Nonfinite and Verbless Clauses PDF

14
Non-finite and verbless

clauses
Rodney Huddleston
1 Preliminaries 1173
1.1 Matters of form and function 1173
1.2 The catenative construction 1176
1.3 To-infinitivals with and without a subject 1178
1.4 The structure of infinitivals 1181
1.4.1 The clause subordinator for 1181
1.4.2 The infinitival subordinator to 1183
1.4.3 Bare infinitivals 1187
1.5 The structure of gerund-participials 1187
1.6 Case of the subject NP in gerund-participials 1191
1.7 Understood subjects 1193
2 The simple catenative construction 1194
2.1 To-infinitivals (I hoped to convince them vs I seemed to convince them) 1194
2.2 Gerund-participials (We enjoyed sailing vs We kept sailing) 1198
2.3 Concealed passives (The house needs painting) 1199
3 The complex catenative construction 1200
3.1 To-infinitivals 1201
3.1.1 The plain-complex construction (I persuaded Liz to go vs I intended Liz to go) 1201
3.1.2 The for-complex construction (I arranged for her to go by bus) 1203
3.1.3 The oblique-complex construction (I signalled to her to move off the road) 1204
3.2 Gerund-participials 1204
3.2.1 Catch vs resent (I caught him doing it vs I resented him doing it) 1204
3.2.2 See (I saw him doing it) 1205
3.2.3 Concealed passives with intervening NP (He needs his hair cutting) 1206
4 The catenative complement as a distinct type of complement 1206
4.1 Non-finites in relation to NP and AdjP complements 1206
4.2 The analysis of auxiliary verbs 1209
4.2.1 The dependent-auxiliary analysis 1210
4.2.2 The catenative-auxiliary analysis 1214
4.3 The traditional distinction between ‘gerunds’ and ‘present participles’ 1220
4.4 Catenative complements, adjuncts, and coordinates 1222
5 Classification of catenative verbs 1225
5.1 Framework of classification 1225
5.2 Class 1: catenative verbs appearing only in the simple construction 1227
5.3 Class 2: catenative verbs appearing in both simple and complex constructions 1229
5.4 Class 3: catenative verbs appearing only in the complex construction 1233
5.5 Index to the classification 1239
1171
Downloaded from https:/www.cambridge.org/core. Boston University Theology Library, on 04 Jun 2017 at 01:31:41, subject to the Cambridge Core terms of use,
available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/9781316423530.015
1172
5.6 Further remarks on the form-types 1240

5.6.1 To-infinitival vs gerund-participial 1240
5.6.2 The minor form-types: bare infinitivals and past-participials 1244
6 Hollow non-finite clauses 1245
6.1 General properties 1245
6.2 Gerund-participials 1246
6.3 Hollow to-infinitivals 1246
7 Non-catenative complements in clause structure 1251
7.1 Subject and extraposed subject 1252
7.2 Object and extraposed object 1255
7.3 Predicative complement 1255
8 Further complement uses of non-finite clauses 1256
8.1 Non-finite complements of adjectives 1256
8.2 Non-finite complements of nouns 1259
8.3 Non-finite complements in the structure of PPs 1262
8.4 Indirect and matrix-licensed non-finite complements 1262
8.5 Interrogative infinitival clauses 1264
9 Non-finite clauses as modifiers and supplements 1264
10 Verbless clauses 1266
11 Further remarks on the interpretation of subjectless non-finites 1268
1173
1 Preliminaries
1.1 Matters of form and function

Non-finite vs finite
Non-finite clauses are distinguished from finites largely but not wholly by the inflectional
form of the verb. Clauses whose verb is a gerund-participle or past participle are always
non-finite, and those whose verb is a preterite or present tense form (or irrealis were)
are always finite: only the plain form appears in both classes of clause (Ch. 3, §1.8). Of
the three main syntactic constructions having a plain-form verb, two are finite and one
non-finite:

[1] i imperative Be patient. [finite]
ii subjunctive It’s essential that he be more careful.
iii infinitival It’s important for him to be more careful. [non-finite]
Infinitivals are distinguished from the two finite constructions by the following
properties:
[2] i Most infinitivals, apart from the complements of modal auxiliaries and support-
ive do, contain the VP subordinator to: this is a clear marker of the infinitival.
ii Unlike imperatives, they do not take auxiliary do in negatives, etc.: compare
Don’t be late and It’s important not to be late.
iii Unlike imperatives, they are almost invariably embedded in a larger clause.
iv Unlike subjunctives, they usually have no subject, and where there is a subject
it appears in accusative (or plain) form, not nominative (compare him in [1iii]
with he in [1ii]).
v Whereas the most common type of subjunctive construction, the mandative,
takes the finite-clause subordinator that, the infinitival subordinator (used only
when a subject is present) is for.
Form-types
On the basis of the inflectional form of the verb we distinguish three main kinds of non-
finite clause: infinitival, gerund-participial, and past-participial. We refer to these
as form-types. They are illustrated by the bracketed clauses of [3] (all of which are
1174 Chapter 14 Non-finite and verbless clauses
complements of the preceding verb):

[3] i Max wanted [to change his name]. [infinitival]
ii I remember [locking the door]. [gerund-participial]
iii His father got [charged with manslaughter]. [past-participial]
Precisely because it is only the plain form that can occur in both finite and non-
finite clauses, it is only for [i] that we need a label that is not simply derived from
the name of the inflectional form of the verb. For [ii–iii] we use gerund-participle
and past participle for the verb-forms, gerund-participial and past-participial for the
clauses.
Subtypes of infinitival: to-infinitivals and bare infinitivals
Infinitivals are subdivided into to-infinitivals and bare infinitivals according to the pres-
ence or absence of the subordinator to :
[4] T O-infinitival bare infinitival
i a. They forced me to sign the petition. b. They helped me move the furniture.
ii a. You ought to sell it. b. You should sell it.
iii a. All I did was to ask a question. b. All I did was ask a question.
The to-infinitival occurs in a very wide range of constructions, whereas the bare infinitival
is very restricted. It occurs primarily as complement to a small number of verbs, but
these include the modal auxiliaries and supportive do, which makes it a very common
construction.
Form-types, auxiliaries, and voice

The modal auxiliaries and supportive do are excluded from all non-finite clauses. Infini-
tivals admit the remaining three auxiliaries:
[5] i I expect [to have finished soon]. [perfect have]
ii I expect [to be working all week-end]. [progressive be]
iii I expect [to be interviewed by the police]. [passive be]
Gerund-participials accept have and passive be, but not progressive be :1
[6] i I regret [having told them]. [perfect have]
ii I resent [being given so little notice]. [passive be]
iii ∗I remember [being working when they arrived]. [progressive be]
The past participle form of the verb has two uses, perfect and passive. Clauses with
a perfect past participle as head occur as complement to auxiliary have, and accept
progressive and passive be, while those with a passive past participle as head accept no
auxiliaries:
[7] i Ed has [seen her]. Ed has [been seeing her]. Ed has [been seen]. [perfect]
ii He had it [checked by the manager]. [passive]
1
Examples with progressive be are occasionally encountered in casual speech: I’ve missed endless buses through
[not being standing at the bus stop when they arrived]. This cannot, however, be regarded as an established
construction in Present-day English.
§ 1.1 Matters of form and function 1175
Subjectless non-finites
The great majority of non-finite clauses have no subject, as in:
[8] i Kim was glad [ to reach home].
ii It has been a pleasure [ meeting you].
iii Anyone [ living nearby]will be evacuated.
iv The sum [ spent on gambling]was extraordinary.
Whereas the subject is an obligatory element in canonical clauses, there are no non-
finite constructions in which a subject is required.2 There are, moreover, many con-
structions where it is impossible to add a subject, as in [iii–iv], or the examples of [4]
above.
It will be evident from this formulation that we take the subject to be an optional element in
non-finite clauses, not an element whose presence is necessary for an expression to qualify as
a clause. That is, the bracketed expressions in [8] are analysed as clauses that consist of just
a VP rather than simply as subclausal expressions with the form of a VP. We take the VP to
be the head of the clause and the presence of a VP is normally sufficient to establish clausal
status. The main exception is with attributive VPs in NP structure:

[9] i our rapidly approaching deadline
[VPs, not clauses]
ii a poorly drafted report
Expansion of the verbs in this construction is virtually limited to adverbial modifiers
preceding the verb: the range of structural possibilities here is quite different from that
found in clauses.
Hollow clauses
In a relatively small number of constructions, almost exclusively to-infinitivals, some
non-subject element is missing:
[10] i The letter isn’t legible enough [for you to read ].
ii The letter isn’t legible enough [to read ].
iii I don’t think they are worth [spending much time on ].
The bracketed clauses here we call hollow clauses. The missing non-subject is normally
recoverable from the matrix clause. In [i–ii] the object of read is recoverable from the
letter, while in [iii] the object of on is recoverable from they. Missing subjects, by contrast,
are not always recoverable from the matrix, as evident from [8ii] and [10ii].
Interrogative and relative to-infinitivals

Interrogative and relative clauses may have the form of to-infinitivals:
[11] i a. I can’t decide whether to go with them. [closed interrogative]
b. He doesn’t know how to placate her. [open interrogative]
ii a. They have funds with which to conduct a survey. [wh relative]
b. Another option for you to consider is renting a caravan. [non-wh relative]
These will receive little attention in the present chapter, as they are covered along with
other subordinate interrogatives (Ch. 11, §5.3.5) and relatives (Ch. 12, §5).
2
The subordinator for cannot occur without a subject, but we do not find constructions where for + subject is
required.
Distribution of non-finite clauses

Non-finite clauses occur as dependent or supplement in a wide range of constructions.3
The major distinction we draw is between non-finites in complement function, and
those in non-complement function (modifiers or supplements); these are illustrated for
infinitivals in [12i–ii] respectively:
[12] i a. His aim was to intimidate us. [comp in clause structure]
b. She is [keen to regain control]. [comp in AdjP structure]
c. I’ve missed [the opportunity to have my say]. [comp in NP structure]
d. She left at six [in order to catch the early train]. [comp in PP structure]
ii a. She left at six to catch the early train. [modifier in clause structure]
b. He’s a charlatan, to put it bluntly. [supplement to a clause]
c. I’ve found [a box to keep the tapes in]. [modifier in NP structure]
The complement functions are similar to those of content clauses, but the non-
complement uses are more varied and frequent than those of content clauses.
1.2 The catenative construction

The functions of complement non-finites in clause structure include the following:
[13] i To underestimate her would be foolish. [subject]
ii I found talking to her quite helpful. [object]
iii I call that taking liberties. [predicative comp]
iv It was natural to be worried. [extraposed subject]
v I found it distressing to see her so ill. [extraposed object]
vi a. She wants to leave the country.
b. She seems to like them. [catenative comp]
c. She hopes to hear from them soon.
In the case of [i–v] the non-finite clause has the same function as expressions of other
kinds, e.g. NPs in [i–iii], finite clauses in [iv–v]. Compare:
[14] i Such behaviour would be foolish. [subject]
ii I found the discussion quite helpful. [object]
iii I call that an outrage. [predicative comp]
iv It was natural that they should be worried. [extraposed subject]
v I found it distressing that she was so ill. [extraposed object]
In the case of [13vi], on the other hand, we take the view that the non-finite clause
cannot be satisfactorily subsumed under a complement type filled by other classes of
expression. The governing verbs want, seem, and hope appear elsewhere with such com-
plements as we see in:
[15] i She wants a holiday. [object]
ii She seems fond of them. [predicative comp]
iii She hopes for an early reply. [comp of prepositional verb]
3
They also occur in some minor main clause constructions, such as polar echoes and certain kinds of directive:
see Ch. 10, §§4.8.3, 9.6.3.
§ 1.2 The catenative construction 1177
We will argue in §4.1, however, that the differences between these do not carry over
to [13via– c], so that the latter cannot satisfactorily be regarded as containing respec-
tively an object, a predicative complement, and the complement of a prepositional verb.
Instead, we analyse the underlined clauses in [13vi] as examples of a distinct type of
complement realised exclusively by non-finite clauses; we refer to them as catenative
complements.
The term ‘catenative’ applies to a large class of constructions where a verb has a
non-finite internal complement. The name reflects the fact that the construction can be
repeated recursively, yielding a concatenation (‘chain’) of verbs:
[16] i I wanted to arrange for Kim to do it.
ii She intends to try to persuade him to help her redecorate her flat.
In [i] we have a chain of three verbs, with for Kim to do it complement of arrange and
to arrange for Kim to do it complement of want. We apply the term ‘catenative’ both
to the non-finite complement and to the verb in the matrix clause that licenses it, so
that want and arrange here belong to the class of catenative verbs. The last verb in the
chain, do, is not a catenative verb as it does not have a non-finite complement. Simi-
larly, the underlined verbs in [ii] are catenative verbs with the non-finite complements
shown in:
[17] catenative verb catenative complement
i intend to try to persuade him to help her redecorate her flat
ii try to persuade him to help her redecorate her flat
iii persuade to help her redecorate her flat
iv help redecorate her flat
Note that persuade and help take an NP complement (an object) in addition to the
catenative complement, namely him and her respectively.
Auxiliary verbs
Auxiliary verbs in their core use as markers of mood, tense, aspect, and voice also belong
in the class of catenative verbs. On this account, She may like it, for example, will be
analysed with like it a non-finite complement of may just as to like it is a complement of
seems in She seems to like it ; we take up this point in §4.2.
Simple vs complex catenative constructions

Catenative constructions may be classified on one dimension according to form-type,
and on another as simple or complex, according as they lack or contain an intervening
NP, i.e. an NP positioned between matrix and subordinate verbs and functioning as
complement of one or other of them.
[18] simple complex
i a. I hope to finish soon. b. I advise you to sell it. [to-infinitival]
ii a. I helped wash up. b. I made them apologise. [bare infinitival]
iii a. I stopped worrying about it. b. I saw them fighting. [gerund-participial]
iv a. I got arrested. b. I had my car stolen. [past-participial]
More precisely, the simple construction has no intervening NP in the active: passives like
You are advised to sell it are subsumed under the complex construction.
Four subtypes of the (active) complex construction can then be distinguished:

[19] i I arranged for them to go by bus. [for-complex]
ii I rely on them to look after themselves. [oblique-complex]
iii I resented their being given such favourable treatment. [genitive-complex]

iv I want them to be happy.
[plain-complex]
v I resented them being given such favourable treatment.
In the for-complex construction the intervening NP is preceded by the subordinator for ;
it is subject of the non-finite but in accusative (or plain) case. The oblique-complex is
found with certain prepositional verbs: the intervening NP is an oblique complement
of the matrix verb, i.e. complement of the preposition that the verb selects. In the
genitive-complex the intervening NP is in the genitive case and functions as subject of
the non-finite clause. Finally, in the plain-complex the intervening NP is in plain or
accusative case, and (depending on factors discussed in §3) either object of the matrix,
as in [iv], or subject of the subordinate, as in [v].4
1.3 To-infinitivals with and without a subject

To-infinitivals with overt subject require the subordinator for
To-infinitivals containing a subject are always introduced by the subordinator for :
[20] i [For them to withdraw now]would be a mistake. [subject]
ii It’s not necessary [for them to wait any longer]. [extraposed subject]
iii The best plan would be [for them to go alone]. [predicative comp]
iv I can think of no solution except [for them to sack him]. [comp of preposition]
If we drop the for we must also drop the subject, and conversely, for cannot appear
without a following subject. Compare [20ii], for example, with:
[21] i It is not necessary [to wait any longer].
ii ∗ It is not necessary [them to wait any longer].
iii ∗ It is not necessary [for to wait any longer].5
Notice, moreover, that the interrogative to-infinitivals shown in [11i] and the wh
relative [11iia] cannot contain a subject:
[22] i a. I can’t decide [whether to go with them]. (=[11ia])
b. ∗I can’t decide [whether (for) us to go with them].
ii a. He doesn’t know [how to placate her]. (=[11ib])
b. ∗He doesn’t know [how (for)us to placate her].
iii a. They have funds [with which to conduct a survey]. (=[11iia])
b. ∗They have funds [with which (for) us to conduct a survey].
The well-formed examples here all exclude for because the prenuclear position is
occupied – by the interrogative subordinator whether, the interrogative phrase how,
or the relative phrase with which. And because an infinitival cannot contain a subject
4
The infinitival version of the plain-complex is often referred to as the ‘accusative + infinitive construction’.
5
There are, however, non-standard varieties where certain catenative verbs may take for without a subject: He !
wanted for to see you.
§ 1.3 To-infinitivals with and without a subject 1179
unless introduced by for the asterisked examples here are as bad in the version without
for as in the one where for illicitly appears.
The catenative construction not an exception

At first glance examples like the following might appear to cast doubt on the rule that
to-infinitivals with a subject require for :
[23] i They arranged for the performance to begin at six. [for required]
ii They expected the performance to begin at six. [for excluded]
iii They intended (for) the performance to begin at six. [for optional]
For is required after arrange, excluded after expect, and optional after intend. The struc-
tural difference between [i] and [ii], however, is not just a matter of the presence or
absence of the subordinator for : in [i] the performance is subject of the infinitival clause,
but in [ii] it is object of the matrix clause. This means that [ii] is not an exception: there
is no for here because the infinitival clause has no subject. And in [iii] the version with
for has the same structure as [i], while the version without for has the performance as
matrix object, like [ii].
Semantically, of course, the performance does belong with begin in [23ii], just as it
does in [i], for the meaning is “They expected that the performance would begin at six”.
Nevertheless, there are several pieces of evidence showing that syntactically it belongs
in the matrix clause, that the performance and to begin at six in [ii] do not combine to
form a single constituent (a clause) but are separate complements of expect : object and
catenative complement respectively.
(a) Passivisation
Examples [23i] and [ii] behave quite differently under passivisation:
[24] i It was arranged for the performance to begin at six.
ii ∗It was expected the performance to begin at six.
iii The performance was expected to begin at six.
In the construction with for, passivisation is accompanied by extraposition, so that for the
performance to begin at six becomes extraposed subject. But where for is absent the sequence
the performance to begin at six does not become extraposed subject: rather it is just the NP
the performance that is promoted to subject. This is strong evidence that the performance is
a separate complement of expect in [23ii] – its object. Note that with intend, where for is
optional, we have two passive versions: It was intended for the performance to begin at six
(matching [24i]), and The performance was intended to begin at six (matching [24iii]). Thus
for marks the beginning of the infinitival clause, with the following NP therefore subject of
the subordinate clause, but where for is absent the NP behaves syntactically as object of the
matrix clause.
It is true that not all catenatives taking an NP without for allow passives like [24iii]:
[25] i They wanted the performance to begin at six.
ii ∗The performance was wanted to begin at six.
However, there is no other relevant difference between want and expect, and given that
passivisation doesn’t provide a necessary condition for objects, we shall not wish to assign
different structures to the want and expect examples. Note, moreover, that want takes for
when there is a preceding adjunct, but does not allow passivisation in the for construction
either:
[26] i They had wanted all along for the performance to begin at six.
ii ∗It had been wanted all along for the performance to begin at six.
This suggests that the deviance of [25ii] is due to a property of the verb want, not to the
structure of the active clause [25i].
The important point, then, is that while there are verbs which exclude passives like [24iii],
there are none that accept those like [24ii]: the latter is grammatically impossible because
the sequence NP + to-infinitival does not form a constituent and thus cannot function as
extraposed subject any more than it can function as an ordinary subject.
(b) Position of adjuncts
In general, adjuncts cannot occur between a verb and an NP object, but they are permitted
between a verb and a clausal complement:
[27] i ∗We expected all along an improvement.
ii We expected all along that things would improve.
In the catenative construction such an adjunct can follow the matrix verb in the construction
with for, but not in the one without:
[28] i a. He arranged at once for the performance to be postponed.
b. ∗He expected all along the performance to be postponed.
ii a. I’d prefer if at all possible for you to do it tomorrow.
b. ∗I’d prefer if at all possible you to do it tomorrow.
As in the passive case, therefore, the performance behaves as an NP object of expect, rather
than as part of a larger clausal complement. The contrast is particularly striking with verbs
like prefer which occur with or without for : if we drop if at all possible from [ii] both versions
are well formed.
(c) Pseudo-cleft
While a for-infinitival can occur as complement in pseudo-cleft and related constructions
the sequence NP + infinitival VP cannot:
[29] i a. ∗What they expected was the performance to begin at six.
b. What they arranged was for the performance to begin at six.
ii a. ∗ All I want is you to be happy.
b. All I want is for you to be happy.
Note again the particularly clear contrast with verbs like want in [ii] that occur in catenative
constructions with and without for.6
(d) With some verbs the intervening NP is obligatory

There is a class of verbs including believe, assume, etc., where the intervening NP cannot be
omitted even when it is coreferential with the subject of the matrix clause:
[30] i Max believed Kim/himself to be in the right.
ii ∗Max believed to be in the right.
6
In the catenative construction want takes for only when the catenative complement is preceded by an adjunct,
as in [26i], but this is sufficient to sanction the for-infinitival in [29iib].
§ 1.4 The structure of infinitivals 1181
This indicates that there is a direct syntactic relation between the matrix verb and the NP: the
NP is an obligatory complement of the verb. Note by contrast that in the for construction,
where the NP is not a complement of the matrix verb, it can normally be readily omitted
(along with for):7
[31] i Max arranged for Kim to see a solicitor.
ii Max arranged to see a solicitor.
(e) The intervening NP may belong semantically in the matrix clause
With some verbs appearing in the plain-complex structure the NP is semantically related to
the matrix verb:
[32] They persuaded the students to cancel the performance.
Here there is no reason at all to suggest that the students to cancel the performance is a single
complement, for the meaning is that persuasion was applied to the students in order that they
should agree to cancel the performance. The only plausible structure is one with the NP and
the infinitival as separate complements, and this structure is then available to accommodate
examples like [23ii] above. We will examine the semantic difference in detail in §3.1.1, but what
is relevant at this point is that [32] is like [23ii] with respect to all the properties discussed in
(a)–(c) above:

[33] i a. The students were persuaded to cancel the performance.
[passivisation]
b. ∗It was persuaded the students to cancel the performance.
ii ∗They persuaded easily the students to cancel the performance. [adjunct]
iii ∗What they persuaded was the students to cancel the performance. [pseudo-cleft]
In passivisation the intervening NP is promoted to subject, adjuncts cannot come between
the verb and the intervening NP, and the intervening NP + infinitival VP cannot form the
complement of a pseudo-cleft.
The conclusion must be that there is no construction where the sequence NP + to-
infinitival, with no preceding for, behaves as a subordinate clause, a single constituent.
1.4 The structure of infinitivals

Two special features of the to-infinitival construction are, firstly, the for that intro-
duces the clause if it contains a subject and, secondly, the to itself that marks the
VP. We consider them here in turn, and then take up briefly the bare infinitival
construction.
1.4.1 The clause subordinator for

Historically, the for which introduces to-infinitivals containing a subject derives from the
preposition for. This prepositional source is reflected in a number of current properties,
but there are nevertheless good reasons for saying that it has come to be reanalysed as a
subordinator.
7
An exception is call: He called for her to be released, but not ∗He called to be released. Note, however, that a
reflexive is also normally excluded: ∗He called for himself to be released.
Syntactic reflection of the prepositional source of for

(a) Case
Accusative rather than nominative forms of personal pronouns are used for the subject:
[34] He arranged for her/∗she to be interviewed first.
(b) Subject must immediately follow for
No clause element can come between for and the subject NP:
[35] i It’s important for you to read the first one immediately.
ii ∗It’s important for the first one you to read immediately.
Note the contrast here between for and the finite subordinator that : It’s important that
the first one you read immediately.
(c) Distribution
There is a very large overlap between the positions where we find for clauses and those
where for PPs appear:
[36] i a. the need for peace b. the need for us to cooperate
ii a. too cold for a swim b. too cold for us to go out
iii a. ready for departure b. ready for us to start
iv a. We arranged for a postponement. b. We arranged for it to be postponed.
Similarly, a for clause is like a for PP in that it can’t generally occur as complement of a
preposition:
[37] a. ∗I’m thinking of for a holiday. b. ∗I’m thinking of for us to leave.
Reanalysis of for as subordinator

The item for that appears with to-infinitivals differs in important ways from the one
that takes NP complements. Although traditional grammars and dictionaries classify it
as a preposition, there are strong grounds for analysing it as a subordinator, with the
following NP functioning as subject of the infinitival VP. The subordinator for does for
infinitival clauses with subjects what the subordinator that does for finite content clauses.
(a) Occurrence of to-infinitivals in non-PP positions

In spite of the similarities illustrated in [36], for clauses commonly appear in places that do
not allow for PPs:
[38] i For you to give up now would be tragic. [subject]
ii It’s rare for the bus to be so late. [extraposed subject]
iii This made it necessary for the meeting to be postponed. [extraposed object]
iv I can’t afford for them to see me like this. [complement of afford]
Most importantly, they occur as subject or as extraposed subject/object, as in [i–iii]: PPs do not
appear in extraposed position and appear as subject only under highly restrictive conditions.
In addition, [iv] shows that the class of verbs taking for clause complements is not limited to
those allowing for PPs: while [iv] is grammatical, ∗I can’t afford for an investigation is not (cf.
also agree and say).
(b) Range of subject NPs
Except for the matter of accusative case, the NPs following for are the same as those which
occur as subject of finite main clauses. Note in particular that dummy there occurs freely
§ 1.4.2 The infinitival subordinator to 1183
here though it can’t occur as complement in a for PP, and the same applies to numerous NPs
that form parts of idioms:
[39] i It’s essential for there to be no misunderstanding on this point.
ii He called for close tabs to be kept on the new recruits.
(c) Constituent structure
It is clear that for + NP + VP forms a constituent – the subject of [38i], for example, is
uncontroversially for you to give up now, and similarly for the other examples cited.8 Within
this sequence there is no reason to say that the NP combines directly with for ; rather the NP
and VP combine to form a clause nucleus showing the same range of contrasts as a main
clause – or as a clause nucleus following the subordinator that :
[40] i It’s necessary [for both your parents to sign the form].
ii It’s necessary [for the form to be signed by both your parents].
iii It’s necessary [for your parents both to sign the form].
The underlined clauses in [i] and [ii] contrast as active vs passive, while that in [iii] contrasts
with [i] with respect to the position of both: the NP following for behaves like an ordinary
clause subject.
(d) Absolute initial position and contrast with that
For must occupy initial position in the subordinate clause. We have noted that it can’t occur
in the interrogative and wh relative constructions [22]: this is because the initial position is
there pre-empted by the interrogative subordinator whether and the interrogative and relative
phrases. Its syntactic role is therefore closely parallel to that of the finite clause subordinator
that, which is likewise excluded from interrogative and wh relative constructions. Like that,
for has no identifiable meaning of its own, but serves as a syntactic marker of a particular
syntactic construction. The functional similarity to that is particularly clear in such pairs as
the following, where we find a direct contrast between finite and infinitival clauses:
[41] i a. It is important that detailed records be kept.
b. It is important for detailed records to be kept.
ii a.That’s the best course that you can take.
b.That’s the best course for you to take.
iii a. In order that the bill may be passed major amendments were made.
b. In order for the bill to be passed major amendments were made.
1.4.2 The infinitival subordinator to

Constituent structure
It is important that to enters into construction with a VP, not just a verb. This is shown
by such data as:
[42] i She wants me to lend him the money, so lend him the money I have to.
ii She wants me to lend him the money, but I don’t have to .
iii I have to lend him the money and find a solicitor for him.
8
Examples like It would be good for us to have a period on our own are ambiguous according as for us is a PP
dependent of good (cf. To have a period on our own would be good for us) or part of the to-infinitival (cf. For us
to have a period on our own would be good ).
Lend him the money acts as a constituent in [i]. It is separated from to and occupies
prenuclear position. In [ii] lend him the money as a whole is ellipted. And in [iii] it is
coordinated with another constituent of the same category, find a solicitor for him. 9
Traditional grammar treats to lend as a form of the lexeme lend, as if to were an
inflectional prefix, comparable to the inflectional suffix that marks the infinitive in such
languages as Latin and French. This is quite inappropriate for English. The evidence
from [42] shows that to is not syntactically in construction with the verb base, let alone
morphologically bound to it.
Syntactic reflection of the prepositional source of to

Like for, to derives historically from the homophonous preposition. This source is again
reflected in the present properties of the construction – though to a much lesser extent
than in the case of for.
(a) Distributional restrictions
To-infinitivals, like PPs consisting of to + NP, do not in general occur as complement to
a preposition:
[43] a. ∗ We’re thinking of to London. b. ∗We’re thinking of to travel by bus.
This is an extension of the point made above in connection with for : the restriction
applies to to-infinitivals whether or not they contain for.
(b) Contrast with prepositions from and against
The prepositional history of to is reflected in the contrast we find between to-infinitival
and from + gerund-participial complements with certain pairs of verbs of opposite
meaning such as encourage vs discourage, persuade vs dissuade. Compare:
[44] i a. I persuaded her to buy it. b. I dissuaded her from buying it.
ii a. I assented to her proposal. b. I dissented from her proposal.
The non-finite complement is comparable to a goal in [ia] (metaphorically, persuasion
moves someone towards an action, and assent to a proposal is likewise seen as movement
towards it) and source in [ib] (dissuasion and dissent involve pulling back or moving
away).
We find a somewhat similar contrast between a to-infinitival and against + gerund-
participial with verbs like warn, though this time to is not used with NP complements:
[45] a. I warned her to stay indoors. b. I warned her against staying indoors.
Here [b] is equivalent to I warned her not to stay indoors, as [44ib] is equivalent to I
persuaded her not to buy it.
However, points (a) and (b) are not nearly sufficient to justify treating infinitival to
as a preposition in Present-day English. It cannot coordinate with any preposition
(∗I don’t want you warning her to or against); its complement cannot coordinate with
9
The validity of coordination data as evidence for constituent structure is discussed in some detail in Ch.
15, §4.6. There are some types of ‘non-basic’ coordination where the coordinates are not constituents in
corresponding non-coordinate constructions: in the American proposal, for example, the American is not a
constituent but it occurs as a coordinate in the American and the Russian proposals. The fact then that we
can say I have to chlorinate and to vacuum the pool does not demonstrate that to chlorinate is a constituent in
I have to chlorinate the pool: it too involves non-basic coordination, in contrast to the basic, and much more
likely, I have to vacuum and chlorinate the pool.
§ 1.4.2 The infinitival subordinator to 1185
the complement of prepositional to (although I agreed [to it] and I agreed [to go] are
both grammatical, ∗I agreed [to it and go] is not); and the phrases it introduces cannot
be systematically substituted for PPs or vice versa. It is quite clear that the distribution
of to-infinitivals has to be described independently of that of PPs.
Reanalysis of to as a VP subordinator
To introduces phrases that function as predicate in clause structure, and all the evidence
is compatible with these phrases being of the same category as phrases functioning as
predicate in canonical clauses. That is, to lend him the money can be assumed to be a VP.
And as already noted, to combines with a VP to make this larger VP. We can therefore
assume the partial structure in [46].
[46] VP
X VP
to lend him the money
To fill out the details we must decide which constituent (to or lend him the money) is the
head of the upper VP. If the constituent labelled X is the head, then it follows from general
principles that it belongs to the verb category, and we would have the structure in [47a] below.
The alternative is that lend him the money is the head. In that case to will have a function in
the VP comparable to that of that, whether, and for in the structure of the clause, namely as
a marker of the subordination, and we will have the structure in [47b].
[47] a. VP b. VP
Head: Comp: Marker: Head:

V VP Subordinator VP
to lend him the money to lend him the money
It is difficult to find compelling evidence to choose between these alternative analyses. It

would not be impossible to maintain [a], if to were assumed to be an auxiliary verb, albeit
a rather problematic defective one with no forms other than the plain form. Some linguists
have defended that view. But the case for to being a VP subordinator is stronger. To has no
meaning independently of the semantic properties of the infinitival complement construction
as a whole. It functions with respect to the VP lend him the money in much the same way
as whether functions in whether she ever lent him the money, or that in that she lent him the
money. It would seem both syntactically and semantically appropriate to place to in the same
category as whether and that, the category of subordinators.
It must be acknowledged that it differs from the clause subordinators in two significant
respects, but those differences are not sufficient to motivate choosing structure [a].
The first difference is that in elliptical constructions, to can stand alone without the lower
VP – i.e. it can be ‘stranded’, as in [42ii] (I don’t have to ). Indeed, in this respect it is like the
auxiliary verbs, for these can be stranded too (cf. but I won’t ). The similarity is strengthened
when we note the strong tendency not to stress secondary forms of auxiliaries stranded by
ellipsis: He MAY have is preferred over ?He may HAVE ; They COULD be is preferred over
?
They could BE ; and parallelling these, You HAVE to is strongly preferred over ?You have
TO . In this respect, to and the secondary forms of auxiliaries appear to function alike.
In response to this argument, however, note that there are special conditions on the
stranding of to that do not apply to clear cases of auxiliaries (see Ch. 17, §7.3, on contrasts
like Not to would be a mistake vs ∗To would be a mistake). Some special mention of to is
necessary either way: either it is the only subordinator that can be stranded under ellipsis of
the constituent it introduces, or it is the only auxiliary verb subject to these special conditions
on stranding. And the stress facts are expressible in a different way: we can say that it is
strongly preferred for a stranded item that is stressed to bear tense.
The second acknowledged difference between to and clause subordinators is that to does
not always occupy absolute initial position in the constituent it marks:
[48] i She taught her children always to tell the truth.
ii I’ll try not to underestimate the opposition next time.
Always and not here belong in the subordinate clause, not the matrix, but they precede the
subordinator to. With clause subordinators this is not the case (He thought always that there
would be some way to work it out has always unambiguously interpreted in the matrix clause,
and He thought that always there would be some way to work it out has it unambiguously in the
subordinate clause – but see [13] in Ch. 11, §3). It is of course unlike the other subordinators
anyway, in that it is a marker of VPs rather than clauses, so what we have to say is that the VP
subordinator allows for various adjuncts in its VP to precede it.
Again, this does make to an unusual subordinator. But there are two further arguments
that militate against its being treated as an auxiliary verb, one that suggests it is unlike heads
in general, and one that weighs specifically against its being a verb. The argument that it is
not a head is that under certain conditions it is omissible without any change in meaning or
grammatical construction type. One such case is [49].
[49] a. All I did was to ask a question. b. All I did was ask a question.
In this respect to is like the finite subordinator that (though the latter is of course very much
more freely omissible). It is not at all like heads, which seem never to be freely and optionally
omissible in this way without a change of construction. Note also that when causative verbs like
make and let are passivised, infinitival to switches from being disallowed in the complement
to being required:
[50] i They made the general public pay for it / ∗to pay for it.
ii The general public was made ∗pay for it / to pay for it.
For a verb to select a different subordinator for its complement depending on whether it is
active or passive seems less strange than for the head verb of the complement to be required
to change.
Finally, the general argument against to being a verb is that there are no counterexam-
ples in English to this very broad generalisation: all verbs can occur as head of a main
clause. Even the highly anomalous verb beware, which has no inflected forms at all, oc-
curs in main clauses like Beware the jabberwock. If admitted as a meaningless and defec-
tive auxiliary verb, the item to would be the unique exception to a principle that holds
for all of English and, as far as we know, for all languages, because it can only appear
in non-elliptical sentences when some other verb is superordinate to it. Being limited to
§ 1.4.3 Bare infinitivals 1187
subordinate clauses is precisely the property that we do expect in subordinators, and is a

property that we expect never to find in verbs.
We conclude that to is a VP subordinator. Our analysis for a to-infinitival clause con-

taining a subject can now be displayed in more detail:
[51] Clause
Marker: Head:
Subordinator Clause
Subject: Predicate:
NP VP
Marker: Head:
Subordinator VP
for you to lend him the money
1.4.3 Bare infinitivals

These lack both the VP subordinator to and the clause subordinator for. They are almost
always subjectless, in which case they consist simply of a VP headed by a verb in the plain
form. There are two constructions, however, where a subject is present:
[52] i Kim tell a lie! Surely not.
ii Rather than Kim give the introductory lecture, why don’t you do it yourself?
Example [i] belongs to the main clause polar echo construction discussed in Ch. 10,
§4.8.3, while subordinate construction [ii] is found only as complement to rather than.
The first is characteristic of informal style, and a personal pronoun would therefore
generally appear in accusative case (Me tell a lie!), though a nominative cannot be entirely
excluded. In [ii] we illustrate a rather rare construction where a nominative pronoun
would be very unlikely – the choice of case for the subject is easily avoided by using
a subjectless construction as in Rather than have Kim give the introductory lecture, why
don’t you do it yourself?
1.5 The structure of gerund-participials

The verb in gerund-participials ends with the suffix ·ing, a suffix which also appears on
nouns and adjectives, so that it is necessary to distinguish (by the criteria given in Ch.
3, §1.4) between a gerundial noun, a gerund-participle form of a verb, and a present-
participial adjective:
[53] i She had witnessed the breaking of the seal. [gerundial noun]

ii a. There’s no point in breaking the seal.
[gerund-participle form of verb]
b. They were entertaining the troops.
iii an entertaining show [present-participial adjective]
Traditional grammar, we have noted, distinguishes between breaking in [iia], a gerund,

and entertaining in [iib], a present participle – but since no verb lexeme has distinct
forms in these constructions there is no basis for saying that they involve different
inflectional forms: our gerund-participle thus covers both gerund and present participle
of traditional grammar insofar as these terms are applied to verbs rather than nouns or
adjectives.
Complement and non-complement uses of gerund-participials

We also take the view that even from a purely syntactic point of view the traditional
distinction between gerundial and participial uses of the verb-form ending in ·ing is not
well motivated: see §4.3. Instead, we distinguish primarily between gerund-participials
with complement function and those in non-complement function (modifiers or sup-
plements):

[54] i a. Telling her father was a big mistake.
[complement]
b. He stopped seeing her.

ii a. Being a foreigner himself, he understood their resentment.
[non-complement]
b. Anyone knowing his whereabouts should contact the police.
In terms of the traditional analysis, the non-complement uses all involve participles,
while the complement uses contain primarily gerunds but also some participles.
Nominal source of the traditional gerund

Historically, the ·ing suffix derives from two distinct sources, corresponding respectively
to traditional grammar’s present participle and gerund. The gerund suffix formed nouns
from verbs – as it still does in what we are calling gerundial nouns, such as the breaking
of the seal ([53i]). In the course of time, however, the syntactic use of this form was greatly
extended, so that it came to combine not just with dependents of the kind associated with
nouns, as in that example, but also with those associated with verbs, as in breaking the seal
in [53iia]. It was this extension that led to the split between nominal and verbal ‘gerunds’,
though the traditional definition of the gerund as a ‘verbal noun’ fails to recognise the
reanalysis of the form as a verb in the latter type.
The nominal source of the ·ing verb that is found in most gerund-participial com-
plements is reflected in certain properties that still hold in the present-day language.
(a) Distributional similarity to NP
The distribution of gerund-participial complements is much closer to that of an NP than
is that of any of the other non-finite form-types, or indeed of finite subordinate declar-
atives. In particular, they freely occur as complement to a preposition and can follow
the verb in subject–auxiliary inversion constructions. Compare the gerund-participials
with the to-infinitivals in:
[55] i a. It’s a matter of breaking the seal. b. ∗It’s a matter of to break the seal.
ii a. Is breaking the seal wise? b. ∗Is to break the seal wise?
With regard to extraposition, gerund-participials fall somewhere between NPs and to-
infinitivals (cf. Ch. 16, §7):
[56] i a. ∗It was silly the breaking of the seal. b. ∗It amused him the breaking of the seal.
ii a. It was silly breaking the seal. b. #It amused him breaking the seal.
iii a. It was silly to break the seal. b. It amused him to break the seal.
§ 1.5 The structure of gerund-participials 1189
Extraposition is normal with to-infinitivals, as in [iii], but not generally possible with
NPs, as illustrated in [i]; with gerund-participials speaker judgements vary, but in general
it tends to be possible over a short VP like the intransitive was silly but not over longer
ones like transitive amused him.
(b) Hybrid constructions
Examples occasionally arise where the dependents are of mixed types:
[57] i a. This constant telling tales has got to stop.
b. Let’s have no more of this bringing food into the computer room.
ii a. There was no telling what he might do next.
b. There’ll be no stopping her.
The relevant heads are double-underlined. The pre-head dependents are characteristic
of NP structure, the post-head ones characteristic of VP structure (except that into the
computer room occurs readily in either).
The examples in [57i] are of somewhat marginal acceptability. Those in [ii] are fully
acceptable in the present-day language, but this use of no with a gerund-participial is
virtually restricted to the existential construction with there ; we don’t get, for example,
∗
No telling what he might do next was possible.
The examples in [57] illustrate the kind of hybrid construction that can arise when a
historical change has not been fully carried through to completion. Such examples resist
elegant description.
(c) Genitive case
The NP preceding the ·ing word can be in genitive case in the verbal construction just as
in the nominal one:
[58] i I resented [his constant questioning of my motives]. [noun]
ii I resented [his constantly questioning my motives]. [verb]
That questioning is a noun in [i] and a verb in [ii] is evident from the contrast between
constant (adjective) and constantly (adverb), and between of my motives (PP) and my
motives (NP – object), but in both we have genitive his. And the characteristic use of
genitive case is of course to mark the dependent of a noun, not of a verb.
Analysis of the genitive NP as subject of the gerund-participial

It might be argued that his constantly questioning my motives in [58ii] should be described
in terms of some kind of nominal–verbal hybrid construction such as we evidently need
for [57], but we believe it is better, for several reasons, to regard the genitive as having
been reanalysed as a clause subject.
(a) Marginal status of the determiner + VP construction
The examples like [57] certainly involve a determiner in construction with a VP as head,
but they are very peripheral to the present language-system by virtue of the questionable
acceptability of [i] and the semi-formulaic nature of [ii]: this construction should not be
allowed to determine our analysis of the much more central and productive construction
with a genitive.
(b) Relation with accusative/plain case
The genitive can be replaced in informal style by an accusative (or plain) case: [I resented]
him constantly questioning my motives (see §1.6 below). In this non-genitive construction
the historical process of changing from noun to verb has been taken a step further: there
is nothing noun-like about the structure here, and its analysis as a clause is unproblematic.
It is then simpler to treat the stylistic alternation between genitive and non-genitive as a
matter of the case of the subject NP than as a major difference between two quite separate
constructions: the preference for the non-genitive in informal style can be seen as regularising
the clausal construction.
(c) Optionality of the genitive NP
The genitive NP can normally be omitted:
[59] i a. I regretted [his leaving the firm]. b. [Your being a shareholder] is important.
ii a. I regretted [leaving the firm]. b. [Being a shareholder]is important.
The significant point here is that the presence or absence of the genitive NP is not like that of
a genitive determiner in NP structure. In I regretted his decision, for example, we cannot drop
his: ∗I regretted decision. This is ungrammatical because decision is a singular count noun and
requires the presence of some determiner. In the gerund-participial construction, however,
the presence or absence of the genitive is like the presence or absence of the subject in a
to-infinitival:
[60] i a. I arranged [for him to leave the firm]. b. [For you to be a shareholder] is essential.
ii a. I arranged [to leave the firm]. b. [To be a shareholder]is essential.
We interpret [59iia] as “I regret my leaving the firm” and [60iia] is interpreted as “I arranged
for me to leave the firm”; and similarly [59iib] and [60iib] may be glossed roughly as “One’s
being a shareholder is important”, “For one to be a shareholder is essential”. Moreover, we
have noted that the genitive can be replaced by a non-genitive: if I regretted [his leaving the
firm] and I regretted [him leaving the firm] are analysed as quite different constructions, with
only the second of them a clause, then which of the constructions would I regretted [leav-
ing the firm] belong to? This problem would be particularly difficult to resolve with those
gerund-participials where it is not possible to include an NP before the verb, as in He didn’t
bother [giving me a copy]. We avoid these problems by treating the optionality of the initial
NP as simply a matter of the optionality of subjects in non-finite clauses.
The structure of the gerund-participial in [58ii] will therefore be as follows:
[61] Clause
Subject: Predicate:
NP VP
his/him constantly questioning my motives
Subject must be in absolute initial position

Whether genitive or not, the subject always occupies initial position in a gerund-
participial clause. It is therefore not possible to prepose elements or to postpose the
subject:
[62] i I resented [them/their going without me].
ii ∗ I resented [without me them/their going].
iii I remember [a troop of boy scouts suddenly appearing over the hill].
iv ∗ I remember [suddenly appearing over the hill a troop of boy scouts].
§ 1.6 Case of the subject NP in gerund-participials 1191
Comparison between genitive marking and the subordinator for

Genitive marking in gerund-participials bears some resemblance to the marking of
to-infinitivals by the subordinator for.10 For marks the clause boundary, so that the fol-
lowing NP must belong in the subordinate clause; the genitive case relates the NP to the
following VP and hence as again belonging in the subordinate clause. In to-infinitivals the
subject must immediately follow for and in gerund-participials the subject must be the first
element: neither construction allows for elements to be fronted to pre-subject position or for
the subject to be postposed.
Nevertheless, the analogy is not a close one, for there are several important differences.
(a) Status and position
For is a separate word belonging to the category of subordinators; it occupies initial position
in what we are calling the marker function. The genitive, by contrast, is an inflectional case.
Compare, then, structure [61] with [51] above.
(b) Restrictions on NP
Genitive case is incompatible with certain types of NP: for example, we can’t say ∗I resented
this’s being made public – only non-genitive this is possible. No such restrictions apply to for.
(c) Omissibility
For cannot in general be omitted unless the following subject is omitted too, and in those
constructions where it is omissible – after such catenatives as intend, prefer, etc. – its presence
or absence correlates with a major difference in structure, the NP belonging in the subordinate
clause when for is present but in the matrix clause when it is absent (cf. §1.4). The genitive
inflection, by contrast, is systematically omissible (subject to considerations of style), and
there is no reason to suppose that its omission has any further effect on the structure. Most
importantly, genitive case is not necessary in gerund-participials for the sequence of NP +
VP to form a constituent. This is evident from such examples as [63], where the status of
the bracketed sequences as constituents is unproblematic in both genitive and non-genitive
versions:
[63] i I have no objection to [their/them taking notes]. [comp of preposition]
ii What he doesn’t like is [Kim/Kim’s taking all the credit]. [predicative comp]
iii [His son’s / His son being a friend of the judge]hadn’t helped at all. [subject]
1.6 Case of the subject NP in gerund-participials

Nominative or plain case in non-complement gerund-participials
Gerund-participials functioning as supplement to a clause may contain a subject; pro-
nouns with a nominative–accusative contrast usually appear in the nominative, with
accusative a somewhat marginal alternant in informal style, while other NPs take plain
case:
[64] i They appointed Max, [he/him being the only one who spoke Greek].
ii [His mother being ill,]Max had to withdraw from the expedition.
This construction is the only one where we find a nominative subject in a non-finite
clause; it belongs to fairly formal style, and hence the informal accusative form in [i] is
10
Modern works that adopt a parallel treatment usually represent the elements as ‘for–to’ and ‘poss–ing’.
unlikely: the alternation here is not like that between It’s I/me, where the accusative is
much more common.
Genitive vs non-genitive subjects in gerund-participial complements

When the clause is functioning as complement, we find both genitive and accusative/plain
case, and our main concern in this section is to investigate the factors favouring one or
other of them. A preliminary point to emphasise is that potential alternation between
these cases arises only when the NP is subject of a gerund-participial clause – compare:
[65] i [I enjoyed] his/∗him reading of the poem. [determiner in NP: genitive required]
ii [I caught] him/∗his reading my mail. [object of matrix: genitive excluded]
iii [I remember] his/him reading my mail. [subject: genitive or accusative allowed]
In [i] reading is a noun: the preceding NP is therefore determiner and has to be in the
genitive. Note then that while Kim didn’t like his singing is ambiguous according as singing
is a noun or a verb, Kim didn’t like him singing is unambiguous, with singing necessarily a
verb.11 In [65ii] him is the object of caught, with reading my mail a second complement:
him is not subject of the gerund-participial clause and therefore cannot be genitive. Only
in [iii], where the pronoun is subject of the subordinate clause, do we have alternation
between the cases. (For the structural distinction between [ii] and [iii], see §3.2.1.)
In constructions of this third kind, where the NP is subject of a gerund-participial
complement,12 the choice between genitive and accusative/plain depends on the follow-
ing factors:
(a) Style
Genitives are more likely to occur in formal than in informal style.13
(b) Type of NP
Some NPs cannot take genitive marking, however formal the style: dummy pronouns
(particularly there); fused-head NPs like this, that, all, some; pronoun-final partitive NPs
like both of them, some of us; and so on. Such NPs cannot occur as determiner in NP
structure but readily appear in non-genitive form as subject of a gerund-participial:
[66] i He resented [there/∗there’s having been so much publicity].
ii I won’t accept [this/∗this’s being made public].
There are others which, while they do not exclude genitive marking in general, dis-
favour it in the subject of a gerund-participial:
[67] i He objected to [the girls / ?the girls’ being given preferential treatment].
ii It involved [the Minister of Transport / ?the Minister of Transport’s losing face].
11
In the noun interpretation we are concerned with the manner of his singing (cf. Kim didn’t like his singing of
this difficult aria), whereas in the verbal interpretation it is a matter of the activity or fact of his singing (cf.
Kim didn’t like him singing obscene songs).
12
We should also add the condition that the subject immediately precede the verb. If it is followed by a supplement,
a non-genitive subject is required: He resented Kim, after only two years, being promoted manager. And the same
applies if the verb is omitted in the gapping construction: He objected to Kim having three tries and Pat only
two. Contrastive stress on the subject also strongly favours a non-genitive: I’ve no objection to KIM doing it.
13
Modern usage manuals generally do not condemn non-genitives altogether (as Fowler did in early work),
though they vary in their tolerance of them, the more conservative ones advocating a genitive except where it
sounds awkward, stilted, or pedantic – by virtue of the type of NP involved (our factor (b)).
§ 1.7 Understood subjects 1193
The regular plural girls in [i] is identical to its corresponding genitive in speech, but
in writing they are distinguished by the apostrophe, and in this construction the form
without the apostrophe is much the more likely. In [ii] the NP contains a post-head de-
pendent PP of Transport, and while such NPs can take genitive marking when determiner
to a noun (the Minister of Transport’s performance), such marking is very unlikely in a
gerund-participial.
More generally, the genitive in a gerund-participial is awkward with NPs of any
significant length or complexity, especially those with post-head dependents. It is most
likely with personal pronouns, and after that singular NPs that refer to people and have
no more than one or two words as pre-head dependents.
(c) Matrix construction
A gerund-participial in subject function is somewhat more likely to select a genitive
than one in other complement functions. There are also differences within the class of
catenative verbs: verbs like appreciate, countenance, mind, etc. (class 2ci of §5.3) take a
genitive more readily than like and hate (class 2bi), while with stop a genitive is not likely
to occur at all (we find She stopped them using it, not ∗She stopped their using it).
1.7 Understood subjects

Most non-finite clauses have no subject, but the interpretation requires that we find an
‘understood subject’: characteristically, the VP represents a semantic predicate and we
need to find the argument that it is predicated of.
We distinguish three main ways in which the interpretation of a subjectless clause is
completed:
[68] i Pat persuaded Kim [to travel by bus]. [controlled interpretation]
ii Pat intended Kim [to travel by bus]. [raised interpretation]
iii It was necessary [to travel by bus]. [non-syntactic interpretation]
In [i] and [ii] we understand it to be Kim who was to travel by bus, but there is a
difference between them. In [i] Kim is associated with two semantic roles, patient of
persuade (Pat applied pressure to Kim) and agent of travel, but in [ii] Kim has only the
latter role: intend expresses a relation between Pat and the situation where Kim travels
by bus. Nevertheless, Kim is syntactic object of the matrix clause in both.
These two cases involve controlled and raised interpretations of the missing subject.
In [68i] the interpretation of the missing subject is controlled by an antecedent in the
matrix clause – in this example by the matrix object, in Kim wanted to travel by bus
by the matrix subject. In [ii] the missing subject is retrievable from the raised comple-
ment in the matrix clause. But [iii] is sharply different in that there is nothing in the
sentence itself to identify the traveller: the understood subject here has to be identified
from the context. It could be the speaker, someone else, or a group including or exclud-
ing the speaker. We say here that the interpretation is determined non-syntactically,
i.e. the missing subject is not linked to an antecedent in some syntactically specified
position.
The distinction between constructions with controlled and raised interpretations is
discussed in detail in §§2–3, and we return at the end of the chapter to that between
controlled and non-syntactic interpretations.
2 The simple catenative construction
Our primary concern in this section will be with the semantic status of the subject: in
some cases it represents an argument of the catenative verb, whereas in others it does
not, having a semantic role only with respect to the non-finite, so that it has the status
of a raised complement. We will examine this distinction first in to-infinitivals and then
in gerund-participials.
2.1 To-infinitivals (I hoped to convince them vs I seemed to convince them)

The semantic difference between the two kinds of subject can be illustrated with the
verbs hope and seem:
[1] i Liz hoped to convince them. [ordinary subject]
ii Liz seemed to convince them. [raised subject]
In [i] Liz is an ordinary subject in that it is an argument of the verb: hope denotes a
psychological attitude on the part of someone to some situation (here Liz’s attitude to
the later, potential, situation where she convinces them). But in [ii] Liz is not an argu-
ment of seem. The meaning is something like “Seemingly, Liz convinced them”; seem
has a modal meaning, serving to make [ii] weaker than the unmodalised Liz convinced
them. Syntactically Liz is subject of seem, but semantically it relates only to the subor-
dinate convince clause, not to seem. Liz in [ii] is then a raised subject: the verb that Liz
relates to syntactically is higher in the constituent structure than the one it relates to
semantically.
In [1i], Liz has two semantic roles, as experiencer of hope and as agent of convince;
but in [ii], Liz has only one semantic role, as agent of convince. The difference is like that
between [68i] and [68ii] of §1.7 (Pat persuaded Kim to travel by bus and Pat intended Kim
to travel by bus), and we will again say that the missing subject of the non-finite clause
has a controlled interpretation in [1i] and a raised interpretation in [1ii]. Verbs like seem
which take a raised complement are called raising verbs.
Evidence for the distinction between an ordinary subject and a raised subject
(a) Relation with finite complement construction
With verbs which take either infinitival or finite complements the semantic relations are
more transparent in the construction with a finite complement. Compare:
[2] i Lizi hoped that shei would convince them. [two arguments]
ii It seemed that Liz convinced them. [one argument]
In [i] hope clearly has two arguments, represented by the subject and the content clause.
The identical indices on Liz and she indicate that we are concerned with [i] in the
interpretation where she has Liz as its antecedent: the double reference to Liz thus makes
transparent that Liz has two semantic roles, as experiencer of hope and agent of convince.
In [ii], by contrast, there is only one reference to Liz, in the subordinate clause: again,
then, the construction shows transparently that Liz has just one semantic role, agent of
convince. There is no syntactic relation between Liz and seem, and this reflects the absence
of any semantic relation between them. The subject it is a dummy element: it satisfies the
syntactic requirement for a subject but is not an argument of the verb. Thus while hope
§ 2.1 To-infinitivals 1195
has two semantic arguments, seem has only one. The semantic difference is the same as
in [1], but this time it is reflected syntactically because raised subjects are found only
with non-finite complements.
(b) Relation with passive infinitivals
Compare now the following superficially similar pairs:

[3] i a. Liz hoped to convince them.
[voice-sensitive: a = b]
b.They hoped to be convinced by Liz.

ii a. Liz seemed to convince them.
[voice-neutral: a = b]
b.They seemed to be convinced by Liz.
Such pairs bring out very clearly the difference between hope and seem, for with hope
[a] and [b] differ sharply in meaning, whereas with seem they are equivalent. In [i] the
content of the hope remains constant (namely, that Liz convince them or, equivalently,
that they be convinced by Liz), but the experiencer is different: Liz in [ia], they in [ib].
The subject of hope has the experiencer role, and changing the subject therefore changes
the meaning. In [ii], however, the subject is not an argument of seem and this is why it is
possible to change the subject without affecting the overall propositional meaning: the
equivalence of [iia] and [iib] matches that found in the main clause pair Liz convinced
them and They were convinced by Liz, or in It seemed that Liz convinced them and It seemed
that they were convinced by Liz. Here, then, the subject has a semantic role relative to
convince, but not to seem. We speak of hope-type verbs as voice-sensitive and of seem-
type verbs as voice-neutral – sensitive or neutral, that is, to a change of voice in the
sequence consisting of their subject and the non-finite VP.
(c) Selection restrictions
Hope imposes selection restrictions on its subject: it must denote some entity capable of
filling the experiencer role, and hence be animate and typically human. But seem does
not restrict its subject: any NP is permitted provided only that it satisfies any selection
restrictions imposed by the subordinate clause verb. Compare, then, the anomaly of [i]
with the acceptability of [ii] in:
[4] i #This news hoped to convince them. [violation of restriction]
ii This news seemed to convince them. [no violation]
This property of hope demonstrates conclusively that there is a direct semantic relation
between it and its subject. And the absence of restrictions with seem is predicted by the
raised subject analysis: if there is no direct relation between the two elements there could
be no comparable restrictions.
(d) Dummy subjects – subjects with no independent meaning.
With raising verbs, the dummy subject NPs there and it are possible:
[5] i There seemed/∗hoped to be enough food available. [dummy there]
ii It seemed/∗hoped to be unwise to dissent. [dummy it]
The subjects here have no identifiable meaning of their own, so the unacceptability of
the hope versions can be seen as following from point (c) above: such NPs could not
satisfy the requirement that the subject denote an animate being capable of filling the
experiencer role.
(e) Choice between simple and complex constructions.

[6] simple complex
i a. Ed expected to die. b. Ed expected Pat to die.
ii a. Ed arranged to go. b. Ed arranged for Pat to go.
iii a. Ed tried to help us. b. ∗ Ed tried (for)Pat to help us.
iv a. Ed seemed to faint. b. ∗ Ed seemed (for)Pat to faint.
Verbs entering into the simple catenative construction vary according to whether they
can also appear in the complex construction – either the plain subtype, as in [i], or the
one with for, as in [ii]. All raising verbs are like seem in being restricted to the simple
construction. This restriction follows directly from the fact that they take raised sub-
jects. In [iva] Ed represents an argument of faint but not of seem: Ed simply refers to the
one who seemingly fainted; if we then add another NP as the subject-argument of faint
there is no semantic role left for Ed, so [ivb] is uninterpretable. This problem doesn’t
arise with the non-raising verbs, many of which allow a choice between the simple and
complex constructions. In [ia] Ed is the experiencer of expect and the theme of die: when
we add Pat in [ib] it takes over the latter role but leaves the former to Ed. Similarly in
[ii]: whether or not we have the subordinator for before the inserted NP is irrelevant
to the issue of the semantic roles. There are nevertheless some non-raising verbs, like
try in [iii], which are incompatible with either version of [b]. In such cases there is often
a semantic explanation for the restriction. Thus trying involves internal effort that is
necessarily directed towards one’s own actions: Ed could try to persuade Pat
to help us or try to make Pat help us, but it is not possible for the trying itself to
apply directly to Pat.14
(f) Matrix passivisation
A few non-raising catenatives allow passivisation, with the infinitival appearing as ex-
traposed subject, but this is quite impossible with raising verbs:
[7] i a. We hope to return to this issue. b. It is hoped to return to this issue.
ii a. We seem to be in danger. b. ∗It is seemed to be in danger.
Hope takes two arguments, and the passive allows the active subject to be left unexpressed,
with the infinitival appearing as extraposed subject; seem takes only one argument, so
passivisation is quite impossible.
Summary of differences between hope-type verbs and seem-type verbs

The six differences between non-raising verbs like hope and raising verbs like seem,
described in (a)–(f) above, are summarised in [8]:
[8] property non-raising verbs raising verbs
(a) Finite complement Two arguments (or N/A) One argument (or N/A)
(b) Infinitival voice Voice-sensitive15 Voice-neutral
(c) Restrictions on subject Yes No
(d) Dummy subject No Yes
(e) Simple/complex choice Yes for many verbs No
(f) Matrix passivisation Yes for a few verbs No
For footnotes 14–15, see next page.
‘N/A’ in (a) means “not applicable”: this is the case when the verb doesn’t allow a finite
complement; (e) and (f) provide a distinguishing criterion only in the case of a ‘yes’
answer, which indicates a non-raising verb, ‘no’ being consistent with either type.
The problem of agentivity
With so many factors involved, the distinction between ordinary and raised subjects is gener-
ally clear, but it is not always so. The main problem involves agentivity, and it arises primarily
with aspectual verbs like begin. These verbs don’t take finite complements, so point (a) is not
applicable. Consider, then, the following examples:
[9] i There began to be some doubt in our minds as to whether he was trustworthy.
ii a. His behaviour began to alienate his colleagues.
b. His colleagues began to be alienated by his behaviour.
iii a. Jill began to unwrap the parcel.
b. ? The parcel began to be unwrapped by Jill.
In [i–ii] begin behaves like a raising verb. We see from [i] that it allows a semantically empty
subject (property (d)), and hence does not impose selection restrictions (property (c)). In
[ii] we have a clear example of voice-neutrality (property (b)). But [iiib] is not an acceptable
alternant to [iiia]. The difference between [ii] and [iii] is that the situation described in the
infinitival is non-agentive in the former, agentive in the latter: compare the corresponding
main clauses His behaviour alienated his colleagues and Jill unwrapped the parcel, where his
behaviour has a non-agentive role, Jill an agentive one. And if the whole situation of Jill’s
unwrapping the parcel is agentive, isn’t the initial phase of it, the beginning, likewise agentive?
If so, Jill in [iiia] will be an ordinary subject, not a raised one – which will account for the
apparent difference in the relation of [a] to [b] in [ii] and [iii]. This reasoning has thus led to
a position where begin can take either an ordinary or a raised subject according to whether
the subordinate situation is agentive or not.
There are good grounds, however, for rejecting this solution to the problem posed by
[9iiib]. The important point to note is that while [iiib] is of low acceptability it does
not describe a different situation from [iiia]: there is no difference in truth conditions
such as we have in [3i] (Liz hoped to convince them vs They hoped to be convinced by
Liz). Indeed, we can find examples involving begin with an agentive infinitival where the
two members of the pair are not only equivalent but also both fully acceptable: Max be-
gan to court Jill and Jill began to be courted by Max (or compare They began to sell the
shares at a discount and The shares began to be sold at a discount). Voice neutrality does
not require that the two alternants be equally likely and acceptable (for this condition
does not hold for many simple active–passive pairs: compare I opened my eyes and ?My
eyes were opened by me), only that they be truth-conditionally equivalent. The data in
[9iii] are thus quite consistent with a unitary treatment of begin as a raising verb in all of
14
Our model verb hope takes for (She was hoping for Kim to return safely), but a finite construction (She was
hoping that Kim would return safely) is much more likely than the complex infinitival, and hence we have used
other examples in [6].
15
Examples are occasionally attested where non-raising verbs are treated as though they were voice-neutral: The
exam papers are trying to be marked by next week (“We are trying to mark them”). There can be no doubt,
however, that they are rare and unsystematic enough for us to be able to dismiss them as mistakes.
[9i–iii], which is much preferable to putting it in both classes, with a great deal of resulting
indeterminacy over the analysis of particular examples.
The issue of agentivity also arises, but with a larger set of verbs, in cases like:
[10] i It’s time you began to relax.
ii It’s essential that you at least APPEAR to be enjoying yourself.
Here begin and appear occur in subordinate clauses where the matrix confers an agentive
interpretation on their subjects: it is a matter of your volitionally bringing about the situations
where you begin to relax and appear to be enjoying yourself. Although agentive subjects
characteristically represent arguments of their verbs, it is reasonable to maintain that they
don’t necessarily do so, for in cases like [10] it is not the verbs begin and appear themselves
that confer agentivity on the subject, but the larger context. These examples are comparable
to the passive given in note 16 of Ch. 4, They advised the twins not to be photographed together,
where advise confers agentivity on the understood subject of the infinitival, although the role
assigned by photograph itself is clearly not agent. The fact that the agentivity in [10] is not
attributable to the verbs begin and appear is quite consistent with you being a raised subject,
i.e. as representing an argument of relax and enjoy rather than begin and appear.
2.2 Gerund-participials (We enjoyed sailing vs We kept sailing)

The distinction between ordinary and raised subjects applies to gerund-participials too:
[11] i Kim enjoyed heckling him. [ordinary subject]
ii Kim kept heckling him. [raised subject]
In [i] Kim is an argument of enjoy, with the semantic role of experiencer. In [ii], however,
there is no direct semantic relation between Kim and keep: [ii] simply says that the
situation in which Kim heckled him recurred over and over again. The parallel with
the infinitival construction is made clearer by the fact that there are both raising and
non-raising catenatives that take either infinitival or gerund-participial complements,
e.g. raising begin and continue, non-raising like and hate.
Three of the six distinguishing properties discussed above apply here quite straight-
forwardly: we therefore first review these summarily together, and then take up the other
three.
Passive infinitivals, selection restrictions, dummy subjects

[12] i a. He enjoyed being heckled by Kim. [=[11i]: enjoy is voice-sensitive]
b. He kept being heckled by Kim. [=[11ii]: keep is voice-neutral]
ii a. #My papers enjoyed blowing away. [violates selection restrictions on subject]
b. My papers kept blowing away. [no selection restrictions on subject]
iii a. ∗There enjoyed being problems with the radio. [dummy subjects excluded]
b. There kept being problems with the radio. [dummy subjects allowed]
These show respectively that the raising catenative keep, like seem and in contrast to
non-raising enjoy and hope, is neutral as to the voice of the non-finite (property (b)),
imposes no selection restrictions on its subject (property (c)), and hence allows dummy
subjects such as there (property (d)).
§ 2.3 Concealed passives 1199
Relation with finite complement construction

Property (a) is less useful as a diagnostic this time because fewer verbs taking a gerund-
participial also allow a finite complement. None of the raising verbs do, but some non-
raising ones do and exhibit the predicted behaviour:
[13] i I regret giving him my address.
ii I regret that I gave him my address. [two arguments]
It is transparent in [ii] that regret has two semantic arguments, represented by the subject
and the content clause. I here has two semantic roles, experiencer of regret and agent of
give.
Choice between simple and complex constructions

With respect to property (e), the gerund-participial construction differs from the in-
finitival in that there are raising as well as non-raising verbs that occur either with or
without an intervening NP:
[14] i a. Ed likes reading aloud. b. Ed likes me reading aloud.
ii a. Ed kept sitting on the stool. b. Ed kept me sitting on the stool.
However, the semantic relation between the two constructions is not the same in the
two cases. With like, which takes an ordinary subject, the difference is the same as in the
infinitival case: the meaning of the verb remains constant and Ed remains experiencer, so
that the difference is simply a matter of who is doing the reading – Ed in [a], me in [b].
But in [ii] the meaning of keep does not remain constant: in [b] it has a causative sense
(“Ed made me stay sitting on the stool”) and in [a] there is an element of interruption
and repetition (Ed repeatedly got on and off the stool) that is missing in [b]. Given this
difference in meaning, [iib] is not inconsistent with a raising analysis of [iia]: Ed is an
argument (agent) of keep in [iib] but not in [iia].
Matrix passivisation
Gerund-participials differ from infinitivals in that they do not readily allow extraposition,
so we focus here on the construction without extraposition. Matrix passivisation of this
kind is predictably quite impossible with raising verbs; with non-raising verbs it tends to
be somewhat marginal, but acceptability can be increased by adding motivating context:16
[15] i ∗Heckling people was kept by Kim.
ii ? Heckling people was enjoyed by Kim.
iii Watching TV is enjoyed by far more people than reading novels.
2.3 Concealed passives (The house needs painting)

With a small number of catenative verbs, notably need, require, deserve, and want, a
gerund-participial may be passive while lacking the usual marking of the passive – we
16
A fully acceptable type of passive is illustrated in Swimming after a heavy meal is not recommended. Recommend
is clearly a non-raising verb, but it differs from enjoy in that the matrix subject is not the understood subject
of the gerund-participial: in Doctors don’t recommend swimming after a heavy meal it’s a matter of people in
general, not doctors, swimming after a meal.
refer to this as the concealed passive construction. Compare:

[16] i a. The house needs to be painted.
[ordinary passive]
b. These books want to be taken back to the library.

ii a. The house needs painting.
[concealed passive]
b. These books want taking back to the library.
The catenative complements in [i] are overtly marked as passives by the auxiliary
be and the past participle forms painted and taken, while those in [ii] lack such marking
although they are interpreted in the same way.17
In the absence of distinctive kinds of dependent there will often be syntactic ambiguity
between a concealed passive gerund-participial and a gerundial noun, but commonly
with negligible difference in meaning:
[17] i The children need coaxing.
ii It’s an attractive feature of avocados that they do not require processing.
In [i] coaxing can be a verb, “to be coaxed”, or a noun, as in They need a little coaxing, but
we understand that the coaxing should apply to the children, so the meaning is effectively
the same as with the verbal reading. Likewise in [ii]: processing can be a verb, “to be pro-
cessed”, or a noun, as in processing of any kind, and in either case it is a matter of processing
the avocados. Similar examples can be formed with training, teaching, and numerous
others.
The concealed passive is to be distinguished from a hollow clause, an active with a
missing non-subject element, object of the verb or of a preposition:
[18] i The article needs checking. [concealed passive]
ii The article is worth reading . [hollow: active]
Note that the first construction is possible only where there is a corresponding ordinary
passive, but this is not so with the second. Compare, then:
[19] i ∗The article was had a careful look at. [ordinary passive]
ii ∗The article needs having a careful look at. [concealed passive]
iii The article is worth having a careful look at . [hollow: active]
The deviance of [ii] here reflects that of the main clause [i], whereas the acceptability of [iii]
matches that of an active main clause like We will have a careful look at the article.
A second point is that the concealed passive can contain the by phrase that appears in
ordinary passives as an internalised complement:
[20] i The article needs checking by the editor. [concealed passive]
ii ∗The article is worth reading by the editor. [hollow: active]
The concealed passive is found only with gerund-participials and in the catenative construc-
tion; the hollow active is more often found with non-catenative to-infinitivals, as in an easy
problem to solve (§6).
3 The complex catenative construction
In the complex construction the matrix and subordinate verbs are separated by an
intervening NP which functions as complement in one or other of the clauses. As with
the simple construction, we look first at to-infinitivals and then at gerund-participials.
17
Want with the sense “need” is more characteristic of BrE/AusE than AmE, and is more likely with a concealed
passive gerund-participial than with a to-infinitival (where there is more danger of confusion with the primary
sense of “desire”).
3.1 To-infinitivals
3.1.1 The plain-complex construction (I persuaded Liz to go vs I intended Liz to go)
In this construction the intervening NP always belongs syntactically in the matrix: it
functions as matrix object, as argued in §1.3 above. Semantically, however, we find a
contrast according as the object represents an argument of the matrix (an ordinary
object) or only of the subordinate clause (a raised object):
[1] i Pat persuaded Liz to interview both candidates. [ordinary object]
ii Pat intended Liz to interview both candidates. [raised object]
In [i] the syntactic structure matches the semantics quite straightforwardly. Persuade has
three complements (Pat, Liz, and the infinitival) and each represents an argument: the
matrix situation involves one who applies the persuasion (Pat), one to whom it is applied
(Liz), and the situation aimed for (that Liz interview both candidates). Liz is thus an
ordinary complement, an argument of the verb which governs it. But in [ii] there is no
such simple relation between syntax and semantics. In particular, Liz is not an argument
of intend: the situation simply involves one who has the intention (Pat) and the content
of the intention (that Liz interview both candidates). With intend, therefore, we have
three complements but only two arguments: Liz is a raised object.
The distinction is parallel to that discussed in §2 for the simple catenative construction:
there we were concerned with the semantic status of the subject, here with that of the
object (the intervening NP). We will therefore extend the term ‘raising verb’ to cover
intend as well as seem – where necessary we can distinguish them as ‘raised object verb’
and ‘raised subject verb’ respectively. The first five factors discussed in §2 for the contrast
between ordinary and raised subjects apply again here with suitable adjustments.18
(a) Relation with finite complement construction

Compare [1i–ii] first with corresponding finite constructions:
[2] i Pat persuaded Lizi that shei should interview both candidates. [three arguments]
ii Pat intended that Liz should interview both candidates. [two arguments]
Even more clearly than with the hope vs seem contrast the switch to a finite complement
brings out the semantic difference between persuade and intend. In [2] persuade has three
syntactic complements, and three semantic arguments, while intend has two of each –
this time there is no dummy it to complicate the picture. In [i] the dual semantic role of
Liz, as patient of persuade and agent of interview, is now transparent, with Liz appearing
as matrix object and co-referential she as subordinate subject.19 With intend the fact that
there is no direct semantic relation between intend and Liz is likewise transparent in the
finite construction, where Liz is not a complement of intend.20
18
The sixth factor, matrix passivisation, does not apply here because in the complex catenative construction it is
the intervening NP, not the infinitival, that is promoted to subject, and this kind of passivisation applies with
ordinary and raised objects alike (though in either case there are some verbs which block it: see §5).
19
There is a slight difference in meaning between the infinitival and finite constructions with persuade. Whereas
[1i] entails that Liz agreed or undertook to interview the candidates, [2i] is a little weaker: she accepted that
there was an obligation on her to do so. But this doesn’t affect the semantic status of Liz : in both cases Pat
applied persuasion directly to Liz, producing a change in her psychological state.
20
We have noted that the subordinator for is like that in marking clause boundaries very clearly, so that the
for-complex construction provides the same kind of evidence for a raising analysis with the few verbs that
enter into both plain- and for-complex constructions. Compare, then, I’d prefer Liz to do it herself and I’d prefer
(b) Relation with passive infinitivals

Compare next the following superficially similar pairs:

[3] i a. Pat persuaded Liz to interview both candidates. [voice-sensitive
b. Pat persuaded both candidates to be interviewed by Liz. a= b]

ii a. Pat intended Liz to interview both candidates. [voice-neutral
b. Pat intended both candidates to be interviewed by Liz. a = b]
In [i] there is an obvious difference in meaning: in [ia] Pat applied persuasion to Liz,
but in [ib] to the candidates. This shows that the object is an argument of persuade. In
[ii], by contrast, [a] and [b] are equivalent, just as they are in the main clause pair Liz
interviewed both candidates and Both candidates were interviewed by Liz. The fact that
we can change the object in this way without affecting the propositional meaning shows
that there can’t be any direct semantic relation between it and the matrix verb.
(c) Selection restrictions

Persuade imposes selection restrictions on its object: except when used in an extended,
metaphorical sense it requires that its object denote a sentient being capable of making
decisions. But intend does not restrict its object: any NP is permitted that is compatible
with the infinitival. Compare the anomaly of [i] with the acceptability of [ii] in:
[4] i # Liz persuaded the spotlight to intimidate Pat. [violation of restriction]
ii Liz intended the spotlight to intimidate Pat. [no violation]
(d) Dummy objects – objects with no independent meaning
Dummy objects are found with intend, but not persuade, and again the unacceptable
versions with persuade can be regarded as special cases, more grammaticalised, of a
violation of selection restrictions:
[5] i Pat intended/∗persuaded there to be one student on the board. [dummy there]
ii Pat intended/∗persuaded it to be easy to obtain a pass grade. [dummy it]
(e) Choice between complex and simple constructions
[6] complex simple
i a. Liz persuaded Pat to leave. b. ∗Liz persuaded to leave.
ii a. Liz intended Pat to read it. b. Liz intended to read it.
iii a. Liz believed Pat to be ill. b. ∗Liz believed to be ill.
We have noted that catenative verbs differ according to whether they enter into both
simple and complex constructions or only one. In §2.1 the contrast was between simple
only (seem) vs both (expect), whereas here it is between complex only (persuade) and
both (intend). In [i] Pat is an argument in the matrix and is obligatory: even if the
understood subject of the infinitival is referentially identical with the subject of persuade
we need the intervening NP, which will be a reflexive, Liz persuaded herself to try again.
In [ii] Pat has no role relative to intend, but is simply agent of read; if we make the agent
of read identical with the experiencer of intend then it can be left understood, as in [iib].
As in §2.1 we find here an imperfect correlation with the ordinary vs raised distinction.
for Liz to do it herself : there is no perceptible difference in meaning and the lack of a direct semantic relation
between prefer and Liz is transparent in the for construction.
§ 3.1.2 The for-complex construction 1203
The main complicating factor (one which has no evident semantic explanation) is that
there is a rather large class of verbs of cognition and saying, like believe, which require an
intervening NP even when there is referential identity with the matrix subject: instead
of [iiib] we have Liz believed herself to be ill.
Summary of differences between persuade and intend

The distinctions between non-raising verbs like persuade and raised object verbs like
intend are summarised in [7]:
[7] property non-raising verbs raising verbs
(a) Finite complement Three arguments (or N/A) Two arguments (or N/A)
(b) Infinitival voice Voice-sensitive Voice-neutral
(c) Restrictions on object Yes No
(d) Dummy object No Yes
(e) Simple/complex choice Normally no21 Yes for many verbs
Property (e) provides a distinguishing criterion only when the answer is ‘yes’, indicating
a raising verb, since a ‘no’ answer is consistent with either type.
3.1.2 The for-complex construction (I arranged for her to go by bus)

This construction is fairly straightforward: for is a subordinator marking the start of the
non-finite clause and is obligatorily followed by the subject NP, in accusative or plain
case. There are no catenative verbs that appear only in this construction. All but a few
are found also in the simple construction with the meaning contrast illustrated in:
[8] i I arranged for her to go by bus. [for-complex]
ii I arranged to go by bus. [simple; control by subject]
In [i] the infinitival has an overt subject, in [ii] a covert one, controlled by the matrix
subject – it’s a matter of my going by bus.
The verbs which don’t follow this pattern appear in contrasting complex construc-
tions:
[9] i a. He pressed for Ed to be admitted. b. He pressed Ed to join the club.
ii a. He called for Ed to be sacked. b. He called on Ed to resign.
In [ia] Ed belongs straightforwardly to the subordinate clause, whereas in [ib] Ed is an
ordinary object of press and controller of the understood subject of join. Ed has one
semantic role in [ia], the ‘admittee’, but two in [b], patient of press and agent of join. The
semantic difference is like that between He intended Ed to be admitted and He persuaded
Ed to be admitted but with press the semantics is reflected directly by the syntax. Other
verbs that follow the pattern of press are push (informal style) and signal; ask combines
the contrasts shown by arrange and press (He asked for Kim to chair the meeting; He asked
to chair the meeting; He asked Ed to chair the meeting – but see, further, §5.3). Call in
[9ii] is similar to press except that in [b] Ed is an oblique, not an object: [9iib] belongs
to the oblique-complex construction.
21
There are one or two cases of non-raising verbs where the intervening NP is omissible, but the semantic effect
is not the same as with raising verbs. For example, the difference between I’ll help you wash up and I’ll help
wash up does not match that between [6iia] and [6iib], for in I’ll help wash up it is a matter of helping some
unspecified person(s).
3.1.3 The oblique-complex construction (I signalled to her to move off the road)
A few prepositional verbs take a subjectless infinitival complement as well as the prepo-
sitional one:
[10] i I signalled to her to move off the road.
ii She relies on him to look after the shop while she’s out.
The oblique NP serves as controller for the missing subject of the infinitival. It cannot
be a dummy pronoun, and hence not a raised complement: ∗I signalled to there to be a
pause ; ∗She relies on there to be a daily delivery of fresh bread.
3.2 Gerund-participials
Here too we need to distinguish according as the intervening NP is or is not an argument of
the matrix verb – but syntactically the plain-complex gerund-participial construction
is not wholly parallel to the infinitival one. We need to recognise three matrix verb types,
not just two.
3.2.1 Catch vs resent (I caught him doing it vs I resented him doing it)
Semantic difference between catch and resent
From a semantic point of view, these verbs are comparable to persuade and intend
respectively, in that the intervening NP is a matrix argument with catch but not with
resent :
[11] i I caught Kim mistreating my cat. [matrix argument]
ii I resented Kim mistreating my cat. [not matrix argument]
Example [i] might be glossed as “I caught Kim in the act of mistreating my cat”: Kim is
the patient-argument of catch. In [ii], however, it wasn’t Kim the person that I resented
but the whole situation in which Kim mistreated my cat. This semantic distinction is
borne out by the familiar kind of evidence:
[12] i a. ?I caught my cat being mistreated by Kim. [voice-sensitive: = [11i]]
b. I resented my cat being mistreated by Kim. [voice-neutral: = [11ii]]
ii a. ∗I caught there being several non-members present. [dummy excluded]
b. I resented there being several non-members present. [dummy allowed]
iii I resented that Kim mistreated my cat. [two arguments]
Catch doesn’t take a finite complement, so only one example is given in [iii], but it
still shows that resent behaves in the predictable way, with the NP appearing in the
subordinate clause and hence transparently an argument of mistreat, not of resent.
Syntactic difference between catch and resent

When we turn to matters of syntax, however, we find that the relation between the two
verbs is not the same as with persuade and intend. A crucial difference between gerund-
participials and to-infinitivals is that a non-genitive NP can function as subject of the
former but not of the latter unless for is present (cf. §§1.4–5):
[13] i What I resented was Kim mistreating my cat.
ii ∗What I intended was Kim to interview both candidates.
iii What I intended was for Kim to interview both candidates.
§ 3.2.2 See (I saw him doing it) 1205
We have no reason, therefore, to analyse the NP following resent as a raised object –

instead we take it as subordinate subject, so that the syntactic structure matches the
semantics:
[14] i I caught Kim mistreating my cat. [matrix argument; matrix object]
ii I resented Kim mistreating my cat. [not matrix argument; subordinate subject]
Evidence for a syntactic difference of this kind is as follows:
(a) Possibility of genitive marking
This can be added only when the NP is subject of the gerund-participial:
[15] i I caught Kim/∗Kim’s mistreating my cat. [matrix object: genitive excluded]
ii I resented Kim/Kim’s mistreating my cat. [subordinate subject: genitive allowed]
(b) Matrix passivisation
The intervening NP can be promoted to subject only if it is matrix object:
[16] i Kim was caught mistreating my cat.
ii ∗Kim was resented mistreating my cat.
With resent it is even marginally possible for the NP together with the gerund-participial
VP to be promoted, especially with genitive marking: ?Kim(’s) being given such an unfair
advantage was deeply resented by everyone.
(c) Choice between simple and complex constructions
[17 complex simple
i a. I caught myself apologising. b. ∗I caught apologising.
ii a. I resent them being exploited. b. I resent being exploited.
Resent enters into both constructions: if the subordinate subject-argument is identical
with that of the matrix it is left unexpressed. With catch the intervening NP is obligatory,
part of the complementation of the verb: a reflexive is therefore required in comparable
cases of identity, as in [ia].
3.2.2 See (I saw him doing it)

See falls in between the sharply distinct patterns found with catch and resent : it is like
resent in that the intervening NP is not a matrix argument, but like catch in that it is
syntactic object of the matrix clause:
[18] matrix argument matrix object
i I caught Kim mistreating my cat. Yes Yes
ii I saw Kim mistreating my cat. No Yes
iii I resented Kim mistreating my cat. No No
Semantically, see involves an experiencer and a stimulus: in [ii] the stimulus is the event
wherein Kim mistreats my cat.22 Generally, when one sees an event one also sees the
participants in the event, so that it will normally be inferred from [ii] that I saw Kim –
but this should not mislead us into thinking that Kim is an argument of see. For note
first that I presumably also saw my cat, but there can be no question of my cat being an
argument of see in [ii]. And, secondly, the intervening NP need not refer to something
22
We ignore at this point the aspectuality: in I saw Kim mistreating my cat the subordinate clause is progressive
in contrast to the non-progressive of I saw Kim mistreat my cat (see §5.4).
visible: I saw the stress of these last few months taking its toll of her.23 Because the intervening
NP is not a matrix argument, see behaves like resent with respect to voice-neutrality and
the occurrence of dummy elements:
[19] i I saw my cat being mistreated by Kim [voice-neutral: = [18ii]]
ii We had seen there developing between them a highly
destructive antagonism. [dummy allowed]
In the syntactic structure, however, the intervening NP belongs in the matrix clause,
as object of see. See is therefore like our model verb intend in taking a raised object.24
The evidence for this analysis is given in:
[20] i I saw Kim/∗Kim’s mistreating my cat. [genitive excluded]
ii Kim was seen mistreating my cat. [matrix passivisation]
iii ∗I saw opening the safe. [simple construction excluded]
These examples show that see behaves syntactically like catch, not resent – compare
[20i–iii] with [15–17] respectively.
3.2.3 Concealed passives with intervening NP (He needs his hair cutting)
Concealed passives are found in the complex as well as the simple construction:
[21] i Your hair needs cutting by a professional. [simple]
ii You need your hair cutting by a professional. [complex]
Semantically your hair is not a matrix argument in either case – we can gloss as “There
is a need for your hair to be cut by a professional” and “You have a need for your hair
to be cut by a professional”. We take it to be a raised subject in [i], a raised object in
[ii]. The syntactic structure is less clear than with see since there is here no possibility of
matrix passivisation. The main reason for taking your hair as matrix object in [ii] is that
genitive marking is completely unacceptable.
4 The catenative complement as a distinct type of complement
4.1 Non-finites in relation to NP and AdjP complements

The traditional classification of subordinate clauses as nominal, adjectival, and adverbial
implies that they can be identified functionally with phrases headed by nouns, adjec-
tives, and adverbs respectively. Certainly there are places where such identifications can
validly be made. The clearest case is that of the subject, the type of complement that is
syntactically most clearly distinguished from others:
[1] i a. The first answer was wrong. b. Pretending you were ill was wrong.
Another is the complement of be in its specifying use, where subject and predicative
complement can be switched:
23
This point is even more evident with feel: I felt him running the feather down my back clearly doesn’t entail that
I felt him.
24
The catenative use of see is to be distinguished from that where the gerund-participial is an adjunct, as in I saw
Kim at the back of the class talking to her neighbour ; the two kinds of gerund-participial can combine: I saw
them walking across the courtyard, arguing vociferously.
§ 4.1 Non-finites in relation to NP and AdjP complements 1207
[2] i a. His goal is total victory. b. His goal is to win at all costs.
ii a. Total victory is his goal. b. To win at all costs is his goal.
The view taken in this grammar, however, is that it is not always possible to identify
the function of subordinate clauses with that of non-clausal constituents such as NPs,
AdjPs, and AdvPs. We argued this case for finite clauses in Ch. 11, §8.2; here we focus on
catenative complements, arguing that they cannot be systematically analysed as objects
or predicative complements.
Irrelevance to to-infinitivals of the distinction between object and oblique

With NPs we have a clear distinction between an object (related directly to the verb) and an
oblique (related via a preposition), but this distinction does not apply to to-infinitivals, for
they cannot occur as complement to a preposition (or at least not to a preposition of the
relevant kind). Compare, for example:
[3] i a. Kim started the riot. b. Kim started to riot.
ii a. Kim remembered the cat. b. Kim remembered to feed the cat.
iii a. They came to an agreement. b. They came to agree on the main points.
iv a. He proceeded to the next task. b. He proceeded to shred the documents.
The underlined complements in the [a] examples are objects in [i–ii] but PP complements
in [iii–iv] (where the NPs an agreement and the next task are related only obliquely to the
verb), but the [b] examples show no such difference. There is then no reason to draw a
functional distinction between the non-finite clauses of [ib/iib] and those of [iiib/ivb], and
no reason to identify their functions with those of the object or oblique NPs in the [a]
examples. The fact that the NP complements of start and remember in [ia/iia] are objects
is not reason enough for saying that the non-finite complements are objects too, because
there are significant syntactic differences between the two types of complement. For exam-
ple, NPs can be promoted to subject through passivisation but to-infinitivals cannot: The
riot was started by Kim but not ∗To riot was started by Kim. Note, moreover, that as far as
the non-finite complements are concerned the significant distinction is not between the
constructions of [ib/iib] and those of [iiib/ivb] but between those with a raised subject, the
start and come examples, and those with an ordinary subject, the remember and proceed
examples.
Irrelevance to non-finites of the distinction between objects and predicatives

With NP complements we have a clear distinction between objects and predicative comple-
ments, but again this distinction is in general irrelevant to non-finite complements:
[4] i a. Kim offered financial advice. b. Kim offered to help.
ii a. Pat needs a large loan. b. Pat needs to consult a solicitor.
iii a. Ed seems a nice guy. b. Ed seems to like him.
Financial advice and a large loan are objects, while a nice guy is predicative complement.
The main grammatical differences between objects and predicative complements concern
passivisation and realisation by an AdjP, but these differences do not carry over to the examples
with non-finite complements. Examples [ia] and [iia] can be passivised, though [iiia] cannot
(Financial advice was offered by Kim; A large loan is needed by Pat ; ∗A nice guy is seemed by
Ed ); but none of the [b] examples can undergo passivisation. And with the NP complements
seem again differs from offer and need in systematically allowing a replacement of the NP a
nice guy by the AdjP nice, but here too there is no equivalent difference with the non-finites.
It is true, of course, that we can replace to like him in [iiib] by nice, but this simply reflects the
fact that seem can take a predicative complement as well as an infinitival one: it doesn’t show
that the latter is a predicative, for the relationship is not the same as with the replacement of a
nice guy by nice in [iiia] (note, for example, that we couldn’t replace the infinitival by an AdjP
in There seems to be a serious misunderstanding here). It follows that there is no valid basis
for generalising to the non-finites the analysis that applies to the NP complements. Again,
the examples have been chosen to bring out the point that the syntactic grouping of verbs in
the [a] examples is different from the semantic grouping in the [b] examples: with infinitival
complements need belongs with seem, not offer, for need and seem take a raised subject while
offer takes an ordinary one (cf. There seems/needs/∗offers to be ample justification for such a
course of action).
Application of the argument to the plain-complex catenative construction

The same points apply to constructions with two internal complements. Consider first:
[5] i a. She taught him Greek. b. She taught him to drive.
ii a. She invited him to her party. b. She invited him to chair the meeting.
In the [a] examples, teach takes two objects, invite an object and a PP complement, but this
distinction is irrelevant to the [b] examples: there’s no evidence for a comparable difference
in structure here.
Similarly with the contrast between objects and predicatives:
[6] i a. I told him the new duties. b. I told him to sweep the floor.
ii a. I consider him a failure. b. I consider him to have failed.
The new duties in [ia] is direct object, a failure in [iia] a predicative complement (with him as
predicand); and at the same time the function of him is different, indirect object in [ia], direct
object in [iia]. But again the syntactic properties distinguishing [ia] from [iia] do not apply
to the catenatives. For example, an indirect object is resistant to fronting in relativisation, etc.
(cf. ∗[the one]whom I told the new duties), but such restrictions do not apply to plain-complex
structures like [6ib] ([the one]whom I told to sweep the floor): the intervening NP here behaves
like a direct object rather than an indirect one. There are semantic differences between [ib]
and [iib], with him a matrix argument in [ib] but not [iib], and the time of the subordinate
situation being later than that of the matrix one in [ib] but not [iib]; but these do not provide
any basis for saying that they differ in syntactic structure in the way that [ia] differs from
[iia].
Varying degrees of likeness to objects

Catenative verbs vary in the extent to which their non-finite complements resemble
objects, but overall the similarities are fairly slight. Relevant factors are as follows:
(a) Passivisation
The catenative complements that are most like objects are those which can be promoted
to subject by passivisation. All are gerund-participials, as in:
[7] i Kim’s leaving early wasn’t mentioned.
ii Going out alone at night isn’t recommended.
All the verbs concerned also take NP objects, and passivise much more readily with an
NP object than with a gerund-participial: the above construction is quite rare and felt
to be rather awkward and marginal.
§ 4.2 The analysis of auxiliary verbs 1209
With a few verbs an infinitival complement can be promoted to extraposed subject:

[8] i It is planned to complete the work in three stages.
ii It is hoped to return to this issue.
This can hardly be regarded, however, as indicative of a significant resemblance to an
object, for clauses with NP objects do not undergo this kind of passivisation, and while
plan allows an NP object hope does not.
(b) Pronouns and clefts
It is not normally found with non-finite complements as antecedent: Yes, I want it, for
example, is not an appropriate response to Do you want to see them?, nor Yes, I avoided
it to Did you avoid implicating your father? Questioning with what is possible with some
verbs: the questions What do you want?, What have you decided?, What do you recom-
mend? might be answered respectively I want to see the manager, I’ve decided to accept
the offer, I recommend leaving things as they are. But this is very restricted, especially
with infinitivals: I’d like to go to Spain, for example, is not an answer to What would
you like?, nor You should stop drinking coffee to What should I stop? Relativisation in a
pseudo-cleft is possible with more verbs (What I’d like is to go to Spain for a few days ;
All I ask is to be allowed to get on with my work in peace) – but again with severe limitations
(∗What they began first was to arrange /arranging the cards in order).25 It-clefts are
marginally possible with some gerund-participials (It’s having to do the job under such
appalling conditions that I resent) but quite impossible with infinitivals (compare It’s
encouragement that she needs and ∗It’s to consult a solicitor that she needs).
In general, gerund-participials tend to be more like objects than to-infinitivals, reflect-
ing the nominal source of the traditional gerund. But there are considerable differences
among them, and (except for the cases defined in §7) it is better to treat gerund-participials
along with other non-finites in the catenative construction than to separate some of them
out and handle them with NP complements as object.
4.2 The analysis of auxiliary verbs

The syntactic class of auxiliary verbs is defined in English by the NICE properties –
occurrence with Negation, Inversion, Code, and Emphasis (see Ch. 3, §2.1). We have
distinguished between core and non-core uses of the auxiliaries: in the core uses they
appear with a following verb in one of its secondary forms (plain form, gerund-participle,
or past participle), whereas in the non-core uses they take an NP, a finite clause, or some
other kind of complement:
[9] core uses non-core uses
i a. He isn’t working. b. He isn’t a liar.
ii a. I haven’t seen it. b. %I haven’t time.
iii a. Would you regret it? b. Would you rather I did it?
25
In some cases a to-infinitival can occur instead of an NP as complement of a pseudo-cleft whose subject contains
a prepositional verb: What she agreed to was his proposal for a cooling-off period / to accept a cooling-off period
orWhat he longed for was her forgiveness / to know he was forgiven. But relatively few verbs behave in this way:
compare, for example, What he applied for was two months’ deferment / ∗to defer his enrolment or What she
decided on was a partial sale of the business / ∗to sell part of the business, and so on.
All auxiliaries have core uses, whereas non-core uses are found only with be, stative have
(in some varieties), and would in the idioms would rather / sooner / as soon.
There are two competing analyses of the core auxiliaries: on the dependent-auxiliary
analysis, they are dependents of a following main verb, whereas in the catenative-
auxiliary analysis they belong to the larger class of catenative verbs which take non-finite
complements. Now that we have examined the catenative construction in some detail,
we will present our arguments for adopting the catenative-auxiliary analysis; we first
set out the case for the dependent-auxiliary analysis and then show why the catenative
analysis is to be preferred.
4.2.1 The dependent-auxiliary analysis

Since the NICE properties are found with a few verbs which clearly take complements, as
in the non-core uses illustrated in [9], these properties cannot provide the justification
for taking the core auxiliaries as dependents rather than heads: the special treatment of
core auxiliaries must be based on other factors.
General motivation for the dependent-auxiliary analysis
These other factors can be brought out by comparing the following constructions:
[10] i Ed had a busy morning: he read the report. [sequence of main clauses]
ii Ed says that he read the report. [finite subordination]
iii Ed asked to read the report. [non-finite subordination]
iv Ed had read the report. [auxiliary construction]
(a) Sequence of main clauses
In [10i] there are two main clauses; they are of equal syntactic status, neither being subordinate
to the other. This relational independence is reflected in the internal independence of the
two clauses. That is, each selects independently of the other for tense, polarity, clause type,
subject, and so on. This is illustrated for polarity in:
[11] i Ed had a busy morning: he read the report. (=[10i]) [positive + positive]
ii Ed had a busy morning: he didn’t read the report. [positive + negative]
iii Ed didn’t have a busy morning: he read the report. [negative + positive]
iv Ed didn’t have a busy morning: he didn’t read the report. [negative + negative]
Similarly for subject selection: as it stands, the subject of the second clause is related anaphor-
ically to that of the first, but the two subjects could be quite unrelated, as in Ed had a busy
morning: the phone rang continuously. And so on: there are no grammatical limitations on
either clause imposed by the other.
(b) Finite subordination
As we move down to [10ii], we find that this internal independence of the two clauses has
diminished, though to a relatively modest extent. Here, that he read the report is a subordinate
clause serving as complement to say: there is therefore a clear structural interdependence in
that one clause is functioning within the structure of the other. By virtue of its subordination,
the read clause loses its potential illocutionary force. That is, in uttering [ii] I do not assert
that he read the report (as I normally do if I utter the simple sentence He read the report): I
don’t make two separate assertions, as in [i], but a single composite assertion. Related to this
is the fact that the choice of clause type in the complement clause is restricted. It would not be
possible to have an imperative clause here. And a closed interrogative is just about excluded
too (?Ed says whether he read the report ) though it would become completely normal if we
§ 4.2.1 The dependent-auxiliary analysis 1211
made the main clause negative or interrogative (Ed doesn’t say / Does Ed say whether he read
the report?), an indication of the interdependence between the two clauses. In other respects,
however, the read clause has the same range of options available as in a main clause: the
options for tense, polarity, subject, for example, are unaffected by the subordination (cf. Ed
says that the boss doesn’t read the reports, which differs in all three of these respects).
(c) Non-finite subordination
Moving from [10ii] to [iii] takes us from a finite to a non-finite subordinate clause, and
non-finites have a reduced range of clausal options open to them. Non-finite clauses tend
to be significantly less explicit than finite clauses: components of meaning that in finites
are directly expressed are in non-finites often left to be derived from the context in which
the clause appears. The non-finite has no inflectional tense – and no possibility therefore of
including a modal auxiliary (the possible contrast in [10ii] between that he may/must/can/will
read the report is not available in [iii]). Instead, the superordinate verb contributes a great deal
to the interpretation of the lower clause: in [iii], for example, the reading is merely potential
and in a time sphere subsequent to the time of asking, whereas in Ed remembered to read the
report it is actual and not temporally separable from the remembering. Like most non-finites,
the read clause of [iii] has no overt subject – but the understood subject is retrievable from
the superordinate subject Ed.
(d) The auxiliary construction
Finally, when we come to [10iv], the range of structural options available is further reduced, so
much so (the argument goes) that it is no longer justifiable to talk in terms of two clauses, one
embedded within the other. Historically it is a two-clause construction, comparable to [iii],
but the two clauses have lost their separate identities, merging together into a single clause.
This evolution has been accompanied by a reinterpretation of the direction of grammatical
dependence, with have now dependent on read, rather than the other way round.
The reduced independence of have and read in [10iv] relative to that of ask and read in
[10iii] is reflected in a number of ways, most notably [12i–iii], which we examine in turn:
[12] i Ask takes an argument subject, whereas have takes a non-argument subject.
ii With ask there is a very clear distinction between negating ask itself and negating the
complement, but have does not follow the same pattern.
iii Ask and its complement show a greater degree of temporal independence than we
find with have and the following verb.
The distinction between argument and non-argument subjects

Both [10iii], Ed asked to read the report, and [10iv], Ed had read the report, have only a single
subject, but there is nevertheless a major difference between them. In the semantic interpre-
tation of [iii] we understand Ed to be an argument of both the verbs – to have two semantic
roles. Ed is the ‘asker’ and also the (potential) ‘reader’. But we cannot similarly attribute two
distinct roles to Ed in [iv]. Ed is here the reader, but not the ‘haver’. Semantically, have relates
to Ed’s reading the report as a whole, not just to Ed. This is the distinction we have discussed
in contrasting hope and seem in §2.1, where we talked of ordinary subject vs raised subject; the
way in which we defined ‘raised subject’, however, assumed a two-clause structure and hence
the term is not appropriate in discussing the dependent-auxiliary analysis: we will simply
say here that ask takes an argument subject, whereas have takes a non-argument subject. The
distinction between the two kinds of subject is reflected in ways that will be familiar from the
earlier discussion.
(a) Ask is voice-sensitive, have voice-neutral

[13] i a. Kim asked to interview the PM.
b. The PM asked to be interviewed by Kim.

ii a. Kim had interviewed the PM.
b. The PM had been interviewed by Kim.
In [i] there is a clear contrast in meaning, with Kim doing the asking in [a], the PM in [b]: ask
is sensitive to the voice of the following infinitival. But in [ii] the propositional meaning is the
same in [a] and [b]: have is voice-neutral. The relation between [iia] and [iib] is just the same
as in the simple active–passive pair Kim interviewed the PM and The PM was interviewed by
Kim. That is what we would expect if have were simply an optional dependent of interview.
(b) Ask imposes selection restrictions on the subject, have does not
[14] i #The knife asked to touch the baby. [selection restrictions apply]
ii The knife had touched the baby. [no selection restrictions]
Example [i] is anomalous because catenative ask imposes selection restrictions on its subject:
normally it is only people who ask to do something. This indicates a direct semantic relation
between ask and the subject: verbs generally impose selection restrictions on their arguments.
Perfect have, by contrast, imposes no such restrictions. In [ii] the knife satisfies the restrictions
imposed by touch (compare #Infinity had touched the baby). This indicates a direct relation
between the subject and touch but not have: again this is what we would expect if touch were
the head element, and have a dependent.
(c) Have allows subjects with no independent meaning, ask does not
[15] i ∗There asked to be a mistake in the proof. [dummy there excluded]
ii There had been a mistake in the proof. [dummy there permitted]
Example [i] is unacceptable because there has no meaning of its own and hence cannot satisfy
the selection restrictions imposed by ask. But [ii] is acceptable because There was a mistake
in the proof is: adding have to this doesn’t affect acceptability, and this too is what one would
expect if it were a dependent of be.
Negation
With ask we find a sharp semantic and syntactic distinction between negating ask and negating
its complement. Negation of the complement occurs most naturally with passives, so we may
contrast [16ia–b], whereas with have we have the single negation [ii]:
[16] i a. She didn’t ask to be included in the survey. [negation of ask clause]
b. She asked not to be included in the survey. [negation of complement]
ii She had not been included in the survey.
The dependent-auxiliary analysis treats She had been included in the survey as a single clause
and hence predicts that there will be just one negation of it, as in [ii].
With the modals, moreover, we find that the clear semantic contrast between [16ia] and
[16ib] may be syntactically neutralised. Compare:
[17] i You may not start yet. [external negation]
ii You must not start yet. [internal negation]
As explained in Ch. 3, §9.2.1, we use the term ‘external negation’ for the interpretation where
the negation has scope over the modal, and ‘internal negation’ for that where the negation
is within the scope of the modal. Thus [17i] is comparable semantically to [16ia] in that may
(here indicating permission) falls within the scope of the negative, so that we have negation
§ 4.2.1 The dependent-auxiliary analysis 1213
of the modality: it says that you do not have permission to start yet. It differs syntactically
from [16ia] because may is an auxiliary and can hence be negated directly, without do-
support. However, [17ii] is comparable semantically to [16ib] in that must (here indicating
requirement or obligation) is outside the scope of the negative. The meaning is “You are
required not to start yet” rather than “You aren’t required to start yet”; what is negative,
therefore, is not the modality but the propositional content that the modality applies to. This
is why [17i–ii] effectively mean the same even though may and must themselves are of course
very different in meaning. In spite of the difference in the semantic scope of the negative,
however, they are syntactically alike, not just superficially in terms of the position of not, but
more fundamentally in terms of their behaviour with respect to the tests for clausal negation
set out in Ch. 9, §1.1; compare, for example:
[18] i a. She didn’t ask to be included in the survey and nor/∗so did your brother.
b. She asked not to be included in the survey and so/∗nor did your brother.
ii a. You may not start yet and nor/∗so may your brother.
b. You must not start yet and nor/∗so must your brother.
In [ia] ask is negated and hence we have matching nor following, whereas in [ib] ask is positive
and hence selects positive so following. But the parallel semantic difference between [iia] and
[iib] is not reflected in this way: [17ii] no less than [17i] is treated as syntactically negative.
This suggests that from a syntactic point of view You must start is a single clause, so that it
has only one negative counterpart.
Temporal specification
Consider the following contrast:
[19] i She asked to read the report on Saturday.
ii She had read the report on Saturday.
Example [i] is ambiguous in that on Saturday can specify the time of the asking or the time
of the (potential) reading. The ambiguity can be resolved in favour of the former meaning by
moving the adjunct: She asked on Saturday to read the report. And it can be resolved in favour
of the second meaning by changing on to next, giving She asked to read the report next Saturday,
since the future meaning of the adjunct is incompatible with the past time meaning of asked.
The point is that the asking and the reading are temporally quite distinct, and we can therefore
add temporal specification relating to either. Or, indeed, to both: She asked yesterday to read
the report on Saturday. But it would not be possible to say ∗She had yesterday read the report
on Saturday. Again, the data suggest that have and read have significantly less independence
than do ask and read, and this can be captured by assigning them to a single clause.
The verb group

Under the dependent-auxiliary analysis, sequences like had read, must start, was inter-
viewed, had been interviewed, etc., constitute single syntactic units, resulting from the
historical reanalysis of earlier catenative constructions. We will call them verb groups
(‘VGps’).26 Within the structure of the VGp, the main verb functions as head and the
auxiliaries as dependents.
26
The term ‘verb group’ is an ad hoc one: if the main verb is head and the auxiliaries dependents, it would
more properly be called a ‘verb phrase’. The ad hoc term reflects our view that the category is not theoretically
justified but may have some practical descriptive value. We prefer, therefore, to reserve the term ‘verb phrase’
for the unit which includes the complements and modifiers of the verb, in accordance with widespread usage
in modern grammars. Traditional grammar generally uses the term ‘verb’ itself for the VGp as well as for the
individual words within it, but it is undesirable to lose the distinction between the word and the larger unit.
The auxiliaries are optional elements, but their order is rigidly fixed. A simplified
structure of the VGp (ignoring do, use, and the have of I have to go) is shown in [20],
where parentheses indicate optionality:
[20] (Modal) (Perfect) (Progressive) (Passive) Main Verb
will have be be take
can write
etc. etc.
Such a structure accounts neatly for the possible combinations of auxiliary verbs and their
relative order. Note, for example, that we have She has been reading (perfect + progressive),
not ∗She is having read (progressive + perfect). The fixed order of the modals and perfect
have similarly handles such data as the following:
[21] i I may have mentioned it yesterday. [internal perfect]
ii I should have mentioned it yesterday. [external perfect]
iii Kim needn’t have written it. [ambiguous]
In [i] the past time expressed by have is associated not with the modality, but with the
proposition that the modality applies to, thus “It is possible that I mentioned it yesterday”
(cf. Ch. 3. §9.9). In [ii], by contrast, the modality – which here involves the concept of the
‘right’ thing to do – falls within the semantic scope of have, for it is a matter of what was
right in the past: “the right thing to do was (or would have been) to mention it yesterday”.
Example [iii] is ambiguous, for it can be interpreted in either way. The more salient reading
follows the pattern of [ii]: “There was no need for Kim to write it yesterday” (with need
inside the semantic scope of have); but it has a second interpretation along the lines of [i]: “It
isn’t necessarily the case that Kim wrote it” (with need now outside the scope of have). The
phenomenon is similar to that involving negative scope illustrated in [17] – where You must
not start yet is interpreted differently from You may not start yet. The development of a VGp
unit with a rigidly ordered syntactic structure results in certain semantic scope distinctions
being left implicit. The VGp selects as a whole for negative polarity and perfect tense, and the
markers of these categories occupy a fixed syntactic position within the structure irrespective
of their semantic scope.27
4.2.2 The catenative-auxiliary analysis

Although the category of VGp is in many ways descriptively very useful, the position
taken here is that there are nevertheless compelling grounds for preferring an analysis
of the modal, tense, aspectual and voice auxiliaries as catenative verbs taking non-
finite complementation. On this account, [10iii–iv], Ed asked to read the report and Ed
had read the report, have essentially the same structure. We will argue that there is no
principled basis for drawing a structural distinction between catenative + complement
and dependent-auxiliary + main-verb constructions.
27
A further case of mismatch between syntactic and semantic scope is to be found in the position of adverbs.
Although the order illustrated in He must always/never have filed the letters represents the usual one, it is also
possible to have the adverb before the auxiliary, as in He never/always must have filed the letters. Must here
has its epistemic sense, roughly “I am forced to conclude”. And it is clear that the frequency adverbs relate
semantically to the filing, not to the epistemic judgement: the meaning is “I am forced to conclude that he
always/never filed the letters”, not “I am always/never forced to conclude that he filed the letters”. Semantically,
then, it is comparable to He promised always to do his best, although the order matches that of He always
promised to do his best: the semantic contrast between the two promise examples is lost with must.
§ 4.2.2 The catenative-auxiliary analysis 1215
The argument has both a negative and a positive side to it. The negative side involves re-
considering the differences between ask and have presented in the last section as supporting
a dependent-auxiliary analysis, and showing that on closer examination they do not in fact
do so. The positive side involves introducing other phenomena which can be handled more
satisfactorily under the catenative analysis.
(a) The distinction between argument and non-argument subjects

The first point we made was that while ask takes an argument subject, have does not. But as
we in fact noted when discussing this difference, it isn’t only auxiliary verbs that take non-
argument subjects: verbs such as seem, appear, begin, tend do so too. The following examples
with seem thus match the above ones with perfect have in [13–15]:28

[22] i a. Kim seemed to intimidate the PM.
b. The PM seemed to be intimidated by Kim.
ii a. The knife seemed to touch the baby. [no selection restrictions]
b. There seemed to be a mistake in the proof. [there permitted]
It is also important to note that the concept of catenative can be applied to adjectives as
well as to verbs. Adjectives like certain, likely, eager, keen take non-finite complements, and
can be chained together recursively in the same way as catenative verbs: cf. She is likely to be
keen to accept or (with a mixture of verbs and adjectives in a chain of four catenatives) She
is likely to at least appear to be keen to try to win. And catenative adjectives show the same
split between those that take argument subjects and those that take non-argument subjects.
Compare:

[23] i a. Kim was keen to interview the PM.
b. The PM was keen to be interviewed by Kim.

ii a. Kim was likely to intimidate the PM.
b. The PM was likely to be intimidated by Kim.
Furthermore, this split is to be found in the class of auxiliaries too: it is not quite true
that all auxiliaries take non-argument subjects, for dare and the would of would rather take
argument subjects:

[24] i a. Neither dare interview the PM.
b. The PM daren’t be interviewed by either.
ii ∗The piano-lid daren’t be open. [selection restrictions apply]
iii ∗There daren’t be any dust on the piano. [dummy there excluded]
What we find, then, is the cross-classification shown in:
[25] argument subject non-argument subject
i dare, would (rather) be, can, have, may, need [auxiliaries]
ii expect, hope, try, want appear, begin, seem, tend [lexical verbs]
iii anxious, determined, keen apt, certain, liable, likely [adjectives]
The contrast in the semantic relation to the subject thus provides no basis for treating
auxiliaries differently from lexical verbs. Moreover, the behaviour of dare and would (rather)
is inconsistent with the dependent-auxiliary analysis: they are clearly heads, not dependents.
(b) Negation
Although the core auxiliary construction normally has just one negation, the possibility does
in fact exist for contrasts in syntactic as well as semantic scope and for negating more than
28
We have changed the non-finite complement of [13] because both Kim seemed to interview the PM and The
PM seemed to be interviewed by Kim sound somewhat unnatural. They are equivalent, nevertheless.
one verb at a time. Compare:

[26] i She can’t always answer his questions, can she? [negation of can]
ii She can always not answer his questions, can’t she? [negation of answer]
iii She can’t always not answer his questions, can she? [negation of both verbs]
There is an equivalence between [i] and It isn’t always possible for her to answer his questions,
is it? : the negation clearly includes the can/possible within its scope semantically, and the
interrogative tag shows that it does so syntactically too – the standard type of tag reverses
the polarity of the clause to which it attaches (cf. Ch. 9, §1.1).
By contrast, [26ii] is equivalent to It is always possible for her not to answer his ques-
tions, isn’t it? (“to refrain from answering”), where the can/possible is now outside the
scope of the negative, hence positive, as reflected in the negative tag. And [iii] is equiv-
alent to It is not possible for her to always not answer his questions, is it? The two neg-
atives here do not of course cancel each other out because they are negating different
verbs; the tag is the same as in [i], because it is determined by the polarity of the can
clause.
The always in these examples makes it easy to see the scope of the negatives. If we omit
it, there is no longer anything in the linear sequence of elements alone to determine the
scope: She + can + not + answer his questions could in principle have either the first or
second interpretation. The second requires a special prosodic reading in which there is a
break after can and the not is closely linked to answer ; in writing cannot would force the
first interpretation, whereas can not would tend to be used for the second. An inflectional
negative always indicates that the negative has syntactic scope over that verb: She can’t answer
his questions unambiguously has the first interpretation.
The same range of possibilities is even available with perfect have:
[27] i He has not always accepted bribes, has he? [negation of have]
ii He has always not accepted bribes, hasn’t he? [negation of accept]
iii He has not always not accepted bribes, has he? [negation of both verbs]
Have is negated in [i] and [iii], where has not is replaceable by inflectional hasn’t, but in
[ii] the have clause is positive (as evident from the tag), and the not belongs syntactically as
well as semantically with accept : “It has always been his practice to not accept (i.e. to refuse)
bribes”.
Examples like [26ii/27ii] are just like the earlier [16iib] (She asked not to be included in
the survey), and the catenative analysis enables them to be handled in just the same way, as
non-finite complement negation, secondary negation in the sense of Ch. 9, §2.3. This is not
possible in the dependent-auxiliary analysis, where they have to be treated as exceptions. The
above contrasts show that the proposed VGp does not in fact form a unit with a single syntactic
negation. It is true that examples like [i] represent the normal pattern, but the catenative
analysis caters more satisfactorily for the range of options that the verbal system makes
available.29 (It should also be borne in mind that in the ordinary catenative construction
negation of the matrix is much more frequent than negation of the complement.) Other
29
Attested examples of the construction shown in [iii] are rare, but the following is a slightly more complex
version of it: Not since 1992 had Sampras not taken at least one among the Australian Open, French Open,
Wimbledon and the US Open. The initial negative phrase negates the have clause, while the second not negates
the take clause.
examples of secondary negation after auxiliaries are given in:

[28] i He will sometimes not answer the phone.
ii You should just occasionally not give everyone the benefit of the doubt.
iii They are always not accepting new orders.
Temporal specification
The same kind of argument applies here. Initially we can distinguish two types of catenative:
Type i are not temporally distinct from their complement, but Type ii are. Some examples
are given in:
[29] i forget/remember (to); begin, continue, stop; manage, try [Type i: non-distinct]
ii ask, expect, intend, promise, want [Type ii: distinct]
With Type i it is not possible to give separate time specifications (of the same kind) or to give
a time specification that conflicts with the time sphere required by the inflectional tense of
the catenative, whereas with Type ii it is:
[30] i ∗This morning it began to rain this evening.
ii This morning he promised to return this evening.
On this dimension, Type ii verbs clearly involve a higher degree of internal independence
between the clauses than do Type i. Of the auxiliaries, however, only passive be and supportive
do belong clearly to Type i – compare:
[31] i At that time he was still arriving tomorrow.
ii He may have seen her yesterday.
iii We can now set out tomorrow.
iv When I arrived she had already left just a few minutes earlier.
v He had left when Kim arrived.
Progressive be belongs to Type i when expressing progressive aspectuality but not with the
futurate meaning seen here in [i]. The modals in general belong to Type ii. In [ii], the yesterday
gives the time of the seeing: it does not relate to the modality expressed by may, as evident
from the fact that the modal is in the present tense. And in [iii] there are separate time
specifications: now relates to can, tomorrow to set out. Perfect have also belongs basically to
Type ii. In [iv] just a few minutes earlier gives the time of her leaving, whereas when I arrived
is associated with the have: it gives the time of orientation to which the leaving is anterior.
Example [v] is ambiguous, in a way which parallels the earlier [19i] (She asked to read the
report on Saturday). In one interpretation when Kim arrived gives the time of his leaving. In
another it is like when I arrived in [31iv], and associated with have rather than leave: it specifies
a time prior to which he had left. (Adding already after had makes this second reading more
salient.) And in both the have and ask cases, moving the adjunct to the front just about forces
its association with the first verb:
[32] i On Saturday she asked to read the report. [specifies time of asking]
ii When Kim arrived he had left. [specifies time of orientation]
The Type ii behaviour of the auxiliaries provides strong support for the catenative analysis,
where the time adjuncts can be assigned to the superordinate or complement clause as
appropriate, just as they are when the catenative is a lexical verb.
Constituency
The dependent-auxiliary and catenative-auxiliary analyses assign different constituent struc-
tures. For He was writing a letter, say, we have (omitting functional labels):
[33] a. Clause b. Clause1
NP VP NP VP
VGp NP V Clause2
VP
he was writing a letter
V NP
he was writing a letter

The major difference is that in [a], the dependent-auxiliary analysis, writing first enters into
construction with the auxiliary was to form the VGp was writing, whereas in [ii], the catenative
analysis, it first enters into construction with its complement a letter to form the VP writing
a letter. Evidence supporting this second constituent structure is as follows:
(a) Preposing
The VP can occupy prenuclear position in the clause:
[34] She said he’d be writing a letter, and [writing a letter he was].
Here writing a letter has been preposed, and it is characteristic of fronting that it applies to
whole constituents.
(b) Coordination
Coordination provides particularly strong evidence for the catenative analysis when it occurs
in combination with negation:30
[35] i You can’t [join a debating society and not speak].
ii He surely hasn’t [read the letter and failed to see its implications].
iii I’m quite sure he didn’t [write the letter and not post it].
Can’t join and can’t join a debating society are clearly not constituents in [i] (as they are
predicted to be under the dependent-auxiliary analysis): it doesn’t say that you can’t join
a debating society, but rather that you can’t ‘join-a-debating-society-and-not-speak’. Thus
join a debating society must form a unit, which coordinates with (and )not speak, with the
resulting coordination as a whole serving as complement to can’t. The same applies to [ii],
where we find coordination within the complement of have. And similarly in [iii], where
the coordination is complement of supportive do. Such coordination would not be possible
without do (∗He wrote the letter and not post it), but introduction of do detaches the tense
inflection from write, making the way for coordination of two plain form VPs.
(c) Position of adjuncts
Adjuncts frequently occur between an auxiliary and the lexical verb: He was slowly writing
a letter. It is difficult to see how this can be accounted for under the dependent-auxiliary
analysis without invoking the concept of discontinuous constituents, whereas the catenative
30
Coordination without negation is not so straightforwardly conclusive, for we can have either He was writing a
letter and listening to the radio (favouring analysis [33b]) or He has written or is writing his letter of resignation
(which might appear to favour [33a], but is best treated as a case of delayed right constituent coordination:
see Ch. 15, §4.4).
analysis allows the construction to be handled with a simpler model of constituency. The ease
with which we can insert elements between auxiliary and ‘main verb’ casts serious doubt on
the idea that they form a syntactic constituent comparable with, say, NP, AdjP, PP, and the like.
Ordering constraints
Under the dependent-auxiliary analysis the fixed ordering of the auxiliaries is catered for by
stipulating that the VGp has the structure shown in [20]. Under the catenative analysis the
order results from certain independently needed constraints.31
(a) Initial position of the modals
This is attributable to their defective morphology: they have no secondary forms and so can’t
appear in the non-finite complement of another verb.32
(b) Final position of passive be
This is due to the fact that it is the following verb that is passivised: the preceding auxiliaries –
the modals, have, and progressive be – don’t have objects and are outside the scope of the
voice contrast. Compare now:
[36] i The king appears to be hated by his subjects.
ii The king may be hated by his subjects.
Be occupies second position among the verbs in [ii] for the same reason as in [i]: may, like
appear, is outside the scope of the passive.
(c) Fixed order of perfect and progressive
Be can’t precede have because of a more general constraint excluding perfect have from the
complements of aspectual verbs:
[37] i ∗He was having read the book.
ii ∗He began having read the book.
iii ∗He stopped having read the book.
Conclusion
The dependent-auxiliary analysis has certain attractions for textual analysis, in that it very
much reduces the amount of embedding that has to be recognised. It may have been raining,
for example, will be a simple clause instead of a complex one with three layers of embedding.
Given the high frequency of auxiliary verbs, this will result in a very considerable ‘saving’
in the analysis of most texts. It is also simpler in that we don’t have to decide which clause
temporal and similar dependents should be assigned to – e.g. whether at noon belongs with
be or with sleep in They were still sleeping at noon. And the cases where we have shown
it to give clearly unacceptable results are textually quite rare. Nevertheless, we believe we
have shown that from a more theoretical perspective the catenative analysis is very much
sounder.
It follows that in [10] above we distinguish just three categories, sequence of main clauses,
finite subordination, and non-finite subordination. Within this last category, catenative
31
In languages where mood, tense, and aspect are marked inflectionally there is some tendency for an aspect
marker to be closer to the lexical base than a tense marker, and for the latter to be closer than a mood marker;
the English syntactic order matches that morphological order.
32
The have of We have to stop and the like is not a member of the syntactic category of modal auxiliaries, and in
particular is not defective. It is therefore not catered for by structure [20] – e.g. it can precede or follow the
perfect (You have to have completed three years of undergraduate study; I have had to put it aside). Nevertheless,
in some varieties it can be used as a core auxiliary (Ch. 3, §2.5.6), and there is no evident reason why it should
then be treated differently from other core auxiliaries.
complementation, we find a considerable range of differences in the degree of internal inde-

pendence between the two clauses, but there is no principled basis for drawing a distinction
between two quite different constructions. The dependent-auxiliary analysis implies that the
auxiliaries have come, over time, to be reanalysed so as to lose their status as verbs taking
non-finite complements: what we have seen, however, is that they have retained too much
syntactic independence for this thesis to be sustainable. We believe, moreover, that the pro-
posed analysis is much more viable in a descriptive framework that recognises the catenative
complement as a distinct kind of complement – there is no question of suggesting that the
non-finite complements of the core auxiliaries are objects or predicative complements.
4.3 The traditional distinction between ‘gerunds’ and ‘present participles’

The verb-forms we are calling gerund-participles are traditionally divided into gerunds
and present participles, as illustrated in:
[38] i a. [Inviting the twins]was a bad mistake.
b. We’re thinking of [giving them one more chance].
[gerund]
c. I remember [seeing them together].
d. She found [talking to Pat]surprisingly stressful.
ii a. Those [living alone]are most at risk.
b. [Not having read his book,]I can’t comment.
c. She is [mowing the lawn]. [present participle]
d. We saw him [leaving the post office].
e. I caught them [reading my mail].
We argued in Ch. 3, §1.4, that there is no justification for making any inflectional dis-
tinction: all the underlined forms belong to a single inflectional category. We call this
form gerund-participle to reflect the fact that it covers the ground of both gerunds and
present participles in other languages. At the level of words, what is important is to
distinguish gerund-participle forms of verbs from nouns (the reading of the poem) and
adjectives (a very inviting prospect). Our aim in this section is to show that even from a
purely syntactic point of view no viable distinction can be drawn between the bracketed
clauses in [i] and those in [ii]: we refer to them all as gerund-participial clauses.
In terms of their function we distinguish between complement and non-complement
gerund-participials. Non-complement ones, we have noted, are traditionally present
participles, while the complement ones include both traditional categories.
A difference in internal form: case of the subject NP

There is one respect in which ‘gerund’ and ‘present participle’ clauses differ in their
internal form: with ‘gerunds’ the subject may take genitive case, with plain or accusative
case a less formal alternant, but with ‘present participles’ the genitive is impossible
and pronouns with a nominative–accusative contrast appear in nominative case, with
accusative an alternant restricted to informal style.33 Compare, then:
[39] i She resented his/him/∗he being invited to open the debate.
ii We appointed Max, he/him/∗his being much the best qualified of the candidates.
33
We simplify here by ignoring constructions involving coordinate pronouns: see Ch. 5, §16.2.2.
§ 4.3 Traditional distinction: ‘gerunds’ vs ‘present participles’ 1221
This difference, however, is obviously relevant only to those constructions where the
non-finite clause can contain a subject: it cannot be used to justify a distinction between
‘gerund’ and ‘present participle’ in the numerous constructions where no subject is per-
mitted. In terms of our analysis, the contrast in the case of the subject is handled by our dis-
tinction between complement and non-complement gerund-participials: genitive case is
restricted to the former, nominative to the latter. If the traditional distinction of ‘gerund’
vs ‘present participle’ is to be maintained, it must be based primarily on properties of
the subjectless construction. But here there is no difference at all in the internal form of
the constructions.
No viable distinction in function

The traditional distinction is based on the idea that ‘gerunds’ are like nouns while
‘present participles’ are like adjectives. In this respect, then, it is a reflection of the more
general practice in traditional grammar of describing subordinate clause constructions
in terms of functional analogies with the parts of speech. We have argued at length
against this practice in the context of the classification of finite clauses (Ch. 11, §8.2) and
of infinitivals (this chapter, §4.1); here we will consider the issue briefly with respect to
catenative complements.
Some of our catenative complements are included in the ‘gerund’ class (e.g. [38ic]),
while others belong with the ‘present participles’ (e.g. [38iid–e]); to make this distinc-
tion effectively requires that we analyse the former as objects and the latter as predica-
tive complements, on the basis of analogies with NPs and AdjPs respectively. But the
contrast between objects and predicative complements is no more applicable to gerund-
participial catenative complements than to infinitival ones. Note first that there are a
significant number of verbs that take adjectival predicatives but do not allow gerund-
participials:
[40] i a. They seemed resentful. b. ∗They seemed resenting it.
ii a. He became remorseful. b. ∗He became feeling remorse.
iii a. This made them hostile. b. ∗This made them wanting to harm us.
Constructions headed by ‘present participle’ verbs (as distinct from participial adjectives,
as in They seemed very interesting, etc.) are not systematically substitutable for predicative
AdjPs: their distribution is not predictable from that of AdjPs but has to be stated directly.
Similarly, the possibility or otherwise of replacing a gerund-participial by an AdjP does
not provide a plausible criterion for distinguishing between predicative complements
and objects:
[41] i a. He kept staring at them. b. He kept calm.
ii a. He went on staring at them. b. ∗He went on calm.
iii a. He stopped staring at them. b. ∗He stopped calm.
The [a] examples belong together, all having an aspectual verb (or verbal idiom in the
case of go on) in the matrix clause. The fact that keep licenses an adjectival predicative
complement while the others do not is irrelevant, as is the fact that keep and stop license
an NP object while go on does not (He kept his cool; He stopped this annoying behaviour).
The catenative construction is not syntactically close enough to constructions with
NP and AdjP complements for us to be able to extend the distinction between object
and predicative complements to the non-finites in a principled way.
Note also that there is an unsatisfactory asymmetry in the traditional treatment of

catenative complements. The object vs predicative complement distinction is applied
to both gerund-participials and infinitivals, but in the former case it provides the basis
for a distinction between ‘gerunds’ and ‘present participles’ whereas no such distinction
is drawn among the infinitivals. In neither case is there any difference in internal form
in the subjectless construction, and the motivation for drawing a distinction within
the gerund-participials but not in the infinitivals is primarily historical: ‘gerund’ and
‘present participle’ have different historical sources, whereas there is only one source
for infinitivals. In a grammar aimed at describing the present state of the language, a
syntactic distinction based on historical factors in this way cannot be justified.
No systematic difference in aspectuality

The traditional ‘present participle’ covers not only forms in modifier or supplement
function but also the form of the verb that occurs with the progressive auxiliary be. It is
evident, however, that ‘present participles’ do not always have progressive meaning, that
there is no systematic difference in aspectual meaning between ‘present participles’ and
‘gerunds’. Compare:
[42] i a. On hearing his cry, she dashed into the garden.
b. Hearing his cry, she dashed into the garden.
ii a. Despite having no TV himself, he was able to see the programme.
b. Although having no TV himself, he was able to see the programme.
In each pair, the verb in the underlined clause is traditionally analysed as a ‘gerund’ in
[a] and a ‘present participle’ in [b], but in neither case is there any difference in aspec-
tuality. In [ia/b] the aspectuality is perfective, while in [iia/b] it is imperfective but not
progressive. Note, for example, that if we substitute a finite clause for the non-finite one
in [iib], the appropriate form is the non-progressive he had no TV himself.
We conclude that there is no difference of form, function, or interpretation that correlates

systematically with the traditional distinction between ‘gerund’ and ‘present participle’.
The distinction introduces an unmotivated complication into the grammar: it is one of
the features of traditional grammar that should be discarded.
4.4 Catenative complements, adjuncts, and coordinates

In a number of cases, catenative complements bear some resemblance to various kinds
of adjunct or to elements in a coordination.
Adjunct of purpose
In general, infinitival clauses functioning as adjunct of purpose are sharply distinct from
catenative complements. Syntactically, they can be preceded by in order and character-
istically can be moved to front position:

[43] i He walked [(in order) to save money].
[purpose adjunct]
ii [(In order) to save money,] he walked.
§ 4.4 Catenative complements, adjuncts, and coordinates 1223
Semantically, they of course express the purpose of some agentive act, and we accordingly
have a very clear ambiguity in:
[44] He swore to impress his mates. [catenative complement or purpose adjunct]
The catenative interpretation is “He swore that he would impress his mates (in some
unspecified way)”, whereas with an adjunct the meaning is “He swore in order to impress
his mates (by swearing)”. In the adjunct reading, swore receives greater phonological
prominence and in writing is likely to be followed by a comma.
With a few verbs, such as wait and hurry, the semantic distinction is at times somewhat
blurred – compare:
[45] i a. She was waiting for his letter to arrive. [catenative complement]
b. She waited a while to make sure he wasn’t coming back. [purpose adjunct]
c. She was waiting to use the photocopier.
ii a. He hurried to reassure her. [catenative complement]
b. He hurried, to prove he wasn’t as slow as she claimed. [purpose adjunct]
c. He hurried to catch the train.
With wait the contrast is clear in [ia–ib]: the catenative complement identifies the event
she was waiting for and the adjunct gives the purpose of her waiting; [ic] can be con-
strued as a catenative construction, but at the same time there is an element of purpose.
Presumably she couldn’t use the photocopier immediately (someone else was using it
or it needed attention), so it was necessary for her to wait. In [ii] hurry is equivalent to
hasten in [iia], where the meaning is that he quickly reassured her, or tried to do so; [iib]
is straightforwardly purposive and readily allows in order and fronting; such fronting is
very unlikely in [iii], but it is probably best treated as also purposive, for we could not
here substitute hasten.
Go is a borderline member of the catenative category:
[46] i She went to the Old Vic to see ‘Hamlet’. [purpose adjunct]
ii She went to see ‘Hamlet’. [?catenative complement]
It is quite clear that in [i] the phrase to the Old Vic is a complement with the role of
goal and the infinitival is an adjunct of purpose. In [ii] (spoken without any prosodic
break after went) the concept of a spatial goal is very much backgrounded and it is
not implausible to regard the infinitival as having been reanalysed as a complement:
we certainly cannot insert in order and do not interpret it as answering the question
Why did she go?
There is a use of be that is restricted to the perfect and has a similar sense to go as in
Jill has been to Moscow (see Ch. 8, §4.3). This be too occurs with an infinitival, as in She
has been to see ‘Hamlet’. The infinitival here is in contrast with a goal such as to Moscow,
and must qualify as a complement. Similar is move, as in The government has moved to
allay fears of a rise in interest-rates. The sense of physical movement is lost, with move
here meaning “take action”; the infinitival is virtually obligatory, and has the character
of a complement rather than an adjunct.
Adjunct of result
To-infinitivals may be used as adjuncts to express a resultant or subsequent situation,
but with some verbs it is not easy to distinguish sharply between this construction and
the catenative one:

[47] i I ran all the way to the station only to find the train had just left.
[adjunct]
ii She opened the curtains to see that the ground was covered in snow.
iii She lived to be ninety / to regret her decision. [complement]
The first two are clearly adjuncts: they are fully optional and there is here no question
of licensing by the verb. In [iii], by contrast, the infinitival can hardly be said to be
optional, for although She lived is acceptable, the interpretation is different than in
[iii] (we would understand it as “She survived”, i.e. “She didn’t die”). Note that while
[i–ii] can be roughly paraphrased with a coordinative construction (. . . but found the
train . . . ; . . . and saw that the ground . . .), there is no comparable relation between [iii]
and She lived and was ninety or She lived and regretted her decision.
We take live therefore to be a catenative. Likewise go on, as in She went on to become
Prime Minister. Less clear are wake and grow up, as in She woke to find he’d gone, He
grew up to be a complete introvert : they commonly occur in this construction, but the
infinitival can be omitted without apparent effect on the meaning of what remains, and
there is again a close relation with a coordinative construction, so that they are best
treated with [47i–ii]. Note then that grow up differs from grow, which is undoubtedly a
catenative in He grew to like it.
Adjunct of cause
A number of verbs appear in the following range of constructions:
[48] i They rejoiced because they had won the war. [adjunct]
ii They rejoiced at their victory.
iii They rejoiced to hear they had won the war.
iv They rejoiced that they had won the war. [complement]
The because phrase in [i] is clearly an adjunct and the finite content clause in [iv] a com-
plement, while [ii–iii] fall towards the boundary between adjuncts and complements.
Other verbs of this kind are: blush, delight, grieve, grin, laugh, marvel, shudder, smile,
tremble (though only grieve and marvel of these readily enter into construction [iv]).
Depictive adjunct
Compare next:
[49] i He came in /went out looking rather pleased with himself. [adjunct]
ii It came /went hurtling through the window. [?complement]
iii She sat/lay/stood reading the newspaper. [adjunct]
The gerund-participial clause is undoubtedly an adjunct in [i], but it is somewhat closer
to a complement in [ii]: it can be freely omitted in [i], but this is not possible in [ii] without
a significant change of interpretation. In [iii] we illustrate a common construction for
§ 5 Classification of catenative verbs 1225
the positional verbs; the non-finite is not easily omissible here – but it is if we add a
locative phrase like on the patio, and is probably better regarded as an adjunct than as a
catenative complement.
Coordination
A link between catenative complementation and coordination is seen with try and go:
[50] i a. I always try and please him. b. Try and not be so impatient.
ii Go get yourself something to eat.
In [i] and is both semantically and syntactically more like a subordinator than a coor-
dinator. As explained in Ch. 15, §2.2.3, and + VP here is best treated as a non-finite
complement, a further form-type beyond those given in §1.1. Go occurs with and in
coordination, as in Go and get yourself something to eat, I went and got myself something
to eat, but under restrictive conditions it appears without and, as in [50ii]. Both verbs
must be in the plain form and as the construction is specific to go it is best treated as a
special case of bare infinitival complementation.
5 Classification of catenative verbs
In this section we present a classification of catenative verbs (together with some cate-
native idioms) based on the analysis given in §§2–4 above.
5.1 Framework of classification

Dimensions of classification
The primary dimension derives from the distinction we have drawn between the simple
and complex catenative constructions. Class 1 verbs appear only in the simple construc-
tion, Class 2 in simple or complex, and Class 3 only in the complex construction. Within
each major class, the first subdivision then concerns the form-types: to-infinitivals, bare
infinitivals, gerund-participials, and past-participials.
Multiple entries
Where a verb has different senses in different constructions, we give it multiple listings,
with subscripts distinguishing the uses. Compare, for example:
[1] i a. He intends to leave at six. b. He intends leaving at six.
ii a. He should try 1 to eat less. b. He should try2 eating less.
Intend has the same meaning in [a] as in [b]: we therefore list it once, in the class of verbs
taking either a to-infinitival or a gerund-participial. With try, on the other hand, we have
a difference of meaning: in [a] it means “endeavour”, in [b] “test the effectiveness of ”.
We therefore list try1 and try2 separately, the former in the class taking a to-infinitival,
the latter in the class taking a gerund-participial.
Passivisation
In considering whether the matrix clause can be passivised we distinguish three kinds
of passive construction, and the symbols shown on the right in [2] will be used as
annotations for verbs taking them:
[2] i a. They advised me to enrol. b. I was advised to enrol. [p]
ii a. We intended (for it) to resume. b. It was intended (for it) to resume. [ PX ]
iii a. We don’t recommend getting b. Getting involved in options trading
involved in options trading. isn’t recommended. [ PG ]
In [i], with advise as the catenative verb, the intervening NP is promoted to subject;
this is the usual case, and where ‘passivisation’ is used without qualification it is to be
understood in this sense. In [ii] an infinitival complement (of intend) is promoted to
extraposed subject, whereas in [iii] a gerund-participial complement (of recommend ) is
promoted to subject proper. With some verbs passivisation (of type [p]) is obligatory,
and with others it is blocked, and these are marked as in:
[3] i a. ∗They said him to be ill. b. He was said to be ill. [+p]
ii a. They wanted him to see it. b. ∗He was wanted to see it. [–p]
Alternation with finite complements
In many cases the non-finite complement has a finite alternant or near-alternant:
[4] i a. I believe him to be ill. b. I believe that he is ill. [ TU ]
ii a. I’d prefer you to do it yourself. b. I’d prefer that you did it yourself. [TP ]
iii a. He decided to resign. b. He decided that he would resign. [TW ]
iv a. They demanded to be heard. b. They demanded that they be heard. [TM ]
v a. She persuaded me to go. b. She persuaded me that I should go. [T S ]
The ‘t’ annotation is mnemonic for the that which occurs (usually optionally) in finite
declarative complements. We distinguish five types, as marked by the subscripts. In [i]
the finite is unmodalised (TU ), in [ii] a modal preterite (TP ), in [iii] it contains modal will
(TW ), in [iv] it is mandative (TM ), in [v] it contains modal should (TS ). As noted earlier
(§3.1.1), the constructions are not strictly alternants in [v].
Further annotations
To reduce the number of classes we add annotations to members of a class instead of
dividing it into two or more smaller classes. The annotations used are as follows:
[5] b Takes bare infinitival (I helped her mend the fuse)
f Takes to-infinitival with for (She asked for it to be postponed)
n Occurs predominantly in non-affirmatives (I don’t mind waiting a little)
ns Non-syntactic interpretation of understood subject (She said to meet at six)
p Matrix passivisation, with further specification as in [2–3]
pp Takes past-participial (He reported them killed)
t Also takes finite declaratives (with comparable sense); subtypes as shown in[4]
In addition, ‘?’ indicates that the verb’s membership of the class is questionable.
§ 5.2 Class 1 1227
Omissions
We avoid setting up one-member or very small classes for verbs which have already been
dealt with in the discussion – e.g. for call, the only verb appearing in for-complex but
not simple constructions.
5.2 Class 1: catenative verbs appearing only in the simple construction

Here we need only two dimensions of subclassification, one based on the form-type of the
non-finite, one on the semantic status of the matrix subject, whether raised or ordinary.
Class 1A: bare infinitival complement (She may know the answer)
[6] can dare 1 do had better may
must need 1 n shall will 1 would rather
This class consists of supportive do, the modal auxiliaries, and the modal idioms.34 All
take raised subjects, except for dare and would rather.
Class 1B: to-infinitival complement

1bi: ordinary subject (Kim decided to leave)
[7] affect aspire choose1 condescend consent
?
contrive dare2 (b) decide TW PX decline deign
demand TM determine TW disdain elect1 endeavour
? ? ?
forget1 get1 go on1 grow hasten
hate1 hesitate know1 learn live
look1 manage move1 offer omit
?
plot prepare1 presume1 pretend TU proceed
refuse regret1 remember1 resolve TW seek
serve stand strain strive struggle
survive swear TW think1 N threaten1 TW trouble 1
try1 undertake TW venture volunteer vow TW
In some cases it is not quite clear whether a for-complex is possible in addition to the
usual simple infinitival; thus contrive and prepare 1 are marked questionable because of
the marginal possibility of examples like ?He had contrived for his son to be admitted to
the course ; ?I hadn’t prepared for them to question me on that issue. Dare2 is a lexical verb,
while dare1 is an auxiliary: see Ch. 3, §2.5.5.
1bii: raised subject (She seemed to like it)
[8] appear TU be1 chance1 TU come1 fail
happen TU have1 have got look2 ought (b)
promise1 prove1 seem TU tend threaten2
turn out TU use
34
Also in this class are (for the varieties of English that have them) the compounds wanna, gonna, hafta, gotta,
etc., which incorporate to into the lexical base; see Ch. 18, §6.3. In the negative can occurs with an idiomatic
use of but meaning “not”: It cannot but improve, “cannot not improve – i.e. must improve”. (This is to be
distinguished from the but in conditionals: if I could but explain how I feel, “if only I could explain how I feel”.)
For ought with bare infinitival and for use (He used to like it), see Ch. 3, §§2.5.4/9 Be 1
(quasi-modal be) is used only in primary forms (There is to be another inquiry). Verbs
marked TU in 1bii take a finite clause in the impersonal construction (It appears I’m
wrong). In general the distinction between raised and ordinary subjects is clear, though
get (It had got to be quite late), go on 1 (She went on to become President of the Union), and
grow (He had grown to love her) are somewhat problematic: their meanings would lead
us to expect them to be raising verbs (e.g. grow here means essentially the same as come),
but they do not readily exhibit raising verb behaviour (? There had grown to be unanimity
between them). Promise and threaten illustrate the ‘bleaching’ (partial loss of primary
meaning) that is sometimes associated with raising verbs. In He promised 2 /threatened 1 to
tell the police we clearly have ordinary subjects, but in The weather promised 1 /threatened2 to
change the meaning of making a (characteristically verbal) promise or threat has been
lost and the meaning is reduced to approximately “look likely”, together with a favourable
or unfavourable view of the likely event.35
Class 1c: to-infinitival or gerund-participial complement

1ci: ordinary subject (I propose to tell / telling her)
?
[9] attempt bother n fear 1 neglect propose PX
scorn
1cii: raised subject (He began to shout / shouting)
[10] begin cease commence continue start1
Class 1D: gerund-participial only

1di: ordinary subject (We avoided being seen)
[11] avoid come2 complete consider1 discontinue
escape evade finish get2 go
postpone practise quit repent resist
resume try2
1dii: raised subject (Kim was writing the introduction)
[12] be 2 end up go on 2 keep1 keep on1 stop1
Most aspectual verbs have raised subjects, relating to the situation as a whole rather than
specifically to the subject-argument, and hence belong in Class 1 Dii, 1 Cii, or 1 Bii; there
are, however, a few that have ordinary subjects, normally with an agentive interpretation:
discontinue, finish, quit, resume (Class 1 Di). Compare, for example, It continued raining
and ∗It discontinued raining.
Class 1E: past-participial only (He got taken by a shark)

[13] be3 have2 get3
The three verbs in this class all have raised construal. Have is the marker of perfect tense,
auxiliary be and lexical get of passive voice.
35
In its primary meaning promise allows an intervening NP and hence belongs in Class 2ai, not 1bi, like threaten.
§ 5.3 Class 2 1229
5.3 Class 2: catenative verbs appearing in both simple and complex

constructions
We include in this class only those verbs where the intervening NP can be omitted without
a change in the meaning of the verb.36
Class 2A: to-infinitival but not gerund-participial

The simple construction has an ordinary subject, which controls the understood subject
of the non-finite, except for help and say, marked ‘NS’ for non-syntactic interpretation.
2 Ai: plain-complex, with ordinary object (She asked to see him ; She asked me to see him)
[14] ask TM (F) beg TM (F) help (B) ns pay (F) petition (F)
pledge TW pray TW (f) promise2 TW request TM train
Most verbs that take an ordinary object in the complex construction do not enter into the
simple construction (cf. They persuaded him to leave, but not ∗They persuaded to leave),
and so belong in Class 3 below. The few that do enter into both simple and complex
constructions, the present Class 2 Ai, exhibit three different relations between them:
[15] i a. Liz asked to leave. b. Liz asked Pat to leave.
ii a. Liz promised to phone at six. b. Liz promised me to phone at six.
iii a. Liz helped to clear up the mess. b. Liz helped me to clear up the mess.
In [i] the simple version [a] has control by subject, while the complex [b] has control by
object: it’s a matter of Liz leaving in [a] and of Pat leaving in [b]. In [ii] both versions have
control by subject, with Liz the one to phone. Finally, with help [b] has control by object
(I cleared up the mess, with Liz’s help), whereas [a] has a non-syntactic interpretation –
though it might also be said to have implicit control by object: we understand Liz to have
helped someone, and this someone cleared up the mess.37
36
It is for this reason that the following verbs are listed separately in Classes 1 and 3 :
Choose : They chose1 not to answer. (1bi) They chose2 Kim to lead the party. (3ai)
Consider : He considered1 resigning. (1di) He considered2 it to be a fraud. (3aii)
Dare : He wouldn’t dare2 (to) go alone. (1bi) I dare3 you to repeat that. (3ai)
Elect : He elected1 to take early retirement. (1bi) They elected2 Kim to lead the party. (3ai)
Fear : He fears1 to go out alone. (1ci) She is feared2 to have drowned. (3aii)
Get : I never got1 to speak to her. (1bi) We got4 them to move / moving. (causative, 3bi)
Have : I have1 to leave now. (1bii) He had3 them paint it black. (3bi)
Keep : He keeps1 interrupting. (1dii) They kept2 him waiting. (3cii)
Know : You know1 not to cut it that way. (1bi) He knew2 it to be impossible. (3aii)
Move : They moved1 to allay her fears. (1bi) What moved2 him to behave so aggressively? (3ai)
Prepare : He prepared1 to attack. (1bi) We’re preparing2 her to take over as head. (3ai)
Prove : It proved1 to be impossible. (1bii) She proved2 it to be impossible. (3aii)
Stop : It has stopped1 raining. (1dii) They stopped2 me taking part. (3ci)
Think : I didn’t think1 to check his credentials. (1bi) He was thought2 to be trustworthy. (3aii)
Trouble : He didn’t trouble1 to close the door. (1bi) May I trouble2 you to close the door? (3ai)
Will : It will1 be over soon. (1a) You can’t will2 her to do it. (3ai)
37
This kind of interpretation does not hold for all instances of help. The problem is that the helper and helpee may
be involved in the subordinate situation in a range of different ways. In [15iiib], Liz and I cleared up together,
and in [a] Liz and some unspecified person(s) did so. In The commotion helped me to escape unnoticed, on the
other hand, it was only me who escaped, the commotion having merely a supportive role. And at the other
extreme, consider The eyebrows help to keep sweat out of the eyes. The eyebrows have the primary role in keeping
the sweat out, so much so that in this case we don’t really reconstruct an understood helpee at all.
No other verbs follow the patterns of help and promise, and indeed [iib] is rather
marginal: many speakers find this unacceptable and it would be much more usual to use
a finite complement here (Liz promised me that she would phone at six). The case with
ask is complicated by the fact that, under very restrictive conditions, control by subject
applies (again somewhat marginally) in the complex construction. Compare:
[16] i Liz asked Pat to be allowed to leave. [control by matrix subject]
ii Liz asked Pat to be photographed with the children. [control by matrix object]
In [i] we understand that Liz asked for permission to leave, but it is only complements
like to be allowed and synonyms that permit matrix subject control in this way. In [ii],
for example, we have a passive infinitival, but it still takes control by object, like the
active [15ib]. Beg, pray, petition, and perhaps request exhibit the same behaviour as ask
(begged Pat to be allowed to leave has matrix subject control, begged Pat to be photographed
does not). Pay is similar but takes control by subject in the complex construction under
somewhat different conditions – compare:
[17] i They paid her $100 to dance naked. [control by object]
ii They paid her $100 to see her dance naked. [control by subject]
Example [ii] is one of those where the non-finite clause falls at the boundary between
catenative complement and purpose adjunct (see §4.3). The complement expressing the
price paid is of course irrelevant to the issue of control: They paid ($100)to see it. Pledge
and train follow the pattern of [15i] with no possibility of control by subject in the
complex construction:38
[18] a. They have pledged to end the fighting. b. She pledged herself to support us.
Ask and the others marked ‘(F)’ also allow for – compare:
[19] a. He asked Pat to be interviewed. b. He asked for Pat to be interviewed.
In [a] Pat represents the goal of ask as well as the patient of interview, but in [b] only
the latter. Note here the contrast between ask and a raising verb like intend: He intended
Jill to be interviewed and He intended for Jill to be interviewed are equivalent, whereas
[19a–b] differ sharply in meaning.
2 Aii: plain-complex with raised object (I expect to finish soon; I expect you to finish soon)
[20] claim TU desire TM (F) expect TW mean1 TM (F) profess TU
reckon wish –P (F)
2 Aiii: for-complex (He longed to return home ; He longed for her to return home)
?
[21] ache agree PX aim PX apply arrange PX TW
be dying burn burst can afford N care
clamour hope PX TW itch long opt
pine say1 TS NS wait yearn
Most of these also take complements of the form for + NP: compare She longed for him
to be dismissed and She longed for his dismissal. The exceptions are agree (They agreed for
it to be postponed ∼ They agreed to/on/∗for a postponement), can afford (I can’t afford for
38
With pledge the complex commonly has a reflexive object, but not invariably: The treaty pledges the Sultan to
co-operate with a democratically elected government.
§ 5.3 Class 2 1231
it to be postponed ∼ I can’t afford a postponement / ∗for a postponement), and informal

say (She said for you to come at six – no comparable construction with NP).
2 Aiv: oblique-complex (He signalled to stand up ; He signalled to us to stand up)
[22] motion signal
These allow the prepositional complement to be omitted but recoverable from the con-
text, with the understood oblique controlling the missing subject of the infinitival. In
He signalled to stand up we understand that he signalled to some person or persons that
they should stand up. The simple catenative construction is to be distinguished from the
non-catenative construction with a purpose adjunct, as in He signalled to show us he was
wounded.
Class 2B: to -infinitival or gerund-participial

Except where otherwise indicated, the simple construction has an ordinary subject and
the plain-complex has a raised object.
2 Bi: both form-types can be simple or complex; genitive allowed (I’d hate to see it ; I’d
hate you to see it ; I hate wasting time; I hate his /him wasting time)
[23] can bear N can stand N hate2 like PP loathe
love prefer T
The genitive version of the complex gerund-participial (He didn’t like my interrupting
him) is generally restricted to formal style, and even there many would use only the
non-genitive. None of these verbs allows passivisation: ∗He’d be hated to see it. Most verbs
of liking and not liking belong in this class, but detest and for the most part dislike are
restricted to the gerund-participial, and hence belong in Class 2 C. For the past-participial
with like, see §5.6.2.
2 Bii: both form-types can be simple or complex; no genitive (I need to read it ; I need
you to read it; My hair needs cutting ; I need my hair cutting)
[24] deserve F? need2 require want1
The gerund-participials here are concealed passives. There can be no passivisation of the
matrix, except with require in the to-infinitival: The form is required to be returned by 1
May. The complex to-infinitival with deserve is somewhat marginal: ? He didn’t deserve
for his request to be turned down.
2 Biii: to-infinitival simple or complex, gerund-participial restricted to one or the
other (I intend to tell /telling her ; I intended him to hear me)
[25] intend PX TM (F) plan PX TW (F) want2 –P PP
With the first two the gerund-participial is permitted only in the simple construction,
whereas with want 2 it is found only in the complex (for It wants washing is a concealed
passive with want 1 “need”):
[26] i a. I intended to read /reading it. [simple]
b. I intended you to read / ∗reading it. [complex]
ii a. I want them standing when the Minister enters.
b. I don’t want you bringing your dog with you.
The gerund-participial with want 2 generally has a progressive interpretation, but in

non-affirmative contexts it can be non-progressive. Example [iia] is equivalent to I want
them to be standing when the Minister enters, contrasting with non-progressive I want
them to stand when the Minister enters. But in [iib] the meaning is “to bring”, not “to be
bringing”.
In the complex to-infinitival intend takes a raised object, plan an ordinary one:
[27] i I intended there to be more time for discussion.
ii We planned the seminar to coincide with her visit.
2 Biv: to-infinitival plain-complex, gerund-participial simple (He admits it to have been
a mistake; He admits breaking it)
[28] ackowledge TU admit TU confess TU deny TU
The simple construction has an ordinary subject, the complex one a raised object. These
verbs belong semantically with the verbs of cognition/saying in Class 3 Aii, but differ
syntactically in allowing a simple construction with a gerund-participial.
2 Bv : to-infinitival plain-complex, gerund-participial simple or complex (I remember
him to be irascible; I remember telling you; I remember his/him telling you)
[29] advise NS TM encourage NS forget2 TU –P recollect TU –P
recommend NS TM remember2 TU report TU PP
Both genitive and non-genitive forms are permitted in the complex gerund-participial,
with the genitive as usual more formal. In the simple construction, advise, encourage, and
recommend have a non-syntactic interpretation of the missing subject, while the others
have subject control. Compare :
[30] i I wouldn’t recommend buying it. [potential buyer unspecified]
ii I remember buying it. [buyer = speaker]
2C: gerund-participial only
All verbs in this class allow a genitive subject in the gerund-participial. The non-finite
complements here are relatively close to objects: except where marked ‘–PG ’ they can
marginally be promoted to subject through passivisation (whereas the intervening NP
never can). There are two subclasses to distinguish.
2 Ci: simple construction has control by subject (You risk being arrested; I won’t risk
them/their seeing us together)
[31] abhor anticipate appreciate begrudge can help N39
celebrate chance2 contemplate countenance defer
?
delay describe detest discuss dislike
?
dread endure enjoy envisage fancy1
foresee imagine1 mention mind N –PG miss
put off recall regret2 relish resent
risk tolerate welcome
39
Non-affirmative can help also occurs with but + bare infinitival: compare I couldn’t help overhearing /
couldn’t help but overhear what you were saying to Jill.
§ 5.4 Class 3 1233
2 Cii: simple construction has non-syntactic interpretation of the missing subject

(It will mean getting up earlier ; It will mean you/your getting up earlier)
[32] advocate deplore PG deprecate PG discourage PG facilitate
fancy2 N include involve –PG justify mean2 –PG
necessitate oppose save –PG suggest support
understand1
5.4 Class 3: catenative verbs appearing only in the complex construction

Class 3A: infinitival but not gerund-participial
3 Ai: plain-complex with ordinary object (She urged me to go)
[33] accustom aid appoint assist authorise
back badger beckon beseech blackmail
bribe bring bring up caution challenge
choose2 coax command commission compel
condemn constrain dare3 defy design
direct discipline drive elect2 empower
entice entitle entreat equip exhort
fit forbid force implore incite
induce inspire instruct invite lead
leave1 make out move2 nag nominate
oblige persuade TS pester prepare2 press (F)
pressure program(me) prod prompt provoke
push (F) remind school second select
sentence spur on stimulate stir summon
teach tell TS tempt thank trouble2 –P
trust urge warn TS will2
This class includes certain verbs of causation such as compel and force which we discuss
with some semantically related verbs under Class 3 B.
3 Aii: plain-complex with raised object (I assumed there to be a mistake in the
instructions)
[34] i accept affirm allege announce argue
ascertain assert assume attest believe
certify concede conceive conclude conjecture
consider2 declare deduce deem demonstrate
discern disclose discover1 establish estimate
fear2 PP find1 gather grant guarantee
guess hold imagine2 intuit judge
know2 (B) TU note presume2 presuppose proclaim
pronounce prove2 recognise represent repute +P
reveal rule rumour +P say2 +P show1
state stipulate suppose surmise suspect
take think2 tip understand2 verify
?
ii allow cause enable let –P B make B
?
order TM PP permit
The ‘B’ annotation for make applies only in the active: in the passive it takes a to-
infinitival:
[35] i They made us feel guilty. [active + bare infinitival]
ii We were made to feel guilty. [passive + to-infinitival]
This class contains a rather large number of verbs of cognition or saying, illustrated in
[34i], and a handful of verbs of permission, ordering, and causation, listed in [34ii]. The
verbs in [34i], except for informal tip, also occur with an unmodalised finite complement:
I assumed that there was a mistake in the instructions.40 All allow perfect infinitivals
(I assumed there to have been a mistake) and, in the absence of perfect marking, the time
of the subordinate situation is the same as that of the matrix. This means that we cannot
have, say, ∗I believe her to win tomorrow’s semi-final corresponding to finite I believe
that she will win tomorrow’s semi-final. In the non-perfect the infinitival normally has a
stative interpretation: be is especially common. Matrix passives are always possible and
tend to be more frequent than actives; with say and others marked ‘+P’ passivisation is
obligatory: He is said to be dying (not ∗They say him to be dying).
Consider now the verbs of permission and ordering in [34ii] (we take up the causatives
in the discussion of Class 3 Bi). Allow appears in examples like:
[36] i Will you allow me to audit your course?
ii We mustn’t allow there to be any repetition of this behaviour.
iii The weather didn’t allow us to finish the game.
Example [i] illustrates a conventional way of requesting permission, which might sug-
gest an interpretation where me is an argument of allow, with the role of recipient of
permission. In [ii], however, allow is clearly behaving like our model raising verb intend:
there is certainly no question here of giving someone permission to do something. This
is even clearer in [iii], where we have an inanimate subject. Allow has a much more
general meaning than “give permission”, the core being something like “not prevent,
make possible, enable”, and it would be difficult to maintain that the construction was
ambiguous in such a way that in He allowed Kim to take all the credit for this achieve-
ment, say, the object is raised (it’s fair to assume that Kim didn’t seek permission to
take the credit) while in He allowed Kim to audit my course we have an ordinary ob-
ject (with Kim seeking and receiving permission). For even where giving permission
is apparently involved we still find equivalence between active and passive infinitival
constructions:

[37] i He allowed the postgraduates to audit the course.
[voice-neutral]
ii He allowed the course to be audited by the postgraduates.
The recipient of permission can be encoded with give permission, but even here it need
not be, so that we have a contrast between He gave them permission to audit the course
(recipient encoded) and He gave permission for them to audit the course (recipient not
encoded). This contrast is not available with allow, which is best treated as a raising
verb in all cases; it does not encode the giving of permission to anyone, though it
40
It has been suggested that the finite and non-finite constructions are not entirely equivalent, that I believe
him to be telling the truth, for example, is more acceptable than I believe him to be lying (by reason of being
compatible with I believe him), but we are sceptical about the validity of a distinction along these lines.
§ 5.4 Class 3 1235
may, in favourable circumstances, trigger an implicature that permission is given to the

object-referent. Permit and let behave in essentially the same way. So indeed do the
modals can and may in their deontic sense, except that here only one argument is
encoded: the raised complement is thus subject, not object. You can come in now may
be interpreted in context as “I give you permission to come in now”, but the giver and
receiver of permission are not encoded as arguments of can.
Order has a meaning comparable in specificity to give permission rather than al-
low : it involves issuing some (normally verbal) order. But the object clearly doesn’t
encode the recipient of the order in examples like He ordered the documents to be de-
stroyed: this has a raised object and is equivalent to the finite construction He ordered
that the documents be destroyed. Nevertheless, order differs from allow in that the cases
where the object clearly does not represent the recipient of the order predominantly
have passive infinitivals.41 Thus whereas we can say He ordered that the data be freely
available to all interested parties, the infinitival #He ordered the data to be available to
all interested parties is anomalous (since it requires us to take the data as recipient of
the order).
With a human object and active infinitival there is a strong implicature that the object
represents the recipient, so that He ordered Kim to unlock the safe, say, is not equivalent
to He ordered that Kim unlock the safe. Command is more straightforward in that it does
not normally allow a raised object with a passive: #He commanded the documents to be
destroyed.
3 Aiii: oblique-complex (Kim appealed to them to release the hostages)
[38] appeal [to] bank [on] count [on] depend [on] keep on2 [at] rely [on]
These take a catenative complement only in combination with a prepositional comple-
ment; in addition call uniquely occurs either in this construction or in the for-complex:
cf. [9iia–b] of §3 (He called for Ed to be sacked ; He called on Ed to resign).
Class 3 B: infinitival, gerund-participial, or past-participial
Verbs in this class all take non-genitives in the gerund-participial. We distinguish two
subclasses on the basis of matrix passivisation.
3 Bi: no matrix passivisation (I got them to talk /talking; I got my car repaired )
[39] get4 have3 B
These belong semantically with the verbs of causation, but occur in a wider range of
constructions than the others, which are found in classes 3 Ai and 3 Aii: we consider them
all together here. The distinction between ordinary and raised complements is seen
in:
[40] i They forced/compelled Kim to unlock the safe. [ordinary object]
ii This caused both of us to overlook the inconsistency. [raised object]
Force and compel (3 Ai) impose selectional restrictions on the object and assign an agen-
tive role to the covert subject of the infinitival: in [i] force/compulsion is applied directly
41
Not exclusively, however: cf. France has ordered nuclear testing to resume. Occasional examples of this kind are
also found with instruct: I instructed prison routine to continue as normal.
to Kim and the verbs clearly belong with our model verb persuade. Cause and enable, on
the other hand, belong equally clearly with the raising verbs (3 Aii): note, for example,
the equivalence between [ii] and This caused the inconsistency to be overlooked by both
of us; the second argument in [ii] is thus the whole event of our overlooking the incon-
sistency, and there is no direct relation between the matrix verb and its object. Make, get,
and have are less clear-cut. Make could substitute with little effect on the meaning both
for force/compel in [i] and for cause in [ii]. It does not readily take a passive infinitival,
but ? Pat made both candidates be interviewed by Kim seems to differ in propositional
meaning as well as acceptability from Pat made Kim interview both candidates, suggest-
ing that it takes an ordinary object, like force/compel. On the other hand, it can take
a dummy object, certainly extrapositional it, as in He made it appear that he had been
acting under duress, where it is difficult to see any direct relation between verb and object.
Make thus seems to allow either control by object or raised construal, with a good deal
of indeterminacy between the two. Get allows a wider range of objects than force/compel
(cf. He finally got the car to start, where the car’s role is non-agentive), but it cannot
take dummy there as object (cf. ∗He finally got there to be a reconciliation). As for voice,
we have:
[41] i He got a specialist to examine his son.
ii He got his son to be examined by a specialist.
iii He got his son examined by a specialist.
Examples [i] and [ii] are not equivalent and get here clearly takes an ordinary ob-
ject (as also with a gerund-participial). But [iii] is not an alternant of [ii]: get takes
a raised object in the past-participial construction (which has no active counterpart).
With have the analogue of [ii] is not acceptable (∗He had his son be examined by a spe-
cialist), and we have equivalence between He had a specialist examine his son and He
had his son examined by a specialist ; this indicates a raised object, which ties in with
the fact that have is also used with a non-causative “undergo” sense: He had the police
call round in the middle of the night to question him about his secretary’s disappear-
ance, where the visit was something that happened to him rather than something he
arranged – and where there would seem to be no direct semantic relation between verb
and object.
3 Bii: matrix passive allowed (I heard them arrive/arriving ; I heard the window broken)
[42] feel TU (B) hear TU (B) notice TU B observe TU (B) overhear (B)
see1 TU (B) watch B
These verbs, together with smell from Class 3 Cii below, are the verbs of sensory per-
ception. All take bare infinitivals, while those where the ‘B’ is parenthesised take a to-
infinitival as well. For see we have the following possibilities:
[43] i a. We saw Kim leave the bank. b. ∗Kim was seen leave the bank.
ii a. We saw Kim leaving the bank. b. Kim was seen leaving the bank.
iii a. We saw Spurs beaten by United. b. ? Spurs were seen beaten by United.
iv a. We saw him to be an impostor. b. He was seen to be an impostor.
We put the to-infinitival last because this does not represent the primary sense: it is not
a matter of sensory perception but of mental inference. In this construction, see behaves
§ 5.4 Class 3 1237
like the verbs of cognition/saying (Class 3 Aii), following their pattern of favouring a
matrix passive and the verb be in the subordinate clause, of allowing the perfect (He was
seen to have altered the figures), and of alternating with the finite construction (We saw
that he was an imposter).
The primary sense, illustrated in [43i–iii], involves two arguments, an experiencer and
a stimulus (the situation perceived): Kim in [i–ii] thus does not represent an argument of
see. We demonstrated this for the gerund-participial construction in §3.2.2, but it holds
for the other form-types too; this is why there is equivalence between [iiia] and We saw
United beat Spurs.
The gerund-participial in [43ii] has progressive meaning: in [i] we saw the whole
event of Kim’s leaving the bank, in [ii] a segment of it – the contrast is the same as that
between Kim left the bank and Kim was leaving the bank. The progressive auxiliary be
cannot be used (∗We saw Kim be leaving the bank), and passive be is likewise omitted to
give [iiia] instead of ∗We saw Spurs be beaten by United.42
The bare infinitival does not allow matrix passivisation, as is evident from [43ib]. The
to-infinitival, however, has a wider range of use in the passive than in the active:
[44] a. ∗We saw Kim to leave the bank. b. Kim was seen to leave the bank.
It is therefore tempting to see [44b] as filling the gap created by the ungrammaticality
of [43ib] (parallel to the case with make: We made Kim leave the bank ; ∗Kim was made
leave the bank; Kim was made to leave the bank). Yet it is doubtful if the sense is quite the
same: [44b] has at least a trace of the cognitive component of meaning noted above for
[43iv]. Compare, for example:
[45] i They had seen him drive, so everyone decided to go by bus.
ii He had been seen to drive, so everyone decided to go by bus.
Notice that [i] is perfectly coherent, but [ii] is not. In [i] they had perceived the event,
and hence the manner of his driving, and we infer that it was the latter that made them
decide to go by bus. But in [ii] it is the fact of his driving that had been registered, and
this doesn’t provide an obvious reason for them to go by bus.
None of the other sense verbs shows quite the same range as see. The closest is feel but
construction [43iii] is here virtually restricted to reflexives or body parts (I felt myself /
my leg grabbed from behind ). With hear and overhear [iv] is virtually excluded in the
active (∗We’d heard him to be an impostor) and in the passive we have again the problem
of distinguishing between the senses of [i] and [iv]: we do not have ∗He was heard to be an
impostor (where see would be quite normal), but only examples like He was heard to lock
the door, which is very close to They heard him lock the door.43 Watch wholly excludes
the to-infinitival, whether active or passive. Notice and observe are hardly possible in
[iii], and notice is also marginal in [iv]. Smell is generally restricted to [ii], and hence is
listed in Class 3 Cii below; it combines predominantly with burn (I can smell something
burning).
42
Get, however, is perfectly possible: We saw Spurs get annihilated by United.
43
With a finite complement hear characteristically involves hearing something said; it also occurs in the fixed
expressions hear say/tell, a simple catenative construction with contextual construal.
Class 3 C: gerund-participial only

3 Ci: genitive possible (They prevented us/our speaking to her)
[46] excuse –P forgive –P pardon –P preclude –P prevent
?
prohibit –P stop2
These verbs have alternants containing prepositions: for in the case of excuse, forgive,
pardon, and from with preclude, prevent, prohibit, stop :
[47] i a. Forgive me/my saying so, but . . . b. Forgive me for saying so, but . . .
ii a. He prevented us/our seeing her. b. He prevented us from seeing her.
Passivisation is restricted to the prepositional construction (He was never forgiven for
abandoning his children), except very marginally with prevent and stop (?She was pre-
vented/stopped writing to us). Stop is a questionable member of this class because it does
not generally allow a genitive in the [a] construction (∗He stopped our seeing her); more-
over it has a wider range of meaning in [a] than in [b]. In the [a] construction we find
the two meanings illustrated in:
[48] i We must stop him coming back tomorrow. [“not allow, prevent”]
ii They stopped us playing before we had finished the first set. [“made us stop”]
Construction [47iib], with from, generally yields the “prevent” meaning; from can thus
be inserted in [48i] but not in [48ii]. In the second sense stop belongs with keep2 and
start 2 below (3 Cii) as a causative of an aspectual verb.
In [47i] it is a matter of forgiving the offence in [a], the offender in [b], but the
difference is negligible. In [47ii] the [a] version involves preventing an event, while [b]
is more problematic. The [b] structure suggests that us is an argument of prevent, but
some speakers allow dummy there here, % We must prevent there from being any repetition
of this error. The semantically empty there cannot be an argument and hence must be a
raised object, but the construction is unique in that the raised object precedes a PP rather
than a non-finite VP; it is probably to be accounted for in terms of a blend between the
[a] and [b] constructions.
3 Cii: no genitive (I caught them/∗their smoking)
[49] catch discover2 depict envy –P find2
keep2 leave2 picture portray see2 –P
set show2 smell start2 –P
There are several subgroups here. Catch, discover, and find clearly take ordinary objects,
as do depict, picture, portray, and show ; with all of these the gerund-participial has
a progressive interpretation. Leave is more problematic: I left them quarrelling among
themselves entails that I left them, suggesting an ordinary object; but The quarrel left me
feeling insecure does not entail that the quarrel left me: the meaning here is approximately
“result in”, suggesting a raised object. Keep, set, and start have causative meanings; they
allow a quite limited range of non-finites, especially the last two: they do not accept
passives (∗He kept them being washed; ∗We started them being vaccinated), and generally
require dynamic situations (It started them talking / ∗liking her). See 2 (which usually
occurs with can) is voice-neutral (I can’t see Kim beating Pat = I can’t see Pat being beaten
by Kim), but differs from see 1 in not allowing passivisation: it is like imagine 1 (2 Ci) and
fancy 2 (2 Cii) except that it is restricted to complex constructions.
§ 5.5 Index to the classification 1239
5.5 Index to the classification

The following index includes all the verbs and verbal idioms mentioned in §§5.2–5.4,
showing the class to which they belong.
abhor 2 Ci accept 3 Aii accustom 3 Ai ache 2 Aiii acknowledge 2 Biv
admit 2 Biv advise 2 Bv advocate 2 Cii affect 1 Bi affirm 3 Aii
agree 2 Aiii aid 3 Ai aim 2 Aiii allege 3 Aii allow 3 Aii
announce 3 Aii anticipate 2 Ci appeal 3 Aiii appear 1 Bii apply 2 Aiii
appoint 3 Ai appreciate 2 Ci argue 3 Aii arrange 2 Aiii ascertain 3 Aii
ask 2 Ai aspire 1 Bi assert 3 Aii assist 3 Ai assume 3 Aii
attempt 1 Ci attest 3 Aii authorise 3 Ai avoid 1 Di back 3 Ai
badger 3 Ai bank 3 Aiii be 1 1 Bii be 2 1 Dii be 3 1 E
be dying 2 Aiii bear 2 Bi beckon 3 Ai beg 2 Ai begin 1 Cii
begrudge 2 Ci believe 3 Aii beseech 3 Ai blackmail 3 Ai bother 1 Ci
bribe 3 Ai bring 3 Ai bring up 3 Ai burn 2 Aiii burst 2 Aiii
can 1 A can afford 2 Aiii can bear 2 Bi can help 2 Ci can stand 2 Bi
care 2 Aiii catch 3 Cii cause 3 Aii caution 3 Ai cease 1 Cii
celebrate 2 Ci certify 3 Aii challenge 3 Ai chance1 1 Bii chance2 2 Ci
choose1 1 Bi choose2 3 Ai claim 2 Aii clamour 2 Aiii coax 3 Ai
come1 1 Bii come2 1 Di command 3 Ai commence 1 Cii commission 3 Ai
compel 3 Ai complete 1 Di concede 3 Aii conceive 3 Aii conclude 3 Aii
condemn 3 Ai condescend 1 Bi confess 2 Biv conjecture 3 Aii consent 1 Bi
consider1 1 Di consider2 3 Aii constrain 3 Ai contemplate 2 Ci continue 1 Cii
contrive 1 Bi count 3 Aiii countenance 2 Ci dare1 1 A dare2 1 Bi
dare3 3 Ai decide 1 Bi declare 3 Aii decline 1 Bi deduce 3 Aii
deem 3 Aii defer 2 Ci defy 3 Ai deign 1 Bi delay 2 Ci
demand 1 Bi demonstrate 3 Aii deny 2 Biv depend 3 Aiii depict 3 Cii
deplore 2 Cii deprecate 2 Cii describe 2 Ci deserve 2 Bii design 3 Ai
desire 2 Aii determine 1 Bi detest 2 Ci direct 3 Ai discern 3 Aii
discipline 3 Ai disclose 3 Aii discontinue 1 Di discourage 2 Cii discover1 3 Aii
discover2 3 Cii discuss 2 Ci disdain 1 Bi dislike 2 Ci do 1 A
dread 2 Ci drive 3 Ai elect 1 1 Bi elect 2 3 Ai empower 3 Ai
enable 3 Aii encourage 2 Bv end up 1 Dii endeavour 1 Bi endure 2 Ci
enjoy 2 Ci entice 3 Ai entitle 3 Ai entreat 3 Ai envisage 2 Ci
envy 3 Cii equip 3 Ai escape 1 Di establish 3 Aii estimate 3 Aii
evade 1 Di excuse 3 Ci exhort 3 Ai expect 2 Aii facilitate 2 Cii
fail 1 Bii fancy1 2 Ci fancy2 2 Cii fear1 1 Ci fear2 3 Aii
feel 3 Bii find1 3 Aii find2 3 Cii finish 1 Di fit 3 Ai
forbid 3 Ai force 3 Ai foresee 2 Ci forget1 1 Bi forget2 2 Bv
forgive 3 Ci gather 3 Aii get1 1 Bi get2 1 Di get3 1 E
get4 3 Bi go 1 Di go on1 1 Bi go on2 1 Dii grant 3 Aii
grow 1 Bi guarantee 3 Aii guess 3 Aii had better 1 A happen 1 Bii
hasten 1 Bi hate1 1 Bi hate2 2 Bi have1 1 Bii have2 1 E
have3 3 Bi have got 1 Bii hear 3 Bii help 2 Ai hesitate 1 Bi
hold 3 Aii hope 2 Aiii imagine 1 2 Ci imagine 2 3 Aii implore 3 Ai
incite 3 Ai include 2 Cii induce 3 Ai inspire 3 Ai instruct 3 Ai
intend 2 Biii intuit 3 Aii invite 3 Ai involve 2 Cii itch 2 Aiii
judge 3 Aii justify 2 Cii keep1 1 Dii keep2 3 Cii keep on1 1 Dii
keep on2 3 Aiii know 1 1 Bi know2 3 Aii lead 3 Ai learn 1 Bi
leave1 3 Ai leave2 3 Cii let 3 Aii like 2 Bi live 1 Bi
loathe 2 Bi long 2 Aiii look1 1 Bi look2 1 Bii love 2 Bi
make 3 Aii make out 3 Ai manage 1 Bi may 1 A mean1 2 Aii
mean2 2 Cii mention 2 Ci mind 2 Ci miss 2 Ci motion 2 Aiv
move1 1 Bi move2 3 Ai must 1 A nag 3 Ai necessitate 2 Cii
need1 1 A need2 2 Bii neglect 1 Ci nominate 3 Ai note 3 Aii
notice 3 Bii oblige 3 Ai observe 3 Bii offer 1 Bi omit 1 Bi
oppose 2 Cii opt 2 Aiii order 3 Aii ought 1 Bii overhear 3 Bii
pardon 3 Ci pay 2 Ai permit 3 Aii persuade 3 Ai pester 3 Ai
petition 2 Ai picture 3 Cii pine 2 Aiii plan 2 Biii pledge 2 Ai
plot 1 Bi portray 3 Cii postpone 1 Di practise 1 Di pray 2 Ai
preclude 3 Ci prefer 2 Bi prepare1 1 Bi prepare2 3 Ai press 3 Ai
pressure 3 Ai presume1 1 Bi presume2 3 Aii presuppose 3 Aii pretend 1 Bi
prevent 3 Ci proceed 1 Bi proclaim 3 Aii prod 3 Ai profess 2 Aii
programme 3 Ai prohibit 3 Ci promise1 1 Bii promise2 2 Ai prompt 3 Ai
pronounce 3 Aii propose 1 Ci prove1 1 Bii prove2 3 Aii provoke 3 Ai
push 3 Ai put off 2 Ci quit 1 Di recall 2 Ci reckon 2 Aii
recognise 3 Aii recollect 2 Bv recommend 2 Bv refuse 1 Bi regret 1 1 Bi
regret2 2 Ci relish 2 Ci rely 3 Aiii remember1 1 Bi remember2 2 Bv
remind 3 Ai repent 1 Di report 2 Bv represent 3 Aii repute 3 Aii
request 2 Ai require 2 Bii resent 2 Ci resist 1 Di resolve 1 Bi
resume 1 Di reveal 3 Aii risk 2 Ci rule 3 Aii rumour 3 Aii
save 2 Cii say1 2 Aiii say2 3 Aii school 3 Ai scorn 1 Ci
second 3 Ai see 1 3 Bii see 2 3 Cii seek 1 Bi seem 1 Bii
select 3 Ai sentence 3 Ai serve 1 Bi set 3 Cii shall 1 A
show 1 3 Aii show 2 3 Cii signal 2 Aiv smell 3 Cii spur on 3 Ai
stand 1 Bi start 1 1 Cii start 2 3 Cii state 3 Aii stimulate 3 Ai
stipulate 3 Aii stir 3 Ai stop 1 1 Dii stop 2 3 Ci strain 1 Bi
strive 1 Bi struggle 1 Bi suggest 2 Cii summon 3 Ai support 2 Cii
suppose 3 Aii surmise 3 Aii survive 1 Bi suspect 3 Aii swear 1 Bi
take 3 Aii teach 3 Ai tell 3 Ai tempt 3 Ai tend 1 Bii
thank 3 Ai think 1 1 Bi think 2 3 Aii threaten 1 1 Bi threaten 2 1 Bii
tip 3 Aii tolerate 2 Ci train 2 Ai trouble 1 1 Bi trouble 2 3 Ai
trust 3 Ai try 1 1 Bi try 2 1 Di turn out 1 Bii understand 1 2 Cii
understand 2 3 Aii undertake 1 Bi urge 3 Ai use 1 Bii venture 1 Bi
verify 3 Aii volunteer 1 Bi vow 1 Bi wait 2 Aiii want 1 2 Bii
want 2 2 Biii warn 3 Ai watch 3 Bii welcome 2 Ci will 1 1 A
will 2 3 Ai wish 2 Aii would rather 1 A yearn 2 Aiii
5.6 Further remarks on the form-types

5.6.1 To-infinitival vs gerund-participial
Which form-type(s) a particular catenative selects must be specified lexically for that
verb: we cannot assign distinct meanings to the form-types and treat the selection as
semantically determined. On the other hand, the selection is not random: verbs with
§ 5.6.1 To-infinitival vs gerund-participial 1241
similar meanings tend to select the same form-types, and where a verb allows both major
form-types we very often find a difference in meaning that is at least partly motivated
by their general characteristics.
We have noted (§1.4) that infinitival to derives historically from the preposition to
and that while it has quite clearly undergone a syntactic change such that in this use it
is no longer a preposition, certain aspects of its infinitival subordinator use reflect its
origin. Prepositional to is characteristically associated with a goal, and a metaphorical
association between to-infinitivals and goals is to be found in the fact that they commonly
involve temporal projection into the future, as with the complements of ask, choose,
consent, hesitate, order, persuade, promise, resolve, strive, tell, threaten, and countless other
catenatives. Linked with this is the modal feature of potentiality. The gerund-participial,
by contrast, is commonly associated with what is current and actual, as in They enjoy
walking, She finished working, He practised speaking with an American accent, and it is
plausible to see this as connected with the nominal source of most gerund-participial
complements. But it must be emphasised that we are talking here of historically motivated
tendencies and associations, not constant elements of meaning.
Catenative verbs that license both form-types

From the point of view of Present-day English the main interest is in those cases where
the same verb allows both form-types. In some cases there is no discernible difference
between the constructions, while elsewhere we find a variety of differences bearing some
relation to the above broad outline.
(a) No discernible difference
This applies with bother, intend, plan, propose :
[50] i a. He didn’t bother to tell us. b. He didn’t bother telling us.
ii a. He intends to leave tomorrow. b. He intends leaving tomorrow.
Note that [50iib] shows that futurity is not incompatible with a gerund-participial.
Attempt also belongs in this group (contrast try in [56] below), though the gerund-
participial is rather rare: She attempted to walk / walking without the stick.
(b) Aspectual verbs
With the subset of aspectual verbs given in Class 1cii, there is again often no percep-
tible difference in meaning: He continued to see / seeing her every Sunday. But, es-
pecially with begin and start, the constructions are not always wholly equivalent or
interchangeable:
[51] i a. I began to understand how she felt. b. ? I began understanding how she felt.
ii a. ? Don’t start to tell me how to run b. Don’t start telling me how to run
my life. my life.
The gerund-participial tends to suggest ongoing activity. In [i] the understand situation
is too static for the gerund-participial: compare She began explaining how she felt. This
is why the gerund-participial is not possible in the existential construction: ∗There began
being some grounds for hope of an improvement. The salient context for [ii] is one where
you have already said something which I interpret as telling me how to run my life and
in this case the infinitival is not appropriate: compare If it starts to rain / raining, bring
the washing in, where either form-type is appropriate.44
(c) Verbs of liking (and not liking)
[52] i a. I like to stay home at weekends. b. I like staying home at weekends.
ii a. I’d like to be a politician. b. I’d like being a politician.
With [i] there are many contexts where [a] and [b] would be equally appropriate, but
there are also some favouring one or the other. Suppose you ask me to go bushwalking
next week-end but I wish to decline: [a] would here be more appropriate than [b].
Conversely if I am currently enjoying a week-end at home [b] is more appropriate
than [a].
The infinitival is more associated with change, the gerund-participial with actuality.
Thus someone who has recently turned forty or got married might say I like being forty
or I like being married. An infinitival would be strange here, suggesting repeated changes
from not being forty or married to being forty or married. In this case the meaning is
close to that of enjoy, which only allows gerund-participials.45 Would like, by contrast,
projects into the future and resembles a verb of wanting, with a strong preference for the
infinitival, as in [iia]; [iib] is possible, but the interpretation is roughly “I’d like/enjoy the
life of a politician”. If we change the example to I’d like to start the meeting a little earlier
this week the gerund-participial becomes quite implausible: I’d like starting the meeting
a little earlier suggests that the starting is itself something to be enjoyed, which is an odd
idea.
Hate with a to-infinitival has an idiomatic use seen in
[53] I hate1 to tell you this, but your battery is flat.
This can be thought of as involving projection into the immediate future: “I’m going to
tell you, though I hate having to do so”. What is special about this use (virtually confined
to the 1st person) is the combination of simple present tense in the matrix and single
dynamic event in the complement – contrast [52ia], where we have repetition of staying
home; other verbs of liking and not liking do not allow this pattern, though it is found
with adjectives: I am happy / ∗like to tell you that you’ve passed your test (cf. also regret in
[56] below).
(d) Memory verbs
[54] a. I remembered1 to lock up. b. I remembered2 locking up.
In [a] the locking up is simultaneous with the remembering – but I remembered some
kind of prior obligation to lock up and hence there is projection into the future with
respect to that implicit earlier time. In [iib], however, I simply remembered some actual
past event. Forget behaves in the same way, but recollect takes only the second form and
meaning.
44
It is sometimes said that the infinitival is preferred in examples like I had just begun to sign / signing the cheque
when he snatched it away, where there is hardly time for ongoing activity, but many speakers find the two types
equally acceptable here.
45
Detest and for the most part dislike likewise are restricted to the gerund-participial.
§ 5.6.1 To-infinitival vs gerund-participial 1243
(e) Gerund-participial expressing progressive meaning

[55] a. I’ve finally got the program to work. b. I’ve finally got the program working.
Here [a] entails that the program now works, [b] that it is now working: the difference of
meaning here is therefore aspectual, with the infinitival non-progressive and the gerund-
participial progressive. Compare also want in [26ii] – though want differs from get in
that we can also have a progressive infinitival: I want them to be standing when the
Minister enters, but not ∗I’ve finally got the program to be working. This semantic contrast
is more often found between the gerund-participial and a bare infinitival, as with the
sense verbs of Class 3 Bii and have3 of 3 Bi. The gerund-participial also has a progressive
interpretation after certain verbs that don’t take infinitivals: catch, discover2 , find2 , smell
(from Class 3 Cii).
(f) Some individual cases
[56] i a. She tried1 to open the window. b. She tried2 opening the window.
ii a. They fear1 to go out at night. b. They fear1 going out at night.
iii a. He scorns to compromise. b. He scorns compromising.
iv a. I regret1 to inform you that . . . b. I regret2 telling her that . . .
v a. They reported him to have left b. They reported his leaving the safe
the safe unlocked. unlocked.
Try1 (“endeavour”) involves effort towards a goal: the opening is only potential; try2 (“test
the effectiveness of ”) indicates actual activity: she opened the window to see whether
this would achieve the desired result. With fear infinitival [a] involves an element of
volition/intentionality: [a] implicates that they don’t/won’t go out, while the gerund-
participial [b] lacks this meaning and is comparable to an NP object (They fear an attack).
Similarly [iiia] conveys that he doesn’t/won’t compromise, while in [b] the complement
is more nominalised (cf. He scorns compromise) and arguably the missing subject is
interpreted non-syntactically, i.e. it is a matter of compromising in general rather than
specifically on his part. Regret usually takes a gerund-participial, describing some actual
present state or past situation, as in [b]; with the to-infinitival, characteristically found
with 1st person simple present followed by a verb of informing (cf. hate in [53] above),
the subordinate situation follows (albeit immediately) the expression of regret. Report
in [v] is one of the few verbs of cognition/saying that allows a contrast between the two
main form-types; the difference is that [b] presents the report as true while [a] is non-
committal – a difference that relates to that between actual and potential, and which
also reflects the greater similarity of the gerund-participial to an NP, for They reported
his failure to lock the safe likewise presupposes that he failed to lock it.
The doubl-ing constraint

Some verbs that license gerund-participial complements cannot themselves occur in the
gerund-participle form when they have such a complement. Compare:
[57] i a. They started quarrelling. b. ∗They are starting quarrelling.
ii a. The lawn needs mowing. b. ∗The lawn is always needing mowing.
iii a. We considered buying one. b. We are considering buying one.
The succession of gerund-participles in [ib/iib] is excluded by what is known as the
‘doubl-ing constraint’. As evident from [iiib], it applies to only a subset of catenative
verbs – a small subset, in fact. The clearest cases are aspectual verbs such as begin, cease,
continue, start, stop, and verbs taking concealed passives, like need in [ii]. We should prob-
ably also include others, such as intend, but there is a good deal of variation between
speakers as to their acceptance of the [b] construction. We noted in §1.1 that gerund-
participials cannot have the progressive auxiliary be as head (∗They accused him of being
running away when the alarm sounded ), and this can be seen as a special case of the
constraint.
5.6.2 The minor form-types: bare infinitivals and past-participials

Bare infinitivals
Only a relatively small number of catenatives take bare infinitivals: the auxiliaries in Class
1 A (the modals and supportive do), the sensory perception verbs (3 Bii), and have, let, and
make among the causatives. In addition a few are found either with or without to: ought,
dare2 , know, and help.
Know takes a bare infinitival only in the perfect and with a special sense:
[58] i I’d never known him (to) lose his temper before.
ii I know him to be thoroughly reliable.
In [i] it’s a matter of knowledge based on more or less direct experience of his losing his
temper. The bare infinitival here is characteristic of BrE: AmE requires to in [i] as well
as [ii].46
With help some speakers restrict the bare infinitival to cases of relatively direct
assistance – compare:
[59] i He helped me (to) finish on time by doing the bibliography for me.
ii He helped me to finish on time by taking the children away for the week-end.
In [i] he actually did some of the work, whereas in [ii] he enabled me to do it myself. But
it is questionable how widely shared such judgements are: many speakers would allow
a bare infinitival in [ii] no less than [i], and there is certainly no clear-cut distinction
between direct and indirect help.
The bare infinitival is virtually restricted to constructions where the matrix is active.
Have and let do not passivise with infinitival complements, while the other transitive
catenatives always take to in the passive: He had never been known to lose his temper before.
With make, to is restricted to the passive: They made her regret it ∼ She was made to regret
it; for the relation between They saw him pocket the key and He was seen to pocket the key,
see the discussion of Class 3 Bii.
Past-participials
In the simple construction these occur with just three verbs (Class 1 E): the perfect marker
have and the passive markers be and get. In the complex construction they are found, as
46
Some speakers of BrE also allow a bare infinitival with find with a sense like “see” or “notice”: Outside you will
find Wren create new green dimensions with sensitive landscaping that creates a community and not just a row of
houses.
§ 6 Hollow non-finite clauses 1245
passives, with a fairly small subset of verbs taking infinitivals:

[60] i Most of the sense verbs (3 Bii) – I heard the window broken.
ii Get 4 and have 3 (3 Bi): She got/had the house painted; I had my wallet stolen.
iii Like from 2 Bi, want 2 from 2 Biii, report from 2 Bv, fear2 and order from 3 Aii; here
the past-participial is an alternant of a passive to-infinitival: He’d like / wants them
(to be) killed humanely; The captain was reported (to have been) killed; They are
feared (to have been)abducted; He ordered it (to be)destroyed.
iv Need2 and want1 from 2bii as an alternant of the concealed passive: He needs/wants
his hair %cut/cutting, but the past-participials are restricted to certain regional
dialects such as Scottish.
Except with report, fear, and order, past-participials are wholly or virtually restricted to
occurrence with an active matrix: ∗They were wanted killed; ?The door was heard slammed.
Past-participial complements are to be distinguished from adjectival passives (Ch. 16,
§10.1.3), which are found with most verbs taking adjectival predicative complements:
He remained mistrusted by his colleagues ; She considered it superseded, etc.
6 Hollow non-finite clauses
6.1 General properties

Hollow non-finite clauses are clauses other than relatives or open interrogatives where
some non-subject NP is missing but recoverable from an antecedent NP or nominal.
The missing NP is normally object of the verb or object of a preposition. In the following
examples, the hollow clause is enclosed in brackets, with the site of the missing NP,
the gap, shown as usual by ‘ ’, and the antecedent that determines its interpretation is
underlined:
[1] i The problem took her only a few minutes [to solve ]. [object of verb]
ii I found her father a very easy person [to get on with ]. [object of prep]
Examples that further illustrate the range of possibilities are as follows:
[2] i That he would do such a thing is hard [to believe ]. [clausal internal comp]
ii Pat is easy [to be intimidated by ]. [comp of passive by]
?
iii An ideal husband is not easy [to be ]. [predicative comp]
∗
iv You won’t find these kids easy [to teach Greek]. [indirect object]
The gap can be object of the preposition by in the passive, though examples of this
kind are very rare. Examples with the gap in predicative complement function, as in
[iii], are at best very marginal. And indirect object function is excluded, as seen in
[iv]: this is comparable to the resistance of the indirect object to being realised as a
gap in open interrogatives, relatives, or preposing (cf. Ch. 4, §4.3). Since the missing
element can be object of either the verb or a preposition, it is possible to find paired
examples that are alike save with respect to which function in the infinitival clause is
missing:
[3] i Serious music is hard [to play on an instrument like this]. [object of verb]
ii An instrument like this is hard [to play serious music on ]. [object of prep]
Construction [ii], however, with the missing element object of a preposition in a transitive
clause, is relatively uncommon. It is most likely when the object is part of an idiom: They
are hard to make sense of / get the better of / do justice to / take advantage of .
Hollow non-finite clauses belong to the minor type of unbounded dependency con-
structions: see Ch. 12, §7.1. They usually have to-infinitival form-type; we examine these
in §6.3 but look first at constructions involving hollow gerund-participials.
6.2 Gerund-participials
These are licensed as complement to the adjectives worth and worthwhile, and to the
preposition for with a purpose sense:
[4] i Your idea is certainly worth [giving some further thought to ].
ii The plan is so unpopular that it wouldn’t be worthwhile [our pursuing ].
iii This knife isn’t very good for [cutting meat with ].
Recall (from §3.2.3) that the superficially similar construction governed by such verbs as
need is a concealed passive not a hollow active, as evident from the possibility of having
a by phrase: The proposal needs [ evaluating by a specialist]. The missing element here,
therefore, is subject, not object.47
6.3 Hollow to-infinitivals

There are six constructions where hollow to-infinitivals are required or permitted.
(a) As complement to predicative adjectives and nouns

[5] i Max is impossible to live with .
ii The assignment was an absolute pain to do .
A sample of adjectives and nouns (or nominals) licensing this construction is given in:
[6] i a. awkward bad boring convenient cumbersome
dangerous depressing desirable difficult dreadful
easy embarrassing essential exciting expensive
fashionable fine good hard ideal
impossible instructive interesting nice odd
painful pleasant safe simple tedious
ticklish tough tricky useful wonderful
b. bastard bitch breeze cinch delight
devil doddle dream embarrassment joy
nightmare pain piece of cake pig pleasure
ii available beautiful fit free frosty
homely pretty ready soft suitable
The adjectives and nouns in [6i] have to do mainly with the ease or difficulty of the
situation described in the infinitival clause or with one’s emotional attitude to it. Note
that impossible belongs in the class but possible does not: That claim is impossible/∗possible
47
Bear occurs with a very limited range of gerund-participials, as in the familiar phrase It doesn’t bear thinking
about; the complement here is probably a hollow clause rather than a concealed passive.
§ 6.3 Hollow to-infinitivals 1247
to substantiate. PPs with similar meanings are also occasionally found: The temptation
was beyond his capacity to resist. A number of nouns used in this construction (inclu-
ding, for example, the first four cited in [ib]) belong to colloquial style. The adjectives
in [ii] are semantically and syntactically less homogeneous. Some are collocationally
quite restricted: The air was frosty to breathe ; They were pretty to look at ; It was soft to
touch.
The main difference between [6i] and [6ii], however, is that the former also license
ordinary to-infinitivals as subject or extraposed subject:
[7] i a. His speech was embarrassing /an embarrassment [to listen to ].
b. It was embarrassing /an embarrassment [to listen to his speech].
ii a. The document is now ready [for you to sign ].
b. ∗ It is now ready [for you to sign the document].
The semantic equivalence between [7ia] and [ib] bears some resemblance to that
between constructions with infinitival and finite clauses as complement to verbs like
seem, discussed in §2.1. Compare, for example:
[8] i a. Her criticism was hard [for Ed to accept ].
b. It was hard [for Ed to accept her criticism].
ii a. Ed seemed [to accept her criticism]. [raised subject + non-finite comp]
b. It seemed [that Ed accepted her criticism]. [dummy subject + finite comp]
We have said that with an infinitival complement seem takes a raised subject – that the Ed of
[8iia], like that of [iib], is not a semantic argument of seem but only of convince. It has been
suggested that a corresponding treatment is appropriate for [ia]. This would be to say that
her criticism is likewise a raised subject, that it doesn’t represent an argument of hard but only
of accept, as is transparently the case in [ib]. On this account her criticism in [ia] and Ed in
[iia] would both be raised subjects of the matrix clause, with her criticism and Ed interpreted
respectively as an object-argument and a subject-argument of the subordinate clause.48
There is, however, an important difference between the two pairs that leads us to reject a
raised subject analysis for [8ia]. The subject of seem in [8iia] can be a dummy element, but
the subject of be hard in [8ia] cannot. Compare [9i–ii] with [9iii–iv]:
[9] i a. It seems to have been Kim who leaked the news.
b. It seems that it was Kim who leaked the news.
ii a. There seems to have been a conspiracy between them.
b. It seems that there was a conspiracy between them.
iii a. ∗It’ll be hard for us to prove to have been Kim who leaked the news.
b. It’ll be hard for us to prove it to have been Kim who leaked the news.
iv a. ∗There will be hard for us to prove to have been a conspiracy between them.
b. It will be hard for us to prove there to have been a conspiracy between them.
We are saying that with an infinitival complement seem has a raised subject: this means that
seem itself imposes no constraints on what can occur in subject function but accepts any
element that is licensed as subject by the infinitival clause predicate. In [ia] it is accepted as
48
In the classical transformational-generative analysis her criticism appears in ‘underlying structure’ in the
subordinate clause and is moved (raised) into the matrix subject position. The rule concerned is commonly
called ‘Tough movement’, tough being one of the adjectives that allows this upward movement of a non-subject
NP.
subject of the seem clause because the infinitival clause belongs to the it-cleft construction,
which in finite clauses takes it as subject, as in [ib]. Hard, however, does not behave in
this way. The infinitival clause for us to prove it to have been Kim who leaked the news is
syntactically and semantically impeccable, but this does not suffice to sanction [iiia]. Similarly
with dummy there. This is allowed as subject of seem in [iia] because it is licensed by the
infinitival complement, which belongs to the existential construction. But again it is not
acceptable as subject in [iva] in spite of the well-formedness of for us to prove there to have
been a conspiracy between them. The conclusion must be that the external complement of hard
+ infinitival is not raised, not licensed purely within the infinitival clause. It must represent
a semantic argument of hard.
In [8ia], therefore, hard + infinitival denotes a property that is predicated of her criticism:
the subject must represent an entity of a kind that such a property can be ascribed to it. Cleft
it and dummy there do not satisfy this requirement. The equivalence between [8ia] and [8ib]
is quite consistent with this account. Her criticism has the property that for Ed to accept it is
hard, but the fact that there is such a property is sufficient to establish that it is hard for Ed
to accept her criticism. And if [8ib] is true it must follow that her criticism had the property
that it was hard for Ed to accept it. Similarly, in the matching pair [3], [i] assigns a property to
serious music, while [ii] assigns a property to the instrument, but each is deducible from the
other. Note also the contrast in such a pair as:
[10] i It has been a pleasure to listen to someone with so much enthusiasm.
ii ? Someone with so much enthusiasm has been a pleasure to listen to .
The underlined NP is completely acceptable as the complement of the preposition in [i], but
it is quite marginal as the subject of [ii]: if we were dealing with a raised subject construction
there should be no such difference.
Potential ambiguity between hollow and ordinary constructions
For many of the items in [6] an infinitival internal complement must always be of the
hollow type. Some of them, however, license ordinary as well as hollow complements,
and there is then the potential for ambiguity:
[11] i They are ready to use . [hollow]
ii They are ready to depart. [ordinary]
iii They are ready to eat ( ). [ambiguous]
In [i] use is transitive: the missing object is recovered from the antecedent they, while
the user is not explicitly indicated. In [ii] depart is intransitive, with they antecedent for
the missing subject – note, then, that an overt subject can be supplied for [i], but not
[ii] (They are ready for you to use , but not ∗They are ready for you to depart). Eat in
[iii] is a dual-transitivity verb, so that they can be antecedent for either a missing object
(cf. The jam tarts are ready to eat ) or the missing subject (cf. The guests are ready to
eat). Ready may be contrasted with, for example, easy, where we have hollow This knife
is easy to cut with , but not ordinary ∗This knife is easy to cut. Other adjectives that
behave like ready include:
[12] available bad fit free good nice
(b) Licensed by an attributive adjective

[13] i London is an easy place to get lost in .
ii The price was a difficult one to better .
§ 6.3 Hollow to-infinitivals 1249
These initial examples bear a clear resemblance to the construction just discussed: they
are comparable to London is easy to get lost in and The price was difficult to better. The
applicability of the adjective is contingent on it being construed with the infinitival: [i]
doesn’t say that London is an easy place, but that London is a place which it is easy
to get lost in. Likewise in [ii] it is not a matter of a difficult price, but of a price that
it would be difficult to better. It makes sense, therefore, to treat the infinitivals here as
indirect complements in the structure of the NP: they are licensed not by the head of the
construction, the noun, but by a dependent of it, the attributive adjective.
Very often, however, the adjective is applicable derivatively to the noun:
[14] i It’s a difficult book to understand.
ii That wasn’t a very sensible remark to make.
iii This was a surprising decision for them to take.
iv It is an extremely stressful and emotional decision for any woman to make.
If a book is difficult to understand, that makes it a difficult book. In [ii] the remark
itself as well as the act of making it will be construed as not very sensible. And likewise
in [iii–iv] both taking or making the decision and the decision itself were surprising
or extremely stressful and emotional. This is a very common type: compare similarly a
difficult person to get on with, an impossible price to pay, an easy problem to solve, a good
book to buy. Note that [iii], for example, is appreciably more likely than, say, This was a
surprising decision for them to criticise, though the latter is certainly grammatical.
The range of adjectives used in the attributive construction is somewhat wider than
that found in predicative function. Compare, for example:
[15] i That’s a stupid book to set as a text for Year 1.
ii ∗ The book was stupid to set as a text for Year 1.
There is likewise no corresponding predicative use of sensible and surprising in
[14ii–iii]; semantically similar adjectives such as clever, unusual, exciting show the same
pattern. Note, however, that the attributive adjective construction usually has the adjec-
tive within an NP that is itself in predicative function, as in all the examples in [13–15i].
We would not say, for example, ∗The catalogue contained several stupid books to set as
a text for Year 1, ∗They are charging us a difficult price to better. This restriction is less
applicable to those cases like [14] where the adjective can be interpreted as applying to
the head noun as well as to the subordinate clause: She is married to a rather difficult guy
to get on with.
This construction is not sharply distinct from that containing a noun postmodified by a
non-wh relative clause. Compare:
[16] i The premier’s health is another significant issue to bear in mind. [relative]
ii That would be an interesting issue to explore. [structurally ambiguous]
In [i] the infinitival is quite independent of the attributive adjective significant, which could be
dropped without affecting acceptability. An approximate gloss is “an issue that we should bear
in mind which is significant”. And [ii] can be interpreted in the same way: “an
issue that we should/might explore which is interesting”. But it can also be interpreted with
the infinitival an indirect complement licensed by interesting: “an issue which it would be
interesting to explore”. Though there are two different syntactic structures, the meanings
they yield are effectively the same. Cases where the present construction yields a clearly
different meaning from the relative are illustrated in [15i] and [14iii]. The relative interpre-
tation of the former would be the implausible “That’s a book that we could/should set as a
text for Year 1 which is stupid”. And what differentiates [14iii] from a relative is that it conveys
that the decision was in fact taken: the infinitival relative generally has a modal meaning,
involving what could or should be done.
(c) As complement to verbs

There are five verbs that take hollow to-infinitivals as internal complement: be and the
transitive cost, need, require, take.
[17] i The decision is for you to make .
ii The car cost over $1,000 to repair .
iii The dispute needed/required a great deal of tact to resolve .
iv The letter took me all morning to write .
With be the hollow clause usually has for + subject, but we also find exceptional cases of
subjectless ones: The house is to let and You are to blame . These are restricted to let
(perhaps also rent) and blame : compare ∗The house is to sell or ∗You are to criticise. The
construction with the transitive matrix verbs shows the same alternation as we illustrated
in [8] for the adjective hard ;49 compare [17ii–iv] with:
[18] i It cost over $1,000 to repair the car.
ii It needed/required a great deal of tact to resolve the dispute.
iii It took me all morning to write the letter.
(d) With dummy there, have (got), with/without, and predicative genitive
[19] i There is/remains Kim [to consider ].
ii Jill has (got) her elderly parents [to look after ].
iii With her elderly parents [to look after ], Jill is finding life somewhat stressful.
iv The money wasn’t yours [to spend ].
The infinitival here has a modal interpretation. In [i–iii] it is a matter of deontic necessity:
Kim needs to be considered, Jill’s elderly parents need to be looked after. In [iv] it is deontic
possibility, and since the matrix clause is negated we understand “You weren’t entitled
to spend the money”.
In the deontic necessity case we again find a close relationship with the relative clause con-
struction. Compare:
[20] There are several assignments to mark / that I have to mark.
The examples in [19] were chosen as ones where the object NP would not accommodate
an integrated relative clause, with Kim a proper noun (in its proper name use) and her
elderly parents containing a genitive determiner and being referentially fully determinate.
The finite relative clause in [20] differs from an ordinary integrated relative, however, and
it is uncertain whether the infinitival should here be included in the relative category or
not.
49
There is also an alternant for [17i] with it : It is for you to make the decision. But this does not belong to the
same construction as [18]. In particular, for you is a PP functioning in the matrix clause, not subordinator +
subject in the infinitival, as is evident from the fact that for is replaceable by up to.
§ 7 Non-catenative complements in clause structure 1251
(e) Complements licensed by too, enough, sufficient(ly)

[21] i My coffee was too hot to drink .
ii I haven’t enough money left to spare for luxuries like that.
iii The proposal isn’t sufficiently developed for us to accept in its present form.
These items also license ordinary infinitivals (He’s too young to go to school): we discuss
both types together briefly in §8.4.
(f) Purpose infinitivals

[22] i I bought them to give to the children.
ii I need it for the children to do their homework on .
Again the purpose clause is not required to have a non-subject gap, and we therefore
postpone further discussion of the construction to §9.
7 Non-catenative complements in clause structure
We have argued in §4.1 that non-finite complements of verbs cannot in general be assigned
the same function as NPs or AdjPs, but are best analysed as constituting a distinct type
of complement, the catenative complement. The cases not covered by the latter are
illustrated in:
[1] i For you to accept liability would be a serious mistake. [subject]
ii It is important to ascertain the cause of the malfunction. [extraposed subject]
iii He considers taking advice beneath his dignity. [object]
iv I thought it better to wait. [extraposed object]
v His aim is to gain control of the company. [subjective predicative comp]
vi I’d call that taking unfair advantage of a beginner. [objective predicative comp]
The subject is an external complement sharply distinguished in English from other types
of complement: there is therefore no problem in distinguishing the infinitival in [i] from
our catenative complements and identifying it functionally with other forms of subject.
The infinitival in [ii] is distinguished from catenatives by virtue of its relationship with
the dummy element it in subject position.
In [1iii] the subordinate clause serves as predicand for the predicative complement
beneath his dignity. The relationship between the gerund-participial and the PP matches
that between NP object and PP in He considers such action beneath his dignity and is
sufficient to distinguish the gerund-participial from a catenative complement and to
enable us to subsume [iii] under the complex-transitive construction. In [iv] we have
another extrapositional construction, and the relationship with it is again sufficient to
distinguish the infinitival from a catenative complement.
Example [1v] is an instance of the specifying be construction, with the infinitival
complement distinguishable from a catenative by its ability to switch function with the
subject (To gain control of the company is his aim), a property it shares with other forms,
such as the PP in The best place is in the garden ∼ In the garden is the best place (Ch. 4,
§5.5.2). We also include various other non-finite complements to be under the predicative
complement function. The catenative uses of be are the progressive (She is working), the
passive (She was nominated for the position of treasurer), and the quasi-modal (Everyone
is to remain seated).
Finally, [1vi] is a further case of the complex-transitive construction, but this time the
gerund-participial is in predicative function. The relationship between it and the object
clearly matches that between adjectival predicative and object in I’d call that unfair. The
distinction between catenative and non-catenative is not so sharp here. We take the
gerund-participial in He kept them waiting, for example, to be a catenative complement,
not a predicative, because the relationship between it and them is not the same as that
between predicative and object in He kept them warm. Compare here the following sets:
[2] i a. I call that unfair. b. That is unfair.
ii a. I call that taking advantage of him. b. That is taking advantage of him.
iii a. He kept them warm. b. They were warm.
iv a. He kept them waiting. b. They were waiting.
The sets do not match because while the be of [ib/iib] is the same (the copula), this is not
so for [iiib/ivb], where the former is the copula and the latter the progressive marker:
warm is a predicative complement but waiting is not.50
7.1 Subject and extraposed subject

(a) To-infinitivals
[3] i a. For you to take the children could seriously endanger our mission.
b.To refuse her request would be unthinkable.
ii a. It embarrassed her to see him so drunk.
b. It would be a good idea for you to consult a solicitor.
Subjects of the infinitival are freely admissible, as in [ia] and [iib], though the subjectless
form is much more frequent.
A sample of items that license infinitival subjects is given in:
[4] i amuse cause cost delight disturb
embarrass occur [to] pay please take
ii easy essential foolish good hard
impossible necessary possible ridiculous usual
iii mistake offence pleasure task way
Verbs, adjectives, and nouns are listed in [i–iii] respectively. With the verbs, the licensing
often involves internal complements as well as the verb head: It took courage to tell them,
but not ∗It took Kim by surprise to tell them; It didn’t occur to me to invite him, but
not ∗It didn’t occur last Tuesday to invite him. The adjectives and nouns head phrases
in predicative complement function, as in [3ib/iib]; this pattern is much more frequent
than the one with verb as licensor ([3ia/iia]). PPs with meanings comparable to adjectives
are also found: out of the question, of considerable interest, and so on.
There is a large overlap between the items that license infinitivals and those that license
declarative finites (content clauses). The latter can be either mandative or non-mandative
(Ch. 11, §7.1.1), and for those that take mandatives there is in general little difference in
50
There is of course an interpretation of [2iib] where be is the progressive auxiliary, but that doesn’t stand in any
significant relation to [2iia], and hence is irrelevant to the argument.
§ 7.1 Subject and extraposed subject 1253
meaning between the infinitival and finite constructions. Compare, for example:
[5] i a. It is important for you to lock up carefully.
b. It is important that you lock up carefully.
ii a. It was necessary for him to walk to school.
b. It was necessary that he walk to school.
Where the finite is non-mandative, it tends to be concerned with facts or propositions
while the infinitival is concerned with situations (actions, events, states, etc.), though in
some cases the difference is again very slight. Compare:
[6] i a. It was good to be back at school.
b. It was good that they were back at school.
ii a. It pleased her to be honoured in this way.
b. It pleased her that she was honoured in this way.
In [ia] the situation of being back at school was good (for whoever it was who was back
at school), while in [ib] the fact of their being back at school was good (for whoever it
was who judged it good – in the default case, the speaker). But there is no perceptible
difference in [6ii].
The distinction is particularly clear with possible as the licensing adjective:
[7] i It was possible for him to walk to school.
ii It was possible that he walked to school.
With an infinitival complement, possible expresses dynamic or deontic modality, whereas
with a finite complement it is epistemic. A rough paraphrase for [ii] would be Maybe he
walked to school: concerning whether the proposition that he walked to school is true,
[ii] says maybe it is. But in [i] the issue is not the truth of a proposition, but his abilities.
Note, however, that the modal adjective necessary does not exhibit the same difference:
the modality is deontic in both constructions in [5ii].
The tendency for non-mandative finites to be associated with facts/propositions and
infinitivals with situations ties in with the two main cases of items restricted to one
or other form of complement. These restrictions are stated in [8], and exemplified
in [9].
[8] i Adjectives concerned with truth or likelihood take declaratives, not infinitivals.
ii Adjectives concerned with the ease or difficulty of doing something take infini-
tivals, not declaratives.
[9] i a. ∗ It was obvious for him to be lying.
b. It was obvious that he was lying.
ii a. It was easy for me to sympathise with her.
b. ∗It was easy that I sympathised with her.
Obvious in [ib] can be glossed as “obviously true”; other adjectives likewise restricted
to finites are likely, probably, certain, clear, evident, apparent – and true and false.51
Conversely, facts or propositions can’t be easy, hence the difference in [ii]. Hard can take
a finite complement, but its sense is then quite different from the sense it has with an
51
These last two can take an infinitival if its verb is say (or a near-synonym): It’s true that he cheated, It’s true to
say that he cheated (with the content clause an internal complement of say, not an extraposed subject of be
true), but not ∗It’s true for him to have cheated.
infinitival:
[10] i It’s hard for them to work twelve hours a day.
ii It’s hard on them that they have to work twelve hours a day.
The situation described in the infinitival is often merely potential rather than actu-
alised, and this is reflected in the frequent occurrence of the infinitival in construction
with would be, where the corresponding non-mandative finite has if, not that :
[11] i a. It was good to invite them both. b. It would be good to invite them both.
ii a. It was good [that you invited them b. It would be good [if/∗that you invited
both]. them both].
In [iia] good is factive: it is taken for granted that you invited them both. This factiv-
ity is inconsistent with the conditional implication of would be, and hence we need a
conditional adjunct in [iib] instead of a factive complement.
(b) Bare infinitivals

These are occasionally found, in informal style, as subject in the reversed version of the
specifying be construction; the internal complement contains the verb do:
[12] i Plead mitigating circumstances is all you can do.
ii Seek professional advice is what we should do.
(c) Gerund-participials
[13] i a.Their reporting him to the manager led to his dismissal.
b. Inviting your uncle was a bad mistake.
ii a. It’s no use his/him asking for special consideration.
b. It has been nice meeting you.
The subject of the gerund-participial is much more often left understood, as in [ib/iib],
than overtly expressed. The overt subject is particularly unlikely in the extraposed con-
struction.
The gerund-participial is considerably less frequent in these functions than the in-
finitival, and there is also a major difference with respect to the extraposed and non-
extraposed constructions: infinitivals are usually extraposed, gerund-participials com-
paratively rarely. To a significant extent, then, the difference between infinitival and
gerund-participial is a matter of information packaging: end position tends to favour
the infinitival while basic subject position tends to favour the gerund-participial.
For the most part, items that license a gerund-participial also license an infinitival,
and vice versa. An infinitival could, for example, be substituted for the gerund-participial
in [13ib] and [ii]. But such substitution would not be possible in [ia]: ∗For them to report
him to the manager led to his dismissal. Compare, similarly:
[14] i Paying off the mortgage last year has put us in a strong position.
ii ∗It has put us in a strong position to pay off the mortgage last year.
The gerund-participial here and in [13ia] denotes a specific, actualised situation, and
although the infinitival is not wholly excluded in such cases (cf. It was a mistake to invite
your uncle), its use is considerably constrained. Note, then, that if we change from actual
to hypothetical, the infinitival becomes acceptable:
§ 7.2 Object and extraposed object 1255
[15] i (For them) to have reported him to the manager would have led to his dismissal.
ii To have paid off the mortgage last year would have put us in a strong position.
The gerund-participial is certainly not restricted to actualised situations (cf. Changing
the arrangements would be very difficult at this stage), but for non-actualised situations
the infinitival will often be required or at least quite strongly preferred, especially where
the non-finite contains a subject:
[16] i a. It would be better for the lecture to be rescheduled.
b. ∗The lecture’s being rescheduled would be better.
ii a. To doubt her word would never have occurred to me.
b. ? Doubting her word would never have occurred to me.
7.2 Object and extraposed object

As explained at the beginning of §7, we take non-finite clauses to be objects, rather
than catenative complements, only when they occur in some distinctively object relation
with some element other than the head verb. The main case is the complex-transitive
construction:
[17] i This made obtaining a loan virtually impossible.
ii I regard solving this problem as my first priority.
Infinitivals cannot occur in this position between verb and predicative complement;
normally, then, they occur in extraposed object position, though it is marginally possible
for them to be preposed in prenuclear position, without extrapositional it :
[18] i I thought it wise to adopt a low profile.
ii For them to sack him we would regard as a gross miscarriage of justice.
A gerund-participial, generally short and simple in structure, can also function as
indirect object in the ditransitive construction, though it is likely to be considered stylis-
tically somewhat awkward:
[19] We’ve been giving moving to Sydney a good deal of thought recently.
7.3 Predicative complement

The construction most clearly distinct from the catenative is the reversible specifying
construction, where the internal complement can be a to-infinitival, a bare infinitival,
or a gerund-participial:
[20] i His intention was (for the meeting) to begin at six [to-infinitival]
ii All I did was print out the table of contents. [bare infinitival]
iii The funniest thing was (Kim) trying to hide in the coal-box. [gerund-participial]
The bare infinitival is restricted to cases where the subject NP contains do in a relative
clause; as we have noted, to can be added here (All I did was to print out . . . ). Since
a subjectless gerund-participial can also be catenative complement to progressive be,
some examples are ambiguous between the specifying and progressive constructions.
This potential ambiguity is exploited in the advertising slogan Our business is working
for you: as a specifying construction this identifies what our business is; with be a marker
of progressive aspect, it says what our business is doing.
In addition a to-infinitival can occur as purpose complement, replaceable by a purpose
PP such as for the purpose of . . . :
[21] The grid is to prevent the cattle from wandering off.
One other construction that belongs here is the idiomatic one illustrated in:
[22] i To discuss melodrama, then, is to raise questions about ‘culture’ itself and the cate-
gories and oppositions by which we conceptualise it.
ii For any German director to attempt to make a film about Josef Mengele, the notorious
Auschwitz concentration camp doctor, is to court controversy.
This construction has a to-infinitival as both subject and internal complement of be. It
differs from the specifying construction in that it cannot be reversed – To raise questions
about culture itself . . . is to discuss melodrama is not equivalent to [i]. The construction
indicates what the situation described in the subject entails or necessarily involves: “The
discussion of melodrama necessarily raises questions about culture itself . . . ”. In general,
it can be paraphrased by means of if or when: If/When a German director attempts to
make a film about Josef Mengele, they necessarily court controversy.
Finally, we have noted that a gerund-participial can function as objective predicative
complement under very restricted conditions, mainly with call as matrix verb, as in
[2iia].
8 Further complement uses of non-finite clauses
8.1 Non-finite complements of adjectives

These are predominantly to-infinitivals. We review these first, and then turn very briefly
to gerund-participials; adjectives do not license past-participial complements.
To-infinitivals
The construction where an adjective is followed by a direct complement – i.e. one
licensed by the adjective lexeme itself – is to be distinguished from a number of super-
ficially similar ones. Compare:
[1] i You are [free to leave when you want]. [direct comp of adj]
ii She’s [too young to go to school]. [indirect comp]

iii She’s [young] to be going to school.
[adjunct in clause structure]
iv I was [mad] to volunteer.
v It would be [foolish]to ignore them. [extraposed subject]
In [ii] the infinitival is a constituent of the AdjP, but is licensed by too rather than by
the adjective young. It is therefore an indirect complement; complements of this kind
are dealt with in §8.4. In [iii–iv] the infinitival is an adjunct, not a complement. It is not
lexically licensed, and though it could not be preposed there is some evidence that it does
not form part of the AdjP. Note in particular that the adjective + infinitival could not
function as postmodifier in NP structure: compare ∗She is one of those young to be going
to school and ∗Anyone mad to volunteer can’t expect much sympathy. While [ii] says that
she is young to a degree higher than that at which she can or should go to school, [iii]
§ 8.1 Non-finite complements of adjectives 1257
says that she is young relative to those who go to school: it is unexpected or noteworthy
that someone as young as she is should be going to school. The meaning of [iv] is that I
was mad in that I volunteered: vounteering was a mad thing to do on my part. Finally,
the infinitival in [v] is extraposed subject (compare the version without extraposition,
To ignore them would be foolish), and as such is not part of the AdjP; see §7.1 for this
construction.
Hollow vs ordinary infinitivals
[2] i Their argument was [impossible [to follow ]]. [hollow]
ii Kim was [anxious [to follow the argument]]. [ordinary]
The first division within the infinitival complements licensed by the head adjective is
between the hollow and ordinary types. Hollow clauses, like to follow in [i], have a gap
in some non-subject function, normally object of a verb or preposition; they have been
discussed in §6, and henceforth in this section we will confine our attention to ordinary
infinitivals, those without such a gap, like to follow the argument in [ii].
Raising and non-raising adjectives
The distinction between catenative verbs like hope and seem, which take respectively
an ordinary and a raised subject, applies also to adjectives functioning as predicative
complement and taking an infinitival complement. Compare:
[3] ordinary subject raised subject
a. Liz was determined to convince them. b. Liz was likely to convince them.
In [i] the subject Liz represents an argument of determined to convince them: the property
denoted by the AdjP is ascribed to Liz. But in [ii] the likelihood applies not to Liz but
to the situation of Liz’s convincing them. The grammatical and semantic differences
noted in our discussion of hope and seem in §3.1 apply in essentially the same way to the
adjectival construction. Compare:
[4] i a. Lizi was determined that shei would convince them. [double reference to Liz]
b. It was likely that Liz would convince them. [single reference to Liz]
ii a. They were determined to be convinced by Liz. [= 3a]
b. They were likely to be convinced by Liz. [= 3b]
iii a. #This news was determined to convince them. [violates selection restriction]
b. This news was likely to convince them. [no violation]
∗
iv a. There is determined to be enough food left. [dummy subject inadmissible]
b. There is likely to be enough food left. [dummy subject admissible]
v a. Liz was determined for them to have a good time. [infinitival admits subject]
b. ∗ Liz was likely for them to have a good time. [infinitival excludes subject]
Because the parallel with the verbal constructions is so close, only a brief commentary
is needed. In [i] the infinitivals are replaced by finite clauses. In [ia] Liz remains subject,
and is the antecedent for a personal pronoun in the convince clause, which is still com-
plement of determined. In [ib], by contrast, the subject is now it and the convince clause
is extraposed subject – compare the non-extraposed That Liz would convince them was
likely, where it is even more transparent that likely has a single argument. The data in [ii]
show that determined is voice-sensitive, while likely is not: [iia] differs in meaning from
[3a] because the determination is ascribed to ‘them’ rather than Liz, while the synonymy
between [4iib] and [3b] shows that the likelihood does not apply to Liz or ‘them’, but
to the situation of Liz convincing them and their being convinced by Liz. In [4iii] the
[a] example is anomalous because determined takes an animate subject, while [b] is ac-
ceptable because there is no direct semantic relation between likely and the subject. This
is why determined does not permit a dummy pronoun like there as subject, while likely
does, as shown in [4iv]. Finally, the examples in [4v] show that with determined but not
likely the infinitival can take a different subject from the matrix clause: the raised subject
in the likely clause belongs semantically in the infinitival clause, and there is therefore
no possibility of adding another subject to the latter.
The adjectives taking a raised subject, besides likely, are as follows:
[5] about apt bound certain due
fated liable set sure wont
The complement is obligatory: omitting it leads either to ungrammaticality (He is wont
to be late but not ∗He is wont) or to a change in the meaning of the adjective (compare She
is sure to win, “It is certain that she will win”, and She is sure, “She is not in any doubt”).
Adjectives taking ordinary subjects are much more numerous. A sample is given in:
[6] able accustomed afraid annoyed anxious
ashamed astonished careful concerned content
curious delighted depressed disgusted disposed
eager F eligible embarrassed fascinated fit
free frightened furious glad happy
hesitant impatient impotent inclined indignant
interested jubilant keen F loath perturbed
poised powerless prepared prompt prone
puzzled qualified quick ready F relieved
reluctant F satisfied slow sufficient F surprised
thankful welcome willing F worried worthy
Many of these do not allow a subject in the infinitival clause, or do so only marginally:
those that most readily accept for + subject are annotated with ‘F’. Compare They are
willing for the proposal to be resubmitted and ∗You are welcome for your children to come
with you.
Able is a somewhat peripheral member of this class. It differs from clear members in that
pairs such as the following do not differ in truth conditions:
[7] i Primary schoolchildren are able to solve these problems.
ii These problems are able to be solved by primary schoolchildren.
Are able to could here be replaced by can, which is a raising verb. Yet able differs from can in
that it does not allow dummy pronouns as subject. Compare:
[8] i There can’t be any progress without goodwill on both sides.
ii ∗There isn’t able to be any progress without goodwill on both sides.
The clear inadmissibility of [ii] indicates that able cannot take a raised subject. The equivalence
of [7i–ii] must then be handled along the lines suggested for the pair given in [8i] of §6.
Example [7i] ascribes a property to primary schoolchildren, not to the situation of their
solving these problems, and [7ii] ascribes a property to the problems. But if the children have
the property that they are able to solve the problems, the problems must necessarily have
§ 8.2 Non-finite complements of nouns 1259
the property that they are able to be solved by the children. The truth conditions of [7i–ii]
must therefore be the same even though they do not belong to the raised subject construction.
Gerund-participial complements
The adjectives busy and worth/worthwhile license complements of this form:
[9] i She was busy [preparing her report]. [ordinary]
ii These objectionsi aren’t worth [bothering about i ]. [hollow]
iii It isn’t worth [taking the matter any further]. [ordinary; impersonal]
Numerous adjectives take gerund-participials as oblique complements, i.e. with a gov-
erning preposition (engaged in preparing her report, keen on playing games, etc.), but busy
takes the gerund-participial directly, as in [i]. Worth and worthwhile take hollow gerund-
participials, as in [ii], where the gap functioning as complement of about is anaphorically
linked to the predicand these objections. They also license ordinary gerund-participials, as
in [iii], where there is no non-subject gap in the bracketed clause. This type is restricted
to the construction with impersonal it as subject; it is comparable to extraposition, but
does not allow the subordinate clause to appear in subject position: ∗Taking the matter
further isn’t worth.
8.2 Non-finite complements of nouns

Many nouns license non-finite complements, all to-infinitivals:
[10] i advice aim application F appointment arrangement F
attempt authorisation F claim command compulsion
consent F decision desire F determination exhortation
failure hope F incitement inducement inspiration
instruction F intention invitation longing F move F
need F obligation offer order F permission F
plan F pledge plot pressure promise
proposal F provocation recommendation refusal reminder
request resolution selection struggle F tendency F
threat undertaking F vow warning will
wish F yearning F
ii ability eagerness F eligibility fitness freedom F
impatience keenness F readiness F reluctance F willingness F
iii chance F concern opportunity F power strength
The great majority of the head nouns are morphologically derived from (or homonymous
with) verbs of matching senses that take the same complementation, as with those in
[i]. Those in [ii] are similarly derived from adjectives, while with those in [iii] the
complementation is not predictable in this way. Compare then:
[11] i a Kim decided to go to Bonn. b. Kim’s decision to go to Bonn
ii a. Pat was eager to help us. b. Pat’s eagerness to help us
iii a. [no relevant counterpart] b. the opportunity to make a quick profit
We include strength in [iii], not [ii], because the adjective strong does not take an in-
finitival complement. Note, then, that the following do not have matching structures or
interpretations:
[12] i She was strong to withstand this pressure.
ii She had the strength to withstand this pressure.
In [i] the infinitival is an adjunct in the clause; the meaning is that she withstood the
pressure, this indicating that she was strong. In [ii] the infinitival is a complement, and
the meaning is “She had the strength necessary to withstand the pressure”.
All the nouns in [10] take subjectless infinitivals, and those with the annotation ‘F’
also allow the construction with subordinator for + subject:
[13] i This provided an opportunity [for them to plan the next step].
ii Permission [for the ceremony to be held in the church itself ]was finally granted.
A number of other nouns in our list could probably also occur with for + subject in the
infinitival, but it is in general a relatively infrequent construction, and judgements as to
whether a given noun could appear here are not always clear-cut.
In the subjectless construction, the antecedent for the missing subject may be found
within the NP (as genitive determiner or within a PP complement), as in [14i], or
outside the NP, as in [ii], and in cases like [iii] there is no antecedent in the sentence
at all:
[14] i a. your /Kim’s promise to help me with my tax return
b. a proposal by the government to introduce a goods-and-services tax
c. the willingness of the other members to agree to the proposal
d. an instruction to the secretary to call an extraordinary meeting
ii a. They gave me instructions [to evacuate the building].
b. I received instructions [to evacuate the building].
c. What I hadn’t expected to receive was an instruction [to evacuate the building].
iii They were discussing a proposal [to introduce a summer semester].
In general, the recovery of the understood subject is determined by semantic princi-
ples, not rules of syntax. But in some cases the matter is more grammaticalised; with
selection, for example, the antecedent is normally required to appear as complement
to of :
[15] i The selection of Judge Carter to head the inquiry is to be welcomed.
ii ∗The selection to head the inquiry hasn’t yet been announced.
iii ∗Judge Carter is their selection to head the inquiry.
Nouns do not take raised complements
In general, clauses with raised complements, licensed by raising verbs or adjectives, do
not have counterparts with the form of NPs:
[16] i a. Kim seemed to be distressed. b. ∗the seeming of Kim to be distressed
ii a. I believe them to be genuine. b. ∗my belief in/of them to be genuine
iii a. They are certain to resent it. b. ∗their certainty to resent it
One exception involves the verb tend, which has the noun counterpart tendency. Corres-
ponding to clausal The tabloids tend to support Labour and the equivalent Labour tends to be
supported by the tabloids we have:
§ 8.2 Non-finite complements of nouns 1261
[17] i a. the tendency for the tabloids to support Labour

b. the tendency for Labour to be supported by the tabloids
ii a. the tendency of the tabloids to support Labour
b. Labour’s tendency /the tendency of Labour to be supported by the tabloids
In [ia/ib] the tabloids and Labour are subjects of the infinitival clauses, and the equivalence
between the whole NPs simply reflects the equivalence holding between the active and passive
subordinate clauses they contain. In [ii], however, the tabloids and Labour are oblique or
genitive complements of tendency: here, therefore, they appear to bear some resemblance
to raised complements. There is, however, a major difference between the NP and clause
constructions. Dummy elements can occur as raised complements in clause structure, but they
cannot occur as oblique or genitive complement of tendency – or of any other noun. There
does not have a genitive form and cannot normally occur as complement of a preposition,
but these general restrictions do not apply to it, and yet this too cannot be used in the relevant
positions when it is a dummy element:
[18] i a. It tends to be the wife who provides this support.
b. ∗its tendency to be the wife who provides this support
ii a. It tends to be more efficient to pay by credit card than by cheque.
b. ∗its tendency to be more efficient to pay by credit card than by cheque
Such data argue that nouns, unlike verbs and adjectives, do not take raised complements.
We take the tabloids and Labour, therefore, to represent arguments of tendency in [17iia–b]
respectively, though not of course in [17i]. The relation between [17ia] and [17iia] is thus the
same as that between [8ib] and [8ia] of §6.3 (It was hard for Ed to accept her criticism and Her
criticism was hard for Ed to accept).
The same applies to failure, derived from the raising verb fail. The verb allows dummy
subjects, but the noun does not:
[19] i It had failed to become apparent, even after a day, what their intentions were.
ii ∗its failure to become apparent, even after a day, what their intentions were
Consider finally the case of order:
[20] i a. They ordered the building to be evacuated.
b. #the order to the building to be evacuated
c. the order for the building to be evacuated
ii a. They ordered the doctor to examine the victims.
b. the order to the doctor to examine the victims
c. the order for the doctor to examine the victims
The verb takes a raised object: the building is not an argument of the verb in [ia]. But with
the noun the complement of to is an argument, with the role of recipient of the order.
Hence the anomaly of [ib]: one doesn’t give orders to buildings. The nominal structure
thus differs from the verbal one in allowing for this role to be explicitly encoded. It does
not have to be, however, for we can also have [ic/iic]: these match the verbal (clausal) con-
struction in not encoding who received the order. Here the building and the doctor function
as subject of the infinitival, not as oblique complement of the noun. Again, then, oblique
complements of the noun are not raised: they have to be interpretable as arguments of the
noun.
8.3 Non-finite complements in the structure of PPs

Gerund-participials
These occur very freely as complement of a preposition:
[21] i I’m looking forward [to (you/your) returning home].
ii [On hearing the news] she immediately telephoned her father.
iii She was reported [as saying that she would appeal against the ruling].
iv [Although claiming to have a Ph.D.,] he didn’t in fact have any degree at all.
Although more often occurs with a finite complement (as in although he claimed to have a
Ph.D.). Such clauses can be reduced to gerund-participials, past-participials, or verbless
clauses: see [3] of §10 for a list of prepositions of this kind.
To-infinitivals
In contrast to gerund-participials, to-infinitivals occur in this function only under very
restrictive conditions. The only prepositional expressions that take to-infinitivals other
than interrogatives as direct complement are purposive in order, and as if/though:
[22] i We got up at five [in order to catch the early train].
ii He raised his hand [as if to defend himself ].
iii She glanced out of the window at the phaeton [as though to say that he was not
the only man to have a new carriage that morning].
The as if /though construction is related to purpose in that it can be glossed as “as if/though
with the purpose/intention of ”. In order allows for + subject, but as if/though does
not.
Past-participials
These occur in the complement of prepositions like although, until, etc., that allow
reduction of a finite complement, and also in comparative clauses (see Ch. 13, §2.1):
[23] i Please remain seated [until requested to board your flight].
ii He had more debts [than previously acknowledged ].
iii The problem turned out to be more serious [than expected ].
8.4 Indirect and matrix-licensed non-finite complements

Infinitivals indirectly licensed by too, enough, sufficient, sufficiently
[24] i a. It is too late [for you to go out now].
b. Enough people turned up [to form a quorum].
c. The instructions weren’t sufficiently clear [for us to be able to assemble it].
ii a. Too good [to miss ]is how I’d describe it.
b. The problem isn’t important enough [to worry about ].
c. Have you had sufficient [to eat ]?
The underlined degree expressions license ordinary infinitivals, as in [i], or hollow
ones, as in [ii]. The complements are indirect in that in constituent structure they
are not dependents of the licensor, but of a head item that is modified by the latter.
In [ia], for example, the infinitival is a dependent of late, not directly of too. The in-
finitival strongly favours end position in the matrix clause: [ib], for example, is strongly
§ 8.4 Indirect and matrix-licensed non-finite complements 1263
preferred over Enough people to form a quorum turned up. The latter is not ungrammatical,
however, and in the specifying be construction [iia], the infinitival cannot be moved to the
end.
The licensors indicate degree relative to some need, purpose, desire, etc. Too expresses
a degree that exceeds the maximum or upper bound consistent with fulfilling the need,
purpose, or desire. In [ia] it is late to a degree higher than the maximum at which you can
or should go out: it follows that you can’t or shouldn’t go out now. Note that excessively
is not substitutable for too: compare My coffee is too/∗excessively hot to drink. Enough and
sufficient(ly) express a degree that is at least as high as the minimum or lower bound.
In [ib], for example, the number of people who turned up was at least as high as that
needed to form a quorum.
Hollow infinitivals indirectly licensed by an attributive adjective
[25] That was a silly thing [to do ].
The infinitival is licensed by silly but in constituent structure is a complement in the NP
headed by thing ; for this construction, see §6.3.
Matrix-licensed complements
A number of prepositions take non-finite complements if the larger construction licenses
them (see Ch. 7, §5.1). There are three main cases.
(a) With prepositions of inclusion or exception including, but, except, save
[26] i He does nothing but/save/except waste people’s time.
ii I couldn’t help but notice her embarrassment.
iii You have no choice but to accept her offer.
iv There’s nothing he wants save to pursue his studies in peace.
v This would achieve nothing except to antagonise some of our supporters.
In [i] the bare infinitival is licensed by do nothing + the preposition of exception: compare
∗
He likes nothing but waste people’s time (in this context we need gerund-participial
wasting) or ∗He does wonderful things but waste people’s time. In [ii] the licensor is the non-
affirmative idiom can help but. The other examples have to-infinitivals, but the licensing
is again a property of the matrix construction, with choice, want, achieve the decisive
elements. Want straightforwardly takes a to-infinitival itself: He wants to pursue his studies.
Achieve does not, but nevertheless to antagonise some of our supporters is to achieve
something, and hence the infinitival is admissible: contrast ∗He found nothing except to
antagonise our supporters.
(b) With the prepositions of comparison as and than
[27] i I’d rather stay at home than go out in this weather.
ii That wouldn’t be as bad as for you to lose your job.
iii They visit the area for such recreational purposes as to attend hockey matches.
In [i] the infinitival is permitted after than because the matrix would rather licenses
a complement of the same form (stay at home). In [ii] the predicate be bad licenses
infinitival subjects (For you to lose your job would be bad ). And in [iii] the infinitival
denotes a purpose.
(c)With as in purposive and resultative complements

[28] i They were asked to stand back so as not to hamper the efforts of the firefighters.
ii Blood for transfusion cannot be chosen so as to exclude every possibility of sensiti-
sation.
iii On this matter their views are so close as to be indistinguishable.
iv His art is such as to render the familiar original and mysterious.
v Will you be so good as to order your men not to molest my maid.
The as here is itself licensed by so or such. In most cases, as + infinitival alternates with
a finite clause optionally or obligatorily introduced by the subordinator that : so (that)
they would not hamper the efforts of the firefighters ; so (that) it will exclude every possiblity
of sensitisation ; so close (that) they were indistinguishable ; such that it renders the familiar
original and mysterious. No such replacement is possible in [v], where be so good as to
is an idiom meaning “kindly”. None of these constructions allows for + subject in the
infinitival.
The so phrase in [28i] is a purpose adjunct, while that in [ii] is a manner adjunct; in
the latter construction so can be separated from the as complement: cannot be so chosen
as to . . . Unlike so + finite clause, so + as + infinitival does not serve by itself as a
resultative adjunct: there is no as + infinitival counterpart of examples like He didn’t
wake up until ten, so that he wasn’t able to see his mother before she went to work.
8.5 Interrogative infinitival clauses

While the distribution of non-interrogative infinitival clauses is very different from that
of content clauses, this is not so with interrogatives. Interrogative infinitival complements
are found in a large subset of the environments where interrogative content clauses are
licensed. Compare, for example:
[29] i a. I don’t know whether I should go. b. I don’t know whether to go.
ii a. She decided what she would do. b. She decided what to do.
iii a. It doesn’t matter what you say. b. ∗It doesn’t matter what to do.
The interrogative phrase cannot have subject function (∗I don’t know who to go first),
and no other subject is permitted either (∗She didn’t say what for me to do). For further
discussion, see Ch. 11, §5.3.5.
9 Non-finite clauses as modifiers and supplements
We turn now, very much more briefly, to non-finite clauses in non-complement function.
Post-head modifier in NP structure

[1] i a. This provides [a solid foundation on which to build]. [wh relative]
b. This provides [a solid foundation to build on]. [non-wh relative]
ii a. [People living near the site] will have to be evacuated. [gerund-participial]
b. I came across [a letter written by my great-grandfather]. [past-participial]
Clauses in this function cannot contain an overt subject. Infinitivals, as in [i], belong
to the class of integrated relative clauses; they are discussed in Ch. 12, §5.
§ 9 Non-finite clauses as modifiers and supplements 1265
Gerund-participials and past-participials are semantically similar to relative clauses: com-

pare people who live near the site and a letter that was written by my great-grandfather.
We do not analyse them as relative clauses since there is no possibility of them contain-
ing a relative phrase (cf. ∗people who living near the site, etc.).
Past-participial modifiers are bare passives (Ch. 16, §10.1.1), as evident from the ad-
missibility of a by phrase in internalised complement function. Gerund-participials can
be active or passive. Passive gerund-participials contrast with the past-participials in
aspectuality as progressive vs non-progressive, but with actives the progressive vs non-
progressive distinction is lost:
[2] voice aspectuality form-type
i people earning this amount active neutralised gerund-participial
ii the amount being earned by Kim passive progressive gerund-participial
iii the amount earned by Kim passive non-progressive past-participial
The active neutralises the distinction between people who are earning this amount and
people who earn this amount. Features of form or context in particular cases may favour or
require one or other kind of interpretation – compare anyone knowing his whereabouts
(“who knows”) and Who’s the guy making all that noise? (“who is making”). The con-
struction itself, however, is quite neutral between the two interpretations. In the passive,
on the other hand, [ii–iii] contrast like the amount which is being earned by Kim and the
amount which is earned by Kim. Note, though, that we are concerned here with the se-
mantic category of aspectuality, not the syntactic category of aspect (see Ch. 3, §3, for this
distinction). It must be emphasised that being in [ii] is the passive auxiliary: as pointed
out in §1.1, gerund-participials do not accept the progressive auxiliary.
Modifiers in clause structure

[3] i They are saving up to buy a washing-machine.
ii They arrived home to find the house had been burgled.
iii He was a fool to say he’d go.
iv Liz was lying by the pool reading a novel.
The infinitival in [i] is an adjunct of purpose, while that in [ii] indicates a resultant or
subsequent situation. In [iii] it indicates the respect in which he was a fool, a reason
or explanation for the judgement. The gerund-participial in [iv] is a depictive adjunct,
giving descriptive information about Liz; note that it is interpreted with progressive
aspectuality: “she was reading a novel”.
Adjuncts that are integrated into clause structure as modifiers tend not to be sharply
distinct from complements: see §§4.4, 8.1.
Supplements
[4] i a. His hands gripping the door, he let out a volley of curses.
b. This done, she walked off without another word.
ii a. Realising he no longer had the premier’s support, Ed submitted his resignation.
b. Born in Aberdeen, Sue had never been further south than Edinburgh.
iii Whether working or relaxing, he always has a scowl on his face.
The underlined non-finites are supplements with the main clause as anchor. Those in
[i] contain a subject, and belong to what is known as the absolute construction, one
which is subordinate in form but with no syntactic link to the main clause. Those in
[ii] have no subject, and are syntactically related to the main clause in that the missing
subject is controlled by the subject of the main clause: it was Ed who realised he no longer
had the premier’s support, and Sue who was born in Aberdeen. In neither [i] nor [ii] is
there any explicit indication of the semantic relation between the supplement and the
anchor. This has to be inferred from the content of the clauses and/or the context. The
natural interpretation of the supplement in [ib], for example, is temporal (“when this
was done”), and of that in [iia] causal (“because he realised . . . ”). Both constructions
allow gerund-participials or past-participials – and also verbless forms, as exemplified
in §10. Example [iii] belongs to the exhaustive conditional construction discussed in
Ch. 11, §5.3.5.
While the missing subject in [4ii] is controlled by the subject of the anchor clause, we
also find supplements where it has to be interpreted non-syntactically:
[5] i To put it bluntly, they’re utterly incompetent.
ii But, judging from their reaction, the decision was a complete surprise to them.
iii Based on the latest inflation data, there’ll be another rate-rise soon.
Such supplements belong to the category of speech act-related adjuncts (Ch. 8, §18): they
are concerned with the manner in which the main assertion is expressed, or the evidence
for it. In [i–ii], the missing subject is understood by reference to the speaker, while with
the past-participial in [iii] it is the prediction of another rate-rise that is based on the
inflation data. The past-participial (which is less clearly established as grammatical than
the others) is more or less restricted to based on; for the relation between the gerund-
participials and deverbal prepositions, see Ch. 7, §2.3.
Non-finites can also serve as supplements to NP anchors:
[6] i Kim and Pat, both of them suffering from hypothermia, were winched into the
helicopter.
ii a. Kate’s proposal – to dismiss the manager – was greeted with dismay.
b. Jim’s hobby – collecting beermats – is taking up all his time.
c. There was only one thing to do: call in the police.
The supplement [i] is of the ascriptive type, comparable to a relative clause (compare
who were both of them suffering from hypothermia). Those in [ii], by contrast, are of
the content-specifying type: see Ch. 15, §5, for this distinction. Note that while proposal
licenses infinitival complements, hobby does not license gerund-participials. The supple-
ments here are thus sanctioned semantically (collecting beermats is a possible hobby),
rather than being lexically licensed.
10 Verbless clauses
We confine our attention here to verbless clauses in dependent or supplement functions

comparable to those realised by non-finite clauses, as described in the main part of the
chapter.
§ 10 Verbless clauses 1267
(a) Complement to with and without

[1] i They were standing against the wall [with their hands above their heads].
ii They were wandering around [without any clothes on].
iii [With the children so sick,] we weren’t able to get much work done.
iv Who is that guy [with his hands in his pockets ]?
The underlined clauses have subject + predicate structure, but with no verb in the
predicate. With and without do not license finite complements, but non-finites are found
in addition to the verbless forms (see §8.3 above). The bracketed PPs function as adjunct
to a clause ([i–iii]) or post-modifier in NP structure ([iv]). With is semantically similar
to have, and without to not have : [i–ii], for example, entail They had their hands above
their heads,They didn’t have any clothes on.52
(b) Complement to prepositions that license reducible clauses

We have observed that a number of prepositions like although usually take finite clause
complements, but allow the complement to be reduced – either to a non-finite clause,
as illustrated in §8.3, or else a verbless one, as in:
[2] i Although no longer a minister, she continued to exercise great power.
ii Once away from home, she quickly learned to fend for herself.
iii He spoke in an injured voice, as though resentful of the fact that she had not given
him proper warning.
iv He can be very dangerous when drunk.
v While in Paris, I visited Uncle Leonard.
A finite clause can be reconstructed by adding a subject and a form of the verb be:
although she was no longer a minister, and so on.
We noted in Ch. 12, §6.4, that when and whenever, while and whilst fall at the
boundary between relative words and prepositions taking content clause complements;
the fact that they enter into the present construction is a feature they share with the
prepositions taking content clause complements. Including them, the governing items
are:
[3] although as if as though if once
though when whenever while whilst
In addition to the construction illustrated in [2], if, when(ever), where(ver), and margi-
nally unless take necessary and possible as complements, as in:
[4] i Don’t hesitate to call me at home [if necessary.]
ii He was anxious to learn and helped me [wherever possible.]
The interpretation of the missing subject derives from a proposition expressed in the
matrix: we can expand as if it is necessary to call me and wherever it was possible to help
me. Note that the locative meaning has been bleached out of where in this example: it
could be replaced by whenever.
52
This use of on applied to clothing is one where the construction with be is somewhat unidiomatic: ? Their
clothes weren’t on.
(c) Supplements
[5] i His face pale with anger, he stormed out of the room.
ii The contestants, some of them primary school children, were kept waiting for two
hours.
iii The Chinese, whether drunk or sober, never kiss in public.
These are verbless analogues of the non-finite supplements given in [4i] and [6i] of §9.
Condensed structures and reanalysis

Verbless clauses may appear ‘condensed’, with the subject consisting of a single word
even when it is a count singular noun:
[6] i Dinner over, they resumed their game of chess.
ii He stood glowering at us, face red with anger.
iii They threw him head first into the pond.
Note that head first in [iii] is integrated into clause structure, having the status of modifier
rather than supplement.
A small number of expressions of this kind can function as predicative complement
as well as adjunct. Compare:
[7] i a. They walked away arm in arm. b. They were arm in arm.
ii a. He stood at the door, hat in hand. b. ∗He was hat in hand.
Predicative complement is not a function that normally accepts verbless clauses, and it
is plausible to suggest that arm in arm has been reanalysed, losing its status as a clause
and coming to be construed as a PP. The head will be in, with the second arm an internal
complement, and the first an external complement: see Ch. 7, §4.3.53 It is also possible
for such PPs to consist of an external complement NP + intransitive preposition, as in
face down (compare She was face down, but not ∗She was palms up). Further examples
are: side by side , back to back , inside out , upside down.54
11 Further remarks on the interpretation of subjectless non-finites
The subject is an obligatory element in canonical clauses, and the interpretation of

clauses which lack a subject requires that the omission be somehow made good. For
non-finite clauses we have distinguished three cases, involving raised, controlled, and
non-syntactic interpretations.
In the raised case, the missing subject is retrievable from the subject or object of the
matrix clause. Raised complements are licensed only by verbs and adjectives; they have
been discussed at length in earlier sections of the chapter and need not be considered
53
The structure proposed bears a significant resemblance to that of the verbless clause. Thus in is also the ultimate
head of hat in hand in [7iia], with hand an internal complement (in hand constituting the predicate) and hat
an external complement (more specifically, the subject).
54
Upside down, a reworking of earlier up-so-down, differs etymologically from the others in that upside
postdates – and is probably derived by back-formation from – the whole phrase. PPs of the type discussed
in Ch. 7, §4.2, with ago and apart as head (e.g. a week ago or this apart) also derive historically from absolute
clauses. Ago derives from agone, related to the past participle of go, while this apart is comparable to the dinner
over of [6i].
§ 11 Further remarks on subjectless non-finites 1269
further here. The relation between the other two types of interpretation, however, does
merit additional discussion.
The distinction between controlled and non-syntactic interpretations

The relation between a missing or covert subject and the controller is a special case
of anaphora. It is thus analogous to the relation between a personal pronoun and its
antecedent: compare Jill expected to finish on time and Jill expected that she would finish
on time (in the interpretation where she is co-referential with Jill ). We can thus use the
familiar notational device of co-indexing: Jilli expected [ i to finish on time]. It is not
only in the case of control, however, that a missing subject may be co-indexed with an
antecedent located elsewhere in the sentence. Compare:

[1] i a. Ii hope [ i to see her next week]. [controlled
b. Jill asked Pati [ i to help her]. interpretation]

ii a. This would involve [ moving to Sydney]. [non-syntactic
b. All Suei has had so far is a request [ i to accept nomination]. interpretation]
In [i] the underlined NPs are the antecedents, and more specifically the controllers,
of the missing subjects. In [iia] there is no antecedent at all, while in [iib] Sue is the
antecedent, but not the controller. It is understood that Sue is the potential accepter
of the nomination, but this interpretation is not determined by the syntactic relation
between Sue and the missing subject. It is arrived at, rather, by semantic inference along
the following lines:
[2] i The matrix clause (with specifying be) entails that Sue has had a request.
ii Sue therefore fills the semantic role of recipient of the request.
iii The understood subject of a non-finite complement to the noun request repre-
sents either the maker of the request (his request to see the files) or the recipient
(I received a request to make a donation to the Scholarship Fund).
iv Only the recipient is expressed in [1iib], and the content of the request makes it
more likely that the understood subject will represent the recipient of the request.
Similar factors are at work in:
[3] i Jilli found it difficult [ i to understand what he was getting at].
ii Maxi admitted it had been a mistake [ i to leave so little time for revision].
In [i] the infinitival is extraposed object in a complex-transitive construction: this makes
it the predicand of the adjective difficult. The missing subject in an infinitival predicand
of difficult is co-indexed with the NP with the semantic role of experiencer, and this
role is associated with the subject of find, i.e. Jill. Contrast this example with This made
it difficult to understand what he was getting at : make assigns the role of causer, not
experiencer, to its subject, so that this time we do not have co-indexing between matrix
and subordinate subjects. In [ii] the infinitival is extraposed subject and predicand of
a mistake: the missing subject represents the one who made the mistake, and admit in
the next higher clause suggests that it was Max who made the mistake. Note, however,
that it doesn’t have to have been Max who left so little time for revision. This is a likely
interpretation but not the only possible one: admitting something doesn’t entail that
one is responsible for it.
Control is the case where the identification of the missing subject can be described
by reference to syntactic functions rather than in terms of semantic roles. One of the
clearest cases is the catenative construction:

[4] i Kimi wants [ i to enter the competition]. [control by matrix subject]
ii Kim wants mei [ i to enter the competition.] [control by matrix object]
Here, as in almost all non-raising cases, the simple construction has control by the subject
of the matrix clause, while the plain-complex construction has control by the object.
To say that control is defined syntactically is not to suggest that it is semantically
arbitrary. On the contrary, it is strongly motivated by the semantics. Compare, for
example:
[5] i Sue told Timi [ i to arrange the interviews]. [control by matrix object]
ii Suei promised Tim [ i to arrange the interviews]. [control by matrix subject]
The infinitival subject is co-indexed with the object of tell and the subject of promise, but
this reflects semantic differences between tell and promise. Tell belongs with the set of
verbs of ‘influence’: it denotes an attempt to influence someone’s behaviour, and the one
(potentially) influenced corresponds to the missing subject of the infinitival complement.
Promise, by contrast, belongs with the verbs of ‘commitment’, and the missing subject
with such verbs corresponds to the one making the commitment. Control, then, is
semantically motivated, but these examples differ from those like [1iib] and [3] in that
the missing subject is anaphorically linked to an antecedent in a specified syntactic
function. Notice, moreover, that the use of promise shown in [5] is quite rare: it is much
more usual to have a simple catenative construction (Suei promised i to arrange the
interviews) or to have a finite complement (Suei promised Tim that shei would arrange
the interviews). In the overwhelming majority of cases, then, the controller for the plain-
complex construction is the matrix object.
A related difference between tell and promise concerns passivisation:
[6] i Timi was told by Sue [ i to arrange the interviews].
ii ∗Tim was promised by Sue i [ i to arrange the interviews].
Tell passivises readily, and it is now the subject that controls the missing subject of the
infinitival. But promise does not passivise: the antecedent for the missing subject in [ii]
is complement of the preposition by, and that is not a syntactically permitted function
for a controller.
Non-finite clauses functioning in NP structure

The interpretation of non-finite clauses functioning in the structure of NPs is deter-
mined very differently. We observed in §8.2 that raised complements do not occur in NP
structure, and the data in [7] show that we do not have syntactic control either:
[7] i Tim was satisfied with [Sue’s i promise [ i to pay the rent]].
ii Tim was satisfied with [the promise by Sue i [ i to pay the rent]].
iii Tim extracted from Sue i [a promise [ i to pay the rent]].
iv Tim will not be satisfied with [a mere promise [ to pay the rent]].
The missing subject of the complement of the noun promise is anaphorically linked to
the NP denoting the one making the promise, irrespective of the syntactic function of
that NP. And indeed, as we see from [iv], there does not have to be any such NP in the
sentence: it can be left to the context to determine the interpretation.
§ 11 Further remarks on subjectless non-finites 1271
As a second example, compare the verb request with the corresponding noun:
[8] i a. Kim requested us i [ i to enter the competition].
b. Wei requested [ i to enter the competition].
ii a. [Sue’s i request [ i to enter the competition]]has not yet been considered.
b. We i received a request from Sue j [ i /j to enter the competition].
c. We still haven’t received any requests [ to enter the competition].
The plain-complex catenative construction [i] has control by the object, with the object
representing the recipient of the request, while the simple catenative [ii] has control by
the subject, with the subject representing the maker of the request. But with the NP
construction there is no control, no anaphoric link between the missing subject and
an antecedent in some specified syntactic function. The antecedent may be in the NP
headed by request, or it may be in the matrix clause, or there may be no antecedent at all,
as in [iic]. Note, moreover, that [iib] is ambiguous, interpretable with either we or Sue
as antecedent: Sue may have been requesting us to enter, or requesting to enter herself.
The verb request is a marginal member of a small class of verbs that allow the normal
syntactic rule of control to be overridden under highly restrictive conditions. In We i
requested them i to be allowed to enter the competition, the antecedent is the subject,
not the object, as in [8ia]. This, however, is a very exceptional departure from the normal
pattern of object control, not possible beyond a very narrow range of cases like to be
allowed (see the discussion of Class 2 Ai in §5.3): it is not comparable with the situation
found with non-finites embedded in NPs.
The anaphoric relation between the antecedent and the missing subject
In examples where the antecedent is a simple NP such as a proper name, the missing
subject can be recovered quite straightforwardly: in Kimi remembered i to lock the
door, for example, it was Kim who locked the door. Where negation or quantification is
involved, however, matters are more complex:
[9] a. No onei intends [ i to harm you]. b. Both of themi hope [ i to speak first].
Example [a] does not mean “No one intends that no one should harm you”, and [b] does
not mean “Both candidates hope that both candidates will speak first”. We understand,
rather, “No one intends that he or she should harm you” and “Each of the two of them
hopes that he or she will speak first”. What is understood is something more abstract
than a repetition of the antecedent: we need to invoke the concept of variables, as in the
informal representations “No one x intends [x to harm you]” and “Both of them x want
[x to speak first]”. This, however, is not a special feature of subjectless non-finites, but
rather a quite general feature of anaphora, and for this reason further discussion can be
left to Ch. 17.
It is also a general feature of anaphora that we can have a sequence of links between
a missing element or pro-form and its antecedent, forming an ‘anaphoric chain’, as in:
[10] J ill i intends [ i to try [ i to mediate between them]].
The missing subject of mediate is controlled by the subject of try, but that itself is missing,
controlled by the subject of the next higher clause, with intend as predicator. Jill is thus
associated with the role of experiencer relative to intend, and agent relative to try and
mediate.

Nonfinite and Verbless Clauses PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Nonfinite and Verbless Clauses PDF

Загружено:

Авторское право:

Доступные форматы

14

Non-finite and verbless

5.6 Further remarks on the form-types 1240

1.1 Matters of form and function

complements of the preceding verb):

Form-types, auxiliaries, and voice

Interrogative and relative to-infinitivals

Distribution of non-finite clauses

1.2 The catenative construction

Simple vs complex catenative constructions

Four subtypes of the (active) complex construction can then be distinguished:

1.3 To-infinitivals with and without a subject

The catenative construction not an exception

(d) With some verbs the intervening NP is obligatory

1.4 The structure of infinitivals

1.4.1 The clause subordinator for

Syntactic reflection of the prepositional source of for

Reanalysis of for as subordinator

(a) Occurrence of to-infinitivals in non-PP positions

1.4.2 The infinitival subordinator to

Syntactic reflection of the prepositional source of to

to lend him the money

Head: Comp: Marker: Head:

to lend him the money to lend him the money

It is difficult to find compelling evidence to choose between these alternative analyses. It

subordinate clauses is precisely the property that we do expect in subordinators, and is a

We conclude that to is a VP subordinator. Our analysis for a to-infinitival clause con-

for you to lend him the money

1.4.3 Bare infinitivals

1.5 The structure of gerund-participials

Traditional grammar, we have noted, distinguishes between breaking in [iia], a gerund,

Complement and non-complement uses of gerund-participials

Nominal source of the traditional gerund

Analysis of the genitive NP as subject of the gerund-participial

The structure of the gerund-participial in [58ii] will therefore be as follows:

his/him constantly questioning my motives

Subject must be in absolute initial position

Comparison between genitive marking and the subordinator for

1.6 Case of the subject NP in gerund-participials

Genitive vs non-genitive subjects in gerund-participial complements

1.7 Understood subjects

2 The simple catenative construction

2.1 To-infinitivals (I hoped to convince them vs I seemed to convince them)

(e) Choice between simple and complex constructions.

Summary of differences between hope-type verbs and seem-type verbs

For footnotes 14–15, see next page.

2.2 Gerund-participials (We enjoyed sailing vs We kept sailing)

Passive infinitivals, selection restrictions, dummy subjects

Relation with finite complement construction

Choice between simple and complex constructions

2.3 Concealed passives (The house needs painting)

refer to this as the concealed passive construction. Compare:

3 The complex catenative construction

(a) Relation with finite complement construction

(b) Relation with passive infinitivals

(c) Selection restrictions

Summary of differences between persuade and intend

3.1.2 The for-complex construction (I arranged for her to go by bus)

Syntactic difference between catch and resent

We have no reason, therefore, to analyse the NP following resent as a raised object –

3.2.2 See (I saw him doing it)

4 The catenative complement as a distinct type of complement

4.1 Non-finites in relation to NP and AdjP complements