Kim, the pattern.

c. Jim has probably seen the yarn and Kim, the pattern.

d. the older and the younger children

3 Some basic linguistic relations

Argumenthood
o Semantic valency
o Transitivity
Modification
Predication
o Expletive it
o Aristotelian versus Fregean predicates
o Expletive there
o Some special cases
Notes
Exercises and problems
Supplementary material
o Adjectives
o Expletive elements in English
o Finiteness in English
o Grammatical relations
o Reference and related notions
o Thematic roles

In Chapters 1 and 2, we presented various pieces of evidence for the existence of syntactic
structure in human language. The facts presented there raise a basic question - what is the basis
of syntactic structure? In this chapter, we introduce three fundamental linguistic relationships
that underlie syntactic structure. Two of these relationships, argumenthood and modification, are
at bottom semantic relationships (although the expression of argumenthood is more constrained
in natural language than purely semantic considerations would dictate), whereas the third,
predication, is purely syntactic.

Argumenthood

Semantic valency

The most obvious factor that determines how vocabulary items combine has to do with their meaning, a
point most conveniently illustrated with verbs. From the point of view of a simple formal semantics, the
verb laugh is a function from entities to truth values, as illustrated in (1). Entities that laugh are
associated with the value T(rue); entities that don't with the value F(alse). In the world described in (1),
Beatrice, Gary, Lukas, and Tina laugh, and Chris and Eva don't.

By convention, entities are indicated by boldface, sets are enclosed in curly brackets, and
ordered pairs are enclosed in angle brackets. It is also conventional to indicate denotations of
expressions by enclosing the expressions in special square brackets. These special brackets
 are not part of the HTML character set, so we use two ordinary square brackets instead.

(1) [[ laugh ]] = { Beatrice → T,

Chris → F,

Eva → F,

Gary → T,

Lukas → T,

Tina →T}

Laugh can combine with a single argument, which denotes an entity. Intuitively, we can think of
arguments as the central participants in a situation. Combining laugh with an argument (say,
Lukas) has a syntactic effect and a corresponding semantic effect. The syntactic effect is to yield
the sentence in (2a). (For simplicity, we disregard the past tense morpheme -ed here and in what
follows.) The corresponding semantic effect is to apply the function in (1) to the argument; that
is, to select the entity denoted by the argument in the function in (1) and to return the associated
value. In the example at hand, the sentence comes out as true, as shown in (2b).

(2) a. Lukas laughed.

b. T

On the other hand, combining Chris with laughed yields Chris laughed with a truth value of F.

In addition to denoting simple functions, verbs can also denote recursive functions. For instance,
a transitive verb denotes a function from entities to a second function, the latter of the same type
as just described for the intransitive verb laugh (a function from entities to truth values). So the
transitive verb invite might denote the function in (3).

(3) [[ invite ]] = { Chris → ( Andrew → T ) ,

David → ( Andrew → T ) ,

Eddie → ( Andrew → F ) ,

Chris → ( Brian →F),

 David → ( Brian →F),

Eddie → ( Brian →T)}

Combining invite with a theme argument (say, David) has the syntactic effect of yielding the
phrase in (4a). As before, the corresponding semantic effect is to select the entity denoted by the
argument in (3) and to return the associated values, as shown in (4b).

(4) a. invited David

b. [[ invited David ]] = { ( Andrew → T ) ,

( Brian →F)}

Further combining invited David with an agent argument (say, Andrew) yields the sentence in
(5a) and the truth value in (5b). This second step in the derivation of a transitive sentence is
exactly equivalent to the first and only step that is necessary in an intransitive sentence.

(5) a. Andrew invited David.

b. [[ Andrew invited David ]] = T

It is important to understand that the order of the arguments in (3) reflects derivational order (the
order in which the arguments combine structurally), not their superficial linear order. Given
purely semantic considerations, it is equally easy to write functions in which derivational order is
congruent with linear order, and you are asked to do so in Exercise 3.1.

Verbs like laugh and invite are instances of one-place and two-place predicates, respectively.
The term predicate here refers to a vocabulary item, with a focus on its capacity to combine with
one or more arguments. The number of arguments that a predicate requires is its semantic
valency.

The relations denoted by predicates can involve more than two arguments. An example of a
three-place predicate is give, which denotes the relation among a set of givers, a set of gifts, and
a set of recipients. Even more complex relations are possible. For instance, rent is a five-place
predicate denoting a relation among landlords or other sorts of owners, tenants, rental property,
amounts of money, and lengths of time (lease terms).

Transitivity

In principle, a predicate's valency might completely determine the syntactic structure that it appears in.
The ungrammaticality of the sentences in (7) would fall out directly from such a system.
(7) a. * Lukas laughed the train. (one-place predicate; superfluous argument)

b. * Andy invited. (two-place predicate; missing theme argument)

The actual situation, however, is more complex. For instance, eat denotes a relation between
eaters and food. It is therefore a two-place predicate, like invite. However, unlike invite, eat has
both a transitive and an intransitive use, as illustrated in (8).

(8) a. Transitive: The children have eaten their supper.

b. Intransitive: The children have eaten.

Notice that the semantic properties of eat remain constant in (8). In other words, (8a) and (8b)
are both interpreted as involving the ingestion of food, even though there is no explicit mention
of food in (8b).

In view of the mismatch between the semantic and syntactic properties of eat in sentences like
(8b), it is useful to distinguish between semantic and syntactic arguments. As mentioned earlier,
we can think of semantic arguments as central participants in a situation. Syntactic arguments, on
the other hand, are constituents that appear in particular syntactic positions (see Chapter 4 for
further discussion). Semantic arguments are typically expressed as syntactic arguments, but the
correspondence between the two is not perfect, as (8b) shows.

We will use the term transitivity to refer to the number of syntactic arguments that a verb
combines with, and we can then divide verbs into three subcategories as in (9).

We are using the term 'transitivity' in a slightly unorthodox way. Traditionally, the term refers to
the number of a verb's objects, which is one less than the number of its arguments. Thus, as the
terms imply, an intransitive takes no objects, and a ditransitive takes two.

(9) Degree of transitivity Number of syntactic arguments

Intransitive 1

Transitive 2

Ditransitive 3
Because of mismatches as in (8), it turns out to be quite rare for verbs to belong to just one
syntactic subcategory. (10) shows some two-place verbs besides eat that can be used either
transitively or intransitively. The slashes separate the arguments from the predicate and each
other.

Basically transitive Intransitive use

(10) a. He / interrupted / the meeting. He / interrupted.

b. Amy / knits / sweaters. Amy / knits.

c. They / are reading / a book. They / are reading.

Conversely, certain one-place verbs can be used not only intransitively, but also transitively, as
illustrated in (11). Notice that the verb and its object in the transitive examples are
etymologically related, or cognate. For this reason, the transitive use of one-place verbs as in
(11) is known as the cognate object construction.

Basically intransitive Transitive use

(11) a. Dennis / died. Dennis / died / a peaceful death.

b. Lukas / laughed. Lukas / laughed / an infectious laugh.

c. Mona Lisa / was smiling. Mona Lisa / was smiling / a mysterious smile.

Further, it is possible to use some basically three-place verbs not just ditransitively, but
transitively and even intransitively.

Basically ditransitive Transitive use Intransitive use

(12) a. We / teach / college students / syntax. We / teach / college students. We / teach.

We / teach / syntax.

b. He / told / me / the whole story. He / told / me. He / better not tell.

He / told / the whole story.

Finally, it is possible to use basically two-place verbs ditransitively.

Basically transitive Ditransitive use

(13) a. I / baked / a delicious cake. I / baked / my friends / a delicious cake.

b. She / sang / a lullaby. She / sang / her baby / a lullaby.

Modification

Events are associated with more or less central participants and properties. The central
participants are the semantic arguments just discussed. Properties of a situation typically taken to
be less central, such as manner, time (point in time, duration, frequency), place (location, origin,
destination), reason (cause, purpose), and so on, can be expressed by modifiers.

Arguments and modifiers both introduce restrictions on the denotation of a predicate, and the
relationships of argumenthood and modification do not differ in this respect. For instance, the
situations denoted by invite David are a subset of those denoted by invite, just as the situations
denoted by laugh uproariously are a subset of those denoted by laugh.

Modifiers of verb phrases are typically adverbial phrases or prepositional phrases, but noun
phrases can serve as modifiers as well (you will be asked to illustrate this latter fact in the
Exercises). In the following examples, the modifier is in italics, and the verb phrase that it
modifies is underlined.

(14) a. Manner: He read the letter carefully.

b. Point in time: They discussed the proposal in the afternoon.

c. Duration: You should keep your tax records for several

years.

d. Frequency: I read the Times quite often.

e. Location: We met in my office.

f. Origin: We set out from Bangalore.

g. Destination: We arrived in Benares.

h. Cause: He threw it away out of spite.

i. Purpose: I sent the message to warn everyone.

Because of their semantically peripheral character, modifiers are syntactically optional. The
converse is emphatically not true, however. Not all syntactically optional constituents are
modifiers; recall from (8b) that semantic arguments aren't always expressed.

Verb phrases are not the only category that can be modified. For instance, nouns are often
modified by adjective phrases, prepositional phrases, or relative clauses.1

(15) a. a very important period

b. a period of great import

c. the car that just turned the corner

Moreover, adjective phrases and prepositional phrases, the quintessential modifiers, can
themselves be modified.

(16) a. very proud of her progress

b. surprisingly good to eat

(17) a. almost in the dark

b. right behind the shed

Predication

The two linguistic relations discussed so far - argumenthood and modification - are basically semantic
notions that are optionally expressed in the syntax. In this section, we introduce a third relation,
predication, which differs from argumenthood and modification in being an irreducibly syntactic
relation. By this, we mean that predication is not always semantically motivated.

Expletive it

In (18a), the italicized that clause functions as the sole syntactic argument of the adjective
evident, on a par with the noun phrase in (18b). (For simplicity, we disregard the copula as
semantically vacuous.)

(18) a. That they are corrupt is evident.

b. Their corruption is evident.

An indication of the semantic equivalence of the two expressions is the fact that they can both
serve as a short answer to the question in (19a).

(19) a. What is evident?

b. That they are corrupt.

Their corruption.

In addition to (18a), a synonymous variant, (20), is available in which the that clause appears at
the end of the entire sentence. The original position of the that clause is occupied by the
expletive pronoun it.

The term 'expletive' means that the pronoun does not refer to a discourse entity in the
ordinary way that pronouns do (see Referential versus expletive it for more discussion
concerning the distinction between these two uses of it).

(20) It is evident that they are corrupt.

Given that the that clause satisfies the semantic requirement of evident for an argument in both
(18a) and (20), the presence of the expletive pronoun in (20) is striking. From a semantic point of
view, it is unnecessary, and one might therefore expect it to be optional. But this is not the case,
as the ungrammaticality of (21) shows.

(21) * Is evident that they are corrupt.

The ungrammaticality of (21) leads us to conclude that there exists a purely syntactic well-
formedness condition requiring all clauses to have a subject.

Earlier, we saw that it is possible for arguments to be semantically necessary and yet not to be
expressed in the syntax. Expletive subjects represent roughly the converse of this situation, being
cases where an expression that is not motivated by semantic considerations is nevertheless
obligatory in the syntax.

Aristotelian versus Fregean predicates

We will refer to the requirement just mentioned as the subject requirement. According to it, every
clause consists of a subject and a predicate (independently of semantic requirements). The term
'predicate' as used here has a different sense than in our earlier discussion concerning argumenthood; it
refers to what remains of a clause when its subject is removed.
 Remember not to confuse 'subject' with 'simple subject.'

For clarity, we can use the term 'Aristotelian predicate' for this sense, since the observation that
all sentences consist of a subject and a predicate goes back to Aristotle (384-322 BCE).
Predication is the relation between a subject and an Aristotelian predicate. So (22a) and (22b)
are two alternative ways of stating the subject requirement.

(22) a. Every clause has a subject.

b. Every clause is an instance of predication.

The sense of 'predicate' that we used earlier, in which the term refers to a single vocabulary item,
is much more recent and can be attributed to one of the founders of modern logic, the
mathematician and philosopher Gottlob Frege (1848-1925). Accordingly, we can use the term
'Fregean predicate' for this sense. What Frege made explicit is that Aristotle's division of a clause
into subject and predicate is simply the first of a potential series of such bifurcations. Just as it is
possible to peel off, as it were, the subject of a clause, leaving the Aristotelian predicate, it is
possible to further peel off any arguments (and modifiers) contained within the Aristotelian
predicate, yielding in the final instance a single vocabulary item, the Fregean predicate.

Fruitful as Frege's analytic insight is, we should not let it obscure the key difference between
subjects and other constituents of a clause: namely, that constituents contained within an
Aristotelian predicate must be licensed by semantic considerations (in other words, these
constituents are present because of semantic considerations), whereas the subject, which is
external to the Aristotelian predicate and combines with it, is required independently of semantic
considerations.

In the following examples, the Aristotelian predicate of the largest clause is in italics, and the
Fregean predicate is underlined. As the increasingly complex sentences show, Aristotelian
predicates are recursive categories. Fregean predicates, on the other hand, not being phrases, are
not.

(23) a. The tabby cat enjoys catnip immensely.

b. They have found that the tabby cat enjoys catnip immensely.

c. My downstairs neighbor suspects that they have found that the tabby cat enjoys catnip
immensely.

Expletive there
In English, further evidence for the purely syntactic character of the subject requirement comes from
the expletive there construction. (24) illustrates an ordinary sentence and the corresponding expletive
there construction (see Adverbial versus expletive there for more discussion concerning the expletive
use of there).

(24) a. Several vexing questions remain.

b. There remain several vexing questions.

Just as expletive it occupies the position that would otherwise be occupied by a clausal subject,
expletive there occupies the position that would otherwise be occupied by a noun phrase subject.
And just as in the case of expletive it, omitting expletive there results in ungrammaticality.

(25) * Remain several vexing questions.

It should be pointed out that not every English sentence has an expletive there counterpart.
Rather, expletive there is subject to a licensing condition (a necessary condition for its
occurrence) that can be stated roughly as in (26).

(26) Expletive there must be the subject of a verb of existence or coming into existence.

In the following examples, the (Fregean) predicate licensing expletive there is highlighted in
green.

(27) a. After their military defeat, there arose among the Plains tribes a powerful spiritual movement.

b. There is a problem.

c. There began a reign of terror.

d. In the end, there emerged a new caudillo.

e. There ensued a period of unrest and lawlessness.

f. There exists an antidote.

g. There follows a section on the care of gerbils.

h. There has occurred an unfortunate incident.

i. There remains a single course of action.

Predicates that aren't verbs of (coming into) existence don't license expletive there. This is the
reason that the following examples are ungrammatical; the non-licensing (Fregean) predicates
are highlighted in red.

(28) a. * There came more than sixty dignitaries.

b. * There continued the same problem.

c. * There rang the mail carrier.

d. * There sang an impressive choir from Russia.

e. * There walked a poodle into the room.

Some special cases

Nonfinite clauses. The instances of predication provided so far have all been finite clauses like
those in (29).

(29) a. He laughed uproariously.

b. It will seem that they won the game.

c. There is a problem.

Nonfinite clauses like those in (30) are also instances of predication; the clauses at issue are set
off by brackets.

(30) a. We expected [ him to laugh uproariously ].

b. We expected [ it to seem that they won the game ].

c. We expected [ there to be a problem ].

At first glance, it might seem preferable to treat the italicized noun phrases in (30) as objects of
expected, rather than as subjects of the embedded nonfinite clause the way we have done.
However, such an approach faces at least two difficulties. First, the relation between the
italicized and the underlined constituents in the nonfinite embedded clauses in (30) is analogous
to the relation between the undoubted subjects and predicates of the finite embedded clauses in
(31).
(31) a. We expected that [ he would laugh uproariously ].

b. We expected that [ it would seem that they won the game ].

c. We expected that [ there would be a problem ].

Second, in (30a), the thematic relation of agent that the noun phrase him bears to the phrase to
laugh uproariously is the same as that between the subject he and its predicate laughed
uproariously in (31a). If him were the object of expected rather than the subject of the nonfinite
clause, we would be forced to admit the otherwise unprecedented pairing of the thematic role of
agent with the grammatical relation of object.

Small clauses. Because of the parallel between the nonfinite and finite embedded clauses in (30)
and (31), it makes sense to treat to in to infinitive clauses as the nonfinite counterpart of a modal
like would. There also exist instances of predication without any overt counterpart to a modal at
all.2 Such instances of predication are called small clauses (the idea behind the name is that the
absence of a modal element makes them smaller than an ordinary clause). (32)-(35) provide
some examples of small clauses; the captions indicate the syntactic category of the small clause's
(Aristotelian) predicate.

(32) a. Adjective phrase We consider [ the proposed solution completely inadequate ] .

b. They proved [ the theory false ] .

(33) a. Noun phrase They called [ the actor a traitor ] .

b. I consider [ Mark Judy's closest collaborator ].

(34) a. Prepositional phrase They made [ him into a star ] .

b. I want [ everyone off the boat ] .

(35) a. Verb phrase (bare verb) God let [ there be light ] .

b. Verb phrase (gerund) I hear [ the cat scratching at the door ] .

Small clauses are typically arguments of verbs, but they can also be arguments of (certain)
prepositions. This is illustrated in (36) for with.

(36) a. Adjective phrase: With the weather much less turbulent, flights were able to resume for the
first time in days.
 b. Noun phrase: With his wife an airline industry lobbyist, the senator's support for the
bailout was hardly surprising.

c. Prepositional With all of their three kids in college, their budget is pretty tight.
phrase:

d. Verb phrase With the parade passing right outside her living-room window, Jenny could
(gerund): not have had a better view of it.

Imperatives. Imperative sentences like (37) appear to lack a subject.

(37) Come over here.

There is reason to believe, however, that they contain a second-person subject comparable to the
pronoun you except that it is silent (the "you understood" of traditional grammar). For one thing,
(37) has the variant in (38) in which the subject is explicitly expressed.

(38) You come over here.

Another reason to assume that all imperatives contain a silent, yet syntactically active subject is
that the grammaticality pattern in (39), where the subject is overt, has an exact counterpart in
(40).

(39) a. You shave { yourself, yourselves. }

b. * You shave you.

c. * You shave themselves.

(40) a. Shave { yourself, yourselves. }

b. * Shave you.

c. * Shave themselves.

Notes

1. The alert reader will notice that in the examples we give, it is verb phrases, adjective phrases,
and prepositional phrases that are modified, but nouns (not noun phrases). You will be able to
explain this apparent asymmetry after reading Chapter 5.
2. The alert reader will notice that small clauses are exceptional in another regard: they are a
further instance of constituents in which the constituenthood tests of Chapter 2 yield false
negative results (at least for most speakers).

Exercises

Exercise 3.1

A. Imagine a language Hsilgne that is exactly like English except that transitive predicates combine first
with the agent, and then with the theme. Does (1) mean the same thing in Hsilgne as it does in English?
Explain, using the discussion in connection with (3)-(5) in the text as a model.

(1) Brian invited David.

B. Are you sure that you speak English and not Hsilgne? What's the evidence for your
conclusion?

Exercise 3.2

A. In your own words, discuss the difference between the terms 'modify' and 'refer'. Feel free to use
illustrative examples, but be as concise as you can.

B. In your own words, discuss the difference between 'modify' and 'predicate'. Feel free to use
illustrative examples, but be as concise as you can.

C. In traditional grammar, the term 'modifier', which refers to any expression that adds
information about some other expression, is often used interchangeably with the term 'adjective'.
By contrast, it is customary in linguistics to distinguish between the two terms and to define
adjectives on the basis of morphosyntactic criteria rather than function. For instance, many
adjectives can appear in a comparative (better, more acceptable) or superlative (best, most
acceptable) form. Moreover, in many languages (though not in English), adjectives are inflected
for case, number, and gender to agree with the nouns that they modify or that they are predicated
of.

Assuming the linguistic rather than the traditional definition of 'adjective', give an example of an
adjective (or adjective phrase) that is not a modifier, and give an example of a modifier that is
not an adjective. No discussion is necessary beyond the examples.

D. This part of the exercise is designed to help you get a better feel for the concept of 'small
clause,' which many students find difficult. Make up three sentences that you think might contain
small clauses. Clearly indicate the sequence that you think is the small clause, and note how sure
you are whether it is a small clause or not. You should feel free to make up three examples about
which you are unsure, since that way, your examples and our feedback on them will be
maximally informative to both of us.

Exercise 3.3

Modifiers are typically expressed by adverbial phrases or prepositional phrases, as in the examples in
the body of the text, but they can also take the form of noun phrases, as in (1).

(1) Manner: They solved the problem another way.

Do your best to provide examples of noun phrase modifiers for the other types of modification
illustrated in the chapter, repeated for convenience in (2) (= (14b-i) in the text). (This is not
necessarily possible for every type.)

(2) a. Point in time: They discussed the proposal in the

afternoon.

b. Duration: She kept their books for five years.

c. Frequency: I read the Times quite often.

d. Location: We met the students in my office.

e. Origin: We set out from Bangalore.

f. Destination: We arrived in Benares.

g. Cause: He threw it away out of spite.

h. Purpose: I sent the message to warn everyone.

Exercise 3.4

A. Assuming the licensing condition on expletive there in (26), repeated here as (1), explain for each of
the grammaticality judgments in (2) whether it is expected or not. The brackets indicate clause
boundaries (of to infinitive clauses or small clauses) and are added for clarity. Assume the judgments
given even if you do not share them.

(1) Expletive there must be the subject of a verb of existence or coming into existence.

(2) a. ok Feynman suspected [ there to be a problem with the O-ring ] .

 b. * Feynman suspected [ there a problem with the O-ring ] .

c. ok There was suspected [ to be a problem with the O-ring ] .

d. * There was suspected [ a problem with the O-ring ] .

B. Is the following argument valid?

The grammaticality contrast in (3) shows that expletive there cannot be an object. In (4), it therefore
cannot be the object of want, but must be the subject of the to infinitive clause.

(3) a. ok There is a fly in the soup.

b. * I dislike there in the soup.

(4) ok I don't want there to be a fly in the soup.

Exercise 3.5

A. Discuss the syntactic difference(s) between the two sentences in (1), focusing on the concepts
introduced in this chapter. (Source: http://www.meredith.edu/grammar/modifier.htm)

(1) a. Winston considered the judges careful.

b. Winston considered the judges carefully.

B. According to the source of the sentences in (1), consider is a linking verb in (1a). This is
nonsense, but what might have led the author of the website to think so?

Exercise 3.6

Using the concepts introduced in this chapter and the supplementary readings, explain why the
following instances of linguistic humor are funny.

(1) a. Greeks Fine Hookers

b. Lawmen from Mexico Barbecue Guests

c. Lawyers Give Poor Free Legal Advice

d. Lung Cancer in Women Mushrooms

(2) Q. What did the Zen master say to the guy at the hot dog stand?
A. Make me one with everything.

(3) It used to be that if someone spilled coffee in their lap, they simply called themselves clumsy.
Today, too many people are calling themselves an attorney. (http://www.mlaw.org/wwl,
accessed 30 Jan 03)

(4) The comedian Dick Gregory tells of walking up to a lunch counter in Mississippi during the days of
racial segregation. The waitress said to him, "We don't serve colored people." "That's fine," he
replied, "I don't eat colored people. I'd like a piece of chicken."
(Steven Pinker. 1994. The language instinct. How the mind creates language. New York: Morrow.
115.)

Problem 3.1

A. Nonfinite clauses like the bracketed sequence in (1) are prima facie counterexamples to the subject
requirement.

(1) I promised [ to come on time ].

Provide as much evidence as you can for the existence of a silent subject in (1) and nonfinite
clauses like it.

B. The availability of (2) in vernacular usage might tempt one to conclude that expletive it is
optional, and that the subject requirement is not absolute.

(2) Seems like they're finally getting somewhere.

Does (2) really show that expletive subjects are optional? Discuss, providing evidence.

Problem 3.2

The expletive there sentences in (1) are acceptable in modern English (though quite formal in style).
Discuss.

(1) a. At the end of the intermission, there sounded a silvery bell.

b. Then the curtain rose, and there waltzed onto the stage an exquisitely, but strangely dressed
apparition.
4 Introducing the X' schema of phrase structure

The X' schema for elementary trees

o Transitive elementary trees
o The X' schema
o Intransitive elementary trees
Deriving simple sentences
Deriving complex sentences
The adjunct relation
o Modification is different
o The need for an adjunction operation
o A typology of syntactic dependents
o More on the distinction between complements and adjuncts
Notes
Exercises and problems
Supplementary material
o Modals and auxiliary verbs in English
o Reference and related notions
o Thematic roles

As was mentioned in Chapter 1, we can represent the individual vocabulary items of a language
as small pieces of syntactic structure, or elementary trees. The idea is to generate phrases and
sentences by composing (and possibly otherwise manipulating) these elementary trees in well-
defined ways.

In this view, vocabulary items are comparable to the atoms of physical matter. Atoms do not
combine into molecules just because they happen to be next to each other; rather, their
combinatorial possibilities are governed by their internal structure (for instance, the number of
electrons on an atom's outermost shell and the relative number of protons and electrons).

Accordingly, in the first part of this chapter, we consider the internal structure of elementary
trees. As in the last chapter, we begin by focusing on how verbs combine with their arguments to
form larger phrases. For the time being, we will treat noun phrases and prepositional phrases as
unanalyzed units, postponing discussion of their internal structure until Chapter 5. We then
generalize the approach developed for verbs and their arguments to the point where we can build
simple sentences as well as complex sentences containing subordinate clauses. In order to derive
sentences, we will find it necessary to introduce a formal operation called movement, which
allows us to represent the fact that constituents can have more than one function in a sentence.

In the second part of the chapter, we turn to the representation of modification. As we will show,
it is not possible to combine modifiers with elementary trees by the substitution operation
introduced in Chapter 1. We therefore introduce a third and final formal operation (besides
substitution and movement) called adjunction. As we will see, it is not only modifiers that are
integrated into syntactic structure by adjunction, but certain semantic arguments as well.
The X' schema for elementary trees

Transitive elementary trees

We begin our investigation of the internal structure of elementary trees by considering how a transitive
verb like ate combines with its two arguments in a sentence like (1).

(1) The children ate the pizza.

From the possibility of pronoun substitution, as in (2), we know that the two arguments are
constituents (specifically, noun phrases).

(2) They ate it.

In principle, the verb could combine with its two noun phrase arguments in either order, or with
both at once. The three possibilities are represented by the schematic structures in (3) (we
address the question of which syntactic category to assign to the nodes labeled by question marks
in a moment).

(3) a. b. c.

However, as we already know from the discussion in Chapter 2, only the representation in (3a) is
consistent with the do so substitution facts in (4).

(4) The children ate the pizza; the children did so.

In other words, transitive verbs combine first with their object. The resulting constituent in turn
combines with the subject.

What is the syntactic category of the constituents that result from these two combinations? In
principle, the result of combining a verb with a noun phrase might be a phrase with either verbal
or nominal properties. But clearly, a phrase like ate the pizza doesn't have the distribution of a
noun phrase. For instance, it can't function as the object of a preposition (even a semantically
bleached one like of) Nor does it pattern like a noun phrase in other respects. For instance, as we
have just seen, the appropriate pro-form for it is not a pronoun, but a form of do so, just as would
be the case if the predicate of the sentence were an intransitive verb. In other words, for the
purposes of do so substitution, the combination of a verb and its object is equivalent to an
intransitive verb (say, intransitive eat); cf. (4) with (5).

(5) The children ate; the children did so.

However, it won't do to simply assign the syntactic category V to the verb-object combination,
on a par with the verb that it contains, since that would leave unexplained the contrast between
(4) and (6) with respect to do so substitution.

(6) The children ate the pizza; *the children did so the pizza.

Notice furthermore that the syntactic category of the verb-object constituent is distinct from the
syntactic category of the constituent that includes the subject. This is evident from the contrast in
(7), which would be unexpected if both constituents belonged to the same syntactic category.

(7) a. We saw the children eat the pizza.

b. * We saw eat the pizza.

In order to represent the facts in (4)-(7), the following notation has been developed. Verbs are
said to project three bar levels, conventionally numbered from zero to two. The lowest bar
level, V0, is a syntactic category for vocabulary items; it is often indicated simply by V without a
superscript. The next bar level is V' (read as 'V-bar'),1 the syntactic category of a transitive verb
and its object. The highest bar level is V" (read as 'V-double-bar'), which is the result of
combining a V' with a subject. For a transitive verb, each bar level corresponds to the number of
arguments with which the verb has combined.

Somewhat confusingly for the novice, the verb's second projection, V", is more often than not
labeled VP. In early work in generative grammar, the label VP was intended as a mnemonic
abbreviation for the verb phrase of traditional grammar and did indeed correspond to that
category. In current phrase structure theory, however, the label that corresponds to the traditional
verb phrase is V', whereas VP includes a verb's subject, which the traditional verb phrase does
not. The idea is that the highest bar level projected by a verb contains all of its arguments. For
clarity, we will avoid using the term 'verb phrase' if possible, but if we do use it, we mean the
traditional verb phrase that excludes the subject (V', not VP). Conversely, when we say VP, we
always mean the projection that contains all of the verb's arguments, not the verb phrase of
traditional grammar.

The fully labeled structure for (1), with the standard labels for the three verbal projections, is
given in (8).
(8)

From (8), we can derive the elementary tree for ate in (9) by 'un-substituting' the two arguments.

(9)

The X' schema

As we show later on in this chapter and in Chapter 5, the basic form of the elementary tree in (9)
can be extended to other syntactic categories. In other words, (9) is an instantiation of a general
phrase structure template, shown in (10) and known as the X' schema (read: X-bar schema) of
phrase structure. X, Y, and Z are variables over syntactic categories.

(10)

A number of standard terms are used in connection with the X' schema. X (= X0) is the lexical
projection of the vocabulary item that it dominates, X' the intermediate projection, and XP (=
X") the maximal projection (sometimes also called phrasal projection). The correspondence
between projections and the bar levels introduced earlier is summarized in (11).

The terms 'intermediate' and 'phrasal' are somewhat misleading, since they both suggest that
the syntactic status of intermediate projections is somehow intermediate between lexical and
phrasal constituents. This is not the case. Intermediate projections are full-fledged phrases, and
'intermediate' simply refers to the position of the projection in the tree structure.

(11) Label Projection Bar level

 X (= X0) Lexical 0

X' Intermediate 1

XP (= X") Maximal (phrasal) 2

The lexical projection X is known as the head of the structure in (10) (the term is sometimes also
used to refer to the vocabulary item dominated by the lexical projection). The three projections
of the head form what we will call the spine of the elementary tree. Following traditional
terminology, the sister of the head - YP in (10) - is called its complement. As we discuss in the
next subsection, elementary trees need not include a complement position.

Note the spelling of complement with e (not i). The idea is that complements complete the
meaning of the head.

The sister of the intermediate projection - ZP in (10) - is called the specifier. Each elementary
tree has at most one specifier, and elementary trees can lack a specifier altogether, as we will see
later on in this chapter. The specifier and complement positions of a head are its (syntactic)
argument positions. In other words, an elementary tree consists of a spine and from zero to two
argument positions.

The terms 'specifier,' 'complement,' and 'argument' can be used to refer to the constituents
that substitute into the positions just described, but they can also refer to the structural
positions themselves. (This is analogous to the way we can use a nontechnical term like 'bowl'
to refer either to the container itself (Hand me the bowl) or to its contents I'd like a bowl (of
soup).) If it is necessary to avoid confusion between the two senses, we can distinguish
between 'specifier position' and 'constituent in specifier position' (and analogously for
'complement' and 'argument').

An important question that arises in connection with the X' schema in (10) is how to represent
predicates with more than two semantic arguments (say, rent or give). The most obvious
approach is to allow elementary trees with more than two complements. Plausible as this
approach may seem, however, it is now widely assumed that syntactic structure is at most
binary-branching (in other words, binary-branchingness is assumed to be a formal universal). If
a predicate has more than two semantic arguments, there are two ways in which the additional
arguments can be integrated into syntactic structure. In some cases (as with rent), the
supernumerary arguments are integrated into syntactic structure by adjunction, an operation
distinct from substitution that we introduce later in this chapter. This case involves a syntax-
semantics mismatch, since a semantic argument ends up occupying a position that is not a
syntactic argument position. In other cases (as with give), the apparently atomic predicate is
decomposed semantically and syntactically into more than one head, thus yielding a total of
more than two argument positions. This second case is discussed in detail in Chapter 7.

Intransitive elementary trees

So far, we have discussed the internal structure of the elementary trees required for transitive
verbs (and transitive categories more generally). In this section, we address the internal structure
of the elementary trees required for intransitive verbs - for instance, intransitive eat. The two
structures in (12) come to mind as possibilities.

(12) a. b.

The trees differ in the presence of an intermediate projection, and (12b) might at first glance
seem preferable because it is simpler (in the sense of containing fewer nodes). Nevertheless, we
will prefer (12a) on the grounds that adopting it results in a simplification of the grammar as a
whole - that is, a simplification not just of the elementary trees themselves, but also of rules and
definitions stated over them. For instance, adopting (12a) allows us to summarize the facts
concerning do so substitution illustrated in (4)-(6) by means of the succinct generalization in
(13).

(13) Do so substitutes for instances of V'.

Given (12b), the corresponding generalization in (14), being a disjunctive statement (a

statement that contains or) is more cumbersome.

(14) Do so substitutes for instances of V' or of V without a complement.

A second, similar reason to prefer (12a) is that it permits the succinct definition of the notion of
specifier in (15a) rather than the disjunctive statement in (15b).

(15) a. Specifiers are sisters of intermediate projections.

b. Specifiers are sisters of intermediate projections or of lexical projections without a

complement.

Notice that we allow verbs that can be used either transitively or intransitively, such as eat, to be
associated with two elementary trees. More generally, we will allow a vocabulary item of any
syntactic category to project one or more elementary trees, as required by its combinatorial
properties.

Deriving simple sentences

We are almost at the point of being able to construct representations of complete sentences, but before
we can, we need to address the syntactic representation of tense. The following discussion relies on the
notion of do support and on the status of modals and auxiliary do as members of the syntactic category
I(nflection); see Modals and auxiliary verbs in English for more details.

In a sentence like (16), the verb waited contains the bound morpheme -ed, which expresses past
tense.

(16) He waited.

If tense morphemes were invariably expressed on the verb in this way, then complete structures
for full sentences could be derived by substituting appropriate structures into the argument
positions of the verb's elementary tree. However, this is not a general solution, because tense is
not always expressed as a bound morpheme on the verb. For instance, in (17), the future tense
counterpart of (16), the future tense is expressed by a free morpheme, the modal will.

(17) He will wait.

Even more strikingly, the past tense in English, though ordinarily expressed as a bound
morpheme on the verb, must be expressed by a free morpheme in do support contexts, as shown
in (18).

(18) a. Emphasis: He did wait.

b. Negation: He didn't wait.

c. Question: Did he wait?

The morphologically variable expression of tense as a free or a bound morpheme raises two
related syntactic questions. First, what is the representation of sentences like (17) and (18a),
where tense is expressed as a free morpheme? (We postpone discussion of negated sentences and
questions until later chapters; see Chapters 6 and 11.) Second, and more generally, how can we
represent all sentences in a syntactically uniform way, regardless of how tense is expressed
morphologically? The reason that we want a syntactically uniform representation is that from a
semantic point of view, both past and future are semantically parallel functions, taking situations
(denoted by VPs) as input and returning as output situations that are located in time, either before
or after the time of speaking.2
We begin by answering the first question in several steps. First, it is clear that (17) and (18a)
share a common predicate-argument structure (predicate used here in the sense of Fregean
predicate). That is, both of these sentences denote a situation in which someone is waiting, with
the sentences differing only as to which point in time the situation holds. We can capture this
commonality by taking the elementary tree for the verb wait in (19a) and substituting an
argument constituent in the specifier position, yielding (19b).

(19) a. b.

Second, in accordance with the general approach to syntactic structure that we have been
developing, modals and auxiliaries, like all vocabulary items, project elementary trees. The
elementary trees for will and auxiliary did are shown in (20).

(20) a. b.

We can substitute the structure in (19b) into each of the elementary trees in (20), yielding (21).3

(21) a. b.

The structures in (21) neatly reflect the semantic relation between tense and situations. The
element in I corresponds to the tense function, the complement of I (VP) corresponds to the
function's input (the situation), and the maximal projection of I (IP) corresponds to the function's
output (the situation located in time). There remains a problem, however: the I element and the
subject of the sentence are in the wrong order in (21). This problem can be solved by introducing
a movement operation that transforms the structures in (21) into those in (22).
(22) a. b.

A few remarks are in order about this operation. Movement is best understood as a convenient
way of representing mismatches between various aspects of a sentence or its constituents.
Specifically, in the case at hand, he satisfies two distinct functions. First, it is a semantic
argument of the verb wait. Second, it is the subject of the entire sentence, whose syntactic head is
I, not V. It is important to recognize that these two functions are distinct. This is clearly shown
by the existence of passive sentences. For instance, in the active sentence in (23a), it is the agent
argument that functions as the subject, whereas in its passive counterpart in (23b), it is the theme
argument.

(23) a. Susie drafted the letter.

Agent Theme

Subject Object

b. The letter was drafted (by Susie).

Theme Agent

Subject Prepositional phrase

In order to clearly express a phrase's multiple functions, we do not simply move the phrase from
one position to another. Instead, movement leaves a trace in the phrase's original position, and
the two positions share an index. In the syntactic literature, indices for movement are represented
by the same alphabetical subscripts as referential indices. For clarity, we diverge from this
practice and use the natural numbers as referential indices, and the lowercase letters i, j, k, and so
on, as movement indices. A constituent and its traces of movement are called a chain. The
elements of a chain are its links. Higher links in a chain are often referred to as the antecedents
of lower ones. Finally, the highest and lowest links in a chain are sometimes referred to as the
chain's head and tail, respectively.

Don't confuse this sense of the term 'head' with the sense introduced earlier in connection with
X' structures. The head of an X' structure is the structure's lexical projection (or sometimes the
vocabulary item dominated by it). The head of a movement chain is the highest constituent in a
 chain; the constituent's X' status is irrelevant). Which sense is meant is generally clear from the
context.

Is the head of the movement chains in (22) a head in the X' sense? No. The reason is that
it is possible to replace it by what is clearly a phrase (say, by the student in the red
sweater). But the head of a chain can be a head in the X' sense, as we will see in Chapter
6 in connection with verb movement.

Finally, we should point out the existence of a special type of chain - what
mathematicians would call a degenerate case. It is perfectly possible for a chain to consist
of a single constituent. This is the case when a constituent hasn't moved. The chain then
contains a single link, which is simultaneously the chain's head and its tail.

We are now in a position to answer the second question posed earlier - namely, how can
sentences be represented in a syntactically uniform way regardless of the morphological
expression of tense? A simple answer to this question is possible if we assume that English has
tense elements that are structurally analogous to auxiliary do, but not pronounced, as shown in
(24); we will use square brackets as a convention to indicate such silent elements.

(24) a. b.

Elementary trees as in (24) make it possible to derive structures for sentences in which tense is
expressed as a bound morpheme on the verb along the same lines as for sentences containing a
modal or auxiliary do. In (25), we illustrate the derivation of He waited.

(25) a. b. c. d.

Substitute predicate-
Select
Substitute argument structure in Move subject from
elementary
argument (25b) into elementary Spec(VP) to Spec(IP)
tree for verb
tree for tense
Deriving complex sentences

This section is devoted to the derivation of sentences that contain complement clauses (also known as
clausal complements). Some examples are given in (26); the complement clauses are in italics.

(26) a. We will ask if she left.

b. They believe that he came.

Although sentences with complement clauses can become unboundedly long (recall the instances
of recursion in Chapter 1), deriving structures for them proceeds straightforwardly along the
lines already laid out. If and that are both complementizers, so called because they have the
effect of turning independent sentences into the complements of a matrix verb, and they project
the elementary trees in (27).

(27) a. b.

Given elementary trees like (27), we can derive the italicized complement clause in (26a) as in
(28).

(28) a. b.

Elementary tree for

verb of Substitute argument
complement clause
 c. d. e.

Substitute (28b) in
Move subject in Substitute (28d) in elementary tree for
elementary tree for
complement clause complementizer (27a)
tense (24b)

The structure in (28e) in turn allows us to derive the entire matrix clause, as in (29).

(29) a. b.

Substitute arguments, including clausal

Elementary tree for matrix clause verb
complement (28e)
 c. d.

Substitute (29b) in elementary tree for

Move subject in matrix clause
modal (20a)

Given representations like (29d), we can now formally characterize recursive structures as in
(30).

(30) a. A structure is recursive iff it contains at least one recursive node.

b. A node is recursive iff it dominates a node distinct from it, but with the same label.

The recursive nodes in (29d) are the higher IP, I', VP, and V' nodes (and no others). Note that a
recursive node need not be the root node of a tree, and that it can be any projection level (XP,
X', or X).

The lower IP, I', VP, and V' nodes are not recursive nodes, since they don't dominate
another instance of the same category.

For a node to be recursive, it is not enough that the tree contains a second instance of the
category somewhere. The first node has to dominate (though not necessarily immediately
dominate) the second one. For instance, none of the NounPhr nodes in (29d) is recursive.

The adjunct relation

Modification is different
The elementary trees introduced so far allow us to represent two of the three basic linguistic relations
discussed in Chapter 3: namely, argumenthood and predication. As we have seen, semantic arguments
of a verb can be expressed as syntactic arguments substituting into one of the two argument positions in
the verb's elementary tree: either the complement position or the specifier position. VPs and IPs can be
treated as arguments (specifically, as complements) of I and C, respectively. And finally, although
predication is not reducible to argumenthood (recall from Chapter 3 that expletive subjects are required
independently of a verb's semantic requirements), subjects occupy specifier positions regardless of
whether they are semantic arguments or not. In other words, predication does not require a structural
relationship uniquely associated with it. An important remaining question is how to represent the
modification relation using the X' schema developed so far.

In principle, modification might resemble predication in not requiring a structural relation of its
own. As it turns out, however, neither of the two head-argument relations (head-specifier, head-
complement) adequately represents the relation between a head and its modifier. As we have
seen, when a verb combines with a complement, the category of the resulting constituent (V') is
distinct from that of the verb (V) (recall the contrast between (4) and (6)), and when the verb and
the complement in turn combine with the specifier, the category of the resulting constituent (VP)
is distinct yet again (recall the contrast in (7)). By contrast, modifying a verb-complement
combination like ate the pizza in (31) does not change the syntactic category of the resulting
constituent, which remains V' (the modifier is in italics).

(31) a. The children ate the pizza.

b. The children ate the pizza with gusto.

This is evident from the do so substitution facts in (32), where either the unmodified or the
modified verb-complement combination can be replaced by a form of do so.

(32) a. The children ate the pizza with gusto; the children did so with gusto.

b. The children ate the pizza with gusto; the children did so.

The same pattern holds for intransitive verbs that combine with a modifier.

(33) a. The children ate with gusto; the children did so with gusto.

b. The children ate with gusto; the children did so.

The do so substitution facts in (32) and (33) motivate the syntactic structure for (31b) that is
given in (34) (for clarity, we focus on the internal structure of the VP, omitting the projection of
the silent past tense element and subject movement).
(34)

The structural relation of the modifier with gusto to the spine of the V projection is known as the
adjunct relation, and the modifier itself is said to be an adjunct. Modifiers are always
represented as adjuncts in syntactic structure. As a result, 'modifier' and 'adjunct' tend to be used
somewhat interchangeably. In this book, however, we will distinguish between the two terms as
follows. We will use 'modifier' when we want to highlight a phrase's semantic function of
qualifying or restricting the constituent being modified. For instance, as we mentioned in
Chapter 3, a verb like laugh denotes the set of entities that laugh. Combining the verb with a
modifier like uproariously yields the expression laugh uproariously, which denotes a subset of
the set denoted by laugh. We will use the term 'adjunct' when focusing on a constituent's
structural position in a tree. As we will see later on in this chapter, it is possible for semantic
arguments to be represented as syntactic adjuncts. This does not change the semantic argument
into a modifier, however!

The need for an adjunction operation

The structure in (34) raises the question of what elementary tree for transitive ate is involved in its
derivation. 'Un-substituting' both arguments and the modifier, as we did earlier for trees containing only
arguments, yields the structure in (35).

(35)

Is the structure in (35) a satisfactory elementary tree? Clearly, allowing it means that our
grammar now contains two elementary trees for transitive ate. At first glance, this doesn't seem
like a serious problem, since we already allow the two elementary trees for transitive and
intransitive ate in (36).

(36) a. b.
But (35) differs in one important respect from the structures in (36): it is a recursive structure.
This has an extremely undesirable consequence: namely, that if we were to derive structures like
(34) by means of elementary trees like those in (35), there would be no principled way to avoid
an unbounded number of such elementary trees. For instance, the derivations of the sentences in
(37), with their increasing number of modifiers, would each require a distinct elementary tree for
drink, and each additional modifier would require an additional elementary trees.

(37) a. We would drink lemonade.

b. We would drink lemonade in summer.

c. We would drink lemonade in summer on the porch.

d. We would drink lemonade in summer on the porch with friends.

But the whole point of a generative grammar is to generate an unbounded set of sentences from a
finite set of elementary expressions and operations. Given this aim, we are compelled to require
elementary trees to be non-recursive structures. This has the consequence that adjuncts cannot be
integrated into larger syntactic structures by substitution, and accordingly, we introduce a
further tree operation called adjunction. For clarity, the operation of interest to us is sometimes
called Chomsky-adjunction, to distinguish it from Joshi-adjunction, a different formal operation
that plays a central role in Tree-Adjoining Grammar (Joshi, Levy, and Takahashi 1975).

For the moment, we will use the adjunction operation to integrate modifiers into syntactic
structures. As we will see in Chapter 6, the adjunction operation is also used for other purposes.
Whatever its linguistic purpose, however, it is always the same formal (= graph-theoretical)
operation: namely, a two-step process that targets a particular node. When the purpose of
adjunction is to integrate a modifier, as it is here, the target of adjunction is an intermediate
projection, indicated in red in (38a). The first step in carrying out adjunction is to make a clone
of the target of adjunction that immediately dominates the original node, as in (38b). The second
step is to attach the tree for the modifier as a daughter of this higher clone, as in (38c).

(38) a. b. c.

Select target of Clone target of Attach modifier as daughter of

adjunction adjunction clone

Deriving the rest of the structure for the entire sentence proceeds as outlined earlier, as shown in
(39).
(39 a b c
) . . .

Substitute (39a) in elementary

Substitute arguments Move subject
tree for tense

For expository clarity, we have chosen to illustrate the derivation of the sentence with adjunction
preceding substitution and movement. However, the order of adjunction with respect to the other
operations is irrelevant.

In concluding this section, we raise a general point concerning intermediate projections and
adjunction to them. Given that words (or syntactic atoms of some sort) combine with one another
to form phrases, any theory of syntax must assume heads and projections. But distinguishing
between two types of phrases (intermediate projections vs. maximal projections) seems
uneconomical, and attempts have therefore been made to eliminate the need for intermediate
projections, along with the possibility of adjunction to them. For instance, given our current
assumptions, sentences like (40) force us to allow adjunction to intermediate projections.

(40) a. [IP They [I' never [I' will agree to that. ] ] ]

b. God let [VP there [V' suddenly [V' be light. ] ] ]

However, if the IP and the small clause VP in such sentences were 'split up' into two separate
projections, it would be possible to eliminate the intermediate projections and to adjoin the
modifiers to maximal projections instead. This is illustrated in (41), where IP has been split into
Agr(eement)P and T(ense)P, and the small clause VP has been split into Pred(ication)P and a
lower VP. We hope to add a chapter one day that discusses such structures.

(41) a. [AgrP They [TP never [TP will agree to that. ] ] ]

b. God let [PredP there [VP suddenly [VP be light. ] ] ]

A related question is whether, even in a system with intermediate projections such as we

continue to assume for the moment, adjunction can indiscriminately target either intermediate or
maximal projections. For simplicity, since we know from (40) that adjunction must be able to
target intermediate projections under our current assumptions, we will consistently represent it as
doing so.4

A typology of syntactic dependents

Each of the three types of syntactic dependents that we have been discussing - complements, specifiers,
and adjuncts - stands in a unique structural relation to the head and to the spine of the head's
projection. Complements and adjuncts are both daughters of intermediate projections, but they differ in
that complements are sisters of heads, whereas adjuncts are sisters of the next higher projection level.
As sisters of intermediate projections, adjuncts resemble specifiers. But again, the two relations are
distinct because adjuncts are daughters of intermediate projections, whereas specifiers are daughters of
maximal projections. These structural relations and distinctions are summarized in (42). We also include
the formal operations that fill or create the positions in question.5

(42) Relation to head Sister of ... Daughter of ... Formal operation

Complement Head Intermediate projection Substitution

Adjunct Intermediate projection Intermediate projection Adjunction

Specifier Intermediate projection Maximal projection Substitution

More on the distinction between complements and adjuncts

Given the table in (42), it is easy to tell whether a constituent is represented in a particular tree
structure as a complement or as an adjunct. However, it is not always self-evident whether a phrase is a
complement or an adjunct as a matter of linguistic fact.

Remember that tree structures are models of linguistic facts. Just because it is possible to build a
tree that represents a certain phrase as a complement of a certain head doesn't mean that the
phrase actually is a complement. In other words, trees can "lie".

The most reliable way to determine the relation of a particular phrase to a verb is to use do so
substitution. If a phrase need not be included as part of the sequence being replaced by do so,
then it is an adjunct. If it must be included, then it is a complement. Using this test, we find that
phrases specifying cause or rationale, time, location, or manner are generally adjuncts, even if
they are noun phrases. Some examples, including the results of do so substitution, are given in
(43); the adjuncts are in italics.
(43) a. Rationale They waited for no good reason, but we did so for a very good one.

b. Duration They waited (for) a day, but we did so (for) a month.

c. Location They waited in the parking lot, but we did so across the street.

d. Manner They waited patiently, but we did so impatiently.

In the examples we have seen in this book so far, semantic arguments are expressed in a
syntactic tree as syntactic arguments (or not at all), and modifiers are expressed as adjuncts. It is
possible, however, for semantic arguments to be expressed in the syntax as adjuncts (this is the
mismatch case mentioned earlier in connection with binary-branchingness). For example, as we
mentioned in Chapter 3, rent, from a semantic point of view, is a five-place predicate, with
arguments denoting landlord, tenant, rental property, amount of money, and length of time (lease
term). Some of these semantic arguments are expressed as syntactic arguments. For instance, in
(44), the phrase denoting the rental property is a complement, as is evident from the results of do
so substitution.

(44) a. Dennis rented the apartment to Lois.

b. * ... and David did so the studio to Rob.

On the other hand, do so substitution shows that the phrase denoting the lease term is an adjunct,
even though lease terms are semantic arguments of rent on a par with rental properties.

(45) a. Dennis rented Lois the apartment for two months.

b. ok ... and David did so for a whole year.

A final word should be said about the correlation between a syntactic dependent's obligatory or
optional character and its status as a complement or adjunct. It is tempting to assume the
biconditional relationship indicated in the penultimate column in (46).6

Wishful Actual
thinking situation

(46) a. If a syntactic dependent is obligatory, then it is a

TRUE TRUE
complement.

b. If a syntactic dependent is a complement, then it is

TRUE FALSE
obligatory.
But as last column indicates, the biconditional relationship doesn't hold. It is true that obligatory
syntactic dependents are complements. For instance, the contrast in (47) is evidence that the
noun phrase following devour is a complement, a conclusion that is borne out by do so
substitution in (48).

(47) Every time I see him, ...

a. * ... he's devouring.

b. ... he's devouring a six-inch steak.

(48) a. He devoured a hamburger and french fries, and I did so, too.

b. * He devoured a hamburger and french fries, and I did so six samosas.

But not all complements are obligatory. The grammaticality of (49a) shows that the phrase
French fries is optional, but the ungrammaticality of (49c) shows that it is nevertheless a
complement.

(49) a. He ate, and I did so, too.

b. He ate French fries, and I did so, too.

c. * He ate French fries, and I did so three samosas.

Although (46b) is false, (46a) does have the consequence in (50) (derived by the modus tollens
rule of propositional logic).

(50) If a syntactic dependent is not a complement, it is not obligatory.

The two valid generalizations in (46a) and (50) can be summarized succinctly as in (51).

(51) a. Obligatory syntactic dependents are complements.

b. Adjuncts are optional.

Notes

1. Why is V' read as V-bar when it contains not a bar, but a prime symbol? The reason is that when the
idea of bar levels was introduced in the 1970s, the various levels were distinguished by horizontal bars
over a syntactic category. The lowest level had no bars, the first level one, and the second two. But back
in the days of typewriters, such overbars were cumbersome to type (you typed the symbol, --* rolled up
the platen a bit, backspaced, typed an overbar *--, repeated from --* to *-- for each overbar, and then
rolled the platen down again the right amount). Overbars are also expensive to typeset, and even today,
they aren't part of the standard character sets for HTML documents such as this one. Therefore, it was
and continues to be convenient to substitute prime symbols for overbars. However, linguists have failed
to update their terminology (terminological inertia), and so the old term 'bar' is still with us.

2. The semantics of tense we are assuming here is oversimplified, but sufficient for our purposes.

3. The representations in (21) look like appropriate representations for the questions Will he
wait? and Did he wait? But notice that they can't be, since they contain unfilled substitution
nodes. Moreover, as we will see in a later chapter, there is reason to postulate a projection above
IP in the representations of questions.

4. Sentences like (i) appear to require adjunction to maximal projections (IP in the present case).

(i) Tomorrow, we will eat pizza.

It has been argued, however, that the apparent adjunct tomorrow occupies the specifier position
of the projection of a head higher than I. We return to this issue in Chapter 14. Analogous
reasoning would extend to examples like (ii).

(ii) Tomorrow, what will we eat?

5. Alert readers will notice that the relations in (42) are defined redundantly. Such readers are
likely to enjoy Problem 4.2.

6. A similarly tempting (but false) biconditional relationship was discussed in Chapter 2.

Exercises and problems

Exercise 4.1

What is the X' status of Fregean and of Aristotelian predicates? You should be able to answer in one or
two brief sentences.

Exercise 4.2

The trees in (1) fail to correctly account for certain grammaticality judgments. What are the judgments?
(1) a. b.

Exercise 4.3

A. Are the italicized phrases in (1) syntactic arguments or adjuncts? Explain. Your discussion
needn't be extensive, but you must include the syntactic evidence (do so substitution facts) on
which you base your conclusions.

(1) a. They waited for us.

b. This program costs twenty dollars.

c. We drove to Denver.

d. We worded the letter carefully.

e. They are behaving very inconsiderately.

f. This volcano might erupt any minute.

B. Using the grammar tool in x-bar 1, build structures for the sentences in (1). Needless to say,
the structures you build should be consistent with the evidence you gave in Part A.

Exercise 4.4

A. Using the grammar tool in x-bar 1, build structures for the sentences in (1). Motivate your
attachment of the lowest argument or adjunct in each sentence (in other words, give the
judgments that lead you to attach the relevant phrase the way you do).

(1) a. They demolished the house.

b. Mona Lisa called the other neighbor.

c. Mona Lisa called the other day.

 d. You will recall that her smile amazed everyone.

e. Most people doubt that Mona Lisa lives in Kansas.

f. My friend wondered if Mona Lisa would come to his party.

B. Indicate all recursive nodes in the structures that you build for (1). You can do this by using
the grammar tool's highlighting feature (see the "Instructions" menu).

Exercise 4.5

A. (1) is structurally ambiguous. Paraphrase the two relevant interpretations. (Don't be concerned
with the referential vagueness of they.)

(1) They claimed that they paid on the 15th.

B. Using the grammar tool in x-bar 1, build a structure for each of the interpretations, indicating
which structure goes with which interpretation.

C. Indicate all recursive nodes in the structures that you build for (1). You can do this by using
the grammar tool's highlighting feature (see the "Instructions" menu).

Exercise 4.6

A. Make up a sentence with two adjuncts. Provide syntactic evidence that the adjuncts are adjuncts
rather than syntactic arguments. Then build the structure for the sentence using the grammar tool in x-
bar 1. Finally, switch the linear order of the adjuncts, and build the structure for the resulting word
order variant of your original sentence.

B. Make up a simple sentence in which one of the semantic arguments of the verb is expressed in
the syntax as an adjunct. Provide evidence that the adjunct is one. Finally, build the structure for
your sentence using the grammar tool in x-bar 1.

Problem 4.1

A. Is it possible for an adjunct to immediately precede a complement? Take into account both
VO and OV languages. (You may find it helpful to use the grammar tool in variable headedness
1.)

B. Is it possible for two adjuncts to be sisters? Explain.

Problem 4.2
As mentioned in the text, it is attractive to assume a theory of phrase structure that does away with the
notion of intermediate projection. However, a theory of phrase structure with only two bar levels runs
up against a serious conceptual problem. What is the problem, and how can it be resolved?

Hint: Redo (42) using two bar levels.

5 Extending the X' schema

Noun phrases
o Parallels and differences between noun phrases and sentences
o Noun phrases as DPs
o More on determiners
o Modification and related issues
Adjective phrases
Prepositional phrases
Crosslinguistic variation in headedness
Notes
Exercises and problems
Supplementary material
o Nouns

In Chapter 4, we introduced the X' schema of phrase structure, according to which lexical items
project an elementary tree consisting of a spine of projections and up to two argument positions.
In this chapter, we extend the X' schema to syntactic categories other than V, I, or C, including
N(oun) (see Nouns for some basic information), D(eterminer), Adj(ective), and P(reposition).
The final section of the chapter illustrates crosslinguistic variation with regard to the order of
heads and complements.

Noun phrases

Parallels and differences between noun phrases and sentences

The X' schema of phrase structure that we introduced in Chapter 4 is a specific expression of a more
general idea - namely, that lexical items of different syntactic categories show significant cross-
categorial parallels. In the history of generative grammar, this idea was primarily based on the cross-
categorial parallels between noun phrases and sentences (Chomsky 1970). In what follows, we review
these parallels, as well as some differences between the two categories.

Argument structure. Early in the history of generative grammar (Lees 1960), it was observed
that sentences like (1a) and noun phrases like (1b) share several important properties.

(1) a. The army destroyed the city.

 b. the army's destruction of the city

The semantically central element of the sentence in (1a) is the verb destroyed, and its semantic
arguments, the agent the army and the theme the city, are both expressed as syntactic arguments
of the sentence. In a parallel way, the semantically central element in the noun phrase in (1b) is
the nominal counterpart of destroy, the noun destruction. Like the verb, the noun is associated
with an agent argument and a theme argument that are both overtly expressed - in this case, as
the possessive expression the army's and the prepositional phrase of the city.

The correspondence in (1) is supported by that between the passive sentence in (2a) and its
passive-like noun phrase counterpart in (2b).

(2) a. The city was destroyed (by the army).

b. the city's destruction (by the army)

In both of these examples, the argument preceding the head is now the theme the city('s), and the
agent argument is expressed by an optional by phrase.

Modification. A further parallel between sentences and noun phrases is that in both categories,
the semantically central element - the verb or the noun - can be modified in similar ways, as
illustrated in (3) and (4).

(3) a. Prepositional phrase She gives money to the organization on a regular basis.

b. Adverb She regularly gives money to the organization.

(4) a. Prepositional phrase her gifts of money to the organization on a regular basis

b. Adjective her regular gifts of money to the organization

Some cross-categorial differences. Sentences and noun phrases also exhibit certain differences.
First, arguments and modifiers are not always expressed in exactly the same way across the two
categories. For instance, the agent argument is expressed as an ordinary noun phrase in a
sentence like (1a), but as a possessive noun phrase in a noun phrase like (1b). In a sentence, the
theme argument is expressed as a noun phrase, but in a noun phrase, it must be part of a
prepositional phrase, usually an of phrase. Finally, although verbs and nouns can both be
modified by prepositional phrases, verbs are modified by adverbs, whereas nouns are modified
by adjectives. In connection with this last difference, notice that adverbs can precede or follow
the verb they modify, whereas adjectives are ordinarily restricted to prenominal position.

(5) a. Adverb The kids regularly donate their old toys.

 b. The kids donate their old toys regularly.

(6) a. Adjective the kids' regular donation of their old toys

b. * the kids' donation of their old toys regular

A further and even more fundamental difference between sentences and noun phrases concerns
the subject requirement. As we saw in Chapter 3, all sentences require a syntactic subject, even
when it does not correspond to a semantic argument, as is evident from the contrast between (7)
and (8).

(7) a. It appears that the manuscript has been found.

b. There exists a solution.

(8) a. * Appears that the manuscript has been found.

b. * Exists a solution.

By contrast, noun phrases never require a subject. For instance, the agent argument of a noun can
be expressed, but it needn't be, as shown in (9).1

(9) a. the committee's criticism of the proposal

b. the criticism of the proposal

What is even more striking is that sentences with expletive subjects have no noun phrase
counterparts. As (10) shows, the very expletive expressions that are obligatory in (7) are
ungrammatical in noun phrases.2

(10) a. * it(s) appearance that the manuscript has been found

b. * there('s) existence of a solution

In summary, noun phrases resemble sentences in that their core categories - nouns and verbs,
respectively - have semantic arguments that can be expressed as syntactic arguments in partly
similar ways. Nouns and verbs can also be modified in largely similar fashion. We do not deny
that the two categories differ fundamentally with respect to the subject requirement. However, in
the remainder of this part of the chapter, our focus will be on how to represent the parallel
aspects of noun phrases and sentences.
Noun phrases as DPs

A striking fact about nouns is that they cannot in general function as arguments on their own, but
must be accompanied by a determiner.

(11) a. * Assignment is not difficult.

b. * You should hand in assignment.

(12) a. { The, this, that } assignment is not difficult.

b. You should hand in { the, this, that } assignment.

We conclude from this that noun phrases are the result of composing two projections, one headed
by the noun and the other by the determiner, as shown in (13).

(13) a. b. c.

Given the structure in (13c), the traditional term 'noun phrase' is a misnomer since noun phrases
are maximal projections of D rather than of N. Because the term 'noun phrase' is firmly
established in usage, we continue to use it as an informal synonym for 'DP'. However, in order to
avoid confusion, we will use the term 'NP' only to refer to the subconstituent of a noun phrase
that is the complement of a determiner. We will never use it to refer to an entire noun phrase
(that is, a DP).

In the simplest case, the elementary tree for a noun consists of just a spine, as in (13b). But like
verbs, nouns can have both complements and specifiers. For instance, depending on which of the
noun phrases in (14) it appears in, criticism is associated with one of the elementary trees in (15).

(14) a. They refuted the criticism.

b. They refuted the criticism of the proposal.

c. They refuted the committee's criticism.

d. They refuted the committee's criticism of the proposal.

(15) a. b. c. d.

In (14a), the phrase the criticism is derived in exactly the same way as the assignment in (13) -
by substituting the NP in (15a) as the complement of the determiner. In (15b), the noun phrase
containing criticism is derived as in (16). For simplicity, we disregard the internal structure of
the PP for the moment.

(16) a. b. c.

Elementary tree for N Substitute theme Substitute (16b) in elementary

(15b) argument tree for D

Noun phrases containing possessive noun phrases, like (14c), require a possessive head 's, which
contains two argument positions (rather than just one like an article). The possessive head is
shown in (17a). The derivation of the entire noun phrase proceeds as follows. First, the agent
argument substitutes into the specifier position of the elementary tree for the noun. The resulting
NP then substitutes into the complement position of the possessive head. Finally, the argument in
Spec(NP) moves to Spec(DP) in a manner analogous to subject movement in sentences.3

(17) a. b. c.

Elementary tree for Substitute agent

Elementary tree for N (15c)
possessive 's argument
 d. e.

Move specifier from Spec(NP) to

Substitute (17c) in (17a)
Spec(DP)

Finally, deriving (14d) involves substituting both the agent and theme arguments in the
elementary tree in (15d). The remainder of the derivation is identical to that of (14c), as shown in
(18).

(18 b c
a.
) . .

Elementary tree for

Elementary tree for N (15d) Substitute arguments
possessive 's (17a)

d
e.
.

Move specifier from

Substitute (18c) in (18a)
Spec(NP) to Spec(DP)
In (19), we repeat the tree for the noun phrase in (18e) side by side with the tree for the
corresponding sentence. In order to underline the topological parallel between the two trees, we
omit the internal structure of the theme DP in (19b). As is evident, apart from the labels for the
syntactic categories, the two-layered structure for noun phrases (NP, DP) presented here is
analogous to the two-layered structure for simple sentences from Chapter 4 (VP, IP).

(19) a. b.

In view of the structural parallel in (19), it is convenient to generalize the notion of subject to
include both Spec(IP) and Spec(DP). Accordingly, we will use the term 'subject movement' to
subsume both movement from Spec(VP) to Spec(IP) and movement from Spec(NP) to
Spec(DP).

The structural parallel between the two trees in (19) is further supported by the following
semantic parallel. In a formal semantics that is simple but sufficient for our purposes, an NP
constituent denotes a set of individuals. For instance, the NP dominating woman denotes the set
of all women, and the NP dominating president of the United States denotes the set of all
presidents of the United States - past, present, and future. Combining an NP with a determiner
like this or those has the syntactic effect of yielding a DP and the semantic effect of picking out a
particular individual (or individuals, in the case of a plural noun) from the set denoted by the NP.
Which particular individuals are actually picked out depends not just on the meaning of the NP
and the determiner, but also on the particular discourse context in which the DP is used. This is
what allows a noun phrase like the cat to refer to different cats in different discourse contexts. In
a similar way, we can think of VPs as denoting situations. For instance, a VP like these cats jump
onto the dresser denotes the set of all situations in which the individuals denoted by these cats
jump onto the piece of furniture denoted by the dresser. Combining a VP with a tense morpheme
in I then picks out one of these situations (at least under some theories of tense). For instance, the
tensed IP These cats jumped onto the dresser picks out one of the situations that occurred before
the time of speaking (how the past tense morpheme combines with the verb to yield jumped is
discussed in detail in Chapter 6). Once again, the particular situation picked out depends in part
on the discourse context, so that the same sentence can be used to refer to more than one
situation.

In concluding this section, we draw attention to the fact that in the noun phrases that we have
considered so far, any constituents in Spec(DP) have expressed arguments of the noun and have
hence undergone subject movement. However, subjects of noun phrases don't necessarily
originate in the NP projection. In a noun phrase like the student's program, for instance, it makes
more sense to treat the student as an argument of the possessive morpheme rather than of the
noun program. Accordingly, we give the noun phrase the structure in (20), where N doesn't
project a specifier and the subject of the noun phrase is substituted directly into Spec(DP). (For
simplicity, we omit the internal structure of the specifier.)

(20)

More on determiners

Subcategories of determiners. Like verbs and nouns, determiners have different degrees of
transitivity. For instance, the definite article the and the indefinite article a(n) are obligatorily
transitive, whereas the demonstratives this and that are optionally so.

(21) a. I'll buy { the, a } book.

b. * I'll buy { the, a. }

(22) a. I'll buy { this, that } book.

b. I'll buy { this, that. }

Certain ordinary pronouns pattern just like demonstratives, as shown in (23), and so we will treat
them, too, as optionally transitive determiners.

(23) a. we Americans, you fool(s)

b. we, you

Finally, ordinary pronouns can also behave like obligatorily intransitive determiners, as shown in
(24).

(24) a. I, he, she, it, they

b. * I idiot, he fool, she linguist, it piece of junk, they traitors

In this connection, recall the warning in Chapter 1 that the term 'pronoun' is potentially
misleading. It suggests that pronouns are a subclass of nouns. If that were so, then pronouns
should combine with articles and demonstratives in the same way that other nouns do. In fact,
however, pronouns behave exactly like complete noun phrases in this regard, as shown in (25).
The facts in (25) thus provide strong evidence for the analysis of pronouns as determiners just
presented.

(25) a. Noun the people, this woman, that addressee

b. Noun phrase (= DP) * the these people, this the woman, that the addressee

c. Pronoun (= pro-DP) * the they, this she, that you

Elementary trees for the various types of determiners that we have just discussed are given in
(26).

(26) a. b. c. d. e. f.

Silent determiners. As shown in (27), plural indefinite count nouns and indefinite mass nouns
are apparently not accompanied by an article, in contrast to their singular or definite
counterparts.

(27) a. ___ cars, ___ apples; ___ rice

b. a car, an apple; the rice

However, we assume for conceptual reasons that the examples in (27a) contain a silent article
that is semantically roughly comparable to the unstressed some in I would like some apples and
some rice. We assume that the silent article has a singular and a plural form, as shown in (28).
The singular form combines with mass nouns, and the plural form with plural count nouns.

(28) a. b.

We have two reasons for assuming the existence of silent determiners. First, this assumption
allows us to minimize the difference between English and a language like Spanish, where the
indefinite article has singular and plural forms that are both overt. The resulting correspondence
between English and Spanish determiners is shown in (29); the plural indefinite articles are in
boldface. For simplicity, we give only the masculine forms of the Spanish determiners.

(29) English Spanish

Sg Pl Sg Pl

Demonstrative this these este estos

that those ese esos

Definite article the the el los

Indefinite article a(n) [indef pl] un unos

Second, assuming the silent determiner allows us to maintain that all noun phrases are DPs.
Sentences like (30) can then all be derived using the single elementary tree for brought in (31).

(30) a. Some butlers brought some tea.

b. Some butlers brought tea.

c. Butlers brought some tea.

d. Butlers brought tea.

(31)

We show the complete structure for the apparently articleless butlers and tea in (32), but in (33)
we simplify the structure for the noun phrases in order to make it easier to focus on the structural
similarity across all four sentences.
(32) a. b.

(33) a. b.

c. d.

In principle, we could take an alternative tack. If it were our goal to assign the least possible
amount of structure (that is, the structures with the fewest nodes) to each sentence in (30), we
would reject the silent determiner in (28) and we would represent butlers and tea using the trees
in (34) rather than those in (32).

(34) a. b.

The alternative structures under discussion for (30d) are given in (35). (35a) is simply (33d), but
with the internal structure of both DPs fully shown.
(35) a. b.

Clearly, the tree in (35b) is simpler than its counterpart in (35a) in the sense of containing fewer
nodes. However, this simplicity comes at the price of a veritable explosion in the number of
elementary trees in the grammar, since every argument position that can be filled by a noun
phrase would need to be associated with two elementary trees (one with a DP substitution node,
and one with an NP substitution node). For instance, instead of the single elementary tree for
brought in (36a), we would need the three additional trees in (36b-d).

(36) a. b. c. d.

More generally, obligatorily intransitive verbs would require two elementary trees rather than
one, obligatorily transitive verbs - the case just illustrated - would require four (2 x 2) rather than
one, and optionally transitive verbs would require six (4 + 2) rather than two (1 + 1). This result
seems unappealing on computational grounds. Moreover, the whole idea of simplifying the
representations of individual sentences is inconsistent with the Chomskyan paradigm of
language. Why? From a Chomskyan perspective, what syntactic theory attempts to model and
understand is grammar in the sense of the mental capacity to generate sentences, not the set of
phrases and sentences that is the output of the grammar. A reasonable working hypothesis is that
the best model for this capacity is the simplest possible grammar. From a Chomskyan
perspective, striving to simplify the representations of sentences at the expense of complicating
the grammar itself is missing the whole point of constructing grammars in the first place!

Modification and related issues

N' as target of adjunction. As we noted in our introductory review of the parallels between
noun phrases and sentences, nouns and verbs can be modified in similar ways. In (37), for
instance, the same prepositional phrase in the hospital modifies the noun stay and the
morphologically related verb stayed.
(37) a. Mike's stay in the hospital

b. Mike stayed in the hospital.

Extending the approach to representing modification introduced in Chapter 4, we can derive the
structure for the noun phrase in (37a) as in (38). (For simplicity, we omit the internal structure of
the proper noun in the specifier; for details, click here).

(38) a. b. c.

Elementary tree Substitute (38b) in elementary tree for

Substitute argument
for N possessive 's

d. e. f.

Select N' as target of Adjoin PP at target of adjunction in

Move subject
adjunction (38e)

Apart from the category labels, the resulting structure in (38f), repeated for convenience as (39a),
is analogous to the structure for the corresponding sentence in (39b).
(39) a. b.

Leftward adjunction. So far, we have discussed modifiers that follow the head, whose
representation involves rightward adjunction. Structures for examples like (40), where the
modifier precedes the head it modifies, can be derived by leftward adjunction, with the results in
(41).

(40) a. Kelly's nervous grimace

b. Kelly nervously grimaced.

(41) a. b.

One substitution. As discussed in Chapter 4, do so substitution allows us to distinguish between

complements and adjuncts in the verbal system. A similar diagnostic is available in the nominal
system - one substitution, which is illustrated in (42).

(42) a. this book on the floor and that one

b. this book on the floor and that one on the table

In the most natural interpretation of (42a), one is interpreted as book on the floor. In (42b), on the
other hand, one is interpreted as simply book. We can represent these facts by assuming that the
first conjunct in both cases has the structure in (43).
(43)

According to (43), the noun book has no complement, and the PP on the floor is an adjunct. The
pro-form one substitutes for instances of N', just as do so substitutes for instances of V'. One
substitutes for the higher N' in (42a), and for the lower N' in (42b).

As in the case of V', adjunction to N' can apply more than once, yielding multiply recursive
structures like (44).

(44)

Restrictions on one substitution. A cautionary note is in order about one substitution. Although
in principle one can substitute for all instances of N', it is subject to two restrictions, which are
important to keep in mind when using one substitution as a diagnostic for syntactic structure. The
first, which makes some sense given its meaning, is that one can substitute only for count nouns,
as illustrated in (45).

(45) a. I have swum in this { ocean, pool, river, } and you have swum in that one.

b. * I have swum in this water, and you have swum in that one.

A second and more mysterious restriction is that one cannot immediately follow the indefinite
article, a cardinal number, a possessive noun phrase, or, for many speakers, the plural
demonstratives these and those. Whatever the exact source of this restriction is, it is very
superficial, since an intervening word renders the ungrammatical (a) examples in (46)-(49)
grammatical.4

(46) a. Indefinite article * I bought a book, and you bought a one, too.
 b. I bought a blue book, and you bought a red one.

(47) a. Cardinal number * I bought { two , ten } books, and you bought { two, ten } ones, too.

b. I bought { two, ten } blue books, and you bought { two, ten } red ones.

(48) a. Possessive * I like { Mary's, her } book, and you like { John's, his } one.

b. I like { Mary's, her } blue shirt, and you like { Mary's, her } red one.

(49) a. Plural demonstrative * I like these books, and you like those ones.

b. I like these blue books, and you like those red ones.

Structural ambiguity. Having introduced N' as a possible target of modification, we are now in
a position to associate structurally ambiguous sentences like (50) with two distinct syntactic
representations.

(50) They ate the pizza in the living room.

(50) has two interpretations, which can be paraphrased as in (51).

(51) a. Verbal modifier It was in the living room that they ate the pizza (though the pizza
interpretation may have started out elsewhere).

b. Nominal modifier It was the pizza in the living room that they ate (though perhaps they
interpretation took it and ate it elsewhere).

On the verbal modifier interpretation in (51a), the prepositional phrase in the living room
modifies the verb ate, and (50) has the structure in (52a). On the nominal modifier interpretation
in (51b), the prepositional phrase modifies the noun pizza, and the sentence has the structure in
(52b).
(52) a. b.

Verbal modifier Nominal modifier

High attachment Low attachment

The structures in (52) are consistent with the results of relevant constituenthood tests. For
instance, substituting the ordinary pronoun it for the pizza and substituting did so for ate the
pizza yields (53a) and (53b), respectively.

(53) a. They ate it in the living room.

b. They did so in the living room.

In both sentences, the prepositional phrase is unambiguously interpreted as a verbal modifier, as

expected given that the pizza and ate the pizza are represented as constituents in (52a), but not in
(52b).

Conversely, in the question-answer pair in (54), the prepositional phrase is unambiguously

associated with a nominal modifier interpretation. Again, this is expected, since the pizza in the
living room is represented as a constituent in (52b), but not in (52a).

(54) What did they eat? The pizza in the living room.

The complement-adjunct distinction in the nominal system. Given the semantic parallel
between the sentence in (55a) and the noun phrase in (55b), it is reasonable to treat the of phrase
in (55b) as a complement of the noun author.

(55) a. This man authors murder mysteries.

b. this author of murder mysteries

That is, the elementary tree for author needed to derive (55b) is as in (56a), and the structure for
the entire noun phrase is (56b).

(56) a. b.

Since one is analogous to do so in substituting for intermediate rather than for lexical projections,
we expect the contrast between (57) and (58), and this accurately reflects the judgment of many
speakers.

(57) a. This man authors murder mysteries, and that woman does so, too.

b. this author of murder mysteries and that one

(58) a. * This man authors murder mysteries, and that woman does so nature guides.

b. * this author of murder mysteries and that one of nature guides

Some speakers, however, accept (58b), or at least do not completely reject it. How can we make
sense of this variation among speakers' judgments? Recall that complements of nouns, unlike
those of verbs, are always expressed as prepositional phrases. This means that the evidence
whether a particular phrase is a complement or an adjunct is murkier in the case of nouns than in
the case of verbs, both for children acquiring the language and for adult speakers. A further,
probably related, complication is that even nouns that are morphologically derived from
obligatorily transitive verbs are themselves optionally intransitive (for instance, compare
consume, destroy, employ with consumer, destroyer, employer). Moreover, the intransitive use of
these nouns might be more frequent than their transitive use. As a result, the mental grammar of
some speakers might include only the intransitive elementary tree in (59a), and not the transitive
elementary tree in (56a). Such speakers would have no way of deriving the structure in (56b), but
they would be able to derive the alternative structure in (59b) by adjoining the of phrase, rather
than by substituting it.
(59) a. b.

For such speakers, author in (58b) would be an N', rather than an N, and so they would accept
(58b) rather than rejecting it as ungrammatical.

Notice furthermore that the intransitive elementary tree in (59a) is available even for speakers
whose mental grammar includes the transitive elementary tree in (56a), since all speakers of
English accept (60).

(60) this author and that one

If some of these speakers allow the of phrase to adjoin into the intransitive elementary tree in
addition to substituting into the transitive one, then they, too, would judge (58b) to be acceptable
(at least marginally so).

Adjective phrases

In this section, we discuss the structure of adjective phrases, beginning with examples like those in (62)
and (63), where the prepositional phrase following the adjective is optional.

(62) a. They are proud.

b. They are proud of their grandson.

(63) a. They are happy.

b. They are happy with their car.

Recall from Chapter 2 that the pro-form so substitutes for adjective phrases. More specifically,
examples like those in (64) and (65) allow us to conclude that the of phrase is a complement of
proud in (64), but that the with phrase is an adjunct of happy in (65).

(64) a. They are proud, and we are so, too.

b. They are proud of their grandson, and we are so, too.

 c. * They are proud of their grandson, and we are so of him, too.

(65) a. They are happy, and we are so, too.

b. They are happy with their car, and we are so, too.

c. They are happy with their car, and we are so with our bikes.

We can represent these facts by associating the two adjectives with the elementary trees in (66)
and by stating that so substitutes for instances of A'.

(66) a. b. c.

Most adjectives in English, like the two just discussed, are optionally or obligatorily intransitive.
A rare case of an obligatorily transitive adjective is fond.5 The contrast in (67) is evidence for the
complement status of the of phrase (recall from Chapter 4 that obligatory syntactic dependents
are complements), and that status is confirmed by the results of so substitution.

(67) a. * They are fond.

b. They are fond of their grandson.

(68) a. They are fond of their grandson, and we are so, too.

b. * They are fond of their grandson, and we are so of him, too.

In view of the facts in (67) and (68), fond is associated with the single elementary tree in (69).

(69)

Prepositional phrases

The syntactic category P corresponds closely to the traditional part of speech of preposition, but is not
identical to it. We address two differences between the syntactic category and the traditional part of
speech in the next two subsections.
 Following standard usage in the syntax literature, we sometimes use the term 'preposition' to
refer to the syntactic category P in contexts where the difference is either clear or immaterial.

Transitivity

The etymology of the term 'preposition' (< Latin prae 'before' and positio 'position') implies that all
prepositions should precede a complement, and English does in fact have a number of obligatorily
transitive Ps, some of which are illustrated in (70). The asterisk outside the parenthesized material is a
conventional way of indicating that the parenthesized material is obligatory.

(70) a. They drove from *(Boston).

b. He's the inventor of *(that gizmo).

c. She dove into *(the water).

d. They jumped onto *(the bandwagon).

But X' theory leads us to expect that there should also be intransitive Ps, and as the examples in
(71) show, this expectation is fulfilled.

(71) a. I've never seen him before (this meeting).

b. Are you for (the proposal) or against (it)?

c. The bird flew { in, out } (the window).

d. It's time to get { off, on } (the train).

e. They jumped over (the ditch).

f. We've been fast friends ever since (that time).

g. She came to (her senses).

h. Have you looked underneath (the sombrero)?

In traditional grammar, Ps that are used intransitively are known as adverbs or particles, rather
than as prepositions, but this terminology goes against the spirit of X' theory, which seeks to
maximize the parallels among categories. From our point of view, there is as little reason for the
syntactic category of a lexical item to depend on its transitivity in the case of a P like since as
there is in the case of a V like eat. In both cases, the intransitive variant has a semantic argument
that is not expressed in the syntax, but is supplied in the course of interpretation, based on the
discourse context.

The elementary trees for of and over are shown in (72), and the full structures for the PPs headed
by them in (70) and (71) are shown in (73). Note the identity of (72c) and (73c).

(72) a. b. c.

(73) a. b. c.

Clausal complements of prepositions

As we saw in Chapter 4, verbs can take either noun phrase complements or clausal complements. (74)
gives a further example.

(74) a. He reported the monkey's dislike of camphor.

b. He reported that the monkey dislikes camphor.

The examples of transitive Ps discussed so far have all had noun phrase complements, but given
the parallel between verbs and prepositions concerning transitivity, we might expect Ps to allow
clausal complements as well. Once again, this expectation is borne out, as shown in (75).

(75) a. Noun phrase complement: { after, before, since } the war

b. Clausal complement: { after, before, since } the war ended

In traditional grammar, Ps that take clausal complements are classified as subordinating

conjunctions (along with if and that), but as in the case of intransitive Ps, we again reject the
traditional approach. First, it is conceptually uneconomical. Specifically, it expresses the
difference between (75a) and (75b) in terms of the syntactic category of the heads (preposition
vs. subordinating conjunction), which redundantly encodes the difference in the syntactic
category of the complement (noun phrase vs. clause). Second, the items in (75) share roughly the
same semantic content, regardless of the categorial status of their complement. In contrast, if and
that are relatively contentless and give the impression of functioning purely as 'grammatical
glue.' In the approach that we are advocating, the contentful and contentless subordinating
conjunctions belong to different syntactic categories (P and C, respectively).

We have not yet said what syntactic category clausal complements of prepositions belong to. At
first glance, examples like (75b) suggest that the answer to this question is IP. The elementary
tree for after in (75b) would then be as in (76), and the elementary trees for before and since
would be analogous.

(76)

There is good reason to believe, however, that clausal complements (specifically, finite clausal
complements) of P are CPs rather than IPs. As illustrated in (77a), the clausal complement of
after and prepositions like it would be headed by a silent counterpart of the complementizer that,
resulting in (77b) as the structure for after the war ended (for simplicity, the internal structure of
IP is omitted).

(77) a. b.

There are a number of empirical arguments for preferring the elementary tree in (77a) over the
one in (76). First, at least one preposition in English allows - indeed, requires - CP complements
headed by an overt complementizer, as shown in (78).

(78) They differ in *(that) they hold sharply opposing views on educational reform.

A second reason for preferring (77a) over (76) is that even though sentences like (79), with an
overt complementizer, are ungrammatical in modern English, such examples occurred freely in
Middle English. Some examples are given in (80).

The thorn character (þ) was borrowed from Old Norse and used in Old and Middle English
 where we use 'th' today. It is still used in Icelandic. The yogh character (ȝ) was used in Middle
English where we use 'g' or 'y'.

(79) * { after, before, since } that the war ended

(80) a. And after þat þis bataile was done, þe Britons assemblede ham, and went þens

'and after this battle was over, the Britons assembled (themselves)' (PPCME2,
cmbrut3,100.3080)

b. Ȝit bifore that Dauith cam to Jerusalem, a new debate roos bitwixe the men of Israel and the
men of Juda

'Yet before David came to Jerusalem, a new debate arose between the men of Israel and the
men of Juda' (PPCME2, cmpurvey,I,11.415)

c. Now, sith that i have toold yow of which folk ye sholde been conseilled, now wol I teche yow
which conseil ye oghte to eschewe.

'Now, since I have told you what kind of people you should be counseled by, now I will teach
you which advice you ought to eschew' (PPCME2, cmctmeli,223.C1.247)

Finally, analyzing clausal complements of prepositions as CPs allows us to treat prepositions in

English in the same way as prepositions in other languages such as French, where it is clear that
the clausal complements are CP complements.6

(81) a. { après, depuis, pendant } la danse

after since during the dance
'{ after, since, during } the dance'
b. { après, depuis } que Jean a dansé; pendant que Jean dansait
after since that Jean has danced while that Jean was.dancing
'{ after, since } Jean danced; while Jean danced'
c. * { après, depuis } Jean a dansé; pendant Jean dansait

In addition to these three empirical arguments, there are also important conceptual reasons to
analyze finite clausal complements of prepositions as CPs rather than IPs. An IP analysis would
force us to complicate the theory of case checking that we present in Chapter 8 as well as the
theory of wh- movement that we present in Chapter 11 and subsequent chapters.

Crosslinguistic variation in headedness

As illustrated in (82)-(88), heads in English precede their complements, and English is therefore said to
be a head-initial language. The headedness of a language (or of a category or lexical item) always refers
to the order of heads and complements. In other words, headedness is determined with respect to the
intermediate projection of elementary trees, not with respect to the maximal projection. For instance,
English determiners can be medial in their maximal projection (the possessive morpheme 's must be
preceded by a DP in Spec(DP)), and English verbs and modals must be medial in their maximal
projections, but they all count as head-initial because they are the leftmost elements in the
intermediate projections of their elementary trees.

(82) a. V They [V' pursued [DP their objective. ] ]

b. She [V' submitted [DP her application. ] ]

(83) a. I They [I' should [VP pursue their objective. ] ]

b. She [I' could [VP submit her application. ] ]

(84) a. C They agreed [C' that [IP they should pursue their objective. ] ]

b. She wondered [C' if [I' she could submit her application. ] ]

(85) a. N the [N' pursuit [PP of their objective ] ]

b. the [N' submission [ PP of her application ] ]

c. Lisa's [N' pride [PP in her work ] ]

(86) a. D [D' the [NP pursuit of their objective ] ]

b. [D' the [NP submission of her application ] ]

c. Lisa [D' 's [NP pride in her work ] ]

(87) a. A She is [A' proud [PP of her work. ] ]

b. He is [A' fond [PP of his children. ] ]

(88) a. P [P' over [DP the next five years ] ]

b. [P' with [DP great fanfare ] ]

But universal grammar by no means prescribes head-initial phrase structure. Rather, many
languages exhibit consistently head-final phrase structure; two such languages are Japanese and
Korean. The examples in (89)-(93) are from Korean. In order to avoid using 'head-final
preposition,' which is an etymological contradiction in terms, linguists have coined the term
postposition for the Ps in (93). The term adposition is a cover term for prepositions and
postpositions (that is, for Ps regardless of headedness). Examples for I and D are missing
because Korean has neither overt modals of the English sort nor overt articles; the abbreviations
in the glosses are explained in the notes,7 but are not crucial for present purposes.

(89) a. V kutul-un [V' [DP mokcek-ul ] chukwuha-yess-ta. ]

they-Top objective-Acc pursue-Past-Decl
'They pursued their objective.'
b. ku-nun [V' [DP ciwonse-lul ] ceychwulha-yess-ta. ]
3.ps.sg-Top application-Acc submit-Past-Decl
'He submitted his application.'
(90) a. C [C' [IP kutul-un mokcek-ul chukwuhayya ha-n ] tako ] tonguyha-
yess-ta.
they-Top objective-Acc pursue must-Pres that agree-
Past-Decl
'They agreed that they should pursue their objective.'
b. [C' [IP ku-nun ciwonse-lul ceychwulhayto toy ] nunci ]
kwungkumha-yess-ta.
3.ps.sg-Top application-Acc submit be-able if
wonder-Past-Decl
'He wondered if he could submit his application.'
(91) a. N kutul-uy [N' [DP mocek-uy ] chukwu ]
they-Gen objective-Gen pursuit
'their pursuit of their objective'
b. ku-uy [N' [DP ciwonse-uy ] ceychwul ]
3.ps.sg-Gen application-Gen submission
'his submission of his application'
c. Lisa-uy [N' [PP il-ey tayhan ] capwusim ]
Lisa-Gen work-in regarding pride
'Lisa's pride in her work'
(92) a. A [A' [DP il-i ] calangsule-un ] saram
work-Nom proud-Mod man
'a man proud of his work'
b. [A [DP aitul-i ] coh-un ] saram
children-Nom fond-Mod man
'a man fond of his children'
(93) a. P [P' [DP taum o nyen ] tongan ]
next five years over
'over the next five years'
b. [P' [DP tay phanphalay-wa ] hamkkey ]
big fanfare-with with
'with big fanfare'

Languages tend to be harmonic with respect to headedness; that is, they tend to be consistently
head-initial or head-final. However, in certain languages, some syntactic categories project head-
initial trees and others project head-final ones. Such mixed phrase structure is found, for
instance, in Dutch and German. The examples in (94)-(99) are from German; the reason that they
are all subordinate clauses is that main clauses in German (and Dutch) involve a complication
that obscures the position of finite verbs (see Chapter 14). As the examples show, V and A are
head-final in German, whereas C, N, D, and P are head-initial. I is missing from the examples
because German lacks modals of the English type, so that there is no conclusive evidence for the
position of I.
(94) a. V dass sie [V' [DP ihr Ziel ] verfolgten ]
that they their objective pursued
'that they pursued their objective'
b. ob sie [V' [DP ihre Bewerbung ] einreichte ]
if she her application submitted
'if she submitted her application'
(95) a. A [A' [DP seinen Prinzipien ] treu ]
his principles-Dat loyal
'loyal to this principles'
b. [A' [PP auf seine Kinder ] stolz ]
on his children proud
'proud of his children'
(96) a. C [C' dass [IP sie ihr Ziel verfolgten ] ]
that they their objective pursued
'that they pursued their objective'
b. [C' ob [IP sie ihre Bewerbung einreichte ] ]
if she her application submitted
'if she submitted her application'
(97) a. N die [N' Verfolgung [DP ihres Ziels ] ]
the pursuit their-Gen objective-Gen
'the pursuit of their objective'
b. diese [N' Treue [PP zu seinen Prinzipien ] ]
this loyalty to his principles
'this loyalty to his principles'
(98) a. D [D' die [NP Verfolgung ihres Ziels ] ]
the pursuit their-Gen objective-Gen
'the pursuit of their objective'
b. [D' diese [NP Treue zu seinen Prinzipien ] ]
this loyalty to his principles
'this loyalty to his principles'
(99) a. P [P' über [DP die nächsten fünf Jahre ] ]
over the next five years
'over the next five years'
b. [P' mit [DP grossem Trara ] ]
with great fanfare
'with great fanfare'

To complicate matters yet further, German allows postpositions, as in (100).

(100) [P' [DP den Fluss ] entlang ]

the river along
'along the river'

And finally, to really liven things up, certain adpositions in Dutch and German can either
precede or follow their complements. This is illustrated in (101) and (102), again for German;
the (a) and (b) examples share the same meaning.

(101) a. [P' wegen [DP des Wetters ] ]

because.of the weather
'because of the weather'
 b. [P' [DP des Wetters ] wegen ]

(102) a. [P' gegenüber [DP der Kirche ] ]

across.from the church
'across from the church'
b. [P' [DP der Kirche ] gegenüber ]

Dutch, too, allows such variation between head-initial and head-final adpositions, and in that
language, it is even accompanied by a systematic meaning difference. Specifically, when
adpositions with variable headedness are postpositions, their meaning is always directional, but
when they are prepositions, their meaning is generally locative. This is illustrated in (103) and
(104) (Kroch 1994).

(103) a. Ik fiets in de straat.

I bike in the street
'I ride my bike in the street.' (locative)
b. Ik fiets de straat in.
I bike the street in
'I ride my bike into the street.' (directional)
(104) a. Ik klim in de boom.
I climb in the tree
'I climb in the tree.' (locative)
'I climb into the tree.' (directional)
b. Ik klim de boom in.
I climb the tree in
'I climb into the tree.' (only directional)

In case the German and Dutch examples just discussed sound exotic, it is worth noting that
English sports two postpositions of its own, as illustrated in (105).8

(105) a. They searched the whole world over.

b. They work the whole week through.

In conclusion, we note that it is not uncommon for languages to undergo phrase structure change.
For instance, the phrase structure of Old English (ca. 800-ca. 1100 C.E.) is reminiscent of that of
modern German and Dutch; in particular, verbs were head-final for most of the Old English
period. (The three languages are closely related historically, so the syntactic similarity is not
surprising.) The first instances of verb-initial phrase structure appeared in late Old English. Early
Middle English was characterized by rampantly variable headedness in the verb phrase (Kroch
and Taylor 2000b), but by ca. 1350, the change from head-final to head-initial verb phrases was
essentially complete in all dialects of Middle English. Since Chaucer lived from 1342 to 1400,
his language is already modern in this respect, though his syntax differs quite strikingly from that
of the modern language in other ways, as we will discuss in later chapters. In the modern
language, only isolated relics of the old verb-final phrase structure survive, like the saying
Indictments do not a conviction make.9
Notes

1. It is not just the expression of agent arguments that is freer in noun phrases than in sentences.
As the contrast between (i) and (ii) shows, the same is true of theme arguments.

(i) a. The mills employed thousands; their practices damaged the environment.

b. * The mills employed; their practices damaged.

(ii) a. an employer of thousands; the damage to the environment

b. an employer; the damage

2. Notice also the related contrast in (i); the construction in (i.a) is discussed in Chapter 9.

(i) a. The manuscript appears to have been found.

b. * the manuscript's appearance to have been found

3. For expository convenience, we show the possessive morpheme 's in D. In possessive

pronouns, however, 's is absent, showing that it is necessary to posit a silent possessive
morpheme in English. This silent morpheme combines with person and number features (first
person singular, first person plural, and so on), and the resulting combination is then spelled out
as the appropriate possessive pronoun (my, our, and so on).

The abstract possessive head just introduced makes available an alternative analysis for ordinary
possessive noun phrases like the the committee's than the one in (17). Under this alternative
analysis, D contains the silent possessive head rather than 's, and the possessive form the
committee's represents the spellout of the noun phrase in Spec(DP) and the silent possessive
morpheme. In other words, under this analysis, the possessive marker 's is introduced in the
mapping from syntax to morphology, rather than in the syntax proper. For expository simplicity,
we will show an analysis with overt 's whenever possible, even though we are not theoretically
committed to it.

4. More evidence for the idiosyncratic character of the constraint against (52a) comes from the
acceptability of (i) (at least in formal registers) (thanks for Sonali Mishra for drawing such
examples to our attention).

(i) not a one, such a one

5. Strictly speaking, this statement is true only of fond in predicative position, not in prenominal
position.

(i) a. Predicative: * Their parents are fond.

b. Prenominal: ok their fond parents

6. In (81), some French speakers prefer or require the subjunctive form of the auxiliary (ait)
rather than the indicative form (a). For present purposes, this variation, which is comparable to
that found in English between If I was a rich man and If I were a rich man, is irrelevant.

7. Acc = accusative (case of direct object), Gen = genitive (possessive case), Decl = declarative
clause, Mod = modifier, Pres = present tense, ps = person, sg = singular, Top = topic of sentence.

8. A further apparent instance of a postposition in English is ago. We address the proper analysis
of ago in Exercise 7.8.

9. Cf. the parallel token in (i).

(i) There had appeared to him something rather fine in his policy of refusing to identify himself in any
way with Sedleigh, a touch of the stone-walls-do-not-a-prison-make sort of thing.
(P.G. Wodehouse. 1974. The world of Psmith. London: Barrie & Jenkins. 114.)

10. Many thanks to Amy Forsyth for example (1) in Exercise 5.8.

Exercises and problems

Be sure to use the grammar tools for this chapter, not the versions of them from the previous
chapter.

Exercise 5.1

What is the syntactic difference between standard and nonstandard them?

(1) This is definitely one of them jobs, man, if you're one of them worriers …
(Overheard at a lunch truck on the southwest corner of 34th Street and Walnut Street,
Philadelphia, PA, 31 August 1999)
Exercise 5.2

A. Formally, the structures in (1) are consistent with X' theory. Empirically, however, they are
unsatisfactory representations because they are inconsistent with certain linguistic judgments. What are
the judgments in question?

In your answer, you should feel free to substitute other determiners for the and singular for
plural nouns.

(1) a. b.

B. The representation in (2) violates X' theory. In addition, (2) doesn't properly represent the
constituenthood of one or more strings in the sentence. Using appropriate substitution tests,
identify all the strings in question. Your answer should include the strings as well as the evidence
that you used to make your decision.

The sentence in (2) is structurally ambiguous. Disregard the ambiguity, and focus on the ordinary
interpretation of the sentence.

(2)

Exercise 5.3
 This exercise assumes that you are familiar with the distinction between noun complement
clauses and that relative clauses. If not, please refer to the information at the end of the
exercise.

A. Subordinate clauses of the type illustrated in (1) are traditionally called noun complement
clauses.

(1) a. The idea that Columbus was the first European to discover America is incorrect.

b. The fact that they are wrong is lost on them.

Are such clauses syntactic arguments of the noun in boldface, or are they adjuncts? In other
words, given the way that the term 'complement' is used in X' theory, is the term 'noun
complement clause' for these clauses a misnomer, or not? Explain, giving the linguistic facts that
you base your decision on.

B. Are relative clauses, illustrated in (2), arguments or adjuncts of the noun they modify?
Explain.

(2) a. The idea that Columbus was working with was incorrect.

b. The fact that they have discovered is important.

Here are two diagnostics for distinguishing noun complement clauses from that relative
clauses. First, stripping away the complementizer that leaves a complete sentence in the case
of a noun complement clause, but something incomplete in the case of a relative clause (it
feels like there is a gap, as indicated by the underlining).

(i) a. Columbus was the first European to discover America.

b. They are wrong.

(ii) a. * Columbus was working with ___.

b. * They have discovered ___.

Second, the complementizer that can generally be replaced by a wh- word in a relative
clause, but never in a noun complement clause.
 (iii) a. * The idea which Columbus was the first European to discover America is incorrect.

b. * The fact which they are wrong is lost on them.

(iv) a. The idea which Columbus was working with was incorrect.

b. The fact which they have discovered is important.

Exercise 5.4

A. Using the grammar tool in x-bar 2, build trees for the (a) examples in (1)-(3).

Don't build structures for the examples after (a); they're just there to provide guidance
concerning the intended interpretation and structure for the (a) examples.

(1) a. the monster's mother's lair

b. the monster

c. the monster's mother

(2) a. the hero of the poem's name

b. the hero

c. the hero of the poem

(3) a. the mother of the monster's dislike of the poem's hero

b. the monster

c. the mother of the monster

d. the poem

e. the poem's hero

B. Using the same grammar tool, build structures for the noun phrases in (4).

(4) a. yesterday's lecture

 b. this week's unseasonably high temperatures

Exercise 5.5

A. Can you think of obligatorily transitive adjectives other than fond?

B. Can you think of plausible candidates for obligatorily intransitive prepositions?

Exercise 5.6

A. What is the syntactic difference between the prepositions in (1a) and (1b)? You should be
able to answer in a sentence or two.

The exercise calls for a syntactic difference, so don't give a semantic difference as your answer.

(1) a. at, despite, during, of

b. along, besides, between, by, plus, under

B. Does with belong with the prepositions in (1a) or in (1b)? Explain. (For fun, you might ask a
few of your friends whether they agree with you. Make sure to pick native speakers of English.)

Exercise 5.7

A. As first mentioned in Chapter 1, human language is characterized by structural ambiguity. For

instance, the noun phrase in (1) has two distinct interpretations, which can be paraphrased as in (2).

(1) the houses on the corner with a sign

(2) a. the houses on the corner that have a sign

b. the houses on the corner that has a sign

Using the grammar tool in x-bar 2, build two distinct structures for the noun phrase in (1), and
indicate which structure goes with which interpretation in (2).

B. Give paraphrases for the two interpretations available for (3), and use the grammar tool to
build the structures corresponding to them, indicating which structure(s) goes with which
interpretation.
(3) I enthusiastically recommend this candidate with no qualifications.

Exercise 5.8

A. The sentences in (1) and (2) are many-ways ambiguous (don't assume that they have the same
number of interpretations).10 Find as many interpretations as you can, clearly describing the
relevant situations you have in mind (see (3) and (4) for model descriptions). Using the grammar
tool in x-bar 2, build trees for each interpretation you find, clearly indicating which tree is
associated with which interpretation.

(1) The officer poked the man in the car with the gun.

(2) The trainer tapped the seal with the ball on its nose.

B. Many-ways ambiguous though (1) is, it cannot be used to describe the situation in (3). Why
not?

(3) There is a car, and outside the car are a man with a gun and an officer. The officer pokes the man
(with something or other).

C. Many-ways ambiguous though (2) is, it cannot be used to describe the situation in (4). Why
not?

(4) There is a seal balancing on its nose, and the trainer taps the seal with a ball.

Exercise 5.9

Recall the syntactically ambiguous expressions from Exercise 1.6, repeated here in (1). Relying on your
more sophisticated knowledge of phrase structure, use the grammar tool in x-bar 2 to build structures
for each possible interpretation of the expressions. Indicate clearly which interpretation each structure
is intended to represent.

If you encounter difficulties, briefly describe them and implement a solution if you can.

(1a) is trickier than it seems.

(1) a. chocolate cake icing

 b. clever boys and girls

c. John will answer the question precisely at noon.

d. Watch the man from across the street.

e. They should decide if they will come tomorrow.

Exercise 5.10

As noted at the end of this chapter, English has undergone a phrase structure change in the course of its
history. In early Old English, V and I were both consistently head-final. Over the course of Old English, I
became head-initial. In other words, in addition to old structures in which I followed VP, new ones
became available in which I preceded VP. By the beginning of Middle English (about 1100), I had become
exclusively head-initial, but V continued to be variably head-final or head-initial. Finally, by about 1350,
V had become consistently head-initial.

Given this historical sketch, use the grammar tool in variable headedness 2 to build trees for all
of the phrase structure variants of (1) that were possible during the course of the history of
English. (For the non-modern stages, simply use modern vocabulary items, but arranged
according to the relevant parameter settings.)

For simplicity, the grammar tool for this exercise implements a grammar in which the nouns in
the sentences below are not associated with any arguments. Your grammar probably agrees
with the grammar in the grammar tool, but don't worry if it doesn't. The focus of the exercise
doesn't concern the internal structure of the noun phrases.

Click here for discussion concerning the representation of proper nouns.

Once you've given the internal structure of a noun phrase for one stage, you don't have
to repeat for the other two.
(1) a. Beowulf will slay Grendel.

b. The hero will slay the monster.

c. The hero of the poem will slay the monster's mother.

Exercise 5.11

A. What is the syntactic category of hiring in (1a) and (1b)? Explain.

(1) a. Kim's impulsive hiring of incompetents is damaging the company.

b. Kim's impulsively hiring incompetents is damaging the company.

B. Using the grammar tool in x-bar 2, build structures for the underlined gerund phrases in (1).
(You don't have to build the structure for the entire sentence.)

Exercise 5.12

Using the grammar tool in x-bar 2, build structures for the sentences in (1). For simplicity, you
can build subtrees, and indicate how they would fit together.

(1) a. The students will solve the problem, though it is difficult.

b. The students will solve the problem, though we acknowledge that it is difficult.

Problem 5.1

If a given string is structurally ambiguous (that is, it has more than possible structural representation), is
it necessarily associated with more than one meaning? Explain, giving examples.

Hint: Think carefully about Exercise 5.7, B.

Problem 5.2

For all speakers of English, the sentence in (1) can have either of the interpretations in (2).

(1) Jane has a big black dog, and Jean has a small one.

(2) a. Jane has a big black dog, and Jean has a small dog.

b. Jane has a big black dog, and Jean has a small black dog.

On the other hand, (3) means only (4a) for most speakers of English. However, some speakers
are able to interpret (3) as (4b) (Radford 1988) (such variable judgments among different
speakers are conventionally indicated by a percent sign).

(3) Jane has a big black dog, and Jean has a brown one.

(4) a. Jane has a big black dog, and Jean has a brown dog.
 b. % Jane has a big black dog, and Jean has a big brown dog.

A. Which of the interpretations of (1) and (3) is problematic? Explain.

B. Can you think of a way of resolving the problem you laid out in your answer to (A)?

Problem 5.3

The structures in (1) are intended to represent the second conjunct in (2).

(1) a. b. c.

(2) the book on the table, and that on the shelf

Discuss the relative merits of the three structures in (1). In other words, what considerations
(whether empirical or conceptual) make each of the structures attractive or unattractive?

Problem 5.4

In (53) of Chapter 1, we mentioned an alternative approach to noun phrase structure than the one
presented in this chapter. According to the alternative approach (updated to be in accordance with the
X' schema), all determiners are intransitive, and all nouns have substitution nodes in the specifier
position for (possibly silent) determiners, as shown in (1).

(1) a. b.

Discuss the relative merits of this approach compared to the one presented in the text. In other
words, what considerations (whether empirical or conceptual) would make one adopt or reject
this alternative approach to noun phrase structure? Assume that N can be intransitive, as in (1b),
but that it can also take complements (to account for expressions like criticism of the proposal).
6 The verb movement parameter

Verb raising: V movement to I

o The French future tense
o The order of adverbs and verbs in French
Tense lowering: I movement to V
o The order of adverbs and verbs in English
o Do support in English
Cues for the acquisition of verb raising
Verb raising and related issues in the history of English
o The loss of verb raising
o A change in the status of not
o The emergence of do support
o The emergence of modals
o Remnants of verb raising in modern English
Notes
Exercises and problems
Supplementary material
o Modals and auxiliary verbs in English
o Node relations

As we saw in Chapter 4, tense in English can be expressed in one of two ways.1 The future tense
is expressed by will, which precedes the verb and is a free morpheme; that is, it can be separated
from the verb and stand alone.

(1) a. We will never watch that show.

b. (Will you watch that show?) We will.

c. * We watch will that show.

The past tense, on the other hand, is expressed by a bound morpheme, ordinarily the suffix -ed,
which combines with the verb to form a morphologically complex word.

(2) a. * We -ed never watch that show.

b. * (Did you watch that show?) We -ed.

c. We watch-ed that show.

This dual expression of tense is typical of the Germanic language family, to which English
belongs. In all of these languages, the future is expressed analytically (as two separate words),
whereas the past is mostly2 expressed synthetically (as a single morphologically complex word).

The syntactic structure for sentences presented in Chapter 4, according to which they are
projections of I, provides a structural locus for the free tense morpheme and is therefore
straightforwardly compatible with the analytic expression of tense. On the strength of the
analogous semantic contribution of free and bound tense morphemes to the meaning of English
sentences, we extended the IP analysis to the synthetic past tense. This extension receives further
support from the fact that in languages like French, the future tense is synthetic, yet semantically
equivalent to its analytic English counterpart. Representing all sentences uniformly as IPs does,
however, raise the question of how tense in I and the verb in V merge to form a complex word
when tense is expressed synthetically. In this chapter, we present two ways in which this merger
can come about: either V moves to I (verb raising) or I moves to V (tense lowering). Both types
of merger are instances of a more general process of head movement, by which one syntactic
head adjoins to another, forming a complex head that subsumes both simple ones. Given
theoretical assumptions that are beyond the scope of this introductory textbook, it is possible to
recast tense lowering in a way that assimilates it to verb raising. The idea is that the verb moves
up the tree in both cases, but that this upward movement is visible only in some cases (verb
raising) and invisible in others (tense lowering). Anticipating this reformulation, we refer to the
choice between verb raising and tense lowering as the verb movement parameter.

In addition to presenting the basic facts of verb raising and tense lowering, we discuss a closely
related and important topic in the grammar of English: the do support that is found in sentences
negated with not (cf. He doesn't like okra with *He not likes okra). We then review
crosslinguistic evidence that verb raising is linked (in ways that are still not fully understood) to
the overt expression of subject-verb agreement, and we discuss the process by which the loss of
agreement morphology in a language can result over time in the loss of verb raising. The chapter
concludes with a case study of the verb movement parameter and related issues in the history of
English. As we will see, the diachronic interplay of the principles of Universal Grammar with
several contingent language-particular developments has resulted in the intricate web of facts
related to the verb movement parameter that characterizes modern standard English.

Verb raising: V movement to I

The French future tense

As we mentioned, the merger of tense and the verb when tense is expressed synthetically can
take place in two directions: either the verb moves up to the tense morpheme, or the tense
morpheme moves down to the verb. We begin with the verb raising case. In this connection, it is
informative to consider the future tense in French, which is formed by attaching suffixes to a
verb's infinitive.

(3) Future tense of Present tense of

 chanter 'to sing' avoir 'to have'

je chanter-ai 'I will sing' j'ai 'I have'

tu chanter-as 'you.sg will sing' tu as 'you.sg have'

il, elle chanter-a 'he, she will sing' il, elle a 'he, she has'

nous chanter-ons 'we will sing' nous avons 'we have'

vous chanter-ez 'you.pl will sing' vous avez 'you.pl have'

ils, elles chanter-ont 'they will sing' ils, elles ont 'they have'

As is evident from (3), the future tense affixes are nearly identical to the present tense forms of
the verb avoir 'to have', the only difference being that the affixes are truncated in the first and
second person plural by comparison to the full two-syllable forms of avoir. This correspondence
suggests that the future tense in French developed via a semantic shift from 'they have to V' to
'they will V'.3 In addition, and more immediately relevant for the present discussion, the
originally free forms of avoir were reanalyzed as bound morphemes.4 The analytic roots of the
synthetic French future tense thus indicate that the two ways of expressing tense (analytic or
synthetic) are not just semantically parallel, but that they are also not as unrelated
morphologically as they seem to be at first glance. In particular, what the French case suggests is
that bound tense morphemes can project syntactic structure on a par with free tense morphemes.
The elementary trees for the future tense suffixes in (3) are then as in (4).

(4) a. b. c. d. e. f.

Given these elementary trees, sentences like (5) can be derived as follows.

(5) Nous chanter-ons une chanson.

we sing fut a song
'We will sing a song.'

We begin with the elementary tree for the verb chanter in (6a) and substitute it as the
complement of the elementary tree for the future tense marker to yield the structure in (6b).
(6) a. b.

Substitute (6a) in elementary tree

Elementary tree for chanter
of future tense suffix (4d)

The synthetic future tense form can then be created by moving the verb and adjoining it to the
left of the tense morpheme. This is shown in the step-by-step derivation in (7).

(7) a. b. c.

Select target of Clone target of Attach V as left daughter of higher

adjunction adjunction clone

The remaining steps of the derivation are identical to the ones that would be required to derive
the corresponding English sentence We will sing a song. These steps (substitution of the subject
and object arguments and subject movement) are shown in (8).

(8) a. b.

Substitute arguments Move subject

Our use of adjunction in building morphologically complex words differs in certain respects
from our earlier use of it, as summarized in (9). In particular, our present use of adjunction is
combined with movement, a fact that is highlighted by the term head movement. Nevertheless,
adjunction consists of the same formal operation in both cases: selecting a target of adjunction,
cloning it, and attaching a suitable constituent as a daughter of the higher clone.

(9) Use of adjunction for ... Modification Head movement

Adjunction structure represents Semantic relation Morphological relation

(between modifier and modifiee) (between stem and affix)

Target of adjunction Intermediate projection Head

Adjoined constituent Maximal projection Head

Movement involved? No Yes

The order of verbs and adverbs in French

The facts of French presented so far are actually consistent not only with an analysis according to which
V raises to I, but also with one in which I lowers to V. However, there is evidence in favor of the verb
raising analysis that is based on the order of verbs and adverbs (Emonds 1978).

As illustrated in (10)-(12), there are certain adverbs in French (in italics) that must ordinarily
precede the main verb of a sentence (in boldface), rather than follow it.

(10) a. Elle va à peine travailler trois heures.

she goes hardly work three hours
'She is going to hardly work three hours.'
b. Mon ami va complètement perdre la tête.
my friend goes completely lose the head
'My friend is going to completely lose his head.'
c. Je vais presque oublier mon nom.
I go almost forget my name
'I'm going to almost forget my name.'
(11) a. * Elle va travailler à peine trois heures.

b. * Mon ami va perdre complètement la tête.

 c. * Je vais oublier presque mon nom.

(12) a. * Elle va travailler trois heures à peine.

b. * Mon ami va perdre la tête complètement.

c. * Je vais oublier mon nom presque.

Be sure to focus on the French grammaticality judgments, especially in (12). Adverbs don't
necessarily behave syntactically like their translation equivalents, as highlighted by the
grammaticality contrast in (i).

(i) a. * perdre la tête complètement

b. ok lose one's head completely

These word order facts reflect the fact that the adverbs in question must adjoin to the left of V',
as shown schematically in (13), rather than to the right.

(13)

In reading the following discussion, bear in mind that our focus is not on the distribution of
adverbs per se. In particular, we are not claiming that all, or even most, adverbs left-adjoin to V'
in French; in fact, there are many that right-adjoin. Rather, the idea is that we will use the
particular subset of adverbs that left-adjoin to V' as a diagnostic tool to determine the position
of finite verbs in French.

Participles behave analogously to infinitives, as shown in (14)-(16).

(14) a. Elle avait à peine travaillé trois heures.

she had hardly worked three hours
'She had hardly worked three hours.'
b. Mon ami a complètement perdu la tête.
my friend has completely lost the head
'My friend completely lost his head.'
c. J'avais presque oublié mon nom.
I had almost forgotten my name
 'I had almost forgotten my name.'
(15) a. * Elle avait travaillé à peine trois heures.

b. * Mon ami a perdu complètement la tête.

c. * J'avais oublié presque mon nom.

(16) a. * Elle avait travaillé trois heures à peine.

b. * Mon ami a perdu la tête complètement.

c. * J'avais oublié mon nom presque.

Moreover, the negative marker pas behaves like an adverb in French.5

(17) a. Nous allons (ne) pas écouter la radio.

we go NE not listen the radio
'We are not going to listen to the radio.'
b. * Nous allons (ne) écouter pas la radio.

c. * Nous allons (ne) écouter la radio pas.

(18) a. Nous (n') avons pas écouté la radio.

we NE have not listened the radio
'We haven't listened to the radio.'
b. * Nous (n') avons écouté pas la radio.

c. * Nous (n') avons écouté la radio pas.

However, when the the main verb of the sentence is finite, the adverb-verb order that is
obligatory with infinitives and participles is ungrammatical.

(19) a. * Elle à peine travaillera trois heures.

she hardly work.fut three hours
'She will hardly work three hours.'
b. * Mon ami complètement perdra la tête.
my friend completely lose.fut the head
'My friend will completely lose his head.'
c. * Je presque oublierai mon nom.
I almost forget.fut my name
'I will almost forget my name.'
d. * Nous (ne) pas écouterons la radio.
we NE not listen.fut the radio
'We won't listen to the radio.'
Instead, the adverb must follow the verb, although it still cannot follow the entire V'.

(20) a. Elle travaillera à peine trois heures.

b. Mon ami perdra complètement la tête.

c. J'oublierai presque mon nom.

d. Nous (n') écouterons pas la radio.

(21) a. * Elle travaillera trois heures à peine.

b. * Mon ami perdra la tête complètement.

c. * J'oublierai mon nom presque.

d. * Nous (n') écouterons la radio pas.

We can make sense of these facts if we continue to assume that the adverbs under discussion
adjoin to the left of V' regardless of the finiteness of the verb that they modify. This gives the
correct adverb-verb order for infinitives and participles, and it also immediately explains the
ungrammaticality of (21). The contrast between (19) and (20) follows straightforwardly as well if
finite verbs move to I to merge with the tense morpheme, as shown in (22).

(22) a. b.

Under an analysis according to which I lowers to V, it is difficult to see how the contrast
between (10) and (11) on the one hand and that between (19) and (20) on the other could be
handled in a principled way. It is these contrasts that lead us to conclude that V raises to I in
French, rather than that I lowers to V.

As (23) and (24) show, the adverb facts for other simple tenses in French are parallel to those for
the future tense.

(23) a. Elle travaillait à peine trois heures.

she work.imperf hardly three hours
 'She used to hardly work three hours.'
b. Mon ami perd complètement la tête.
my friend lose.pres completely the head
'My friend completely loses his head.'
c. J' oublie presque mon nom.
I forget.pres almost my name
'I am almost forgetting my name.'
d. Nous (n') écoutions pas la radio.
we NE listen.imperf not the radio
'We weren't listening to the radio.'
(24) a. * Elle à peine travaillait trois heures.

b. * Mon ami complètement perd la tête.

c. * Je presque oublie mon nom.

d. * Nous (ne) pas écoutions la radio.

On the strength of this evidence, we extend the verb raising analysis to these other tenses as well.

Tense lowering: I movement to V

The order of verbs and adverbs in English

Let's now turn to English and investigate simple-tense verbs, using exactly the same diagnostic
that we did in French - namely, the position of adverbs. As in French, certain adverbs in English
obligatorily precede nonfinite verbs.

(25) a. They will { always, never } apply.

b. They have { always, never } applied.

c. They are { always, never } applying.

(26) a. * They will apply { always, never. }

b. * They have applied { always, never. }

c. * They are applying { always, never. }

But unlike French, these adverbs precede the main verb of a sentence even when the verb is
finite.
(27) a. They { always, never } applied.

b. * They applied { always, never. }

The ungrammaticality of (27b) means that the verb raising analysis that is successful for French
cannot be extended to English. But recall the second option that we mentioned earlier: that tense
and the verb might merge in the other direction, by means of I lowering onto V. The simplest
way of implementing this idea would be to have -ed project an elementary tree, as in (28a). The
past tense marker would then take a VP complement, as shown in (28b), and rightward
adjunction of the tense marker to the verb would result in the structure in (28c).

(28) a. b. c.

Elementary tree for bound Substitute VP in Lower tense morpheme from

morpheme (28a) I to V

Unsatisfactory analysis

But although such an analysis would allow us to derive regular past tense verbs, it doesn't extend
to irregular past tense forms like brought, sang, taught, and so on. In order to derive both regular
and irregular past tense forms in a uniform way, we will therefore assume a silent past tense
morpheme as in Chapter 4. It is this silent morpheme that lowers onto the verb. A question that
remains open is the exact form of the verb that merges with tense. On the one hand, V might
dominate a form that is already inflected for past tense, as in (29a). On the other hand, V might
dominate the bare form of the verb, as in (29b).

(29) a. b.

The idea is that structures like (29b) are passed on to a morphological component of the
grammar, which contains rules for how to spell out the terminal nodes of syntactic structures.
According to these rules, the past tense morpheme in English is ordinarily spelled out as -ed.
With irregular verbs, however, it is the entire combination of verb and tense that is spelled out in
more or less idiosyncratic fashion. Thus, the regular watch + [past] is spelled out as watched,
whereas the irregular sing + [past] is spelled out as sang. Although the choice between the two
approaches in (29) is not completely straightforward, we prefer the second approach for the
following reason. According to the first approach, the morphological component of the grammar
generates verb forms bearing certain properties, or features, including tense. These verb forms
then project elementary trees in the syntax that combine with other elementary trees, possibly
yielding ungrammatical structures. For instance, a present tense I might take a VP complement
headed by a past tense form. In order to rule out structures with such feature mismatches, it
would be necessary to institute a special checking procedure, either as part of tense lowering
itself or as a sort of quality control on the structures resulting from it. The second approach
avoids the need for such a procedure. The idea is that terminal nodes dominated by V contain no
tense features of their own, thus eliminating the possibility of feature mismatches in the syntax.
When the syntactic structures are passed on to the morphological component, the tense-verb
combinations are simply spelled out appropriately according to the morphological rules of the
language.

Note, incidentally, that a morphological component is necessary not just in tense-lowering

languages like English, but in verb-raising languages like French as well. As discussed in the
previous section, the future tense in French is formed for regular verbs by combining the future
tense morpheme with a verb's infinitive. In the case of irregular verbs, however, what combines
with the tense morpheme is not the infinitive, but a special stem. For instance, the future tense of
être 'to be' is formed with the stem ser-, yielding the future tense forms je ser-ai 'I will be', tu ser-
as 'you will be', and so on. The approach in (29b) can be extended to this case straightforwardly.
The idea is that in a syntactic structure like (30), the morphological rules of French spell out the
terminal nodes être + -ai as serai rather than as *êtrai.

(30)

Do support in English

In this section, we turn to an apparently idiosyncratic and quirky consequence of the fact that English
has tense lowering - namely, the do support that is necessary in sentences negated with not. In order to
clearly show the conditions under which do support takes place, we will contrast sentences containing
not with ones containing other negative elements, such as never, which don't require do support.

In vernacular English, never often functions as simple sentence negation, without its literal
meaning of not ever.
(31) a. Did you get a chance to talk to Tom at the party?

b. i. Nope, I never did.

ii. Nope, I didn't.

But despite their functional equivalence in contexts like (31), the negative elements not and
never exhibit a striking syntactic difference: not obligatorily triggers do support, whereas never
doesn't. (All forms of do in this section are to be read without emphatic stress.)

(32) a. * He not applied.

b. He { did not, didn't } apply.

(33) a. He never applied.

b. * He did never apply.

In order to explain this puzzling fact, we will develop an analysis of do support that relies on two
main ideas: first, that never and not are integrated into the structure of English sentences in
different ways, and second, that Universal Grammar allows tense lowering (and head movement
more generally) only under certain structural conditions.6

A syntactic difference between never and not. As shown in (34), never is intransitive and
hence a maximal projection in its own right, whereas not is transitive and hence a head, rather
than a complete phrase.

(34) a. b.

There are several pieces of evidence for this distinction. The first comes from negative
inversion, a construction reminiscent of the so am I construction discussed in Chapter 2 in
connection with the constituenthood of adjective phrases. (35a) shows an ordinary negative
sentence, and (35b) shows its negative inversion counterpart, in which the negative constituent
(in boldface) has moved to the beginning of the sentence, and the subject (underlined) has
inverted with the auxiliary (in italics).

(35) a. They would appreciate no present more than another novel by Wodehouse.
 b. No present would they appreciate more than another novel by Wodehouse.

We discuss the structure of sentences with inversion in Chapter 14.

An important property of this construction is that the material preceding the auxiliary must be a
maximal projection. Thus, in contrast to the DP no present in (35b), the head of the DP, the
negative determiner no, cannot undergo negative inversion on its own.

(36) * No would they appreciate present more than another novel by Wodehouse.

Bearing in mind this fact about negative inversion, consider the canonical and negative inversion
sentences in (37).

(37) a. They will never tolerate this mess.

b. Never will they tolerate this mess.

(38) illustrates the beginning of the derivation of (37a). (38a) is the structure for the positive
sentence corresponding to (37a) (where irrelevant, we omit the internal structure of maximal
projections). Adjoining never as a verbal modifier then yields (38b).

(38) a. b.

As noted earlier, we discuss the structure for sentences with inversion in Chapter 14, but what is
important for now is that never in the canonical variant is a maximal projection, and hence a
candidate for negative inversion.

Now consider the not variant of (37a) in (39).

(39) They will not tolerate this mess.

Making the reasonable assumption that I can take either NegP or VP complements, we can give
(39) the structure in (40).
(40)

Given this structure, not on its own is not a maximal projection, and so not, like no but unlike
never, should not be able to undergo negative inversion. As (41) shows, this expectation is
confirmed.

(41) * Not will they tolerate this mess.

A second piece of evidence for the status of not (and its variant n't) as a head comes from the fact
that it optionally raises and adjoins to I, forming a complex head that can exhibit morphological
irregularities. For instance, shall-n't and will-n't are spelled out as shan't and won't, respectively.
Such irregularities are typical of what is possible when two heads combine. Although the
direction of movement is different, we have seen comparable examples in connection with
irregular past tense forms in English, where the combination of two heads like sing and [past] is
spelled out as sang. Other well-known examples of the same phenomenon include the
idiosyncratic spell-outs for preposition-determiner combinations like those in (42).

(42) a. French de + le > du; de + les > des; à + le > au; à + les > aux
of the.m.sg of the.pl to the.m.sg to the.pl
b. German an + dem > am; in + dem > im; zu + dem > zum; zu + der
> zur
to the.m.dat.sg in the.m.dat.sg to the.m.dat.sg to
the.f.dat.sg
c. Italian con + il > col; in + il > nel; su + il > sul
with the.m.sg in the.m.sg on the.m.dat.sg

A constraint on tense lowering. We turn now to the second piece of our solution to the puzzle
presented by the contrast between (32) and (33), repeated here as (43) and (44).

(43) a. * He not applied.

b. He { did not, didn't } apply.

(44) a. He never applied.

 b. * He did never apply.

The idea is that tense lowering (though not verb raising) is subject to the locality condition in
(45).

(45) a. When a head A lowers onto a head B, A and B must be in a local relation in the sense that no
projection of a head distinct from A and B intervenes on the path of branches that connects A
and B.

b. An element C, C distinct from A and B (and projections of A and B), intervenes between two
elements A and B iff A (or some projection of A) dominates C and C (or some projection of C)
dominates B.

It is important to understand that intervention is defined not in terms of linear precedence, but in
terms of the structural relation 'dominate.' This means that the place to look for whether the
locality condition in (45) is satisfied or violated is not the string of terminal nodes beginning with
A and ending with B, but the path of branches that connects A with B in the tree.

The structure for (44a) is given in (46). In this structure, tense lowering is consistent with the
locality condition in (45), since adjoining never at V' results in the adverb being too low in the
tree to intervene between I and V. (In other words, AdvP isn't on the green path from I to V.)

(46)

In the structure in (47a), on the other hand, tense lowering violates the locality condition because
the red projections of Neg intervene on the path between I and V, indicated in green. As a result,
only the do support variant of (47a) is grammatical, which is shown in (47b). Although the
intermediate and the maximal projections of Neg intervene between I and V in (47b) as well,
forms of do, being free morphemes, don't need to lower onto the verb. Since the locality
constraint is a constraint on tense lowering, not a constraint on syntactic trees in general, (45) is
irrelevant and hence not violated in (47b).
(47) a. b.

Cues for the acquisition of verb raising

In this section, the Icelandic characters eth (capital Ð, lowercase ð) and thorn (capital Þ,
lowercase þ) represent the voiced and voiceless 'th' sounds in this, eth and thin, thorn,
respectively.

Our discussion so far has treated movement from V to I and from I to V as two symmetrical
parametric options provided by Universal Grammar. However, the languages in which the two
options have been studied in greatest detail - the Germanic and Romance languages - suggest
that they are ranked and that it is the V-to-I option that is preferred, all other things being equal.7

Of course, we need to take into account that in this case, as in life generally, all other things
aren't equal. Among the Germanic and Romance languages, we can distinguish two groups,
which have to do with the expression of subject agreement on finite verbs.8 All of these
languages resemble English in distinguishing three grammatical persons and two grammatical
numbers (singular and plural). In principle, therefore, a language might have six (= 3 x 2) distinct
agreement morphemes, one for each person-number combination. In languages like Italian and
Spanish, this is exactly what we find, and French makes up to four distinctions. In Germanic, no
language makes six distinctions, but Icelandic makes up to five and Yiddish makes four. The
agreement paradigms for these rich agreement languages are illustrated in (48). Square brackets
enclose material that is silent.

We focus on the number of distinctions that are made in speech, because that is what children
hear. They only learn to read and write later on, once language acquisition is virtually complete.

(48) Verb paradigms in rich agreement languages

Italian Spanish French Icelandic Yiddish

 'I speak' 'I speak' 'I will speak' 'I say' 'I say'

1 sg parl-o habl-o parler-ai seg-i zog

2 sg parl-i habl-as parler-a[s] seg-ir zog-st

3 sg parl-a habl-a parler-a seg-ir zog-t

1 pl parl-iamo habl-amos parler-on[s] segj-um zog-n

2 pl parl-ate habl-áis parler-e[z] seg-ið zog-t

3 pl parl-ano habl-an parler-on[t] segj-a zog-n

By contrast, the mainland Scandinavian languages (Danish, Norwegian, and Swedish) exhibit no
agreement morphology at all, even with a verb like 'be', which in English preserves agreement
distinctions that are not expressed elsewhere in the language. For ordinary verbs, English
expresses only one distinction in the present tense and none at all in the past tense. (49) gives
some paradigms for these poor agreement languages.

(49) Verb paradigms in poor agreement languages

Danish Swedish English

'I throw' 'I am' 'I throw' 'I am' 'I throw' 'I am'

1 sg kaster er kaster är throw am

2 sg " " " " " are

3 sg " " " " throw-s is

1 pl " " " " throw are

2 pl " " " " " "

 3 pl " " " " " "

In rich agreement languages, tense merges with the verb in the same way as we have already
seen for French; that is, the verb (in boldface) raises to I and hence precedes adverbs and
negation (in italics). This is illustrated for Icelandic and Yiddish in (50) and (51). The examples
are in the form of subordinate clauses because main clauses in Germanic introduce a
complication - briefly mentioned for Dutch and German in Chapter 5 and discussed in detail in
Chapter 14 - that eclipses verb movement to I.

(50) a. Icelandic að Jón keypti { ekki, aldrei, raunverulega } bókina

that Jón bought not never actually book.def
'that Jón { didn't buy, never bought, actually bought } the
book'
b. Yiddish az zey redn ( nit, avade, mistome } mame-loshn
that they speak not certainly probably mother-tongue
'that they { don't, certainly, probably } speak Yiddish'
(51) a. * að Jón { ekki, aldrei, raunverulega } keypti bókina

b. * az zey { nit, avade, mistome } redn mame-loshn

In poor agreement languages, on the other hand, tense merges with the verb by lowering onto it,
just as in English (in those contexts that don't require do support); in this case, the finite verb
follows adverbs and negation. (52) and (53) illustrate this for Danish and Swedish.

(52) a. Danish at Peter { ikke, ofte } drikker kaffe om morgenen

that Peter not often drinks coffee in morning.def
'that Peter { doesn't drink, often drinks } coffee in the
morning'
b. Swedish att Ulf { inte, faktiskt } köpte boken
that Ulf not actually bought book.def
'that Ulf { didn't buy, actually bought } the book'
(53) a. * at Peter drikker { ikke, ofte } kaffe om morgenen

b. * att Ulf köpte { inte, faktiskt } boken

We note in passing that the syntax of sentences containing negation is simpler in mainland
Scandinavian than it is in English. Negation patterns like other adverbs, and mainland
Scandinavian has no do support. This is consistent with the status of sentence negation in
Scandinavian as a maximal projection, as evidenced by its ability to participate in negative
inversion.9

(54) a. Swedish Inte vet jag var hon bor.

not know I where she lives
 'I don't know where she lives.'
b. Icelandic Ekki veit ég hver hun byr.
not know I where she lives

We know of no rich agreement languages in which I lowers to V. Related to this is the fact that
languages that lose rich agreement also tend to lose verb movement to I over time. Although we
do not know why this correlation between richness of agreement and verb raising should hold, it
suggests that children acquiring a language prefer the parametric option of verb raising over the
tense lowering alternative, but only if they are able to detect sufficient cues for it in the sentences
that they hear. In Germanic and Romance, the cues for the verb raising option are twofold: on the
one hand, richness of agreement, and on the other, the word order that results from verb raising
(finite verb > adverb). If the language being acquired has rich agreement, then the cues for the
verb raising option are extremely robust. This is because virtually every sentence that the child
hears contains the agreement cue, which is further reinforced by the word order cue in those
sentences that contain adverbs. Under these conditions, children acquire the verb raising option
without difficulty. On the other hand, given a language with poor agreement and without cues
from word order, the idea is that children are simply unable to acquire the verb raising option.

What happens in a language in which agreement is being lost? In such a language, agreement
first becomes variable (that is, some sentences contain agreement, whereas other do not) and then
is lost entirely. Thus, the cues from rich agreement become less frequent over time, and children
acquiring the language become increasingly dependent on the word order cue. But since not
every sentence contains adverbs of the relevant sort, the cues for the verb raising option in a
language that is losing rich agreement are nowhere near as robust as in a language with stable
rich agreement. This means that although it is possible in principle for children to acquire the
verb raising option, at least some children might acquire the tense lowering option instead (all
other things being equal). Such children would no longer produce sentences in which the finite
verb precedes the adverb. Instead, they would produce adverb-verb orders, which are errors from
the point of view of the verb raising grammar, but the only option that the tense lowering
grammar generates. Thus, the relative frequency of the word order cue would decrease yet
further, in turn decreasing the chance of other children acquiring the verb raising option. Such a
feedback mechanism would predict an overall tendency over time for the verb raising option to
disappear from the language. During a period of transition, the old parametric option might
continue to be used alongside the new one - for instance, in formal usage. But for speakers who
have acquired the tense lowering option in early childhood, the verb raising option would never
be as natural as tense lowering, and so the new parametric option would tend to supplant the old
one even in formal usage.

It is possible to track these developments in some detail in the history of the Scandinavian
languages. In Swedish, agreement begins to be lost in the 1400s, and the earliest examples of
tense lowering are from the late part of that century. During a transition period from 1500 to
1700, both verb raising and tense lowering are attested, sometimes even in the same text (as in
the (b) examples in (55) and (56)).

(55) a. Verb raising at Gudz ord kan ey vara j honom

that God's word can not be in him
 'that God's word cannot be in him'
b. när thet är ey stenoghth
when it is not stony
'when it is not stony'
(56) a. Tense lowering om den dristigheten än skulle wara onågigtt uptagen
if that boldness yet would be amiss taken
'if that boldness would yet be taken amiss'
b. wm annar sywkdom ey krenker nokon
if another illness not afflict someone
'if someone isn't afflicted with another illness'

Finally, after 1700, the verb raising option in Swedish dies out completely.

The geographically more isolated Faroese seems to be at the very tail end of the same change.
Agreement has weakened in Faroese, and speakers do not ordinarily produce verb raising
sentences. However, when asked to give grammaticality judgments, many speakers accept both
word orders in (57), characterizing the verb raising variant in (57b) as archaic.

(57) a. Tense lowering Hann spur, hvi tad ikki eru fleiri tilikar
samkomur.
(vernacular) he asks why there not are more such
gatherings
'He asks why there aren't more such gatherings.'
b. Verb raising (archaic) Hann spur, hvi tad eru ikki fleiri tilikar samkomur.

Interestingly, there is at least one dialect of Swedish, the dialect of Älvdalen, that has retained
agreement (the paradigm for kasta 'throw' is: 1, 2, 3 sg kast-ar, 1 pl kast-um, 2 pl kast-er, 3 pl
kast-a). In this dialect, as we might expect, verb raising is the only option, and tense lowering,
unlike in standard Swedish, is ungrammatical.

(58) a. um du for int gar ita ia firi brado

if you get not done this before breakfast
'if you don't get this done before breakfast'
b. fast die uar int ieme
if they were not home
'if they weren't home'
c. ba fo dye at uir uildum int fy om
just because that we would not follow him
'just because we wouldn't follow him'

Verb raising and related issues in the history of English

The characters eth (capital Ð, lowercase ð) and thorn (capital Þ, lowercase þ) were borrowed
from Old Norse and used in Old and Middle English where we use 'th' today. The yogh
character (ȝ) was used where we use 'g' or 'y' today.
This section gives a brief review of the history of the verb movement parameter in English.10 As
we will see, this part of the grammar of modern English is the culmination of one of the most
complicated chapters in the entire history of the language, and it reflects several distinct but
interlocking developments, which include:

the loss of agreement and verb raising,

a change in the status of not from intransitive (like never) to transitive,
the emergence of do support,
the emergence of a distinct class of modals, and
the retention of verb raising in the case of two lexical items (have and be).

The loss of verb raising

In Middle English, the period of the language that lasted from about 1150 to 1500, verbs exhibited
roughly as much person-number agreement as in modern French, as illustrated in (59). Silent letters are
enclosed in square brackets.

(59) Verb tense paradigms in two dialects of Middle English and two tenses of French

Southern Midlands French 'I sing' French 'I will sing'

1 sg sing-e sing-e chant-[e] chanter-ai

2 sg sing-est sing-est chant-[es] chanter-a[s]

3 sg sing-eþ sing-eþ chant-[e] chanter-a

1 pl " sing-en chant-on[s] chanter-on[s]

2 pl " " chant-e[z] chanter-e[z]

3 pl " " chant-[ent] chanter-on[t]

Given its richness of agreement, we would expect Middle English to exhibit verb raising, and so
it did. As the examples in (60)-(62) show, the finite verb moved to I across both adverbs and
negation, just as it does in French, Icelandic, and Yiddish.

(60) a. always he weneth alwey that he may do thyng that he may nat do. (PPCME2,
cmctmeli,222.C1.193)
 'he always thinks that he can do things that he can't do'

b. for þe Britons destroiede alwai þe cristen peple þat seynt Austyne hade baptisede
(PPCME2, cmbrut3,98.2951)
'for the Britons always killed the Christians that St. Austin had baptized'

c. þe ȝong man resortyd alwey to þe preste (PPCME2, cmkempe,57.1270)

'the young man always resorted to the priest'

(61) a. never for God ... ȝeueþ neuer two tymes to-geder (PPCME2, cmcloud,20.115)
'for God ... never gives two times together'

b. and y ne11 seiȝ neuer þe ryȝtful for-saken (PPCME2, cmearlps,44.1880)

'and I have never seen (lit. saw never) the righteous forsaken'

c. he thought he sawe never so grete a knyght (PPCME2, cmmalory,180.2434)

'he thought he had never seen so great a knight'

d. for þey synneden neuere. (PPCME2, cmwycser,234.204)

'for they never sinned.'

(62) a. not This emperour Claudius was so obliuiows þat, sone aftir he had killid his wyf, he asked
why sche cam not to soper. (PPCME2, cmcapchr,49.534)
'This emperor Claudius was so oblivious that, soon after he had killed his wife, he asked
why she didn't come to supper.'

b. He mad eke a precept þat no Jew into Jerusalem schuld entre, but Cristen men he
forbade not þe entre. (PPCME2, cmcapchr,52.604-605)
'He also made a law that no Jew should enter into Jerusalem, but he did not forbid
Christians from entering (lit. the entry).'

c. Ich ne hidde nouȝt þy mercy (PPCME2, cmearlps,49.2107)

'I did not hide your mercy'

d. Bott I sawe noght synne. (PPCME2, cmjulnor,60.289)

'But I did not see sin.'

e. but he wythdrowe not hir temptacyon (PPCME2, cmkempe,16.321)

'but he did not withdraw her temptation'

f. but Balyn dyed not tyl the mydnyghte after. (PPCME2, cmmalory,69.2361)
'but Balyn did not die till the midnight after.'
In the course of Middle English, several syntactic developments took place that culminated in the
complex grammar of modern English with respect to the verb movement parameter. First, by
1500, the beginning of Early Modern English, the agreement system of Middle English was
simplified, and as we would expect given what we know of the history of Scandinavian, verb
raising was lost as well. For instance, between 1475 and 1525, the frequency of verb raising
dropped from roughly 65% to 10%. In the case of adverbs, the loss of verb raising simply led to
the modern word order adverb > finite verb, as is evident from the translations for (60) and (61).
But the effects of the loss of verb raising in the case of negation were quite a bit more
complicated and involved two further changes: a change in the status of not and the emergence
of do support. We discuss these changes in turn.

A change in the status of not

Negative inversion. There is good evidence that in early Middle English not was an ordinary
adverb on a par with never and French pas. Like never and negative phrases throughout the
history of English, it could undergo negative inversion.

(63) a. & nohht ne stannt itt stille (PPCME2, cmorm,I,125.1079)

and not NE stood it still
'and it didn't stand still'
b. Acc nohht ne mihht itt oppnenn hemm Þe ȝate off heoffness blisse
(PPCME2, cmorm,I,142.1171)
and not NE might it open them the gate of heaven's bliss
'and it could not open the gate of heaven's bliss for them'

In the absence of further developments, we would therefore expect the loss of verb raising in
ordinary sentences to result in a word order change from verb > not to not > verb, as happened in
mainland Scandinavian. However, in contrast to negation in Scandinavian, not in the course of
Middle English went from being an ordinary adverb to being a head (recall the discussion of its
status as a head in connection with our discussion of do support). As a result, the modern English
counterparts of (63) are ungrammatical (as we already illustrated in (41)).

(64) a. * Not did it stand still.

b. * Not could it open the gates of heaven's bliss for them.

Adjunction to I'. There is a further piece of evidence that not changed from a phrase to a head in
the course of Middle English. In early Middle English, not could adjoin not just to V', but also to
I'.

(65) a. Þatt Jesuss nohht ne wollde Ben borenn nowwhar i þe land (PPCME2,
cmorm,I,122.1052)
that Jesus not NE wanted be born nowhere in the land
'that Jesus did not want to be born anywhere in the land'
b. ða þinges ðe hie naht ne scolden ȝiuen. (PPCME2, cmvices1,139.1728)
the things that they not NE should give
 'the things that they shouldn't give'

In this respect, not resembled never and other adverbs, which have preserved this ability to this
day, as shown in (66).12

(66) a. Middle English he swore þat Saxones neuer shulde haue pees ne reste (PPCME2,
cmbrut3,69.2088)

'he swore that the Saxons never should have peace or rest'

b. Modern He { always, never } will admit his shortcomings.

English

However, in contrast to the other adverbs, not lost the ability to adjoin to I' in the course of
Middle English, with the result that the counterparts of (65) are ungrammatical in Modern
English.

(67) a. * the things that he not should give

b. * that Jesus not would be born anywhere in the land

The emergence of do support

The reanalysis of not from an ordinary adverb to a head was essentially complete by 1400,13 and
shortly thereafter, the first examples of the contracted form n't are attested, as we might expect.
Agreement began to weaken around this time. What consequences did this have for children
acquiring sentences containing not in early Middle English? On the one hand, the rich agreement
cues for verb raising were weakening, but on the other hand, tense lowering was impossible in
sentences containing not given that not was a head. In other words, in the absence of any other
developments, the loss of verb raising in sentences containing not would have resulted in a
situation in which children were acquiring a grammar unable to generate ordinary negative
sentences!

One can imagine a number of different resolutions to such an impasse, each of them representing
a particular possible accident of history. For instance, speakers might have begun using the
adverb never to take over the function of the negative head not. In fact, this did happen in the
vernacular, as we saw in (31), but it never became the only way of expressing negation.
Alternatively, children might have managed to acquire verb raising solely on the strength of the
word order cue in sentences containing not. The idea is that these sentences might have become
particularly salient because of the impossibility of their verb lowering counterparts. As we will
see, this actually happened in connection with the auxiliary verbs have and be. In the case of
ordinary verbs, however, what actually happened in the history of English was something that
depended on an unrelated development in the language that had taken place in the 1200s: the
development of the verb do into an auxiliary element.
Like many languages, Middle English had a construction involving a causative verb and a lower
verb, in which the lower verb's agent could be left unexpressed.14 We first illustrate this
construction, which has since been lost from English, for French and German in (68). The
causative verb is in boldface, and the lower verb is in italics.

(68) a. FrenchEdouard a fait assembler une grande armée.

Edward has made assemble a great army
'Edward had a great army assembled.'
(lit. 'Edward had (someone) assemble a great army.')
b. German Eduard liess ein grosses Heer versammeln.
Edward let a great army assemble
'Edward had a great army assembled.'

In Middle English, two different causative verbs were used in this construction depending on the
dialect. The East Midlands dialect use do, as illustrated in (69), whereas the West Midlands
dialect used make. In other words, the West Midlands equivalent of (69a) would have been
(using modern spelling) Edward made assemble a great host.

(69) a. Middle English (East Kyng Edwarde dede assemble a grete hoste (PPCME2,
Midlands) cmbrut3,112.3377)
'King Edward had a great army assembled'
(lit. 'King Edward had someone assemble a great army.')

b. This Constantin ded clepe a gret councel at Constantinople (PPCME2,

cmcapchr,81.1483)
'This Constantine had a great council called at Constantinople'
(lit. 'This Constantine had (someone) call a great council at
Constantinople')

c. He ded make ferþingis and halfpenies, whech were not used before
(PPCME2, cmcapchr,128.2962)
'He had farthings and halfpennies made, which weren't used before'
(lit. 'He had (someone) make farthings and halfpennies')

Now in many discourse contexts, causative sentences like He had a great army assembled are
used more or less interchangeably with simple sentences like He assembled a great army. As a
result, in situations of dialect contact, it was possible for West Midlands speakers (those with
causative make) to misinterpret sentences with causative do from the East Midlands dialect as
just another way of saying a simple sentence. Based on this misinterpretation, they might then
themselves have begun to use do, but as an auxiliary verb bleached of its causative content rather
than as a causative verb (for which they would have continued to use their own make). Since the
border between the East and West Midlands dialects runs diagonally through England, the
chances of dialect contact and of the reinterpretation and adoption of do as an auxiliary verb
were good. In any event, it is West Midlands speakers who first used do as an auxiliary verb.
Once the auxiliary use was established, it could then have spread to other dialects, especially in
big cities like London, where people came from many different dialect backgrounds and where
dialect distinctions were leveled as a result.

What is important from a syntactic point of view is that auxiliary do occurred rarely before 1400.
However, when agreement weakened and verb raising began to be lost, auxiliary do was
increasingly pressed into service since it allowed negative sentences to be generated by the verb
lowering grammar.

The emergence of modals

Auxiliary do must either have entered the language as a modal (that is, a member of the syntactic
category I), or have been reanalyzed as one very early on, since if it had been a V, it would have
had to combine with tense and thus would have run afoul of exactly the locality constraint that it
actually helped to circumvent. In any event, auxiliary do was one of a growing number of modals
in Middle English that developed out of an earlier class of auxiliary verbs. Historically, these so-
called premodals belonged to a special class of verbs with morphological peculiarities, and
some of them were already syntactically special from the very beginning of Middle English. For
instance, the forerunners of must and shall never occur as nonfinite forms in Middle English.
Children acquiring these two premodals would therefore have had no evidence that they moved
from V to I as opposed to belonging to the category I, and so they might already have been
modals in early Middle English.

Consider now the effect of the loss of verb raising on the status of any premodals that were still
members of the syntactic category V. In particular, consider a structure like (70) (we assume that
the premodals, just like modals, took VP complements).

(70)

In the outgoing verb raising grammar, the finite modal can combine with tense even in the
presence of negation because verb raising is not subject to the locality constraint on tense
lowering. For examples like (71), this yields a schematic derivation as in (72).

(71) yef sho wil noht do it (PPCME2, cmbenrul,31.1035)

'if she will not do it'
(72) a. b. c.

The reason that we represent the verb as raising first to Neg and then I, rather than as skipping
Neg and raising directly to I, is because Middle English allows questions like (73), where the
negated verb inverts as a constituent with the subject.

(73) Wil noht sho do it?

In the incoming tense lowering grammar, structures containing not are ordinarily rescued by do
support. But in contrast to sentences containing ordinary verbs, do support in a structure like (72)
might plausibly have been ruled out on the grounds that auxiliary do inherited a constraint from
causative do that is given in (74).

(74) The complement of a causative construction cannot be headed by an auxiliary element (a

premodal, modal, or auxiliary verb like have or be).

Notice that the constraint on causative verbs in (74) is not specific to Middle English; its effects
in modern English and German are illustrated in (75) and (76). The causative verb is in italics,
and auxiliary elements are in boldface.

(75) a. No auxiliary The coach had the players run.

b. Auxiliary * The coach had the players be running.

c. * The coach had the players have run.

(76) a. No auxiliary Der Trainer liess die Spieler laufen.

the coach had the players run.
'The coach had the players run.'
b. Auxiliary * Der Trainer liess die Spieler am Laufen sein.
the coach had the players at.the running be
'The coach had the players be running.'
c. * Der Trainer liess die Spieler gelaufen sein.
the coach had the players run.part be
'The coach had the players have (lit. be) run.'
 d. * Der Trainer liess die Spieler laufen { können, wollen. }
the coach had the players run be.able want
'The coach had the players { be able, want } to run.'

Again, various ways out of this impasse are conceivable. For instance, the constraint in (74)
might have been relaxed for auxiliary do. What actually happened, however, was that any
remaining premodals were reanalyzed as modals along the lines of must, shall, and auxiliary do.
The schematic structure for (72) after the reanalysis is shown in (77).

(77)

After this reanalysis, sentences like (78), with nonfinite forms of premodals like cunnen and
mowen, both meaning 'be able to', ceased to be possible in English (at least in the standard
language).

(78) a. he schuld cun best rede þe booke (PPCME2, cmkempe,4.52)

'He should be able to read the book best.'

b. I shal not conne wel goo thyder (PPCME2, cmreynar,14.261)

'I won't be able to go there easily.'

c. and hij shul nouȝt mow stonde (PPCME2, cmearlps,19.765)

'and he shall not be able to stand'

d. Noo man shall mow resyst thy power in all thy lyfe. (PPCME2, cmfitzja,A3R.28)
'No man shall be able to resist your power in all your life.'

Remnants of verb raising in modern English

Despite the overall loss of verb raising in the history of English, verb raising is still possible with two
verbs in English, have and be. These two verbs, which did not belong to the premodals, have functioned
as both auxiliary verbs and main verbs throughout the history of the language. The two uses are
illustrated for modern English in (79) and (80); auxiliaries are in boldface and main verbs are underlined.
For more detailed discussion of the morphological and syntactic properties of have and be, see Modals
and auxiliary verbs in English.

(79) a. Auxiliary verb Perfect I have read that chapter.

 b. Progressive I am reading that chapter.

c. Passive That material is treated in the next chapter.

(80) a. Main verb: I have that book.

b. This chapter is difficult.

We begin by considering these verbs as auxiliaries in structures like (81) (we assume for
simplicity that the elementary trees for auxiliary verbs don't have specifiers, but the assumption
isn't crucial in what follows).

(81)

As just discussed in connection with modals, tense lowering is impossible in a structure like (81)
because not intervenes between tense and the verb, nor can the structure be rescued by auxiliary
do given the constraint suggested in (74). This is exactly the situation in which the premodals
were reanalyzed as instances of I. In the case of the premodals, this reanalysis was possible
because hardly any of them ever occurred as nonfinite forms. But an analogous reanalysis in the
case of auxiliary verbs was precluded because nonfinite auxiliary have and be occurred very
often in Middle English. Some examples are given in (82) and (83); again, the auxiliary verbs are
in boldface and the main verbs are underlined. In addition, the element in I (modal or premodal),
which guarantees the nonfiniteness of the auxiliary verb, is in italics.

(82) a. y shulde haue axede of here no more (PPCME2, cmbrut3,19.562)

'I should have asked no more of her'

b. and after he wolde haue conquerede al Scotland and Walys (PPCME2, cmbrut3,23.686)
'and afterwards he would have conquered all Scotland and Wales'

c. And Gutlagh wolde haue went into his countree (PPCME2, cmbrut3,25.728)
'And Gutlagh would have gone into his country'

(83) a. Bot euensang sal be saide wid foure salmes (PPCME2, cmbenrul,18.626)
'But evensong shall be said with four psalms'
 b. the wordes of the phisiciens sholde been understonden in this wise (PPCME2,
cmctmeli,226.C2.365)
'the words of the physicians should be understood in this way'

c. A sone, Josias bi name, schal be born to the hous of Dauith (PPCME2, cmpurvey,I,13.518)
'A son, Josias by name, shall be born to the house of David'

Again, of course, various ways of resolving this impasse are conceivable. For instance, the
constraint preventing do from occurring with auxiliary elements might have been relaxed.
However, what actually happened in the history of English is that children acquired the verb
raising option with precisely these two lexical items. As a result, the order of auxiliary have and
be with respect to negation in modern English parallels that in French.

English French

(84) a. Verb raising We have not read the book. Nous (ne) avons pas lu le
livre.
we NE have not read the
book
b. We are not invited. Nous (ne) sommes pas invités.
we NE are not invited
(85) a. No verb * We (do) not have read the * Nous (ne) pas avons lu le livre.
raising book.

b. * We (do) not be invited. * Nous (ne) pas sommes invités.

(86) schematically illustrates the derivation of the English examples. (86a) is identical to (81),
and as in the analogous structure for modals in (70), the verb raises to I via Neg.

(86) a. b. c.

Let us now turn to the main verb uses of have and be. We begin with have. Because of the
homonymy of main verb have and auxiliary have, main verb have for a time behaved
syntactically like auxiliary have, raising from V to I and otherwise exhibiting the syntactic
behavior of a modal, as illustrated in (87).
(87) a. Negation without do support He hasn't any money; you haven't any wool.

b. Question formation without do support Has he any money; have you any wool?

In present-day usage, however, the pattern in (87) has become archaic in American English and
is on the wane even in British English. It has been replaced by the pattern in (88), where main
verb have exhibits the syntax of an ordinary verb, not that of a modal.15, 16

(88) a. Negation with do support He doesn't have any money; you don't have any wool.

b. Question formation with do support Does he have any money; do you have any wool?

Finally, we consider main verb be, which exhibits richer agreement than any other verb in
English. Strikingly, it is also the only main verb in English that continues to raise to I.

(89) a. No do support This chapter isn't difficult.

b. Is this chapter difficult?

(90) a. Do support * This chapter doesn't be difficult.

b. * Does this chapter be difficult?

Notes

1. In what follows, we focus on the past tense since the present tense is not overtly marked at all
in English. The -s of the third person singular expresses subject agreement rather than present
tense (Kayne 1989).

2. Yiddish and the southern German dialects from which it developed are exceptions in this
regard. In these languages, the synthetic simple past has been completely replaced by the analytic
present perfect (Middle High German ich machte 'I made' > Yiddish ikh hob gemakht, literally 'I
have made').

3. A very similar shift occurred in English from 'they have to V' to 'they must V'. Such semantic
shifts, with concomitant changes in morphological status (see Note 4), are very common across
languages.

4. Such reanalysis might be the source of much, if not all, inflectional morphology. In many
cases, especially in languages that are not written, the sources of the inflections would be
obscured by further linguistic changes, primarily phonological reduction. Consider, for instance,
the development of the future tense in Tok Pisin, an English-based contact language that
originated in the 1800s and that has become the national language of Papua New Guinea. In
current Tok Pisin, particularly among speakers who learn it as a first language, the future marker
is the bound morpheme b-. We are fortunate to have written records of Tok Pisin from the late
1800s, and so we happen to know that this morpheme is the reflex of the adverbial phrase by and
by, which the earliest speakers of Tok Pisin frequently used to indicate future tense. Without
these records, a derivation of b- from by and by would be speculation at best.

5. Historically, the negative marker in French was ne, and pas, literally 'step', was an intensifier
without negative force of its own. Modern English has comparable intensifiers, as in I don't want
to do it { one bit, at all. } In the course of the history of French, ne, being phonologically weak,
was often elided in speech, and pas was reanalyzed as carrying negative force. In modern
French, ne is characteristic of the formal language, and in some spoken varieties, such as
Montreal French, ne hardly ever occurs. In the present discussion, we disregard ne, treating it as
an optional, semantically meaningless particle.

6. Do support raises some of the thorniest problems in English syntax, and no completely
satisfactory analysis of it exists as yet. So although our analysis is adequate to explain the
contrast between (32) and (33), it does not address many other puzzling facts that have been
discovered in connection with do support.

7. The discussion in this section is based on data and ideas in Barnes 1992, Falk 1993, Holmberg
and Platzack 1995, Platzack 1988, Roberts 1993, and Vikner 1995.

8. In what follows, we do not consider verb-final languages like German or Dutch. Evidence for
verb movement to I in these languages would have to come from adverbs that right-adjoin to V',
with the finite verb then moving rightward across the adverb. However, for reasons that are not
yet understood, right-adjunction to V' does not seem to possible in verb-final languages.

9. For some reason, negation cannot participate in negative inversion in Danish, perhaps because
it cannot bear prosodic stress.

10. The discussion in this section is based on data and ideas in Frisch 1997, Kroch 1989, Roberts
1993, Rohrbacher 1993, and the Penn-Helsinki Parsed Corpus of Middle English, second edition
(Kroch and Taylor 2000a).

11. Early Middle English had a negative particle ne, etymologically cognate with French ne and
syntactically comparable to it. See Note 5. The Middle English particle was lost between 1200
and 1400.

12. The possibility of adjoining adverbs to I' complicates the assignment of structures to
sentences with adverb-verb word order once verb raising begins to be lost. This is because they
could be instances of the old verb raising grammar, with the adverb adjoined at I', or instances of
the new verb lowering grammar, with the adverb adjoined at either I' or V'. In any particular
sentence, it isn't possible to tell which is the right structure. But in a corpus of sentences, it is
possible to correct for the complication introduced by the possibility of adjunction to I', because
the frequency of adjunction to I' has remained stable from Early Middle English until today
(about 15% with never). This means that frequencies of adverb-verb order appreciably over 15%
in a corpus can reliably be attributed to the verb lowering grammar.

13. Not continued to be available as an adverb with a low frequency into the 1600s. The evidence
for this is the existence, though rare, of negative sentences in Early Modern English of the
modern mainland Scandinavian type, with not preceding a finite verb, as in (i).

(i) a. they deafe mens' eares, but not edify.

b. he that filches from me my good name robs me of that which not enriches him.

c. Safe on this ground we not fear today to tempt your laughter by our rustic play.

These sentences are linguistic hybrids in the sense that they contain the adverbial not
characteristic of early Middle English, but instantiate the tense lowering parameter characteristic
of modern English. As the use of adverbial not finally dies out completely in the 1600s, so do
sentences of the type in (i).

14. A construction related to the agentless causative construction discussed in the text that was
retained a bit longer in the history of English is that in (i).

(i) a. They heard say that the English had won the battle of Agincourt.
'They heard someone say that ..., they heard it said that ...'

b. They heard tell of the wages of sin.

'They heard someone tell ...'

The construction in (i.a) survives in the nominalized form hearsay.

15. The syntactic divergence between auxiliary and main verb have is exactly comparable to that
between auxiliary and main verb do.

16. The replacement of (87) by (88) is complicated by the existence in both American and
British English of the have got pattern illustrated in (i), where have serves as an auxiliary verb
rather than as the main verb. Sutherland 2000 studies the competition among all three variants
(have with and without do support and have got) in both dialects of English.

(i) a. He hasn't got any money; you haven't got any wool.

b. Has he got any money; have you got any wool?

Exercises and problems

Exercise 6.1

Using the grammar tool in verb movement, build structures for the Middle English sentences in (1).

A note on spelling: u and v were used interchangeably in Middle English.

The data raise certain issues beyond the ones concerning verb movement. For instance,
is never before a constituent? Is such a determiner or an adjective? What about
numerals? Solve the issues as best you can, and briefly describe the issues and justify
your solutions. Assume that you can bring evidence from Modern English (or other
languages, for that matter) to bear on the structures you are building for the Middle
English sentences.
(1) a. Engist knew neuer before þat lande
'Engist never before knew that land.'

b. she saide she had neuer company of man worldely

'She said that she never had the company of any worldly man.'

c. Seynt Edmond vsyd euer after that prayer to his lyvys ende
'Saint Edmund afterwards always used that prayer till the end of his life.'

d. sche had euyr mech tribulacyon tyl sche cam to Iherusalem

'She always had much tribulation till she came to Jerusalem.'

e. I knewe never such two knyghtes

'I never knew two such knights.'

f. thes two gyauntes dredde never knyght but you

'These two giants never feared any knight but you.'

Exercise 6.2

So-called mandative verbs, such as require and suggest (but not say or think), take subjunctive
complement clauses.

(1) a. I will require that he { come, *comes, *came. }

b. I required that he { come, *comes, *came. }

The structure for the grammatical alternative in (1a) is given in (2).

(2)

Is the subjunctive element that heads these complement clauses a silent tense element or a silent
modal (corresponding roughly to should)? Explain. Take into account the facts in (1) and (3), the
results of negating the complement clauses in (1) and (3), and any other facts that you find
relevant.

(3) a. I will say that he { *drive, drives, drove } a Mazda.

b. I said that he { *drive, drives, drove } a Mazda.

Exercise 6.3

A. Using the grammar tool in verb movement, according to which not is a head that takes any
syntactic category as its complement, build two structures for (1) that are both consistent with the
locality condition on tense lowering discussed in this chapter. (Don't build structures for the
material in parentheses.)

(1) She not only wrote the letter (but she sent it).

B. Now build a structure for (2).

(2) She didn't only write the letter (but she sent it).

Be sure that the structures you build for (1) and (2) are consistent with the contrast in (3).

(3) a. * She not wrote the letter.

 b. She didn't write the letter.

Exercise 6.4

African American English (AAE) distinguishes two instances of be: so-called habitual be, the
focus of this exercise, and ordinary be. Both types of be can be used as main verbs or auxiliaries.
We pose the exercise after describing the semantic and morphological differences between the
two types of be. The data are based on Green 1998.

Habitual be has no counterpart in standard English. It is used to describe situations that are
generally true, as illustrated in (1).

(1) a. Main verb The coffee be nasty at that joint.

'The coffee is always/usually bad at that place.'

b. Auxiliary The baby be sleeping when they call.

'The baby is always/usually sleeping/asleep when they call.'

Ordinary be resembles standard English be in its use.

Unlike be in standard English, ordinary be can be silent in the present tense in AAE, as
indicated by the parentheses in (2). In this respect, AAE resembles languages like Hebrew and
Russian. Habitual be cannot be silent. We mention these facts for completeness only. For the
purposes of the exercise, disregard the silent option.

(2) a. Main verb This coffee (is) nasty.

'This coffee is bad.' (as a one-time occurrence)

b. Auxiliary The baby (is) sleeping.

'The baby is sleeping.' (now)

The two types of be also differ morphologically, as shown in (3).

(3) Ordinary be Habitual be

I am be
 you is "

he/she/it " "

we " "

y'all " "

they " "

Given the facts in (1)-(3) and the further fact in (4), take a stab at what the emphatic, negated,
and interrogative versions of (1) are. Assume that the grammars of AAE and standard English
are identical unless you are forced to assume the contrary by the facts in (1)-(4).

(4) Assume that AAE, unlike standard English, doesn't have person agreement in the present. In other
words, AAE has I, he played; I, he play; I, he did; I, he do.

Exercise 6.5

A. Explain the grammaticality contrast in (2). If necessary, invent a new syntactic category for so to
belong to.

Not every speaker shares the judgments in (1) below. If you are such a speaker, simply describe
the judgments as stated, without worrying about the discrepancy in judgments.

(1) A (challenging B): You're lying; you didn't go to the movies.

(2) a. B (responding to the challenge): I did so go to the movies.

b. B: * I so went to the movies.

B. Does the so in (2) have the same syntactic properties as the so in (3)? Explain briefly.

(3) So did I.

Exercise 6.6
One of the Korean sentences in (1) is ungrammatical because it violates a principle of Universal
Grammar. Which sentence is it, and why is it ungrammatical? If you wish, you can use the
grammar tool in Korean negation to build structures for the sentences in (1).

Assume that hay-essta is a morphologically simple head of category I (despite containing the
same -essta as mek-essta). This is exactly parallel to the way that we treat inflection on
auxiliary do in English.

The case morphemes for nominative and accusative case are included for
completeness. Disregard them for the purposes of the exercise.

The data are somewhat simplified in that they do not reflect a syntactic process called
scrambling. As a result, native speakers of Korean will find more than one of the
sentences in (1) unacceptable.
(1) a. Chulswu-ka pap -ul mek-essta.
Chulswu-Nom meal-Acc eat-Past
'Chulswu ate the meal.'
b. Chulswu-ka pap -ul mekci ani hay-essta.
Chulswu-Nom meal-Acc eat not do -Past
'Chulswu did not eat the meal.'
c. Chulswu-ka pap -ul mek-essta ani.
Chulswu-Nom meal-Acc eat-Past not
'Chulswu did not eat the meal.'
d. Chulswu-ka an pap-ul mek-essta.
Chulswu-Nom not meal-Acc eat-Past
'Chulswu did not eat the meal.'

Exercise 6.7

Explain the contrast in (1).

(1) He said he would give away all his money, ...

a. and give away his money he did.

b. * and gave away his money he.

Problem 6.1

You are an archaelogist living in the 31st century C.E., and your work involves deciphering and analyzing
linguistic relics from an ancient North American empire. The oldest sentence types that you have, from
the very beginning of the 21st century, are illustrated in (1) and (2).

(1) a. He is never late.

 b. He isn't late.

(2) a. She never regretted her extravagances.

b. She didn't regret her extravagances.

Based on data for the verb be, given in (3)-(5), there is evidence of three later stages of the
language (not necessarily presented in chronological order).

(3) a. He never bees late.

b. He doesn't be late.

(4) a. He bees never late.

b. He bees not late.

(5) a. He never bees late.

b. He not bees late.

There are only two sentence types attested in connection with ordinary verbs, the type in (2) and
that in (6) (where (6a) is identical to (2a)).

(6) a. She never regretted her extravagances.

b. She not regretted her extravagances.

A. What are the properties of the grammmars that generate the sentence types in (3)-(5)? Is it
possible to arrange the grammars in chonological order? Explain briefly. (The radiocarbon dating
machine or whatever archaeologists are using in the 31st century to date the media bearing the
sentences has gone on the blink, and so you are forced to arrange the data based in internal
linguistic evidence alone.)

B. Which of the variants in (2) and (6) goes with which of the variants in (3)-(5)? Explain
briefly.

Problem 6.2

In the analysis presented in the text, adverbs like pas 'not' obligatorily left-adjoin to V', and the order
finite verb - adverb is derived by moving the finite verb leftward across the adverb. An alternative
analysis is conceivable in principle, which dispenses with verb movement, and according to which
adverbs like pas can either left-adjoin or right-adjoin to verbal projections. Assess the relative merits of
the analysis presented in the text and the alternative analysis just described.

Problem 6.3

Certain English dialects, including dialects in the southern United States, allow the so-called double
modal construction illustrated in (1).

(1) a. I might can come to your party

b. I might could come to your party

The second modal in the construction is essentially restricted to can or could. Propose an
analysis of the construction.

7 VP shells

Double-object sentences
o The structure of ordinary causative sentences
o Parallels between causative sentences and double-object sentences
o Abstract verb movement
Double-complement sentences
o Give and send
o Put
o Persuade
The causative alternation
o Manner of motion verbs
o Get
Further issues
o Locality constraints on idioms
o Small clauses revisited
Notes
Exercises and problems
Supplementary material
o Thematic roles

In Chapter 4, we briefly mentioned the binary-branching hypothesis - the idea that syntactic
nodes have at most two daughters. At first glance, this hypothesis seems incompatible with the
existence of double-object sentences in natural language, illustrated for English in (1).

(1) Travis will give Betsey the receipts.

In such sentences, the verb appears to be associated with three semantic arguments (agent,
recipient, theme), and it looks like the recipient (Betsey) and the theme (the receipts) both need
to be represented as complements of the verb.1 In this chapter, we present a proposal for how to
make double-object sentences consistent with the binary-branching hypothesis. The proposal
hinges on the fact that ditransitive verbs like give can be semantically decomposed into a
causative part and a remainder whose meaning differs according to the verb in question. Some
examples are shown in (2).

(2) feed = cause to eat

give " get

lend " get (temporarily)

show " see

teach " learn

The decomposability of the verbs in (2) suggests deriving sentences like (1) from schematic
structures like (3), where the uppercase predicates CAUSE and GET indicate abstract verbal
heads.

(3)

The structure in (3) accommodates the same three arguments as the original sentence in (1), but
since there are now two heads, neither of them needs to be associated with more than one
complement - exactly as required by the binary-branching hypothesis. Because one VP is
embedded directly under another, structures like (3) are known as VP shells.2

In order to motivate the VP shell treatment of double-object sentences, we begin by discussing

ordinary causative sentences (ordinary in the sense that the causative verb is overt). After
showing that causative verbs take a VP small clause complement, we present some striking
parallels between causative sentences and double-object sentences in Japanese.

Strictly speaking, according to the VP shell analysis, there are neither double-object verbs nor
double-object sentences. However, these terms are so well established that we will continue to
use them for expository convenience. We will use the term 'double-object verb' to refer to a
 complex V head that is associated with three semantic arguments, and the term 'double-object
sentence' to refer to sentences containing such a verb and two DPs bearing the thematic roles of
recipient and theme.

We then turn to the details of the structure in (3); in particular, we propose that the lower verbal
head adjoins to the higher one, yielding a complex verb that is spelled out depending on the
content of the lower head. This extends an idea already introduced in Chapter 6, where we said
that the combination of sing and past is spelled out as sang. In a similar way, we are saying here
that, for instance, the combination of CAUSE and GET is spelled out as some form of give.

Having presented the core components of the VP shell analysis, we extend it to cover several
other important cases in English. We first consider the variant of (1) given in (4), where the order
of the recipient and theme arguments is reversed and the recipient argument is expressed by a PP
rather than a DP.

(4) Travis will give the receipts to Betsey.

We will refer to DP-PP sentences like (4) as double-complement sentences. Again, we use this
term strictly for expository convenience and do not intend to imply a ternary-branching
structure.

We then discuss the ditransitive verbs put and persuade as well as verbs that participate in the
causative alternation illustrated in (5).

(5) a. The ball dropped.

b. The children dropped the ball.

The final section of the chapter addresses two issues related to VP shells. The first issue arises in
connection with a proposed constraint on idioms according to which they must be constituents.
At first glance, idioms like give someone the creeps and throw someone to the wolves violate this
constraint because they appear to be discontinuous. However, just as the VP shell analysis allows
us to maintain the binary-branching hypothesis in the face of double-object and double-
complement structures like (1) and (4), so, too, does it allow us to maintain the structural
constraint on idioms in the face of apparently discontinuous idioms. The second issue concerns
small clauses. Having motivated the VP shell analysis with reference to causative small clauses,
we conclude the chapter with a discussion of the structure of small clauses more generally.

Double-object sentences

The structure of ordinary causative sentences

We begin our exploration of VP shells by considering ordinary causative sentences like the one in (6),
where the semantic notion of CAUSE is overtly expressed by the verb let.

(6) God let there be light.

Recall from Chapter 3 that expletive there must be licensed as the subject of a verb of existence
(be in (6)). It follows from this that the sequence there be light forms a small clause, a minimal
instance of predication (minimal because unlike an ordinary clause, it doesn't contain any overt I
element). It is this small clause that serves as the complement of let, as shown in (7).

(7) a. b.

The treatment of there be light as a constituent is motivated not only in syntactic terms (with
reference to the licensing requirement on expletive there), but also by the intuition that let takes
two semantic arguments, an agent (expressed by the matrix subject) and a situation (expressed by
the small clause).

A related piece of evidence that causatives like let takes small clause complements comes from
sentences like (8).

(8) John let it slip that the president's schedule had changed.

The it in (8) is the expletive it discussed in Chapter 3, which is associated with that clauses. Like
expletive there, expletive it must be a subject, and therefore the sequence it slip that ... must be a
small clause.

Parallels between causative sentences and double-object sentences

In certain languages, causative sentences and double-object sentences exhibit noteworthy

parallels. One such language is Japanese, where the case-marking of arguments is strikingly
similar in both sentence types. Case is discussed in more detail in the next chapter, but all that is
important for present purposes is that different case particles in Japanese mark certain
grammatical relations. Specifically, subjects are generally marked with the case particle -ga, as
distinct from direct objects, which are marked with the particle -o.3
(9) a. Taroo-ga hasit-ta koto
nom run past that
'(the fact) that Taroo ran'
b. Taroo-ga ringo-o tabe- ta koto
nom apple acc eat past that
'(the fact) that Taroo ate an apple'

Embedding a sentence under a causative verb has the following effects on case marking. When
the complement sentence is intransitive, the matrix subject is marked with -ga, as usual, but the
subject of the complement clause is marked with -o, as shown in (10). (This is analogous to what
happens in English in They ran and We made them run.)

As the hyphens indicate, the causative verb -(s)ase is a bound morpheme. We return to this fact
shortly.

(10) Hanako-ga Taroo-o hasir-ase- ta koto

nom acc run caus past that
'(the fact) that Hanako made Taroo run'

Given (10), one might expect embedding a transitive sentence under a causative to lead to the
case-marking pattern in (11). The object of the lower clause is expected to be marked with -o
because it is an object (as in (9b)), and the subject of the lower clause is expected to be marked
with -o, too, by analogy to (10). (Again, this would be analogous to what happens in English in
They chased him and We made them chase him.) As it turns out, however, the case-marking
pattern in (11) is ungrammatical, violating what is known in the Japanese syntax literature as the
double -o constraint, which prohibits the occurrence of more than one -o-marked noun phrase per
sentence.

(11) * Hanako-ga Taroo-o ringo-o tabe-sase-ta koto

nom acc apple acc eat caus past that
Intended meaning: '(the fact) that Hanako made Taroo eat an apple'

Rather, when a transitive sentence is embedded under a causative verb, the subject of the lower
clause must be marked with a distinct particle, the dative case marker -ni.

(12) ok Hanako-ga Taroo-ni ringo-o tabe-sase-ta koto

nom dat apple acc eat caus past that
'(the fact) that Hanako made Taroo eat an apple'

What is of interest to us now is that the -ga -ni -o case-marking pattern in (12) recurs in double-
object sentences, as shown in (13a). We underline the parallel by paraphrasing (13a) with the
ordinary causative construction in (13b).

(13) a. Hanako-ga Taroo-ni hon- o mise- ta koto

nom dat book acc show past that
'(the fact) that Hanako showed Taroo a book'
 b. Hanako-ga Taroo-ni hon- o mi- sase- ta koto
nom dat book acc see caus past that
'(the fact) that Hanako made Taroo see the book'

The identical case-marking pattern is exactly what the VP shell proposal leads us to expect, since
the relevant structures, given in (14), are analogous. (Bear in mind that the combination of the
abstract morphemes MIRU 'see' and -(S)ASE in (14a) is spelled out as a form of miseru 'show'.)

The complete structures of the sentences under discussion would of course also include
projections of I (past tense) and C (the complementizer koto). These projections are almost
always omitted in this chapter, though, for expository clarity.

(14) a. b.

Abstract verb movement

The Japanese causative exhibits a further property that is important for the VP shell analysis. Unlike the
English verb let, Japanese -(s)ase is a bound morpheme. As they stand, therefore, the VP shells in (14)
are not yet grammatical. The missing step is for the verb of the lower VP to adjoin to the causative
morpheme. This V-to-V movement is motivated by the same considerations as the V-to-I movement
discussed for French in Chapter 6; in both cases, a verb moves up the tree in order to "support" a bound
morpheme. The result for (14b) is shown in (15).

(15)
In view of the semantic and case-marking parallels between causative and double-object
sentences, it makes sense to extend the overt verb movement in (15) to the double-object case. In
other words, we will assume that verb movement applies to (14a), just as it does to (14b),
yielding (16) as the final VP shell for (13a).

(16)

Finally, we assume that the VP shells for English double-object verbs are analogous to the ones
that we have just motivated for Japanese. (17) shows the VP shell structures, before and after
verb movement, that we are assuming for the English counterpart of (13a). From a structural
point of view, the only difference between the Japanese structures and their English counterparts
is the direction in which V takes phrasal complements.

(17) a. b.

Why do we left-adjoin (rather than right-adjoin) SEE to CAUSE in English? The reason is that
we treat CAUSE by analogy to a suffix like -ify (cf. magn-ify, not *ify-magn).

(18) gives the VP shell for our original English double-object sentence in (1), both before and
after abstract verb movement, and (19) gives the structure for the entire sentence. In the
corresponding present-tense or past-tense sentence, the tense morpheme would lower onto the
complex V head, and the resulting head would be spelled out as give(s) or gave.

(18) a. b.
(19)

Double-complement sentences

Give and send

Many double-object sentences have a double-complement counterpart in which the order of the
recipient (red) and theme (blue) arguments is reversed and the recipient is expressed as a PP rather
than a DP.

(20) a. Travis gave Betsey the receipts.

b. Travis gave the receipts to Betsey.

At first glance, double-complement sentences seem to be completely synonymous with their

double-object counterparts and to stand in a one-to-one correspondence with them. Indeed, early
on in generative grammar, it was held that any double-complement sentence could be
transformed into a double-object sentence by an operation that was known as Dative Shift (recall
that the recipient argument is marked by a dative particle in Japanese). However, certain subtle
semantic restrictions on the two sentence types have been observed that have led this view to be
abandoned (Green 1974, Oehrle 1976, Jackendoff 1990). For instance, recipients in double-
object sentences, but not in double-complement sentences, are constrained to be animate.4

Double-object sentence Double-complement sentence

(21) a. Travis sent Betsey the receipts. (22) a. Travis sent the receipts to Betsey.

b. * Travis sent the post office box the b. Travis sent the receipts to the post office
receipts. box.

This effect is so strong that noun phrases that can be interpreted as inanimate in a double-
complement sentence are forced in the corresponding double-object sentence into an animate
interpretation, if that is possible. For instance, Philadelphia might be interpreted metonymically
as the people at the Philadelphia office.
(23) a. Travis sent the receipts to Philadelphia. (ambiguous between metonymy and location reading)

b. Travis sent Philadelphia the receipts. (only metonymy reading)

What the facts in (21)-(23) suggest is that ascribing exactly the same thematic role (that of
recipient) to the first DP in a double-object sentence and to the PP in a double-complement
sentence is not quite correct. Rather, the PP in a double-complement sentence designates a path
along which the theme moves. Accordingly, we will represent double-complement sentences
using VP shells in which CAUSE takes a small clause complement headed by GO. GO selects as
its complement a directional PP, designating the theme's path. The endpoint of the path can be
expressed by either a recipient (as in (22a)) or a location (as in (22b)). We give the structures that
we are assuming shortly.

This move of carefully distinguishing between recipients and locations is supported by the
parallel between (21)-(23) on the one hand and the corresponding simple 'get' and 'go' sentences
in (24) and (25) on the other.

Parallel to double-object
Parallel to double-complement sentence
sentence

(24) a. Betsey got the receipts. (25) a. The receipts went to Betsey.

b. * The post office box got the b. The receipts went to the post office box.
receipts.

c. Philadelphia got the receipts. c. The receipts went to Philadelphia. (ambiguous

(only metonymy reading) between metonymy and location reading)

Let us now spell out in detail how the pattern of judgments in (21)-(25) can be made to follow
from distinguishing between recipients and locations. For convenience, we introduce the notion
of argument array, by which we simply mean an unordered list of semantic arguments that are
associated with a (possibly abstract) Fregean predicate. As we already saw in Chapter 4 in
connection with optionally transitive verbs like eat, predicates can be associated with more than
one argument array. (26) gives argument arrays for eat as well as for the abstract predicates GET
and GO that are of interest here.

(26) a. eat { agent, theme };

{ agent }

b. GET5 { recipient, theme }

c. GO { recipient, theme };
 { theme, location }

As we know from Chapter 4, semantic arguments are mapped onto (= associated with) positions
in syntactic structures. As it turns out, this mapping is not one-to-one. For instance, in the case of
GET, the argument array { recipient, theme } is mapped onto the structure in (27a). In the case of
GO, the same array is mapped onto the structure in (27b).

(27) a. b.

Given the two structures in (27), the acceptability of (24a) and (25a) follows straightforwardly.
So does the acceptability of (21a) and (22a), which simply reflect the embedding of the
structures in (27) under CAUSE.

As we have just seen, a single argument array can be mapped onto more than one syntactic
structure. Conversely, a single syntactic structure can be associated with more than one argument
array. In particular, the location argument in a { theme, location } argument array can occupy the
same structural position as the recipient argument in (27b), as shown for GO in (28).

(28)

It is this structure that underlies (25b) and its causative counterpart in (21b), and the location
interpretations of (25c) and (23a).

As it turns out, there is no mapping between the { theme, location } argument array and the
structure in (29).6

(29)
This is the reason for the unacceptability of (24b) and its causative variant (21b) and for the
unavailability of a location reading in (24c) and (23b). (For completeness, we must also assume a
semantic constraint that prevents locations from serving as recipients.)

At this point in our discussion, let us return to the animacy constraint stated earlier (in
connection with (21) and (22)), according to which the recipient in a double-object sentence
must be animate. More generally, the constraint would lead us to expect that any recipient of
GET must be animate. As (30) shows, however, the constraint is not actually correct.

(30) a. Tina gave the cabinet a fancy handle.

b. The cabinet got a fancy handle.

Having distinguished between recipients and locations allows us to summarize the preceding
discussion as in (31).

(31) The first object in a double-object sentence must be a recipient and cannot be a location.

Statistically speaking, recipients tend to be animate, and it is this tendency that was enshrined as
a categorical assertion in early work on the topic.

In the double-complement examples presented so far, the path complement is headed by a

transitive P. Of course, as we would expect given X' theory, the projection of an intransitive P
can serve as a path complement as well (in traditional grammar, what we here call intransitive Ps
would be called adverbs).

(32) Travis sent the receipts { here, there } .

Notice that the pro-PPs here and there must refer to locations. As expected, the double-object
counterpart of (32) in (33) is unacceptable.7 The contrast between (32) and (33) is exactly
analogous to the contrast between (22b) and (21b).

(33) * Travis sent { here, there } the receipts.

From what we have said so far, it is clear that not every double-complement sentence has a
double-object counterpart (double-complement sentences with location arguments don't).
However, since both recipients and locations can designate the endpoint of a path, it might still
be the case that every double-object sentence has a double-complement counterpart. But this
turns out not to be true either. The reason is that the path argument in a double-complement
structure imposes a semantic requirement of its own on the theme: namely, that the theme
undergo a transfer from one end of the path to the other. Themes in double-object sentences, on
the other hand, needn't satisfy a path-related requirement. This explains how there can be double-
object sentences like (34), whose double-complement counterparts are awkward at best.

Double-object sentence Double-complement sentence

(34) a. The scandal gave the reporter an idea. (35) a. * The scandal gave an idea to the reporter.

b. Bright lights give Amy a migraine b. * Bright lights give a migraine headache to
headache. Amy.

In English, the experiencer of an idea or a headache is treated as a recipient, and since it is

perfectly possible for ideas or migraine headaches to be the result of certain causes, the double-
object sentences in (34) are acceptable. The reason that the double-complement sentences are
unacceptable is that the idea and the headache are taken to arise within somebody's head without
having traveled there along some path. Once again, as expected, the simple 'get' and 'go'
sentences in (36) and (37) are parallel to (34) and (35).

Parallel to double-object sentence Parallel to double-complement sentence

(36) a. The reporter got an idea. (37) a. * An idea went to the reporter.

b. Amy got a migraine headache. b. * A migraine headache went to Amy.

Notice that even contagious diseases don't travel along a path. Instead, they spread (occupying
their original location in addition to the new location). This explains the contrast between (38)
and (39).

(38) a. Jerry gave Amy his cold.

b. Amy got a cold.

(39) a. * Jerry gave his cold to Amy.

b. * A cold went to Amy.

In concluding this section, we should point out that we have implicitly focused on the similarities
between send and give. Not surprisingly, of course, the two verbs do not behave completely
identically. In particular, the argument array associated with send can contain either a recipient
or a location, whereas that associated with give must contain a recipient argument, not a location
argument. Another difference between give and send is that send, by virtue of its irreducible
meaning, imposes a path requirement on the theme even in a double-object sentence. This
explains the contrast between (38a) and (40).
(40) * Jerry sent Amy his cold.

Put

Another double-complement verb is put, which, like send, is associated with the argument array { agent,
theme, location }. Put differs from send, however, in that its lower VP shell is headed by BE, which
selects a purely locative complement (rather than the directional complement selected by GO).

(41)

Unlike give or send, put is not associated with the argument array { agent, recipient, theme }. Put
therefore appears in double-complement sentences, but not in double-object sentences, as shown
in (42).8

(42) a. Amy put the books { on the shelf, there }.

b. * Amy put { the shelf, there } the books.

Persuade

In the examples of VP shells that we have considered so far, the lowest complement has been VP
(causatives), DP (double-object verbs), or PP (double-complement verbs). In persuade, we have a case of
a VP shell in which the lowest complement is a clause (CP), which can be either finite or nonfinite, as
illustrated in (43).

(43) a. Finite: We persuaded him that he should do it.

b. Nonfinite: We persuaded him to do it.

(44) gives the VP shell for the finite case.

(44)

We defer discussion of the nonfinite case until Chapter 9, which is devoted to discussing
nonfinite complement clauses.

The causative alternation

Manner of motion verbs

This section extends the VP shell approach to the so-called causative alternation, illustrated in (45).

(45) a. The ball dropped. ~ The children dropped the ball.

b. The ball rolled down the hill. ~ The children rolled the ball down the hill.

c. The boat sank. ~ The explosion sank the boat.

In the intransitive variant on the left, the verbs drop, roll, and sink designate a manner of motion,
and the subject expresses a theme argument. In the transitive variant on the right, the subject is
an agent or cause initiating the motion, and the theme argument follows the verb. These facts
follow straightforwardly if we assume that the structure for the transitive variant contains the
intransitive variant embedded under CAUSE, as shown in (46) and (47). For clarity, we show the
shell structures both before and after any instances of movement that apply.

(46)

(47) a. b.
Notice that in the predicates under discussion, the simple and the causative variant are both
spelled out using the same phonological form. For instance, both DROP in (46) and CAUSE +
DROP in (47b) are spelled out as the same phonological form drop. Any verbal heads for which
this is true (not necessarily a manner of motion verb) are said to participate in the causative
alternation. Conversely, however, even though give is semantically the causative corresponding
to get, give and get are not said to participate in the causative alternation.

Get

Although give and get are not causative alternants in the sense just defined, get on its own does
participate in the causative alternation. In other words, the combination CAUSE + GET can be spelled out
not only as give, but also as get, as illustrated in (48).9

(48) a. Betsey got the receipts.

b. Travis got Betsey the receipts.

The argument structure for GET in (48a) is already familiar from (27a) and is repeated here as
(49a). Embedding (49a) under CAUSE results in (49b). For simplicity, we show only pre-
movement structures in what follows.

(49) a. b.

As we know from our earlier discussion, the { recipient, theme } argument array can also be
mapped onto a DP-PP structure. In the case of give and send, this alternative structure contains
the abstract head GO. (50a) shows that GET can head the same structure as GO, and (50b) shows
that this GET, too, participates in the causative alternation. The necessary structures are given in
(51).

(50) a. The receipts got to Betsey.

b. Travis got the receipts to Betsey.

(51) a. b.

For simplicity, in our earlier discussion of give and send, we associated GET with the single
argument array { recipient, theme }. In fact, like GO, GET is also associated with the argument
array { theme, location }. Again, as in the case of GO, this second argument array maps onto a
structure that is identical to (51a), except that the lowest complement is a location rather than a
recipient. The resulting structure and its causative alternation counterpart are shown in (52).

(52) a. b.

The sentences in (53) illustrate this argument array.

(53) a. The receipts got to the post office box.

b. Travis got the receipts to the post office box.

Sentences with causative get pattern as expected with respect to the distinction between
recipients and locations, as illustrated in (54).

(54) a. * Travis got { the post office box, here } the receipts. (cf. (21b), (33))

b. Travis got Philadelphia the receipts. (only metonymy reading) (cf. (23b))

c. Travis got the receipts to Philadelphia. (ambiguous between metonymy and location reading)
(cf. (23a))
 d. Tina got the cabinet a fancy handle. (cf. (30a))

Moreover, just as in the case of GO, the theme in a DP-PP structure headed by GET must
undergo transfer.10

(55) a. * An idea got to the reporter. (cf. (37a))

b. * A { migraine headache, cold } got to Amy. (cf. (37b))

Further issues

Locality constraints on idioms

It has been traditional in generative grammar to (attempt to) impose a locality constraint on idioms
along the lines of (56) (locality constraints are so called because they make reference to relatively small,
or local, domains).

(56) All parts of an idiomatic expression must together form a constituent.

The motivation for (56) is the desire to impose a structural restriction on what can count as an
idiom in natural language, thereby preventing arbitrary combinations of words and phrases from
having idiomatic readings. For instance, (56) prohibits idioms like the made-up example in (57),
because blue and hop don't by themselves form a constituent.

(57) a. The blue lunch at Bitar's hops.

Intended meaning: 'The lunch at Bitar's is unusually large.'

b. They've bred a strain of blue drosophila that hops.

Intended meaning: 'They've bred a strain of drosophila that is unusually large.'

c. The great apes all have blue brains that hop.

Intended meaning: 'The great apes all have unusually large brains.'

In many cases, the constraint in (56) is trivially satisfied. For instance, red tape 'bureaucratic
difficulties' is an NP, the Big Apple 'New York City' is a DP, and kick the bucket 'die' or let the
chips fall where they may 'disregard the consequences of one's actions' are VPs. There are even
idioms that consist of entire clauses, like The shit hit the fan. Crucially, however, there shouldn't
be any idioms consisting of discontinuous chunks. At first glance, therefore, idioms like those in
(58) seem to pose a problem for the locality constraint in (56).
(58) a. give someone the creeps 'make someone uneasy'

b. throw someone to the wolves 'sacrifice someone'

However, just as the VP shell analysis allows us to preserve the binary-branching hypothesis in
the face of prima facie counterevidence, it also allows us to preserve the locality constraint on
idioms in the face of idioms like (58). This is because the VP shell analysis allows us to say that
what is idiomatic in (58) are the underlined VPs in (59).

(59) a. CAUSE someone GET the creeps

b. CAUSE someone GO to the wolves

Strong evidence for the decomposition in (59) is the existence of the related idioms in (60).

(60) a. get the creeps

b. go to the wolves

In addition, since heads form constituents with their complements but not with their specifiers,
potential idioms such as those in (61) are predicted not to be possible.

(61) a. the { creeps, wolves } GET someone

b. the { creeps, wolves } GO to someone

This elegantly explains the unacceptability of sentences like (62) and (63) (on their intended
idiomatic interpretation).

(62) a. * The creeps got me.

b. * The wolves went to Felix.

(63) a. * Oscar threw the wolves Felix. (= CAUSE the wolves GET Felix)

b. * Crazy people give the creeps to me. (= CAUSE the creeps GO to me)

Small clauses revisited

We motivated the assumption of VP shells with reference to causative small clauses like (6),
repeated in (64).

(64) God let [ there be light ] .

As we know from Chapter 3, small clauses can also contain predicates headed by syntactic
categories other than V. (65) gives some examples.

(65) a. AP They proved [ the solution completely inadequate ] .

b. DP They consider [ her a friend ].

c. PP They made [ him into a star ] .

Stowell 1983 proposed that all small clauses have a uniform structure, illustrated for (65b) in
(66).

(66)

According to this analysis, the small clause (Aristotelian) predicate (underlined in the examples
above) is an intermediate projection. The entire small clause (in brackets) is the predicate's
maximal projection, and the subject (in italics) is the maximal projection's specifier and the
predicate's sister. Stowell's analysis is attractive because it treats small clauses as structurally
analogous to ordinary clauses. The only difference between the two clause types concerns
whether the clause contains a projection of I. Nevertheless, the analysis cannot be maintained for
DP small clauses because it fails to accommodate the minimal variant of (65b) in (67).

(67) They consider [ her Tanya's friend ].

Here, the DP predicate contains a possessor, which under Stowell's analysis would compete with
the small clause subject for Spec(DP) (Heycock 1991).

In order to maintain binary branching, the structure for examples like (67) must include an
additional head, which we take to be a silent counterpart of the copula be. We give the structure
for (67) in (68a), and our revised structure for (66) in (68b). Notice that both structures preserve
the property of treating small clauses and ordinary clauses as structurally parallel.
(68) a. b.

Based on the parallel acceptability of the DP small clauses in (69) and the AP and PP small
clauses in (70) and (71), we propose to extend the structure in (68) with silent BE to the
remaining small clauses in (65).

(69) a. DP They consider [ Dean the Democratic Party's best hope ] .

b. With [ Dean the Democratic Party's best hope ] , ...

(70) a. AP They consider [ the unemployment figures ominously high ] .

b. With [ the unemployment figures ominously high ] , ...

(71) a. PP They consider [ the patient out of danger ] .

b. With [ the patient out of danger ] , ...

BE is not the only predicate that can head small clauses. (72) illustrates small clauses that are
headed by as, and (73) gives the structure for (72c).

(72) a. They regard [ Dean as the Democratic Party's best hope ] .

b. They regard [ the unemployment figures as ominously high ] .

c. They regard [ the patient as out of danger ] .

(73)
An apparent problem for the analysis just proposed is the fact that let can take VP small clause
complements headed by ordinary verbs, including ordinary be, but not ones headed by silent BE.

(74) a. We let [ Martha be Lukas's buddy ] .

b. * We let [ Martha Lukas's buddy ] .

Conversely, with doesn't allow small clauses headed by ordinary verbs, but does allow ones
headed by silent BE.

(75) a. * With [ Martha be Lukas's buddy ] , ...

b. With [ Martha Lukas's buddy ] , ...

This problem is less serious than it appears, however, since heads are able to subcategorize not
only for the syntactic category of their complements, but to specify that category's head as well.
We know this because of examples like (76), where a head selects not just a PP complement, but
a PP complement headed by a particular preposition.

(76) a. faith { in, *at, *on, *to } your ability

b. rely { on, *at, *in, *to } someone

Notes

1. In traditional grammar, the recipient and theme are taken to be the verb's indirect and direct
object, respectively.

2. The idea underlying the VP shell analysis goes back to Chomsky 1955 and was taken up in
Larson 1988, 1990 (see also Jackendoff 1990). The treatment in this chapter is indebted to that in
Harley 2002, though not identical to it in all details.

3. In addition to marking the grammatical relations like subject or direct object, Japanese also
marks discourse functions such as topic. In Japanese main clauses, topic -wa marking overrides
subject -ga marking. Because of this, it is customary to illustrate -ga marking using subordinate
clauses, as we do in what follows.

4. This statement is not quite correct. A more adequate version is given in (31).
5. The astute reader will observe that GET, like GO, is also associated with the argument array {
theme, location } (as in The receipts got to the post office box). We return to this fact later on in
the chapter.

6. We remain agnostic as to why there is no mapping between the argument array { theme,
location } and the structure in (29). For present purposes, it is sufficient that the mapping does
not exist, even if it could exist in principle. A much more interesting claim, of course, would be
that the structure in (29) itself - and hence the mapping - is ruled out by some grammatical
principle.

7. The alternation in (i) - specifically, the well-formedness of (i.b) - is only an apparent exception
to the statement in the text.

(i) a. Amy sent the mail { back, off } .

b. Amy sent { back, off } the mail.

Back and off are so-called particles, which can behave like ordinary PPs, as in (i.a), but also
more like bound affixes, as in (i.b). A detailed analysis of the syntax of particles is beyond the
scope of this discussion, but evidence for their differing syntactic status in (i) comes from
contrasts as in (ii).

(ii) a. Amy sent the mail right { back, off } . (cf. right to the CEO)

b. * Amy sent right { back, off } the mail.

8. Again, alternations as in (i) are only apparent exceptions to the statement in the text and reflect
the status of on and back as particles; see fn. 7.

(i) a. Amy put her sweater (right) { on, back } .

b. Amy put (*right) { on, back } her sweater .

9. We are being a bit sloppy here. The CAUSE + GET combination that gets spelled out as give
isn't actually completely identical to the one that gets spelled out as get. In other words, there are
slightly different heads GET-1 and GET-2, with give being the spellout of CAUSE + GET-1, and
get the spellout of CAUSE + GET-2. Evidence that the lower heads differ slightly is the fact that
give and get don't have exactly the same distribution (cf. Jerry { gave, *got } Amy a cold).

10. Strictly speaking, the sentences in (55) are grammatical, but only if the themes are interpreted
as entities able to undergo physical transfer. The relevant readings become more salient if the
indefinite article is replaced by the definite article. The idea might then refer to an idea contained
or expressed in a book, and the cold might refer to a cold virus contained in a test tube being sent
from one lab to another.

Exercises and problems

Exercise 7.1

Find five double-object or double-complement verbs not mentioned in the chapter and suggest a
semantic decomposition for them. Are there double-object or double-complement verbs that are not
amenable to the causative decomposition proposed in the chapter?

Exercise 7.2

Make up one short sentence for each of the double-complement verbs give, send, put, and
persuade, and use the grammar tool in x-bar 2 to give complete structures for them.

Exercise 7.3

A. Using the grammar tool in vp shell spines, propose structures for each of the following
euphemisms. Assume that German and Latin are head-final.

die kill

(1) a. German um-kommen (lit. around-come) um-bringen (lit. around-bring)

b. Latin per-ire (lit. through-go) per-dere (lit. through-place)

c. Latin inter-ire (lit. between-go) inter-facere (lit. between-make)

Exercise 7.4

Many of the judgments in this exercise follow straightforwardly given the analysis in the
chapter, but explaining some of the judgments will require you to extend the analysis. In this
connection, it is helpful to read fn. 9.

A. Explain the contrast in (1) (or as much of it as you can).

(1) a. Travis { sent, got } the receipts to the post office box.
 b. * Travis gave the receipts to the post office box.

B. Explain the contrast between (2) and (3) (or as much of it as you can).

(2) Jerry got Amy a present.

(3) a. * Jerry got Amy a cold.

b. * The scandal got the reporter an idea.

c. * Bright lights get Amy a migraine headache.

C. Explain the pattern of judgments in (4) and (5) (or as much of it as you can).

(4) a. The couch got a shove.

b. The movers gave the couch a shove.

c. * The movers got the couch a shove.

(5) a. * A shove got to the couch.

b. * The movers gave a shove to the couch.

c. * The movers got a shove to the couch.

D. Explain the contrast in (6) (or as much of it as you can).

(6) a. Crazy people give me the creeps.

b. * Crazy people get me the creeps.

E. Explain the pattern of judgments in (7) and (8) (or as much of it as you can).

(7) a. The surgeon gave the patient the finger. (ambiguous between literal and idiomatic reading)

b. The surgeon gave the finger to the patient. (unambiguously literal)

(8) a. The surgeon got the patient the finger. (unambiguously literal)
 b. The surgeon got the finger to the patient. (unambiguously literal)

Exercise 7.5

Explain the contrast between (1) and (2).

(1) God let there be light.

(2) a. * They consider there light.

b. * With there light, we can start trekking.

Exercise 7.6

Using the grammar tool in x-bar 2, build structures for (1).

For simplicity, leave the internal structures of noun phrases or adjective phrases unspecified.
The grammar tool for the exercise doesn't allow you to build the internal structures very
easily, and they are irrelevant to the point of the exercise.

(1) a. They kept the president's arrival a secret.

b. They kept the president's arrival very secret.

Exercise 7.7

For each of the trees that you draw for this exercise, include a paraphrase for the interpretation
that the tree represents.

A. Using the grammar tool in x-bar 2, build structures for each interpretation of the following
structurally ambiguous headlines. Unlike in the chapter, give full IPs where necessary.

For simplicity, treat compound nouns (e.g., NBA referees) as simple nouns without internal
structure. Treat the gerund form in (1c) as a simple verb without morphological structure
(growing rather than grow + -ing).

(1) a. Lawyers Give Poor Free Legal Advice

 b. Young makes Zanzibar stop

c. Complaints About NBA Referees Growing Ugly

B. Using the grammar tool in x-bar 2, propose structures for the intended interpretation of (2)
and for a structurally possible (but let us hope unintended!) cannibalistic interpretation.

(2) "I want to make you my favorite sandwich."

(Holly Hughes. 2003. Best food writing 2003. New York: Marlowe. 167.)

C. Using the grammar tool in x-bar 2, propose structures for the two salient interpretations of the
punchline in (3) (the customer reading and the Zen reading). For simplicity (contrary to the
solution for Exercise 5.9, (1d)), you can treat the imperative clause as a bare VP.

(3) Q. What did the Zen master say to the guy at the hot dog stand?

A. Make me one with everything.

Exercise 7.8

A. Off the top of your head, propose an elementary tree for ago. Does the elementary tree differ from
other elementary trees of the same syntactic category?

You won't necessarily be able to build the tree you want with the grammar tool for this
chapter.

B. Now look up ago and its etymology in the Oxford English Dictionary. Using the grammar
tool in x-bar 2, build the elementary tree for ago that is consistent with the etymology that you
find.

C. This part of the exercise is not closely related to the material covered in this chapter, but you
will need the results to complete (D). Is the syntactic category of the quantifier many D or (say)
Adj? Give the evidence on which you base your answer.

Hint: Reread the discussion of two tymes to-geder in the solution to Exercise 6.1.

Don't give an answer based on the meaning of many. The meaning won't decide this question
for you, since some quantifiers, like no or some, are determiners, and others, like few or two, are
not.
D. On the basis of your results from (B) and (C), use the grammar tool in x-bar 2 to give the
structure for the sentence in (1).

(1) Mark's family lived there many years ago.

Problem 7.1

There seem to be no ditransitive nouns, adjectives or prepositions. Is this an accident?

Problem 7.2

For some speakers, the second clause in (1) contradicts the first. For others, (1) is semantically coherent.

(1) They sent a rocket to Uranus, but it never arrived.

An apparently unrelated fact is that, for some speakers, (2a) entails that the students learned
syntax, whereas (2b) doesn't have that entailment. For other speakers, the sentences in (2) are
synonymous.

(2) a. The instructor taught the students syntax.

b. The instructor taught syntax to the students.

Can you suggest a (single) explanation for these judgments concerning send and teach?

8 Case theory
Old version (Fall 2006) - new version coming soon

A first look at case

o The basic purpose of case
o Case government
o Synthetic versus analytic case marking
Case features
Case licensing
o Spec-head licensing
o Head-spec licensing
o Head-comp licensing
o Nonstructural conditions on case licensing
o The dative-accusative distinction
Case agreement (coming eventually...)
Notes
Exercises and problems
Supplementary material
 o Grammatical relations

This chapter is devoted to a discussion of case, a morphosyntactic property of noun phrases. We

begin by illustrating the basic purpose of case, which is to identify a noun phrase's function or
grammatical relation in the sentence (for instance, whether a noun phrase is a subject or object).
We also show that particular lexical items can impose morphological case requirements on noun
phrases, a phenomenon known as case government. We then turn to how case is expressed
across languages, focusing on various Indo-European languages (the language family to which
English belongs). Universal Grammar allows case, just like tense, to be expressed either
synthetically (as suffixes on nouns) or analytically (by means of prepositions or other syntactic
heads that take an entire noun phrase as their argument). As we will see, English allows both
ways of expressing case (just as it allows both ways of expressing tense in watch-ed and will
watch). It is possible to describe both expressions of case in a unitary way by treating case as a
feature on a noun phrase that is checked by a head. As we will show, case checking is subject to
structural as well as nonstructural licensing conditions.

A first look at case

The basic purpose of case

In order to understand the purpose of case in human language, it is useful to consider languages
in which constituent order is not as fixed as it is in English. In German, for instance, unlike
English, the subject of an ordinary declarative clause needn't precede the verb, as shown in (1)
and (2) (we discuss the structure of German sentences in more detail in a later chapter; for now,
only the variable constituent order is of interest). In the examples, boldface indicates the subject,
and italics indicates the object.
(1) a. German Der Mann sieht den Hund.
the man sees the dog
'The man sees the dog.'
b. Den Hund sieht der Mann.
the dog sees the man
same as (1a), not the same as (2a)
(2) a. Der Hund sieht den Mann.
the dog sees the man
'The dog sees the man.'
b. Den Mann sieht der Hund.
the man sees the dog
same as (2a), not the same as (1a)

Since German speakers can't rely on constituent order to identify subjects and objects, how is it
possible for them to keep track of which constituent expresses which grammatical relation? The
answer is that grammatical relations are encoded in German in terms of morphological case
marking. In particular, the subjects of finite clauses in German appear in a particular form called
the nominative case, whereas objects generally appear in the accusative. (3) gives a
morphological analysis of the noun phrases in (1) and (2).

(3) a. d- er Mann, d- er Hund

the nom man the nom dog
 b. d- en Mann, d- en Hund
the acc man the acc dog

Notice that in (3), the distinction between nominative and accusative case is marked once: on the
head of the noun phrase (the determiner).

In certain exceptional cases in German, case distinctions are marked redundantly: on the
determiner as well on the noun. This is illustrated in (4). ∅ indicates a zero nominative suffix; -
(en) is the optional accusative suffix.

(4) a. Nominative d- er Bär- ∅, d- er Student-∅

the nom bear nom, the nom student nom
b. Accusative d-en Bär-(en), d-en Student-(en)
the acc bear acc the acc student acc

The redundant case marking in (4) is a historical relic from an earlier stage of German where this
pattern was more extensive. In certain languages, redundant case marking on the determiner and
the noun is the norm. This is illustrated for modern Greek in (5).

(5) a. Modern Greek O andr-as vlepi t- o skil-o.

the.nom man nom sees the acc dog acc
'The man sees the dog.'
b. O skil-os vlepi t- on andr-a.
the.nom dog nom sees the acc man acc
'The dog sees the man.'

Finally, case can be marked solely on the noun. This is illustrated in (6) for Latin, a language
without articles.

(6) a. Latin Av- us can-em videt.

grandfather nom dog acc sees
'The grandfather sees the dog.'
b. Can-is av- um videt.
dog nom grandfather acc sees
'The dog sees the grandfather.'

To summarize the discussion in this section: noun phrases can be case-marked either on the
determiner, or on the noun, or redundantly on both. But regardless of the particular pattern, case
marking has the same basic purpose: it visibly expresses a noun phrase's function in a sentence.

Case government

In many languages, a noun phrase's particular morphological case depends not only on its
function in the entire sentence, but also on which particular lexical item it is most closely
associated with. For instance, in German, the object in a sentence appears in the dative or the
accusative,1 depending on the verb, as illustrated in (7) and (8).
(7) a. Dative ok { d- em Hund, d- er Frau } helfen
the dat dog the dat woman help
 'to help the { dog, woman }'
b. Accusative * { d- en Hund, d- ie Frau } helfen
the acc dog the acc woman help
(8) a. Accusative ok { d- en Hund, d- ie Frau } unterstützen
the acc dog the acc woman support
'to support the { dog, woman }'
b. Dative * { d- em Hund, d- er Frau } unterstützen
the dat dog the dat woman support

In traditional grammar, the verb is said to govern the case of the object. For instance, helfen
'help' governs the dative, unterstützen 'support' governs the accusative, and so on. An attractive
hypothesis is that the morphological case that a verb governs correlates with the verb's meaning,
the idea being that variation in case government as illustrated in (7) and (8) correlates with
(possibly subtle) differences in the semantics of helfen 'help' and unterstützen 'support'. One idea
that comes to mind, for instance, is that unterstützen 'support' is a simple transitive verb, whereas
helfen reflects the spellout of a VP shell CAUSE someone GET help. Although we will not work
out this idea in full in this chapter, we present some related considerations concerning dative and
accusative case-marking in VP shells later on in the chapter.

Case government in Latin is illustrated in (9). As in German, each particular verb governs the
case of its object, but in Latin, the choice of case ranges over three cases - dative, accusative, and
ablative.

(9) a. Dative { femin-ae, *femin-am, *femin-a } { sub- venire, suc-

currere }
woman dat acc abl under-come under-run
'to help the woman'
b. Accusative { femin-am, *femin-ae, *femin-a } ad-iuvare
woman acc dat abl to-support
'to support the woman'
c. Ablative { femin-a, *femin-ae, *femin-am } frui
woman abl dat acc enjoy
'to enjoy the company of the woman'

In both German and Latin, prepositions resemble verbs in governing the case of their
complement. In German, prepositions govern the accusative, the dative, or (rarely) the genitive;
in Latin, they govern the ablative or the accusative.

(10) a. German durch d- ie Tür, bei d- er Kirche, während d- es

Krieges
through the acc door by the dat church during the gen war
'through the door, by/near the church, during the war'
b. Latin de sapienti-a, ad rip- am
about wisdom abl to shore acc
'about wisdom, to the shore'

Finally, in both German and Latin, certain prepositions can govern more than one case. In such
cases, the accusative marks direction, and the other case (dative in German, ablative in Latin)
marks location.
(11) a. German in { d- ie, *d- er } Bibliothek schicken; in { d- er, *d- ie
} Bibliothek arbeiten
in the acc the dat library send in the dat the
acc library work
'to send into the library, to work in the library'
Latin in { bibliothec-am, *bibliothec-a } mittere; in { bibliothec-a,
*bibliothec-am } laborare
in library acc abl send in library abl
acc work
'to send into the library, to work in the library'

Synthetic versus analytic case marking

In the languages that we have been discussing so far, case is expressed synthetically, by means of
morphologically complex words. But Universal Grammar also allows noun phrases to be marked
for case analytically. The case marker is then not an affix, but a relatively independent syntactic
head. We illustrate these two options of expressing case in connection with a brief overview of
case in the Indo-European language family, to which English belongs.

Proto-Indo-European (PIE), the reconstructed ancestor of the Indo-European language family

(which includes English) which was spoken thousands of years ago, had eight cases, which were
expressed synthetically. The nominative marked the subject of finite clauses, the accusative and
dative (and perhaps other cases) marked objects (depending on the verb, as just discussed), and
the genitive indicated possession. The PIE ablative indicated the source of movement (as in I
drove from Chicago), the locative was used for locations (as in I used to live in Chicago), and the
instrumental marked instruments or means (as in He cut it with his pocketknife). Finally, the
vocative was used to address persons (as in Hey, Tom, come on over here).

The original PIE case system is essentially preserved in Sanskrit, although the distinction
between the ablative and the genitive is somewhat obscured because ablative and genitive forms
were often homophonous in Sanskrit. Such homophony among two or more case forms is called
case syncretism. Among living languages, the PIE system is best preserved in the Baltic
languages (Latvian and Lithuanian) and some Slavic languages (for instance, Czech and
Ukrainian). In these languages, the genitive and the ablative have merged completely, leaving
seven cases. In other words, in the history of these languages, case syncretism affected all forms
of the genitive and the ablative, not just some of them, and so children learning the language no
longer had any evidence anywhere in the language for distinguishing between the two cases.
Several other Slavic languages, including Russian, have in addition almost completely lost the
vocative, leaving six cases. In Latin, the PIE ablative, instrumental, and locative merged into a
single case, called the ablative, which serves all three functions, also leaving six cases. In
Ancient Greek, the ablative, instrumental, and locative were lost, leaving five cases. Old English
had five cases as well, having lost the ablative, locative, and vocative; in addition, the
instrumental had mostly merged with the dative. Another Germanic language, modern German,
retains four cases: nominative, dative, accusative, and an increasingly moribund genitive. The
developments just sketched for Indo-European are summarized in (12), where "R" and "---"
indicate retention and loss, respectively.
 PIE, Baltic, Old
Other Ancien Germa
Sanskri some Latin Englis
Slavic t Greek n
t Slavic h

Nominative R R R R R R R

Dative R R R R R R R

Accusative R R R R R R R

Genitive R merged merged R R R R

as as
(12 Ablative R genitiv genitiv --- --- ---
) e e
merged
Locative R R R as --- --- ---
ablativ mostly
Instrumenta e merged
R R R --- ---
l with
dative

Vocative R R --- R R --- ---

Number of
8 7 6 5 4
distinct cases

(13) shows the complete case paradigms for the Latin nouns femina 'woman' and avus
'grandfather'. These two nouns are each representative of two distinct declensions, or word
classes. Latin had a total of five such word classes, each of which was characterized by unique
endings for combinations of case and number. For instance, dative singular is marked by -ae on
femina and by -o on avus. In the remaining three declensions, the same combination happens to
be marked by the same suffix, namely -i (distinguishing three remaining declensions, rather than
collapsing them into one, is motivated by other distinctions in the paradigms). For more details,
take a look at Allen and Greenough's New Latin Grammar, available through the Perseus project.

(13) a- declension o- declension

Latin 'woman' 'grandfather'
Sg Pl Sg Pl

Nominative femin-a femin-ae av-us av-i

Genitive femin-ae femin-arum av-i av-orum
Dative femin-ae femin-is av-o av-is
Accusative femin-am femin-as av-um av-os
Vocative femin-a femin-ae av-e av-i
 Ablative femin-a femin-is av-o av-is

As (13) shows, Latin exhibited some case syncretism. For instance, the genitive and the dative
singular are homophonous for femina 'woman', the dative and ablative singular are homophonous
for avus 'grandfather', and the dative and the ablative plural are homophonous for both nouns.

In the descendants of Latin, the Romance languages, case continues to be expressed synthetically
on pronouns. For instance, the distinction between dative and accusative pronouns is illustrated
for French in (14). (Note that unstressed pronouns in French are clitics; unlike full noun phrases,
they precede the verb they are construed with.)

(14) a. Je veux leur parler.

I want 3.pl.dat talk
'I want to talk to them.'
b. Je veux les voir.
I want 3.pl.acc see
'I want to see them.'

With full noun phrases, however, the same distinction is expressed analytically by the presence
or absence of the case marker à.

(15) a. Je veux parler à vos voisins.

I want talk your neighbors
'I want to talk to your neighbors'
b. Je veux voir vos voisins.
I want see your neighbors
'I want to see your neighbors.'

The case marker à is etymologically related to the spatial preposition à 'to', but is distinct from it.
This is demonstrated by the fact that the pro-form for phrases in which à is a spatial preposition
is not leur (or lui in the singular), as in (14a), but y, just as it is for other spatial prepositions like
dans 'in' or sur 'on'.

(16) a. Nous avons envoyé le vin à Toulouse; mon ami habite à Paris.
we have sent the wine to Toulouse my friend lives in Paris
'We sent the wine to Toulouse; my friend lives in Paris.'
b. Nous y avons envoyé le vin; mon ami y habite.
we there have sent the wine my friend there lives
'We sent the wine there; my friend lives there.'
(17) a. Le cadeau se trouve dans mon sac; nous avons mis le cadeau sur
la table.
the present refl finds in my bag we have put the present on
the table
'The present is (literally, finds itself) in the bag; we put the
present on the table.'
b. Le cadeau s' y trouve; nous y avons mis le cadeau.
the present refl there finds we there have put the present
'The present is there; we put the present there.'
As mentioned earlier, Old English had five cases, which are illustrated in (18) for three
declensions. As is evident, case syncretism is more extensive in Old English than in Latin.

Masculine Feminine Neuter

(18)
Old English 'fox' 'learning' 'animal'
Sg Pl Sg Pl Sg Pl

Nominative fox fox-as lar lar-a deor deor2

Genitive fox-es fox-a lar-e lar-a deor-es deor-a
Dative fox-e fox-um lar-e lar-um deor-e deor-um
Instrumental fox-e fox-um lar-e lar-um deor-e deor-um
Accusative fox fox-as lar-e lar-a deor deor

In the course of Middle English (1150-1500), the old genitive case suffixes were lost, and their
function was taken over by a syntactic head - the possessive determiner 's (in the plural, the
possessive is spelled out as a silent determiner that is orthographically represented as an
apostrophe). The old synthetic genitive case is illustrated in (19). Recall that the thorn character
(þ) corresponds to modern English 'th'.

(19) þe king-es suster of France (cmpeterb, 59.593)

the king-gen sister of France
'the king of France's sister'

Although the change itself is not yet fully understood, it is clear that the modern possessive
marker is no longer a synthetic case suffix on a noun (N) (king), but rather analytically case-
marks an entire noun phrase (DP) (the king of France). This is clear from the fact that it follows
postnominal material like the prepositional phrase of France in the translation of (19). The
difference between the old synthetic genitive suffix and the analytical possessive determiner that
replaced it emerges even more sharply from the contrast in (20), where the possessive determiner
obligatorily follows an element that is not even a noun. For clarity, the noun phrase that is case-
marked by the possessive determiner is underlined in (20b); the entire sequence in (20b) from the
to cat is of course also a noun phrase.

(20) a. * the guy's that I used to go out with cat

b. the guy that I used to go out with 's cat
We ordinarily think of the possessive form of singular noun phrases as containing 's.
Under the analysis just given, however, the nominative, possessive, and objective case of
a full noun phrase are all homophonous in Modern English, and the determiner 's in the
king's is a case marker on a par with the preposition of in of the king.

Although the possessive is marked analytically on full noun phrases, it continues to be spelled
out synthetically on pronouns (recall the similar analytic/synthetic split between full noun
phrases and pronouns in French). Much as the combination of a verb like sing and a silent past
tense morpheme is spelled out as sang, a pronoun like we (or more precisely, the feature
combination first person plural) and possessive case is spelled out as our.

Beginning in late Old English (ca. 1000 C.E.), the distinction between the dative and the
accusative weakened, and the distinction was lost completely in the course of Middle English
(1150-1500). In what follows, we will refer to the case that resulted from the merger as the
objective. The distinction between nominative and objective case continues to be expressed
synthetically in modern English on most ordinary pronouns, as illustrated in (21).

(21) Nominative Objective

1 sg I me
2 sg, pl you you
3 sg m, f, n he, she, it him, her, it
1 pl we us
3 pl they them

As the table shows, with the two pronouns you and it, the distinction between the nominative and
the objective has been lost, and this is also true for full noun phrases. Finally, it is worth noting
that despite the efforts of prescriptive grammarians to keep a distinction alive between
nominative who and objective whom, the two forms have merged as who. James Thurber has a
diabolically witty essay on the topic.

Case features

In this section, we introduce some concepts and syntactic conditions that enable us to derive the
distribution of the various case forms of noun phrases in English and other languages. We begin
by introducing the notion of case feature.

Consider the contrast between (22) and (23).

(22) a. ok They will help her.

b. ok She will help them.
(23) a. * Them will help she.
b. * Her will help they.

Why are the sentences in (23) ungrammatical? The answer is that noun phrases in English are
subject to the requirements in (24).

(24) a. Subjects of finite clauses appear in the nominative.

b. Objects appear in the objective.
As is evident, both of the subjects in (23) are objective forms, and both of the objects are
nominative forms. Each of the sentences in (23) therefore contradicts the requirements in (24) in
two ways.

Now compare the examples in (22) and (23) with those in (25).

(25) a. You will help her.

b. She will help you.

As we saw in (21), they and she exhibit distinct forms for the nominative and objective, whereas
you doesn't. But because case syncretism between the nominative and the objective is not
complete in English (in other words, because at least some pronouns still have distinct forms for
the two cases), we will treat you as a nominative form in (25a), equivalent to they and she, but as
an objective form in (25b), equivalent to them and her. For the same reason, we treat the noun
phrase my big brother as a nominative form in (26a) and as an objective form in (26b).

(26) a. My big brother will help her.

b. She will help my big brother.

In order to disambiguate instances of case syncretism like you and my big brother, it is useful to
associate each noun phrase in a language with a case feature. Each case feature has a value that
is selected from among all the various case forms in that language (regardless of whether the
case forms are expressed synthetically or analytically). In English, for instance, a case feature
can assume the value "nominative", "possessive," or "objective". In Russian, a case feature has a
choice among six values (nominative, genitive, dative, accusative, locative, instrumental). If we
need to represent a noun phrase's case feature, we can do so by means of labels as in (27) and
(28).

(27) a. [DP-nom They ] will help [DP-obj her. ]

b. [DP-nom You ] will help [DP-obj her. ]
c. [DP-nom My big brother ] will help [DP-obj her. ]
(28) a. [DP-nom She ] will help [DP-obj them. ]
b. [DP-nom She ] will help [DP-obj you. ]
c. [DP-nom She ] will help [DP-obj my big brother. ]

Case licensing

Earlier, we said that the purpose of case is to encode a noun phrase's function in the sentence. In
order to make the notion of function more precise, we can think of each noun phrase in a
sentence as being licensed by (= linked to) some syntactic head. A common way of putting this is
to say that the case feature on a noun phrase needs to be checked against a corresponding case
feature on the case-licensing head. In English, case-licensers must be either verbs or
prepositions, but there are languages that allow adjectives and nouns to be case-licensers as well.
If the case features on the two participants in a checking relationship don't match up (say, one is
nominative and the other is accusative) or if they don't stand in a one-to-one relationship (say,
the case feature on a head ends up checking case features on more than one noun phrase), then
the sentence is ungrammatical. On the other hand, if every case feature in a sentence stands in a
proper relationship with a matching partner, then all is well with the sentence as far as case
theory is concerned. A question that immediately comes to a syntactician's mind is whether case
checking is subject to structural constraints. If so, we are of course interested in providing as
general a formulation of those constraints as possible.

There is reason to believe that there is more than one type of case checking. We can distinguish
between case licensing, which holds between a noun phrase and a head external to the noun
phrase (say, a verb or preposition), and case agreement, which holds within a noun phrase (say,
between a determiner and a noun). In the current version of this book, we will discuss only case
licensing. In this section, we motivate various conditions (primarily structural, but also
nonstructural) on the relationship between the two participants in a case-licensing relation. In the
first half of the section, we present three structural configurations in which case licensing is
possible: the specifier-head configuration, the head-specifier configuration, and the head-
complement configuration. Beginning in the 1990's, attempts have been made to simplify the
theory of case licensing by identifying a single case-licensing configuration. For instance, it has
been proposed that complements of verbs are not directly licensed in the head-complement
configuration, but that the complement moves to the specifier of a silent head, and that case is
uniformly licensed in the specifier-head configuration. The following discussion will remain
somewhat agnostic on this point. However, we will show that all three of the configurations
mentioned above are almost identical from a topological point of view. In the second half of the
section, we discuss three further nonstructural conditions on case licensing: biuniqueness,
exocentricity, and matching.

Spec-head licensing

In what follows, it's important to distinguish carefully between finite clauses on the one
hand and finite verbs on the other. In English, finite clauses are clauses that can stand on
their own. The clauses in (i)-(iii) are finite; the ones in (iv) are not.
(i) Finite clause Finite tense + finite verb I [pres] do that; he [pres] does
that;
I [past] did that; he [past] did
that.
(ii) a. Finite clause Finite tense + finite auxiliary + I [pres] am doing that; he [pres]
nonfinite verb is doing that;
(present participle) I [past] was doing that; he [past]
was doing that.
b. Finite clause Finite tense + finite auxiliary + I [pres] have done that; he [pres]
nonfinite verb has done that;
(past participle) I [past] had done that; he [past]
had done that.
(iii) Finite clause Modal + nonfinite verb I will do that; he will do that.
 (infinitive)
(iv) Nonfinite Nonfinite verb, no finite to do that; to be doing that; to
clause auxiliary or modal have done that

Finite verbs are ones that aren't participles or infinitives (see Verb forms and finiteness in
English for details). A finite clause always contains some finite Infl element, either a
finite tense morpheme (i, ii) or a modal (iii). A finite tense morpheme in turn is always
associated with a finite verb (i) or a finite auxiliary (ii). A modal, on the other hand, is
always associated with an infinitive.

From this it follows that if a clause contains a finite verb or a finite auxiliary, the clause
itself is finite. But if a clause contains a nonfinite verb, it needn't itself be nonfinite. If it
contains a modal, it is finite (iii); only if it doesn't is it nonfinite (iv).

We begin by considering how case is licensed on the subjects of sentences. Since subjects of
sentences start out life as specifiers of verbs, one's first impulse might be to propose that
nominative case is checked by V. Although we will end up rejecting this approach, let us pursue
it for the moment in order to show why it is unsatisfactory. The proposal is that what checks the
nominative case of He (or more precisely, its trace in Spec(VP)) is the finite verb understands in
(29a) and the bare (nonfinite) form understand in (29b). This putative checking relationship is
indicated by the red boxes. (We further assume that DPs whose case feature is checked are free
to move on to other positions in the sentence.)

(29) a. b.

Finite clause, finite verb form Finite clause, nonfinite verb form
(to be revised) (to be revised)

Now if verbs were able to check nominative case, regardless of whether they are finite or
nonfinite, we would expect the nonfinite verb in the lower IP in (30) to be able to check
nominative case on the lower he, on a par with the nonfinite verb in (29b).3
(30) *

Intended meaning: He claims that he understands Hegel.

However, (30) is completely ungrammatical. We therefore reject the idea that nominative case is
checked by V. We conclude instead that it is checked by finite I. The contrast between (29) and
(30) then follows directly since I is finite in (29) ([pres], does), but not in (30) (to).

Notice, by the way, that the ungrammaticality of (30) isn't due to semantic anomaly, since the
intended meaning is both expressible and semantically well-formed, as indicated by the gloss to
(30). Neither is the ungrammaticality of (30) due to the split infinitive, since (31) is as
ungrammatical as (30).

(31) *

One might attempt to rescue the idea that nominative case is checked by finite V by replacing
(32a) with (32b).

(32) a. Nominative case is checked by finite I.

 b. Nominative case is licensed by finite V where possible (that is, in clauses that contain a
finite V), and by finite I otherwise.

Although there is no empirical argument against (32b), we reject it because it violates conceptual
economy. Our reasoning is as follows. A finite V in a clause implies a finite I (in the form of a
silent tense morpheme). The converse is not true, however. Although a finite I in a clause is
consistent with a finite V, as just stated, it is also consistent with a nonfinite V (the finite I might
be a modal). Clauses with finite I thus form a proper superset of clauses with finite V. This
means that (32a) and (32b) are empirically equivalent. However, the statement in (32b) is
unnecessarily more cumbersome and therefore less preferable.

The upshot of the discussion so far is that the head that checks nominative case in English is
finite I, and that the licensing configuration for checking nominative case in English is the
specifier-head configuration. This is shown in (33) (which supersedes (29)).4 The term
'specifier' is generally abbreviated to 'spec' (read as 'speck').

(33) a. b.

Finite clause, finite verb form Finite clause, nonfinite verb form
(final) (final)

Nominative case is not the only case to be licensed in the spec-head configuration in English. So
is possessive case. Here, the case-checking head is the possessive determiner ('s or its silent
plural variant), as discussed earlier.

In possessive constructions like (i), there are two noun phrases: a lower one (the
possessor) and a higher one (the entire noun phrase that contains both the possessor and
the thing possessed).

(i) (ii)

It is important to keep in mind that each of the two noun phrases has a case feature of its
 own that needs to be checked. The lower DP has a possessive case feature. The higher DP
generally has a nominative or an objective case feature, but it might itself bear a
possessive case feature if it is part of an even larger possessive construction, as it is in (ii).

(34) illustrates the spec-head configuration in its general form. The nodes that bear the case
features that need to be checked are the head X and its specifier YP. The path between the two
nodes is indicated in red; we return to some properties of this path in connection with the two
remaining case-licensing configurations that we discuss (head-spec, head-complement).

(34)

Head-spec licensing

A second configuration that licenses case checking is head-spec licensing, which we motivate
on the basis of sentences like (35a).
(35) a. He expected her to dislike him.
b. He expected that she would dislike him.

In both sentences, what is expected is a state of affairs (= a proposition). Given the semantic
parallel between the two sentences, it is reasonable to suppose that expect in (35a) takes a single
complement (the entire italicized sequence her to dislike him), rather than a sequence of two
complements (the DP her and some constituent dominating to dislike him). Assuming that to is a
nonfinite structural counterpart of finite would leads us to give (35a) the structure in (36).

(36)

An additional reason for treating the noun phrase following expect as the subject of a
complement clause rather than as the object of the matrix clause concerns sentences containing
expletive there. Recall from Chapter 3 that expletive there is licensed as the subject of a clause
containing a verb of (coming into) existence. If we treat the DP immediately following expect as
a subject, the parallel between (37a) and (37b) is expected and straightforward (as is the parallel
between (37) and (35)).

(37) a. He expected [IP there to be a fly in his soup ] .

b. He expected that [IP there would be a fly in his soup ] .

On the other hand, if we were to treat the postverbal DP as an object, we would have to
complicate our statement of how expletive there is licensed. Moreover, even if we succeeded in
formulating a descriptively adequate licensing condition, we would still forfeit the structural
parallel between (37a) and (37b).

Having motivated the structure in (36), let's now return to our main concern: how objective case
is licensed on the embedded subjects in (35a) and (37a). Consider the schemas in (38).

(38) a. b.

Spec-head licensing Head-spec licensing

Notice that the head-spec configuration in (38b) is the mirror image of the spec-head
configuration in (38a), already familiar from (34), in the following sense. In both cases, the case-
licensing configuration can be characterized as in (39).

(39) A case-licensing configuration is defined as follows:

a. a head X
b. the nonterminal node closest to X (i.e., the intermediate projection X')
c. a node closest to X' that is distinct from X
d. the specifier of the node in (c)

The difference between spec-head and head-spec licensing simply concerns the direction that the
path takes in (39c). Spec-head licensing chooses the mother of the head's intermediate
projection; head-spec licensing chooses the daughter.

Given (39), we can say that objective case is checked on the complement subject in an ECM
construction by the matrix verb in the head-spec configuration.

It is standard to refer to the construction in (35a) and (37a) as the Exceptional Case-
Marking (ECM) construction. Given the analysis that we have just presented, the term is a
 bit of a misnomer. If the construction is indeed exceptional, it is not for structural reasons,
but because of the crosslinguistic rarity of heads that take IP complements and are also
able to check objective case. Because the term is prevalent in the literature, we will
continue to use 'ECM construction' to refer to the construction in question and 'ECM verb'
to refer to any verb with the two properties just mentioned (takes IP complement, able to
check objective case).

ECM constructions are not the only ones where case is checked in a head-spec configuration.
The same configuration is also relevant for the constructions discussed in Chapter 7, VP shells
and small clauses. In a language like English, which does not distinguish between a dative and an
accusative case, but has only a single objective case, case checking proceeds along exactly the
same lines as described above. In (40a) (= (3) of Chapter 7), for instance, the head of the higher
VP checks objective case on the specifier of the lower VP. In (40b) (= (7a) of Chapter 7), let
checks objective case on the small clause subject there.

(40) a. b.

In languages with a dative-accusative distinction, case checking in VP shells and small clauses is
a bit more involved than in English, and we therefore defer discussion of these constructions in
these languages until the end of the chapter.

In concluding our discussion of the head-spec configuration, let us briefly return to nominative
case checking in English. In the previous section, we argued that nominative case is licensed in
Spec(IP) by the spec-head configuration. If this is so, then subject movement in English can be
derived from considerations of case checking. In other words, the subject must move from
Spec(VP) to Spec(IP) because nominative case can't be checked in its original position.
However, the availability of head-spec licensing opens up the alternative that nominative case is
checked in the head-spec configuration. The case-checking head continues to be finite I, for the
reasons discussed earlier. If this possibility is correct, then subject movement in English must be
derived from considerations other than case theory, such as predication. Given the word order
facts of English, it is very difficult to determine which of the two possibilities just outlined is
correct. Currently, many generative syntacticians take the (somewhat odd) position that
nominative case is checked in the spec-head configuration, but that subject movement is
motivated by considerations of predication.

Head-comp licensing

A third and final case licensing configuration arises in connection with simple transitive
sentences like (41).
(41) a. He expected her.

b.

Here, objective case on her is checked by the verb expected in the head-complement
configuration, schematically indicated in its general form in (42).

(42)

Notice that the head-complement configuration is a subconfiguration of the head-spec

configuration just discussed. This means that a general structural constraint on case licensing,
subsuming all three configurations discussed so far, can be formulated as in (43).

(43) Structural licensing condition:

The nodes bearing the case features in a case-checking relationship as well as the nodes
on the path connecting them must all be a (not necessarily proper) subset of the set of
nodes in (39).

The head in a case-licensing relationship always corresponds to the node specified in (39a). The
noun phrase corresponds to either (39c) (head-comp licensing) or (39d) (spec-head licensing,
head-spec licensing).

Nonstructural conditions

In what follows, we further illustrate the structural licensing condition on case checking in (43),
and we introduce three additional, nonstructural conditions on case-licensing: biuniqueness,
exocentricity, and matching.

First, consider (44), where we treat their as the spellout of they and possessive 's.

(44) a. He expected their approval.

 b.

In (44), objective case on the higher boxed DP is checked by the verb expected, being licensed
by the head-comp relation between them. Possessive case on the lower DP is checked by the
possessive morpheme 's, being licensed by the spec-head relation between them. So far, so good.

However, a question that arises in connection with the structure in (44) is what rules out (45)
(with the same intended meaning as (44)), where the objective case feature on expected checks
the objective case feature on them in the head-spec licensing configuration.

(45) * He expected them approval.

b.

The answer is as follows. Assume the case-checking relationship between expected and the lower
boxed DP them. This leaves the higher DP with a case feature that must be checked. In principle,
expected might check the case feature on the higher DP in the head-complement configuration,
but then a single case feature (the one on expected) would then be checking more than one case
feature in the rest of the sentence. Conversely, any case feature on the silent determiner would
not get to participate in case-checking. Because (45) is ungrammatical, we conclude that case-
checking is subject to a condition as in (46).

(46) Biuniqueness condition:

Case features on heads and noun phrases stand in a one-to-one relationship.
Is there any head other than expected that the higher DP in (45) could enter into a case-licensing
relationship with? The only head that is close enough is the higher DP's own head, the silent
determiner. In particular, just as the head-spec configuration is the mirror image of the spec-head
configuration, so the relation between the higher DP and its head would correspond to the mirror
image of the head-complement relationship.5 However, assuming a case-licensing relationship
between a phrase and its own head is not sensible given that the purpose of case is to signal the
relationship between a noun phrase and the rest of the sentence. In other words, we will impose a
further condition on case licensing along the lines of (47).

(47) Exocentricity condition:

Case licensing is a relationship between a head and an 'outside' noun phrase (that is, a
noun phrase distinct from any projection of the case-checking head).

For completeness, let us note that (45) is impossible even if we were to assume that the silent D
bears no case feature. This is because D heads that do not bear case features (like a or the in
English) don't license specifiers. In other words, there are no elementary trees of the form in
(48), which would be needed to derive the higher DP in (45b).

(48)

An important joint consequence of the biuniqueness and exocentricity conditions is given in (49).

(49) Minimality condition on case licensing

When a case-checking head has the possibility in principle of entering into a case-
licensing relation with either of two noun phrases, it is the minimal configuration (the one
involving the shorter path) that is the grammatical one.

The third and final nonstructural condition on case licensing is one already mentioned informally
at the very beginning of our discussion of case licensing. For ease of reference, we now give it a
name.

(50) Matching condition:

A case feature on a head and the corresponding case feature on a noun phrase must match
in value.

In the remainder of this section, we illustrate the interplay of the various conditions that we have
proposed, both structural and nonstructural, with reference to the German examples in (51). The
verb kennen 'know' governs the accusative, and the preposition mit 'with' governs the dative.
(Unbelievable as it may seem, German speakers, including children learning the language, really
do pay attention to the tiny difference between dem and den, and have been doing so for
centuries!)
(51) a. d- en Mann mit d- em Hut kennen
the acc man with the dat hat know
'to know the man with the hat'
b. * d- em Mann mit d- en Hut kennen
the dat man with the acc hat know
c. * d- en Mann mit d- en Hut kennen
the acc man with the acc hat know
d. * d- em Mann mit d- em Hut kennen
the dat man with the dat hat know

The schematic structure for all four verb phrases is given in (52) (recall from Chapter 4 that
verbs are head-final in German, whereas (most) prepositions are head-initial).

(52)

In (51a), kennen checks accusative case with the higher DP, and mit checks dative case with the
lower DP, each in the head-comp licensing configuration. In other words, each head checks the
case feature of the DP closest to it.

(51b) is ruled out because it violates the matching condition in (50). Specifically, even though
kennen and the higher DP would stand in a legitimate licensing configuration (head-comp), the
accusative case feature of kennen doesn't match the dative case feature on the higher DP.
Moreover, the accusative case feature of kennen is unable to check the matching accusative case
feature on the lower DP, because the verb and the lower DP are too far apart (a checking
relationship between these two nodes would violate the structural licensing condition in (43)).
Analogous considerations hold for mit and its potential checking relationships with the lower and
higher DPs, respectively.

(51c) is ruled out as follows. Case checking on the higher DP is unproblematic; accusative case
is checked by kennen in the head-comp configuration. However, case cannot be checked on the
lower DP. Checking accusative case with mit in the head-comp configuration would violate the
matching condition (dative and accusative don't match), and checking accusative case on the
lower DP with kennen would violate both the structural licensing condition and the biuniqueness
condition on case checking. (51d) is ruled out for analogous reasons.

The dative-accusative distinction

In this section, as promised, we return to the issue of case checking in languages that, unlike
English, distinguish dative and accusative case.

(53) gives a double-complement sentence in German.6

(53) dass ich den Roman an den Jungen schicke

that I.nom the.acc novel to the.acc boy send
'that I am sending the novel to the boy'

(54) gives the structure for (53); for completeness, we give the structures both before and after
GO moves to CAUSE.

(54) a. b.

From (53), we conclude that CAUSE checks accusative case, and we would therefore expect the
recipient in the double-object counterpart of (53) to appear in the accusative case as well. But
(55) shows that the recipient must instead appear in the dative case.

(55) a. dass ich dem Jungen den Roman schicke

that I.nom the.dat boy the.acc novel send
'that I am sending the boy the novel'
b. * dass ich den Jungen den Roman schicke
that I.nom the.acc boy the.acc novel send

The structure for (55) is shown in (56); once again, we give both pre- and post-movement
structures.
(56) a. b.

The ungrammaticality of (55b) is reminiscent of the double o constraint of Japanese mentioned

in Chapter 7; recall the contrast in (57).

(57) Hanako-ga Taroo-ni ringo-o tabe-sase-ta koto

nom dat apple acc eat caus past that
'that Hanako made Taroo eat an apple'
* Hanako-ga Taroo-o ringo-o tabe-sase-ta koto
nom acc apple acc eat caus past that

However, the situation in the two languages is not completely identical; indeed, the case marking
facts for the German counterpart of (57) are exactly the reverse of those in Japanese.

(58) a. dass der Stefan den Manfred einen Apfel essen liess
that the.nom the.acc an.acc apple eat made
'that Stefan made Manfred eat an apple'
b. * dass der Stefan dem Manfred einen Apfel essen liess
that the.nom the.dat an.acc apple eat made

The challenge facing us is how to make sense of three separate and apparently contradictory
case-marking facts:

1. the alternation between accusative and dative case-marking on the lower specifier in (53)
and (55),
2. the parallel constraint on double accusative marking in (55) and (57), and
3. the contrasting case-marking pattern between (57) and (58).

So far, we have been assuming that when a head and a noun phrase occur in some case-licensing
configuration, this state of affairs both licenses the noun phrase's occurrence in its particular
syntactic position (spec or comp position) and determines the particular case that appears on the
noun phrase (nominative, accusative, etc.). Let us now weaken this latter assumption somewhat.
In particular, we will allow the case that appears on a noun phrase to be only partially determined
by the case features of the head that licenses its position in the structure; the case can also reflect
further details of the structure, including the case features of other heads. In (53) and (54), case
licensing proceeds as before. GO takes a PP complement and has no case feature. Not
surprisingly, therefore, when GO adjoins to CAUSE, there is no effect on the accusative feature
of CAUSE, which we will assume gets shared by the V node formed by adjunction (the V that
dominates both GO and CAUSE in (54b)). In (55) and (56), on the other hand, GET has an
accusative feature of its own. What we propose is that once GET adjoins to CAUSE, the
presence of the case feature on GET is able to change the value of the case feature on CAUSE
from accusative to dative. This dative feature then percolates up to the V node formed by
adjunction (the V dominating both GET and CAUSE in (56b)). Because small clauses are
structurally analogous to VP shells, moving the lower verb tabe- 'eat' to the higher causative -
sase- in the Japanese causative has the same effect, changing the accusative case feature on -
sase- to dative. This still leaves us with the case-marking contrast between (57) and (58). What
could it be due to? Recall that in the previous chapter, we motivated verb movement in the
Japanese causative on the grounds that the causative morpheme -sase- is a bound morpheme.
The German verb lassen 'let', on the other hand, is not a bound morpheme and there is no reason
to assume that the lower verb moves to it. We can therefore derive the contrast between (57) and
(58) by permitting case features to be changed in the way that we have just proposed only in
connection with the movement of a case-checking head. This is schematically illustrated in (59)
(headedness irrelevant).

(59) a. b.

No verb movement Verb movement

Double accusative case marking Dative-accusative case marking

Case agreement

(coming eventually...)

Notes

1. A very small number of German verbs governs a third case, the genitive. We don't discuss
these verbs here, because they are felt to be archaic.

2. Note how the -s-less plural of deer, which is exceptional in modern English, goes back to Old
English, where it was simply the ordinary plural form for the declension to which deor 'animal'
belonged.

3. The structure in (30) is analogous to that of its grammatical counterpart, He claims to

understand Hegel. Details of the structure (for instance, the presence of the CP) are motivated in
the next chapter.
4. The spec-head configuration is also frequently, though somewhat misleadingly, referred to as
spec-head agreement. The reason for this is that subjects and verbs of sentences, which are in the
spec-head configuration in the VP, agree in number (the man runs/*run; the men run/*runs). The
reason that the term is misleading is that morphological agreement doesn't necessarily imply a
spec-head configuration. For instance, determiners agree in number with the head of their NP
complement (that woman/*women; those women/*woman), but the D and the N aren't in a spec-
head configuration.

5. The relation between the silent determiner and the higher boxed DP in (44) is a mirror image
of the head-comp relation in the following sense. The path between the silent determiner and its
NP complement involves a first segment from D to D' and a downward turn at D' to give the
second segment from D' to NP. Now imagine taking an upward turn at D'. The resulting second
path segment ends at the higher boxed DP.

6. We use subordinate clauses because German main clauses involve a complication, already
mentioned in Chapter 3, that is irrelevant here. See Chapter 14 for details.

Exercises and problems

Exercise 8.1

According to the analysis in the text, why are the sentences in (1) ungrammatical?
(1) a. * He claims to he understand Hegel.
b. * He claims he to understand Hegel.

Exercise 8.2

A. Using the grammar tool in ***, build a ternary-branching structure for (1) (= (37a)) along the
lines that was mentioned, but rejected, in the text.

(1) He expected there to be a fly in his soup.

B. Given the ternary-branching structure, how would the licensing condition on expletive there
have to be reformulated?

Exercise 8.3

In the chapter, we stated that nouns and adjectives aren't case-licensers in English. Provide
evidence for that statement. One piece of evidence for each category is sufficient.

Exercise 8.4

A. Using the grammar tool in x-bar 2, build structures for the sentences in (1).
 For the purposes of this exercise, disregard tense lowering.

Originally a preposition, for in modern English can also be a complementizer. Assume

that the complementizer retains the case-licensing ability of the preposition.
(1) a. I waited for her.
b. I waited for there to be a sale.
c. It would be convenient for daycare to be available for the children.
d. It would be convenient for the parents for daycare to be available.
e. I suspect the class to be difficult.

B. How is case checked on each of the DPs in (1)? Your answer should include which case is
checked, by what head, and in what configuration.

Exercise 8.5

As (1) illustrates, there are no ECM adjectives or nouns in English. Is this a statistical accident,
or is there a deeper reason?

(1) a. * I was expectant there to be a problem.

b. * the expectation there to be a problem

Exercise 8.6

A. Using the grammar tool in Welsh case checking, build structures for the Welsh sentences in
(1) (data from Borsley and Roberts 1996:19, 31).
(1) a. Gwelai Emrys ddraig.
see.conditional Emrys.nom dragon.obj
'Emrys would see a dragon.'
b. Disgwyliodd Emrys i Megan fynd i Fangor.
expected Emrys.nom to Megan.obj go.infinitive to Bangor.obj
'Emrys expected Megan to go to Bangor.'

B. How is case checked on each of the noun phrases in (1)? (As in English, nominative case
cannot be checked in nonfinite clauses in Welsh.) Your answer should include which case is
checked, by which head, and in which licensing configuration.

C. Proto-Indo-European (the ancestor of Welsh) is reconstructed as having had rich agreement,

and so it presumably had verb raising. The Celtic languages, which are descendants of Proto-
Indo-European and to which Welsh belongs, have lost agreement, yet they still exhibit verb
raising. Why didn't the loss of agreement lead to the loss of verb raising in Celtic as it did in
Mainland Scandinavian?

Exercise 8.7
Given the discussion in the textbook so far, exactly one of the following statements is true.
Which is it? Briefly explain your choice.
(1) a. All subjects are agents.
b. All agents are subjects.
c. All subjects check nominative case.
d. All noun phrases that check nominative case are subjects.

Exercise 8.8

Both sentences in (1) are intended to have the same meaning. In a sentence or two, explain why
they contrast in grammaticality.

(1) a. ok It appears that they may solve the problem.

b. * Theyi appear that ti may solve the problem.

Problem 8.1

On the one hand, German appears to have a double accusative constraint ((55b) is
ungrammatical). On the other hand, it appears not to ((58a) is grammatical). Can you resolve the
paradox?

Problem 8.2

A. Use the grammar tool in x-bar 2 to build structures for the gerunds in (1) and (2). You can
reuse structures for (1a,b) if you have already built them in connection with Exercise 5.8. On the
basis of the structures you build, explain how case is checked on the subjects of the gerunds (the
noun phrases in boldface). Your answer should include which case is checked, by what head, and
in what configuration.

(1b) and (2a) are not identical.

(1) a. I disapprove of Kim's impulsive hiring of incompetents.

b. I disapprove of Kim's impulsively hiring incompetents.
(2) a. I disapprove of Kim impulsively hiring incompetents.
b. I'm concerned about there not being time.
c. I watched them running down the street.

B. Why are the sentences in (3) ungrammatical? Build trees if necessary, but where possible you
can explain your answer with reference to trees that you have built for (A).

(3) a. * I disapprove of Kim's impulsive hiring incompetents.

b. * I disapprove of Kim impulsive hiring of incompetents.
C. Some speakers accept the gerunds in (4), though not the one in (5). Explain how case is
checked on the subject of the gerunds in (4), providing the usual details, and also explain what
rules out (5).

(4) a. Kim impulsively hiring incompetents is unfortunate.

b. There not being time is unfortunate.
c. Them running down the street is unfortunate.
(5) * Kim impulsive hiring of incompetents is unfortunate.

D. Can the analysis that you propose in (C) be extended to cover the facts in both (B) and (C)?
Why or why not?

Problem 8.3

In the text, we list several conditions on case checking: the structural licensing condition (43),
the biuniqueness condition (46), the exocentricity condition (47), and the matching condition
(50). Is it possible to eliminate at least one of these? For example, is it possible to derive the
biuniqueness condition from the structural licensing condition and the exocentricity condition?

9 Nonfinite clausal complements

Old version (Fall 2006) - new version will change only minimally, so feel free to read this one.

Selectional restrictions
Subject control
o Evidence for two clauses
o Deriving subject control sentences
Raising
o A detour
o Nonthematic subject positions
o Deriving raising sentences
o Tend and occur
o Promise
Object control
More nonthematic subjects
o Subject idiom chunks
o Weather it
o Summary
Notes
Exercises and problems

So far in this book, we have come across three types of clausal complements: finite clausal
complements, ECM complements, and small clauses. (By 'clausal complement,' we mean any
complement that contains a subject and a predicate.) We briefly review the three types in (1)-(3);
the complement clause is bracketed, and any Infl element in it is in boldface.

The verb forms in (1b,c) are nonfinite, but the head of the complement clauses is a finite
morpheme (a finite tense morpheme in (1b) and a modal in (1c)). As a result, the entire
complement clause is finite. For more information, see Finiteness.

(1) a. We heard [ that the children [past] danced ] .

b. We heard [ that the children [pres] are dancing ] .

c. We heard [ that the children can dance ] .

Like finite complement clauses, the ECM complements in (2) contain an Infl element - namely,
to - but it is nonfinite.

(2) a. We expected [ the children to dance ] .

b. We expected [ the children to be dancing ] .

Finally, small clauses as in (3) contain no Infl element at all.

(3) a. We saw [ the children dance ] .

b. We saw [ the children dancing ] .

Despite their diversity with respect to how and whether Infl is realized, the complement clause
types illustrated in (1)-(3) all have one property in common: namely, the presence of an overt
subject (here, the children). But English also permits nonfinite complement clauses in which a
subject is not overtly expressed (although one is understood). For instance, dance, the verb in the
apparently subjectless complement clauses in (4), has an understood agent.

(4) a. Subject control: The children agreed [ to dance ] .

b. Raising: The children seemed [ to dance ] .

More particularly, this agent is interpreted as being identical to the referent of the matrix subject
the children. Yet unlike (2), where the matrix clause and the complement clause each have their
own subject (we, the children), the sentences in (4) contain only a single overt subject, the one in
the matrix clause. In this chapter, we argue that the nonfinite complements in (4) contain a
structural subject position that is filled by a silent element, and we argue further that the silent
element in question is not the same in the two examples. Rather, we distinguish between subject
control, as in (4a), and raising (sometimes called subject-to-subject raising), as in (4b). In a
subject control sentence like (4a), the complement subject position is filled by a silent
pronominal element PRO, which is coreferential with the referent of the matrix subject. In other
words, we give (4a) the structure in (5a); note the parallel with (5b), where the complement of
agree is finite and where the complement subject position is not PRO, but an ordinary personal
pronoun (the indices on the children PRO, and they are intended to represent coreference).

(5) a. [The children]1 agreed [ PRO1 to dance ] .

b. [The children]1 agreed [ that they1 would dance ] .

The idea behind the term 'subject control' is that the matrix subject fixes, or controls, the
reference of PRO. Notice that the parallel between PRO and overt pronouns in (5) is not
complete. Specifically, PRO in (5a) must be coreferential with the matrix subject, whereas the
pronoun they in (5b) can but needn't be, as succinctly summarized in (i).

(i) a. The children1 agreed [ PRO1,*2 to dance ] .

b. The children1 agreed [ that they1,2 would dance ] .

For this reason, only sentences with nonfinite complements can count as instances of
subject control.

Raising sentences differ from subject control sentences in that their matrix subject position starts
out empty and the complement subject moves up to fill it. Their derivation is schematically
illustrated in (6).

(6) a. Before raising: _____ seemed [ the children to dance ] .

b. After raising: [The children]i seemed [ ti to dance ] .

The assumption that the matrix subject position starts out empty is supported by the fact that
when seem takes a finite counterpart, this same position is filled by expletive it.

(7) Itexpl seemed [ that the children danced ] .

Again, for a sentence to count as an instance of raising, the complement clause must be
nonfinite, as in (6), since it is only then that the complement subject raises into the matrix
clause. (7) contains the same matrix predicate as (6), but the complement subject never
 moves out of its clause.

The chapter also addresses so-called object control. The difference between subject and object
control is illustrated by the contrast between (4a), repeated as (8a), and (8b). In both sentences,
the understood agent of the complement verb dance is the discourse entity referred to by the
phrase the children. But in (8a), that phrase - the controller - is the matrix subject, whereas in
(8b), it appears to be the matrix object.

(8) a. The children agreed to dance.

b. We persuaded the children to dance.

The treatment of object control that we present relies on the VP shells introduced in Chapter 7.
Specifically, we decompose object control predicates into CAUSE and an appropriate subject
control predicate, roughly along the lines of (9).

(9) We [past] CAUSE [ the children agree [ to dance ] ] .

Notice that the children is not actually an object in (9), but rather the subject of the small clause
complement of CAUSE. However, as is customary in the literature, we will continue to use the
term 'object control' as a purely descriptive label for sentences like (8b).

A note on terminology. We will refer to the class of (Fregean) predicates to which agree belongs
as subject control predicates. Similarly, we refer to the class of predicates like seem as raising
predicates, and to the class of predicates like persuade as object control predicates.

The term 'raising predicate' is potentially confusing. It is not the verb itself that undergoes
movement. Rather, as schematically indicated in (6) and as we will see in more detail below, it is
the complement subject that raises into the matrix clause. A better term for the class of
predicates in question would be 'raising triggers.' But we continue to use the term 'raising
predicate' because it is standard in the literature.

A leading role in the analysis of nonfinite complementation is played by expletive there. The
final section of the chapter shows that expletive there belongs to a larger class of nonthematic
subjects.

Selectional restrictions
Before addressing the topics of main concern to us in this chapter, we need to introduce the concept of
selectional restrictions. Selectional restrictions are conditions that a (Fregean) predicate imposes on
one or more of its arguments, depending on its meaning. For instance, drink imposes a selectional
restriction on its theme argument to the effect that the theme argument must refer to a liquid (or an
amount of liquid).

(10) a. Amy drank the { lemonade, #sandwich } .

b. Lukas drank a whole { quart, #piece } .

Elapse selects a subject that refers to an explicitly quantified amount of time.

(11) { Two hours, the shift, #two liters, #Larry } elapsed without further incident.

The felicitous use of the verb murder requires (among other conditions) that both the agent and
the theme arguments refer to humans. By contrast, kill imposes weaker selectional restrictions,
requiring only that the agent and theme arguments refer to living beings.

(12) a. The { paramilitary, #bomb, #avalanche } murdered { her husband, #the olive tree, #her house }.

b. The { paramilitary, bomb, avalanche } killed { her husband, the olive tree, #her house }.

Two points are important to keep in mind concerning selectional restrictions. First, notice that we
are using pound signs, rather than asterisks, in (10)-(12); in other words, we are treating the ill-
formedness of sentences that violate selectional restrictions as semantic/pragmatic deviance, not
as ungrammaticality. This approach is consistent with the fact that selectional restrictions can be
deliberately flouted for special effect. For example, the ordinary (literal) meaning of lap up is 'to
eagerly drink up (used especially of animals)'. Based on this meaning, we would expect it to
select a nonhuman animate agent and a liquid theme. But although both restrictions are violated
in (13), the sentence does not come across as deviant.

(13) The little girl lapped up her teacher's praise.

Rather, the violation of the selectional restrictions signals to the hearer that the sentence is
intended to be taken not literally, but figuratively (here, as an instance of metaphor). (14)
summarizes the kind of reasoning that a hearer of (13) would go through; the reasoning process
itself is ordinarily not explicit, but subconscious and lightning-quick.

(14) The little girl lapped up her teacher's praise?? Whoa there, that's complete nonsense!

It's only nonhuman animals that lap up things. And then, whatever they're lapping up has
 to be liquid, not something abstract like praise.

But the speaker seems to know English and be compos mentis, so what could they have
possibly meant by what they said?

I guess what they must have meant is that the attitude of the little girl towards her teacher's
praise resembles the eagerness with which a thirsty animal laps up some welcome liquid.
In distinguishing figurative from literal uses of language, don't let yourself be confused by the
fact that in the vernacular, the adverb literally is routinely used to qualify figurative statements.
So we often hear people say things like My boss literally hit the roof. In other words, literally has
come to mean figuratively!

Don't, by the way, conclude from examples like (13) that selectional restrictions are in force only
intermittently (in force when language is used literally, but not in force when language is used
figuratively). Rather, it is precisely the fact that selectional restrictions are always in force that
prompts a hearer of (13) to go through a reasoning process like (14) and to come up with an
interpretation in which the selectional restrictions are met in the metaphorical interpretation.

A second point to keep in mind concerning selectional restrictions is that the criteria for set
membership that the restrictions are based on are not always crystal clear. In other words, sets
like liquid things, animate beings, potential murderers, or potential murder victims, and so on,
are somewhat fuzzy around the edges. Speakers might disagree, for instance, about whether the
sentences in (15) are deviant; the disagreement would concern whether the selectional
restrictions on murder might, on the basis of recent advances in the understanding of animal
intelligence, be relaxed to include members of species other than Homo sapiens.

(15) a. The { chimpanzee, dolphin } murdered the explorer.

b. The explorer murdered { the chimpanzee, dolphin } .

Fortunately, for our purposes in this chapter, locating the exact boundary between cases that
meet selectional restrictions and ones that violate them will not be necessary. The important
thing is that selectional restrictions exist, and that there are sentences in which they are clearly
met and ones in which they are clearly violated.

Subject control

Evidence for two clauses

Having introduced selectional restrictions, we now use them to show that subject control sentences
contain two separate clauses, each with their own subject. We begin by showing that in finite
complement counterparts of subject control sentences, which incontrovertibly contain two clauses, like
(16), both the matrix and the complement verbs impose separate selectional restrictions on their
respective subjects. (For simplicity, we omit referential indices in what follows; unless otherwise noted,
the intended interpretation is always the one where the complement subject is coreferential with the
matrix subject.)

(16) The children agreed that they would dance.

We then show that subject control sentences like (17) pattern just like their finite complement
counterparts with respect to the selectional restrictions imposed by the two verbs.

(17) The children agreed to dance.

We begin with (18) and (19), where agree takes a finite complement clause. In (18), we have
taken care to satisfy the selectional requirements of the complement clause (wet selects some
physical object as its argument, and get imposes no further selectional restrictions of its own).
We can therefore be sure that the acceptability contrast in (18) is due to the selectional restriction
imposed by the matrix verb agree, which selects human subjects.

(18) a. The children agreed [ that they would get wet ] .

b. # The { horses, trees, rocks } agreed [ that they would get wet ] .

Conversely, in (19), we have taken care to satisfy the selectional restriction imposed by agree.
Here, the acceptability contrast is due to the selectional restrictions imposed by the various
complement verbs.

(19) a. The children agreed [ that they would speak Twi ] .

b. # The children agreed [ that they would { elapse, evaporate } ] .

If subject control sentences contain two clauses, as we are proposing, each with their own
subjects, they ought to behave analogously to (18) and (19), and this is in fact exactly what we
find in (20) and (21).1

(20) a. The children agreed [ PRO to get wet ] .

b. # The { horses, trees, rocks } agreed [ PRO to get wet ] .

(21) a. The children agreed [ PRO to speak Twi ] .

 b. # The children agreed [ PRO to { elapse, evaporate } ] .

One last thing. Not all subject control predicates allow a finite complement paraphrase.

(22) a. The children tried [ PRO to learn Twi ] .

b. * The children tried [ that they would learn Twi ] .

But even for subject control predicates like those in (22), we assume a biclausal structure with a
PRO subject for the lower clause. This is because, just as in (20) and (21), the subject control
verb and the lower verb impose separate selectional restrictions on their respective subjects. We
ask you to provide the relevant evidence in Exercise 9.1A.

If you have been paying close attention to the examples, you may have noticed that agree seems
to have two slightly different meanings, depending on whether it takes a finite or a nonfinite
complement. When it takes a finite complement, the finite complement expresses a proposition,
and agree means something like 'assent to.' In this case, the subject of the complement clause
need not be coreferential with the matrix subject.

(23) The students agreed [ that the problem was difficult ] .

We also have the semantic intuition that someone besides the agreer shares the belief in the
proposition; in other words, agree takes an optional semantic argument, expressible in the syntax
by a with phrase. So, for instance, if Sam were the last person on Earth, we might describe him
as believing that there were no other people, but it would be odd to describe him as agreeing that
there were no other people. On the other hand, when agree takes a nonfinite complement, it
means something like 'commit oneself to a course of action.' Here, part of the irreducible
meaning of agree seems to be that the agreer and the agent of the predicate of the lower
complement must be identical. As a result, a sentence like (24) is ungrammatical (or perhaps
only semantically deviant).

(24) * The students agreed [ the problem to be difficult ] .

Here, too, there is a sense of an optional argument - an entity to whom the agreer has an
obligation to. Now, so far, we have phrased things as if the semantic differences that we have
described are associated with agree itself. However, a more attractive hypothesis is that agree
has exactly the same meaning in both cases, and that the differences in meaning come about as a
result of the different semantic properties of the complement that agree is combining with.
Under this approach, agree would denote a commitment between a rational being and the kind of
thing that the CP complement refers to, in the presence of another rational being. In the finite
complement case, the CP expresses a proposition. A natural way to interpret commitment
between a rational being and a proposition is as intellectual assent to the proposition, and a
natural scenario is that the co-present rational being shares the belief in the proposition. We have
seen that nonfinite IPs can express propositions (I expect there to be problem is synonymous for
our purposes with I expect that there is a problem), but let us assume that nonfinite CPs (for
some reason) cannot refer to propositions, but only to events or actions.

It is tempting to say that nonfinite complements of subject control predicates are VPs (rather
than CPs). But that would leave the presence of to unexplained. Moreover, there are subject
control predicates that take indirect questions, as illustrated in (i).

(i) They decided whether to buy the house.

A natural way to interpret commitment between a rational being and an action is as a

commitment to seeing to it that the action is carried out, and a natural scenario is that the co-
present rational being is someone to whom the agreer has an obligation. A crucial question that
we will have to leave unresolved here is why the matrix subject controls the subject of the lower
predicate in this case. In other words, what accounts for the oddness of (25a) despite its near-
synonymy with (25b)?

(25) a. # At the meeting, the higher-ups agreed for someone else to do the job.

b. At the meeting, the higher-ups agreed to delegate the job to someone else

Deriving subject control sentences

After this excursion into semantic aspects of control, let us now consider the syntactic representation of
subject control sentences, which is straightforwardly analogous to the representation of their finite
complement counterparts. The elementary tree for agree is the same for both cases and is given in (26).

(26)

Substituting a finite CP complement headed by that at the CP substitution node would yield
structures for sentences like (16). Substituting a nonfinite CP complement headed by a silent
complementizer yields structures for subject control sentences like (17). In what follows, we
illustrate the derivation of (17) in detail.
(27) a. b. c.

Substitute PRO in
Substitute (27a) as complement Substitute (27b) as
specifier position of
of to and subject movement complement of silent C
lower verb

Substituting the structure in (27c) as the complement of the control verb yields (28a), which in
turn becomes the complement of the matrix I element, yielding (28b).

(28) a. b.

Substitute (27c) as complement of control Substitute (28a) as complement of matrix

verb I

Finally, moving the matrix subject yields (29a). The structurally analogous tree for the finite
complement counterpart of (29a) (= (16)) in shown for comparison in (29b).
(29) a. b.

In concluding our discussion of subject control, we must point out that we will leave an
important question about PRO unresolved: namely, how its case feature (if any) is licensed.
There is evidence from languages like German and Icelandic that PRO is able to bear the same
case features that overt subjects do. Nevertheless, PRO and overt noun phrases are in
complementary distribution;2 in other words, the positions that PRO can appear in are ones from
which overt noun phrases are barred, and vice versa. It has therefore been proposed that PRO
does not bear a case feature at all (or that PRO bears a case feature unique to it - so-called null
case - which is checked by nonfinite I in the spec-head configuration). Although this approach
does not address the crosslinguistic facts and leaves it mysterious why the case properties of
PRO and overt noun phrases should differ, it does have the advantage of straightforwardly
capturing the distributional difference between PRO and overt noun phrases.

Raising

A detour

Let us turn now to raising sentences like (4b), repeated here as (30).

(30) The children seemed [ to dance ] .

At first glance, it seems as if we could simply treat such sentences on a par with subject control
sentences. But that would leave us without an explanation for the contrast in (31) - in particular,
for the grammaticality of (31b).

(31) a. * There agreed to be a problem.

b. ok There seemed to be a problem.

The analysis in the previous section does correctly rule out (31a), to which we assign the
structure indicated schematically in (32).

(32) * There agreed [ PRO to be a problem ] .

Given the structure in (32), the sentence is ruled out for two reasons. First, expletive there is not
licensed because it does not occupy the specifier position of a verb of (coming into) existence
(this is the reason that we mark (31a) as ungrammatical rather than as just semantically deviant).

It is true that the sentence contains the there licenser be, but there never substitutes into its
specifier position. The predicate whose specifier position there does substitute into, namely
agree, is not a there licenser (*There agreed some students).

Second, expletive there fails to satisfy the selectional restriction of agree, which, as we saw
earlier, selects human subjects. So (32) is a fine representation because it correctly rules out
(31a) as ungrammatical. But by the same token, if we give (31b) the analogous structure in (33),
we incorrectly expect (31b) to be as ungrammatical as (31a).

(33) Incorrect structure: There seemed [ PRO to be a problem ] .

The reason, once again, is that there is not licensed in the representation in (33). The fact that
seem is not a there licenser is demonstrated in (34) (recall Exercise 3.4).

(34) * There seemed a problem.

Be careful not to confuse the class of raising verbs with the class of there licensers. Raising verbs
like seem don't themselves license there, as we see in (34).

Nonthematic subject positions

At this point, notice that the representations in (32) and (33) are both ruled out because there isn't
licensed in the matrix clause. However, only in (32) is the selectional restriction of the matrix
verb violated. It turns out that a crucial difference between agree and seem is that seem imposes
no selectional restrictions. We can see this by replacing agree in (20) with seem; the
acceptability contrast between (20a) and (20b) disappears in (35). (By contrast, replacing agree
with seem in (21) has no effect on the contrast. After reading this section to the end, you will be
able to explain this fact, and you are asked to do so in Exercise 9.1B.)

(35) a. ok The children seemed to get wet.

 b. ok The { horses, trees, rocks } seemed to get wet.

Another noteworthy property of seem is that its specifier position is not (and, in fact, must not
be) associated with any thematic role. It is true that seem takes an argument: what for lack of a
better term we will call the proposition argument.3 However, this argument cannot be expressed
in the specifier position, as we can see in the finite complement counterparts of raising
sentences.4

(36) a. It seemed that the problem was hard.

b. * That the problem was hard seemed.

To summarize: the subject position of seem is semantically defective in the sense that it is
associated neither with selectional restrictions nor with a thematic role. We will refer to such a
subject position as nonthematic.

Deriving raising sentences

Of course, despite being superfluous from a semantic point of view, nonthematic subject
positions are nevertheless syntactically obligatory (recall the subject requirement introduced in
Chapter 3). This makes possible the following analysis of the grammaticality of (31b) and, more
generally, of all subject raising sentences. We begin by deriving the complement clause. Note
how the eventual matrix subject there is licensed as a specifier of main verb be in (37a). We
assume that the subject moves from Spec(VP) to Spec(IP) in (37c) in order to provide the
complement clause with a subject in accordance with the subject requirement.

In the case of a nonexpletive matrix subject, as in The children seemed to dance, the children
would substitute into the specifier of dance, thereby becoming associated with that verb's
agent role.

(37) a. b.

Substitute eventual matrix subject as Substitute (37a) as complement of to and

specifier of lower verb subject movement
We now substitute (37b) into the elementary tree for seemed in (38a).

(38) a. b.

Before proceeding with the derivation, a few words about the elementary trees in (38) are in
order. First, note that both elementary trees in (38) contain a specifier position. Though
semantically unnecessary, as discussed above, this position is motivated by the syntactic
obligatoriness of expletive it in small clauses.

(39) a. They made [ it seem that there was a problem. ]

b. * They made [ seem that there was a problem. ]

Second, the syntactic category of the clausal complement is IP in (38a), whereas it is CP in

(38b), which we would use if we wanted to derive the finite complement counterpart of (31b) (It
seemed that there was a problem). The reason that raising predicates, in contrast to control
predicates, require different elementary trees depending on the finiteness of their complement
has to do with certain structural conditions that must be satisfied by traces of movement (but not
by PRO). We simply mention the existence of these conditions here; their exact character and
their motivation go beyond the scope of this textbook.

Substituting the clause in (37c) as the complement of the elementary tree for raising seem in
(38a) yields (40a), which in turn becomes the complement of the matrix I element, yielding
(40b).

(40) a. b.

Substitute (37c) as complement of raising verb Substitute (40a) as complement of matrix I

At this point in the derivation, the option arises in principle of substituting expletive it in the
matrix Spec(VP) and moving it to the matrix Spec(IP). In fact, this is what we would do if the
complement of seem were finite. In the case of a nonfinite complement, however, this step yields
the hopelessly ungrammatical (41).

(41) * It seemed there to be a problem.

Why is (41) ungrammatical? The reason is that there bears a nominative case feature that cannot
be checked in the lower clause. (How we know that the case feature is nominative is left to you
to determine in Exercise 9.1C.) The case feature can't be checked in the lower Spec(IP) (its
position in (41)) because nonfinite I is unable to check case at all. The case feature also can't be
checked in the lower Spec(VP) (its position before subject movement in the complement clause)
because V doesn't check case in the spec-head configuration.

In principle, seemed could check objective case on there in the head-spec configuration just as
expect and other ECM verbs do. However, the ungrammaticality of (41) shows that raising
verbs are in fact unable to do so. Why they should differ from ECM verbs in this way is
obviously something that needs to be explained. It has been proposed that there is a
correlation between a verb's ability to assign a thematic role to its specifier and its ability to
check objective case. This correlation, known as Burzio's generalization, is discussed further
in Chapter 10.

Since the complement subject's case feature cannot be checked within its own IP, it is forced to
move via the matrix Spec(VP) to the matrix Spec(IP), as shown in (42a). In this final position,
nominative case is checked by the finite I of the matrix clause in the spec-head configuration.
For comparison, the structure of the finite complement counterpart is shown in (42b); here, each
of the two subjects checks nominative case with its own finite I.

(42) a. b.
It is the movement of the subject from the complement clause to the matrix clause in (42a) that is
known as raising. Raising, like ordinary subject movement, targets the subject. In both cases, an
element that is licensed in a lower specifier position moves to a higher specifier position. In the
case of expletive there, the licensing is by an appropriate verb; in the case of nonexpletive
subjects, a natural assumption is that they need to be licensed as arguments of the lower verb. In
both cases of movement, any licensing relations are maintained by the trace of movement. And
finally, in both cases, the lower position is not a case-checking position, whereas the higher one
is. The one difference between raising and ordinary subject movement is that in the case of
raising, the subject moves out of the IP in which it originates.

Tend and occur

As was the case with subject control verbs, certain raising verbs are able to take finite complements in
addition to nonfinite ones, whereas others are restricted to nonfinite complements. Seem, as we have
seen, belongs to the first type. Tend, as shown in (44), belongs to the second type.

(43) a. ok There seem to be huge traffic jams during rush hour.

b. ok It seems that there are huge traffic jams during rush hour.

(44) a. ok There tend to be huge traffic jams during rush hour.

b. * It tends that there are huge traffic jams during rush hour.

Despite the contrast between the (b) examples, seem and tend are both raising verbs; what is
crucial is that the (a) examples, in which the subject of the complement clause moves out of its
clause, are both grammatical.

There are also verbs with the converse pattern of tend.

(45) a. It occurred to me that there is a solution.

b. * There occurred to me to be a solution.

Such verbs have a nonthematic subject position, just like seem and tend. However, they are not
considered raising verbs, since their complement subjects cannot move out of the clause they
originate in, as was mentioned in the introduction.

Promise

Promise has the noteworthy property of behaving either as a subject control predicate or as a raising
predicate. As a subject control predicate, promise means something like 'vow' and selects rational
agents as subjects. This promise can take either finite or nonfinite complements.
(46) a. The { children, #horses } promised [ to eat their oatmeal ] .

b. The { children, #horses } promised [ that they would eat their oatmeal ] .

On this interpretation, the matrix clause can contain manner adjuncts that modify promise (notice
how promise can be replaced by vow in this examples).

(47) a. The children softly promised [ to eat their oatmeal ] .

b. The children obediently promised [ to eat their oatmeal ] .

The grammaticality of (47a) is particularly important, since the only licenser for softly is promise
(*the children were soft); in (47b), it could be argued that obediently is licensed as a property of
the children (ok-the children were obedient).

But promise can also have a 'weaker' meaning; on this interpretation, a sentence like (48a) can be
paraphrased as (48b).

(48) a. This filly promises to win the race.

b. All available evidence indicates that this filly will win the race.

On this interpretation, modifying the matrix predicate by a manner adverb as in (47) is as deviant
in the original promise sentence as in the paraphrase.

(49) a. # This filly { softly, obediently } promises to win the race.

b. # All available evidence { softly, obediently } indicates that this filly will win the race.

Notice, moreover, that in the alternative paraphrase in (50), the presence of expletive there in
matrix subject position indicates that the position is nonthematic.

(50) There is every indication that this filly will win the race.

The nonthematic character of the matrix subject position for this interpretation of promise is
borne out by the grammaticality of (51).

(51) There promises to be a new version by spring.

From these facts, we conclude that the proper representation for a sentence like (51) must be the
raising structure schematically indicated in (52).

(52) Therei promises [ ti to be a new version by spring ] .

Promise sentences with nonthematic subjects, on the other hand, are ambiguous between a
raising analysis and a subject control analysis (as long as they contain no disambiguating
adverbs). Which reading is prominent depends, as always, on the discourse context. In a sentence
like (46a), the prominent interpretation, and the only one considered so far, is the subject control
interpretation represented in (53).

(53) The { children, #horses } promised [ PRO to eat their oatmeal ] .

However, (46a) also has the raising interpretation represented in (54a), which can be paraphrased
as in (54b).

(54) a. [ The { children, horses } ]i promised [ ti to eat their oatmeal ] .

b. There was every indication that the { children, horses } would eat their oatmeal.

Notice that under this interpretation, the contrast between children and horses that is due to the
selectional restrictions imposed by subject control promise disappears.

There is at least one other verb in English that clearly has the same property as promise - namely,
threaten. You are asked to provide evidence for this assertion in Exercise 9.1D.

Object control

In this section, we present an analysis of object control predicates like persuade. The analysis is
extremely simple. According to it, object control predicates are VP shell structures in which a subject
control predicate is embedded under a causative predicate (Larson 1988). Recall from Chapter 7 that we
already have a VP shell analysis of persuade when it takes finite complements, as in (55).

(55) We persuaded the children that they should dance.

The requisite VP shell structure is shown in (56).

(56)

In extending the analysis of the finite complement case to the object control case, we will make a
slight revision and replace BELIEVE by COMMIT. This is because BELIEVE is a predicate that
can combine with propositions but not with actions, whereas COMMIT is general enough to
combine with either propositions or actions, along the lines discussed earlier in connection with
agree. We are not claiming, incidentally, that COMMIT is exactly identical with agree. The two
predicates differ in that COMMIT does not take an optional argument referring to a co-present
rational being. In other words, we can persuade someone that the moon is made of green cheese
without necessarily sharing that belief ourselves, and we can persuade someone to do the dishes
without their incurring an obligation to us to do so. Given this slight change, we are now in a
position to derive (57).

(57) We persuaded the children to dance.

The derivation of the nonfinite clause to dance is exactly the same as in the case of a subject
control sentence; (58a) is identical to (27c). (58a) substitutes as the complement of COMMIT,
and substituting the children in the specifier position of COMMIT yields (58b).

(58) a. b. c.

Substitute (58a) as
Reuse (27c) complement of abstract Substitute specifier in (58b)
subject control verb
Notice how the apparent matrix object the children is not actually an object, but rather a subject
of a small clause complement; in a moment, its objective case feature will be checked by
CAUSE in the head-spec configuration. (58c) substitutes as the complement of CAUSE, yielding
(59a), and then substituting the matrix subject we in the specifier position of CAUSE and
abstract verb movement yields (59b).

(59) a. b.

Substitute (58c) as complement of Substitute specifier in (59a) and abstract verb

CAUSE movement

Finally, the VP in (59b) substitutes into the elementary tree for the matrix I, and the matrix
subject undergoes subject movement. For simplicity, we omit these last steps of the derivation.

The analysis just presented is straightforwardly consistent with the contrast between ECM and
object control illustrated in (60).

(60) a. We expected there to be a problem.

b. # We persuaded there to be a problem.

The structures for the two sentences are shown in (61).

(61) a. b.

In (61a), expletive there is licensed by originating as the specifier of main verb be in the
complement clause. (61b), on the other hand, is ruled out for exactly the same reasons as (32),
repeated here as (62).

(62) # There agreed [ PRO to be a problem ] .

First, expletive there is not licensed, since neither COMMIT nor agree are verbs of existence.
Second, COMMIT, like agree, selects rational beings as subjects, and expletive there fails to
satisfy this selectional restriction.

More nonthematic subjects

In distinguishing among the various verb classes discussed in this chapter, we have relied heavily on the
distribution of expletive there (or, to put it another way, on the grammaticality or ungrammaticality of
sentences containing expletive there). As it turns out, expletive there is not the only nonthematic
subject (= subject that is not associated with a thematic role). In this final section of the chapter, we
present two further instances of nonthematic subjects: so-called subject idiom chunks and weather it.

Subject idiom chunks

In Chapter 7, we introduced a constraint according to which idioms must form a constituent, and we
mentioned the existence of clausal idioms like The shit hit the fan. Subject idiom chunks are simply the
subjects of such clausal idioms. Some further examples of clausal idioms are given in (63); the subject
idiom chunks are italicized.

(63) a. The cat is out of the bag.

 b. The fur will fly.

c. The jig is up.

d. The pot is calling the kettle black.

Subject idiom chunks share two important properties with expletive there. First, just as expletive
there must be licensed by a verb of existence, the subjects in (63) have whatever idiomatic force
they have only in connection with the rest of the idiom, but not otherwise.

The relation between subject idiom chunks and their predicates is actually even stronger than
the relationship between expletive there and its licensers, since predicate idiom chunks also
have no independent idiomatic meaning of their own, whereas verbs of existence can occur
independently of existential there.

For instance, neither cat in (63a) nor pot in (63d) have a metaphorical sense of secret or
hypocrite, respectively, in other syntactic contexts. So sentences as in (64) have only literal
interpretations.

(64) a. The cat is safe with her.

(can't mean: 'The secret is safe with her.')

b. # Fortunately, the pot retracted that piece of duplicitous slander.

(can't mean: 'The hypocrite retracted that piece of duplicitous slander.')

Second, presumably because they are not interpreted literally, subject idiom chunks don't seem to
be associated with any thematic role, and so they can occupy the nonthematic subject position of
raising predicates. As a result of these two properties, contrasts as in (65) are expected.

(65) a. # The cat agreed [ PRO to be out of the bag ] .

b. [The cat]i seems [ ti to be out of the bag ] .

(65a) is ruled out both on a literal and an idiomatic reading. Agree selects rational beings as
subjects and is therefore incompatible with the cat either as a literal or as an idiomatic
(nonthematic) subject. In addition, the cat isn't licensed as an idiom chunk in the representation
in (65a) because it doesn't form a constituent with the rest of the idiom. By contrast, both
readings, and in particular the idiomatic one, are possible in (65b). This is expected, since the
matrix subject originates in the complement clause, forming a constituent with the remainder of
the idiom.
Note that examples like (66) do not invalidate the diagnostic value of subject idiom chunks in
distinguishing between subject control and raising predicates.

(66) The cat wanted [ PRO to be out of the bag ] . (only literal interpretation)

Here, want imposes less strict selectional restrictions on its subject. Since the subject needn't be
rational (only have a reasonably well-developed nervous system), the sentence is grammatical,
unlike (65a). However, unlike in (65b), the matrix subject position isn't nonthematic and the
matrix subject doesn't move out of the lower clause, so the sentence has only a literal
interpretation.

Weather it

The third type of nonthematic subject is weather it, the subject of verbs of precipitation.

(67) ok It is { hailing, pouring, raining, sleeting, snowing. }

As with subject idiom chunks and their predicates, the licensing relationship between weather it
and their predicates is mutual: not only is weather it licensed by weather verbs, but the weather
verbs are in turn themselves licensed by weather it, as shown in (68).

(68) * The { air, atmosphere, environment, precipitation, sky, weather } is { hailing, pouring, raining,
sleeting, snowing. }

Given the nonthematic character of weather it, contrasts as in (69) are expected.

(70) a. # It decided to rain at night.

b. It tends to rain at night.

Unexpectedly, given what we have said so far, sentences like (71) are not that unacceptable.

(71) ? It's trying to rain; it finally managed to rain.

What is going on here? A simple explanation is that the selectional restrictions of try and manage
are being violated at a literal level (recall the discussion of the little girl lapping up her teacher's
praise), prompting the hearer to conceptualize the weather as an animate being.

Summary
We have seen that expletive there is licensed as the subject of verbs of (coming into) existence. Subject
idiom chunks and their predicates stand in a mutual licensing relationship, as does weather it with
weather predicates. Because of their special licensing requirements, none of these subjects is licensed
as the subject of a subject control predicate (or as the apparent object of an object control predicate).
Nor can a control predicate's selectional restrictions not be met by a nonthematic subject. By contrast,
raising predicates neither interfere with the licensing of nonthematic subjects (which takes place in a
lower clause) nor do they impose selectional restrictions that the nonthematic subjects cannot meet. It
is precisely because of their semantic defectiveness that they are able to act as grammatical catalysts,
allowing licensing relations that are normally confined to the same clause to extend across clause
boundaries.

As we have seen, the special properties of nonthematic subjects make them useful diagnostics to
distinguish subject control from raising predicates (and, mutatis mutandis, ECM from object
control predicates). The relevant judgments are summarized in (72); for convenience, we also
include the judgments for manner adverbs discussed in connection with promise.

Subject control Raising

(72) Expletive there * ok

Subject idiom chunk # (or only literal) ok (both idiomatic and literal)

Weather it # (or metaphorical) ok

Manner adverbs ok *

Notes

1. It might occur to a careful reader that an alternative approach to the facts in (20) and (21) is
possible, according to which control sentences contain a single subject (not two, as in the text),
which must simultaneously satisfy the selectional restrictions of both the higher and the lower
verbs. Regardless of whether such an approach might be worked out in detail for English, we do
not adopt it, since it cannot be extended to handle control constructions universally. In particular,
the approach in question fails for Icelandic, as we briefly describe in what follows.

In contrast to English (and most other languages), Icelandic has certain verbs whose subjects
appear in some non-nominative case (genitive, dative, or accusative), even in finite clauses. The
analysis of these so-called 'quirky case' subjects is beyond the scope of this textbook, but it is
well established that they are true subjects (despite the lack of subject-verb agreement) (see, for
instance, Zaenen, Maling, and Thraínsson 1985, Sigurðsson 1991, and the many references
therein). (i) gives examples of Icelandic finite clauses with an ordinary nominative subject and
with a quirky case subject. The subjects are in boldface. Note that the underlined quantifiers
agree in case with the subjects; this fact will be important directly.
(i) a. Ordinary nominative Strákarnir komust allir i skóla.
the-boys (nom) got all (nom pl m) in school
subject 'The boys all got to school.' (Sigurðsson 1991:***,
(**))
b. 'Quirky dative' subject Strákanum leiddist öllum i skóla.
the-boys (dat) was-bored all (dat pl m) in school
'The boys were all bored in school.' (Sigurðsson
1991:***, (**))

(ii) shows that the clauses in (i) can be embedded under a subject control verb (here, vonast til
'hope for'). As in English, the subject of the embedded clauses is silent, but note that the
quantifiers continue to exhibit the same case that they did in (i).

(ii) a. Embedded ordinary Strákarnir vonast til að PRO komast allir

í skóla.
subject the-boys (nom) hope for Comp get all (nom pl
m) in school
'The boys hope to all get to school.' (Sigurðsson
1991:***, (**))
b. Embedded quirky Strákarnir vonast til að PRO leidhast ekki öllum
í skóla.
subject the-boys (nom) hope for Comp be-bored not all
(dat pl m) in school
'The boys hope to all not be bored in school.'
(Sigurðsson 1991:***, (**))

In particular, in (ii.b), the quantifier must appear in the dative. From this, we conclude that the
silent subject of the lower clause in (ii.b) checks quirky dative case in (ii.b), just as it did in the
finite clause in (i.b). The fact that the matrix and embedded subjects don't bear the same case
feature in (ii.b) provides conclusive evidence that control constructions are indeed biclausal,
since a single noun phrase cannot check more than one case (even though it might satisfy more
than one selectional restriction at the same time).

2. The statement in the text is an oversimplification. In fact, there is a bit of overlap in the
distribution of PRO and overt noun phrases - for instance, the subject position of gerunds.

(i) a. [ PRO going out with him ] would bother me.

b. [ { Kim's, Kim } going out with him ] would bother me.

3. For simplicity, we focus on the proposition argument of seem and disregard the optional
experiencer (It seems to me that you've solved the problem). Including the latter in our
considerations would not affect our conclusions.

4. Even clearer evidence that the specifier position at issue is the Spec(VP) associated with seem
(and not, say, some higher specifier position, such as Spec(IP)) comes from small clauses like (i).
(i) a. They made [ it seem [ that the problem was hard ] ] .

b. * They made [ [ that the problem was hard ] seem ] .

Exercises and problems

Exercise 9.1

A. As mentioned in the text, certain subject control predicates, like try, cannot take finite complements.
Provide evidence that in subject control sentences containing these predicates, both the matrix
predicate and the complement predicate impose their own selectional restrictions, thus motivating a
biclausal analysis even for such sentences.

B. Explain the acceptability contrast in (1).

(1) a. The children seemed to learn Twi.

b. # The children seemed to { elapse, evaporate }.

C. What is the evidence that the case feature on there in (2) is nominative?

(2) There seemed to be a problem.

D. Show that threaten is both a subject control verb and a raising verb.

Exercise 9.2

A. The premises of the following argument are correct, and the conclusion itself may be correct, but the
argument is invalid. Where is the fallacy?

(1) is grammatical; (2) is ungrammatical. Therefore, manage must be a subject control predicate, not a
raising predicate.

(1) She managed to solve the problem.

(2) * There managed to solve the problem.

B. Using the sample answers in (2)-(4) as a model, determine whether the verbs in (5) are subject
control predicates, raising predicates, or neither. For the purposes of this exercise, use only active
verb forms.
 Note that sometimes two pieces of evidence are necessary to conclusively determine a verb's
status.

Evidence: Conclusion:

(2) a. ok There chanced to be Chance is a raising verb.

an opening. (2a) is conclusive evidence; (2b) corroborates.

b. ok It chanced that there

was an opening.

(3) a. * There slipped out to Slip out (like occur) is neither a raising verb nor a control verb.
be a problem. The ungrammaticality of (3a) shows that it is not a raising verb.
(3b) shows that its subject position is nonthematic, so it is not a
b. ok It slipped out that control verb, either.
there was a problem.

(4) a. * There resolved to be Resolve is a subject control verb.

a solution. (4a) shows that it is not a raising verb. (4b) (on the intended
reading, with expletive it) shows that its subject position is
b. * It resolved that there thematic.
would be a solution.

(5) agree, aspire, attempt, be, beg, cease, choose, claim, come, commence, continue, dare, demand,
deserve, desire, determine, elect, end up, endeavor, expect, fail, forget, happen, have, hope,
intend, look, mean, need, neglect, plan, pledge, prefer, presume, pretend, proceed, prove,
purport, remember, request, start, strive, swear, tend, train, try, volunteer, vow, wish, yearn

C. Subject control predicates and raising predicates can belong to other syntactic categories than
V. The predicates in (6) are adjectives, those in (7) are participles and it's not always completely
clear whether they are adjectives or verbs, and about in (8) is a preposition. As in (B), determine
which class each of these predicates belong to, giving the evidence on which your decision is
based.

(6) afraid, anxious, apt, certain, content, eager, ecstatic, evident, fortunate, glad, happy, hesitant,
liable, likely, lucky, necessary, possible, ready, reluctant, sorry, sure, unlikely

(7) bound, delighted, destined, determined, embarrassed, excited, fated, going, inclined, itching,
 jonesing, prepared, scared, (all) set, supposed, thrilled

(8) about

D. Using (9)-(11) as a model, determine whether the verbs in (12) are ECM verbs or object
control verbs. Once again, use only active verb forms. For the purpose of this exercise, do not
worry about how you would semantically decompose any object control verbs that you find.

Evidence: Conclusion:

(9) ok I assumed there to be a Assume is an ECM verb.

problem.

(10) a. ok I convinced John to take Convince is an object control verb.

the job.

b. * I convinced there to be a
problem.

(11) a. ok I noticed that the problem Notice doesn't allow nonfinite complements; therefore it is neither
was difficult. an ECM verb nor an object control verb.

b. * I noticed the problem to

be difficult.

(12) acknowledge, advise, allow, anticipate, ask, beg, blackmail, challenge, command, consider,
convince, corral, dare, deem, determine, discover, encourage, enjoin, expect, fear, find, forbid,
get, help, instruct, invite, know, order, perceive, permit, predict, pressure, prompt, prove,
provoke, remind, report, request, require, tell, tempt, urge, warn

Exercise 9.3

For this exercise, be sure to provide the evidence (grammatical or ungrammatical sentences) on
the basis of which you decide that a particular verb belongs to a particular class. Not much in the
way of discussion is required beyond that.

A. As was mentioned in this chapter, raising verbs are logically distinct from there licensers.
There are, however, some verbs that belong to both classes (this is comparable to a single person
belonging to two distinct clubs). Can you think of any?
B. As we mentioned in this chapter, there are raising verbs that cannot take finite complements.
Can you think of other verbs besides tend with this property?

C. Expect is an ECM verb. Which other class(es) of verbs discussed in this chapter does it belong
to?

Exercise 9.4

A. Using the guidelines from Exercise 9.2, determine which class the matrix predicates in (1) belong to.
In (1e,f), you will also have to decide which syntactic category the matrix predicate belongs to; briefly
explain your decision.

(1) a. They failed to be on time.

b. They aspired to get the job.

c. They reminded him to solve the problem.

d. They aren't hesitant to move.

e. They are fated to get the job.

f. They are about to graduate.

B. Using the grammar tool in x-bar 3, build structures for the sentences in (1). Provide suitable
decompositions for any object control predicates.

Exercise 9.5

Explain why the following sentences are ungrammatical. The intended meaning is (2). If convenient, you
can use the grammar tool in x-bar 3 to build structures for the sentences, but you don't need to. The
subscripts in (1b) are intended to indicate that the two instances of they refer to the same discourse
entity.

(1) a. * There seems that they like caviar.

b. * They1 seem that they1 like caviar.

c. * Theyi seem that ti like caviar.

d. * Caviari seems that they like ti.

(2) a. It seems that they like caviar.

 b. They seem to like caviar.

Exercise 9.6

A. You have been asked to review an article for Linguistic Inquiry by Professor Richard Gerneweis, in
which he concludes on the basis of the contrast in (1) that volunteer is a control verb. What is wrong
with his argument?

(1) a. ok Amy volunteered to do the job.

b. * There volunteered to do the job.

B. In your review, you graciously provide the conclusive evidence that Professor Gerneweis
should have provided himself.

Exercise 9.7

B. Explain why the sentences in (1) are ungrammatical. The intended meaning is (2). If
convenient, you can use the grammar tool in x-bar 3 to build the relevant structures, but you
don't need to. Assume that expletive it substitutes directly into Spec(IP), rather than moving
there from Spec(VP).

(1) a. * It seems Jackie to have solved the problem.

b. * Jackie seems that has solved the problem.

(2) a. ok Jackie seems to have solved the problem.

b. ok It seems that Jackie has solved the problem.

Problem 9.1

Based on the analysis of object control verbs in this chapter, explain the contrast between (1) and (2).

(1) I persuaded them to come.

(2) a. * I was persuasive them to come.

b. * I was persuasive of them to come.

Problem 9.2
Some of the predicates discussed in this chapter take either a finite or a nonfinite clausal complement;
others take only a nonfinite complement. Finally, some predicates that in principle might take a
nonfinite complement don't.

(1) a. The children agreed that they would dance; the children agreed to dance.

b. It appears that the bear is hibernating; the bear appears to be hibernating.

(2) a. The children tried to dance; *the children tried that they would dance.

b. The bear tends to hibernate in winter; *it tends that the bear hibernates in winter.

(3) It is evident that there is a problem; *there is evident to be a problem.

The analysis presented in the chapter is not detailed enough to account for these facts. Suggest
how the analysis could be appropriately extended or revised.

11 Wh- movement in English

A movement analysis of questions

o Complementation
o Why a silent complementizer?
o Case checking
o Direct wh- questions
Constraints on wh- movement
o The island constraints
 The apparent unboundedness of wh- movement
 A typology of islands
o The Comp-trace effect
Relative clauses as instances of wh- movement
o Wh- relative clauses
o That relative clauses
o Doubly marked relative clauses
o Zero relative clauses
Notes
Exercises and problems
Supplementary material
o Questions

In this chapter, we introduce a type of movement that differs from the ones discussed so far
(subject movement, subject raising, passive). First, it affects not just DPs, but maximal
projections of many syntactic categories. Second, the landing site for the moved constituent is
outside of IP. Because this type of movement is involved in the derivation of wh- questions, it is
known as wh- movement.

In this chapter, we present the evidence for a wh- movement analysis of questions. We then
describe two constraints on wh- movement that have given rise to much discussion in the field:
the island constraints (Ross 1967) and the Comp-trace effect (Perlmutter 1971). In this chapter,
we focus on a description of the constraints, postponing attempts at analysis until a later chapter.
In the final section of the chapter, we extend a wh- movement analysis to the various types of
relative clauses found in English.

A movement analysis of questions

Complementation

As the presence of the complementizer if in the indirect question in (1) shows, the verb wonder takes a
CP complement.

(1) They wonder if the lions will devour the wildebeest.

The elementary trees for wonder and if are given in (2a,b), and the entire tree for (1) is given in
(2c).

(2) a. b. c.

Now consider the indirect question in (3), which begins with a wh- phrase (a maximal
projection) rather than with a complementizer (a head).

(3) They wonder which wildebeest the lions will devour.

Let's adopt the null hypothesis that wonder is associated with the same elementary tree in (3) as
in (1) - namely, with (2a). Since (3) contains no overt complementizer, the CP tree that
substitutes into the complement node of the elementary tree for wonder must then be the
projection of a silent complementizer. For reasons to be given shortly, we take this
complementizer to be a silent counterpart of that. In deriving the tree for (3), a further difficulty
remains concerning the wh- phrase which wildebeest. On the one hand, the wh- phrase must be
the object of devour, just as in (1), because devour is obligatorily transitive. But on the other
hand, the wh- phrase precedes the subject of the subordinate clause rather than following the
verb. As usual when we are confronted with a mismatch of this sort, we invoke movement in
order to allow a single phrase to simultaneously play several roles in a sentence. Specifically, we
will have the wh- phrase originate as the sister of the verb whose object it is and then move to
Spec(CP). This allows us to accommodate the word order in (3), while maintaining that devour is
a transitive verb regardless of what clause type (declarative or interrogative) it happens to occur
in. The resulting structure for (3) is shown in (4).

(4)

The argument just presented is based on the obligatorily transitive character of devour, but it can
be extended straightforwardly to other syntactic relations - for instance, modification. Recall that
the way that we have chosen to represent the modification relation is to adjoin the modifier at the
intermediate projection of the modifiee. In (5a), the adverb phrase unbelievably quickly modifies
the verb devour, and so it adjoins at V'. In (5b), the corresponding wh- phrase how quickly also
modifies devour, so it needs to adjoin to V' as well, but it precedes the subject. Again, the
mismatch between the position where the phrase is interpreted and where it is pronounced can be
resolved by moving the modifier, as shown in (6).

(5) a. The lions will devour the wildebeest unbelievably quickly.

b. They wonder how quickly the lions will devour the wildebeest.
(6)

Why a silent complementizer?

Let's turn now to the question of why we treat the complementizer in (4) and in (6) as as a silent
counterpart of that. There are several reasons. First, Middle English (1150–1500) routinely
allowed (though it did not require) overt that as the syntactic head of indirect wh- questions. The
examples in (7) are from the Penn-Helsinki Parsed Corpus of Middle English. (7c,d) show that
that alternated with its silent counterpart even in the usage of individual authors (here, Chaucer).

(7) a. he wiste wel hymself what that he wolde answere (cmctmeli.m3, 219.C1.75)
'he himself knew well what he would answer'

b. for ye han ful ofte assayed … how wel that I kan hyde and hele thynges
(cmctmeli.m3,221.C1.149)
'for you have very often determined how well I can hide and conceal things'

c. I wolde fayn knowe how that ye understonde thilke wordes and what is youre sentence
(cmctmeli.m3,227.C2.408)
'I would like to know how you understand these same words'

d. And forther over, it is necessarie to understonde whennes that synnes spryngen, and how they
encreessen (cmctpars.m3, 296.C1b.354)
'And moreover, it is necessary to understand where sins come from, and how they increase'

e. Now shal ye understonde in what manere that synne wexeth or encreesseth in man.
(cmctpars.m3, 297.C2.392)
'Now you shall understand in what manner sin grows or increases in man.'

f. The fifthe circumstaunce is how manye tymes that he hath synned … and how ofte that he hath
 falle. (cmctpars.m3, 323.C1.1501)
'The fifth circumstance is how many times he has sinned … and how often he has fallen.'

Second, contemporary Belfast English resembles Middle English in this respect (Henry
1995:107) (this is the same variety that you were introduced to in Exercise 1.2).

(8) a. I wonder which dish that they picked.

b. They didn't know which model that we had discussed.

Third, wh- phrases followed by that continue to be attested in the unplanned usage of speakers of
modern standard English. Over the years, we have collected the examples in (9) (Radford
1988:500 has a similar collection).

(9) a. I realized how interesting that it was.

(Clara Orsitti, interviewed by Vicky Barker, World Update, National Public Radio, 25 January
1999)

b. Most of my colleagues were amazed how quickly that I recovered.

(advertisement for Temple University Hospital, WRTI, 24 November 1999)

c. It could be that that is why that they were understood. (high-low-high intonation on why)
(Joanna Labov, doctoral dissertation defense, 4 May 2000)

d. These recounts will determine how much of a pick-up that we will have," said Democratic
National Committee Chairman Joe Andrew."
(http://www.cnn.com/2000/ALLPOLITICS/stories/11/08/house.races/)

e. to find out what kind of a house that she was looking for
(Amy Forsyth, in conversation, 23 September 2004)

f. I don't know what floor that it is

(overheard at the Down Home Diner, Reading Terminal Market, Philadelphia, PA, 15 January
2005)

g. I would tell them how that they should solve their problems (high-low-high intonation on how)
(overheard at 4th and Pine Streets, Philadelphia, PA, 5 March 2005)

h. I realize how many people that I've forgotten to put on my list.

(overheard at 2044 Fairmount Avenue, Philadelphia, PA, 13 August 2006)
i. He talks about how profound that he thinks it is.
(overheard at Whole Foods, 10th and South Streets, Philadelphia, PA, 13 October 2006)

j. I don't know what kind of food that they've been used to.
(overheard at Fairmount Pet Shoppe, 2024 Fairmount Avenue, Philadelphia, PA, 22 October
2006)

k. It doesn't matter how long that I have in the program.

(overheard at 2044 Fairmount Avenue, Philadelphia, PA, 29 October 2006)

l. I just didn't realize how much that I beat myself emotionally to a pulp.
(overheard in East Dorset, VT, 10 November 2006)

m. ... tells them which of these things that they're dealing with.
(Mark Steedman, Linguistics Speaker Series talk, University of Pennslvania, 30 November 2006)

n. I never really realized how much time that they devote

(overheard in Lake Harmony, PA, 3 November 2007)

o. I don't care what office that you have

(overheard in Lake Harmony, PA, 3 November 2007)

p. That way we know how long that they were out.

(Laura McRae, in conversation, Philadelphia, PA, 16 November 2007)

q. ... we're still waiting to hear for Senator Clinton to tell us what precise foreign policy experience
that she is claiming.
(Barack Obama, early March 2008,
http://voices.washingtonpost.com/44/2008/03/02/obama_seeks_to_reverse_the_que.html,
accessed 1 June 2009)

r. You know, Aasif, I've heard some people say that Iraqi deaths - civilian deaths especially - are
down because of how successful that these militia death squads have been in actually
segregating Iraqi society.
(Jon Stewart, Petraeus Hearings, The Daily Show, 9 April 2008)

s. ... discuss what subcommittee that they're on.

(Ken Hodge, in conversation, 19 June 2008)

t. Are you prepared to make a public declaration of which of these two candidates that you are
prepared to support?
(Tom Brokaw, Meet the Press, 19 October 2008)
 u. My youngest son, I never knew how affected that he was.
(overheard at 17th and Sansom Streets, 29 August 2009)

v. What's important is that we teach our children the fundamentals now so that they can be
successful in whatever nightmare scenario that they may find themselves in.
("Are violent video games adequately preparing children for the apocalyse?",
http://www.theonion.com/content/video/are_violent_video_games, accessed 24 Oct 2009)

w. I want to start off by telling folks here how inspired that I've been by your continued resilience in
the wake of the floods that devastated this region a few years back
(Barack Obama, www.nytimes.com/2010/03/26/us/politics/26obama-text.htm, accessed 29
March 2010)

Finally, sequences of wh- phrase + overt complementizer in indirect questions occur in

languages other than English. The complementizer in question is generally the counterpart of
that, but the counterpart of if is attested as well.

(10) Bavarian (Bayer 1983-4:212, (8a-d))

a. I woass ned wer dass des toa hod.

I know not who that that done has
'I don't know who did that.'
b. ... wos dass ma toa soin.
what that we do should
'... what we should do.'
c. ... wann dass da Xaver kumt
when that the comes
'... when Xaver is coming.'
d. ... wiavui dass a kriagt.
how much that he gets
'... how much he gets.'
(11) Dutch (den Besten 1989:23, (21b))

... welk boek (of) hij wil lezen

which book if he wants read
'... which book he wants to read'

Case checking

A further argument for wh- movement, similar to the one based on complementation, can be
constructed on the basis of case theory. Consider the contrast in (12), which reflects the fact that the
finite I in the complement clause checks nominative, not objective case.

(12) a. She thinks he will come.

 b. * She thinks him will come.

Now consider the contrast in (13), where the nominative form who is grammatical in standard
English, but the objective form whom is not.

(13) a. I wonder who she thinks will come.

b. * I wonder whom she thinks will come.

How is case checked on the nominative form who in (13a)? We would like to maintain the
generalization that nominative case is checked in a spec-head configuration with finite I. This
rules out the both of the closest I nodes. As shown in (14), who is not in the required spec-head
configuration with either of them; rather, these nodes check nominative case with they and she.

(14) They [pres] wonder [CP whoi [IP she [pres] thinks [CP [IP will come ] ] ] ] .

The solution to this case-checking puzzle is once again to invoke movement. As shown in (15),
nominative case on who can be checked by will before who moves to the Spec(CP) of the
intermediate clause. In this way, every nominative noun phrase in the sentence is checked in the
right configuration, and a one-to-one relation is maintained between case features and case
licensers.

(15) They [pres] wonder [CP whoi [IP she [pres] thinks [CP [IP ti will come ] ] ] ] .

The reason that (13b) is ungrammatical is that objective case cannot be checked in the subject
position of the lowest clause. Under a movement analysis, then, the contrast in (13) is parallel to
that in (12) - a simple and intuitively appealing result.

The facts just discussed illustrate an important difference between wh- movement on the one
hand and subject raising and passive on the other. In subject raising and passive, the noun phrase
undergoing movement originates in a non-case-checking position and moves to a subject
position, motivated either by considerations of case-checking or by the subject requirement. By
contrast, a noun phrase undergoing wh- movement moves from a case-checking position to
Spec(CP), a non-case-checking position.

A related difference concerns the categorial status of constituents undergoing wh- movement.
Constituents undergoing subject movement or passive are noun phrases, but ones undergoing
wh- movement can be maximal projections of other categories, as we have already seen. (16)
provides three further examples.

(16) a. I wonder [AdjP how experienced ] they should be.

 b. I wonder [AdvP how quickly ] the lions will devour the wildebeest.

c. I wonder [PP under which shell ] he hid the pea.

Direct wh- questions

Having argued that wh- phrases move to Spec(CP) in indirect questions, we assume for uniformity that
direct questions like those in (17) are derived by wh- movement as well.

(17) a. [AdjP How experienced ] should they be?

b. [AdvP How quickly ] will the lions devour the wildebeest?

c. [PP Under which shell ] did he hide the pea?

d. [DP Which wildebeest ] will the lions devour?

As is evident from comparing the direct questions in (17) with their indirect question
counterparts, movement of the wh- phrase to Spec(CP) in direct questions is accompanied by a
second instance of movement. In particular, whatever occupies I moves to C, with the additional
proviso that if I is occupied by a silent tense morpheme, as it is in (16c), it is replaced by the
corresponding form of do, as in (17c). We postpone more detailed discussion of head movement
to C to a later chapter. For the moment, we will simply assume that direct questions are
projections of a silent morpheme that occupies C and that expresses interrogative force,
represented in what follows by [?]. The structure that we assume for direct questions is
illustrated for (17d) in (18). As usual with head movement (recall the case of verb movement
from V to I discussed in Chapter 6), head movement to C involves both movement and
adjunction.

(18)

Constraints on wh- movement

The island constraints

Apparent unboundedness of wh- movement. Given that wh- phrases in direct and indirect
questions occupy their surface position as a result of movement, the question arises of how far a
wh- phrase can move from the position where it is interpreted. Examples like (19) suggest that
the distance is in principle (that is, apart from performance considerations such as limitations on
memory) unlimited, or unbounded. Examples like (19b-e), where a wh- phrase moves out of the
CP where it originates, are called long-distance wh- movement (also known as long movement
or nonlocal movement).

(19) a. [CP Whati was he reading ti ] ?

b. [CP Whati did he say

[CP that he was reading ti ] ] ?

c. [CP Whati does she believe

[CP that he said

[CP that he was reading ti ] ] ] ?

d. [CP Whati are they claiming

[CP that she believes

[CP that he said

[CP that he was reading ti ] ] ] ] ?

e. [CP Whati do you think

[CP that they are claiming

[CP that she believes

[CP that he said

[CP that he was reading ti ] ] ] ] ] ?

A typology of islands. However, contrary to what the pattern in (19) suggests, Ross 1967 argued
that wh- movement is not in fact unbounded. For instance, although wh- movement out of that
clause complements to verbs is completely acceptable, as shown in (19b-e), wh- movement out
of that clause complements to nouns is not, as shown in (20) and (21). For clarity, the heads
associated with the complement clauses are underlined in the next few examples.

Noun complement:
(20) a. He made the claim [ that he has met Subcomandante Marcos ] .

b. * [ Who ]i did he make the claim [ that he has met ti ] ?

(21) a. He mentioned the fact [ that he had run into Julia Roberts ] .

b. * [ Which celebrity ]i did he mention the fact [ that he had run into ti ] ?

Particularly striking is the contrast between (20b) and (21b) on the one hand and the essentially
synonymous examples in (22) on the other.

(22) a. [ Who ]i did he claim [ that he has met ti ] ?

b. [ Which celebrity ]i did he mention [ that he had run into ti ] ?

Ross introduced the term island to refer to constructions that do not allow a wh- phrase to
'escape' from them (that is, metaphorically speaking, the wh- phrase is marooned on the island).
Besides complement clauses to nouns, Ross identified several other types of islands:

indirect questions,
relative clauses,1
sentential subjects,
possessive noun phrases, and
coordinate structures.

In the remainder of this section, we illustrate each of these types of island in turn. Our aim for the
moment is purely descriptive. In other words, the internal structure of the islands is not our primary
focus, and we postpone the obvious question of why islands have the effect on wh- movement that they
have until a later chapter. For clarity, the islands are indicated by underlining.

(23) illustrates the island character of indirect questions.

Indirect question:

(23) a. They have forgotten which problem they should solve by Fourier analysis.

b. * Howi have they forgotten which problem they should solve ti?

Be sure to interpret How in (23b) as modifying the complement verb solve, as indicated by the
trace, not the matrix verb forgotten. In other words, a possible answer to (23b) is by Fourier
analysis, but not by succumbing to Alzheimer's.
In (23b), it is important to distinguish the two instances of wh- movement: that of which problem
and that of how. Which problem moves from its original position as complement of solve to the
Spec(CP) of the complement clause. It is this movement - grammatical on its own - that creates
an island for any further wh- movement, preventing how from moving "off island" to the
Spec(CP) of the matrix clause. In other words, wh- movement is grammatical within the confines
of an island, but not beyond its boundaries.2

(24) and (25) illustrate the island character of relative clauses and sentential subjects.

Relative clause:

(24) a. They met someone who knows Julia Roberts.

b. * [ Which celebrity ]i did they meet someone who knows ti?

Sentential subject:

(25) a. That he has met Subcomandante Marcos is extremely unlikely.

b. * Whoi is that he has met ti extremely unlikely?

Finally, (26) and (27) illustrate the island character of possessive noun phrases and of coordinate
structures. The ungrammaticality of these questions is particularly striking because they are so
much shorter than the grammatical questions in (19b-e).

Possessive noun phrase:

(26) a. She bought Jonathan's book.

b. * [ Whose ]i did she buy ti book?

Coordinate structure:

(27) a. They ordered tiramisu and espresso .

b. i. * [ Which dessert ]i did they order ti and espresso?

ii. * [ Which beverage ]i did they order tiramisu and ti?

The Comp-trace effect

Another constraint on wh- movement is the so-called Comp-trace effect. Notice first that long-distance
wh- movement of complements and adjuncts is unaffected by whether the complement clause is
headed by an overt complementizer or a silent one (indicated in the following examples by ø).

(28) a. [ Which friends ]i did they say { that, ø } they saw ti ?

b. [ Which way ]i did they say { that, ø } they would fix the leaky faucet ti ?

By contrast, long-distance movement of subjects is possible only with a silent complementizer.

The presence of an overt complementizer immediately preceding the trace of wh- movement is
ungrammatical; hence the name of the effect.

(29) a. [ Which friends ]i did they say ø ti saw them?

b. * [ Which friends ]i did they say that ti saw them?

There is some variation among English speakers with regard to the status of (29b) (Sobin 1987).
But even speakers who judge (29b) to be acceptable report a Comp-trace effect in connection
with movement out of indirect questions. As we have just seen, indirect questions are islands,
and so long wh- movement is not completely acceptable to begin with. However, it has been
observed that indirect questions introduced by whether or if tend to give rise to relatively weak
island effects; in the examples in (30), this weak effect is indicated by ?*.

(30) a. ?* [ Which friends ]i did they worry whether they snubbed ti ?

b. ?* [ Which way ]i did they wonder if they could fix the leaky faucet ti ?

Analogous long subject movement is illustrated in (31), which is worse than (29b) and (30) for
all speakers and thus provides evidence for the existence of a Comp-trace effect even for those
speakers who accept (29b).

(31) * [ Which friends ]i did they worry whether ti snubbed them?

Relative clauses as instances of wh- movement

Wh- relative clauses

As (32) and (33) show, there is a striking parallel in English between questions and wh- relative clauses:
both are introduced by wh- phrases.
(32) a. Who moved in next door? (33) a. the people who moved in next door

b. Who(m) did you see? b. the people who(m) you saw

c. Where did you meet them? c. the place where you met them

d. Which do you prefer? d. the movie which you prefer

e. Whose parents did you meet? e. the girl whose parents you met

This parallel suggests a wh- movement analysis for relative clauses - an idea that is reinforced by
the fact that wh- relative clauses exhibit the entire range of island effects, as illustrated in (34).
As before, the relevant islands are underlined.

In (34c), be careful to distinguish the two instances of wh- movement: the lower one within the
interviewed clause, which is grammatical and creates the relative clause island, and the higher
one within the dislike clause, which is the one that causes the ungrammaticality.

(34) a. Noun complement: * the revolutionary whoi I don't believe the claim that he has met ti

b. Indirect question: * the method [ by which ]i they have forgotten which problem they
should solve ti

c. Relative clause: * the revolutionary whoi I dislike the journalist who interviewed ti for
CNN

d. Sentential subject: * the addiction whichi that he admitted ti nearly destroyed his career

e. Possessive noun * the girl whosei you met ti parents

phrase:

f. Coordinate structure: * the dessert whichi you ordered espresso and ti

* the dessert whichi you ordered ti and espresso

Relative clauses also exhibit the Comp-trace effect. The examples in (35)-(38) are parallel to
those in (28)-(31).

(35) a. the friends who(m)i they said { that, ø } they saw ti

b. the way whichi they said { that, ø } they would fix the leaky faucet
 ti

(36) a. the friends whoi they said ø ti saw them

b. * the friends whoi they said that ti saw them

(37) a. ?* the friends who(m)i they worried whether they snubbed ti

b. ?* the way whichi they wondered if they could fix the leaky faucet ti

(38) * the friends whoi they worried whether ti snubbed them

The facts just reviewed follow straightforwardly if we assume that wh- relative clauses are
structurally parallel to questions. The wh- relative pronoun moves to Spec(CP), and the syntactic
head of the clause is a silent complementizer, just as in an indirect question. The structure of the
relative clause in (33b) is given in (39a), and adjoining the relative clause so that it modifies the
noun people yields the structure in (39b).

(39) a. b.

That relative clauses

In addition to wh- relative clauses, English also has that relative clauses, as illustrated in (40).

(40) a. the people that moved in next door

b. the people that you saw

c. the place that you met them

d. the movie that you prefer

Structurally, that relative clauses are completely parallel to wh- relative clauses. But in contrast
to wh- relative clauses, it is the complementizer that is overt in that relative clauses, and the wh-
phrase that is silent.3 The structures corresponding to those in (39) are given in (41).

(41) a. b.

The alert reader will have noticed that that relative clauses with a gap in subject position do not
exhibit the Comp-trace effect. Thus, the local wh- movement in (40a), with the structure in (42a),
contrasts with the nonlocal wh- movement in (42b). In both cases, the gray font indicates the
silent wh- phrase.

(42) a. the people [CP whoi that ti moved in next door ]

b. * the people [CP whoi that I think [CP that ti moved in next door ] ]

This contrast between local and nonlocal movement has given rise to many attempts at
explanation, some of which we will review in a later chapter, but none that we know of is
entirely satisfactory.

Doubly marked relative clauses

Given the discussion so far, we would expect to find relative clauses with an overt wh- element in
Spec(CP) combined with an overt complementizer, as in (43).

(43) the people who(m) that you saw

Such doubly marked relative clauses are judged to be unacceptable in modern standard English.
However, just like doubly marked indirect questions, they are attested in Middle English and in
vernacular varieties of other languages, as shown in (44) (from the Penn-Helsinki Parsed Corpus
of Middle English) and (45).
(44) a. thy freend which that thou has lorn (cmctmeli.m3, 218.C1.31)
'your friend that you have lost'

b. the conseil which that was yeven to yow by the men of lawe and the wise folk (cmctmeli.m3,
226.C2.373)
'the counsel that was given to you by the men of law and the wise folk'

c. the seconde condicion which that the same Tullius addeth in this matiere (cmctmeli.m3,
228.C1.429)
'the second condition that the same Tullius adds in this matter'

d. for hire olde freendes which that were trewe and wyse (cmctmeli.m3, 237.C2.799)
'for her old friends who were loyal and wise'

e. the fire of angre and of wratthe, which that he sholde quenche (cmctpars.m3, 308.C2.859)
'the fire of anger and wrath, which he should quench'

Bavarian (Bayer 1983-4:213, (10a,b))

(45) a. der Hund der wo gestern d' Katz bissn hod

the dog who that yesterday the cat bitten has
'the dog that bit the cat yesterday'
b. die Frau dera wo da Xaver a Bussl g'gem hod
the woman who.dat that the a kiss given has
'the woman that Xaver gave a kiss'

What these facts suggest is that doubly marked wh- movement constructions (both relative
clauses and indirect questions) are grammatical (= well-formed from a purely structural point of
view). However, it may be that the movement of the wh- phrase and the overt complementizer
are both taken as markers of clausal subordination, and perhaps a stylistic constraint has
developed against redundant marking that keeps these constructions from occurring in the
modern literary varieties of English and other languages.4

Zero relative clauses

Given the availability of silent wh- elements and silent complementizers in English, we would expect to
find relative clauses that are not introduced by any overt element at all. Such zero relative clauses (also
known as contact relative clauses) are indeed possible in English, as shown in (46).

(46) a. the people ___ you saw

b. the place ___ you met them

c. the movie ___ you prefer

Given the grammaticality of (46), the status of the zero relative variants of subject relative
clauses like (33a) and (38a), repeated in (47a,b), is puzzling. In general, these are unacceptable,
as shown in (47c).

(47) a. the people who moved in next door (are from Illinois)

b. the people that moved in next door (are from Illinois)

c. * the people ___ moved in next door (are from Illinois)

However, structurally analogous examples are attested in English, as illustrated in (48).

(48) a. Everybody ___ lives in the mountains has an accent all to theirself. (Christian and Wolfram
1976, front matter)

b. Three times a day some nurse ___ looks like Pancho Villa shoots sheep cum into my belly.
(Hiaasen 1995: 248-249)

(49) and (50) give further examples from Belfast English (Henry 1995:125) and American
English, classified by linguistic environment.

(49) a. Existential there clause: There's a shortcut ___ takes you to the shops.

b. It cleft: It was John ___ told us about it.

c. Copular construction: John is the person ___ could help you with that.

d. Introduction of discourse I met a man ___ can speak five languages.

entity in object position:

(50) a. Existential there clause: "Thanks for the hurricane, there's a hundred fifty thousand houses
in Dade County ___ need new roofs," he began. (Hiaasen
1995:110-111)

b. It cleft: 'Tis grace ___ hath brought me safe thus far (Amazing grace)

c. Copular construction: You're the second guy this month ___ wants to take out trade in
this bizarre fashion. (Wagner 1986:119)

He's the one ___ inspected the damn things. (Hiaasen 1995:5)
 d. Introduction of discourse how come we have … a pink-haired punk granddaughter ___ got
entity in object position: the manners of a terrorist? … Wears somethin' ___ makes the
garage door flap up? (Wagner 1986:81)

The proper analysis of zero subject relative clauses is not entirely clear and goes beyond the
scope of this textbook. On the one hand, it has been argued that such relative clauses are
grammatical, but avoided for processing reasons (Bever and Langendoen 1971, especially
Section 5; see also Doherty 1993). The idea is that zero subject relative clauses are liable to be
misinterpreted as the predicate of the matrix clause, at least when they modify subjects, as they
do in (48).5 On the other hand, Henry 1995 argues that zero subject relative clauses are not true
relative clauses.

Notes

1. Ross 1967 treated relative clauses together with noun complement clauses, subsuming the
noun phrases containing them under the rubric of complex noun phrases. Given our emphasis on
the complement-adjunct distinction, we prefer to distinguish the two types of island.

2. In more precise structural terms (discussed in more detail in a later chapter), an island is a
constituent dominated by some node. In (23), the node in question is the CP dominating the
indirect question, but for generality, let's refer to the relevant node as the island node. Wh-
movement is grammatical as long as the moved wh- phrase (the head of the movement chain) is
dominated by the island node, but becomes ungrammatical once that is no longer the case.

3. For some speakers, relative clause that has developed from a complementizer into a relative
pronoun. Such speakers produce a possessive form of that that is analogous to whose, as
illustrated in (i).

(i) "... we want them to bring a product to market that's time had not yet come," said Ray Farhung, a
Southern California Edison official.
(Bill Vlasic, "Cool Contest", The Detroit News, p. 1D, January 10, 1993; from a Linguist List posting
by John Lawler)

4. It is clear that stylistic constraints of the sort that we postulate exist. For instance, stylistically
"higher" contexts favor the presence of an overt complementizer in contexts where the grammar
does not require one (Kroch and Small 1978).

(i) I think (that) they are coming.

The modern literary language apparently favors marking of subordinate clauses that is explicit
yet non-redundant.
5. In the sentence-processing literature, constructions that invite misparsing are known as
garden-path constructions. A classic example, involving a so-called reduced relative clause, is
(i.a). The corresponding zero subject relative clause, shown in (i.b.), would include a form of the
passive auxiliary be.

(i) a. The horse raced past the barn fell.

b. The horse was raced past the barn fell.

Exercises and problems

Exercise 11.1

You do not need to indicate the internal structure of your friend.

A. Using the xbar 4 grammar tool, build structures for both of the direct questions in (1).

(1) a. Pied piping In which house does your friend live?

b. Preposition stranding Which house does your friend live in?

B. Using the xbar 4 grammar tool, build structures for the complex sentences in (2).

(2) a. Pied piping I forget [ in which house your friend lives. ]

b. Preposition stranding I forget [ which house your friend lives in. ]

Exercise 11.2

You do not need to indicate the internal structure of your friend.

A. Using the xbar 4 grammar tool, build structures for the noun phrases in (1).

(1) a. Pied piping the house [ in which your friend lives ]

b. Preposition stranding the house [ which your friend lives in ]

B. Using the same grammar tool as in (A), build structures for the noun phrases in (2)-(5).

Build structures for all examples, including the ungrammatical ones.

Feel free to build a single structure for several structurally parallel examples. Indicate
clearly which sentences each structure is intended to represent, and how the variants
differ.

Treat whose as the conventional orthographic representation of who + 's.

In (3), the position of silent elements is indicated by underlining. In the other examples,
the position of silent elements is not explicitly indicated.
(2) a. Wh- relative clause ok the guy [ who they met ]

b. That relative clause ok the guy [ that they met ]

c. Zero relative clause ok the guy [ they met ]

(3) a. Wh- relative clause ok the guy [ whose parents ___ they met ]

b. That relative clause * the guy [ ___'s parents that they met ]
Intended meaning: (3a)

c. Zero relative clause * the guy [ ___'s parents ___ they met ]
Intended meaning: (3a)

(4) a. That relative clause, pied piping * the house [ in that your friend lives ]
Intended meaning: (1a)

b. That relative clause, preposition stranding ok the house [ that your friend lives in ]

(5) a. Zero relative clause, pied piping * the house [ in your friend lives ]
Intended meaning: (1a)

b. Zero relative clause, preposition stranding ok the house [ your friend lives in ]

C. Formulate a single structural generalization that accounts for the ungrammaticality of (3b,c),
(4a), and (5a). The generalization is independent of the constraints introduced in the chapter.
Hint

Exercise 11.3
According to the analysis of relative clauses presented in this chapter, that relative clauses like the
italicized sequence in (1) have the structure in (2a). However, since the wh- phrase is silent, an
alternative analysis of (1) is possible in principle, according to which the wh- phrase remains in its
original position, as shown in (2b). Provide evidence that the movement analysis in (2a) is preferable.

(1) the people that you saw

(2) a. b.

Movement No movement

Exercise 11.4

Make up silent lexical items as needed for (1).

In (2), you can ignore the initial PP In "What Women Want".

Free free to build tree fragments rather than full structures as long as you describe how
they fit together.

For each of the following structurally ambiguous sentences, provide a paraphrase for each
reading. Using the xbar 4 grammar tool, build structures for each of the interpretations,
indicating clearly which structure goes with which interpretation. (1) is taken from Pinker
1994:119). (2) was supplied by a student in this class from an Internet movie database. (3) is
from http://www.slate.com/id/76886/, accessed 8 April 2010.

(1) a. Reporter's telegram: How old Cary Grant?

 b. Grant's reply: Old Cary Grant fine.

(2) In "What Women Want," Mel Gibson plays a man who develops the ability to understand what
women are thinking after a freak accident.

(3) I remember meeting a mother of a child who was abducted by the North Koreans right here in
the Oval Office.

Exercise 11.5

Using the xbar 4 grammar tool, build the structure for (1).

(1) It's the kind of thing that people who like that kind of thing like.

Exercise 11.6

A. In addition to the finite indirect questions in (1), English also has nonfinite ones, as illustrated in (2).
Using the xbar 4 grammar tool, build structures for all of the nonfinite indirect questions in (2), including
the ungrammatical (2c).

(1) a. They know [ who they should invite. ]

b. They know [ which topic they should talk about. ]

c. They know [ who should speak. ]

(2) a. They know [ who to invite. ]

b. They know [ which topic to talk about. ]

c. * They know [ who to speak. ]

Intended meaning: (1c)

B. Why is (2c) ungrammatical?

Exercise 11.7

Using the xbar 4 grammar tool, build the structure for the noun phrase containing the
nonstandard relative clause in (1).

(1) a product that's time has come

Problem 11.1

In addition to finite relative clauses, English also has nonfinite ones, as illustrated in (1)-(3).

(1) a. a guy [ for us to hire ] cf. a guy [ that we can hire ]

b. * a guy [ us to hire ] cf. a guy [ we can hire ]

c. a guy [ to hire ]

(2) a. an office [ in which to work ] cf. an office [ in which they can work ]

b. * an office [ in ___ to work ] cf. * an office [ in ___ they can work ]

c. an office [ which to work in ] cf. an office [ which they can work in ]

d. an office [ ___ to work in ] cf. an office [ they can work in ]

(3) a. a guy [ to fix the sink ] cf. % (I know) a guy [ can fix the sink ]

b. * a guy [ who to fix the sink ] cf. a guy [ who can fix the sink ]

On the basis of the above examples and others of your own devising, compare the syntax of
nonfinite relative clauses with that of their finite counterparts. The following are examples of
questions to consider:

How does the distribution of overt wh- phrases differ in finite and nonfinite relative clauses?
In what respects is the distribution of overt wh- phrases independent of finiteness?
Why are the nonfinite relative clauses in (1b), (2b), and (3b) ungrammatical?

12 Subjacency and the ECP

Old version (Fall 2006) - version for S07 unchanged

Subjacency
o Two possible derivations for long-distance wh- movement
o IP as a barrier to wh- movement
o DP as a barrier to wh- movement
o The coordinate structure constraint revisited
The Empty Category Principle (ECP)
o Antecedent government
o Lexical government
Further issues and refinements
o Is subjacency an independent principle?
o Movement out of ECM complements
o Movement out of DP
 Exercises and problems
Supplementary material
o Node relations

The motivation for formulating constraints on wh- movement has always been to identify as wide-
reaching and general principles of the human language faculty as possible. Thus, in the early 1960s, even
before Ross's discovery of the island constraints, Chomsky proposed the so-called A-over-A constraint.
According to this constraint, wh- movement is ungrammatical out of recursive structures in which one
instance of a category immediately dominates another, as indicated in (1).
A-over-A constraint:

(1) * XPi ... [A ... [A ... ti ... ] ... ]

Given then-current assumptions concerning phrase structure (X' theory hadn't yet been
introduced), the A-over-A constraint is illustrated by the contrast in (2).

(2) a. [PP Up [PP to what age ] ]i can children learn a language without an accent ti ?

b. * [PP To what age ]i can children learn a language without an accent [PP up ti ] ?

Against the backdrop of what was known about wh- movement in the early 1960s, the island
constraints discussed in Chapter 11 represented significant empirical progress. But from a more
theoretical point of view, the island constraints are simply a list of stipulations, and they sharply
raise the question of whether it is possible to reduce them to fewer, deeper structural principles
(ideally, a single such principle). In this chapter, we discuss two influential proposals with this
aim: the subjacency condition and the Empty Category Principle (ECP). As we will see, the
reduction of the constraints on wh- movement to more general principles has proven a
considerable challenge to syntactic theory, and one that persists to the present day. In particular,
no satisfactory overarching framework has yet been found that subsumes the entire range of
island constraints. The island constraints themselves therefore remain as an empirical benchmark
against which to measure any theoretical proposal concerning constraints on wh- movement.

Subjacency

In this section, we present the subjacency condition, an ambitious attempt by Chomsky 1973 to
subsume the island constraints under a single structural principle.

Two possible derivations for long-distance wh- movement

When we consider examples of long-distance wh- movement like those in (3), two possible
derivations come to mind.
(3) What did he say that he was reading?

On the one hand, the wh- phrase might move from the position in which it is interpreted,
however deeply embedded that is, to the sentence-initial Spec(CP) position in one fell swoop,
yielding a wh- movement chain with two links, as in (4). On the other hand, wh- movement
might take place in more than one step. The first step takes the moved constituent from its
original position to the nearest Spec(CP), and each subsequent step takes it to the next higher
Spec(CP). This derivation of (3), which involves two steps and yields a wh- movement chain
with three links, is shown in (5).

(4) (5)

The derivation in (5) is known as a cyclic derivation (the idea being that each successively higher
clause (= CP) forms a separate cycle in the derivation of the entire sentence), and the derivation
in (4) is accordingly known as noncyclic. Notice that the noncyclic and cyclic derivations in (4)
and (5) differ in the presence of an intermediate trace, which is highlighted by a box in (5).

IP as a barrier to wh- movement

On the basis of grammatical instances of long-distance wh- movement like (6a-d) (cf. (19c-e)) of Chapter
11, it is impossible to decide which of the two alternatives just presented is correct, or even whether a
choice must be made between them.

Parentheses indicate intermediate traces that are posited in a cyclic, but not in a noncyclic,
derivation.

(6) a. What-i did he say [ (t-i) that he was reading t-i ? ]

 b. What-i does she believe [ (t-i) that he said [ (t-i) that he was reading t-i ? ] ]

c. What-i are they claiming [ (t-i) that she believes [ (t-i) that he said [ (t-i) that he was reading t-i ? ]
]]

d. What do you think [ (t-i) that they are claiming [ (t-i) that she believes [ (t-i) that he said [ (t-i) that
he was reading? ] ] ] ]

However, the existence of syntactic islands forces us to choose the cyclic alternative. For
instance, consider the ungrammatical question in (7a) (= (23b) of Chapter 11 and its cyclic
derivation in (7b).

Remember to interpret how in (7a) as modifying solve, not forgotten, as indicated by the lowest
trace.

(7) a. * Howi have they forgotten [ which problem they should solve ti ] ?

b.

If wh- movement were able to occur in one fell swoop, then there would be nothing to stop long-
distance wh- movement in (7a), and the question should be grammatical, contrary to fact. But the
ungrammaticality of the question can be made to follow from the assumption that wh- movement
is cyclic. Specifically, let's assume that wh- movement is subject to the condition in (8), and that
IPs form barriers to movement, as indicated by the boxes in (7b).

Subjacency condition:
(8) In a chain formed by movement, the path connecting two neighboring links must not contain more
than one barrier (in other words, on the path between A and B, there is at most one barrier C such
that A c-commands C and C dominates B).

The condition in (8) has the consequence that a wh- constituent can move out of an IP that
dominates it just in case an empty local Spec(CP) is available or can be generated as an
intermediate landing site. By local Spec(CP), we mean the specifier of a CP whose head is a
sister of the IP in question. In the absence of such a landing site, as in (7b), wh- movement is
correctly ruled out as ungrammatical.

Notice that the ungrammaticality of (7a) depends on the indirect question containing two wh-
phrases: how and which problem. The representation in (7b) assumes that which problem moves
before how does (note the order of the indices), thereby preventing the complement Spec(CP)
from serving as an escape hatch for how. It is also necessary to rule out an alternative
derivation, according to which the constituent that moves first is how. In this case, the
complement Spec(CP) is empty, and how can move through it up to the matrix Spec(CP), as
shown in (9). Notice that only one IP barrier (indicated in green) intervenes between any pairs of
links in the movement chain, so that this part of the derivation does not violate the subjacency
condition in (8).

(9) a. [IP they have forgotten [CP howi [IP they should solve which problem ti ]

b. [CP howi have [IP they forgotten [CP ti [IP they should solve which problem ti ]

But now the intermediate trace of how blocks the movement of which problem into the lower
Spec(CP). As the contrast in (10) shows, this movement is necessary for the complement clause
to be interpreted properly as an indirect question. As a result, the derivation begun in (9) fails,
and (7a) continues to be ruled out as desired.

(10) a. ok They have forgotten which problem they should solve.

b. * They have forgotten they should solve which problem.

DP as a barrier to wh- movement

If the only barriers in English were IPs, then wh- movement out of noun complements and left
branch structures like those in (11) (= (20b) and (26b) of Chapter 11) would be expected to be
grammatical, contrary to fact. The structures are given in (11); the green IP nodes are intended to
indicate that the derivation is consistent with subjacency under the (incorrect) assumption that
the only barriers are IPs.
(11) a. * Whoi did [IP he make [DP the claim [CP ti that [IP he has met ti ] ] ] ] ?

b. * Whosei did [IP she buy [DP ti book ] ] ?

However, the ungrammaticality of (11) can be derived straightforwardly if the set of barriers in
English includes not only IPs, but also DPs. This is illustrated in (12), where red indicates
barriers that cause subjacency to be violated. As before, green indicates barriers that are
consistent with subjacency.

In determining whether subjacency is violated, we can consider the relevant movement chains in
either top-down or bottom-up fashion. Here and in what follows, we have chosen bottom-up.
This choice has the consequence that the barriers that cause subjacency to be violated in (12)
are IPs, not DPs. This doesn't mean, though, that DPs aren't barriers! If they weren't, we'd be
back to the representation in (11).

(12) a. * Whoi did [IP he make [DP the claim [CP ti that [IP he has met ti ] ] ] ] ?

b. * Whosei did [IP she buy [DP ti book ] ] ?

Given the theoretical character of the subjacency condition, it should come as no surprise that it
has an empirical consequence that goes beyond the range of facts that it was intended to explain.
Specifically, it leads one to expect any movement out of a noun phrase, not just movement of a
left branch, to be ungrammatical. (13b) gives an example that behaves as expected.

(13) a. He dropped a book about information theory.

b. * Whati did [IP he drop [DP a book about ti ] ] ?

Notice, by the way, that the unacceptability of (13b) cannot be attributed to preposition
stranding, since the pied piping counterpart of (13b) remains unacceptable, as expected given the
structure in (14).

(14) * [ About what ]i did [IP he drop [DP a book ti ] ] ?

As we will see later on in our discussion of the ECP, the predictions made by subjacency in
connection with movement out of DPs, while correct in the case of (13b) and (14), are too strict.
In other words, subjacency (as it stands, with all instances of IP and DP as barriers) incorrectly
rules out grammatical instances of wh- movement.

The coordinate structure constraint revisited

In the previous section, you may have noticed the omission of a potential piece of evidence for the
barrierhood of DP - namely, violations of the coordinate structure constraint like those in (15) (= (27b,c)
of Chapter 11).

(15) a. * [ Which dessert ]i did they order [ ti and espresso ] ?

b. * [ Which beverage ]i did they order [ tiramisu and ti ] ?

Although the internal structure of coordinate phrases is not well understood, it is clear that a
coordinate noun phrase like tiramisu and espresso is a recursive structure consisting of a DP that
dominates two further DPs, as in (16).

(16) [DP [DP tiramisu ] and [DP espresso] ]

Given this structure, the ungrammaticality of the examples in (16) follows from subjacency, as
shown in (17).

(17) a. * [ Which dessert ]i did [IP they order [DP ti and espresso ] ] ?

b. * [ Which beverage ]i did [IP they order [DP tiramisu and ti ] ] ?

However, not all examples that violate the coordinate structure constraint also violate
subjacency. For instance, in (18) and (19), where syntactic categories other than DP are
coordinated, only a single IP barrier intervenes between the moved phrase in Spec(CP) and its
trace.

Take care to read the (b) examples without an intonation break before the conjunction.

(18) a. [PP [PP On which day ] and [PP in which year ] ]i were [IP you born [PP ti ] ] ?

b. * [PP On which day ]i were [IP you born [PP [PP ti ] and [PP in which year ] ] ] ?

c. * [PP In which year ]i were [IP you born [PP [PP on which day ] and [PP ti ] ] ] ?

(19) He said he would get out the vote and win the election, and ...

a. [VP [VP get out the vote ] and [VP win the election ] ]i, [IP he did [VP ti ] ] .

b. * [VP get out the vote ]i, [IP he did [VP [VP ti ] and [VP win the election ] ] ] .
 c. * [VP win the election ]i, [IP he did [VP [VP get out the vote ] and [VP ti ] ] ] .

A further example of this sort is shown in (20b).

(20) a. They have [VP [VP peeled the cucumbers ] and [VP chopped up the onions ] ] .

b. * [ Which vegetables ]i have [IP they [VP [VP peeled ti ] and [VP chopped the onions ] ] ] ?

As (21) shows, (20b) violates a parallelism constraint known as the across-the-board (ATB)
constraint, according to which movement of a constituent out of a coordinate structure must
affect all co-conjuncts simultaneously.

(21) [ Which vegetables ]i have [IP they [VP [VP peeled ti ] and [VP chopped ti ] ] ] ?

The exceptional behavior of coordinate structures with regard to subjacency (expected to be

grammatical, yet in fact not so) and the additional special restriction imposed upon them (the
ATB constraint just mentioned) strongly suggest that the coordinate structure constraint is sui
generis, and that it should not be grouped together with the other island constraints. Accordingly,
most proposals to reduce the island constraints to more general principles make no attempt to
include the coordinate structure constraint, and we, too, will make no further mention of it.

The Empty Category Principle (ECP)

The island constraints and the subjacency condition that subsumes them correctly account for much,
but not all, of the spectrum of relevant facts concerning wh- movement. For instance, Ross himself
observed that it is more acceptable to move complements out of indirect questions than it is to move
adjuncts. Subjects (also non-complements) behave like adjuncts, giving the pattern in (22).

(22) a. ? [ Which problem ] ]i have they forgotten howj they should solve ti tj?

b. * Howi have they forgotten [ which problem ]j they should solve tj ti?

c. * [ Which problem ]i have they forgotten [ how ]j ti should be solved ti tj?

In order to account for the contrast between complements and non-complements with respect to
long-distance wh- movement, it has been proposed that traces of movement must satisfy a
condition distinct from subjacency, the so-called Empty Category Principle (ECP). Early
formulations of the ECP (Aoun, Hornstein, and Sportiche 1982, Huang 1982, Lasnik and Saito
1984) were disjunctive; that is, they consisted of two mutually exclusive conditions. In the
course of the 1980s, attempts were made to reformulate the ECP in conjunctive terms; that is, as
two conditions that traces of movement must satisfy simultaneously. In our view, these attempts
have not been successful, because they continue to impose different conditions on the movement
of complements and of non-complements. That is, even if the ECP itself is no longer formulated
in a disjunctive way, the disjunction it contained is not resolved, but simply appears elsewhere in
the proposals in question. For instance, Rizzi 1990 distinguishes two ways of establishing a
legitimate antecedent-trace relation, one for complements, and one for non-complements. We
will therefore continue to maintain a disjunctive version of the ECP, specifically (23).

Empty Category Principle (ECP):

(23) a. A trace of movement must be properly governed.

b. A trace of movement is properly governed iff

i. it is antecedent-governed, or

ii. it is lexically governed.

Antecedent government

We discuss the two conditions on traces in turn. The notion of antecedent government is
defined in (24) (see Node relations for a definition of binding).

Antecedent government:

(24) A antecedent-governs B iff

i. A binds B, and

ii. at most one barrier intervenes on the path between A and B.

As is evident, clause (ii) of the definition of antecedent government recapitulates the subjacency
condition, and it is easy to see that the antecedent government clause of the ECP therefore
enforces cyclic movement. This of course derives the ungrammaticality of (22b,c), but leaves the
relative acceptability of (22a) unexplained. However, as we will see directly, (22a), though
violating the antecedent government condition of the ECP, satisfies the alternative lexical
government condition.

Lexical government

The notion of lexical government to be presented in what follows relies on the concept of lexical
government domain (which is based on the concept of g(overnment)-projection proposed in
Kayne 1984). The term 'governor of XP' in (25a) refers to the head that stands in a head-comp
configuration with XP.1
 The term 'govern' is used here in a slightly different sense than it was in Chapter 8. There,
'govern' referred to a morphological requirement imposed by a head on a noun phrase
expressing one of the head's arguments. Here, the term 'govern' makes no reference to
morphology, but refers instead to a purely structural relation, the head-comp relation.

Lexical government domain:

(25) YP is a lexical government domain for XP iff YP is the maximal projection of

i. the governor of XP, or

ii. the governor of a lexical government domain for XP.

As is evident, the definition in (25) is recursive. We begin by considering the nonrecursive case
in (i), which is very simple. Consider the configuration in (26), where we take the DP
complement of the preposition as XP and the PP as YP.

(26)

In (26), the governor of the DP is the preposition, and so the PP (the preposition's maximal
projection) is a lexical government domain for the DP.

Now consider the more complex structure in (27), where we continue to take the DP complement
of the preposition as XP, but it is now VP that is the YP.

(27)

Is VP a lexical government domain for the lower DP? (25.i) is not met, since the head of VP, V,
does not govern the DP in question. But V does govern PP, which we determined to be a lexical
government domain for DP in (26). Therefore, VP is a lexical government domain for DP in (27)
by (25.ii).

Given (26) and (27), it is now easy to see that any lexical government domain can be extended
simply by substituting it as a complement of a higher head; that head's maximal projection is
then in turn a lexical government domain. Thus, all the structures in (28) are lexical government
domains for the lowest complement DP.
(28) a. b.

c. d. e.

Lexical government itself can then be defined as in (29).

Lexical government:

(29) A lexically governs B iff

i. A is a proper governor of B,

ii. there is an antecedent C that binds B, and

iii. C is contained in (= dominated by) a lexical government domain for B.

The notion of 'proper governor' in (29.i) is introduced in light of crosslinguistic contrasts like that
between (30) and (31).
(30) a. English Whoi did you talk with ti ?

b. Swedish Vem har du talat med ti ?

who have you talked with
(31) a. French * Qui as- tu parlé avec ti ?
who have you talked with
b. German * Wem hast du mit ti gesprochen?
who have you with talked

The idea is that only a proper subset of governors (= heads) is 'strong' enough to license a trace
by lexical government. Exactly which heads belong to the set of proper governors can vary by
language. For instance, prepositions are proper governors in English and Swedish, but not in
French or German. As a result, the traces in (30) are lexically governed, whereas those in (31)
are not, despite the analogous configurations in both cases. It should be emphasized that proper
government figures only in the definition of lexical government itself, not in the definition of the
concept of lexical government domain. So although the French preposition à 'to' cannot itself
license a trace, it can license the extension of a lexical government domain. This is shown by the
contrast in (32) (Kayne 1984:167). As in (30) and (31), green and red indicate heads that are and
are not proper governors, respectively.

(32) a. * Qu'i est-ce qu' elle tient à ti ?

what is it that she holds to
'What is she keen on?'
b. ok Qu'i est-ce qu' elle tient à faire ti ?
what is it that she holds to do
'What is she keen on doing?'

At first glance, the notions of lexical government and lexical government domain might seem to
permit any complement to undergo wh- movement. But the definition is more restrictive than
that. Consider, for instance, the contrast between the (b) examples in (33) and (34), where the
constituent undergoing wh- movement is the complement of admit in both cases.

(33) a. It can be difficult to admit one's weaknesses.

b. What can it be difficult to admit?

(34) a. To admit one's weaknesses can be difficult.

b. * What can to admit be difficult?

The structure of (33b) is given in (35); the successive lexical government domains for the DP
dominating the trace are indicated by boxes.
(35)

Since the maximal lexical government domain for the trace, the matrix CP, contains the trace's
antecedent (what), (33b) satisfies the lexical government clause of the ECP, and its
grammaticality is expected. Notice, incidentally, that lexical government is satisfied regardless of
the presence of an intermediate trace in the lower Spec(CP). That is, even if the trace were not
antecedent-governed, (33b) would satisfy the disjunctive version of the ECP assumed here.

By contrast, the lexical government domain for the wh- trace in the structure for (34b), given in
(36), extends only as far as the complement IP.

(36)

The reason that it extends no further is that the subject clause (the lower IP) is not governed (in
other words, it is not a complement). Since the infinitival subject clause does not dominate the
trace's antecedent (what), the trace of wh- movement, though governed by admit, is not lexically
governed. Moreover, two IPs intervene on the path between the trace and its antecedent, and so
the trace is not antecedent-governed either. (34b) is therefore correctly predicted to be
ungrammatical.

When there is an independent reason for the complement clause to be a CP (when it is an indirect
question, for instance), wh- movement within the bounds of the complement clause is possible,
as shown in (37).

(37) a. Which of one's weaknesses to admit can be a tricky question.

b.

We leave it open here why the complement clause in (36) is an IP rather than a CP.

Further issues and refinements

Is subjacency an independent principle?

Recall the facts that motivated the proposal of the ECP - namely, the contrast in (22), repeated here in
annotated form as (38).

(38) a. ? [ Which problem ] ]i have [IP they forgotten [CP howj [IP they should solve ti tj ] ] ] ?

b. * Howi have [IP they forgotten [CP [ which problem ]j [IP they should solve tj ti ] ] ] ?

c. * [ Which problem ]i have [IP they forgotten [CP howj [IP ti should be solved ti tj ] ] ] ?

Given the role that the notion of barrier plays in both antecedent government and subjacency, it
would be desirable to eliminate subjacency as a separate condition by subsuming it under the
antecedent government clause of the ECP. Is this feasible? In light of (38a), the answer must
unfortunately be 'no.' The matrix CP in (38a) is a lexical government domain for the trace (cf. the
configuration in (28e)), and in the absence of subjacency, (38a) should therefore be completely
acceptable. The degree to which it is not, then, provides evidence in favor of maintaining
subjacency as a separate constraint on wh- movement. The acceptability contrast between (38a)
and (38b,c) could then be attributed to the violation of only one principle in the former case, but
of two in the latter.
Movement out of ECM complements

Given their status as sisters of intermediate projections rather than of heads, subjects are not in a
position to be lexically governed. In order to satisfy the ECP, subject traces must therefore be
antecedent-governed. This is possible in cases of local movement (that is, movement to a local
Spec(CP)), but not in cases of true (= noncyclic) long-distance movement, as shown by the contrast in
(39).

(39) a. [ Which problem ]i should [IP ti be solved ti ] ?

b. * [ Which problem ]i have [IP they forgotten [CP howj [IP ti should be solved ti tj ] ] ] ?

However, if (39b) violates both clauses of the ECP, this raises the question of how subjects of
ECM complements are able to satisfy the ECP (or for that matter, subjacency), given that two
barriers intervene between the antecedent and its trace in both (39b) and (40).

(40) [ Which candidate ]i do [IP you expect [IP ti to get the job ] ] ?

A standard proposal is to include the head-spec configuration as an instance of proper

government, along with the head-comp configuration (Kayne 1984). This would allow the trace
in (40) to satisfy the lexical government clause of the ECP. But this proposal neither explains
why such examples are completely acceptable (suggesting that they satisfy both the ECP and
subjacency), nor why it is possible to move not just subjects out of ECM complements, but
adjuncts as well, as in (41).

(41) Howi were [IP you expecting [IP them to solve the problem ti] ] ?

An alternative that immediately comes to mind is simply that nonfinite IPs do not count as
barriers, but this proposal fails to account for the contrast in (42), which essentially parallels that
in (38).

(42) a. ? [ Which problem ]i have [IP they forgotten [CP howj [IP to solve ti tj ] ] ] ?

b. * Howi have [IP they forgotten [CP [ which problem ]j [IP to solve tj ti ] ] ] ?

However, a minor revision to the proposal that nonfinite IPs aren't barriers makes it empirically
adequate (though we know of no independent motivation for the revision). We will say that IPs
that are governed by V do not count as barriers. This has the result that the traces in the sentences
in (40) and (41) satisfy the antecedent government clause of the ECP as well as subjacency. (43)
shows our revised assumptions about these sentences; IPs that are not barriers are highlighted in
blue.
(43) a. [ Which candidate ]i do [IP you expect [IP ti to get the job ] ] ?

b. Howi were [IP you expecting [IP them to solve the problem ti] ] ?

Movement out of DP

As mentioned earlier, subjacency goes beyond the original island constraints in ruling out any
movement out of DP. However, in many cases, examples of such movement are completely
unexceptionable. This is puzzling given the representations in (44).

(44) a. Whati did [IP you { read, write } [DP a book about ti ] ] ?

b. Whoi did [IP you take [DP a picture of ti ] ] ?

c. [ { Which, how many } states ]i do [IP you know [DP the capitals of ti ] ] ?

(45) gives some further, naturally-occurring examples (the struck-out which in (45a) is included
for clarity; it is silent in the original).

(45) a. When I was a little boy, he teased me about a temporary but intense devotion I had to Gene
Autry, the singing cowboy--a devotion whichi [IP I would make [DP some lame attempt [CP ti [IP to
justify ti ... ] ] ] ]
(Calvin Trillin. 1996. Messages from my father. New York: Farrar, Strauss, and Giroux. 43-44)

b. arcane technical wisdom [ of which ]i [IP he has scarcely [DP a glimmering of [DP an understanding
ti ] ] ]
(Jeremy Campbell. 1982. Grammatical man. Information, entropy, language, and life. Simon and
Schuster. 260-261)

The most acceptable examples of this type involve movement out of DPs without a possessor, so
it might be proposed that a further refinement of the notion of barrier is in order. In particular, let
us assume that branching DPs are barriers, whereas nonbranching ones are not. Then the
examples in (44) and (45) all satisfy subjacency, as indicated by the revised representations in
(46) and (47), whereas an example like (48) would continue to violate it.

(46) a. Whati did [IP you { read, write } [DP a book about ti ] ] ?

b. Whoi did [IP you take [DP a picture of ti ] ] ?

c. [ { Which, how many } states ]i do [IP you know [DP the capitals of ti ] ] ?
(47) a. a devotion whichi [IP I would make [DP some lame attempt [CP ti [IP to justify ti ... ] ] ] ]
(Calvin Trillin. 1996. Messages from my father. New York: Farrar, Strauss, and Giroux. 43-44)

b. arcane technical wisdom [ of which ]i [IP he has scarcely [DP a glimmering of [DP an understanding
ti ] ] ]
(Jeremy Campbell. 1982. Grammatical man. Information, entropy, language, and life. Simon
and Schuster. 260-261)

(48) ? [ Which building ]i are [IP they protesting [DP the city's demolition of ti ] ] ?

Finally, given the contrast between (38a) and (38b,c), we would expect movement out of DP that
violates both subjacency and the ECP to be strongly unacceptable. Movement out of left
branches satisfies this expectation.

(49) * [ Whose ]i are [IP they protesting [DP ti demolition of the building ] ] ?

An important remaining puzzle is why some instances of movement out of nonbranching DPs are
perfectly acceptable, whereas others (even ones that satisfy the ECP) are not. In our view, an
explanation for the successive decrease in acceptability of examples like (50) should be sought
not in syntactic, but in pragmatic factors. We indicate this explicitly by using the pound sign, the
sign for pragmatic infelicity.

(50) a. What did he { read, write } a book about?

b. # What did he sell a book about?

c. ## What did he drop a book about?

(cf. ok What did he drop a hint about?)

We begin our attempt to explain the pattern in (50) by noting that any question is associated with
a so-called existential presupposition. The presupposition expresses the backdrop of knowledge
against which the question is raised, and the question itself solicits information that is missing in
the questioner's knowledge store. For instance, the question in (51a) is associated with the
presupposition in (51b).

(51) a. { What, what book } is he reading?

b. There is { something, some book } that he is reading.

Now consider the existential presuppositions in (52), which range from ordinary to implausible.
(52) a. There is a topic such that he { read, wrote } a book about that topic.

b. # There is a topic such that he sold a book about that topic.

c. ## There is a topic such that he dropped a book about that topic.

(cf. ok There is a topic such that he dropped a hint about that topic.)

If we make the plausible assumption that, all other things being equal, the acceptability of a
question matches the plausibility of the presupposition with which it is associated, then questions
that are perfectly well-formed from a grammatical point of view might nevertheless be judged as
unacceptable if they are associated with a highly implausible presupposition. This, then, would
account for the range of acceptability in the questions in (50), despite their structural parallelism.

Notes

1. The structural notion of government (to be distinguished from the morphological notion of
case government is defined as in (i).

(i) A governs (= is the governor of) B iff

a. A is a head,

b. B is a maximal projection, and

c. A and B are sisters (= mutually c-command each other).

Exercises and problems

Exercise 12.1

At first glance, the structure in (1a) seems preferable to that in (1b) because it is simpler in the sense of
postulating fewer nodes. It is standardly argued, however, that the structure in (1b) with a silent
complementizer is preferable. What is the motivation for the argument?

(1) a. He thinks [IP they have read War and Peace. ]

b. He thinks [CP [C that ] [IP they have read War and Peace. ] ]

Exercise 12.2
This exercise extends Exercise 5.12.

English has two sorts of though clauses: ordinary ones that do not involve movement, as in (1),
and ones that do, as in (2). The construction in (2) is often referred to as the though preposing
construction.

The term though preposing is potentially confusing. What is preposed is not though itself, but
some constituent in the though clause. In other words, though preposing is XP preposing that is
licensed by though.

(1) a. Ordinary (non-movement) We will solve the problem, though it is difficult.

b. We can solve the problem, though we believe it is difficult.

(2) a. Though preposing (movement) We will solve the problem, difficult though it is.

b. We can solve the problem, difficult though we believe it is.

A. Using the grammar tool in xbar 4, build structures for the sentences in (1) and (2). For
simplicity, you can build chunks and indicate how they go together.

B. Using the same grammar tool as in (A), build structures for just the though clauses in (3).
Omit the material in parentheses.

(3) a. Difficult though I wonder why the problem is, (I don't know for sure.)

b. Difficult though the students enjoy problems which are, (they can't always solve them.)

C. Which, if any, of the though clauses in (3) are expected to be ungrammatical given the
principles covered in this and previous chapters? Explain.

D. (4) is ungrammatical. Why?

(4) * Difficult though the problem is very, (the students will solve it).

Exercise 12.3

A. Corresponding to the declarative clauses in (1), we have the direct questions in (2).

(1) a. They would prefer for her to teach the course.

(2) a. Which course would they prefer for her to teach?

b. * Which teacher would they prefer for to teach the course?

A. Using the grammar tool in movement principles, build structures for both of the questions in
(2).

B. Can the judgments in (2) be derived from the principles introduced in this chapter? Your
explanation should be succinct, but specific. For instance, if a question violates subjacency or the
antecedent government clause of the ECP, indicate which barrier causes the subjacency
violation. If a question violates the lexical government clause of the ECP, indicate which clause
in the definition of lexical government is violated.

C. Is the specifier position in the elementary tree that the grammar tool postulates for the
complementizer for necessary?

Exercise 12.4

A. Consider the sentences in (1). Does the for phrase adjoin at N' or at V'? Explain, providing
evidence from pronoun substitution and do so substitution.

(1) a. They baked a cake for Marie.

b. Who did they bake a cake for?

B. Using the grammar tool in movement principles, build a structure for the question in (1b) that
is consistent with your answer to (A).

C. Is the acceptability of (1b) consistent with the principles introduced in this chapter? Explain.

Exercise 12.5

A. Is the prepositional phrase in (1) is a complement or an adjunct of review? Explain.

(1) this review of the book

B. Using the grammar tool in movement principles, build trees for the following questions.

The grammar tool provides a choice of two elementary trees for review. Use the one that is
consistent with your answer to (A).
(1) a. Which book did you see a review of in the Times?

b. Which book did you expect to see a review of in the Times?

c. Which book did a review of appear in the Times?

d. Which book did you expect a review of to appear in the Times?

C. Record your judgments concerning the questions (use "ok," "?," and "*" as your options).

D. For each of the questions, briefly explain whether it obeys subjacency and the ECP. Assume,
as in Further issues and refinements, that neither nonfinite IPs governed by V nor nonbranching
DPs are barriers. How do the predictions of the model of wh- movement developed in this
chapter mesh with your judgments from (B)?

Exercise 12.6

Consider the sentence in (1).

(1) Already Agassiz had become interested in the rich stores of the extinct fishes of Europe, especially
those of Glarus in Switzerland and of Monte Bolca near Verona, of which, at that time, only a few
had been critically studied.
(Encyclopaedia Britannica Online, Agassiz, (Jean) Louis (Rodolphe). Accessed 27 August 1999.)

A. Using the grammar tool in ***, build structures for the (simplified) subpart of (1) in (2).

In building the relative clause in (2), you may find it helpful to consider the declarative clause
Only of few of them had been critically studied.

(2) the fishes of Europe, of which only a few had been critically studied

B. Why did you put had where you did in the structures you built in (A)?

C. The grammar tool forces you to attach of Europe as an adjunct of fishes rather than as a
complement in (A). Provide evidence in favor of the attachment.

D. Is the trace of wh- movement in (2) antecedent-governed? In your answer, indicate clearly
which nodes (if any) are barriers that intervene between the trace and its antecedent. Assume the
definition of barriers in Further issues and refinements.
E. Is the trace of wh- movement lexically governed? Explain. In your answer, indicate clearly all
(if any) lexical government domains for the trace.

F. Which (if any) clauses of the ECP does the trace of passive movement satisfy in the structure
you built for (A)? As in (D) and (E), your answer should include all (if any) barriers that
intervene between the trace of passive movement and its antecedent, and all (if any) lexical
government domains for the trace of passive movement.

Exercise 12.7

A. Which (if either) of the adjectives in (1) is a control adjective, and which (if either) is a
raising adjective? Your answer must include the sentences (and associated judgments) that you
use as the basis for making your decision.

(1) a. Kim was ecstatic to get the job.

b. Kim is certain to get the job.

B. Based on your answer to (A), use the xbar 4 grammar tool to build structures for both of the
sentences in (1).

C. Using the same grammar tool, build the structure for the variant of (1b) in (2).

(2) Certain to get the job though Kim is, ...

Assume that nonfinite IP sisters to Adj are not barriers, on a par with nonfinite IP sisters to V.

D. For each trace of movement in (C), briefly explain whether it satisfies the ECP. For each case,
your answer should include which clause of the ECP is satisfied, or whether both clauses are
satisfied. (Don't consider verb movement.)

Exercise 12.8

A. Using the xbar 4 grammar tool, build the structure for the phrase in (1). Treat every as a
determiner.

(1) a letter of which every line was an insult (from Jane Austen)

B. Assuming that non-branching DPs are not barriers to wh-movement, is (1) consistent with the
ECP? Explain. Once again, your answer should include which clauses of the ECP are violated or
satisfied.
Problem 12.1

A. What is your acceptability judgment concerning wh- movement out of gerund clauses as in (1b)?

(1) a. The cats give the impression of wanting more food.

b. What do the cats give the impression of wanting?

B. Using the xbar 4 grammar tool, build a structure for (1) that reflects your judgment from (A)
in light of the syntactic principles introduced in this chapter. You will need to decide what
syntactic category to assign to the gerund clause.

C. Explain how your judgment from (A) can be made to follow from the structure you propose in
(B).

15 Binding theory: Syntactic constraints on the interpretation of noun phrases

Coreference and coindexing

Hellan 1988
o The co-argument condition
o The predication condition
o The tensed IP condition
o Strict vs. non-strict co-arguments
Extending Hellan's binding theory to English
o The co-argument condition
o The predication condition
o The tensed IP condition
Chomsky 1981
o Principle A
o Principle B
o Principle C
Notes
Exercises and problems (to be added)

This chapter is devoted to binding theory, the part of syntactic theory that is concerned with how the
interpretation of noun phrases is constrained by syntactic considerations. For the purposes of binding
theory, it is useful to distinguish several types of noun phrases: full noun phrases (the question, the
student that asked the question, and so on), ordinary pronouns (I, you, they, and so on), reflexive
pronouns (myself, yourself, themselves, and so on) and the reciprocal pronoun each other.

We discuss two alternative approaches to binding theory. The first approach, due to Hellan 1988,
was proposed on the basis of Norwegian, a language with a unusually rich set of pronouns. We
first discuss the Norwegian facts and then extend Hellan's analysis to English. The second
approach, due to Chomsky 1981, was proposed on the basis of English and does not cover the
full range of Norwegian facts. However, it includes an important condition on the distribution of
ordinary noun phrases that is missing from Hellan's binding theory.

Coreference and coindexing

The primary function of a noun phrase like Bill Clinton, my two cats, the king of France, Santa Claus, or
colorless green ideas is to refer - that is, to stand for a particular discourse entity, its referent. As the
examples show, referents can be entities in the actual world, entities in some possible world, or even
entities that could not possibly exist in principle. One of the characteristic features of human language is
the absence, in general, of a one-to-one relation between noun phrases and referents. On the one hand,
it is possible to use different noun phrases to refer to the same referent. The classic example is the fact
that the expressions the morning star and the evening star both have the same referent - the planet
Venus. On the other hand, the same noun phrase can have different referents. For instance, my
apartment, used either by the same person at different points in time or by different persons at the
same point in time, can refer to lodgings of vastly different size and attractiveness in completely
different locations. Similarly, Lois's checking account balance can refer to widely differing dollar
amounts.

In general, then, determining the intended referent of an expression requires recourse to a

particular discourse context (who is speaking when, to whom, and so on). But there are certain
expressions whose interpretation is particularly context-dependent - namely, pronouns. For
instance, it is perfectly felicitous to introduce a new topic in a conversation with a friend using
(1a) (provided that the speaker and the friend have in common a single acquaintance by the name
of Vanessa). But replacing the proper noun Vanessa with a pronoun, as in (1b), in the same
context (that is, as a new topic of conversation) is decidedly odd.

(1) a. I ran into Vanessa the other day.

b. I ran into her the other day.

On the other hand, if Vanessa has already been mentioned in the discourse, then (1b) is perfectly
felicitous, as illustrated in the mini-discourse in (2).

(2) A: Have you seen Vanessa recently?

B: Yup. I ran into her the other day.

Pronouns, then, in contrast to ordinary noun phrases, are referentially dependent on some
antecedent in the discourse. The term 'antecedent' is potentially misleading. Since it is derived
etymologically from Latin ante-cedens 'one who walks before', it suggests that antecedents are
required to precede a referentially dependent expression. However, precedence is neither a
necessary nor a sufficient condition for antecedenthood. For instance, a sentence like Zelda's
brother likes herself is ungrammatical even though Zelda precedes herself.. A less misleading
term might be 'referential anchor,' but we continue to use the 'antecedent' because it is the
standard term in the literature.

It is possible for a sentence to contain more than one possible antecedent for a pronoun. For
instance, in (3), he can refer to either Tim or Tom.

(3) Tim told Tom that he needed some time off.

In any particular discourse context, of course, one interpretation may well be favored over the
other, given what is known about Tim, Tom, their respective work loads, and so on. In case the
antecedent for he must be explicitly specified in an unambiguous way, this can be achieved as in
(4).

(4) a. Tim told Tom that he, Tim, needed some time off.

b. Tim told Tom that he, Tom, needed some time off.

When (3) has the interpretation in (4a), he and Tim are said to corefer. On the alternative
interpretation in (4b), it is he and Tom that corefer.

It is convenient to introduce a notation to represent coreference relations. Let us begin by

asssociating any noun phrase with a referential index. In the syntactic literature, it is standard to
use the letters of the alphabet, beginning with i (for 'index'), both as referential indices and to
indicate movement. For clarity, however, we will use the natural numbers as referential indices.
We now introduce the following convention. In order to indicate an interpretation where two
expressions refer to the same discourse entity (in other words, where they corefer), we assign the
same index to both expressions, or coindex them. On the other hand, in order to indicate an
interpretation in which two expressions refer to distinct discourse entities (that is, where they do
not corefer), we assign distinct indices to each of the two expressions. Such expressions are said
to be contraindexed. In neither case are the specific indices important - only whether the indices
are the same or not. For instance, both indexings in (5) represent the interpretation in (4a), and
both indexings in (6) represent the interpretation in (4b).

(5) a. Tim1 told Tom2 that he1 needed some time off.

b. Tim1097 told Tom18 that he1097 needed some time off.

(6) a. Tim1 told Tom2 that he2 needed some time off.

b. Tim380 told Tom7 that he7 needed some time off.

(7) gives a further grammatical indexing for the string in (3). Of course, in any particular
discourse, this indexing is pragmatically felicitous only if a discourse entity with the index 3
(say, Tim's brother Mike) has already been mentioned.

(7) Tim1 told Tom2 that he3 needed some time off.

Reference and indexing must be carefully distinguished. Reference relations are actual linguistic
relations that we have intuitions about. For instance, we have the intuition that Tim and him can
corefer in (8a), but not in (8b).

(8) a. Tim thinks that everyone admires him.

b. Tim admires him.

By contrast, indices are a notational device intended to represent arbitrary reference relations; the
indices themselves have no independent linguistic or psychological status. It is perfectly possible
to assign referential indices to noun phrases so as to represent interpretations of a sentence that
are possible in principle, but ungrammatical in fact. Two such ungrammatical indexings are
illustrated in (9).

(9) a. * Tim1 admires him1.

b. * He1 admires Tim1.

The proposition (= state of affairs) that both (9a) and (9b) are trying to express is not inherently
semantically anomalous. It can be expressed perfectly well by the grammatical sentence in (10).

(10) Tim1 admires himself1.

Notice furthermore that what makes the sequences in (9) ungrammatical is the index assignment.
The same sequences of words as in (9) are grammatical sentences when associated with the
indices in (11).

(11) a. ok Tim1 admires him2.

b. ok He1 admires Tim2.

In other words, the grammaticality of a sentence is always determined with reference to a

particular intended interpretation.1
Representations like those in (9) and (11) are generally abbreviated as in (12). The descending
numerical order of the indices on the object noun phrases is intended to unambiguously indicate
the scope of the asterisk.

(12) a. ok Tim1 admires him2,*1.

b. ok He1 admires Tim2,*1.

Why can't we use sentences like (9a) or (9b) to express the proposition that is expressed
grammatically in (10)? This is exactly the type of question that is addressed in the rest of this
chapter.

Hellan 1988

The co-argument condition

English makes a distinction between reflexive pronouns ending in -self (myself, yourself, and so on) and
the corresponding ordinary pronouns (I, you, and so on). Norwegian, too, distinguishes between
pronouns with forms containing selv 'self' and ones without. As (13) and (14) show, selv is in
complementary distribution with zero (that is, where a form with selv can occur, a form without selv
cannot, and vice versa).

In what follows, we restrict our attention to reflexive pronouns with local or long-distance
antecedents (that is, with an antecedent in the same clause as the reflexive pronoun or in a
higher clause, respectively). Many languages also have inherently reflexive pronouns, which are
special in that they don't function as an argument of their predicate, even though they occupy a
syntactic argument position. Inherently reflexive pronouns are rare in English (to behave oneself,
to perjure oneself), but much more common in other languages.

(13) a. Jon1 beundrer seg selv1.

admires SEG self
b. * Jon1 beundrer seg1.
'Jon1 admires himself1.'
(14) a. * Jon1 bad oss beundre seg selv1.
asked us admire SEG self
b. Jon1 bad oss beundre seg1.
'Jon1 asked us to admire him1.'

In an effort to understand what determines the complementary distribution, let us consider the
derivation of the sentences. Both sentences require the elementary tree in (15) for the verb
beundre(r) 'admire(s)'.
(15)

Observe now that in the derivation of (13a), seg selv and Jon both substitute into the elementary
tree for beundrer. In the derivation of (14a), on the other hand, the subject position of the
elementary tree for beundre is filled by PRO, which is referentially dependent on the matrix
object oss 'us', whereas Jon substitutes into the elementary tree for the matrix predicate bad
'asked'. The structures resulting from substitution in the two cases are shown in (16a) and (16b).

(16) a.

b. i. ii.

Given this derivational difference, we can define a notion of co-argument as in (17a) and propose
the condition governing the distribution of selv in (17b).

(17) a. Definition of co-argument (to be revised):

A and B are co-arguments iff they substitute into the argument positions (specifier or
complement) of the same elementary tree.

b. Co-argument condition (to be revised):

Selv and its antecedent (more precisely, the minimal DP nodes dominating them) must be co-
arguments.

One type of syntax/semantics mismatch. While adequate for (13) and (14), the definition and
condition in (17) does not extend to examples like (18).

(18) a. Jon1 snakket om seg selv1.

talked about SEG self
b. * Jon1 snakket om seg1.
'Jon1 talked about himself1.'
(19) a. * Jon1 bad oss snakke om seg selv1.
asked us talk about SEG self
 b. Jon1 bad oss snakke om seg.
'Jon1 asked us talk about himself1.'

As is evident, the distribution of selv in (18) and (19) is parallel to that in (13) and (14).
However, the grammatical (18a) violates the condition in (17b) because the DPs dominating seg
selv and its antecedent do not substitute into the same elementary tree. Rather, seg selv
substitutes into the elementary tree for om 'about', whereas its antecedent Jon substitutes into that
for snakket 'talked'.

One way of addressing this problem is to introduce a semantically-based notion of argument,

based on the idea that situations can be conceptually decomposed into events (activities, states,
qualities, and so on), on the one hand, and different types of participants in such events (agents,
experiencers, beneficiaries, goals, etc.), on the other. In general, the participants in an event
denoted by a lexical item correspond to the substitution nodes in the elementary tree associated
with that lexical item. Indeed, this correspondence is so general that substitution nodes are often
equated with participants in a situation, with both being referred to as arguments. However,
locutions like snakker om DP 'talk about DP' provide evidence that syntactic arguments
(constituents of the elementary tree projected by a lexical item) must be distinguished from
semantic arguments (participants in the event denoted by the lexical item). In such cases, there is
a mismatch between the syntax and the semantics. Specifically, the DP, being a complement of
the preposition, stands in no local syntactic relation with the verb. Nevertheless, it is a semantic
participant in the 'talking about' activity. For the purposes of binding, the prepositional
complement's status as a semantic argument is sufficient to license selv. Based on the facts
presented so far, it is therefore possible to revise (17b) to read as in (20b).

(20) a. Semantic co-argument:

A and B are semantic co-arguments iff they denote co-participants in the same situation.

b. Co-argument condition (to be revised):

Selv and its antecedent must be semantic co-arguments.

It should be noted that the status of a particular phrase as a semantic argument is not always easy
to determine, and that there is some individual variation regarding the relevant judgments. This is
reminiscent of the situation that we encountered in connection with the distinction between
syntactic arguments and adjuncts discussed in Chapter 4.

Another type of syntax/semantics mismatch. As we have just seen, a noun phrase can be a
semantic argument of a head denoting an event without being one of its syntactic arguments.
Also possible is the converse state of affairs, in which a noun phrase counts as a syntactic
argument of a head without being its semantic argument. We illustrate this second type of
syntax/semantics mismatch with reference to perception verb complements. Consider (21).

(21) Jon1 så seg selv1 sitte i stolen.

saw SEG self sit in chair.def
'Jon1 saw himself1 sit in the chair.'
As the tree for (21) in (22) shows, the matrix verb så 'saw' takes as its complement a VP small
clause, so Jon and seg selv are not co-arguments in the sense of either (17a) or (20a).2

(22)

In fact, seg selv is not a semantic argument of så at all, since the complement subject position
can be occupied by expletive det, the Norwegian counterpart of expletive there.

(23) Vi så det sitte ett spøkelse i stolen.

we saw there sit a ghost in chair.def
'We saw a ghost sit in the chair.'

The facts discussed so far motivate extending the concept of syntactic co-argument as in (24).

(24) Syntactic argument:

A and B are syntactic co-arguments iff they are locally head-licensed by the same head C.

The notion of local head-licensing was introduced earlier (though not by that name) in
connection with case licensing. There, we distinguished three configurations in which case was
licensed: spec-head, head-spec, and head-complement. These structural notions are also relevant
for binding theory. In (16), Jon and seg selv are co-arguments because they stand in the spec-
head and head-complement relation, respectively, with beundrer 'admires'. In (23), (the trace of)
Jon and seg selv are co-arguments because they stand in the spec-head and head-spec relation,
respectively, with så.

Our final revision of the condition on selv is given in (25).

(25) Co-argument condition (final version):

Selv and its antecedent must be semantic or syntactic co-arguments.

Note that the statement of the condition is not yet ideal. The fact that (25) contains a disjunction
suggests that we do not have a complete understanding of the notion of co-argument.

The predication condition

In addition to the distinction between elements with and without selv, Norwegian distinguishes
between the two third person singular forms seg and ham. Like seg, ham occurs on its own as well as
before selv. As (26) and (27) show, seg and ham are in complementary distribution.

Note that the noun phrases in (26) and (27) are all semantic arguments. Therefore, since the co-
argument condition in (25) is satisfied, the contrast in (26) and (27) must be due to independent
conditions governing the distribution of seg and ham.

(26) Jon1 snakket om { seg / *ham } selv1.

talked about SEG HAM self
'Jon1 talked about himself1.'
(27) a. Vi snakket med Jon1 om { ham / *seg } selv1.
we talked with about HAM SEG self
'We talked with Jon1 about himself1.'
b. Vi fortalte Jon1 om { ham / *seg } selv1.
we told about HAM SEG self
'We told Jon1 about himself1.'

The difference between the two cases is that the (potential) antecedent is a subject in (26), but
not in (27). Further evidence that the antecedent of seg must be a subject comes from (28), which
shows that what is crucial is the antecedent's status as a subject of predication.

(28) Vi gjorde Jon1 glad i { seg / *ham } selv1.

we made fond of SEG HAM self
'We made Jon1 fond of himself1.'

The facts in (26)-(28) are accounted for by the condition in (29).

(29) Predication condition:

The antecedent of seg must be a subject of predication, whereas the antecedent of ham must
not be.

The co-argument condition in (25) and the predication condition in (29) are independent of one
another. This has the result that the domain in which the referentially dependent elements under
discussion can appear can be partitioned without overlap as in (30).3

(30) Antecedent co-argument?

Antecedent subject of predication? Yes No

Yes seg selv seg

 No ham selv ham

The tensed IP condition

The contrast in (31) shows that seg is subject to a final condition, given in (32).

(31) a. Jon1 bad oss forsøke å få deg til å snakke pent om seg1.
asked us try to get you towards to talk nicely about SEG
'Jon1 asked us to try to get you to talk nicely about him1.'
b. * Jon1 var ikke klar over at vi hadde snakket om seg1.
was not aware over that we had talked about SEG
'Jon1 was not aware that we had talked about him1.'
(32) The minimal tensed IP dominating seg must also dominate its antecedent.

Strict vs. non-strict co-arguments

As we saw earlier, selv is subject to the co-argument condition in (25), repeated in (33).

(33) Co-argument condition:

Selv and its antecedent must be semantic or syntactic co-arguments.

The contrast in (34) might at first glance be taken to indicate that the reciprocal pronoun
hverandre 'each other' is not subject to such a condition.

(34) a. * Jon1 traff noen venner av seg selv1.

met some friends of SEG self
'Jon1 met some friends of his1.'
b. [ Jon og Marit ]1 traff noen venner av hverandre1.
and met some friends of each other
'[ Jon and Marit ]1 met some friends of each other's1.'

Nevertheless, as (35) shows, the distribution of hverandre is not completely unconstrained.

(35) ?* [ Jon og Marit ]1 leste mine bøker om hverandre1.

and read my books about each other
Intended meaning: 'Jon read my book about Marit, and Marit read my book
about Jon.'

It is possible to account for the contrast between (34b) and (35) by introducing a distinction
between strict and non-strict co-arguments. Strict arguments are defined as in (17) or (20), but
non-strict co-arguments are more loosely related. Consider the two configurations in (36).
(36) a. b.

In (36a), we will say that the absence of a specifier in BP allows B to extend its argument-taking
domain up to the first maximal projection that is 'closed off' by a specifier (in this case, AP). In
such a domain, which we will call a projection chain of B, the arguments of the heads in the
projection chain count as co-arguments in a non-strict sense. In (36b), on the other hand, the
presence of Spec(BP) prevents the formation of such a projection chain.

The similarities and differences in the distribution of selv and hverandre can now be succinctly
stated as in (37).

(37) a. Selv and its antecedent must be strict co-arguments.

b. Hverandre and its antecedent must be co-arguments (possibly non-strict ones).

We illustrate the application of the condition in (37b) to (34b) and (35) with reference to the
structures in (38).

(38) a. b.

(38a) is a (recursive) instantiation of the projection chain configuration in (36a). None of the
maximal projections in the object DP is closed off by a specifier, and the projection chain of the
preposition that governs hverandre therefore extends to VP. As a result, (the trace of) Jon og
Marit is a non-strict co-argument of hverandre and can serve as its antecedent. (38b) turns out to
be an instantiation of the projection chain configuration as well. But here, the projection chain of
the P containing hverandre extends only to the object DP, which is closed off by mine. The
ungrammaticality of the sentence can then be derived from the feature mismatch (number and
person) between hverandre and its potential antecedent mine.

Extending Hellan's binding theory to English

As discussed earlier, Norwegian distinguishes between two third person pronouns: seg and ham, which
can both stand alone or be combined with the morpheme selv, yielding four referentially dependent
forms (seg, seg selv, ham, and ham selv). English lacks the distinction between seg and ham and has only
two referentially dependent forms (exemplified in what follows by him and himself). The question that
guides the discussion in this section is how these forms divide up the syntactic territory covered by the
four forms of Norwegian.

The co-argument condition

Since English -self is cognate with Norwegian selv, it is reasonable to assume that English reflexive
pronouns are subject to a condition analogous to the co-argument condition in (25). Such a condition is
given in (39a). The companion condition on ordinary pronouns in (39b) makes explicit the
complementary distribution between the two types of pronouns.

(39) a. Reflexive pronouns and their antecedents must be (semantic or syntactic) co-arguments.

b. Ordinary pronouns and their antecedents must not be co-arguments.

The conditions in (39) are borne out by the data in (40) and (41).

(40) a. John1 admires { himself1 / *him1 }.

b. John1 talked about { himself1 / *him1 }.

c. We made John1 fond of { himself1 / *him1 }.

(41) a. John1 asked us to admire { him1 / *himself1 }.

b. John1 asked us to talk about { him1 / *himself1 }.

c. John1 made us fond of { him1 / *himself1 }.

In (37a), we proposed that the antecedents of Norwegian selv must be strict co-arguments, and so
we are led to wonder whether this is true of English reflexive pronouns as well. Evidence bearing
on this question comes from examples like (42), which has the structure in (43).
(42) John1 bought some books about himself1.

(43)

Himself, a semantic argument of books, has no strict co-arguments (the PP dominating it has no
specifier). But for precisely this reason, (43) is an instance of a projection chain. In particular,
given that the entire object DP contains no specifiers, the projection chain of about, the head
governing himself, extends to VP, making himself and (the trace of) John non-strict co-
arguments. Since (42) is grammatical, we conclude that the antecedent of himself need only be a
non-strict co-argument. In this respect, then, himself differs from Norweian seg selv and
resembles hverandre 'each other.'

As expected, if the projection chain of about is closed off by a specifier of the object DP, then it
is that DP that must contain an antecedent that agrees with himself in the relevant grammatical
features (number and gender). This is shown by the contrast in (44).

(44) a. Mary bought John1's book about himself1.

b. * John1 bought Mary's book about himself1.

Finally, the contrast between (42) and (45) on the one hand and (44) and (46) on the other
confirms that reflexive and ordinary pronouns are in complementary distribution.

(45) * John1 bought some books about him1.

(46) a. * Mary bought John1's book about him1.

b. John1 bought Mary's book about him1.

The predication condition

A further question that arises in comparing the English pronouns with their Norwegian counterparts is
whether himself covers the syntactic territory of seg selv alone or of both of the selv forms (seg selv and
ham selv). (41) and (46) show that the antecedent of him, unlike that of ham, can be a subject of
predication, and (44a) suggests that the antecedent of himself needn't be (we say 'suggests' rather than
'shows' because the status of prenominal genitives with respect to predication is a bit murky). This
evidence is consistent with the view that himself corresponds to both of the selv forms, and that English
reflexive and ordinary pronouns are not subject to a predication condition analogous to the one in (29).
This view is clearly confirmed by the facts in (47), where the antecedent of himself is not a subject of
predication.

(47) a. We talked with John1 about { himself1 / *him1 }.

b. We told John1 about { himself1 / *him1 }.

The tensed IP condition

Since the tensed IP condition governs the distribution of seg versus ham and English does not distinguish
between these two elements, we would expect the tensed IP condition to be irrelevant in English. This
conclusion is confirmed by the parallelism between (48a) and (48b).

(48) a. John1 asked us to try to get you to talk nicely about { him1 / *himself1 }.

b. John1 was not aware that we had talked about { him1 / *himself1 }.

In summary, then, English himself corresponds closely to the two selv forms of Norwegian, seg
selv and ham selv. The correspondence is not perfect, however, because the antecedent of
English himself, unlike that of the selv forms, needn't be a strict co-argument.

Chomsky 1981

We have presented an approach to binding theory that was developed in order to account for the
particularly rich data from Norwegian. In this section, we compare this approach with the standard
binding theory of Chomsky 1981, which was developed on the basis of morphologically simpler data of
English. Chomsky's binding theory contains three conditions (or principles, as they are more commonly
referred to), which govern the distribution of reflexive and reciprocal pronouns, ordinary pronouns, and
full noun phrases, respectively. We present each of these principles in turn.

Principle A

Consider the English binding facts in (49).

(49) a. [ Zelda ]1 helped [ herself ]1.

b. [ Zelda's sister ]1 helped [ herself ]1.

c. * [ Zelda's ]1 sister helped [ herself ]1.

In Hellan's binding theory, (49a) is grammatical because herself and its antecedent Zelda are
syntactic co-arguments. The same is true of herself and Zelda's sister in (49b), and so it is
grammatical as well. In (49c), on the other hand, the intended antecedent Zelda is not a co-
argument of herself, and so the sentence is ungrammatical under the intended interpretation.

Chomsky 1981 derives the grammaticality pattern in (49) on the basis not of co-argumenthood,
but on the basis of the structural relation of c-command, defined as in (50).

(50) A c-commands B iff (= if and only if)

a. neither A nor B dominates the other, and

b. the first branching node that dominates A also dominates B.

The structures for the sentences in (49) are shown in (51). In all three of the structures, the
intended antecedent of the anaphor herself and the anaphor itself are boxed in red. The first
branching node dominating the intended antecedent is boxed in black. Notice now that the black-
boxed nodes dominates that anaphor in (51a,b), but not in (51c). In other words, the anaphor is c-
commanded by the intended antecedent in (51a,b), but not in (51c).

(5 a b c
1) . . .

Intende
d
anteced
ent c- yes yes no
comman
ds
anaphor
 ?

These configurational relations suggest the condition in (52).

(52) Principle A (to be revised):

An anaphor must be c-commanded by a coindexed antecedent.

The term anaphor in Chomsky's usage refers to reflexive and reciprocal pronouns. This usage is
potentially confusing because in general linguistic usage, anaphora refers to referential
dependence regardless of morphological form. In other words, ordinary pronouns and even full
noun phrases can count as anaphors in this wide sense. Here, we will use the restricted sense of
the term, in keeping with Chomsky's usage.

(52) is generally expressed more succinctly as in (53a), where the notion of 'binding' is defined
as in (53b)

(53) a. An anaphor must be bound.

b. A binds B iff

i. A c-commands B, and

ii. A and B are coindexed.

As it turns out, (53a) is a necessary but not sufficient condition on anaphors in English, since it
fails to account for data involving complex clauses like (54).

(54) John1 thinks that Mary2 will help { herself2 / *himself1 }.

The grammaticality of the variant with herself is unproblematic (herself is bound by Mary ). But
given (53a), the ungrammaticality of the variant with himself is surprising, because himself is
bound by John.

In Hellan's version of the binding theory, John is ruled out as an antecedent of himself by the co-
argument condition in (39a). In Chomsky's version, a similar effect is achieved by introducing
the notion of governing category, a locality domain within which an anaphor must be bound.
Since the definition of governing category is fairly complex, we will work our way up to it in
several steps, motivating each complication of the definition in turn.
(55) Principle A (final version):
An anaphor must be bound within its governing category.

(56) Governing category (to be revised):

The governing category for an expression is the lowest IP that contains that expression.

(55) correctly describes the contrast in (54), but is unable to account for contrasts as in (44),
repeated as (57).

(57) a. Mary bought John1's book about himself1.

b. * John1 bought Mary's book about himself1.

In both sentences, the lowest IP that contains himself (the only IP in the sentence) also contains
John, in accordance with the definition of governing category in (56). The ungrammaticality of
(57b) is therefore unexpected.

A first step in accounting for such contrasts lies in making reference to DPs in addition to IPs in
the definition of governing category, as in (58).

(58) Governing category (to be revised):

The governing category for an expression is the lowest IP or DP that contains that expression.

But now (58) overshoots the mark by incorrectly ruling out sentences like (42), repeated as (59).

(59) John1 bought some books about himself1.

A further refinement of the definition of governing category yields (60).

(60) Governing category (to be revised):

The governing category for an expression is the lowest IP or DP that contains that expression, and
a specifier.

The reference to a specifier in (60) is the formal counterpart to the distinction between strict and
non-strict co-arguments in Hellan's approach to the binding theory. There is a difference,
however, in what the two approaches regard as the unmarked domain within which an anaphor
must be bound. In Hellan's approach, the availability of non-strict co-argument antecedents
extends a binding domain that would otherwise be smaller. By contrast, in Chomsky's approach,
the addition of clause (ii) in (60) restricts a binding domain that would otherwise be larger.
There remains one final revision to make to the definition of governing category. The revision is
motivated by sentences like (61a), which have the structure in (61b).

(61) a. Joan1 believes herself1 to be indispensable.

Let's consider in detail why (61) poses a difficulty for the definition in (60). As is evident from
(61b), the lowest IP or DP that contains the anaphor and a specifier is the complement IP
(indicated by the box). (The anaphor and the specifier turn out to be the same node, but nothing
in (60) rules this out.) But since the complement IP contains no antecedent for the anaphor, (61a)
is incorrectly predicted to be ungrammatical.

In order to accommodate sentences like (61a), the governing category must be extended in just
the right way to allow matrix subjects to act as antecedents in sentences like (61a), but not in
ones like (54). This is achieved by the formulation in (62).

(62) Governing category (final version):

The governing category for an expression is the lowest IP or DP that contains (i) that expression,
(ii) a specifier, and (iii) the expression's case-licensing head.

Note that the addition of clause (iii) in (62) has the same effect as Hellan's extension of the
notion of syntactic argument in (24) (recall the discussion of the second type of syntax/semantics
mismatch discussed earlier in connection with the distribution of Norwegian selv).

Principle B

We turn now to the distribution of pronouns.

As noted earlier, reflexive and reciprocal pronouns are combined in the Chomskyan tradition
under the rubric of anaphors. This leaves the unqualified term 'pronoun' to refer to ordinary
personal pronouns (I, you, he, and so on). Unless otherwise noted, we follow this usage in what
 follows.

Replacing the anaphors in (49) by pronouns yields the sentences in (63).

(63) a. * [ Zelda ]1 helped [ her ]1.

b. * [ Zelda's sister ]1 helped [ her ]1.

c. [ Zelda's ]1 sister helped [ her ]1.

The resulting grammaticality pattern, which is the converse of that in (49), suggests the condition
in (64).

(64) Principle B (to be revised):

A pronoun must be free (= not bound by an antecedent).

As in the case of our first formulation of Principle A, the formulation of Principle B in (64) is not
ye quite inadequate. This time, however, the condition errs on the side of caution, incorrectly
ruling out grammatical sentences like (65).

(65) John1 thinks Mary will help him1.

Since John c-commands him (along with everything else in the sentence), (64) incorrectly leads
us to expect (65) to be ungrammatical.

The fact that the antecedent and the pronoun are in different clauses in (65) suggests
reformulating (64) to incorporate the concept of governing category, as in (66).

(66) Principle B (final version):

A pronoun must be free in its governing category.

As is evident, both Hellan's and Chomsky's approaches to the binding theory agree that anaphors
and pronouns are in complementary distribution.

Principle C

Consider the sentences in (67).

(67) a. * She1 treats Mary1 well.

b. * She1 claims that they treat Mary1 well.

c. * She1 claims that we know that they treat Mary1 well.

In Hellan's approach to the binding theory, (67a) is ruled out by the condition in (39b), according
to which an ordinary pronoun and its antecedent cannot be co-arguments. It is not, however,
ruled out by Principle B, the counterpart to (39b) in Chomsky's binding theory, since the pronoun
is free. In order to rule out sentences like (67a), Chomsky's binding theory therefore contains a
third principle that governs the distribution of full noun phrases (referred to in Chomsky's usage
as r(eferential)-expressions).

(68) Principle C:
R-expressions must be free.

Principle C is reminiscent of Principle B, but differs from it in that the anti-binding requirement
is not relativized to a binding domain. The absolute character of the anti-binding requirement in
Principle C is borne out by the ungrammaticality of (67b,c).

The ungrammaticality of the sentences in (67b,c) means that an anti-binding condition on full
noun phrases also needs to be incorporated into Hellan's approach to binding theory. This is
because the co-argument condition in (39b), which ruled out (67a), fails to apply to (67b,c). They
must therefore be ruled out by separate means. The necessity for an independent anti-binding
condition on full noun phrases is underscored by sentences like (69).

(69) * We made herself1 treat Mary1 well.

In Chomsky's binding theory, (69) is ruled out by both Principles A and C. But in Hellan's
version of the binding theory, (69) satisfies the co-argument condition in (39a), the equivalent of
Principle A, because Mary and herself are syntactic co-arguments (treat locally head-licenses
Mary). It is therefore only with reference to an anti-binding condition on full noun phrases that
(69) can be ruled out as required.

In traditional grammar, the examples in (67) and (69) would all be classified as instances of
cataphora, instances of referential dependence in which the antecedent follows the referentially
dependent element. It is worth pointing out explicitly that not all instances of cataphora are
ungrammatical. Examples like (70), where the referentially dependent element precedes, but
does not c-command the antecedent, are grammatical, as expected under Principle C.

(70) If he1 calls, tell John1 I'll be back in an hour.

 The term 'anaphora' is generally used to subsume both cataphora and what might be called
strict anaphora, where the antecedent precedes the referentially dependent element. According
to the metaphor underlying the terms, discourse flows along like a stream, with temporally
earlier elements located upstream from later ones. The antecedent is then upstream of the
referentially dependent element in strict anaphora (< Greek ana 'up, against the current'), but
downstream in cataphora (< Greek kata 'down, with the current').

A final point needs to be made about Principle C. We pointed out earlier that Principle C is an
absolute requirement in the sense that it makes no reference to binding domains. It is absolute in
a further sense as well: it makes no reference to whether the binder is a referentially dependent
element. Thus, Principle C rules out (71) on a par with (67a).

(71) Mary1 treats Mary1 well.

It has been argued, however, that under certain discourse conditions, sentences structurally
parallel to (71) are in fact acceptable, as illustrated in (72a). Two further acceptable violations of
Principle C from unedited usage are given in (72b,c).

(72) a. Nobody likes Oscar. Even Oscar1 doesn't like Oscar1.

b. Phil1 said that if he2 came by, Phil1 would show him2 around. ([email protected]
to [email protected])

c. Luke1 thinks that everyone has as much integrity as Luke1 has. (overheard at the White Dog
Cafe, Philadelphia, PA, 17 Feb 01)

However, equivalent sentences in which the full noun phrase is bound by a referentially
dependent element are degraded.

(73) a. * Nobody likes Oscar. Even he1 doesn't like Oscar1.

b. * He1 said that if { he2 / Sean2 } came by, Phil1 would show him2 around.

c. * He1 thinks that everyone has as much integrity as Luke1 has.

Although judgments regarding contrasts of the type illustrated by (72) and (73) can be delicate, it
seems reasonable to weaken Principle C to include reference to the status of the binder. This is
done in (74).
(74) Weakened version of Principle C:
Full noun phrases must not be bound by a referentially dependent element.

Notes

1. The same point arises in other connections as well. For instance, the question in (i) is
grammatical on the interpretation in (ii.a), but not on that in (ii.b).

(i) How have they forgotten which problem they should solve?

(ii) a. Howi have they forgotten ti which problem they should solve? (By succumbing to Alzheimer's.)

b. * Howi have they forgotten which problem they should solve ti? (By using Fourier analysis.)

2. The Trees program doesn't support the special character å, so we substitute the conventional
Scandinavian orthographic variant aa in the trees.

3. Certain difficult cases are set aside; for further discussion, see Hellan 1988, Chapter 3.