The Grammar of English

Non-Finite Clauses

Reading: Huddleston & Pullum (2005), A Student's Introduction to English Grammar, chapter 13

Structure and function What time and location do you associated with the clause for Madonna to write a novel? There's no answer to that: no time, no place. It hasn't happened, anywhere; it might never happen; though in principle it could. But the thing is, the construction isn't designed to tie down the situation to a time and place. The ‘for NP to VP’ construction is non-finite in the sense of having no temporal or spatial bounds, the way infinity has no temporal or spatial bounds. That's the way the traditional notions ‘finite’ and ‘non-finite’ arose.

When you want to take a potential situation, like that of Madonna writing a novel, and tie it down to a particular time and place, you have to choose a tense, and maybe a temporal or locative adjunct or two, and you get a finite clause like Last year Madonna wrote a novel. That expresses a proposition, and makes a claim (in this case the claim is, I believe, false; but false claims are claims too). Non-finite clauses aren't supposed to make claims that can be true or false, and they usually couldn't be said to do that, though they can participate in expressing such claims.

There are four kinds of non-finite clause in English:

to-infinitivalto write a novel    plain form
bare infinitivalwrite a novel plain form
gerund-participial   writing a novel gerund-participle
past-participialwritten a novel past participle

A few notes on the special form of non-finite clauses:

According to CGEL, there are two special subordinators that occur introducing non-finite clauses: to and for.

To is a real problem. I argued in an earlier lecture that its behavior fits very nicely with the hypothesis that it is a defective auxiliary verb that has only one form, the plain form. And it occurs at the beginning of the VP when there is a subject, not before the subject. Yet its semantic inertness, and occasional optionality (He helped to clear up = He helped clear up), and its failure to occur in independent clauses all make it look like a subordinator.

It's fairly tricky to determine that for is a subordinator and not a preposition, but that is the judgment we reached when writing The Cambridge Grammar. We could be wrong here too, but we did base our decision on evidence. We believe the following:

When a non-finite clause is subjectless, there has to be a way of understanding its predicate. A subjectless clause serving as a complement generally gets its meaing determined by properties of the lexical head that it is a complement of (I'll illustrate immediately below, and return to it next time). But there are also non-finite clauses used as adjuncts. That is how we get to our next topic: dangling modifiers.

The problem of dangling modifiers

Non-finite clauses can function as adjuncts, and this means that when they have no subject there is a possible difficulty about understanding them: non-finite complement clauses are under specific syntactic restrictions that say how the missing subject is to be filled in, but adjuncts aren't.

For example, with try the complement subject can only be understood as being a property of the same entity that the matrix clause VP is a property of, so in Bugsy says Tony thinks Lou is gonna try to escape tonight, the only escapee we are talking about is Lou; it cannot mean that Lou is going to try to spring Bugsy out of prison.)

Adjuncts aren't under the same requirement, so there is a special problem if the subject is missing. This leads to the dangling problem.

The problem of dangling is generated by the need hearers and readings have to find a predicand. Most commonly they look to the matrix clause subject. In some cases that produces unintendedly hilarious misunderstandings. In other cases it somehow passes almost without notice. Every serious style handbook has an article on the topic with examples. Look up "dangling" (it may direct you to somewhere else, like "modifier, dangling"). Some colleagues and I have been collecting examples for some time. Here are a few.

  1. Once having achieved this stage, the prognosis is poor, although even with microalbuminuria cardiovascular complications occur frequently. (, found by Jason Grafmiller 3/07)
  2. President Johnson, in particular, intuited such weakness [in his advisers]. He could butter a prospective adviser like a turkey. But once under his sway, his abuse was considerable. Michael Powell, "Managing Up, Down and Sideways", NYT Week in Review, 10/7/07, p. 3:

  3. Ben Goldhirsh founded GOOD Magazine and Reason Pictures in 2004. Starting from the back of a small office in Los Angeles, momentum has been building steadily as a growing team works to craft their vision of GOOD. ["Our very brief history" on the website for GOOD Magazine:]

  4. In looking through the tests, it was difficult to assign a classification of "ungrammatical" or "purely stylistic variation" to each question... [Draft of a student paper, 10/9/07]

  5. While not offering any opinion as to whether Mr Rudd would make a better prime minister than Mr Howard, it is clear that Labor is in the box seat going into the final week of the campaign. [Weekend Australian 17-18 Nov 07: 18]

  6. The surprising thing about this discussion, at least sitting here in snowy Madison, Wisconsin, is that nobody has mentioned this as a basic characteristic of restaurant names.

  7. [Magnet Island] is as pretty as I remember it from my childhood, when a trip here in the family boat was a welcome reprieve from the stifling Townsville heat. As a teenager, the island held even more allure. [Sunday Mail, Discover: 27.01.08]

  8. Dear Prof. Pullum,

    As staff member of the Department of Linguistics and English Language of the University of Edinburgh, it is with great pleasure that I write to inform you about the Summer Sessions being offered by the ICD Academy of Cultural Diplomacy, Berlin. These sessions may be of particular interest to students looking to build their resumes during the summer break or those coming to Germany on exchange. . . .

    [Email from the Insitute for Cultural Diplomacy]

Catenatives, raising, and dummies

Catenative complements are probably the most frequently encountered type of non-finite clausal internal complement in English. Here are some of their properties:

There isn't really a limit to how many catenative complements you could chain together. Here's an example with a chain of at least eleven verbs, a matrix verb and ten catenative complements:

They [ seemed to [ want to [ try to [ claim to [ have [ avoided [ appearing to [ hope to [ manage to [ aspire to [ own anything. ] ] ] ] ] ] ] ] ] ] ]
In fact if we treat to as some kind of uninflectable and meaningless verb, it has a chain of 18 verbs:
They [ seemed [ to [ want [ to [ try [ to [ claim [ to [ have [ avoided [ appearing to [ hope [ to [ manage [ to [ aspire [ to [ own anything.] ] ] ] ] ] ] ] ] ] ] ] ] ] ] ] ] ]

Catenative complements sometimes co-occur with (and follow) an object NP associated with the matrix verb. We will refer to a clause where there isn't an object NP as the simple catenative construction; when there is an object, we have a complex catenative construction. With at least one verb, help, we can illustrate both: They helped clear it up illustrates the simple catenative and They helped us clear it up illustrates the complex catenative.

In matrix clauses immediately containing a catenative complement we have to distinguish the cases of raised subjects from the case of ordinary subjects. This is an important topic, and not always easy to grasp at first. As a simple first approximation, the raised subject constructions are the ones that allow dummy subjects and idiom chunk subjects. The ordinary constructions don't. These contrasts illustrate:

It seems to have been raining.     *It hopes to have been raining.
It seems to be all right to smoke.     *It hopes to be all right to smoke.
It still seems to amaze everyone the things she can do.     *It still hopes to amaze everyone the things she can do.
There seems to be nothing that we want to change.     *There hopes to be nothing that we want to change.
Close tabs seem to be kept on immigrants.     *Close tabs hope to be kept on immigrants.
Lip service seems to be paid to democracy there.     *Lip service hopes to be paid to democracy there.
Strings seem to be pulled to get him regular contracts.     *Strings hope to be pulled to get him regular contracts.

Here is another test, a more more complicated. If you

you should get a synonymous result in the case of a raising construction (because which NP is "raised" into the matrix clause makes no semantic difference given that the matrix verb is not transitive).

You should get a non-synonymous result in the case of an ordinary construction (because which NP is the matrix subject makes a big difference to the semantics).

I'll illustrate with the names Arthur and Betty:

Arthur seemed to convince Betty.     Arthur hoped to convince Betty.
Betty seemed to be convinced by Arthur.     Betty hoped to be convinced by Arthur.

In the second case it is crucial to know who was the hoper, Arthur or Betty: one of them was hoping for it to become true that Arthur would convince Betty, and which one that was is a very significant question. But in the first example it merely seemed to be the case that Arthur convinced Betty, which is the same thing as saying that it seemed to be the case that Betty was convinced by Arthur. All the action is in the subordinate clause, and semantically it doesn't much matter whether it is made passive or not.

These are the two key tests to be used in distinguishing ordinary from raised subjects. Altered versions of them also work for distinguishing ordinary from raised objects. For example:

We hold you to be responsible.     We urge you to be responsible.
We considered it to be stupid to resist.     *We urged it to be stupid to resist.
I believe there to have been foul play.     *I forced there to have been foul play.

Hold and believe take raised objects, and whether those objects are dummies or not doesn't matter; urge and force take ordinary objects, which have to denote entities that it is sensible to exhort or compel to engage in certain actions.

The passive test works too. We choose a transitive clause like the doctor examine the baby as the complement clause, and try out both the active and the passive variant after a raised-object verb and an ordinary-object verb:

We believed the doctor to have examined the baby.     We urged the doctor to examine the baby.
We believed the baby to have been examined by the doctor.     ??We urged the baby to be examined by the doctor.

With believe, a raised-object verb, the semantic roles of the doctor and the baby are purely within the subordinate clause, and whether we make that clause passive or not doesn't matter much for the truth conditions. But urge takes an ordinary object, which must denote the person at whom the urging is directed. And one tends to devote one's urging toward people who known how to do things, so ??We urged the baby to be examined by the doctor seems bizarre, and not at all like the very sensible We urged the doctor to examine the baby.

Linguistic literature note: There is a large literature on the issue of raising verbs and the ordinary kind, also known as control verbs. Perhaps the most important such work is Paul Postal's book On Raising (MIT Press, 1974), which is entirely devoted to establishing that the right description of these verbs involves a transformational rule that raises object NPs out of complement clauses and into matrix object positions.

In the terms assumed here, the issue is about the grammatical functions of certain NPs and the semantic interpretations assigned to them, rather than about rules moving objects around, but the same arguments are relevant in each case. For example, Postal points out that manner adverbs are allowed between a verb and its clause complement, but never between a verb and its object:

I believe firmly we have a real chance to get this contract.
*I believe firmly my wife.
So, he points out, we need simply look at this example to get a new argument regarding the syntax of sentences with believe:
*I believe firmly my wife to be innocent.
It is ungrammatical, so that suggests that my wife is an object — and that my wife to be innocent is not a clause. The ungrammatical structure looks like this:
    /  \
   /    \
Subj:  Pred:
 NP     VP______
 |      /|\     \
 |     / | \     \
 |    V Adv \     \
 |    |  |   \     VP
 |    |  |    \    |\
 |    | firmly \   | \
*I believe      NP |  \
                /\ to  \
               /__\     \
              my wife    VP
                        /  \
                     be innocent
But the grammatical structure looks like this:
    /  \
   /    \
Subj:  Pred:
 NP     VP
 |      /|\
 |     / | \
 |    V Adv \
 |    |  |   \ 
 |    |  |    \ 
 |    | firmly \
 I believe      Clause  
                 / \
                /   \    
             that   Clause 
                    /  \
                   /    \
                 my wife is
There is no prohibition against an adverb preceding a clause in the VP.

Verbless clauses

There are a very few verbless constructions in English that should probably be regarded as clauses. One is seen in the underlined part of With you in Colorado, it's very lonely, or While there, he caught pneumonia, or They found a photo of him without his shirt on, or The residents, many of them elderly, were forced to flee.

Possibly these should be analyzed as having non-VP predicates:

    /  \ 
   /    \
Subj:  Pred:
 NP     PP
 |      /\
 |     /  \
 |  Head: Comp:
 |    P    NP
 |    |    |
 |    |   Head:
 |    |    N
 |    |    |
you  in Colorado

It isn't clear. They may not all have similar analyses. Perhaps you in Colorado might be an NP with a PP modifier:

    /  \
   /    \
Head:   Mod:
  N     PP
  |     /\
  |    /  \
  | Head: Comp:
  |   P    NP
  |   |    |
  |   |   Head:
  |   |    N
  |   |    |
 you in Colorado
But surely medical help unavailable seems like a clause, with a subject and a predicate.

If there are indeed verbless clauses, there have to be various constraints to prevent them from turning up elsewhere, as main clauses or complements.

Last updated Thu 1 Nov 2012 09:51:43 EDT by GKP.