Professional Documents
Culture Documents
Clause Structure, Complements, and Adjuncts: February 2020
Clause Structure, Complements, and Adjuncts: February 2020
net/publication/339106818
CITATIONS READS
0 5,465
1 author:
Patrick Duffley
Laval University
87 PUBLICATIONS 328 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Patrick Duffley on 07 February 2020.
This chapter discusses central concepts in clause structure such as the notion of ‘head’
and the distinction between ‘complements’ and ‘adjuncts’. The different ways in
which constituency and dependency grammars approach clausal heads are examined,
and modern analyses are juxtaposed with the grammatical tradition. Regarding the
distinction between complements and adjuncts, two important issues are discussed: its
relevance to the clause’s semantic structure and how to determine obligatoriness, the
the challenge of explaining complement choice with individual verbs, either by means
Keywords
Chapter 17
Patrick Duffley
17.1 Introduction
When we speak, we do not simply utter disconnected sequences of words, each of Mis en forme : P
which is a stand-alone unit. Words are organized into coherent groups called phrases
(for example, the beautiful baby is a noun phrase composed of the article the, the
adjective beautiful, and the noun baby); phrases are grouped into clauses and clauses
into sentences (in the sentence The parents spoke quietly because the baby was
asleep, there are two clauses— – a main clause, The parents spoke quietly because the
baby was asleep, and a subordinate clause, because the baby was asleep). A clause is
made up of phrases constituting a subject and a predicate. The subject is what the
clause is about (e.g., the parents); the predicate denotes a property attributed to the
Clauses can be finite or nonfinite. A finite clause has a predicate whose core
element is a verb form conjugated for person and tense, as in The baby was asleep. A
nonfinite clause has a predicate whose core element is a verb form which is not
conjugated for person or tense, as in The baby being asleep, the house was quiet.
There are four basic types of nonfinite clauses in English, to-infinitive clauses (For
the baby to be asleep already was unusual), bare infinitive clauses (What! Me not be
on time for a meeting!), past participle clauses (The dishes done, they returned to the
living room) and -ing clauses (Things being what they are, we had better not press the
This chapter provides an overview of various ways in which the internal structure
of clauses can be analyszed. It will address questions such as what determines clause
type (headedness), which components of a clause are obligatory and which are
optional (complements versuss. adjuncts), and how the predicate constrains the choice
2
A central concept in clause structure is the notion of ‘head’, a term which goes back
endocentric structures in terms of the relation between the head and the overall
resulting word-group does not belong to the class of any of its immediate constituents
(the sentence John ran is made up of John, a noun, and ran, a verb, but is itself
hand, the resulting phrase belongs to the class of one of its components, designated as
its head. In coordinate endocentric constructions such as boys and girls, the construct
belongs to the class of both of the coordinated constituents, giving rise in this case to
whole phrase belongs to the same class as the head, here the adjective poor. The
few years before Bloomfield by Otto Jespersen (1924 [1975]: 115–116): a junction
such as the ‘adjective + noun’ sequence new-born dog forms one single
denomination, “a composite name for what conceivably might just as well have been
called by a single name,” i.e. puppy. A nexus such as the ‘noun + verb’ sequence
Dogs bark “contains two ideas which must necessarily remain separate.”.
In the 1970s, Noam Chomsky proposed that the distinction between exocentric
approach that amounts to treating sentences as extended verb phrases.2 The internal
1
I = Inflection.
2
This gave rise to what is called X-bar Theory (see Chomsky 1970).
3
structure of the sentence thus construed has sometimes been analysed as ternary (NP–
Aux–VP),3 as in Radford (1981: 38), but the recent tendency has been to view it as
binary (NP–I′), with I′ being constituted by I + VP (see Aarts and Haegeman 2006:
139).
As regards the question of determining the head of a clause, two main lines of
thought are found in the literature, represented by constituency grammar on the one
hand (see Borsley, this volume) and dependency grammar on the other (see Herbst,
that shares morphosyntactic features with the construct as a whole (Zwicky 1985: 6).
Heads are needed in constituency grammar to account for the percolation of features,
whereby higher-level constructs inherit features from the component that dominates
the whole (Zwicky 1985: 10–11; Bauer 1990: 21): when a noun is the head of a
phrase, for example, the whole phrase behaves syntactically like a noun. In
treats as a single constituent, e.g., [x y z]: y and z might both depend on x; x and y
could depend on z; x on y and y on z, etc. (see Matthews 1981: 84–85). Hudson (1993:
272–273) observes that “virtually any assumptions about heads are compatible with
the basic assumptions of constituent structure.” Thus, although X-bar theory and
3
NP = Noun Phrase; Aux = Auxiliary Verb; VP = Verb Phrase.
4
Head-Driven Phrase Structure Grammar4 are both constituency grammars, in the
former the mother and the head-daughter of a clause must have the exact same
features, whereas in the latter they are required to share only some features (Gazdar et
al. 1985: 52). Seuren (1998: 226) observes that for a generative-type model
constituency trees fit better with the theoretical syntactic machinery of the grammar
because “the tree structure itself is part of the formulation of the syntactic
Modern dependency grammars are built on the model of Tesnière (1959) and,
like X-bar theory, take the verb as the head of the sentence, viewing the latter’s
argument-structure of the verb.5 This entails that subjects, direct objects, and indirect
objects are seen as dependent upon the verb. Hewson (1991b: 2364) points out that
the verb was viewed as dependent upon its subject, this dependency being marked by
its agreeing grammatically with the subject. Thus for a traditional grammarian like
Jespersen (1924: 100), the subject is a primary, as is the noun houses in the NP big Commenté [B1]: AQ: quote as above? (1924 [1975]) and as
per the biblio?
houses, and the verb a secondary, like the adjective big in the same construction. Commenté [PD2R1]: As per the biblio.
be superior to valency grammar in that it is able to explain: (1) why the verb agrees
4
Head-Driven Phrase Structure Grammar (HPSG) is a generative theory in which the lexicon
is more than just a list of entries but is richly structured and determines much of the syntax.
5
An argument is an expression needed to complete the meaning of a predicate. For example,
the verb like requires two arguments, a ‘liker’ and ‘something/someone liked’.
5
with its subject and not with its direct or indirect objects, a fact which is mysterious if
all three of these elements are placed on the same level as valency-slot fillers or
‘complements’ of the verb, as in Huddleston and Pullum (2002: 215); (2) why a finite
verb may or may not have an object, but typically cannot exist without a subject (e.g.,
Traditional grammar would also seem to have the advantage over both
perceived semantic structure of clauses, in which one feels that the predicate says
something about its subject, and not merely that the subject forms a unit with the
predicate or fills a valency deficit in the latter. As Searle (1969: 119) points out, “the
subject serves to identify an object;6 the predicate, if the total illocutionary act is one
been identified.”. Cruse (1986: 105–106) observes in this respect that the predicate
“displays at least one of the characteristics of a semantic dependent, and that is that it
such as Hirtle (2007a: 121), who utilizes the notion of applying a semantic import
relation.7
Grammar (1988: 194–196; see Cchapter 8) treats Fred as having two heads in (1):
6
Searle uses the term ‘object’ here in its everyday sense to refer to an entity.
7
Hirtle uses the term ‘incidence’ to characterize this import/support relation.
6
(1) Fred keeps talking to Mary about himself/*ourselves.
Mis en forme : EXT
This is because Fred is felt to be the subject of both the verb keeps, which agrees with
Classical generative grammar accounts for why Fred is felt to be the notional subject
of both verbs by ‘subject-to-subject raising’: Fred starts out as the subject of talking
in the lower clause and is raised to the role of subject of keeps in the main clause.
Both the double-head and the raising accounts are problematic, however. The raising
analysis is based on the claim that keep does not assign any theta-role8 to its subject,
i.e. Fred is not represented as a ‘keeper’ in (1) above, unlike in Fred kept my pen. The
validity of this claim depends crucially, however, on how one defines the semantics of
Webster’s dictionary (1961), it can cogently be argued that in (1) Fred is the one who
persistently remains engaged in the activity of talking to Mary about himself and so
does have a theta-role with respect to the keeping. Of course, by itself this verb does
not make sense as a predicate of Fred (*Fred keeps), but this is simply because its
meaning requires the specification of the activity in which the keeper remains
engaged: since the keeper role is defined as the continuing occupation of the talker
role, the former cannot be conceived independently of the latter. As Achard (2001: 9)
8
Theta-roles (i.e. thematic roles, also known as semantic roles) are referential relations that
arguments have with respect to the action denoted by the verb, e.g., agent, patient,
7
points out with regard to the raising verb seem in the French equivalent of Marie
As for the double-head analysis, it remains a surface observation that keeps and
talking are understood to share a common subject, but why this should be the case is
not explained. Duffley (2006: 113–115) argues that in structures like (1) the -ing form
sequence keeps talking evokes the idea of the subject maintaining itself within the
ongoing activity denoted by the gerund-participle, which obviously implies that the
Hudson (2003a) extends the double-head analysis also to control verbs10 such as
try in (2) below, claiming that the difference between raising and control is not
9
Subject complements (also known as subjective predicative complements) are adjectival or
nominal elements that are related to the subject through the verb, as is the adjective quiet in
He kept quiet until they left. The adjective evokes here the state in which the subject
10
Control verbs occur in constructions in which the understood subject of a predicate is
determined by another expression in a higher clause. Thus Fred stopped laughing is a case
of subject control, as the subject of the control verb stopped is also the understood subject
of the subordinate verb laughing; Fred forced Joe to do it, on the other hand, manifests
object control, because the object of the control verb forced is understood to be the subject
8
Mis en forme : EXT
According to Hudson’s syntactic analysis of (1) and (2), both involve the grammatical
function of ‘sharer’, according to which the complement shares its subject with the
matrix verb, but, unlike the raising verb keep, the control verb try assigns a theta role
to its subject, and so Fred is understood to be not just the talker but also the ‘tryer’. As
with the Word Grammar approach to raising, however, this account fails to actually
explain the sharing of the subject Fred by tried and talk. Duffley (2014) argues that a
an explanation and predicates his account on the fact that the to-infinitive is not the
direct object of try, but rather a phrase indicating the goal of the trying. This analysis
is based on the contrast with the gerund-participle, which is claimed to be a true direct
object with this verb, as illustrated by (3) and (4) below in which only the -ing form
(3a) a. I have already tried talking to Mary, but she refuses to change her mind.
(3b) b. I have already tried that, but she refuses to change her mind.
(3c) c. Talking to Mary has already been tried, but she refuses to change her mind.
(4a) a. I have already tried to talk to Mary, but she won’t even meet me.
(4b) b. *I have already tried that, but she won’t even meet me.
(4c) c. *To talk to Mary has already been tried, but she won’t even meet me.11
11
Mair (1990: 105) points out that “although it is convenient to regard the infinitival complements of
the attempt-class of verbs [i.e. monotransitives] as objects, matrix verbs cannot be passivized in the
normal way”:
9
Mis en forme : EXT
This account is consonant with the fact that to has the same meaning when it occurs
with the intransitive verb strive in I strove to talk with Mary as it does with try in I
tried to talk with Mary: in the same way as to talk with Mary represents the goal of
the striving in the first sentence, it represents the goal of the trying in the second.
Thus in (2) above try implies a movement towards a goal, and to evokes the infinitive
talk as the terminal point of this movement. Given this configuration of meanings, it
is perfectly logical that the prospective realizer of talk should be understood to be the
same as the actual realizer of try. Control (e.g., try + to-infinitive) and raising (e.g.,
analysis such as that proposed by Hudson, which leaves no room for the intervention
Some authors such as Langacker (1997, 2000), working in the Cognitive Commenté [B3]: AQ: please check, Langacker 2000
appears to be missing from the biblio.
Grammar framework (see Taylor, this volume), employ both the notion of Commenté [PD4R3]: Here is the reference: Langacker,
Ronald W. 2000. Grammar and Conceptualization. Berlin
and New York: Mouton de Gruyter.
dependency, seen as more fundamental, and that of constituency, seen as a reflection
The same observation is made by Matthews (1981: 181–182) and Visser (1969: 1312). Their
observations are confirmed by the absence of any attested examples of the passive with transitive
verbs construed with the to-infinitive in the 100-million-word British National Corpus.
12
The distinction between raising and control has also been abandoned in the latest version of
generative grammar, Minimalism (see Boeckx et al., Hornstein and Nunes 2010), due to
the fact that this theory allows arguments to accumulate two or more theta-roles along their
movement-trajectory through a syntactic structure, which means that control can be treated
10
of “the basic psychological capacity for grouping” (1997: 9). Constituency is viewed
by Langacker as emergent rather than basic, reflecting the order in which a complex
grouping based on linear contiguity” (2000: 269–270). In contrast to generative Mis en forme : Surlignage
theory, in Cognitive Grammar constituency “is not invoked for the definition of
grammatical relations like subject and object,” which rather than being defined
correspondences between subparts of semantic structures” (Langacker 2000: 148– Mis en forme : Surlignage
150). Thus the subject of the clause is defined as the “trajector,”, i.e. the figure or
most salient element within the relational profile denoted by the verb (Langacker Mis en forme : Surlignage
2000: 33, 150). The direct object for its part corresponds to the “landmark” or
secondary figure. However, defining the subject merely as the most salient element
within a relational profile does not do justice to its semantic relation to the predicate,
nor does treating the direct object as a less salient secondary substructure distinct
from the subject-trajector. Langacker himself in another publication defines the direct
secondary figure” (1991: 323), and depicts it as “lying downstream from the subject
along an action chain or with respect to some abstract analog of energy flow” (1991:
13
An example of conceptual overlap would be the relation between the noun box and the
adjective old in the phrase old box: the adjective has a conceptual slot for an entity
possessing the quality that it denotes and the noun specifies the slot-holder.
11
is insufficient of itself to adequately describe the nature of subjects and direct objects
in English.
For Langacker, the head of a construction is the component that “imposes its own
profile as the composite structure profile” (2000: 212–213). Thus in the noun phrase Mis en forme : Surlignage
the table near the door, the head of the phrase is table, and in the prepositional phrase
near the door, the head is near. This echoes Bloomfield’s original definition of the
head as the determinant of the form-class of the whole phrase. The head of a clause is
(1997: 7). Thus in Alice likes Bill, the verb like profiles a relationship between a
trajector (Alice) and a landmark (Bill), and so corresponds to the head of the clause in
the sense of providing the essential conceptual content that characterizes the profiled
relationship (see also Langacker 2015). This analysis, like that of valency grammar,
Another important issue in the area of clause structure is the distinction between
complements and adjuncts. In its most general sense, a complement is a word, phrase,
1997: 75). Thus in He opened the door the noun phrase the door is necessary to
complete the meaning of the verb opened and is therefore its complement. Huddleston
and Pullum (2002: 219) argue that the most important property of complements is that
they require the presence of a head that licences them. The dependency between
subcategorized according to what kind of complements they take; for example, open
12
allows a noun object while look does not (*He looked the door), but instead requires
In certain approaches (e.g., Pollard and Sag 1994; Huddleston and Pullum 2002;
Miller 2011), the subject of the verb is treated as a verbal complement, based on the
removal does not affect the grammaticality of the remainder, as for example the
locative prepositional phrase in Mary met Joe at church. Besides subjects and direct
objects, other commonly recognized types of complement are indirect objects (Joe
gave Mary a bouquet), subject complements (Mary was ecstatic), object complements
(Mary found the room warm) and prepositional-phrase complements (He looked at
complements a property of the subject at the time expressed by the verb, object
complements a property of the object at the time expressed by the verb, and
Osborne et al., Putnam and Gross (2011: 328–329) point out that a dependency
grammar places adjuncts on the same level as complements. Langacker (2000: 212– Mis en forme : Surlignage
214), for his part, distinguishes between complements and “modifiers,”, the former
being components that elaborate a substructure of their head, and the latter
components whose substructure is elaborated by their head. Thus in the PP near the
i.e. it identifies the secondary figure with respect to which the relationship of
proximity is defined, and in the NP the table near the door the PP near the door is a
13
identifying the entity that is represented as being in a relation of proximity to the
door. One might wonder, however, why a similar analysis to that applied to
headedness on the clausal and sentential levels is not invoked on the phrasal level in
the table near the door: if the table is the trajector of near, and the door its landmark,
should not the preposition near be analyszed as determining the composite structure
profile as a relationship of proximity between two entities, i.e. as the head of the
overall phrase?
Two basic issues arise with respect to the distinction between complements and
adjuncts: (1) its relevance to the semantic structure of the clause and (2) how to
Grimshaw and Vikner (1993: 143) observe that “it is generally held that while
example of an optional one is the object in Adam was talking politics. A typical
lunch. Grimshaw and Vikner argue that there is a class of adjuncts that are obligatory,
namely those occurring with the passives of accomplishment verbs (note the contrast
in acceptability between This house was built last year and *This house was built).
Goldberg and Ackerman (2001) propose that obligatory adjuncts such as these can be
accounted for by pragmatic requirements, in this case the need for the utterance to
have an informational focus. Thus This house was built does not provide significant
information about the house, since we know that all houses are built. This observation
raises the very important question of the contribution made to the determination of
14
structure.14 Thus if one measures obligatoriness by the yardstick of the speaker’s
whole point of what the speaker wanted to say in uttering Mary met Joe at church. In
virtue of what principle can it be treated as optional? The ultimate idea behind the
the verbal event, called for by the very nature of the predicate,15 while the former are
essentialness versus accidentalness outside of a context: thus, for example, the verb
addressee; however in a use such as (5) there are but two arguments and there is no
(5) The author tells the story using the third person.
Mis en forme : EXT
complements: mobility and stackability (Aarts 2011: 105). Stackability concerns the
possibility of having more than one adjunct in a clause (Yesterday the sun was just as
it is in India). This is not distinctive of adjuncts, however, as one can also have two
complements functioning as object in the same clause (I sold [Gabriel]1 [my car]2 for Commenté [B5]: AQ: should this read 'objects'?
Commenté [PD6R5]: Yes.
14
The fact that one could accept This house was built in a fairy-tale, as in This house was built, but that
one just appeared out of nowhere, serves to confirm the importance of pragmatic considerations for
this question.
15
The technical term for this dependency of complements on predicates is ‘licencing’ (see
Huddleston and Pullum 2002: 219); the predicate eat is inconceivable without something
15
$2500).16. As regards mobility, complements can also sometimes occupy different
positions within the clause (I don’t need advice/Advice I don’t need), although there is
less freedom in this respect than with adjuncts. Some authors hold therefore that no
diagnostic criteria have emerged that will reliably distinguish adjuncts from
As regards the semantic structure of the clause, one may well wonder what the
optional or obligatory character of one of its components tells us about the relation
between this component and the rest. The idea behind the complement/adjunct
head, without which the latter would be incoherent, while an adjunct merely adds a
further characterization to its head, restricting the latter to a proper subset of its
denotation (see Dowty 2003: 34). Thus unleavened bread is a subset of bread, but the
latter is perfectly conceivable without the adjunct if one wishes to refer to the whole
category, whereas *they destroyed seems incomplete all by itself and calls for the
specification of some object that underwent the destruction expressed by the verb.
However, in the NP extremely heavy rocks the adverb extremely denotes a subset of
heavy and so qualifies as an adjunct, and yet one would not want to equate the
semantic relation between the adverb extremely and the adjective heavy with that
16
One might argue here that ‘complement’ is merely an umbrella term denoting very different
grammatical functions and that direct and indirect objects do not have the same
grammatical function whereas all adjuncts do. This opens up the debate about what
Unfortunately, space does not allow this issue to be addressed here. Suffice it to say that
from the semantic point of view there is just as much difference between a temporal and a
16
between the adjective heavy itself and the noun rock just because both denote a subset
of their head (for instance, extremely denotes the way heavy applies to the rocks, but
heavy does not denote the way the rocks are rocks). Even less would one want to
semantically equate the ‘adjective-noun’ relation between the adjunct unleavened and
bread to the ‘circumstantial-verb’ relation between slept and on the couch in He slept
on the couch. Thus while the distinction between complements and adjuncts is
obligatoriness may be an indication that certain elements require the presence of the
The term ‘complement’ is also sometimes used in a more restricted sense to refer only
to core arguments of main predicates (see Dixon and Aikhenvald 2006: 1; Achard
2007: 782–783). In this view, the presence of the preposition to in a sentence such as
complement.17 Here one of the most important tasks has been motivating the
distribution of complement forms with respect to the main verbs with which they
occur. Talmy Givón (1980, 2001b) has made an influential proposal of a hierarchy of
17
When there is no preposition, the noun phrase following the verb is considered a
17
complements,18 which claims a correlation between the morphosyntactic
independence of the complement and the conceptual distance between its event and
that denoted by the matrix verb. In English, Givón’s scale of event integration places
deverbal nouns above the gerund-participle, and the latter above the to-infinitive, with
There are significant problems with this proposal. Firstly, the distinguishability of the
different ranks on the scale is called into question by the fact that certain verbs are
construable with more than one complement construction with no significant change
in communicative import:
Furthermore, as far as the control relation between the subject of the main verb and
the complement is concerned, the scale runs in exactly the opposite direction to that
18
To cite just a few studies that refer to Givón’s binding hierarchy, see Foley and VanValin
18
showing practically exclusive subject control readings, with its subject being nearly
always identical to that of the main verb (e.g., He likes to control/He needs to
control). The gerund-participle, for its part, varies between subject and non-subject
controlling). The deverbal noun is indifferent to control and often allows both
readings with the very same matrix verb (He needs control could mean either ‘he
integration. Thus the bare infinitive has no semiological material intervening between
it and the main verb, and disallows temporal dislocation between the matrix and the
complement, while the to-infinitive, which has greater syntactic distance between the
matrix verb and the infinitive due to the interpolation of to, does not:
The reality of complement usage is more complex than this, however, as the verb
force is like make in not admitting temporal dislocation, but is construed nonetheless
Moreover, temporal integration does not correlate very well with the purported
and the gerund-participle’s events can be dislocated either towards the future or
19
Mis en forme : EXT
(9a) a. I anticipate losing some money on this deal/a net loss on this deal.
(9b) b. I regret making a mistake in your address/the mistake in your address.
Mis en forme : EXT
In addition, the to-infinitive can be temporally contiguous to the main verb’s event:
Furthermore, Givón (2001b: 59) places perception verbs considerably lower on the
however, both involve temporal integration, and syntactically both are construed with
On what basis are they placed on opposite ends of the scale of semantic binding
between matrix and complement verbs? And if they are so different semantically, why
are both construed with the very same complement form? The identity of syntactic
construction is explained in Duffley (1992: 29–30, 65–66), who argues that in I saw
Joe cross the street the perception cannot be conceived as existing prior to the action
perceived; similarly, since in I made Joe open the door the verb make denotes the
effectively producing the effect of Joe opening the door prior to the realization of the
20
Further problems are raised by Givón’s treatment of the ‘manipulation verbs’
make and have (2001b: 47). Based on the contrast between (12a) and (12b), he argues
manipulation’:
(12a) a. *He wasn’t in the room when she entered, so she made him come in.
(12b) b. He wasn’t in the room when she entered, so she had him come in.
(Givón 2001b: 45)
Mis en forme : EXT
Contrary to Givón’s claims, however, make does not always imply “direct contact
(13) Unfortunately several of bin Laden’s top lieutenants did escape our military in
Iraq and Afghanistan and we’ve never found them. One of them was Osama’s
right hand man, a cunning and ruthless murderer named Dr Fahad Manjikian.
He makes bin Laden look like a Boy Scout leader.
(https://books.google.ca/books?isbn=1479704539) Commenté [B7]: AQ: I was unable to access this link;
please check.
Commenté [PD8R7]: You could use the reference to the
book: Cauthen, John V. 2012, Chasing the wind. Exlibris
Here the idea is simply that of the effect produced on the image one has of bin Laden (ebook), p. 29.
Mis en forme : EXT
by the comparison between him and Dr Manjikian. Moreover, have does not always
involve an intermediary:
The idea signified by this verb is simply that of ‘having someone under one’s control
in the production of an effect’ (Duffley 1992: 72), which explains why there is an
or because the action requested is so trivial that compliance can be taken for granted
21
(e.g., She had me take a seat). This also explains why there is no impression of
subsequence between the having and the complement event, and consequently no to
All this points to the complexity of the inner workings of clause structure in
English. Abstract scales such as Givón’s binding hierarchy do reflect the semantic and
of what each of the components of such constructions means and how its meaning fits
Pullum 2002: 219–220). Verbs that take no object complements are classified as
intransitive (I left), while verbs with one or two objects are classified as
licenced. This occurs with the verbs mention and allude, where the first licences a
direct object (She mentioned the problem) while the second requires a prepositional
phrase with to (She alluded to the problem). Subcategorization frames can also be
observed with nonfinite verbal complements: thus the verb want licenses only the to-
infinitive, enjoy only the gerund-participle, and like both the to-infinitive and the
gerund-participle.
22
The subcategorization approach raises a certain number of questions, however.
Are subcategorization features merely syntactic tags attached to items in the lexicon,
or does the meaning of the verb play a role in determining what sort of
complementation it can take? Regarding verbs such as like, which admit more than
one type of complement, is this a mere syntactic fact about this lexical item or does it
correlate with a distinction on the level of meaning? Does the term ‘complement’
Regarding the first question of whether the meaning of the main verb plays a role
number of authors that the to-infinitive is associated with a specific type of meaning
which has been described by the notions of “hypothesis” (Bolinger 1978b: 10), “mere
potentiality for action” (Quirk et al. 1985: 1191), and “futurity” (Wierzbicka 1988:
188). In the light of the full range of infinitival usage, Duffley (1992: 89) broadens
this description so as to cover cases such as She managed to open the door, proposing
that the notion of movement expressed by to leading from the matrix verb’s event to
the infinitive’s entails that the latter is represented as somehow subsequent to the
former. This relation can take the form either of ‘subsequent potentiality’, as in She
tried to open the door, where the infinitive’s event is understood to be future and
opening the door. In certain cases the subsequence holds exclusively between ideas in
the mind, as in The wall appears to be crooked, where the meaning of to applies to the
relation between appearances and what the latter indicate as probably being the case,
with the appearance of the wall being construed as leading to the attribution of
23
crookedness to it. (Note the existence of motion/direction metaphors with other
expressions in this notional domain, e.g., All the evidence points to a cover-up.)19
with Huddleston and Pullum’s (2002: 1240) conclusion that infinitival and gerundive
selects must be specified lexically for that verb: we cannot assign distinct meanings to
the form-types and treat the selection as semantically determined.”. While it has been
shown above that this statement is not accurate with respect to the infinitive, whose
reference point (Kruisinga 1931: 259; Freed 1979: 72–73; Hewson and Bubenik 1997: Commenté [B9]: AQ: this ref appears to be missing from
the biblio. Please check.
Commenté [PD10R9]: Here is the reference: Kruisinga,
Etsko. 1931. A Handbook of Present-Day English, Part II.
19 English Accidence and Syntax, vol. 1, 5th ed. Groningen:
This purely notional form of subsequence is also discernible in uses such as:
P. Noordhoff.
In (i), knowledge allows Max to be connected to the possession of a criminal record; in (ii)
going to church is conceived as a product of habit; in (iii) showing up at the same bar that
20
This term designates a verb that takes as a complement another verb, the latter being
24
5–6; Hirtle 2007a: 61–62; Egan 2008: 129). When put into relation with another
This is explicitly stated by Wierzbicka (1988: 62), who speaks of “sameness of time”
Another cognitive grammarian, Hamawand (2002: 100), defines the -ing form of the
verb as denoting “ungrounded ongoingness” and places uses such as (28) under this
“simultaneity”, not realizing that there is a flat contradiction between simultaneity and
anteriority. The situation is further complexified when one considers uses such as
(17):
denotes its event as an interiorly homogenous entity construed as the direct object of
25
the main verb (Duffley 2006: 36–37). Thus an action which is enjoyed will be
understood to produce pleasure during its realization; an action which is regretted will
be understood to have taken place before the feeling of regret of which it is the cause;
moment of the recommending. In other words, any temporal relation between the
gerund-participle’s and the matrix verb’s event is an inference based on the meaning
of the matrix verb and the fact that the gerund-participle represents ‘that which is
verb-ed’. Some matrix verbs even imply no particular temporal relation with their
complement:
(18) Is there any specific word to describe walking the same path from one place to
another? (english.stackexchange.com)
Mis en forme : EXT
Since any verb that makes sense with a direct object denoting an event can be
construed with an -ing complement, it is impossible to tie the selection of the gerund-
participle to any specific type of semantic content in the matrix. Huddleston and
Pullum (2002) are thus vindicated for the gerund-participle, but not for the to-
infinitive.
Regarding the second issue of verbs such as like which admit more than one type
distinction on the level of meaning which can be observed in minimal pairs such as:
26
Mis en forme : EXT
These show that like can be construed either as a positive disposition leading the liker
to try to achieve a goal (19a), or as a positive feeling derived from some action that
the liker is performing on a regular basis (19b). Similarly, try can be construed either
machine to stop. And remembering can be taken either as a trigger leading to action
(21a) or as the recalling of some action performed in the past (21b). It is interesting to
note that the more basic notion of remembering, which refers to the simple fact of
for action, whereas the notion of recalling, which involves calling something back
This illustrates the fact that pinpointing the precise reason why some verbs admit
more than one type of complementation, while others do not, requires careful research
As for the final question of whether the term ‘complement’ itself corresponds to
any semantic category, the answer is unfortunately negative. The above discussion
shows that the relation of to-infinitive complements to the main verb is very different
from that of gerund-participle complements. The latter have the function of direct
object denoting ‘that which is verb-ed’, and any temporal relation they are felt to have
with the matrix is an inference based on the interaction between this function and the
lexical content of the main verb. The to-infinitive, on the other hand, is related to the
27
matrix by the meaning of the preposition to, which places it in some form of
subsequence with respect to the main verb’s event. The term ‘complement’ covers
both of these forms, whose semantic relationship to the matrix verb is significantly
different.
17.6. Conclusion
This chapter has covered the topic of clause structure from a range of different angles.
The definition of ‘head’ has been examined and the various approaches to clause
constituency grammar, the basic structure of the finite clause is binary, with the cut
being made between NP and I′. Dependency grammar comes in two varieties, the
traditional one in which there is a binary division between subject and predicate, and
the valency model that represents the clause as a constellation of elements all
dependent on the verb. The question of double dependencies has also been raised in
control verbs was proposed. The difference between complements and adjuncts has
been discussed and important points regarding licencing and optionality and their
relationship to semantics and pragmatics have been made. Givon’s influential binding
and manipulation verbs. The final section has dealt with subcategorization, showing
that the meaning of the main verb plays a crucial role in determining the type of
complementation it is compatible with, and that verbs which allow more than one
28