Constituent structure......Syntax

Constituent structure

4.0 Introduction

In chapter 1 two distinct aspects of syntactic structure were introduced; they
were labelled ‘relational structure’ and ‘constituent structure’. The previous two chapters
were devoted to relational structure, and in this chapter the focus is on the other aspect
of syntactic structure, constituent structure. The constituent structure of a sentence is
concerned with the units into which the words in a sentence are grouped, which are
the constituents, and their hierarchical organization. Constituent structure of the kind
to be discussed in this chapter is also sometimes referred to as phrase structure. In
section 1.1, the following provisional constituent structure was given for the sentence
The teacher read a book in the library.
(4.1) [S [NP The [N teacher]] [VP [V read] [NP a [N book]] [PP [P in] [NP the [N library]] PP]
VP] S]
In the next section, constituency tests will be discussed in detail, and another method
for representing constituent structure will be introduced, namely tree diagrams. In the
analysis of the constituent structure of sentences, it is necessary to break sentences
down into their various constituents and establish the form classes found in them and
also to develop rules which will specify the constituent structure of sentences. Breaking
sentences down into their constituents is known as parsing, while the specification
of their structure involves the formulation of phrase-structure rules. In chapter 5,
grammars based on phrase-structure rules will be constructed.
Constituent structure is purely formal in nature; that is, it is based upon the syntagmatic
and paradigmatic properties of elements and groups of elements rather than
their meaning. This is not to say that there may not be semantic properties associated
with constituents, which there surely are. Rather, it means that the determination of
constituents is based on tests which refer to the cooccurrence and substitution properties
of elements and groups of elements and which make no reference to meaning.
‘Constituent’ and ‘form class’ are closely related notions. That is, lexical classes
like noun, verb, etc. and the phrases they head can be considered form classes or constituents,
depending upon the analytic task at hand. Formal criteria for form classes can
be stated in terms of constituents’ syntagmatic and paradigmatic properties. They are
summarized in (4.2).
Constituent structure
111
(4.2) Formal criteria for form classes
a. Internal structure: a particular form class contains certain types of elements
but not others.
b. External distribution: a particular form class has a unique range of morphosyntactic
environments in which it can occur.
The internal structure criterion is paradigmatic in nature: within particular types of
form classes, certain types of elements may replace each other but not other types.
For example, a pronoun can replace the elements in an NP but not in a PP, VP or AdjP.
The external structure criterion, on the other hand, is syntagmatic in nature: each type
of form class has a unique set of possibilities for cooccurring with other elements in
morphosyntactic environments. For example, an AdjP can cooccur with a noun inside
an NP or with a copular verb within a VP but not with an adposition in a PP. An NP,
on the other hand, can occur in all three of these grammatical contexts.
In the next section the primary tests for constituency will be investigated, and
constituent-structure (phrase-structure) tree representations will be introduced. In
section 4.2, the notion of form class will be reexamined, and in section 4.3 a more complex
schema for conceiving of constituent structure will be presented. In section 4.4 the
constituent structure of some of the types of complex sentences introduced in chapter 2
will be explored, and in the final section some of the grammatical phenomena from
chapter 2 will be reanalyzed in constituent-structure terms.
4.1 Constituents and their formal representation
For syntactic analysis it is necessary to have a set of tests which will permit
the analyst to uncover the constituent structure of sentences of the language being
studied. This is known as immediate constituent analysis. In addition, explicit representation
of phrase structure is an important part of many theoretical approaches in
syntax. In this section tests for constituency and immediate constituent diagrams for
representing the structure of sentences will be introduced.
4.1.1 Tests for constituency
There are three primary tests for constituency: substitution, permutation and
coordination. The substitution criterion entails that only a constituent can be replaced
by another element, usually a pro-form, i.e. a pronoun for nouns, a pro-VP for VPs, or
a pro-PP for a PP. In (4.3b)–(4.3e) possible substitutions in a sentence like the one in
(4.3a) are given.
(4.3) a. The new teacher read a short book in the library.
b. She read a short book in the library. ‘she’ replacing ‘the new teacher’
c. The new teacher read it in the library. ‘it’ replacing ‘a short book’
d. The new teacher read it there. ‘there’ replacing ‘in the library’
d′. The new teacher read it in there. ‘there’ replacing ‘the library’
e. The new teacher did. ‘did’ replacing ‘read a book in
the library’
An introduction to syntax
112
In (4.3b)–(4.3d) an NP is replaced by the appropriate pronoun, she in (b) and it in (c)
and (d). The pronoun replaces the whole NP, not just the N; this can be seen in the
impossibility of *the new she or *a short it. There can be a pro-PP, replacing a PP, as
in (d), or it can be a pronoun, substituting for an NP when it is the object of a locative
preposition, as in (d′). (Note that while in/on/under/over/behind there are all fine,
*with there is impossible.) When there functions pronominally, it must replace the
whole NP and not just the N, as the ungrammaticality of *in the there clearly shows.
Finally, in (4.3e) did functions as a pro-VP and replaces the entire VP. It is a pro-VP
and not a pro-verb, because it cannot replace the verb alone, as the impossibility of
*The new teacher did a short book in the library attests.
What is important for this criterion is not only the possible substitutions, as in
(4.3b)–(4.3e), but the impossible ones as well. There is no possible pro-form *glarf in
English which could replace the sequence of teacher read in (4.3a), yielding *The new
glarfed a short book in the library. Similarly, there is no possible pro-form *wug which
could replace the sequence read a short, yielding *The new teacher wugged book in the
library. The possible substitutions correspond to the constituents in (4.1).
The second criterion is permutation, that is a constituent may occur in different
positions in a sentence while retaining its structural unity. This can be seen in the
alternative forms of (4.3a), exemplified in (4.4), (4.6), (4.7) and (4.8).
(4.4) a. In the library, the new teacher read a short book.
b. ?The library, the new teacher read a short book in.
c. *In, the new teacher read a short book the library.
In (4.4a) the PP in the library occurs at the beginning of the sentence; the preposition
alone cannot occur at the beginning of the sentence, as in (4.4c). This is evidence that
PP is a constituent. There are instances in which less than a full PP can permute, as in
(4.4b), an example of preposition stranding. It is illustrated further in (4.5).
(4.5) a. The teacher gave a book to the student.
b. The student the teacher gave a book to.
b′. To the student the teacher gave a book.
c. Who did the teacher give a book to?
c′. To whom did the teacher give a book?
An important difference between the PP in (4.4a) and the one in (4.5a) is that the latter
is a syntactic argument, the indirect object, while the former is an adjunct. While the
conditions on preposition stranding are very complex, it is generally true that it is more
acceptable with argument PPs than with adjunct PPs.
(4.6) a. A short book the new teacher read in the library.
b. *A short the new teacher read book in the library.
c. *Short book the new teacher read a in the library.
d. *Book the new teacher read a short in the library.
In the examples in (4.6), the NP following the verb, a short book, appears in initial
position, and again this is only possible if the entire NP occurs initially; it is not
possible to have just the head noun or the head noun plus one but not all of its modifiers
in initial position with its (other) modifiers occurring later in the clause, as in (4.6c) and
Constituent structure
113
(4.6d), nor is it possible for the modifiers to occur initially with the head noun later in
the clause, as in (4.6b).
(4.7) a. The new teacher wanted to read a short book in the library, and read a
short book in the library she did.
b. *The new teacher wanted to read a short book in the library, and read she
did a short book in the library.
c. ?The new teacher wanted to read a short book in the library, and read a
short book she did in the library.
The sentences in (4.7) involve VP-preposing (also known as VP-fronting), and the
verb alone cannot occur in initial position, as (4.7b) shows. Interestingly, the verb plus
the direct object NP seem to form a possible constituent, as (4.7c) shows; this will be
discussed further below.
(4.8) a. A short book was read by the new teacher in the library.
b. *The new book was read a short by teacher in the library.
c. *A short was read book by the new teacher in the library.
d. *Book was read a short by the new teacher in the library.
e. *A short book the new was read by teacher in the library.
f. *A short book teacher was read by the new in the library.
As discussed in chapter 2, in an English passive construction the direct object NP
of the corresponding active sentence appears as the subject and the subject NP of the
corresponding active occurs as the object of the preposition by, if it occurs at all, as in
(4.8a). What is relevant for this discussion is that this alternation involves whole NPs
and not subparts. In (4.8b), the head noun of the erstwhile direct object NP a short book
has replaced the head noun of the subject NP, while its modifiers remain in postverbal
position; in addition, the head noun of the erstwhile subject NP appears as the object of
the preposition by. In this example, only the head nouns have changed their positions
and functions, and the result is quite ungrammatical. In the remaining examples, a head
noun alone or modifiers alone have permuted, with predictable ungrammatical results.
Thus, in all of the different permutations in (4.4)–(4.8), it is whole constituents that
change function or position in every instance, and they are for the most part the same
constituents that were identified by the substitution test in (4.3).
The final test is coordination: only constituents may be linked, usually by a coordinate
conjunction, to form a coordinate structure. This is illustrated in (4.9)–(4.11).
(4.9) a. in the table and under the chair [PP PP and PP]
b. on the table and the desk P [NP NP and NP]
c. on and under the table [[P P and P] NP]
d. *on the and under a table.
d′. *on the big and under a small table
(4.10) a. the happy boys and the angry girls [NP [NP ART ADJ N] and
[NP ART ADJ N]]
b. the happy boys and angry girls ART [? [? ADJ N] and [? ADJ N]]
c. the happy boys and girls ART ADJ [N N and N]
d. *the happy and the angry boys
An introduction to syntax
114
(4.11) a. Kim read a book at home and wrote a poem [VP [VP V NP PP] and
in the library. [VP V NP PP]]
b. Kim read a book and wrote a poem in the [VP[? [? V NP] and
library. [? V NP]] PP]
c. Kim can read and write a poem. [V V and V] NP
d. *Kim read a and wrote the poem.
In the examples in (4.9)–(4.11) the whole phrase forms a coordinate PP, NP or VP in
(a) and the heads form coordinate Ps, Ns and Vs in (c). In the (b) example in (4.9)
coordinate NPs form a coordinate object of the preposition. In the (b) sentences in
(4.10) and (4.11), on the other hand, the constituents being coordinated are neither heads
nor whole phrases; they are apparently a new type of intermediate constituent; this
appears to be the same constituent that the permutation identified in (4.7c). This will be
discussed in section 4.3 below. Finally, all of the ungrammatical examples involve
conjoining a sequence of words which is less than a full constituent, i.e. a preposition
and NP modifier in (4.9d) and (4.9d′), an article plus adjective in (4.10d), and a verb
plus an NP modifier in (4.11d).
One might well wonder, how many tests must a group of words pass before it can
be considered a constituent? Ideally, it should pass all three, but minimally it must pass
at least one. All of the constituents discussed thus far pass all three. These tests can be
used to diagnose instances of structural ambiguity. Consider the ambiguous sentence
in (4.12a).
(4.12) a. Robin decided on the train.
b. On the train, Robin decided.
c. The train was decided on by Robin.
The issue here is whether on the train is a constituent or not. The sentence in (4.12a)
can have either of two meanings, depending upon whether on is analyzed as part of
the prepositional verb decide on, which means ‘choose’, or is a preposition heading
a prepositional phrase, in this case on the train. The two readings are ‘Robin made
the decision while on the the train’ (on as head of PP) and ‘Robin chose the train’
(decide on as prepositional verb). The ambiguity can be resolved by the permutation
test. If the PP appears at the beginning of the sentence, as in (4.12b), then only the PP
reading is possible. Since only constituents can be preposed in this way, on must form
a constituent with the train, and therefore the prepositional-verb reading is ruled out.
If the sentence is passivized as in (c), then on and the train are not part of the same
constituent but decide and on are, and therefore only the prepositional-verb meaning
is possible.
In chapter 3 there were examples of discontinuous NPs; the example from Kalkatungu
is repeated in (4.13).
(4.13) a. ‡a-ci japacara-qu kuva-ji oaji quar-· maoqa. Kalkatungu
1sg-dat clever-erg father-erg kill snake-abs mob-abs
‘My clever father killed the snakes.’
b. ‡a-ci kuva-ji oaji quar-· maoqa-· japacara-qu.
1sg-dat father-erg kill snake-abs mob-abs clever-erg
‘My clever father killed the snakes.’ (*‘My father killed the clever snakes.’)
Constituent structure
115
In (4.13b) an adjective modifying the ergative noun kuva-ji ‘father’ occurs separated
from it at the end of the sentence. Does this mean that ‹a-ci japacara-qu kuva-ji ‘my
clever father’ is not a constituent? Not necessarily. First, it can be replaced by a
pronoun, as in (4.14).
(4.14) ki-ji oaji quar-· maoqa-·.
3sg-erg kill snake-abs mob-abs
‘He killed the snakes.’
Second, it can permute freely in the sentence as a unit.
(4.15) a. ‡aci japacaraqu kuva-ji oaji quar maoqa.
b. ‡aci kuvaji oaji quar maoqa japacaraqu.
c. T_uar maoqa oaji ‹aci japacaraqu kuvaji.
d. T_uar maoqa ‹aci kuvaji oaji japacaraqu.
e. ‡aci japacaraqu kuvaji quar maoqa oaji.
‘My clever father killed the snakes.’
Third, it can be coordinated with another NP to form a coordinate NP, as in (4.16).
(4.16) a. ‡a-ci japacara-qu kuva-ji maxapai-qu-jana oaji quar-· maoqa-·.
1sg-dat clever-erg father-erg woman-erg-and kill snake-abs mob-abs
‘My clever father and the woman killed the snakes.’
b. ‡a-ci kuva-ji maxapai-qu-jana oaji quar-· maoqa-· japacara-qu.
1sg-dat father-erg woman-erg-and kill snake-abs mob-abs clever-erg
‘My clever father and the woman killed the snakes,’ or ‘My father and the
clever woman killed the snakes.’
When the adjective appears separated from the conjoined NP, the result is ambiguous,
with the adjective being interpreted as modifying either of the ergative nouns. Thus,
despite the fact that the modifier does not have to occur adjacent to the noun it modifies,
the NP still functions as a unit and passes the constituency tests. This shows that
in some languages the elements making up a constituent do not have to occur adjacent
to each other; however, the typical case is for the elements making up a constituent to
be adjacent to each other.