Lecture 9

Linguistics 522

Movement

Lecture 9

NP Movement

Passive

We begin by dealing with the analysis of passive sentences like the following:

The car will be put in ther garage.

We argue for the following analysis:

[_NP e] will be put the car in the garage.
The car will be put [_NP e] in the garage.

We propose that such sentences are derived by a passive-transformation. The D-Structure is (1), the S-structure (2). Note that the subject is empty in D-structure. Unless something happens, this D-structure will produce an ungrammatical S-structure, since English does not allow empty subjects:

* will be put the car in the garage.

Thus, Passive saves an otherwise doomed D-Structure.

This is supposed to be a universal theory. Is this structure motivated for any other languages? The answer is yes. Some languages, such as German, have what is called an impersonal passive in which something like the D-structure proposed above can be a surface form, with something like a dummy there replacing the original subject:

There will be put the car in the garage.

Note that in our English sentence the D-structure is not the D-structure of any active sentence. We claim that every passive underlyingly has an empty subject. Active sentences, on the other hand, have overt NP subjects:

_NP

_Infl

_VP

Consider a passive paraphrase of this sentence:

The car will be put in the garage by John.

The D-Structure for the passive is quite different from the D-structure of the active:

_NP

_V'

_NP

_PP

Note the by-phrase is not put there by transformation; it's not derived at all. It's there in D-structure as an adjunct. Note also: At no point in the derivation of this passive is John represented as an "underlying subject." So Passives are not derived from actives in this theory. There are several differences between the D-structures of passives and actives:

Passive sentences have an empty subject, actives do not.
Passive sentences have a passive verbal form in D-structure, actives have an active verb form. This means that passive verb forms are found in the lexicon; they are not derived by transformation.

There is also one very important similarity:

Passive verb forms have the same subcategorization frames as their active counterparts, and they are filled in exactly the same way in D-structure. We turn to arguments for this apparently crazy analysis:

Connectivity		We account for why passive verbs have the same subcat frame as active verbs, minus NP: John will put the car in the garage. * John will put the car. [PP obligatory for active] The car will be put in garage. The car will be put.[PP obligatory for passive] Note we can't just allow put* another subcat frame, which omits the direct object. This is only possible with the passive form: * John will put in the garage. Having the movement rule allows just one passive frame. And predicts exactly when you get just the PP complement. Note also you can't just allow empty NPs to be generated willy nilly: * The car was put the bicycle in the garage. It is clear why this can't happen after passivization. Nothing can be inserted into a position an NP has moved from. What is less clear on our account is why this can't be a D-structure. The long answer is for volume II. The short answer is that although the passive form of the verb has the the same subcategorization frame as the active form, it differs semantically. The passive form of the verb put assigns no role to its subject position (there is no putter in its semantics). We assume roles are assigned at D-structure. This is one of the functions of D-structure: To assign roles to arguments in their canonical syntactic positions. Structures that have overt NPs which are not assigned any role at D-structure are ruled out. * The car was put the bicycle in the garage. The exact technical reasons for this involve the interaction of the Projection Principle and the Theta- Criterion [introduced p.391], and will be spelled out in volume II.
Funny Elements		Idiom chunks. The FBI kept close tabs on Bill. Close tabs were kept on Bill. The students paid little heed to the professor's warnings. Little heed was paid to the professor's warnings. The inspector took note of the suspect's nervousness. Note was taken of the suspect's nervousness. Note that these idiom chunks have very limited distribution: * I don't like talking about tabs/ surveillance. * Note/Attention makes me nervous. He expects complete * heed/obedience. In our analys of passive, passivized idiom chunks would start out in directy object position in D-structure. Thus there distribution in passives in D-strctuture is exactly the same as it is for actives. So if idiom chunks are checked at D-structure, we have a simple account of their distribution in D-structure, and our passive rule accounts for their distribution at S-structure.
Selection Restriction		Sincerity frightens John. * John frightens sincerity. John is frightened. [just like the active when John is Direct Object] * Sincerity is frightened. [just like the active when sincerity is Direct Object] [_NP e] is frightened John. * [_NP e] is frightened sincerity. At D-structure, these are in Direct Object position. So if we apply selection restrictions at D-structure, all these facts have a uniform account: frighten takes animate direct objects. Our passive rule then accounts for where the various arguments selection has applied to may occur at S-structure.

Connectivity

We account for why passive verbs have the same subcat frame as active verbs, minus NP:

John will put the car in the garage.
* John will put the car. [PP obligatory for active]
The car will be put in garage.
*The car will be put.[PP obligatory for passive]

Note we can't just allow put another subcat frame, which omits the direct object. This is only possible with the passive form:

* John will put in the garage.

Having the movement rule allows just one passive frame. And predicts exactly when you get just the PP complement.

Note also you can't just allow empty NPs to be generated willy nilly:

* The car was put the bicycle in the garage.

It is clear why this can't happen after passivization. Nothing can be inserted into a position an NP has moved from. What is less clear on our account is why this can't be a D-structure.

The long answer is for volume II. The short answer is that although the passive form of the verb has the the same subcategorization frame as the active form, it differs semantically. The passive form of the verb put assigns no role to its subject position (there is no putter in its semantics).

We assume roles are assigned at D-structure. This is one of the functions of D-structure: To assign roles to arguments in their canonical syntactic positions. Structures that have overt NPs which are not assigned any role at D-structure are ruled out.

* The car was put the bicycle in the garage.

The exact technical reasons for this involve the interaction of the Projection Principle and the Theta- Criterion [introduced p.391], and will be spelled out in volume II.

Funny
Elements

Idiom chunks.

The FBI kept close tabs on Bill.
Close tabs were kept on Bill.
The students paid little heed to the professor's warnings.
Little heed was paid to the professor's warnings.
The inspector took note of the suspect's nervousness.
Note was taken of the suspect's nervousness.

Note that these idiom chunks have very limited distribution:

* I don't like talking about tabs/ surveillance.
* Note/Attention makes me nervous.
He expects complete * heed/obedience.

In our analys of passive, passivized idiom chunks would start out in directy object position in D-structure. Thus there distribution in passives in D-strctuture is exactly the same as it is for actives. So if idiom chunks are checked at D-structure, we have a simple account of their distribution in D-structure, and our passive rule accounts for their distribution at S-structure.

Selection
Restriction

Sincerity frightens John.
* John frightens sincerity.
John is frightened. [just like the active when John is Direct Object]
* Sincerity is frightened. [just like the active when sincerity is Direct Object]
[_NP e] is frightened John.
* [_NP e] is frightened sincerity.

At D-structure, these are in Direct Object position. So if we apply selection restrictions at D-structure, all these facts have a uniform account: frighten takes animate direct objects.

Our passive rule then accounts for where the various arguments selection has applied to may occur at S-structure.

Finally, note that passive may occur in exceptional clauses and in small clauses

John is considered [_S [_NP e] to be a fool.].
John is considered [_SC [_NP e] to be a fool.].

Raising (or NP Movement) Constructions

We now turn to sentences like:

John seems to be unhappy.

The analysis we propose for these is the following:

D-structure: [_NP e ] seems [ [_NP John ] [_Infl to ] [_VP be unhappy.] ]
S-structure: [_NP John ] seems [ [_NP e ][_Infl to ] [_VP be unhappy.] ]

In other words, the NP John moves from being subject of to be unhappy to being subject of the main clause. Because John gets raised in the tree, these are called raising or NP-movement constructions.
One claim made is that, underlyingly, John is subject of the downstairs clause. Let's test this with some of our subjecthood tests. Then we'll look for evidence of connectivity. That means complementary distribution, selection restrictions, funny subjects:

Reflexives

John seems to me to have perjured himself.
* John seems to me to have perjured myself.
This is strong evidence that, at some level of structure, John belongs in the underlying clause as subject. The proposed analysis does this.
[_NP e ] seems [_S [_NP John ] [_Infl to ] [_VP have perjured himself.] ]
Moreover the fact that John moves lets us capture another kind of reflexive fact:
[_NP e ] seems to himself [_S [_NP John ] [_Infl to ] [_VP be smart.] ]
After movement we get:

John seems to himself to be smart.
So by being subject of the main clause as well, after movement, the NP is entitled to another reflexive.
Note that we could say that reflexivization operates across S-bar boundaries (just as we said about Passive]. But this would get the facts wrong:
* John seems to me to have perjured myself.

Agreement This, like the reflexivization argument, is based on a subject property limited single clauses. Subjects agree with predicate nominals in the sam clause:

They expect he is the president/*presidents.
He seems to them to be the president/*presidents.
Again assuming starts out in the same clause with the predicate nominal allows a uniform account.

Selection
restrictions

John seemed the understand the problem.
* My pet amoeba seemed to understand the problem.
My pet amoeba seems to be reproducing.
? My kettle seems to be reproducing.
My kettle seems to be boiling over.
? My pet amoeba seems to be boiling over.
In each case it seems to be the embedded clause verb that determines what the subject can be. The main clause verb seem on the other hand seems to be selectionally promiscuous. Whatever is okay downstairs is okay with it.
It's useful to compare seem with try

My kettle seemed to boil over.
? My kettle tried to boil over.
Kettles can boil over but they can't try to boil over. The verb try really places selection restrictions on its subject. In contrast the verb seem seems not to.

Verbs that
swing two
ways
Note that some verbs that are raising predicates take exceptional clauses and some take small clauses:

John needs to be happy.
John became happy.
* John needs happy.
* John became to be happy.
But some verbs are BOTH:

John seems [_S t_i to be happy]
John seems [_SC t_i happy]
The right way to describe this is to say that seem subcategorizes both an exceptional S and an SC.
Funny
Subjects
The promiscuity of seem with respect to subjects is further brought out when we look at idiom chunks:

The jig seems to be up.
The cat really seems to be out of the bag.
Tabs seem to have been kept on this gentleman for quite a while.

Again the key idiom words have limited distribution:

The jig has reached its endpoint. [No idiomatic meaning.]
? I wish you would stop divulging that cat.
* Tabs is/are a tricky business.

It's again useful to compare seem with a verb like try:

* The jig try to be up.
The cat really tried to be out of the bag.[no idiomatic meaning.]
* Tabs tried to be kept on this gentleman for quite a while.
None of these subjects are compatible with trying, which does assign some kind trying role to its subject.
Looking at the sentences with seem again:

The jig seems to be up.
The cat really seems to be out of the bag.
Tabs seem to have been kept on this gentleman for quite a while.
In all cases the construction that licenses the idiom is the downstairs one.
Ditto for dummy subjects.

It seems to be obvious that no one will come. [obvious takes dummy it.]
It seems to be raining.[rain takes dummy it.]
It seems to be time to leave.[time takes dummy it.]
There seems to be a problems here.[be an X takes dummy there.]
Again, trying try doesn't work:

* It tried to be obvious that no one will come.
* It tried to rain.
* It tried to be time to leave.
* There tried to be a problem here.

So what we're finding is evidence of the following:

seem (unlike some other verbs with infinitivals) places no selection restrictions on its subjects.
seem assigns no theta role to its subjects.
Our movement analysis of seem captures exactly these facts. We assume theta-roles are assigned and selection restrictions are applied at D-structure. But seem has no subject at D-structure in our analysis:
D-structure: [_NP e ] seems [ [_NP John ] [_Infl to ] [_VP be unhappy.] ]

S not Sbar
We assume that in raising constructions the lexical head subcategorizes for an S. So seem is "exceptional" like believe in that it goes with an S, not an Sbar. One reason: no complementizer allowed.

* John seems for to be a fool.
Another more theory-internal reason is that we have in mind a general restriction, that will hold for both passive and NP movement, that movement doesn't cross Sbar boundaries.

Passive
test
All of these tests should remind you of the exceptional clause cases:

We believe [_S [_NP John ] [_Infl to ] [_VP be unhappy.] ]
[_NP e ] seems [_S [_NP John ] [_Infl to ] [_VP be unhappy.] ]

In both cases the NP John seemed at first belong to the main clause, in both cases we proposed an analysis that denied it, and claimed that it was subject of the embedded clause. In the exceptional clause case, it was evidence such as passivization that made John appear to belong to the upstairs clause:
John was believed to be happy.
So it can end up separated from what we think of as its true verb Be.
In the raising construction case, our principle evidence that John belonged to the main clause was that it started out separate from the embedded verb:

John seemed to be happy.
It came right before the main verb seem, not right before be.
In both cases, we decided the underlying truth differed from the appearances.
Both analyses share the property that the NP John receives no theta-role from the main verb. This explains how the distribution of funny elements is determined exclusively by the embedded verb.
In both cases, passivizing the embedded clause shouldn't make much of the meaning difference, because none of the roleplayers in the embedded clause play a role in the main clause:

John expected the doctor to examine Mary.
John expected Mary to be examined by the doctor.
The doctor seems to have examined John.
John seems to have been examined by the doctor.

Facts
About
Raising
Firtst, we get raising out of small clauses as well:

John seems angry at himself.
* John seems to me angry at myself.
What is angry at himself here? Well it contains a reflexive controlled by John, so want an analysis on which it is a clausemate with John. We also know there are two clauses here because of (2).
This leads to the small clause analysis:
[_NP e ] seems [_SC [_NP John ] [_AP angry at himself.] ]
That is, John angry at himself is a verbless, infl-less ("small") clause.
Note we can also use "funny" subject arguments: here:
It seems obvious the analysis is right.

PRO vs.
Exceptional
clause
We now have two different competing analyses for simple(-looking) infinitival constructions:

PRO (or control) analysis (try)
[_NP John] [_VP tried [_S' [_Comp e ] [_S [_NP PRO] [_Infl to] [_VP be brave]]]]
Raising analysis (seem):
[_NP e ] seems [_SC [_NP John ] [_AP angry at himself.] ]
Remember they differ in two ways: in whether they involve movement (the raising analysis does, the PRO (or control) analysis doesn't); and in whether there is an S or S' complement after the verb.
Take a brand new construction:

John is apt to smile inappropriately.
Now how do we know which it is, raising or control?
If you have to apply one quick and dirty test to leap to a conclusion, use the there construction:

There is apt to be an earthquake.
This is a funny subject, licensed by the verb "to be" in the embedded clause. So it should work with a verb that assigns no role to its subject, which indicates a raising verb.
As expected, this doesn't work with a control verb:
* There tried to be an earthquake.

You should also be able to use idiom chunks and other dummy subjects:

Tabs are apt to be kept on Commie sympathizers.
It is apt to be obvious that John is drunk.
* Tabs tried to be kept on Commie sympathizers.
* It tries to be obvious that John is drunk.
And keep in mind the passive test:

My dog is apt to attract fleas.
Fleas are apt to be attracted by my dog. [paraphrases]
My dog tries to attract fleas.
? Fleas try to be attracted by my dog. [ not paraphrases]
In drawing your final analysis, be sure to remember apt is an adjective:
[_NP e ] [_VP [_V' is [_AP[_A' apt [_S [_NP John ] [_Infl to] [_VP attract fleas]]]]]]

Chains
You can also see what's coming after the last homework. The final is apt to involve long complicated strings of words with long complicated derivations.
Take an example:
John is certain to be considered a genius.
What's the answer?
First we want the S-structure. We know John is the subject. We next consider the relationship of that NP to certain
Remember that certain may be a raising or a control preidcate. For a quick guess, try a dummy-there test:

There is certain to be an earthquake.
Guess: This is a raising predicate. A dummy subject is okay! Tentative structure:

[John_i] [_Infl is_j ] [_VP t_j [_AP certain [_S t_i [_Infl to ] .....
Note that I have indicated the verb movement for the verb be.
We now move to a downstairs clause. Next we note that be considered is a passive form of the verb:

BE + Passive Particple
be + considered
Can consider be followed by an NP? Which is what passive needs to operate? Yes:

They considered John a genius.
Note that in order to decide the properties of a verb in complex sentences I ALWAYS look at it in simple sentences.
But what is this structure? This is a small clause structure. Evidence:

They considered John alone a genius.
Recall that alone works best with subjects. Structure for simple example:

They considered [_SC John a genius]
Final structure for entire example:

[John_i] [_Infl is_j ] [_VP t_j [_AP certain [_S t_i [_Infl to ] [ be [_VP considered [_SC t_i a genius]]]]]

Reflexives		John seems to me to have perjured himself. * John seems to me to have perjured myself. This is strong evidence that, at some level of structure, John belongs in the underlying clause as subject. The proposed analysis does this. [_NP e ] seems [_S [_NP John ] [_Infl to ] [_VP have perjured himself.] ] Moreover the fact that John moves lets us capture another kind of reflexive fact: [_NP e ] seems to himself [_S [_NP John ] [_Infl to ] [_VP be smart.] ] After movement we get: John seems to himself to be smart. So by being subject of the main clause as well, after movement, the NP is entitled to another reflexive. Note that we could say that reflexivization operates across S-bar boundaries (just as we said about Passive]. But this would get the facts wrong: * John seems to me to have perjured myself.
Agreement		This, like the reflexivization argument, is based on a subject property limited single clauses. Subjects agree with predicate nominals in the sam clause: They expect he is the president/presidents. He seems to them to be the president/presidents. Again assuming starts out in the same clause with the predicate nominal allows a uniform account.
Selection restrictions		John seemed the understand the problem. * My pet amoeba seemed to understand the problem. My pet amoeba seems to be reproducing. ? My kettle seems to be reproducing. My kettle seems to be boiling over. ? My pet amoeba seems to be boiling over. In each case it seems to be the embedded clause verb that determines what the subject can be. The main clause verb seem on the other hand seems to be selectionally promiscuous. Whatever is okay downstairs is okay with it. It's useful to compare seem with try My kettle seemed to boil over. ? My kettle tried to boil over. Kettles can boil over but they can't try to boil over. The verb try really places selection restrictions on its subject. In contrast the verb seem seems not to.
Verbs that swing two ways		Note that some verbs that are raising predicates take exceptional clauses and some take small clauses: John needs to be happy. John became happy. * John needs happy. * John became to be happy. But some verbs are BOTH: John seems [_S t_i to be happy] John seems [_SC t_i happy] The right way to describe this is to say that seem subcategorizes both an exceptional S and an SC.
Funny Subjects		The promiscuity of seem with respect to subjects is further brought out when we look at idiom chunks: The jig seems to be up. The cat really seems to be out of the bag. Tabs seem to have been kept on this gentleman for quite a while. Again the key idiom words have limited distribution: The jig has reached its endpoint. [No idiomatic meaning.] ? I wish you would stop divulging that cat. * Tabs is/are a tricky business. It's again useful to compare seem with a verb like try: * The jig try to be up. The cat really tried to be out of the bag.[no idiomatic meaning.] * Tabs tried to be kept on this gentleman for quite a while. None of these subjects are compatible with trying, which does assign some kind trying role to its subject. Looking at the sentences with seem again: The jig seems to be up. The cat really seems to be out of the bag. Tabs seem to have been kept on this gentleman for quite a while. In all cases the construction that licenses the idiom is the downstairs one. Ditto for dummy subjects. It seems to be obvious that no one will come. [obvious takes dummy it.] It seems to be raining.[rain takes dummy it.] It seems to be time to leave.[time takes dummy it.] There seems to be a problems here.[be an X takes dummy there.] Again, trying try doesn't work: * It tried to be obvious that no one will come. * It tried to rain. * It tried to be time to leave. * There tried to be a problem here. So what we're finding is evidence of the following: seem (unlike some other verbs with infinitivals) places no selection restrictions on its subjects. seem assigns no theta role to its subjects. Our movement analysis of seem captures exactly these facts. We assume theta-roles are assigned and selection restrictions are applied at D-structure. But seem has no subject at D-structure in our analysis: D-structure: [_NP e ] seems [ [_NP John ] [_Infl to ] [_VP be unhappy.] ]
S not Sbar		We assume that in raising constructions the lexical head subcategorizes for an S. So seem is "exceptional" like believe in that it goes with an S, not an Sbar. One reason: no complementizer allowed. * John seems for to be a fool. Another more theory-internal reason is that we have in mind a general restriction, that will hold for both passive and NP movement, that movement doesn't cross Sbar boundaries.
Passive test		All of these tests should remind you of the exceptional clause cases: We believe [_S [_NP John ] [_Infl to ] [_VP be unhappy.] ] [_NP e ] seems [_S [_NP John ] [_Infl to ] [_VP be unhappy.] ] In both cases the NP John seemed at first belong to the main clause, in both cases we proposed an analysis that denied it, and claimed that it was subject of the embedded clause. In the exceptional clause case, it was evidence such as passivization that made John appear to belong to the upstairs clause: John was believed to be happy. So it can end up separated from what we think of as its true verb Be. In the raising construction case, our principle evidence that John belonged to the main clause was that it started out separate from the embedded verb: John seemed to be happy. It came right before the main verb seem, not right before be. In both cases, we decided the underlying truth differed from the appearances. Both analyses share the property that the NP John receives no theta-role from the main verb. This explains how the distribution of funny elements is determined exclusively by the embedded verb. In both cases, passivizing the embedded clause shouldn't make much of the meaning difference, because none of the roleplayers in the embedded clause play a role in the main clause: John expected the doctor to examine Mary. John expected Mary to be examined by the doctor. The doctor seems to have examined John. John seems to have been examined by the doctor.
Facts About Raising		Firtst, we get raising out of small clauses as well: John seems angry at himself. * John seems to me angry at myself. What is angry at himself here? Well it contains a reflexive controlled by John, so want an analysis on which it is a clausemate with John. We also know there are two clauses here because of (2). This leads to the small clause analysis: [_NP e ] seems [_SC [_NP John ] [_AP angry at himself.] ] That is, John angry at himself is a verbless, infl-less ("small") clause. Note we can also use "funny" subject arguments: here: It seems obvious the analysis is right.
PRO vs. Exceptional clause		We now have two different competing analyses for simple(-looking) infinitival constructions: PRO (or control) analysis (try) [_NP John] [_VP tried [_S' [_Comp e ] [_S [_NP PRO] [_Infl to] [_VP be brave]]]] Raising analysis (seem): [_NP e ] seems [_SC [_NP John ] [_AP angry at himself.] ] Remember they differ in two ways: in whether they involve movement (the raising analysis does, the PRO (or control) analysis doesn't); and in whether there is an S or S' complement after the verb. Take a brand new construction: John is apt to smile inappropriately. Now how do we know which it is, raising or control? If you have to apply one quick and dirty test to leap to a conclusion, use the there construction: There is apt to be an earthquake. This is a funny subject, licensed by the verb "to be" in the embedded clause. So it should work with a verb that assigns no role to its subject, which indicates a raising verb. As expected, this doesn't work with a control verb: * There tried to be an earthquake. You should also be able to use idiom chunks and other dummy subjects: Tabs are apt to be kept on Commie sympathizers. It is apt to be obvious that John is drunk. * Tabs tried to be kept on Commie sympathizers. * It tries to be obvious that John is drunk. And keep in mind the passive test: My dog is apt to attract fleas. Fleas are apt to be attracted by my dog. [paraphrases] My dog tries to attract fleas. ? Fleas try to be attracted by my dog. [ not paraphrases] In drawing your final analysis, be sure to remember apt is an adjective: [_NP e ] [_VP [_V' is [_AP[_A' apt [_S [_NP John ] [_Infl to] [_VP attract fleas]]]]]]
Chains		You can also see what's coming after the last homework. The final is apt to involve long complicated strings of words with long complicated derivations. Take an example: John is certain to be considered a genius. What's the answer? First we want the S-structure. We know John is the subject. We next consider the relationship of that NP to certain Remember that certain may be a raising or a control preidcate. For a quick guess, try a dummy-there test: There is certain to be an earthquake. Guess: This is a raising predicate. A dummy subject is okay! Tentative structure: [John_i] [_Infl is_j ] [_VP t_j [_AP certain [_S t_i [_Infl to ] ..... Note that I have indicated the verb movement for the verb be. We now move to a downstairs clause. Next we note that be considered is a passive form of the verb: BE + Passive Particple be + considered Can consider be followed by an NP? Which is what passive needs to operate? Yes: They considered John a genius. Note that in order to decide the properties of a verb in complex sentences I ALWAYS look at it in simple sentences. But what is this structure? This is a small clause structure. Evidence: They considered John alone a genius. Recall that alone works best with subjects. Structure for simple example: They considered [_SC John a genius] Final structure for entire example: [John_i] [_Infl is_j ] [_VP t_j [_AP certain [_S t_i [_Infl to ] [ be [_VP considered [_SC t_i a genius]]]]]