Linguistics 522

The Structure of Clauses

Chapter 6 (Carnie)

Finite and non-finite clauses

Present and
past forms

The following clauses have main verbs in the present tense:

  • John walks.
  • John and Mary walk.

The following clauses have main verbs in the past tense:

  • John walked
  • John and Mary walked.
Number
Agreement

Both present and past tense verb forms vary according to the number of the subject. In the present tense, regular verbs do. In the past tense only the irregular verb be does:

  • singular subject: John walked
  • plural subject: John and Mary walked.
  • singular subject: John was a clown.
  • plural subject: John and Mary were clowns.
Finite
Verbs

We call the present and past tenses of verbs and modals finite forms.

  • All finite verbs agree with their subjects for number.
  • All finite verbs assign nominative case to their subjects.
Finite
Clauses
We call any clause wose main verb is a finite verb a finite clause.
Non-finite
verbs

Verbs that aren't either present or past and don't agree with their subjects for number are called non-finite forms. English has three non-finite forms.

  • present participle: walking
  • past participle: walked
  • infinitive: to walk
What we'll call the bare infinitive form is just the infinitive form without the the to:
  • bare infinitive: walk

All ordinary verbs have the full inventory of non-finite forms. Modals never have non-finite forms:

  • * maying
  • * to may
  • * mayed
  • * toed/toing (attempt to form nonfinite forms of infinitivalT marker "to")
Non-finite
clauses
We call any clause whose main verb is a non-finite verb a non-finite clause.
  • present participle: Mary kept John walking
  • past particple: Mary had her dog walked every day.
  • infinitive: Mary would like for John to walk.
  • (so-called) bare infinitive: Mary made John walk.
Tense and
number
tests
Non-finite verbs don't show tense or agreement:
  • Mary would like for John to walk.
  • * Mary would like for John to walks.
  • * Mary would like for John to walked.
Remember present tense plural forms (walk) can look just like non finite forms (walk). To show a verb is tensed and prove a clause is finite, put the verb into the past tense or use a singular subject in the present tense.

Modal
Test

A non-finite clause will not allow modals, since modals lack non-finite forms:

  • * Mary would like for John to may walk.
  • * Mary had John mayed walk every day.
  • * Mary kept John maying walk.
Subject
Test

Only non-finite clauses can be subjectless:

  • John keeps smiling.
  • John wants to smile.
  • John felt frozen.
Another way of putting this is that only non-finite verb phrases can occur without subjects.
Nominative
Case

Finite clauses assign nominative case to their subjects:

  1. I left.
  2. *Me left.
  3. *My left.
The interesting property, which gives us another test for finite clauses, is that only finite clauses assign nominative case to their subjects.
  • * Mary would like for I to walk.
  • Mary would like for me to walk.
  • * Mary had I walk every day.
  • Mary had me walk every day.
  • * Mary kept I walking.
  • Mary kept me walking.
  • Mary believed that I was coming.
  • * Mary believed that me was coming.

Clause Constituents

Complementizers
Two analyses

Complementizers (shown in italics) are clause-introducing particles.

  • We believe that the mayor will win the election.
  • We all wondered whether the mayor would win the election.
  • I for one would prefer for the mayor to win the election.
Empty
complementizers

In this text we have developed a theory which allows trees with empty elements.

In English, overt complementizers occur only in embedded clauses. But we will say that all clauses have complementizers and that some of these complementizers are covert or empty.

What does this mean?

What we're saying really is that there's a complementizer position in the clause, and it always has an effect, whether it's filled or not. We have also claimed that other things besides complementizers may fill this position (T_>C movement in Chapter 8).

We now present some arguments for this view, first to arguments for main clause complementizers.

Main
Clause
Complementizers 		 

Many languages have main clause complementizers in particular constructions.

  1. Question particles. These are often clause-initial and clause-final. Examples in (30) p. 296.
  2. Exclamative constructions. "What a traet!" (French)
      [French example]
  3. Optative constructions. This is an imperative-style construction that means, roughly, "let it be so." Examples (34), p.297.
      [French example]

    In English we find that auxiliary inversion is always in complementary distribution with the occurrence of complementizers (the two never occur together). Some examples:

    1. semi-indirect speech
      1. John wondered whether he would get a degree. (ordinary embedded question)
      2. John wondered would he get a degree (semi-indirect speech)
      3. * John wondered whether would he get a degree.
    2. subjunctives with inversion:
      1. One must be vigilant, whether it be at home or abroad. (subjunctive with complementizer)
      2. One must be vigilant, be it at home or abroad.(subjunctive with complementizer)
      3. * One must be vigilant, whether be it at home or abroad.(subjunctive with complementizer)
    These curious facts are explained if we assume that inverted auxiliaries fill the complementizer position, and that once the position is filled, it is no longer available to be used by a complementizer.

    Note that this explanation depends on talking about a structural position in the clause. Once we adopt the idea of a fixed complementizer position, then the additional idea that the position may be filled or unfilled does not seem so strange.

Coordination
Test for
CP 		 

We might still object that we've argued that inverted S's and S's with overt complementizers should be CPs, but not that S's with neither inversion nor overt complementizers are. In other words, what motivates ever having an empty Comp? Answer: Coordination of S's withc omplementizers with those without complementizers:

  • John believes the bill is well-crafted and that the president will act to support it.
  • There is good reason to intervene in this instance, but do we have the resources to carry it through?

Complementizer
Classes 		 

We classify complementizers two ways, according to whether they take questions or declaratives, and according to whether they take declaratives or questions.

  1. Mary should know that the president will intervene.
  2. * Mary should know that the president to intervene.
  3. * Mary should know that to intervene.
  4. Mary should know whether the president will intervene.
  5. * Mary should know whether the president to intervene.
  6. Mary should know whether to intervene.
  7. Mary should know who the president will choose.
  8. Mary should know if the president will intervene.
  9. * Mary should know if the president to intervene.
  10. * Mary should know if to intervene.
  11. * Mary should is anxious for the president will intervene.
  12. Mary should is anxious for the president to intervene.
We describe the complementizer properties with the features ± FIN and ± WH (question or non-question)
  1. whether: + WH, ± FIN
  2. if: + WH, + FIN
  3. that: - WH, + FIN
  4. for: - WH, - FIN

Analysis of TP

T
We assume rules for TP like this one:
  • TP => NP T'
  • T' => T VP
This only works for finite clauses, because modals are always finite. To generalize to non-finite clauses, we propose a category T (for "tense"), which will dominate both modals and the infinitival to. These two instances of T look like this:
  1. for [TP [NP John] [T' [T to] [VPleave] ] ]
  2. that [TP [NP John] [T' [T will] [VPleave] ]]
Thus the category T can be both finite and non-finite.

Arguments that to should be conflated with modals:

  1. Both allow VP ellipsis
    1. Anyone who wants to [VP Ø ] can read Proust.
    2. She wants to read Prous and she will [VP Ø ].
  2. to never co-occurs with modals:
    1. * to can go vs to be able to go
    2. * to must go vs to have to go
  3. To can precede negation:
    1. He wanted to not smile.
    2. He vowed to not give in.
    3. He may not go.
    4. He may not give in.

We noted above two important properties restricted to finite clauses:

  1. Only finite clauses assign nominative case to their subject.
  2. Only in finite clauses do subjects agree with verbs.
We will link both these properties to finite T. (to T with the feature +TNS)
  1. An NP which is sister to +TNS T' is assigned Nominative case.
  2. An NP which is sister to a +TNS T' must agree with it in number and person.
Here "NP which is sister to T" is just a structural definition of "subject."

Something that needs to be spelled out is how a subject can agree with T in number and person. In order for this to happen, the T node must be specified for number and person as well.

  1. T ==> [Alpha TNS, Alpha Agr]
  2. + TNS ==> ± PAST
  3. + AGR ==> [Beta Number, Gamma Person]
These are rules not for spelling what T dominates (like phrase-structure rules), but for what features it can be made up of. Here the Greek letters (Alpha, Beta, Gamma) stand for either + or - with TNS and Agr and different number and person values with Number and Person. The first rule says that T consists of a TNS (tense) and AGr (agreement) feature and that these have to have the same value; either they're both + or they're both minus. The second rule says that + TNS can be spelled out as + or - PAST, and that + AGR is spelled out as some number value and some person value.
Complement-
izers
and T

As noted above, complementizers seem to determine whether the T of the following S is finite or non-finite;

  1. We are anxious for John to leave.
  2. * We are anxious for John should leave.
  3. We are anxious that John should leave.
  4. * We are anxious that John to leave.
  5. We wonder whether/if John will come.
  6. We wonder whether to come.
  7. *We wonder if to come.
Empty T
has affixes
We assume that T may be empty too. But even when empty it is still specified for finiteness. WE account for this by saying that, in English, the verbal affix starts out in T and lowers onto the verb.

Our analysis of ordinary tensed clauses without modals is that they have T nodes specified +TNS.

  • [S [NP John] [T[+TNS] e] [VPleaves] ]
This means that a full CP for a finite main clause has two empty nodes:
  • [CP [[Comp[+FIN, -WH] e] [TP [NP John] [T' [T[+TNS] e] [VPleaves] ] ]]
-ing,-en  

We have two kinds of verbal forms out of the picture thus far, present participles and past participles:

  1. present: break + ing = breaking
  2. past: break + en = broken
We are treating the inflections -ing and -en differently from other inflections on verbal forms (person and number). Person and number inflections like -s and -ed are dealt with in T. But -ing and -en must be dealt with elsewhere. Is this right?

Arguments for this.

  1. None of the verbal forms that show up under T inflect for -ing and -en.
    1. *mayen, *maying
    2. *toen, *toing
    3. *shoulden, shoulding
  2. We assume one verb in T at a time and -ing and -en forms can appear when something already fills T:
    1. It would be a pity for you [T to] be working.
    2. It would be a pity for you [T to] have been working

Note that this treatment of -ing and -en as special goes hand in hand with treating the aspectual auxiliaries have and be as distinct from the other auxiliaries. For one thing they can co-occur with overt elements in T, unlike other elements of T.

  1. Mary would prefer for John [T to] be seeing a dentist.
  2. Mary would prefer for John [T to] have seen a dentist.
  3. * Mary would prefer for John [T to] may see a dentist.
  4. Mary [T may] have seen a dentist.
  5. Mary [T will] be seeing a dentist.
  6. * Mary [T will] may see a dentist.

We assume that have and be undergo V->T movement, exceptionally in English. This explains of certain unifying properties of auxiliaries (have+be+ modals):

  1. Undergoes inversion: Has Mary read Proust?
  2. Undergoes VP-ellipsis: Anyone who can read Proust already has [VP Ø ]
  3. Can be followed by negative element, unlike most verbal elements:
    1. John may not go.
    2. * John may go not.
    3. John did not go.
    4. * John not went. [did obligatory in negative sentences with no mdal.]
    5. * John went not.
    6. John wants to not go.
    7. John has not been going.
    8. John has been not going.
We note this and move on.

Subjectless clauses

We turn to the topic of clauses that appear to lack a subject.

PRO

Some clauses appear to be subjectless.

  1. The president isn't sure whether to intervene.
  2. Mary is anxious to leave.
It should be obvious what to do here. In (1) in particular, we have a complementizer, which is diagnostic of S-barhood, but no overt subject.

We posit that the subject is an empty element. This time we're going to write the empty element as PRO instead of e. Don't attach too much meaning to this. It's the position in the tree that tells us what kind of empty element we have, not its spelling. Spelling empty subjects a particular way makes the trees easier to read. So in the above examples, we propose to have empty subjects:

  • The president isn't sure [S' [Comp whether] [S [NP PRO ] [T to] [VP intervene.] ]]

    The so-called PRO-analysis of these constructions is not uncontroversial. The alternative analysis is that to intervene is a VP and that there is no clause following sure. We turn to some arguments for the PRO analysis.

Agreement
Argument
for PRO
Predicate nominals agree with the subjects of their immediate clauses:
  • They are unsure whether the president should be a candidate.
  • * They are unsure whether the president should be candidates.
Note that even though they is plural it is not what the predicate nominal a candidate agrees with. It is too far away. The predicate nominal agrees with the subject of its own clause.

  • The presidenti isn't sure [S' [Comp whether] [S [NP PROi ] [T to] [VP be a candidate/*candidates.] ]] Here the predicate nominal must be singular as well. We can retain our generalization about agreement if we say the predicate nominal agrees with PRO the subject of its own clause, and that PRO takes the president as an antecedent.

Reflexive
Argument
for PRO
Reflexives agree with the subjects of their immediate clauses in number, gender, and person. This is called clausemate agreement:
  • They are unsure whether Johni should show himselfi/*herselfi/*yourselfi.
  • * Theyi are unsure whether John should show themselvesi.
  • Theyi are unsure whether John should show themi.
Note in (2) that the plural reflexive may not agree with plural they. It is too far away. The reflexive agrees only with the subject of its own clause. In (3), we see, in contrast, that a nonreflexive pronoun may be coreferential with a non clausemate, so that clausemate agreement is a special property of reflexives.

  • The president isn't sure [S' [Comp whether] [S [NP PRO ] [T to] [VP show himself.] ]] Here the reflexive must match the president. We can retain our generalization about agreement if we say the reflexive agrees with PRO the subject of its own clause, and that PRO takes the president as an antecedent, as indicated by the subscripts.

Exceptional
Clauses

Consider:

  • I don't expect you to win.
We analyze expect in example (1) as what is called an exceptional clause verb. That is, we assume the following structure for the VP expect you to win:
  • [V' expect [S [NP you ] [T to] [VP win]]]
Note that up until now we have assumed that all embedded clauses are S-bars. Now, exceptionally, we are going to allow a certain very special class of embedded Ss.

We base this on a series of arguments.

Funny
Subjects

There is a class of subjects with very special distributions, which can occur with exceptional clause verbs. First we have special occurrences of it and there which are non-referential:

  1. There is a Santa Claus.
  2. Santa Claus exists.
  3. It was obvious that she was embarassed.
  4. That she was embarassed was obvious.
The it in (1) does not have an antacedent. The sentence is really a paraphrase of (2); we say the it functions as a dummy subject, satisfying a syntactic requirement of English that all finite clauses have overt subjects, even when that subject contirbutes nothing to the semantics. Something similar seems to be going on in (3), which has (4) as a paraphrase.

English also has a number of fixed or idiomatic expressions in which the subjects have a very unusual meaning.

  1. The cat is out of the bag. [ a secret is out.]
  2. The chips were down. [The crisis had arrived.]
  3. The shit hit the fan. [The consequences began to grow very complicated.]
  4. The chickens came home to roost. [Because of past actions which were unfair or evil, bad things began to happen.]

Having established the existence of these special subjects, we note their occurrence after expect

  1. Dummy it and there:
    • We expect it to be obvious that there is a hazard.
    • We expect it to rain tomorrow.
    • We expect there to be a school dance.
  2. Idiom chunks:
    • We expect all the chips to be down in the fourth quarter.
    • We expect the cat to be out of the bag by then.
    • We expect a few chickens to come home to roost in any close election.
    If nonreferential it and the chips were direct objects in these examples, as they would be in PRO analysis of this construction, we would expect these sentences all to be anomalous, because these peculiar ("funny") noun phrases do not occur as subjects of the constructions they are restricted to.
Passive
argument

The following two sentences are near synonyms:

  • We expect the doctor to examine Mary.
  • We expect Mary to be examined by the doctor.
If Mary and the doctor were direct objects, as they would be in a PRO analysis of this construction, we would expect these sentences to differ in truth conditions. Notice how, on the incorrect analysis, different NPs would be the direct object of expect:
  1. [V' expect [NP Mary ] [S' [Comp e] [S [NP PRO ] [T to] [VP be examined by the doctor.] ]]
  2. [V' expect [NP the doctor ] [S' [Comp e] [S [NP PRO ] [T to] [VP examine Mary.] ]]
On this analysis, Mary and the doctor would play a meaningful role (they would bear a theta role, in the language of Chapter 7) with respect to the expecting.
PRO versus
exceptional
clause

Consider:

  • John will force you to help me.
We analyze force as a verb taking an S-bar complement clause verb. Thus, force you to help me will include a PRO construction:
  • [V' force [NP you] [S' [comp e] [S [NP PRO ] [T to] [VP help me]]]
We base this on the following evidence:
  1. Dummy its and there are unacceptable:
    • * We forced it to be obvious that there is a hazard.
    • * We forced it to rain tomorrow.
    • * We forced there to be a school dance.
  2. Idiom chunks are as well:
    • * We forced all the chips to be down in the fourth quarter.
    • * We forced the cat to be out of the bag by then.
    This is explained on the PRO analysis because NPs like nonreferential it and idiom chunks are forced to be direct objects of the verb force, and therfore do not occur as subjects of the constructions they are restricted to:
  3. The following two sentences are not synonyms:
    • We forced the doctor to examine Mary.
    • We forced Mary to be examined by the doctor.
    On the PRO analysis Mary and the doctor are direct objects:
    1. [V' force [NP Mary ] [S' [Comp e] [S [NP PRO ] [T to] [VP be examined by the doctor.] ]]
    2. [V' force [NP the doctor ] [S' [Comp e] [S [NP PRO ] [T to] [VP examine Mary.] ]]
    We would expect these sentences to differ in truth conditions, because Mary and the doctor play different meaningful roles with respect to being forced.
Adverbial
Modifiers

Adverb distribution provides an additional argument for the contrasting analyses of force and expect:

  1. John forced Mary cruelly [S' C [S PRO to yield.]]
  2. * John expected [SMary cruelly to yield.]
Here cruelly can modify the upstairs verb with force but not expect, as is expected with the assigned structures.
Emphatic
Reflexives

Emphatic reflexives, controlled by clausemate subjects, are anomalous with the exceptional clause analysis and work (more or less) with the PRO analysis, as expected given the structures:

  1. John forced Mary himself [S' C [S PRO to yield.]]
  2. * John expected [SMary himself to yield.]

Small clauses

We introduce a new syntactic category SC for a variety of clausal constructions that lack T.
Small
Clauses 		 

Consider:

  • I consider John intelligent.
The maximal V' in example (2) has the following structure:
  • [V'consider [SC [NP John] [AP intelligent]]]
We analyze John intelligent as a small clause. Arguments follow.
Funny
Subjects 		 

The small clause analysis shares with the exceptional clause analysis the following important property: The NPs following the verb in these construction are analyzed as the subjects of a downstairs clause and not as objects of the upstairs verb. Therefore subjecthood tests should apply to both constructions.

  1. We considered it obvious that she was lying. [nonreferential it]
  2. We considered it time to leave. [nonreferential it]
  3. We considered the chips down. [Idiom chunk]
In all cases, these special elements have their distributions determined by the embedded construction, not the upstairs verb. This argues for their being downstairs subjects and against their being upstairs direct objects.
Funny
Preposing 		 

We also claim John in John intelligent is not a direct object of consider, a claim supported by the following odd preposing argument attributed to Chomsky (1981d) in your text:

  • * Consider John though you may intelligent, some people still think he's a pretty untrustworthy guy.
  • Give John though you may a large donation, you will still find him a pretty elusive guy.
Unlike other preposing tests, this one does not presuppose the "preposed" element "consider John" is a constituent, merely that what preposes with the verb in this construction must be a direct object.
Subject
Tests 		 

We are treating John in the above example as the subject of a small clause. Therefore we ought to find subject properties:

  1. We considered John responsible himself. [subjects can usually control floating emphatic reflexives]
  2. This year we considered not many candidates eligible. [not many usually works with subjects.]
No PRO  

Consistent with both the small clause and exceptional clause analysis, neither complementizers nor PRO are possible:

  • * We consider whether /if/that/for John intelligent.
  • * We consider PRO intelligent.
No T  

Though the construction following consider in (2) shares many properties with an exceptional clause construction, we can argue against its being analyzed as an exceptional clause construction on the basis of the fact that T is never allowed:

  1. * Nobody can consider John may intelligent.
  2. * Nobody can consider John to intelligent.