5. The structure of clauses So far we have been talking of sentences

5. The structure of clauses
So far we have been talking of sentences as the maximal linguistic units that syntax deals
with. However, we may look at sentences from two different points of view: 1) as independent
units of communication; and 2) as complex entities, constructed from smaller entities. Under
the second viewpoint – which is the one relevant when it comes to syntactic analysis – the
term clause is usually introduced in addition to sentence. So a clause is a complex linguistic
entity which has a particular internal structure, and some clauses may further be used as
independent units of communication. In that case they are usually called main clauses. The
distinction betwen these two terms may be seen as one manifestation of the general one
between material and function, and may be illustrated by an analogy:
Topic for discussion: material vs. function
Say you want to start a restaurant, from scratch. You buy a plot of land and call in the
builders. Now, from your point of view, what they are going to build is a restaurant.
From the point of view of the builders, however, it is probably just another building.
They have to use the same kinds of things to build your restaurant as they would to
build your private home (bricks, tiles, planks, frames, doors, sheets of glass, pipes,
fittings, etc., etc.). So the complex structure that you – from your functional point of
view – call restaurant, the builders from their material, or structural, point of view
call building. Applying the same two perspectives to complex linguistic expressions,
sentence would correspond to restaurant, clause to building.
The purpose of doing the tests of substitution, movement and deletion is to identify the
‘building blocks’ out of which a particular linguistic construction has been built, and
nothing more. Yet we saw towards the end of the last section that there are differences
between the constituents, differences so far unaccounted for. Just as a house is built out
of different kinds of elements, a clause is built from different kinds of constituents.
Let us stay with the house-building comparison for a little while. The list of names for
elements that go into the construction of a house can also be seen as a list of materials,
or categories of things. These are all things that must be used. Yet you - as the builder are free to choose particular subtypes from within each category - say red bricks rather
than yellow bricks - as long as they will serve the proper constructive purpose of
bricks. In the same way, you are free to use whatever words you like as constituents in
the building of a clause, as long as the constituents serve the proper constructive
purposes. They must be of a material capable of doing the job. Obviously there are
positions in the total structure of a house where bricks go more naturally than others.
You use bricks for walls, not for roofs, for example. In the same way, there are
positions in a clause that more naturally call for one kind of constituent than another. If
you don’t follow the accepted conventions for building a house - if you put solid doors
in all the window-frames, for example - you still get a house, but an ‘ungrammatical’
house! In the same way, you get ungrammatical sentences if you don’t use constituents
of the proper categories in particular structural positions. But what are these categories?
And what are the functions they perform? For practical reasons we begin with the
second question.
23
We return to this terminological question in 5.5., after the introduction of the notion of
analytical level.
5.1. Syntactic functions
If you have done any grammatical analysis before, you are no doubt familiar with terms like
subject and object, perhaps also indirect object and predicate. You may not be aware,
however, that these terms are names for syntactic functions. Nor may you be aware that there
are in fact other syntactic functions. This section will introduce and discuss these matters in
relation to the examples already used in preceding sections.
5.1. 1. Subject and predicate
We established above in section 4. that the minimal grammatically complete sentence1 in
English consists of two constituents, like this:
[20]
S
Constituent 1
Constituent 2
someone
the radio
a button on the radio
someone on the button
the touch of a button
turned
was turned
We said that the constituents in each of the two columns could substitute for each other, and
therefore formed a paradigm. The important point about a paradigm is that all its members
have the potential to perform the same syntactic functions. Since it is constituents that perform
syntactic functions, and since we have just two constituents in [20], it follows that we are here
looking for two syntactic functions. The function performed be Constituent1 is called subject
whereas that performed by Constituent 2 is called predicate. Since the sentences in [20] are
the smallest possible grammatically complete sentences in English, it follows that every
English sentence must contain a constituent functioning as subject and one that functions as
predicate. We can extend the table in [20] to include this information:
1 For historical reasons we shall continue to use the letter ‘S’ as an abbreviation not only for ‘sentence’ but
now also for ‘clause’ . We shall also continue to use the term ‘sentence’ instead of ‘main clause’.
24
[20]
S
Constituent 1
Constituent 2
Subject
Predicate
someone
the radio
a button on the radio
someone on the button
the touch of a button
turned
was turned
Now, there are two different ways to describe these syntactic functions, illustrated by the first
entry someone turned. We could either say that [someone] is the subject of [turned] - or we
could say that [someone] is the subject of the sentence someone turned. Notice that in either
case the term subject applies to a relation between two things: someone is the subject of
either turned or someone turned. We shall adopt the second formulation and say that the
functional relation subject-of holds between a given constituent and the sentence of which it
is a constituent. The reason for this will become clearer as we come to discuss other syntactic
functions. In the same way, the second constituent, [turned], is the predicate of the sentence
someone turned.
Now, since both subject and predicate are relationships between a constituent and the sentence
in which it appears, we say that these functions belong to the sentence level of analysis.
5.1.2. Adjuncts
Consider now a sentence like the radio turned by a touch of that button. We have just
established that the radio turned is a complete and grammatical sentence, with [the radio]
functioning as subject and [turned] functioning as predicate. What about the constituent [by a
touch of that button]? By the deletion test it is an optional constituent. Nevertheless, when it is
present, it has a syntactic function, a function that we shall call adjunct. It is sometimes also
called adverbial, but we shall avoid that because of its similarity with the categorial term
adverb.
Subject, predicate and adjunct are the only three syntactic
functions at the sentence level of analysis.
But the adjunct - in virtue of its optional status - has a wider range of positions than the other
two. In fact, a constituent functioning as adjunct may appear before the subject, after the
predicate and even between subject and predicate:
25
[21]
S
Constituent 1
Constituent 2
Constituent 3
Constituent 4
Constituent 5
Adjunct
Subject
Adjunct
Predicate
Adjunct
By a touch of the radio
that button
-
turned
-
-
The radio
by a touch of turned
that button
-
-
The radio
-
by a touch of
that button
turned
There are, in other words, at most five structural positions at the sentence-level of analysis,
only two of which must be filled. That sentences can nevertheless be very long indeed is due
to the fact that the constituents that perform these functions are highly complex. We can
illustrate this claim by giving a table for the last sentence here:
[22]
S
Constituent 1
Constituent 2
Subject
Predicate
That sentences can nevertheless be
very long indeed
is due to the fact that the constituents that
perform these functions are highly complex.
We are very far from the ultimate goal of giving a description of this sentence down to word
level. Clearly, we need to look more closely at the kind of material that these constituents are
made of, and carry on the analysis on that. This is a matter of assigning the constituents to
proper categories.
The only categories that we have met with so far are the major and minor lexical categories the word classes (check back to 2.1. if you are in doubt about these). But we have just taken
great pains to establish that words and constituents are not the same - in particular, that a
constituent may consist of more than one word. So the categories that constituents belong to
must be different from word classes. Yet it would be natural to expect a connection between
word classes and the classes into which sentence constituents are divided, for constituents
contain words. And this is precisely the case.
5.2. Phrases
The commonly recognized term for a group of words that can have a syntactic function is a
phrase. In classical terminology this corresponds to the term syntagma, which is carried
26
across in the Danish tradition as a syntagme, a term clearly related to the notion of syntax
itself. Although English could easily have used the term ‘syntagm’, it has adopted phrase
instead, and so shall we. So sentence constituents are classified as phrases of various kinds.
As we have already seen several times, a sentence constituent may consist of one or more
words. If it only consists of one word, then necessarily that word must belong to some word
class. This is the basis for the recognition of various kinds of phrases. We get the following
picture:
[23] a.
b.
c.
d.
e.
a Noun Phrase (NP) contains at least a N
a Verb Phrase (VP) contains at least a V
an Adjective Phrase (AP) contains at least an A (does not occur at sentence level)
an Adverb Phrase (AdvP) contains at least an Adv
a Prepositional Phrase (PP) contains at least a P
These are the five kinds of phrase that we shall be operating with.2 In addition to these five
types of phrases there is one more construction type that may be a constituent of a sentence namely the sentence itself! As we shall see, this has farreaching consequences, not only for
analysis, but for explaining our ability to use and understand very long and apparently highly
complex sentences (like the one in [23]).
On the basis of this discussion we can expand the tables to include categorial information as
well, disregarding the entries with someone for the moment:,
[25]
S
2
Constituent 1
Constituent 2
Subject
Predicate
NP
VP
someone
the radio
a button on the radio
someone on the button
the touch of a button
turned
was turned
As you can see, only half of the word classes are associated with a phrase: there is no ‘conjunction phrase’,
no ‘pronoun phrase’, no ‘determiner phrase’, no ‘proper name phrase’, no ‘auxiliary phrase’. The reasons
for this are varied, and in some cases accidental. To mention just a few: there are no ‘pronoun phrases’ or
‘proper name phrases’ because pronuns and proper names in all relevant syntactic respects behave like
Noun Phrases - so they are usually considered subtypes of NP. However, recent syntactic reseach does
recognize what might be interpreted as ‘Conjunction Phrase’ (CP), various types of ‘Auxiliary Phrases’ (IP
for ‘Inflectional Phrase’; TP for ‘Tense Phrase’; AgrP for ‘Agreement Phrase’, etc.), and various types of
‘Determiner Phrase’ (DP; also QP for ‘Quantifier Phrase’, DegP for ‘Degree Phrase’, etc.) - but we
disregard them here, maintaining a more conservative and traditional terminology.
27
[26]
S
Constituent 1 Constituent 2 Constituent 3
Adjunct
Subject
Adjunct
PP
PP
NP
By a touch of
the radio
that button
the radio
by a touch of
that button
the radio
-
Constituent 4 Constituent 5
Predicate
Adjunct
PP
VP
turned
turned
-
turned
by a touch of
that button
It is extremely important to grasp the difference between functional terms like subject and
predicate on the one hand and categorial terms like NP and VP on the other. It is the difference between function and material that was discussed in relation to house-building above.
The constituent [the radio], for example, is always a NP because it contains a N - but it is not
always a subject. It depends on how it is used in a sentence. When we descibe a constituent as
a NP, we describe it in terms of its internal, material composition. When we describe the same
constituent as a subject, we describe it in terms of the job it happens to be performing in the
sentence under analysis. Since it is the same constituent that can be described from these two
points of view, there is of course a relationship between categorial and functional terms, a relationship which is often called realization: in the sentence the radio turned, the subject is said
to be realized by the NP [the radio], and the predicate is said to be realized by the VP [turned].
Look now at the entries with someone that we suppressed in the table. The word someone is a
pronoun. This name suggests some kind of affinity with noun - and the Danish term stedord
(as well as the Latin prefix pro-) makes it fairly clear what that affinity might be: a pronoun
is a word that may ‘stand in’ for a noun. Actually, this is not entirely correct (except for the
English pronoun one). What pronouns ‘stand in’ for are rather NPs (the first entry) or a
category that we haven’t introduced yet, but soon will (the second entry). It is this functional
equivalence which led us to say that pronouns (and proper names) are subtypes of NP above,
even though they do not contain a noun.
Somewhat similar remarks could be brought to bear on the sentence in table [23], repeated
here:
28
[27]
S
Constituent 1
Constituent 2
Subject
Predicate
?
VP
That sentences can nevertheless be very
long indeed
is due to the fact that the constituents that
perform these functions are highly
complex.
Its subject is: That sentences can nevertheless be very long indeed. What kind of category is
this? Try to delete the first word: That sentences can nevertheless be very long indeed. The
result looks quite like a sentence, with [sentences] as subject and the rest as predicate. The
word that (a conjunction) then indicates that the sentence following it has a special syntactic
status. Rather than being an independent, selfcontained sentence, it is ‘reduced’ to being a
constituent within a larger sentence. However, since it performs the function of subject in the
larger sentence and since the typical category to perform this function is NP, such sentences
are sometimes called nominal sentences (and even NPs!).
29
Towards the end of section 4.2. - in the discussion of the possibilitites of ‘reducing’ the
sentence the radio turned on even further - it was suggested that the two constituents [the
radio] and [turned] were ‘more’ obligatory than the constituent [on] - or perhaps ‘obligatory’
in a different sense. It turns out to be the latter. The two sentence functions subject and
predicate are obligatory in English (as opposed, for example, to Spanish, where only the
predicate function is). Therefore, the two constituents that actually perform these jobs in a
given sentence are obligatory. That it is the subject function which is obligatory is borne out
by the use of the words it and there as ‘dummy’ subjects in many contexts, like it is raining,
there was much shouting and grinning.
Notice that in each table, the function as predicate is performed by a VP. In contrast, the
function as subject is performed by NP in [25] and [26], but by S in [27]. The function as
adjunct is performed by PP in [26], but it can also be realized by S, NP, or AdvP. We return to
these matters again.
Brief summary and look ahead
− Sentence constituents are identified by the tests of movement, substitution and
deletion.
− Once identified, they are assigned a syntactic function at sentence level: either
subject, predicate or adjunct.
− There must be one and only one subject, and one and only one predicate per
sentence. There may be a varying number of adjuncts at various positions.
− When their syntactic function is determined, each constituent is assigned to a
syntactic category.
− Relevant categories at sentence level are NP (including Pro), VP, PP, AdvP,
and S.
The sentence level is the top level of analysis, and the analysis is complete when we have
reached word level. So the next few sections will be concerned more particularly with the
notion of analytical level itself, as well as with the identification of and justification for any
intermediate levels of analysis.
5.3. The notion of analytical level
Look at the two sentences in
[28] a.
b.
someone turned
someone was turned
As we have seen, [28] a. is in itself a complete, grammatical sentence. It has two constituents
of which the one functioning as subject is realized by the (Pro subtype of ) NP [someone],
while the other functioning as predicate is realized by the VP [turned]. Now, in [28] b. we
have the same two words - but we also have a third word, was.
By the standard test of substitution, [was turned] in [28] b. is a VP which functions as
predicate. It is a complex constituent, so it requires analysis along precisely the same lines as
those we have so far only used at sentence level - only now we have moved ‘down’ to the
30
phrase level of analysis. The first constituent of the VP [was turned] is the Aux [was]. The
second constituent is the V [turned]. Illustrating this in the familiar way we get
[29]
S
Subject
Predicate
NP
VP
someone
?
?
Aux
V
was
turned
The question marks indicate that Aux and V have syntactic functions within VP, but that we
haven’t said anything about them. Before we can do that we need to discuss two things:
different ways of representing syntactic information and how to recognize analytical
levels.
5.3.1. Analytical levels: Representing syntactic information
The tables like that in [29] that we have been using in order to show as clearly as possible the
analytical results of the more theoretical discussion of constituency, categories and functions
are one way of representing syntactic information, and they have been widely used (especially
in Danish grammar). It shows that the word someone is a NP and the subject of S, that was is
an Aux and turned a V, both with so far unexplained functions within the VP was turned,
which is the predicate of S. It does not show, however, that NP, VP, Aux and V are category
symbols, while Subject and Predicate are functional terms, and it does not show that only the
functional terms are relational. Although all of this is syntactically significant information, it
is clear that this mode of representation only makes some items of syntactic information
explicit while suppressing others. Another way to represent the information just stated might
be the following:
[30] [S [NP(Subject) someone NP] [VP(Predicate) [Aux(?) was Aux][V(?) turned V] VP] S]
This method - called ‘labelled bracketing’ - is more difficult to ‘read’ than the tables although
it makes explicit precisely the same information. One reason for this is that it is linear - onedimensional - as opposed to [29], which is hierarchical, or two-dimensional. Since syntactic
structure in some sense is hierarchical - small elements being put together to form larger
elements that are treated as units for the construction of still larger elements, and so on - it is
easier to ‘read’ a representation of it which directly reflects this fact.
Since the goal of syntactic analysis is to make explicit
as many aspects of syntactically relevant information
as possible, we should choose a form of representation
which attains this goal. From among several such systems of representation we shall employ the one most
widely used in the Anglo-American tradition - the
31
tree. A tree is a structure consisting of nodes and branches, like this:
This is a tree with five nodes and four branches. Now, by convention, the nodes represent
categories, so we give them category names (NP, N, VP, V, Aux, etc.), whereas the branches
represent functional relationships. Suppose we wanted to make these functional relationships
explicit by naming the branches in an appropriate way. At the same time we replace the
circles indicating the nodes by the appropriate category names:
The words become the ‘leaves’ of the tree,
which has its ‘root’ (S) on top (as opposed to
normal trees). As you can see, there is now a
clear distinction between category and function
terms, and it is made clear that function terms
are always relational: a syntactic function is a
relationship between one category and the
category of which it is a constituent. This is a
consequence of the so called ‘single mother’
convention, which states that a node can only
be connected with one node above it. Another
important convention for the drawing of trees
forbids branches to cross. We have already
introduced the names of the functions Subject
and Predicate, but not the names of the two functions Modalization and Head. They will be
explained in section 6.2.
5.4. Objects and their analytical level
Having in this way introduced a new notation for representing syntactic information by means
of tree-structures, we can now return to the main theme. Look at this triplet of sentences:
[32] a. someone turned
b. someone had turned
c. someone had turned the radio
By the substitution test we can establish that had turned is the predicate of [32] b. So,
although the word had does not appear in our original list of 12 words, we can nevertheless
include it as a member of the same paradigm as was.
The next question then is what to do about the radio in [32]c. It contains a N (radio), so it is a
NP. But what is its syntactic function?
We have claimed that there are only three different syntactic functions relevant to what we so far without discussion - have called the sentence level of analysis, subject, predicate and
adjunct. From these premises, there are two possible answers to the question of what syntactic
function the radio has in [32] c: either it is an adjunct, or it has a function that does not belong
to the sentence level of analysis. Although it is by no means impossible for a NP to perform
the syntactic function of adjunct, there is another criterion that adjuncts are obliged to meet: it
is optional. In other words, we should be able to leave out [the radio] from [32] c. under
preservation not only of grammaticality but also of meaning, in the precise sense explained in
4.3. We can’t, in this case. The result would be [32] b. which is grammatical, alright, but
which does not preserve meaning. The conclusion must be that the radio has a syntactic
function that does not belong to the sentence level of analysis.
32
The tree for [32] b. is quite similar to [Tree 2] above, except for had instead of was as Aux:
We still haven’t talked about the functions performed by Aux and V (see section 6.2.). In the
meantime the question is how to accommodate
[29 c], given [Tree 3]. In particular, where we
should attach the NP node that dominates the
radio and thus indicate its analytical level. One
way to find this out is to examine if [the radio]
forms a constituent with any of the material
already in [Tree3]. Consider in this connection the
following data:
[33] a. I hadn’t turned the radio but someone had turned the radio
b. *I hadn’t turned the radio but someone had turned the radio
c. *I hadn’t turned the radio but someone had turned the radio
This is a slightly more complicated variant of the deletion test, by which we have conjoined
two sentences by means of the conjunction but. The trick is that even in normal, everyday
conversation we don’t repeat information that is already given. The assumption behind this
version of the deletion test, however, is the same as that behind the standard version: what we
leave out is a whole constituent at some level of analysis. As appears from [33] the only
grammatical result of this test is [a] - the example in which we have left out [turned the radio].
Leaving out either [had turned the radio] or [had] + [the radio] yields ungrammatical results.
Consequently, [turned the radio] forms one constituent at some level of analysis. The tree
which reflects this is [Tree4]:
As you can see, [turned] alone is still a V
with the function of Head, and [had] is still
an Aux. But the VP realizing the predicate
is now [had turned the radio] and a new
categorial node marked by three question
marks has been introduced to capture the
fact that [turned the radio] is an
independent constituent within VP, in
accordance with the deletion test just discussed. In addition, we now have two unexplained functions to account for - the one
performed by the new category (???), and
the one performed by NP (the radio). Let
us concentrate on the last for the moment.
5.4.1. (Direct) object
This function - performed by a NP inside a VP - is called object - sometimes also direct object. The Danish term is genstandsled. Traditionally, the object has always been considered
33
particularly closely ‘linked’ to the verb (indeed, verbs are subclassified in terms of how many
objects they take) - and this tree-structure makes it clear why: V forms a constituent with the
NP that functions as object. Also traditionally, objects are mentioned in the same breath as
subject. This is understandable from one point of view, but not from another. Like subject,
object is a relational term, as indeed are all functional terms. But - as we have just seen whereas subject is a function at sentence level, object is a function at a lower level of analysis
(the level marked by ??? in [Tree4]). But what name should we give to this level? Put
differently, what category name should we substitute for ??? in [Tree4]? This will emerge
from the next Topic for discussion box.
Topic for discussion: Recognition af analytical levels
The Bar-level
So far we have been operating with the sentence level, the phrase level and the word level of
analysis. These were all recognized and accepted as necessary to syntactic analysis by the
classical grammatical tradition. However, evidence of the sort presented in [29] has made it clear
that a level intermediate between word and phrase is needed to give an explicit description of the
structure of English and many other languages. Nevertheless it wasn’t introduced into general
syntactic theory until 1970 - under the term bar-level. Because of its youth, it is the most
controversial of the analytical levels. It wasn’t recognized by classical grammar, though
‘semiclassical’ terms like ‘nominal’ and ‘verbal’ indicate that it was nevertheless sometimes felt
to be needed.
The term bar stems from the original notation that was used to indicate a bar-category - that is,
a category belonging to the bar-level of analysis. And that notation again was designed to show
the affinity of the bar-categories with the lexical categories N, V, A, P on the one hand and their
associated phrasal categories NP, VP, AP, PP on the other. The original notation was simply the
names of the lexical categories with a short bar ( - ) above it. However, since it is difficult to
write such a compound symbol on a standard word processor, it was soon changed into
something else, namely: N´, V´, A´, and P´. These are the bar-categories usually recognized, and
the ones we shall use. In addition (and for slightly different reasons, which will be explained in
due course) we shall also recognize S´ . They are pronounced N-bar, V-bar, A-bar, P-bar and
S-bar. In Danish, the term bar has been translated into bjælke, and the order has been reversed,
so we have bjælke-N, bjælke-V, bjælke-A, bjælke-P og bjælke-S.
Now, let ‘X’ stand for any one of either ‘N’, ‘V’, ‘A’ or ‘P’. We then generally have three
distinct analytical levels, XP, X´ and X, for any value of X. For this reason the theory explained
in this box goes by the name of ‘the X-bar theory of syntax’. By convention, the bar-level is
recursive, which means there may be 0, 1, 2 or more bar-categories in a single phrase.
5.4.2. Indirect object
We mentioned in passing in the discussion box above that verbs are traditionally subclassified
in terms of how many objects they take. Turn - the V we have in our lexicon - can, as we have
seen throughout, take either 0 or 1 object.The verb give, in contrast, is the standard example in
grammatical textbooks of a verb that takes two objects:
34
[34] someone had given the radio a button
Never mind that this may be an odd thing to say in most normal circumstances - what matters
is that the sentence is grammatical and you understand it (that is why you find it an odd thing
to say!). Based on our discussion so far, the tree that would reveal the syntactic structure of
[34] would be [Tree5]:
Comparing this with [Tree 4] you will see that [the radio] has changed function - from Direct
to Indirect Object. The Direct Object in [29 a] is the NP [a button]. How do I know that?
An exhaustive explanation will have to rely on matters we haven’t discussed yet, in particular
the interplay between semantic and syntactic functions. So for now we’ll just give first a
somewhat dogmatic explanation, then a purely syntactic one.
Try to apply the deletion test to the two objects in turn, yielding
[29] b. someone had given the radio
c. someone had given a button
Are these grammatical? Yes, they are. Do they preserve meaning relative to [29 a]? No - and
yes. In the unlikely, Alice in Wonderland-like, situation of a radio giving a birthday party, you
could use either [a] or [c] to describe the same situation of someone presenting the radio with
a button - only [c] requires the additional condition that it must already be established from
the context who receives the gift. In contrast, [b] could not be so used, for it can only mean
that the radio is the thing given. The dogmatic explanation now is that we cannot have a
sentence containing an indirect object unless it also contains a direct one. We cannot - in
short - delete the direct object, hoping that our listener will interpret the remaining object as
indirect. He won’t! (Except in a very few cases, like May I serve you tea now?).
The purely syntactic explanation is based on the movement test. Consider
[29] d. someone gave a button to the radio
e. someone gave the radio to a button
35
Again, both are grammatical, but only [d] preserves meaning relative to [a]. Indirect objects
are so called because they have greater freedom of movement than direct objects. But notice
that you can only move the indirect object if you ‘convert’ it from a NP to a PP introduced by
to (in some cases for).
5.5. The analytical level of sentence revisited: The notions of Clause and Recursion
Before leaving the subject of analytical levels for the time being we return briefly to the
sentence level. You may recall that so far we have taken intuitive knowledge of sentences and
words as our basic assumptions, and we have also said that the sentence - by definition - is the
largest unit of syntactic description.
Now, it may appear strange that there is in fact no upper limit to how large the largest unit of
description may be. But it is a fact, for reasons given below. So far, however, our main
concern has been the opposite: How small can it be? This has concerned us because it is a
question about the necessary and sufficient number of constituents that form a structure that
may function as a sentence. It is the smallest such structure that is usually called a clause
(there is no appropriate Danish term for it, but it is best translated as sætningsstruktur). Quite
often the clause is taken to define its own analytical level, between sentence- and phrase-level.
This is quite understandable, for much of syntactic and general grammatical analysis is in fact
only relevant to clause ‘level’. Notions like subject, predicate, finiteness, etc. are readily
explained in relation to clauses instead of sentences. Nevertheless, sentences are recognized as
syntactic units, because they have predictable structures (as opposed to paragraphs, chapters,
and texts). So, although sentences may contain multiple clauses just like texts may contain
multiple sentences, there is a difference. We can always give a purely structural description
of multiple-clause sentences without recourse to meaning, which we cannot do with texts.
The relationship between the two ‘levels’ of sentence and clause is the important one of
recursion - the fact that the same kinds of structure can be found at different levels of analysis:
a sentence is a structure which can be analysed into constituents, some of which are
themselves classifiable as sentences!
There are two main types, already recognized by classical grammatical theory: coordination
and subordination. In structural terms we have the two configurations in [Tree6] and [Tree7]:
36
The S´(S-bar) in [Tree7] is the symbol of subordination, and Conj is a conjunction. Although
the encircled S’s in these configurations are usually called clauses, the fact remains that they
are to be analysed in precisely the same terms as any other S, as you can see.
We have stated that the ‘smallest’ grammatically complete sentence in English contains two
constituents, one functioning as subject, the other as predicate. In addition, there may be one
or more adjuncts. But how then to explain grammatically ‘complete’ expressions like
[35] a. run!
b. help!
c. yes.
d. no.
e. damn!
f. oh?
Of these, [a] and [b] are usually classified as imperative sentences, [c] and [d] as sentence
substitutes, and [e] and [f] as exclamations (or even exclamatory sentences!).
37
These various points all lead to a rather different way to account for the distinction between
sentence and clause, not in terms of a difference in descriptive levels, but in terms of the
standard distinction between category and function. The term clause in fact applies to a
particular kind of grammatical structure - that is, a structure which can be analysed into at
least a subject and a predicate. We are looking at the structure from the point of view of its
internal composition. The term sentence, on the other hand, applies to any structure which is
capable of independent communicative function. It is only from this latter point of view that
the expressions of [29] can be called sentences. They are communicatively independent
expressions, but they do not have the internal structure with a subject and a predicate that
would qualify them as clauses.
5.7. Summary of main points
o Syntactic analysis is concerned with revealing the syntactic structure of sentences.
o A sentence ultimately consists of words, each of which belongs to a particular lexical
category: the major categories N, V, and A and the minor ones Pro, Det, Aux, Adv, P,
Conj, and Proper Name.
o However, the sentence is not built directly from words, but from groups of words.
o It is these groups that are the immediate constituents of sentences. A sentence
constituent may contain from one up to a great many words.
o The initial step in the analysis of a particular sentence is to identify its constituents.
o The methods used for this involve three supplementary tests, movement, substitution,
and deletion. All three tests presuppose the notion of grammaticality.
- The movement test determines the constituency boundaries and hence help you
determine what the constituents are.
- The substitution test will yield constituents that may perform the same syntactic
functions. They are said to form a paradigm.
- The deletion test will help you distinguish between obligatory and optional
constituents. It presupposes preservation of (truthfunctional) meaning.
o Having identified the constituents, you assign each to a particular syntactic function.
o A syntactic function is the relation between a category and the larger category of
which it is a constituent.
o In consequence of this, syntactic functions define a number of analytical levels.
o There are four analytical levels: sentence-level, phrase- level, bar-level and word
level.
o The syntactic functions belonging to the sentence level of analysis are Subject,
Predicate and Adjunct.
o The syntactic functions of direct and indirect object belong to the bar-level of
analysis.
o [There are syntactic functions at phrase-, bar- and word level that we haven’t yet
mentioned; see section 6].
o Having assigned each constituent to a particular function, you assign each of them to a
phrasal category, or to S´ (cf. last point on the page).
o The phrasal categories are NP, VP, AP, AdvP, and PP.
o A NP contains at least a N, a VP at least a V, etc.
Exception: Pronouns and Proper Names are regarded as subtypes of NP, not
because they contain N, but because they have the same syntactic potential as NPs
that do contain N.
o Sometimes a sentence has a sentence as a constituent. This is known as Recursion.
38
Important points
o Distinction between category and function. When you describe a
constituent as a member of a category (say NP), you do so from the
point of view of the material it contains (in this case N). When you describe the same constituent as a subject, you do so from the point of
view of the syntactic job it does in the sentence.
o Sentence vs. clause. The distinction between category and function is
usually reserved for phrases, bar-categories and words. But it can also
be applied to (structures that we intuitively call) sentences. When
described in terms of its internal composition (the categorial view) such
a structure is a clause if it contains at least two constituents one of
which is subject, one of which is predicate. When described in terms of
its function as an independent unit of communication, then it might be
called a sentence, where ‘sentence’ is then used in a narrower sense
than the intuitive one above.
o Representation of syntactic information. Carrying through the
analysis of a sentence has little purpose if we haven’t got a notational
system that will make the result of the analysis clear. Among many
such systems we have adopted a system that represents syntactic
information in the form of tree-structures. Let us finish this summary by
drawing a tree for one of the sentences that we have been using for
illustration throughout. The following sections will then deal with the
internal structure of the various phrases that make up a clause.
39
40