Summary

Introduction to generative grammar

This is the study of syntax. Although there are many ways to approach this study,
we will take a primarily «generative» approach. 1 This chapter makes use of some of
the core notions of generative grammar which are most useful for empirical and
pedagogical purposes without concerning itself with the now strongly theoretical
and quite abstract aspects of this approach. A fundamental tenet of generative
grammar is that not only do words occur in a linear order but they also enter into
hierarchical relationships with one another within coherent units known as
«constituents».

Constituents are the proper subparts of sentences, as we will see below. Generative
grammar inherited the notion of constituents from the approach to grammar which
preceded it called «American Structural Linguistics». The form of sentence analysis
used by the American structuralists was known as «immediate constituent analysis».
In a book entitled Syntactic Structures originally published in 1957, which
inaugurated generative grammar, Noam Chomsky argued that immediate constituent
analysis, though not wrong, was insufficient since it dealt only with the surface
order .

Visiting relatives can be tiresome.

D-structure, though their S-structures are the same. In the classical «standard»
form of generative grammar, it was argued that syntax consisted of two types of
rules, phrase structure rules and transformations. 2 Phrase structure rules account
for the form of D-structures, which are simple , active , declarative , and
positive sentences. The phrase structure rules specify what is a constituent of
what, or in other words, they reveal the hierarchical structures of sentences.

Transformations perform various syntactic operations on the output of the phrase

structure rules to produce surface structures. It is the view that in order for a
speaker to acquire a language, the rules of the grammar must be finite in number.
Finally, generative grammar is concerned exclusively with the form of sentences,
distinguishing between «grammatical» and «meaningful». For example, the sentence
Colorless green ideas sleep furiously is grammatical, though not necessarily
meaningful, while the sentence Furiously sleep ideas green colorless is
ungrammatical since it violates the rules of syntax.

Constituents

Constituents may themselves be complex, containing other constituents. Note that

the same sequences of words may not always function as a constituent. It is the
context which determines whether a particular sequence forms a constituent or not.
«in the winter» is a constituent in the second and third sentences, but not in the
first.

In the first sentence «in» forms a constituent with «took». The sequence of words
«beautiful flowers» is a constituent in I received beautiful flowers for my
birthday but not in Though they are beautiful, flowers cause me to sneeze.
Constituents can be identified by a number of different «constituency tests».
Constituency tests are based on the principle that only entire constituents may be
manipulated by syntactic operations.

A phrase structure grammar of english

We will now attempt to construct a phrase structure grammar of English, a set of
rules that will «generate» any possible sentence of English. In reality, our
grammar will be incomplete, partial, possibly wrong, but it will be illustrative.
When trying to write rules to generate all possible but no impossible structures in
English, it is reassuring to remember that no one has yet written a complete and
flawless generative grammar of English.

3.1The form of phrase structure rules

A phrase structure grammar consists of a set of ordered rules known as rewrite

rules, which are applied stepwise.

A {B,C} or A

Whether the mutually exclusive items are written on one line separated by commas or
on separate lines does not matter, as long as they occur within braces.

In every phrase structure rule, there must be an initial symbol, a first left-hand
symbol, such as A above. Thereafter, every symbol appearing on the left has already
been introduced on the right-hand side. They immediately dominate lexical items, or
words.

3.2 Subject and predicate

In constructing our phrase structure grammar of English, we begin with the initial
symbol

S = sentence. We all have an intuitive idea of what counts as a sentence. It is a

tenet of both traditional and generative grammar that S consists of two
constituents: the subject and the predicate .

Note the different structures serving the function of subject in the following
sentences

The chocolate was later confiscated House plants add to humidity.

by his counselors.
The woman with the yellow hair They hoped to make it before midnight.
clapped her hands together with joy.

There are two tests for identifying subject

subject–auxiliary inversion, or the «question test», e.g. Was later confiscated?;

and
the «tag question» test, in which the pronoun in the tag agrees with the subject in
gender, number, and person, e.g. is possible to overdo jargon, isn’t ?

The noun phrase can be expanded in many different ways .

Noun phrase4

Table 7.1. Expansions of NP

The noun is the only obligatory element in the first seven expansions of NP below
and serves as head; the other elements are all optional. The adjective or adjective
phrase precedes the N and the prepositional phrase follows the N; both serve as
modifiers of the noun , expressing a quality of the noun, answering the question
«which dogs?».
.  In current generative theory, phrases are called «projections» , making more
precise or definite the phrase that follows. Det includes quite a diverse set of
grammatical words: demonstratives , articles , wh-words , possessives , and
quantifiers . We can write a rule for Det as follows:

Poss and Q. Note that Poss includes both possessive adjectives such as my and
possessive nouns such as John’s or Sally’s. It may even include an entire noun
phrase, as in that angry man’s We can account for this phenomenon by the following
rule:
In the last two cases in Table 7.

*The fierce Goldy in the yard

Our rule for NP, therefore, must indicate the optionality of Det, AP, and PP and
the mutual exclusiveness of Pro and PN with the other elements.

A preliminary structure for the NP the large dogs might be the following

Here, Det is shown as sister of both AP and N. In fact, the determiner really
relates to the rest of the noun phrase as a whole, not the AP and N separately. For
this reason , we will introduce the intermediate category of N-bar :
.

NP ; however, the determiner form sometimes differs slightly from the pronominal
form, as in My dog is gentle vs. Mine is a gentle dog.
N-bar consists of N and it modifiers .

And our revised structure for the large dogs would be the following

I like large dogs rather than small ones.

The ones that I like best are Rottweilers.
.  We are adopting here – in part – a formalism called «X-bar theory». In this
theory, all phrasal categories are seen as having the same structure, namely:
XP is called a «maximal projection» and X– is called an «intermediate projection».

Premodifier

According to X-bar theory, the grammar has only binary branching nodes. You will
notice that some of our nodes have three or more branches. In this way and others,
we depart from the strict theory .

Numerals, too, present a similar problem. In one respect, they would appear to be
determiners , but they also seem to fill the position predeterminers. We will not
be accounting for numerals here.
To this point in our grammar, the only function for NP that we have examined is
that of subject ; numerous other functions will be recognized later in this chapter

The adjective phrase can be expanded in several different ways .

Adjective phrase

Table 7.2. Expansions of AP

.  We are also not accounting for structures such as all of or some of .
.  We are ignoring adjectives that can follow the noun, as in the people
responsible for the budget or the members present, or even the pronoun, as in
someone responsible.

degree words cannot be modified by other adverbs. Degree words express a quality,
intensity, or degree of the following adjective or adverb; in other words, they
function, like determiners, as specifiers of the head.10
HINT: A test to distinguish between degree adverbs and general adverbs is to see if
the adverb can be preceded by very or quite. If it can, then it is a general adverb
; if it cannot, then it is a degree adverb .
There are two problems with this test. The first is that some adverbs can be both
general adverbs and degree adverbs . Thus, pretty is a degree adverb in the sense
‘somewhat’ and cannot be preceded by another degree word. If it is – i.e. very
pretty – it becomes a general adverb and has the sense ‘attractive’.
The second problem is that although participles which occur before the noun are
general adjectives, they often cannot be modified by Deg , though they can be
modified by AdvP . Note that one cannot say very married , but one can say very
happily married.
We can see that in all cases above, A is the obligatory element and head of the
phrase; all of the other elements are optional. The elements preceding the A are
modifiers or specifiers, but the PP following bears a different relationship to the
A; it serves as complement. Although we will indicate it as optional in our rules,
it is not optional if a complement-taking adverbial structure such as aware of,
afraid of, curious about, obvious to, or angry at is selected. Note that the PP
does not express a quality or degree of the
A but rather «completes» it; the A serves as governor of the PP.

The initial tree structure for the NP the very fiercely barking dog might be the
following: the Deg Adv A very ercely barking

.  Participles which occur after the noun will be accounted for in Chapter 9.
However, this structure incorrectly shows very, fiercely, and barking as sisters,
all modifying dog, that is, a ‘very dog’, a ‘fiercely dog’, and a ‘barking dog’.
Obviously, this is not what is being said, but rather «very fiercely» is modifying
«barking», and «very» is modifying «fiercely». As we will see in the next section,
the sequence constitutes an adverb phrase, so the correct structure is the
following: very ercely

We have seen above that the AP is introduced in our phrase structure rules under
N–, functioning as modifier of the noun; we now have a structural definition for
the «attributive» position of the adjective mentioned in Chapter 5: it is the
adjective dominated by N–.
Later in this chapter we will also introduce AP under the verb phrase, functioning
as complement of the verb; this is the «predicative» position. Note that if an
adjective has a complement, it can only occur in predicative position in English
unless it is compounded .
.  Although a nested structure with an intermediate category A¯ – dominating A
and PP – is warranted here, we will not introduce it because it is not important
for our purposes.
It is quite common for more than one adjective to occur as modifier of the N, as in
the long, blue, silken scarf.
AdvP, as in the very long, quite pale blue, silken scarf. To account for this
possibility, we must introduce a modification to our rule for NP which permits more
than one AP in a single NP. One way to do so would simply be to allow for multiple
APs in a «flat» structure as follows:

That is, rather than being a ‘scarf which is silken and blue and long’, it is a
‘silken scarf which is blue’ and a ‘blue silken scarf which is long’. Note that
changing the order of the adjectives produces unnatural phrases: ? blue long silken
scarf, ? silken blue long scarf.
3.4
Other sources

Generative grammar, a precisely formulated set of rules whose output is all (and
only) the sentences of a language—i.e., of the language that it generates. There
are many different kinds of generative grammar, including transformational grammar
as developed by Noam Chomsky from the mid-1950s. Linguists have disagreed as to
which, if any, of these different kinds of generative grammar serves as the best
model for the description of natural languages.

Generative grammars do not merely distinguish the grammatical sentence of a

language from ungrammatical sequences of words of the same language; they also
provide a structural description, or syntactic analysis, for each of the
grammatical sentences. The structural descriptions provided by a generative grammar
are comparable with, but more precisely formulated than, the analyses that result
from the traditional practice of parsing sentences in terms of the parts of speech.