Complement clauses

Every language has verbs which introduce direct speech, reporting the actual words which may have been uttered. Thus:

(18) ‘Roosevelt has won another election,’ he announced

(19) She told me: ‘(You) pick up the towel!’

(20) ‘Put your bag in the top locker!’ he instructed me

(21) ‘Is the Saab back yet?’ Mary asked

(22) ‘Who left the window open?’ John asked

(23) ‘Mary shall lead the parade,’ Captain Smee decided

(24) ‘Jane was late again this morning,’ the office boy mentioned on Tuesday

(25) ‘John is a fool,’ declared Mary

Verbs such as announce, tell, instruct, ask, mention and declare always refer to some speech event. Decide does not have to, but it can be used in this way, as in (23).

Many languages, including English, have alternative ‘indirect speech’ constructions in which what was said is coded as a subordinate clause— called a ‘complement clause’—in syntactic construction with the verb of speaking. (In fact the phenomenon of complement clauses covers a good deal more than indirect speech, as we shall show.)

English has a variety of complement clauses. The most straightforward involves placing that before the ‘speech clause’, as in:

(18a) He announced [that Roosevelt had won another election]

(19a) She told me [that I should pick up the towel]

(20a) He instructed me [that I should put my bag in the top locker]

(23a) Captain Smee decided [that Mary should lead the parade]

(24a) The office boy mentioned on Tuesday [that Jane had been/was late again that morning]

(25a) Mary declared that [John was a fool]

Examining these carefully we see that present tense in direct speech (has won in (18), is in (25)) becomes past in indirect speech (had won in (18a), was in (25a)) if the main clause is in past tense. Present tense would be retained if the verb of speaking were also in present—compare ‘Mary is winning,’ John says and John says that Mary is winning. The actual past (was late) of direct speech in (24) may be either retained or replaced by previous past (had been late) in (24a). This tense replacement in indirect speech is called ‘back shifting’—see Quirk and Greenbaum (1973: 342V.).

Where the direct speech is an order or resolution, as in (19), (20) and (23), then the modal should is introduced in the THAT complement clause, as in (19a), (20a), (23a). Finally, deictic elements which were originally oriented to the pragmatic situation of the direct speech must be reoriented to the situation of the verb of saying: your in (20) becomes my in (20a); the optional—but always implicit—you in (19) becomes I in (19a); and this morning in (24) must be replaced by that morning in (24a).

Complement clauses which code questions begin with a wh- word in place of that, e.g.

(21a) Mary asked [whether/if the Saab was back yet]

(22a) John asked [who had left the window open]

If the direct question was of the polar variety (expecting ‘yes’ or ‘no’ as answer) then the complement clause will begin with whether or if; if the direct question began with who, whom, whose, what, which, how, why, where or when then this is retained at the front of the complement clause. Once again, present in direct speech becomes past in indirect (is in (21) corresponding to was in (21a) ) and actual past becomes previous past (left in (22), had left in (22a) ) if the main verb is in past tense.

A third variety of indirect speech construction is exemplified by:

(24b) The office boy mentioned on Tuesday [Mary’s having been late again that morning]

Here the subject of the complement clause takes possessive ending ’s and the first word of the VP is in -ing form (there is no tense inflection). Note that the previous imperfective auxiliary construction having been late in (24b) corresponds to actual perfective was late of (24).

A further construction type involves placing for at the beginning of the complement clause and to before the VP, which loses its tense inflection (the first word of the VP appearing in base form):

(23b) Captain Smee decided [for Mary to lead the parade]

Note that although the direct speech in (23) includes a modal shall, no corresponding form (shall or should) can be included in the complement clause of (23b). We refer to (23b) as a ‘Modal (FOR) to complement construction’.

The same construction type is used when a direct speech order, as in (19) and (20), is coded into indirect speech. Here the for must be omitted, giving a complement clause marked just by to; the subject of the complement clause is now also object of the main clause (here, me):

(19b) She told me to pick up the towel

(20b) He instructed me to put my bag in the top locker

If the direct speech concerns a judgement (which generally involves the copula be), as in (25), then this may be coded into indirect speech through a second type of to complement clause, which we call ‘Judgement to’:

 (25b) Mary declared John to be a fool

There is one further type of complement clause that involves to; it has a wh- word before the to, e.g.

(20c) He instructed me where to put my bag

This is not an exact correspondent of the direct speech sentence (20), but rather a description of it.

Many verbs of speaking can have an NP, giving the content of the speech act, in object function, e.g.

(18c) He announced [the election result]

(21c) Mary asked [a tricky question]

(23c) Captain Smee decided on [the order of procession]

(24c) The office boy mentioned [Mary’s latest misdemeanour]

Announce, ask, decide (on) and mention are transitive verbs. They may occur in opposition to a statement of direct speech, which is in lieu of a constituent in object function. Or they can take a syntactic object, which may be either an NP (as in (18c), (21c), (23c), (24c) ) or a complement clause (as in (18a), (21a), (22a), (23a/b), (24a/b) ). The verbs tell and instruct (in (19) and (20) ) already have an NP object and this can be followed by direct speech or an indirect speech complement clause; it is also possible to have a further NP in place of the complement clause, e.g. She told me my instructions, He instructed me in the proper procedure.

Verbs from a number of other semantic types, in addition to the SPEAKING type, may have a complement clause as alternative to an NP, in object or subject function, etc., e.g.

(26a) I believed John/John’s story

(26b) I believed that John told the truth

(26c) I believed John to have told the truth

(27a) The exam results delighted Mr Smith

(27b) His daughter’s having gained first class honours delighted Mr Smith

(27c) That his daughter had gained first class honours delighted Mr Smith

(28a) Mary began her lunch at noon

(28b) Mary began eating her lunch at noon

(28c) Mary began to eat her lunch at noon

It will be seen that some verbs allow several varieties of complement clause to fill a certain functional slot (sometimes with a substantial difference in meaning), e.g. wish, like tell, accepts both THAT and Modal (FOR) to clauses; mention and delight take THAT and ING; begin takes ING and TO; and plan and remember take all of THAT, ING and Modal (FOR) to. Other verbs only accept a single variety of complement clause, e.g. ensure with THAT, want with TO, and finish with ING.

There are of course a number of semantic types all of whose verbs demand an NP (not a complement clause) for each functional slot, e.g. hit, take, put, break, give.

The types which take complement clauses in object, subject, or some other functional slot are BEGINNING, TRYING, HURRYING, DARING, WANTING, POSTPONDING, MAKING, HELPING, SEEM, MATTER, ATTENTION, THINKING, DECIDING, LIKING, ANNOYING, ACTING, HAPPENING, COMPARING, RELATING and of course SPEAKING. We will deal fairly thoroughly with this topic, describing the meaning of each variety of complement clause, their syntactic statuses and the possibilities for coreferential omission etc., and the semantic basis for which verb occurs with which complements. The remainder here gives a preliminary sketch of the syntax of the seven varieties of complement clause in English.

(A) THAT complement clauses are exemplified in (18a), (19a), (20a), (23a), (24a), (25a), (26b) and (27c); and (B) wh- clauses in (21a) and (22a). Both of these have the full structure of a main clause, with obligatory subject, obligatory tense on the first word of the VP, and the full range of VP possibilities (modals, imperfective be, previous have, passive be). A THAT or wh- complement can have its subject coreferential with subject or object of the main clause, but it may never be omitted, e.g. I promised you that I would go; I promised you that you could go.

Besides introducing one variety of complement clause, that also functions as a deictic noun (Did you see that?), as a demonstrative determiner (Did you see that car?) and to introduce restrictive relative clauses, e.g.

(29) Everyone believed [the man [that had hired you]]

Note that here the relative clause that had hired you is a constituent of the object NP the man that had hired you. When (29) is passivized the whole object NP becomes passive subject:

(29a) [The man [that had hired you]] was believed (by everyone)

We mentioned that colloquial styles of English allow a restrictive relative clause to be moved from a subject NP to the end of the main clause, as in:

(29b) The man was believed (by everyone), [that had hired you]

Compare the relative clause marked by that in (29) with a complement clause introduced by that, as in:

(30) Everyone believed [that the man had hired you]

Here the clause that the man had hired you is object of believe, and can become passive subject:

(30a) [That the man had hired you] was believed (by everyone)

A THAT clause in (underlying or derived passive) subject function can be extraposed to the end of the main clause, its subject slot before the predicate being then filled by the impersonal form it:

(30b) It was believed (by everyone) [that the man had hired you]

Thus, although (29) and (30) have similar surface form, differing only in the order of the man and that, they do have entirely different syntactic structures and derivational possibilities.

Whether, if or a wh- word introducing a WH- complement clause can never be omitted. The complementizer that may often be omitted from a THAT clause which immediately follows the predicate, e.g. I think (that) he’s stupid, It was believed (that) the foreman had hired you, or from a THAT clause which immediately follows the object NP of a verb like promise or threaten (a verb for which the main clause object is not expected to be coreferential with the complement clause subject), e.g. I promised Mary (that) John could go. That can never be omitted from a complement clause in subject relation (and so not from (27c) or (30a) ) and is seldom omitted from a complement clause which follows the object NP of a verb for which there is an expectation that main clause object will be coreferential with complement clause subject, e.g. not from I instructed Mary that John should go.

That is unlikely to be omitted from a post-predicate complement clause if an adverb, a linker or a peripheral NP intervenes between it and the predicate—thus, that is generally retained in It was believed implicitly that the foreman had hired you and in It was believed, however, that the foreman had hired you. And that would generally be retained in (30b) if the agentive phrase by everyone were included. The determining factors on omission of that from a complement clause which immediately follows a predicate are largely stylistic—it is more likely to be omitted in casual than in formal speech, and more likely to be omitted if the reference of the complement clause is to some minor item of information rather than an important piece of reportage. Compare He announced (that) it was eggs for breakfast, where that is quite dispensable, with (18a) He announced that Roosevelt had won another election, where it would be unusual not to include that. A few verbs will almost invariably include that in a following complement clause, and this is because they carry a formal aura of meaning, e.g. require, propose, undertake, order, request (note that these verbs are likely also to include should in their THAT clause and the should can be omitted—see below).

Of the verbs which take THAT complements, there are a number which typically include a modal in the complement clause. The identity of the modal depends partly on the semantics of the main verb and partly on the choice of subject for the complement clause, e.g. I wish that I could . . . but I wish that you would ...

There is a group of verbs that commonly take should; indeed, the meaning of the verb implies obligation, requiring should in the complement clause. It is thus scarcely surprising that the should is generally omitted, producing what appears to be a tenseless, modal-less THAT clause (sometimes, unhelpfully, called ‘subjunctive’), e.g. order, command, suggest, propose, insist, require, and:

(31) She demanded that he (should) empty the bin

(32) He recommended that we (should) be told

(Note that should is not omittable from THAT complements with other types of verbs, i.e. not from She decided/knew/believed that he should do it.) Partial justification for saying that She demanded that he empty the bin involves an underlying should comes from the fact that empty is here in base form, which would be expected after a modal. Compare this with She says that he empties the bin, where the verb empties is in tense form, and must be the initial element of its VP.

Wh- words in English may have a number of functions: (a) they may be used as interrogatives, e.g. Who did you see?, What is that?; (b) they can introduce a wh- complement clause, as in (21a), (22a); (c) they may introduce relative clauses, e.g. I rather like the man who you married; (d) they can be the fusion of part of an NP and the introducer to a relative clause within that NP, e.g. she always takes what(ever) money I earn.

All of who, whom, whose, what, which, how, why, where and when have functions (a) and (b). All except how and what occur in function (c) (what is in fact used with this function in some low-prestige dialects, e.g. I like the car what you bought; for other dialects which effectively replaces what, functioning as NP head whereas in (a), (b) and (d) it is generally modifier to an NP head). In function (d) we find all except whose and why; -ever can be added to those wh- words that occur in (d), usually carrying a difference in meaning (compare (8a/b) and (9a/b)). Whether and if only have function (d), introducing complement clauses. If also functions as a clause linker. Whether is the only form in English whose sole function is to mark a complement clause.

Complement clause constructions and ‘fused’ restrictive relative clause constructions can appear very similar, as with:

(33) He ate what they put on the plate (relative clause construction)

(34) He knew what they put on the plate (complement clause construction)

They can be distinguished on semantic grounds. What in (33) is a fusion of that which—he ate ‘that’, which is something concrete, and whose referent is further specified by the relative clause which they put on the plate. Thus, ate in (33) has an NP (which includes a restrictive relative clause) as object; knew in (34) has a complement clause what they put on the plate as object—the speaker knows a fact, referred to by this clause. Note that know could alternatively have a THAT clause as object, e.g. He knew that they put apples on the plate; but eat could not—*He ate that they put apples on the plate is nonsensical.

(C) ING complement clauses are exemplified by (24b), (27b), (28b). The VP does not show tense inflection; instead, its first word is in -ing form. It may not include a modal, but can include aspect markers have or have plus be and/or passive be. The subject of an ING complement clause may be different from main clause subject and is then sometimes marked by possessive ending ’s (or, if a pronoun, it is in possessive form). But the subject can be coreferential with main clause subject, and is then usually omitted from the complement clause (since the VP in an ING clause is not tensed, it does not have to be preceded by a subject), e.g.

(35) I remember John’s/your winning the lottery

(36) I remember (?*my) winning the lottery

A fair number of verbs form a derived noun by the suffixation of -ing. It is important to distinguish between an NP with such a deverbal noun as head, as in (37), and an ING complement clause with the corresponding verb as predicate head, as in (38):

(37) I admired Mary’s singing of ‘Salty Dog’ in church

(38) I admired Mary’s singing ‘Salty Dog’ in church

There is a meaning difference—(38) states that I admired the fact that she did it (Mary’s temerity in giving voice to a bawdy song in a sacred place); (37) states that I admired the manner in which she sang (her syncopated style, etc.).

There are concomitant syntactic differences:

.The object in (37) has the structure of an NP, with Mary’s as determiner (the could be used instead) and with the preposition of introducing the post-head NP ‘Salty Dog’. In (38) Mary is the subject of the predicate singing (Mary’s here could not be replaced by the), with the object ‘Salty Dog’ immediately following the transitive verb.

.The noun singing in (37) could be modified by an adjective, e.g. quiet singing, whereas the verb singing in (38) would be modified by an adverb, e.g. singing quietly.

Of course a shorter version, I admired Mary’s singing, is ambiguous—the NP-as-object reading implies that I admired the way she sang, and the complement-clause-as-object reading indicates that I admired the fact of her singing.

The ’s on the subject of an ING complement clause is often omitted, especially in the most informal speech, and then the series I or unmarked form of the pronoun is used, i.e. him, her, etc., not he, she, etc. The ’s tends to be dropped most often from an inanimate subject, e.g. I don’t like the washing machine(?*’s) shuddering every time it hits ‘spin’ on the cycle. The ’s can be dropped from (24b), (27b), (35), (38) and from:

(39) I hate that man (’s) watching Mary

It was mentioned that there are constructions which appear to involve a reduced version of a restricted relative clause. Thus, corresponding to I hate that man who is watching Mary, we can have:

(40) I hate that man watching Mary

Examples (39) and (40) have quite different meanings: (39) states that I hate the fact of that man’s watching Mary, while (40) states that I hate a particular man, whose identity is specified to the hearer by the fact that he is watching Mary—the reason I hate him may be unconnected with his watching Mary.

When the ’s is omitted (39) coincides in form with (40) and I hate that man watching Mary is thus ambiguous between the readings ‘NP with restrictive relative clause relating to object’ and ‘complement clause as object’.

There are two kinds of complement clause that are similar in form but quite different in meaning. The first is introduced by for, includes to immediately before the complement clause verb and has a ‘Modal-type’ meaning (e.g. I intended for Mary to win). The complement clause subject may be omitted when it is coreferential with an NP in the main clause, and for is then automatically dropped (e.g. I intended to win). But for may also be omitted, in certain semantic circumstances, when the complement clause subject is retained (e.g. I intended Mary to win). The second kind has to immediately before the complement clause verb, but no for, and carries a ‘Judgement’ meaning (e.g. I consider Mary to be beautiful). Like ING clauses, both of these kinds of TO complement may include have or be auxiliaries, but not a modal or tense inflection.

(D) Modal (FOR) TO complement clauses are exemplified by (23b) and:

(41) [For John to have been so foolhardy] scares me

(42) I would love it [for Mary to sing ‘St Louis Blues’]

(43) Everybody agreed [for John to give the funeral oration]

The subject of a Modal (FOR) TO clause (together with the for) may be omitted when it is coreferential with a core NP of the main clause. The syntactic conditions for this omission are:

Scheme I is illustrated by I longed for Mary to win, I longed to win; II by For John to have to work Sundays annoyed Mary, To have to work Sundays annoyed Mary; III by I urged all the teachers for their pupils to take care crossing the street, I urged all the teachers to take care crossing the street. (There is a single addition to III, which only holds in some dialects of English. Promise will omit the subject of a post-object complement clause when it is coreferential with main clause subject, not with main clause object, e.g. I promised John for my charlady to clean his apartment, and I promised John to clean his apartment, from underlying I promised John for me to clean his apartment.)

A Modal (FOR) to clause may drop the for with the complement clause subject still being retained; this then becomes syntactic object of the main verb. Compare:

(44) I chose for Mary to lead the parade

(45) I chose Mary to lead the parade

In (45) Mary is object of the main clause verb chose (and can become passive subject, e.g. Mary was chosen to lead the parade). Note the difference in meaning between (44) and (45). It would be appropriate for me to say (44) if, as organizer of the parade, I communicated my decision to other people who would in turn inform Mary. But if I stood before the participants for the parade and pointed to Mary as I made the choice, then (45) would be the appropriate construction to use. (There is fuller discussion of the semantic role of for later)

For may only be omitted (and the complement clause subject retained) when it immediately follows a transitive verb. This correlates with the syntactic shift we have just observed—that the complement clause subject then becomes surface syntactic object of the preceding main clause verb. That is, for may not be omitted in any of the following circumstances:

(i) when the complement clause is in subject relation in the main clause, e.g. For John to be out so late worries Mary;

(ii) when the complement clause follows a main clause object, e.g. It worries Mary for John to be out so late, I told the captain for his men to clean the latrines;

(iii) when the complement clause follows an adjective, e.g. It is usual for a man to open the door for a lady .

(iv) when the complement clause follows it, e.g. I hate it for John to pick his nose in public (for may be omitted only when it is also dropped, e.g. I hate John to pick his nose in public, not *I hate it John to pick his nose in public);

(v) when the complement clause follows an adverb, e.g. I chose quite deliberately for Mary to lead the parade.

Some main verbs may not omit the for when complement clause subject is retained, e.g. decide, offer, remember, know. Some may optionally omit for, with a meaning difference, e.g. choose (as in (44/5) ) and propose, intend, mean, desire, wish. A third set generally omit for when the complement clause immediately follows the main verb, e.g. want in I want you to go, but will retain it when the verb and complement clause are separated, as when an adverb comes between them, e.g. I want very much for you to go or in constructions such as What I want is for you to go; other verbs behaving like this are need, order and urge. A final set must always omit the for (and thus cannot include an adverb immediately after the main clause verb), e.g. force, cause, allow, permit.

There are also verbs that take what is essentially a Modal (FOR) to complement clause but demand that the complement clause subject be identical to main clause subject and thus omitted (together with for), e.g. begin (as in (28c) ), try, hasten.

All of the four kinds of complement clause described so far—THAT, WH-, ING and Modal (FOR) TO—may fill object slot in the main clause and can then become passive subject, as in (30a) and:

(46) [Where he put it] is not known

(47) [John(’s) winning the lottery] is fondly remembered by all the office staff

(43a) [For John to give the funeral oration] was agreed (on)

(Many clauses with a Modal (FOR) to clause as object, and some with complement clauses of other kinds as object, do not readily passivize. This is for semantic—and not syntactic—reasons)

As already mentioned, once a Modal (FOR) to complement drops the for, then its subject becomes main clause object and it is this that becomes passive subject, not the whole complement clause (see (45) ).

THAT, WH- and Modal (FOR) TO clauses in (underlying or derived passive) subject function may be extraposed to the end of the main clause, with it filling the subject slot. Thus (30b) and:

(46a) It is not known [where he put it]

(41a) It scares me [for John to have been so foolhardy]

(43b) It was agreed [for John to give the funeral oration]

Extraposition is equally possible for a Modal (FOR) TO clause in subject relation that has its subject (and the preceding for) omitted under identity with the main clause object:

(48) [To have to give the welcoming speech] terrified me

(48a) It terrified me [to have to give the welcoming speech]

Note that a complement clause may not usually be extraposed over another complement clause. That is, a clause in subject relation is not open to extraposition if there is also a complement clause in object relation to the verb, e.g. corresponding to:

(49) [That Pegasus won the race] indicates [that Bellerophon is a fine rider]

it is not possible to say:

(49a) *It indicates [that Bellerophon is a fine rider] [that Pegasus won the race]

ING clauses are generally not extraposable. However, we can get what is called ‘right dislocation’, in which the two clauses are set off by contrastive, appositional intonation (and it refers to ‘Mary’s singing in church’), as in:

(38a) It was admired, Mary’s singing ‘Salty Dog’ in church

But this is a different grammatical phenomenon from extraposition.

There are just a few verbs that omit to from a Modal (FOR) to complement in post-predicate position. They are make, let and the causal sense of have from the MAKING type (compare They made John go with They forced John to go, which retains the to); a subset of ATTENTION verbs, which take a complement similar in form to Modal (FOR) TO (e.g. They saw John swim); and just know from the THINKING type (e.g. I’ve never known him ask such a question before). Note, however, that the to must be included in the corresponding passive, e.g. John was made to go, John was seen to swim.

(E) Judgement to complement clauses are exemplified by (25b), (26c) and:

(50) I noticed John to be sleeping/incompetent/badly bitten

All Judgement TO constructions involve a transitive verb, an object NP (which is simultaneously subject of the complement clause), and to followed by the complement clause predicate. This predicate most often includes be, which can be imperfective auxiliary, passive auxiliary, or copula, although other possibilities do occur (as in (26c) where the predicate begins with have).

Judgement TO constructions appear syntactically similar to a Modal (FOR) TO clause in object relation, with for omitted but complement clause subject retained, e.g. I told John to sleep. There is a critical semantic difference: if there is a THAT complement corresponding to a Modal (FOR) TO clause then it will generally include a modal, e.g. I told John that he should sleep; but if there is a THAT complement clause corresponding to a Judgement TO complement then it cannot include a modal, e.g. I noticed that John was sleeping.

To does of course have other functions in the grammar of English— introducing a recipient NP (give it to John) or an NP referring to a destination (go to Moscow). It is also a (preferred) short form of the clause linker in order to, as in (51), which on the surface looks similar to the complement clause construction (52):

(51) He went to bathe

(52) He likes to bathe

There are profound syntactic differences. Go in (51) is an intransitive verb and its clause is linked by (in order) to with the clause bathe. Like in (52) is a transitive verb (e.g. He likes the seaside) and here to bathe is a complement clause; alternatives would be He likes bathing, He likes Mary to bathe, etc. The semantic type of the first verb in a construction like (51/2) enables one to infer the kind of construction that is involved.

(F) WH- TO could perhaps be regarded as a subtype of Modal (FOR) to complements. A wh- word (other than why or if) simply precedes the to (the VP has the same possibilities as in ING, Modal (FOR) to and Judgement TO clauses). Thus (20c) and:

(53) I don’t know who to blame

(54) They remembered where to look

(55) I’ll choose when to go

WH- TO clauses do not have any independent subject; their subject must be coreferential with the subject of the main clause or with its object (e.g. I told John who to see). A wh- element such as who (as in (53) ) or what can refer to any constituent of the main clause except its subject.

(G) FROM ING complements are parallel to a variety of Modal (FOR) to clause, in object relation, that omits for but retains the complement clause subject. Compare:

(56) John forced him to open the door

(57) John prevented him (from) opening the door

(58) John persuaded him to see the doctor

(59) John dissuaded him from seeing the doctor

Him has the same syntactic status in all of (56)–(59); it is underlying subject of the following verb, and is also object NP for the main verb, which may become passive subject, e.g.

(56a) He was forced to open the door

(57a) He was prevented from opening the door

The VP in a FROM ING clause has its first verb in -ing form. It may include previous have, imperfective be, and passive be but not a modal or any tense inflection (this is the same as for ING and the various kinds of TO complement). Interestingly, the from may optionally be omitted following a verb of negative causation (prevent, stop, save) when used in the active, but is retained in the passive—compare (57) and (57a). It can never be dropped after a verb from the SPEAKING type (forbid, discourage, dissuade, prohibit). The semantic conditioning of from omission.

ING, FROM ING and all kinds of TO complement clause may not have tense inflection in their VP. How do they convey information about time reference, which is coded through tense in main clauses and in THAT and WH-complements? It appears that the auxiliary element have, which normally indicates previous aspect, here doubles for past tense and previous aspect; actual present is marked by the absence of have. Compare the THAT constructions in:

(60) I believe that Mary eats mangoes

(61) I believe that Mary ate the mango (yesterday)

(62) I believe that Mary has eaten the mango (that you brought)

(63) I believe that Mary had eaten the mango (that you left there)

with the corresponding TO constructions:

(60a) I believe Mary to eat mangoes

(61a/62a/63a) I believe Mary to have eaten the mango

Here have in the TO construction corresponds to all of actual past (in (61)), previous present (in (62)) and previous past (in (63)). (This is similar to the ‘back-shifting’ of indirect speech, mentioned at the beginning) An identical correspondence applies in ING clauses—all of I mentioned that Mary ate/has eaten/had eaten the mango correspond to I mentioned Mary’s having eaten the mango; see also (24b).

The auxiliary be carries over its imperfective meaning into all varieties of complement clause, e.g. I notice that Mary is eating mangoes (every time we go past) and I notice Mary to be eating mangoes (every time we go past).

There are a number of instances in English where a verb base plus preposition effectively functions as a complete lexical verb, e.g. decide on, hope for, complain about, rely on, agree on, object to. The preposition is retained before an NP object or an ING complement clause as object but must be omitted—by an automatic rule of English grammar—before complementizers that, for and to (no verb of this kind occurs with a FROM ING complement clause). The preposition may be omitted or retained before a complementizer beginning with wh-.

(64a) He decided on John

(64b) He decided that John would be captain

(64c) He decided (on) who would be captain

(65a) She complained about the interruption

(65b) She complained about John’s interrupting her recital

(65c) She complained that John interrupted her recital

(66a) I’m hoping for a promotion

(66b) I’m hoping to get promoted

(66c) I’m hoping that I’ll get promoted

A number of HUMAN PROPENSITY adjectives may take a preposition and then an NP or complement clause, e.g. afraid of, sorry about, proud of, jealous of/about, careful about. Once again, the preposition is retained before an NP or an ING complement, but dropped before that, for or to, e.g.

(67a) John is afraid of the dark

(67b) John is afraid of going out in the dark

(67c) John is afraid to go out in the dark

(67d) John is afraid that it will be too dark

(68a) Mary is sorry about her mistake

(68b) Mary is sorry about making an error in the calculation

(68c) Mary is sorry that she made an error in the calculation

Table 2.5 summarizes some of the more important syntactic features of complement clauses. It will be seen that (A) THAT and (B) WH- have almost identical syntax; they differ only in the minor detail, under 10, that a

preposition may optionally be retained before wh-. Varieties (E) Judgement TO, (F) WH- TO and (G) FROM ING also show strong similarities.

ING clauses stand apart from the others. The fact that an ING clause is not readily extraposable from subject position, in 9, and that it does not easily take a preceding adverb, in 6, suggests that ING clauses function more like NPs than do other varieties of complement clause. Be that as it may, an ING clause is quite different in structure from an NP, as we saw when comparing (37) and (38).