Video 15.2 Subject to Object Raising and Object Control

Welcome to syntax, a generative introduction of 4th edition.

My name is Andrew Karni. I'm a professor of logistics at the University of Arizona.

I'm the author of your textbook, and I'll be needing you through this series of video tutorials.

In the last video, we talked about subject to subject raising, sentences like,

Jean is likely to leave, and subject control, sentences like Jean is reluctantly.

The two differ primarily in terms of data grids.

And Jean is likely to leave. Jean receives a federal from leave, but not from likely.

And that DP moves from the embedded clause into the main clause for case reasons.

In the second sentence, Jean is reluctant to leave.

Jean appears to be getting two-thater roles in violation of the data criterion.

So we claim that Jean does not move, instead Jean is generated in the higher clause.

And there's a second DP, it just happens to be null, called Pro in the embedded clause.

And Pro bears the agent role of leave, and Jean bears the experience or role of reluctant.

In this video, we're going to talk about a parallel set of sentences.

We're going to talk about what is called subject to object raising and object control.

Subject to object raising, which is also sometimes called exceptional case marking or ECM,

is when you have an element that is an embedded subject that is appearing with accusative case,

as if it is part of the higher clause and getting accusative case from want.

And indeed, that's the analysis we're going to give it.

We're going to treat these sentences like John wants bill to leave,

as if bill is generated in the downstairs clause and moves to the upstairs clause for case.

That movement is back to us because of the two positions are side by side.

But there's several reasons for us to want to do this.

One is thematic.

So a bill here is not getting a theta role from want.

John does not want bill.

What John wants is bills leaving the embedded clause.

So bill is only getting a theta role from leave.

It's not getting a theta role from want.

But it is clearly because of the case marking functioning as the direct object to want.

So in order to get accusative case on the embedded subject,

you want to raise the subject into the specifier of the agarope that we discussed in the previous chapter,

which is the position where when you do object shift,

the object's move into that position.

And that position will also give us a place for these raised subjects that are functioning as objects.

So let's look at how that would work.

So here's a tree demonstrating subject to object movement.

Let's break down the various bits.

At the top of the tree, we have a voice phrase that introduces the agent or the experienceer of want.

So it probably an experience or so gene is wanting.

And the voice phrase here introduces that element and gives it its theta role.

That DP has to move into the specifier of TP to satisfy the EPP properties of this higher tense.

And also the case properties of this DP.

It means to be in this position in order to get a nominal case.

The verb want is going to undergo short head movement.

You'll recall that we claim this happens in English.

That a verb root will sort of do two quick prompts from Agarot into the voice phrase.

And it does so to get in front of the object's position, which is right here.

That's what we argued in the previous chapter.

Now what's going on in the embedded clause?

So want takes as its argument a CP.

And that CP can be non-finite, just like the embedded clause in a subject to subject raising construction is non-finite.

And that means that there is no case position here in the specifier of TP.

Now I've cheated a little bit by already moving bill into the specifier of this position.

I did so because this position here needs to be filled because of the EPP.

The other reason you would hop into this position is because of the minimal link condition,

which says you have to move to the closest potential position for the element to check its case.

But this is not a case position.

So neither was the specifier of the voice phrase from which it came.

But it moves here anyways for EPP reasons.

But it still doesn't have case in this position.

So it's going to have to move.

Well fortunately we have a case position.

And that case position is right here.

This is the the specifier of agarrow phrase, which is introduced because we've used an active voice, which means an object is possible.

So there's here we have a DP without any case.

And here we have a position where it could get case.

It just happens to be a queueset of case versus nominal of case.

So here this DP is moving up into this position into the object position in order to get a queueset of case.

So how do we know that we're actually moving into object position, the specifier of agarrow?

There's a couple of pieces of evidence.

First of all, these arguments take a queueset of case.

So they take the case that would be associated with the object position of the verb want.

So that's the first piece of evidence.

The second piece of evidence comes from binding theory.

So we're called that anophores must find their antecedent within the CP that dominates them within their binding domain.

So when you have an anophore you look up the tree and you find the first CP that dominates you.

And there had better be an antecedent that's within that CP.

And that antecedent of course has to be co-indexed with you and seek him and you.

So if we look at our sentences here, John wants himself to leave.

This sentence is okay with himself.

And if this element himself is still downstairs in the embedded clause,

then its antecedent is too far away.

It's antecedent is in the higher clause.

But if we've moved himself into the object position of the main clause,

then the antecedent is within the same CP.

So himself looks up the tree and finds the first CP that dominates it, which is the one that's right on top.

Not one also dominates John.

So anophores tell us that this John and this himself actually have to be quite close together.

The data from pronouns shows the same thing.

So we know that pronouns cannot be co-indexed with an antecedent in the same clause.

So that's what we see with these two sentences.

So John wants him where the same index is ungramadical.

It's only grammatical if him refers to somebody else.

And that's the typical property of pronouns.

It's also worth noting that this pronoun does not behave as if it's in an embedded clause.

When a pronoun is within an embedded clause,

it can either be co-indexed or it can be free.

So he with an eye or he with a j is fine.

And that's not true in these sentences.

The j sentence is fine whether not co-indexed.

But the i sentence is terrible.

So these data show us that this element here, the himself with the him,

is in the object position of the higher clause.

Case shows this and binding theory shows this.

So that's consistent with our analysis.

Okay, next we're going to talk about a second kind of control.

We just talked about a second kind of raising.

We had done previously subject to subject raising.

And we just learned about subject to object raising.

In the previous video, we talked about subject control,

sentences like John is reluctant to leave.

But in this video, we're going to talk about a second kind of control.

Now here, if we sort of think about the thematic relations,

you'll see that Robert is bearing two theta roles effectively.

Robert has been persuaded.

So is the theme of persuade.

And Robert is also the agent of leave.

So when you result in this apparent violations of the theta criterion,

that's a signal right away that you probably have a control construction.

If we think about the theta graded for persuade,

it requires an agent, the persuader, which is gene here.

It requires somebody who is persuaded, Robert.

And then it requires a clause, which expresses what Robert was persuaded of.

But that clause appears to be missing a subject.

And so we've used pro here instead to mark that.

That's how, remember from the previous set of videos that we didn't,

that we proposed pro in order to escape the apparent violation of the theta criterion.

So that Robert gets a theta role from persuade, pro gets a theta role from leave.

And then Robert and pro are linked together by some kind of principle of like the binding theory,

where Robert determines what the value of pro is.

So this is object control.

So let's talk about what happens in the theta grads in object control.

So what we have is a verb like persuade, takes an agent as in gene persuaded,

Robert to leave. The agent here is gene.

Takes a theme, who is the person who is persuaded,

in this case, that's Robert.

And it takes an embedded clause as the thing that Robert is persuaded about.

So that's the CP.

The verb leave takes an agent and here in these theta grads that agent is reflected by pro.

And that's how we get around the theta criterion.

Because by having pro here, we have a situation where Robert doesn't actually get two theta roles.

Now we had lots of evidence for the distinction between subject to subject raising and subject to object raising.

We were able to work out the theta grads, like we just did for object control.

But we also had evidence from extra position where we had an exponent of subject and from a causal subjects.

Now the reason you can't use those tests here is of course because this is about the object.

And those two tests are specifically about the subject position, whether or not you can have an exponent of the subject position or put a clause in the subject position.

So those tests are not going to work for us.

However, our idiom tests will work.

So you'll recall that an idiom like the shit hit the fan in English is ambiguous.

It can have both an idiomatic meaning, meaning bad things happened.

And it can have a literal meaning where there's x-rement hitting rotating fan blades.

And the issue at hand is does it retain the idiomatic meaning once you've done either racing or looking at control.

If it retains the idiomatic reading, then we can deduce that that object has actually started in the embedded clause,

where the embedded clause is taken as the idiom and interpreted as such.

And then you just move the subject out.

If you have object control, the object never starts out in the embedded clause.

You have a pro in the embedded clause, so you will not get the idiomatic reading.

The idiomatic reading will be impossible.

So we look at some data here and we'll see that indeed subject to object racing verbs like want.

You retain the idiomatic interpretation.

So John wants the shit to hit the fan.

Can indeed mean John wants bad things to happen.

By contrast, John persuaded the shit to hit the fan.

Cannot have that interpretation. It's just a bizarre sentence.

It suggests that the shit can be persuaded of something.

So that loses that meaning.

So one quick way you can distinguish subject to object racing predicates from object control predicates is by using this idiom test.