Introduction:

The Chomskian Perspective on

Language Study

Contents

Introduction

1 Linguistics: the science of language

2 The native speaker: grammaticality and acceptability

2.1 Descriptive adequacy
2.2 Grammaticality and acceptability
2.3 The grammar as a system of principles

3 Knowledge of language
3.1 The poverty of the stimulus
3.2 Universal grammar
3.3 Parameters and universal grammar
3.4 Language learning and language acquisition
3.5 The generative linguist
4 The new comparative syntax
4.1 Principles and parameters: a recapitulation
4.2 The pro-drop properties
4.3 Relating the properties
4.4 Agreement and pro drop
-

5 Purpose and organization of the book

5.1 General purpose
5.2 Organization

6 Exercises
Introduction

The aim of this book is to offer an introduction to the version of generative

syntax usually referred to as Government and Binding Theory. 1 I shall not
dwell on this label here; its significance will become clear in later chapters of
this book.
Government-Binding Theory is a natural development of earlier versions
of generative grammar, initiated by Noam Chomsky some thirty years ago.
The purpose of this introductory chapter is not to provide a historical survey
of the Chomskian tradition. A full discussion of the history of the generative
enterprise would in itself be the basis for a book.2 What I shall do here is
offer a short and informal sketch of the essential motivation for the line of
enquiry to be pursued. Throughout the book the initial points will become
more concrete and more precise.
By means of footnotes I shall also direct the reader to further reading
related to the matter at hand. Much of the primary literature will be hard to
follow for the reader who has not worked his) way through the book, but
I hope that the information will be useful for future reference.

1 Chomsky (199::1.) himself expresses reservations about the label 'Government and
Binding Theory' and refers to the theory we are concerned with here as the 'Principles
and Parameters Theory'. The latter term is more comprehensive in that it covers
work done in the Government and Binding tradition as developed in the present
book, and also work done in a recent dev�lopment in the generative framework
usually referred to as the Minimalist Program (Chomsky, 1992). Since the label
Government and Binding Theory or its abbreviation GB-theory is widespread we
continue to use it here to refer to the generative work initiated by Chomsky's book
Lectures on Government and Binding (1981a). The term allows us to distinguish
the approach developed here from the more recent approach in the Minimalist
Program.
2 For a survey of the development of the theory see van Riemsdijk and Wdliams
(1986). This work should be accessible once chapter 7 has been covered.
The reader will find a good introduction to generative grammar in general in­
troductions to linguistics such as Akmajian, Demers and Harnish (1979), Fromkin
and Rodman (1988, 1992), Lighdoot (1982), Smith and Wilson (1979), etc. These
works should be accessible at this point. For more advanced introductions the
reader is referred to Chomsky (1965, 1981a, b, C, 1 982, 1986a, 1988, 1991), but
reading them should be postponed until after chapter 7 of this book, at which
point we shall have covered most of the technical issues that are discussed.
1 My use of the pronoun bis for referents which may be either male and female
follows the conventions of English grammar and I hope that the female readers of
this book will not feel offended by it.
4 Introduction to Government and Binding Theory

1 Linguistics: The Science of Language

When asked to indicate one prominent feature that distinguishes human beings
from animals, many would probably say that this feature is 'language'. Even
though animals may have c ommunication systems, none of these systems is
as rich or as versatile as the language used by humans. Language is human­
specific.4 This means that an understanding of the mechanisms of human
language may lead us to understand, at least partly, what it is that distin­
guishes human beings from animals. Linguistics, the study of language, gives
us an insight into the human mind.
Leonard Bloomfield defined linguistics as the science of language (Bloom­
field, 1 935). �ike all scientists, linguists will aim at formulating the general
principles to account for the d ata with which they are faced. Linguists try to
formulate generalizations about linguistic data, i.e. language.s
There are various ways of approaching the study of language. I assume the
reader is familiar with the traditional view of language study, where the focus
is often on the study of one specific language, say English. A linguist studying
English will try to characterize the principles that determine the formation
of English sentences. The goal will be to provide a systematic description of
English sentence formation, the grammar of English The description will
.

have to acco unt for data such as the following:

la Agatha Christie has written many books.

1b I don't like detective stories.

The sentences in (1) are well formed. They contrast with the sentences in
(2), which are ill formed.

2a * Agatha Christie many books written has.

2b * 1 detective stories like.

Well formed English sentences are constructed according to the grammar

of English: they are grammatical. The sentences in (2) are not formed according

In their introduction to linguistics Akmajian, Demers and Hamish (1979) present

a fairly comprehensive discussion of the differences between human language and
animal language.
Robins (1967) and Newmeyer (1980, 1983) offer good surveys of the development
of linguistics. These books will offer a broader background to situate the theory
we are discussing here in its historical context.
 Introduction: The Chomskian Perspective on Language Study 5

to the grammar of Engl ish : they are ungrammatical, as indicated by the

aster isks.
When writing a grammar, the linguist will not stop at merely listing exam­
ples with the appropriate grammaticality judgements. A simple catalogue of
sentences may be an interesting basis for discussion but it cannot be the
ultimate goal of scientific research. In addition to describing the data, the
linguist will formulate general principles which will be a pplicable to further
data. Informally, a l ing uis t might account for the ungrammaticality of (2), for
instance, by proposing that in English verbs precede their direct objects. A
first hypothesis might be that English sentences are constructed according to
the SVO pattern: subject precedes verb, verb precedes object. Let us call this
the SVO hypothesis. Having formulated this hypothesis on the basis of a
limited set of data, the linguist will test it on the basis of further data. The
SVO hypothesis will lead him to predict, for instance, that (3a) and (3b) are
grammatical; but as it stands, the hypothesis also predicts that (3c) and (3d)
are ungrammatical: the objects, detective stories and which stories respec­
tively, precede the subjects:

3a Jeeves is baking a cake.

3b Jo hn has bought a n ew car.

3c Detecti ve stories, I don't like.

3d Which stories do you l ik e?

Either the SVO. hypothesis itself will have to be modified in the light of the
data in (3c) and (3d) or one or more extra principles are needed which inter­
act with the original hypothesis to account for the grammaticality of (3c) and
(3d). We might, for exampl e, formulate a rule of topicalization which moves
a direct objc:<;t to the b egi nning of the sentence to account for (3c ) . In addition
we might formulate a rule for q uest ion formation which (i) moves the
questioning element (which stories) to the initial position of the sentence, and
(ii) inverts subject and a uxil iary (do) (cf. (3d)).
The total of all the rules and principles that have been formulated with
respect to a language constitutes the grammar of that language. A grammar
of a language is a coherent system of rules and principles that are at the basis
of the grammatical sentences of a language. We say that a gr ammar generates
the sentences of a language.
A first requirement for any grammar is that it provides a ch aracteriz at ion
of the language it describes, i.e. the grammar must be able to distinguish
those strings of words which are sentences of the language from those which
are not sentences of the language in question. Such a grammar will be obser­
vationally adequate.
6 Introduction to Government and Binding Theory

2 The Native Speaker: Grammaticality and

Acceptability

2. 1 Descriptive Adequacy

Not only linguists have the ability to judge English sentences. Every native
speaker of English knows intuitively that the sentences in ( 1 ) and (3) are
acceptable and that those in (2) are not. Moreover, every native speaker of
English produces a large number of grammatical sentences and understands
the English sentences that he comes across. The native speaker may not be
able to formulate the general principles that underlie the sentences he pro­
duces, but he has an unconscious or tacit knowledge of such principles; he
has internalized a grammar of the language. The native speaker's tacit
knowledge of the grammar of his language is the focus of enquiry for the
linguist working in the Chomskian tradition. We say that a grammar reaches
descriptive adequacy if, in addition to describing the data, it provides an
account for the native speaker's intuitions.
Let us consider some examples. We have proposed that (3c) and (3d) could
be generated by a process that moves the direct object leftward to the begin­
ning of the sentence. Now consider the examples in (4), which are not ac­
ceptable (hence the asterisk):

4a "Detective stories, I wonder if he likes.

4b "Where do you wonder if he lives?

To account for the unacceptability of (4a) we might propose that the process
which moves the direct object in (3c) must be constrained: the direct object
cannot move across if.
Similarly, when we consider (4b) we might propose that the rule of ques­
tion formation must also be constrained: the questioning element (where) must
not move across if. At this point we have reached observational adequacy: we
provide a description of the facts. However, if we stop at this point we are
missing a significant generalization. The ungrammaticality of (4a) and (4b) is
due to the same constraint. A descriptively adequate grammar will not simply
provide an analysis for (3c) and (3d) and for the deviance of (4a) and (4b),
but it will try to capture the relation between (4a) and (4b) and formulate a
general principle to explain why both (4a) and (4b) are felt to be unaccept­
able. Such a principle may be that no element in English must be moved
across if. This general principle will also lead us to predict that the .examples
.'
in (5) are ungrammatical, whereas those in (6) are grammatical:
 Introduction: The Chomskian Perspective on Language Study 7

Sa .. Where do you wonder if Emsworth has hidden the Empress?

5b .. Which detective do you wonder ifEmsworth will invite for Sunday lunch?
5c " To Bill, I wonder if he will give any money.

6a Where has Emsworth hidden the Empress?

6b Which detective will Emsworth invite for Sunday lunch?
6c To Bill, he won't give any money.

The general constraint which blocks movement of an element across if will

be taken to be part of the native speaker's internal grammar.
A descriptively adequate grammar will not only describe the linguistic data,
but it will contain the general principles and processes that enable the native
speaker to produce and interpret sentences in his language and decide on the
acceptability of sentences. Such a grammar is an explicit formulation of the
tacit linguistic knowledge of the native speaker, his internal grammar.
The shift of focus from language itself to the native speaker's knowledge
of language is the major feature of the Chomskian tradition. Both the gen­
erative linguist and the traditional linguist will be constructing grammars, i.e.
general systems that underlie the sentences of a language. But the generative
linguist conceives of his grammar as a rdlex of the native speaker's compe­
tence. The grammar is a representation of the speaker's internal linguistic
knowledge.

2.2 Grammaticality and Acceptability

At this point we turn to the notions of 'grammaticality' and 'acceptability'.

'Grammaticality' is a theoretical notion. A sentence is grammatical if it is
formed according to the grammar of English as formulated by the linguist.
'Acceptability', on the other hand, is the term which characterizes the native
speaker's intuitions about the linguistic data. Consider (7):

7a Bill had left. It was clear.

7b [That Bill had left] was clear.
7c It was clear [that Bill had left].
7d Once that it was clear [that Bill had left], we gave up.
7e Once that [that Bill had left] was clear, we gave up.

( 7a) contains two independent sentences. In (7b) the bracketed sentence

Bill had left is the subject of the complex sentence that
Bill had left was clear.
8 Introduction to Government and Binding Theory

We say that Bill had left is a subordinate clause. It is introduced by that, a

subordinating conjunction. Similarly, in (7c) that Bill had left is a subordinate
clause. In (7d) the sentence (7c) is a subordinate clause in a complex sentence.
A grammar must generate complex sentences in which one clause is part of
another one.
Let us turn to (7e). The sentence is odd for most native speakers: it is not
acceptable. However, this sentence is formed according to the same principle
that we posited to account for the formation of (7b)-(7d), i.e. that one sen­
tence may become part of another sentence. Hence (7e) would be grammatical,
though it is not acceptable.
Faced with intuitions such as that for (7e) the linguist might. decide to
modify the grammar he has formulated in such a way that sentence (7e) is
considered to be ungrammatical. He may also decide, however, that ( 7e) is
grammatical, and that the unacceptability of the sentence is due to independ­
ent reasons. For instance, (7e) may be argued to be unacceptable because the
sentence is hard to process. In the latter case the unacceptability is not strictly
due to linguistic factors but is due to the more general mechanisms used for
processing information.
The native speaker who judges a sentence cannot decide whether it is
grammatical. He only has intuitions about acceptability. It is for the linguist
to determine whether the unacceptability of a sentence is due to grammatical
principles or whether it may be due to other factors. It is the linguist's task
to determine what it is that makes (7e) unacceptable. This entails that there
may be disagreement between linguists as to whether certain unacceptable
sentences are grammatical or not. The disagreement is not one of conflicting
judgements of the sentence (although these may also exist), but it is one of
analysis. The linguist will have to determine to what degree the unacceptability
of a sentence is to be accounted for in terms of the grammar. All the linguist
has to go by, though, is the native speaker's intuitions about language, and
these, as argued above, are the result of the interaction between his internal
grammar and other factors.
In this book we focus on the linguistic knowledge of the native speaker.
We restrict our attention to his internal grammar. Obviously, the interaction
between the grammar and other mental processes is also an interesting area
of research, but it is not the topic of this book.

2.3 The Grammar as a System of Principles

One approach to formulating a grammar of a language would be to suppose

that the speaker's internal knowledge of English, i.e. his internal grammar, is
no more than a huge check-list of grammatical sentences. Speakers could be
 Introduction: The Chomskian Perspective on Language Study 9

thought to 'check' any sentence they come across against this internal inven­
tory. Sentences which match a sentence in the list would be said to be gram­
matical, those that do not are ungrammatical. Depending on the degree of
deviance of such ungrammatical sentences we could rank the sentences for
ungrammaticality. A grammar of a language would then be simply a list of
sentences. But it must be immediately obvious that listing all the grammatical
sentences of a language is an impossible task and also that it misses the point.
Cataloguing all the grammatical sentences of English is first of all impos­
sible because there is an infinite number of English sentences. In addition,
there are other objections to such a listing enterprise. We stated above that
linguistics is the scientific study of language. From such a perspective the list­
ing of linguistic data is not enough. We expect general principles to explain
the data.
For the generative linguist who tries to p�ovide a representation of the
native speaker's internal knowledge of a language a mere listing of sentences
would never achieve descriptive adequacy: it could never account for the
native speaker's knowledge of the language. Human beings - in our example
speakers of English - have finite memories: we often forget things we have
heard. Given that the capacity of our memories is finite, it would be absurd
to claim that human beings are able to store all potential sentences of the
language, an infinite set. It is thus in�onceivable that the native speaker's
internal linguistic knowledge is an inventory of sentences. We must assume
that human be ings are somehow equipped with a finite system of knowledge
which enables them to construct and interpret an infinite number of sen­
tences. This finite system of principles is what we referred to loosely above
as . the internal grammar of the language. The generative linguist will try to
render explicit the finite system of principles that make up the native speaker's
competence. In our example, the principle which prohibits moving elements
across if will be able to account for the unacceptability of (4) and (5).

3 Knowledge of Language

3. 1 The Poverty of the Stimulus

A speaker's knowledge of a language is largely unconscious. It is form ally
represented as a grammar. The grammar of a language generates the sen­
tences of a language and assigns to each sentence a set of representations
which provide the formal characterization of some of the properties of the
sentence (semantic, syntactic, morphological, phonological, etc. ). It is the
linguist's task to render explicit the internal grammar of the speaker of a
10 Introduction to Government and Binding Theory

language. In order to construct such an explicit grammar of a language, the

linguist can rely to some extent on data taken from usage, the output of the
speakers. However, usage data are inevitably an incomplete source of in­
formation. The sentences actually produced by a speaker are only a fragment
of the sentences he could have produced. In order to arrive at a charac­
terization of the speaker's potential, the linguist can also rely on the speaker's
knowledge of the language, i.e. on his capacity to evaluate linguistic expressions
in that particular language. For instance, speakers of English intuitively know
that (8a) is an acceptable sentence and that (8b) is not:

8a She has invited Louise to her house.

8b "Has invited Louise to her house.

Informally we will say that (8b) is unacceptable because the subject is missing.
For some reason, to which we return in more detail in chapter 8, the grammar
of English requires that finite sentences like (8a) have an overt subject. The
grammar of Italian differs from that of English, as seen in (9). In (9a) the sub­
ject of ha invitato is expressed, in (9b) it is not realized:

9a Lei ha invitato Louisa a casa.

she has invited Louisa at home
9b Ha invitato Louisa a casa.

We will achieve descriptive adequacy if our grammar is able to provide an

explicit characterization of the general principles of sentence formation in
English. This grammar will, for instance, impose the overt realization of the
subject pronoun in (8b).
Now another important and fascinating question arises: we would like to
understand how native speakers of a language, in our example English, come
to possess the knowledge of their language. We say that a theory reaches
explanatory adequacy if it can account for the fact that the principles of the
internal grammar can get to be known by the speakers, i.e. if it can account
for language acquisition.
The problem of language acquisition has often been summarized in terms
of the problem of the poverty of the stimulus. Our linguistic capacity, for
instance our knowledge of English, goes beyond the evidence we have been
exposed to in our childhood. The linguist wants to account for the fact that
the linguistic competence is attained in spite of important inadequacies in the
stimulus, the linguistic experience. Three types of inadequacies are standardly
 Introduction: The Chomskian Perspective on Language Study 11

referred to in the literature. First, we do not just come across grammatical

sentences: everyday use of language contains slips of the tongue, hesitations,
incomplete sentences, etc. Second, the experience, i.e. the stimulus, is finite,
and we end up being able to produce and process an infinite number of sen­
tences. Third, we acquire knowledge about our language for which we have
no overt or positive evidence in the experience. For instance, consider the
following sentences:

lOa I think that Miss Marple will leave.

lOb I think Miss Marple will leave.

l la This is the book that I bought in London.

1 1b This is the b ook I bought in London.

12a Who do you think that Miss Marple will question first?
12b Who do you think Miss Marple will question first?

On the basis of the examples in ( 1 0)-(12) the child learning English might
well conclude that the conjunction that is optional; the data in (10H12) suggest
that that can always be present and that it can a lways be absent. However,
this conclusion would not be correct:

Ba "Who do you think that will be questioned first?

Bb Who do you think will be questioned first?

In the sentences in (13), the conjunction that must not be present. It is hard
to see how the child can infer this from evidence to which he is exposed .
Observe also that children are not explicitly taught that (Ba) is ungrammatical.
The problem can be summarized by saying that there is a gap between the
data we are exposed to, the input, and our knowledge we achieve, the output;
the stimulus underdetermines the knowledge we ultimately attain. This means
that we cannot simply represent the acquisition of knowledge of language in
terms of the schema (14a). The triggering experience, i.e. exposure to lin­
guistic data, is not sufficient to allow a child to construct the grammar of his
language.

14a Exposure
Triggering experience ----7 Grammar of X
12 Introduction to Government and Binding Theory

3.2 Universal Grammar

Given that neither formal teaching nor overt evidence seems to be the source
of the native speaker's intuitions, it is proposed that a large part of the native
speaker's knowledge of his language, i.e. the internal grammar, is innate. The
idea is that human beings have a genetic endowment that · enables them to
learn language. It is this innate capacity for language learning common to all
human beings that the generative linguist tries to characterize. Of course, it
would be unreasonable to posit that some individuals - those that wUl be­
come native speakers of Englis� - are born with a specific grammar of English
and that others - those that will end up speaking Italian as their first lan­
guage - are born with the gra mmar of Italian readily stored in their minds.
Human beings with normal mental faculties are able to learn any human
language. The innate linguistic endowment must be geared to any human
language and not to just one.
Let us discuss some examples informally in order to provide an outline of
the proposal. We have introduced one generalization about English: the SVO
hypothesis. The data in ( 7) lead us to formulate another hypothesis: any
grammatical English sentence can apparently be embedded and become a sub­
ordinate clause in a complex sentence. Let us refer to this as the embedding
principle.

15 Embedding principle6
A grammatical sentence can become a subordinate clause in a complex
sentence.

The embedding principle tries to render explicit part of the tacit knowledge
of the native speaker. This principle would be taken to be part of the gram­
mar of English, hence available to the native speaker. But this principle is not
one that is particular to the grammar of English, it is not language-specific.
Rather, the embedding principle is part of the grammar of all human lan­
guages. Thus in French too we find sentences such as ( 1 6a) embedded in
( 1 6b):

1 6a Maig�et a abandonne l'enquete.

Maigret has abandoned the enquiry.
1 6b Lucas a annonce que Maigret a abandonne I'enquete.
Lucas has anno unced that . . .

As the reader will see later, the embedding principle is not in fact part of our
grammar. The fact that sentences can b e embedded can be deduced from the
principles of sentence formation discussed in chapters 1 and 2.
 Introduction: The Chomskian Perspective on Language Study 13

Readers familiar with other languages will be able to check that the embed­
ding principle applies in those languages too.
The embedding principle is a universal principle. Principles that hold of
all languages are said to be part of universal grammar, or UG for short.
Informally, UG is a system of all the principles that are common to all human
languages, this means languages as different as English and Italian or
Japanese.
A hypothesis adopted by generativists of the Chomskian tradition is pre­
cisely that universal grammar is innate to the human species. UG is a genetic
endowment: we are born equipped with a set of universal linguistic princi­
ples. To quote Chomsky himself: 'Universal grammar may be thought of as
some system of principles, common to the species and available to each
individual prior to experience' (1981b: 7).
If we assume that there is such an innate linguistic endowment the task of
attaining the knowledge of a specific grammar, say English, is facilitated. Some­
one learning English would not have to learn the embedding principle. It is
innate; it is part of the genetic endowment.7
Universal grammar is the basis for acquiring language. It under:lies all
human languages. All and only human beings are equipped with UG and they
are all able to learn languages. Other systems (say, dogs or television sets) are
not equi pped with UG and therefore will not be able to learn human lan­
guages. The linguistic endowment characterized as UG is species-specific.

3.3 Paramet�rs and Universal Grammar

The innate linguistic endowment UG is not sufficient to enable us to speak

a language. If all that is needed was UG then human beings would be able
to speak any language wherever they were born and in whatever circlim­
stances they grew up. The native language is that spoken by the child's
immediate environment. It would be inconceivable, for instance, that a child
growing up in a community where only English is spoken could become a
native speaker of Japanese. Human beings usually master one language with
native competence and they have a hard time learning other languages later
in life. It is a well-known fact that achieving complete mastery of second or
third languages in adulthood is exceptional.
While certain grammatical principles are universal, there is 'also a lot of
variation between different languages. The grammar of English differs in

The reader may wonder why, if the principle is innate, children do not start using
complex sentences straight away. However, it is conceivable that the development
of the internal grammar interacts with a general maturation process. We leave this
problem aside here.
14 Introduction to Government and Binding Theory

important respects from that of, say, Japanese. Hence, if you 'know' the gram­
mar of English, this will not entail that you 'know' the grammar of Japanese.
In (1) we illustrated some simple English sentences and we saw that English
sentences exhibit SVO word-order. In Japanese, on the other hand, the object
precedes the verb; Japanese is SOY:

17 John-ga Mary-o but-ta.

John-particle Mary-particle hit-past
(Kuno, 1973: 3)

English and Japanese are similar in that sentences contain elements such
as subjects, objects and verbs. But they differ in the way these elements are
ordered linearly. The SVO hypothesis, which we postulated as part of English
grammar, cannot be an absolute linguistic universal: it is part of the grammar
of English (and of other languages) but not of that of Japanese. It is language­
specific. How does a child learn that English has the SVO pattern? We could
envisage the following scenario. The linguistic endowment UG makes available,
among other things, the notions 'subject', 'object', 'verb'. Let us propose for
the sake of the argument (cf. chapter 2, for a different view, though) that
these are universal concepts, available in all human languages. Subject, verb
and object will have to be linearly ordered. When learning a language the
child will have to decide which is the word-order characteristic of his language.
One option is to say that in fact word-order variation between languages is
due to a primitive difference between these languages: it is a parameter along
which English and Japanese vary. Languages could be said to vary with
respect to the word-order parameter: UG provides the binary choice OV or
VO, and individual languages opt for one setting of the parameter or an­
other. We might say that the different word-orders of English and Japanese
are directly correlated with the word-order parameter: English has the setting
where the object follows the verb, Japanese has the opposite setting for the
parameter. The child learning English will have to fix the parameter for the
VO setting, the child learning Japanese will have to fix the parameter for
the OV setting. For each case exposure to transitive sentences in the lan­
guage should enable the child to perform the setting.
Other ways of accounting for word-order variation may come to mind.
The reader may recall that we suggested that the sentence-initial position of
the direct object in (3c) and in (3d) above were due to a fronting operation
which moves the object leftward. It is then in fact conceivable that the same
kind of leftward movement could be invoked to account for the word-order
found in Japanese. Say, for instance, that we propose that UG initially makes
only one order available for a verb and its objects, namely the VO order. It
 Introduction: The Chomskian Perspective on Language Study 15

could then be said that in Japanese a movement operation can shift the object
to the left across the verb, resulting in the ordering QV. We have seen that we
need such movement operations independently. The parameter distinguishing
English and Japanese would then be expressed in terms of the availability of
a particular leftward movement which can move the object to a position
between the subject and the verb. Again the child who is learning Japanese
will have to determine that the movement is available in Japanese, while the
child learning English would assume that it is not.
Whichever option is chosen to account for word-order variation - and the
debate is still very much open, we return to it in chaptet 2 the child learning
-

a language must construct an internal grammar for that language. To achieve

this task he uses, on the one hand, the universal notions and principles of UG
and the choices that it makes available, and on the other hand he uses the
data of his linguistic experience, in our example the English sentences he
hears. Sentences such as those in ( 1 ) will provide evidence to the child that
in English subject precedes verb and verb precedes object. A sentence such as
that in ( 1 7) will enable the child exposed to Japanese data to decide that
Japanese has SOV.
Exposure to linguistic material is an essential ingredient in the child's learn­
ing process. The child will need the linguistic experience to start constructing
the internal grammar of his language and thus to attain the knowledge of
a language. Without exposure the child would not be able to construct his
internal grammar. UG is crucial in the organization of the primary linguistic
experience. UG guides the way the child will interpret and organize the
language he is exposed to. We have now postulated two properties of UG:

(i) UG contains a set of absolute universals, notions and principles which

do not vary from one language to the next.
(ii) There are language specific properties which are not fully determined by
-

UG but which vary cross-linguistically. For these properties a range of

choices is made available by UG.

Absolute universal principles are rigid and need not be learnt. But even
with respect to the mastery of language-specific properties very little 'learn­
ing' is involved under the hypothesis outlined above. For those principles that
are parametrized, the options available are determined by UG. Attaining
linguistic knowledge consists in fixing the parameters.
From this point of view, we conclude that the mastery of a language is not
really the result of learning. Rather, being equipped with UG (with its
parameters) and exposed to a language, the child cannot but construct the
grammar of the language he is exposed to. For this reason the term 'learning'
is often replaced by the term 'acquisition'.
 16 Intro duction to Government and Binding Theory

In addition, the exposure to language will also equip us with a vocabulary,

the words of the language to which we are exposed. Even if we have an
innate knowledge of the principles of language we must inevitably learn the
lexicon of the language, the words and their meaning, in order to be able to
put this knowledge into operation. Thus an English child will have to learn
all the words in the sentences above, and indeed many more. And we go on
learning new words throughout our lives. Similarly a French child will learn
the French lexicon, etc. 8
To sum up: human beings are born equipped with some internal uncon­
scious knowledge of grammar: UG. UG is a set of universal principles of
language, some of which are rigidly fixed, some of which parametrized. Via
the input of the experience of one particular language this knowledge can be
implemented. The acquisition process is 'triggered' by the exposure, the child's
linguistic experience.
Exposure will also enable the child to learn the vocabulary of the language.9
The view of language acquisition in terms of parameter setting is the basis of
current work in the generative tradition. The theory is sometimes referred to
as the 'Principles and Parameters Theory' (cf. fn. 1 ) .

3.4 Language Learning and Language Acquisition

Our ability to speak a language is based partly on the innate principles and
parameters available in UG, partly on the triggering experience of exposure
to a specific language. On the basis of these components we develop a gram­
mar of one (or more) specific languages: the core grammar of such a lan­
guage.
Schematically we can represent the generative view of language acquisition
as follows:

14b
Triggering UG
Core grammar
experience � (with -
Language X
Language X parameters)

8 The acquisition of the vocabulary of a language is also a matter of interest. For

some introductory discussion the reader is referred to Lightfoot (1982: 121-2).
9 The reader will find interesting discussion of language acquisition for instance in
Deprez and Pierce ( 1 993), Hermon (1 992), Lightfoot (1981, 1982, 1989, 1991,
1993), Radford ( 1 990), WexIer and Manzini, (1987). For more general discussion
see also Chomsky ( 1 9 8 1a, b, c), and the literature cited there. Most of these
references might be hard to read at this stage and the reader is advised to postpone
reading these works until he has worked through chapters 1-7 of this book.
 Introduction: The Chomskian Perspective on Language Study 17

The exposure to some language, say English, will activate the innate prin­
ciples of universal grammar. The child will fix the choices to be made for the
language in question, for instance, that the object follows the verb. He will
also learn the vocabulary of the language. To quote Chomsky:

Endowed with these principles, a system provided with adequate experi­

ence will develop a grammar of the peculiar and specific sort characteristic
of human language . . . Lacking these principles, a system will develop no
grammar or some different system. The telephone exchange, for example,
has 'heard' much more English than any of us, but lacking the principles
of universal grammar . . . it develops no grammar of English as part of
its internal structure. ( 1 98 1 b: 8 )

B y the age o f six a child exposed t o English will have constructed the
grammar of his language. This does not mean that no further development
of his knowledge of language is possible. For instance, we go on learning new
words throughout our lives. In addition we also learn certain less usual
constructions of the language. These exceptional or marked parterns of the
language are not taken to be part of the core grammar of the language, they
belong to the marked periphery of the grammar and may be acquired later.
The native speaker will also have to learn all of the social or cultural con­
ventions associated with his language, for instance, that certain words belong
to a very high style whereas others are informal. These conventions are not
part of the grammar, they belong to the more general domain of human
behaviour.
The aim of generative syntacticians is to develop a theory of language that
is a model of the acquisition of language. Linguists want to provide an expli­
cit formulation of the three components of (14b): (i) the principles of UG and
the parametric variation across languages; (ii) the triggering experience needed
to activate the principles of UG; and (iii) the core grammar of specific lan­
guages as it derives from these interacting components. A theory that can
account for these three components will be said to have reached explanatory
adequacy.

3.5 The Generative Linguist

The research programme as sketched here briefly and roughly is one that has
been motivating linguistic research for the past thirty years and has given rise
to many challenging results. The programme is indeed still developing.
It may be useful to repeat that the ultimate aim of generative linguistic
18 Introduction t o Government and Binding Theory

theory is not to describe the details of one specific language, but rather to
formulate the underlying principl�s that determine the grammars of human
languages. These grammars are seen as representations of the native speaker's
knowledge. In the course of their enquiry, linguists will examine data drawn
from individual languages, of course, but the investigator will always bear in
mind the interacting components in ( 14b).
The generative linguist who tries to characterize knowledge of a language,
say English, will wish to do two things: (i) he needs to determine what pro­
perties of English are universal; and (H) what properties are English-specific
and how these relate to the parameters of UG.
It must by now have become clear that by simply looking at English and
only that, the generative linguist cannot hope to achieve his goal. All he can
do is write a grammar of English that is observationally and descriptively
adequate but he will not be able to provide a model of the knowledge of the
native speaker and how it is attained. The generativist will have to compare
English with other languages to discover to what extent the 'properties he has
identified are universal and to what extent they are language-specific choices
determined by universal grammar. Even when his main concern is some
aspect of the grammar of English the linguist will have to go outside this one
language and engage in contrastive work.
Work in generative linguistics is therefore by definition comparative. Gen­
erative linguists often do not focus on individual languages at all: they will
use any human language to determine the general properties of UG and the
choices it allows. Data from a dialect spoken by only a couple of hundred
people are just as important as data from a language spoken by millions of
people. Both languages are human languages and are learnt in the same way.

4 The New Comparative Syntax

4. 1 Principles and Parameters: A Recapitulation

When we look at the development of generative syntax in the last twenty-five

years one important tendency that can be isolated is a marked return to
comparative approaches. The comparative approach is obviously not the
creation of generative grammar: it finds a clear precedent in the nineteenth­
century comparative approaches to language study (d. Robins, 1 967).
The main goal of nineteenth-century comparative grammar was historical,
i.e. that of establishing relations of parenthood and kinship across languages.
The goal of the comparative approach in the generative tradition is
psychological, i.e. that of accounting for the knowledge of language. As we
 Introduction: The Chomskian Perspective on Language Study 19

have already seen, the following questions are asked: (i) What is knowledge
of language? (ii) How is it acquired? The latter question focuses on the issue
of how much of our linguistic knowledge is determined by experience and
how much is due to a predetermined mental mechanism (cf. ( 14b) ).
In order to determine how a specific language (say English) is acquired and
how language in general is acquired we have to determine to what extent the
properties of languages vary from one language to another, i.e. to what extent
the properties are language-specific, and to what extent they are invariant
across languages. Properties of language that vary cross-linguistically will be
learnt by the speaker as a result of exposure to some specific linguistic envir­
onment: the fact that Italian allows the subject pronoun to be absent can be
learnt through exposure to this language, for instance. Speakers who are
repeatedly confronted with subjectless sentences such as . (9b) will be able to
infer that in the language they are exposed to the subject can be omitted. On
the other hand, properties which are shared by all languages might well be
taken to be part of UG, the predetermined linguistic competence of the human
mind. Comparative studies of languages will play a crucial role towards pro­
viding us with answers to these questions, i.e. what is a universal and what
is language-specific. In the present section we focus on the parametric vari­
ation between languages and try to clarify the notion of parameter.
Parameters are postulated to expla� cross-linguistic variation. We should
not assume, though, that each observed difference between one language and
another corresponds to one parameter. The comparative study of languages
has revealed that the properties with respect to which languages vary tend to
organize themselves in clusters which are stable across languages and which
allow us to arrive at a typology of languages. If a language has property X,
it will also have property Y and property Z. The parametric approach will
have to explain why certain properties co-occur.

4.2 The Pro-drop Properties

In order to illustrate this let us look at one of the better known parameters
which has been postulated to account for the difference between English (8)
and Italian (9). Recall that Italian differs from English in that the former,
though not the latter, allows the subject of a finite clause to remain unexpressed.
The parameter which distinguishes languages like English which do not allow
a su bj ect pronoun to be omitted and those like Italian which do is referred
to as the pro-drop parameter. (For detailed discussion see, among others,
Rizzi, 1982a, 1986a; Jaeggli and Safir, 1989.) Italian is a pro-drop language,
English is not. That the subject prono un can be omitted is not the only
property to distinguish pro-drop languages like Italian from non-pro-drop
20 Introduaion to Government and Binding Theory

languages like English. In Italian, the overt subject can occupy a post-verbal
position; this option is not available in English:

1 8a E arrivato Gianni.
is arrived Gianni
'Gianni has arrived.'
1 8b "Is arrived John.

19a Ha telefonato sua moglie.

has telephoned your wife
'Your wife has phoned.'
1 9b ""Has telephoned your wife.

In Italian a subject of a subordinate clause can be moved to the main clause

domain across the overt conjunction cbe, corresponding to that; in English
this is not possible: if a subject is moved then the clause from which the
subject has been moved cannot be introduced by a conjunction (cf. the dis­
cussion of (13) above). The correlation between the data in (20) and the pro­
drop phenomenon is due to Perlmutter ( 1971).

20a Chi credi che abbia telefonato?

who believe (2sg) that have (subj) telephoned
'Who do you think has called ?'
20b "Who do you think that has telephoned?
20c Who do you think has telephoned?

In Italian subjects of weather verbs such as rain are necessarily omitted, in

English such subjects must be realized by a pronoun.

21a ( " Cio) piove.

(it) rains (3sg)
'It is raining.'
21 b " (It) is raining.

Consider now the following:

22a Che Louisa non partica e chiaro.

that Louise not will leave is clear
'That Louisa will not leave is clear.'
22b That Louise will not leave is clear.
 Introduction: The Chomskian Perspective on Language Study 21

In (22) the subordinate clauses che Louisa non partira and that Louise will
not leave function as the subjects of the sentences. The sentences can be para­
phrased if we move the subordinate clause to a final position: in Italian the
position vacated by the moved sentence remains empty, it cannot be blocked
up by a pronominal element as illustrated in (23a). In contrast, in English we
must stick in a pronoun it to fill the vacated subject position, as shown in (23b).

23a ( "" Gio) e chiaro che Louisa non partira.

it is clear that Louisa will not leave
23b "" (It) is clear that Louisa will not leave.

These contrasts listed above are not autonomous properties of the languages
in question, all can be related to the option which allows the subject pronoun
to be omitted in Italian.

4.3 Relating the Properties

We started from the empirical obse�ation that the subject pronoun can
apparently be omitted. Observe that in Italian, the pronominal subject can also
be overt; the overt realization of the subject pronoun has some semantic or
pragmatic effect: for instance it signals contrast or it focuses on the subject:

24 Lei parte e lui arriva.

she leaves and he arrives

When no contrast or no special focus on the subject is needed the pronoun

is absent. This could be derived from some general consideration of economy:
we might say that the non-expression of the subject pronoun requires less
effort than when the pronoun is present, and that therefore the subject will
only be present when the added effort of the overt expression has some yield.
Subject pronouns appear only when it is impossible to leave them out.
The obligatory absence of the subject pronoun of weather verbs in (21a)
can be related to the principle of economy suggested above: it is hard to see
how a subject of a weather verb could have a contrastive function. This
means that there will never be a reason to use the pronoun in Italian. A
similar approach can be suggested for (22). When we move the subject clause
in Italian the vacated position can be empty and it has to remain empty. Why
should this be ? We have already seen that the subject position in Italian need
22 Introduction to Government and Binding Theory

not be filled, it can be empty. In English the subject position cannot be empty
so we stick in a pronoun. It in the English example (23 b) does not contribute
anything to the meaning of the sentence, it cannot be contrasted or focused.
But in Italian, subject pronouns are only used with a contrastive or emphatic
function, so there will never be any motivation for inserting a pronoun in the
Italian equivalent of (23 b), (23a).
Let us turn to the examples with post-verbal subjects, ( 1 8 ) and ( 1 9) . All
English sentences must have subjects. This does not mean, though, that the
subject must necessarily be a referential expression, as the following example
illustrates:

25 There arrived three more students.

In (25) the subject position is occupied by the element there. There is related
to an indefinite post-verbal subject. Let us say that there fills up the position
vacated by an indefinite subject (we return to this in chapters 2 and 9). The
essential point is that there cannot be contrastive or emphatic in (25). In the
Italian examples in ( 1 8a) and (19a) we also have a post-verbal subject. Since
in general Italian does not need a full pronoun to occupy the vacated subject
position (23a), we do not need a filler for the subject position in such exam­
ples as ( 1 8 ) and (19).
The data in (20) might at first sight seem puzzling. It is generally accepted
that one cannot move a subject from the position to the immediate right of
the conjunction (that in English); (20b) suggests that this is possible in Italian.
However, we cannot base our judgements on a superficial comparison of two
sentences in two languages. We need to consider the way these sentences are
formed, their derivation. On the basis of the data in ( 1 8 ) and (19) we are led
to conclude that the subject NP in Italian may appear either pre-verbally or
post-verbally. Hence (20b) has two possible derivations, schematically repre­
sented in (26 ):

26a Chi credi che -- abbia telefonato?

26b Chi credi che abbia telefonato --?

In the representation (26a) chi originates in the position to the immediate

right of che, in (26b) it originates in the post-verbal position, a position also
available for subjects, as seen in (20b). Now it is known that in Italian, as
in English, nothing bans the leftward movement of post-verbal material across
a conjunction.
 Introduction: The Chomskian Perspective on Language Study 23

27a Who do you think that John will invite -- ?

27b Chi credi che Gianni invitera --?

The general principle which bans extracting material from a position to the
immediate right of a conjunction can now be maintained for the grammar of
English AND for the grammar of Italian. In the Italian �entences where this
principle would appear to have been violated, the language uses the alterna­
tive derivation whereby the subject is moved from a post-verbal position.
The correlations established here for the contrast between a pro-drop
language like Italian and a non-pro-drop language like English can extend
straightforwardly to Spanish, for the first group (28), and French for the
second (29);

28 Spanish
28a Baila bien.
dances (3sg) well
'He dances well.'
28b Uego Maria ayer a los doce.
arrived Maria yesterday at noon
'Mary arrived at noon yesterday.'
Uaeggli, 198 1 : 1 39)
28c (Quien dijiste que vino.
who did you say that came
'Who did you say came?'
Uaeggli, i 98 1 : 145)
28d Me parece que Juan tiene hambre.
me seems that Juan has hunger
'It seems to me that Juan is hungry.'
Uaeggli, 1981: 146)

29 French
29a " (Elle) dance bien.
(she) dances well
'She dances well.'
29b .. Arrivait Marie hier a midi.
arrived Marie yesterday at noon
29c "Qui dis-tu que viendra?
who say you that will come
29d " (II) me semble que Jean a faim.
(it) me seems that Jean has hunger
'It seems to me that Juan is hungry.'
24 Introduction to Government and Binding Theory

4.4 Agreement and Pro-drop

The reader may observe that the possi bility of omitting a pronoun subject
correlates with another property o f the languages examined and which is
particularly obvious when we compare Engl ish and Italian. If we look at the
present tense paradigms for the verb inflection for these languages we observe
a striking contrast:

30 English Italian
l sg I speak io parlo
2sg you speak tu parli
3sg she speaks lei parla
1pl we speak nOl parliamo
2pl you speak voi parlate
3pl they speak loro parlano

2 forms 6 forms

In the case of Italian, every number/pe rson combination has a different end­
ing; as a result the inflectional paradigm distinguishes all six persons uniquely.
There is no possibility of confusion: the ending of the verb immediately
identifies the subject. One could say that such inflectional systems are rich.
In contrast, the English system has only one distinctive form, that for the
third person singular; all other persons are unmarked morphologically, the
bare stem is used, which is also identical to the imperative and to the infini­
tive. In the literature, an attempt is made to correlate the inflectional para­
digm of the language with the pro-drop parameter (cf. Jaeggli and Safir (1989) ),
Rizzi ( 1 986a), Taraldsen ( 1980). Languages which have rich inflection are
often pro-drop languages.lo Intuitively this correlation is expected: when the
verb inflection is rich we can recover the content of the subject by virtue of
the inflection and the pronoun would not add information. In languages with
poor inflection the verb inflection does not suffice to recover the content of
the subject and the pronoun is needed. We return to this issue in chapter 8.
The inflectional system of French is relatively poor and French is not a pro­
drop language; Spanish is a pro-drop language and has rich inflection.
The approach above suggests that a number of properties of languages and

10 Gilligan ( 1 987) studies a sample of 100 languages from various language families
and reporrs 76 languages with agreement which allow for the subject pronoun to
be absent, against 17 languages without agreement and which allow the subject
to be absent.
 Introduction: The Chomskian Perspective on Language Study 25

language types can be reduced to a unique elementary difference between

their grammatical systems. The analysis of the pro-drop parameter, originally
developed on the basis of Romance languages in the late 1970s, has led to
what we can refer to as the new comparative syntax. A related development
is the study of dialect variation, which has become strongly prominent in the
1980s; another promising line of research is that in the area of historical
syntax. Diachronic developments of languages are interpreted again in terms
of the Principles and P arameters model, diachronic changes consisting in re­
settings of one or more parameters (cf. Lightfoot, 1979, 1991). In this b ook,
the comparative approach is more prominent in chapter 8, which discusses
non-overt elements; in chapter 9, where we discuss cross-linguistic variation
in q uestio n formation, and in chapter 1 1 , which concerns verb movement.

5 Purpose and Organization of the Book

5. 1 General Purpose

In this book I provide a survey of some of the main results of generative

research over the past thirty years. The book is not meant for the absolute
beginner. The reader is expecte d to have some background in linguistics,
specifically in syntax. He should, for instance, be able to parse sentences and
be familiar with the tree diagram representation, and with the basic terminol­
ogy of syntax. Notions such as sentence, clause, noun, verb, subject, object,
etc., are presupposed. I assume therefore that the reader has had some in­
trod uctory course to syntax or that he has read some introductory works.u
However, in order to guarantee that we have a common starting-point, I shall
often recapitulate the basic notions . It will also be shown how traditional
concepts are used and reinterpreted within the generative framework.
The aim of the book is to offer a general int roduction . I shall not go into
all the complexities and details of ongoing research. Rather, I wish to famil­
iarize the reader with the basic concepts used. I hope that the book will
encourage the reader to turn to the primary literature himself and discover
some of the more intricate problems . The references in the footnotes will
provi de indications for further reading.
Although the examples in the book will be taken primarily from English,

11
I am thinking of works such as Akmajian and Heny ( 1 975), Akmajian, Demers
and Harnish (1978), Burton-Roberts ( 1 986), Fromkin and Rodman (1 98 8 ) ,
Huddleston ( 1 976), Jacobs and Rosenbaum ( 1 970), Smith and Wilson ( 1979 ),
Wekker and Haegeman ( 1 985) to mention only a few.
26 Introduction to Government and Binding Theory

this book is not a grammar of English English is used as just one example
.

of human language and we shall often discuss other languages We shall try
.

to decide what sort of internal grammar native speakers of English have at

their disposal and to determine what it is that makes a sentence acceptable
or unacceptable, what sort of grammatical principles can be advocated and
to what extent these are universal or language-specific. In some sense we are
like linguistic detectives. The linguistic data are like the clues a detective is
given when starting his enquiry. He has to piece these data together, con­
struct hypotheses, check these and ultimately he may discover the explana­
tion for the evidence he has assembled. To remind the reader of this task I
have chosen to illustrate the data with examples in which literary detectives
play a prominent role. At the end of the book I hope that the reader will have
become a competent linguistic detective himself.

S.2 Organization

The book is divided into twelve chapters. The first ten chapters provide the
basic outline of the theory. The last two chapters highlight some recent
developments of the theory. Each chapter is followed by a one-page summary
and by a set of exercises. The exercises have a dual purpose. First, they will
enable the reader to check if he has understood and assimilated the basic
concepts introduced in the chapter. The empirical range of the discussion is
broadened: many exercises will include a discussion of data drawn from lan­
guages other than English.
Second, the exercises will be used to draw the reader's attention to theoretical
or empirical problems not touched upon in the chapter. Often a problem
introduced by way of an exercise in an earlier chapter is then picked up in
the discussion of a later chapter. Alternatively, the exercises will direct the
reader to areas for further reading or for further research.
Footnotes will mainly be used to direct the reader to further reading. The
footnotes will also indicate at which point in the book the reader should be
able to tackle the literature in question.

6 Exercises

Exercise 1
Consider the following sentences. None of them is fully acceptable but
they vary in their degree of deviance. If you are a native speaker of
 Introduction: The Chomskian Perspective on Language Study 27

English try to rank the sentences for acceptability. Wherever you can,
try to construct an acceptable sentence modelled on the one you are
judging. If you are not a native speaker of Engl i sh you may attempt to
carry out the task described above but it may be difficult. Another way
of approaching this exercise is to ask some native speakers to do the
exercise and compare their answers.

Which man do you knowwhat John will give to?

2 Which man do you wonder when they will appoint?
3 Who do you wonder which prese nt will give?
4 Which p resent do you wonder who will give?
5 Which man do you wonder whether John will invite?
6 Which man do you wonder wh eth er will invite John?
7 Which man do you wonder what will give to Joh n?
8 Which man do you wonder when will invite John?

Native English speakers are basically in agreement on the ranking of

sentences 1 -8. The judgements form u lated are not the result of formal
tuition. English grammar classes do not pay attention to sentences like
1 -8. It is quite likely that speakers have never come across such
sentences. In other words, they have not acquired the intuitions on the
basis of overt evidence. On the contrary, given that the Sentences
above are judged as unacceptable, one does not expect them to be
part of the linguistic data that we are exposed to.
On the basis of the judgements, try to classify the examples and
formulate some principles that might account for the relative accept­
ability. You may find the discussion of examples (3), (4), (5) and (6) in
the te� of some help. In chapter 7 and following we shall discuss the
sentences above and similar ones. We shall assume that they are
ungrammatical and we shall attempt to formulate the rules and princi­
ples at work.

Exercise 2

If you are a native speaker of a language other than English translate

the sentences in exercise 1 in your own language, keeping as close
to the English models as you can, and rank them for acceptability. Try
to formulate some prinCiples to explain the degree of acceptability.
If you have access to judgements on the English data and on data
28 Introduction to Government and Binding Theory

in other languages, try to check if the same degree of acceptability of

the examples could be explained by the same principle(s) .

Exercise 3

When reading section 3 the reader will have noted that there are
certain uses of English which allow the omission of the subject and in
which text example (8b) would have been grammatical. The following
are attested examples.

1 a A very sensible day yesterday. Saw no one. Took the bus to

Southwark Bridge. Walked along Thames Street; sa w a flight of
steps down to the river. . . , Found the strand of the Thames, under
the warehouses . . Thought of the refugees from Barcelona
. .

walking 40 miles, one with a baby in a parcel. . . Made a circuit:

.

discovered St Olave's Hart Street.

(Woolf, 1 985: 203-4)
1 b The poor little boy wont say whats the matter. He takes no interest
in anything. Wont turn and wave to her . . . drudges on at Latin.
(Woolf, 1 985: 1 1 7)
1 c Brilliant could have stayed all day.
Brill must come again.
-

Could see everything from wheelchair.

(Quotes from Visitors book 1 991 , The Green, Beaumaris, Angle­
sey North Wales)

Even a superficial glance at these examples shows us that all of the

italicized verbs have one property in common: the subject is missing.
In (1 a) and (1 c) the first person subject is omitted, in (1 b) it is the third
person. The omission of the subject in certain types of English is
observed in traditional descriptions (Quirk, et al. 1 985: 896-7). Such
examples are relatively easy to come by in certain registers of Eng­
lish, which we could roughly characterize as belonging to abbreviated
writing. We do not have to look for attested examples of usage to
discuss such data; every native speaker of English will be able to think
of relevant examples and even non-native speakers will quickly pick up
this type of ellipsis in the appropriate register.
All the attested examples are instances where the subject of a root
clause is omitted. By root clause we mean a clause which is not sub­
ordinate to another clause. The following variants on sentences drawn
from Virginia Woolf's diary are unacceptable:
 Introduction: The Chomskian Perspective on Language Study 29

2a I must work, as *(1) told Sally G . . . (Woolf, 1 985: 38)

2b I don't think *(1 ) need lie quaking at night . . . (Woolf, 1 985: 38)
2c I find this moming that *(1) interrupted the crisis of that London
Group meeting . . . (Woolf, 1 985: 9)

Another property that we find is that attested examples never occur

in questions. In the examples in (3) drawn from usage data, the subject
pronoun cannot be omitted.

3a And what could *(we) do . . . (Woolf, 1 985: 1 9)

3b What can *(1) say . . . (Woolf, 1 985: 3)
3c Now who is *(she) . . . (Woolf, 1 985: 1 5)
3d What shall *(1) write . . . (Woolf, 1 985: 40)

The absence of such examples in subordinate questions is expected

if the omission of the subject is a root phenomenon.

4 and this will show how hard �(I) work . . . (Woolf, 1 985: 1 3)

The subject also never is and in fact cannot be omitted when it is

preceded by a non-subject:

5a The next book *(1) think of calling Answers to

Correspondents . . . (Woolf, 1 985: 3)
5b Such twilight gossip *(it) seemed . . . (Woolf, 1 985: 8)
5c This story *(1) repeated to Duncan last night . . . (Woolf, 1 985: 9)
5d And there *(1) was in the rush of an end . . . (Woolf, 1 985: 1 1 )

When a negative constituent is preposed, resulting in a word-order

where the auxiliary precedes the subject, the subject pronoun cannot
be omitted.

6a Seldom have *(1) been more completely miserable than I was

about 6.30 last night . . . (Woolf, 1 985: 8)
6b Never have *(1) worked so hard at any book . . . (Woolf, 1 985: 1 6)
6c Nor do *(1) wish even to write about it here . . . (Woolf, 1 985: 44)
30 Introduction to Government and Binding Theory

Finally observe that only subject pronouns are omitted: objects are
not omitted. There is not a single example in Woolf's diary of the
omission of an object and the omission of me in (7) leads to an un­
acceptable sentence.

7 This led *(me) to imagine any number of catastrophes . . . (Woolf,

1 965 : 9)

At first sight one might be tempted to conclude that this variety of

English exhibits a manifestation of the pro-drop phenomenon dis­
cussed in section 4. Evaluate this proposal. You should draw on the
English data given above, on the Italian data given in (8) and (9), and
on your own intuitions. Using the argumentation introduced in section
4 try to state your argument as systematically and as explicitly as
possible.

8 Credo che sia gia partito.

I believe that be (subj) already left
'I think that he has already left.'

9a Dove e?
where is (3sg)
Where is he?'
9b Che vuoi?
what want (2sg)
What do you want?'
9c Questo libro non 10 voglio.
this book non it want (1 sg)
'This book, I don't want it.'

Readers whose first language is another non pro-drop language are

encouraged to consider the question of the omissibility of the subject
in abbreviated registt;lrs (diaries, informal notes) in their native
language.
For a discussion of the omission of the subject in English the reader
is referred to Haegeman (1 990) and to Rizzi (1 992a). The latter paper
relates the phenomenon of omission of the subject in the diary register
to data drawn from acquisition.