We now know that spec-TP is empty at D-structure, and that subjects raise to spec-TP from where they are selected. This chapter addresses further issues in phrasal movement. We will mainly be concerned with the similarities and differences between the following sentences.

The sentences of course mean different things. But they consist of the same linear sequence of words. A subject (Khufu), a finite verb (seems/wants), a nonfinite verb (to run) and an adverb (fast). The sole difference is the finite verb. In (1a), the verb is seems and in (1b) the verb is wants.

Despite the surface similarities, we’ll discover in this chapter that the two sentences have wildly different syntactic structures, all of which are related to the following core difference:

That is, want c-selects for a subject and assigns it a thematic role. Khufu in (1b) is an argument of want and gets the thematic role Experiencer. Seem on the other does not select for a subject and does not assign it a thematic role. Instead, in (1a), Khufu is selected by the infinitive and “raises” into the main clause from the lower clause.

The difference between the two verbs is sketched below. We’ll provide a more precise analysis later. The important thing to keep in mind is the starting place of the subject. For verbs like seem, the subject starts inside of the lower clause. For verbs like want, the subject starts outside of the lower clause.

We call predicates like seem raising verbs, and we call predicates like want  control verbs.

All of the following diagnostics illustrate this same difference: that the starting place of the subject of a raising verb is different than the starting place of the subject of a control verb.

Note that while the data is given in English, these diagnostics have cross-linguistic validity. All else equal, you can use these diagnostics to determine whether a verb embedding a nonfinite clause belongs to the class of raising or control verbs.

Raising vs. control

All of the following diagnostics allow you to determine if something is a raising or a control verb. To apply the diagnostic, you mechanically use the data given.

Expletives subjects I.

Raising verbs often permit an alternation between a referential subject and an expletive subject. Control verbs don’t.

How do we know that it in (4a) is an expletive? Well, it doesn’t appear to have semantic content. It doesn’t refer to anything. For instance, you can’t ask the question, *What seems that Khufu runs fast?

These data support the idea that seem doesn’t s-select for a subject, because that position can be filled by a semantically “vacuous” element.

Note that many languages don’t (obviously) have expletive subjects. For instance, in pro-drop languages like Spanish and Swahili, the subject position is obligatorily empty.

Again, this supports the idea that such sentences involve a semantically empty subject, since it cannot be referred to using an overt pronoun.

Idiom chunks.

Some phrases are “non-compositional” in that the meaning isn’t predictable from the parts. These are called idioms or idiom chunks.

We find a distinction between raising and control verbs with respect to idiom chunks.

What we observe is that with raising verbs, the idiom meaning is retained. With control verbs it isn’t. The a. examples sound acceptable because we can interpret the entire idiom chunk together. The control verbs lose the idiomatic meaning because the subject is actually a subject of the higher verb.

Note that this diagnostic is concerned with interpretation. If we get the non-literal (idiomatic) reading after inserting the verb – as we do in (9a) and (10a)  – then we conclude that  the subject must have raised from the lower clause, as this is where the rest of the material from the idiomatic expression resides. This same idea explains why the idiomatic meaning is not retained with control verbs; the subject the cat in (9b) can only be interpreted as an actual animal that wants out of a bag (literal reading) because it was never in the lower clause, and therefore cannot be interpreted with the idiomatic material.

Expletive subjects II.

English has another expletive subject, there. It’s only possible to use expletive there with a particular class of verbs, namely, verbs which assert existence, like the copula be.

Again we see a distinction between raising and control verbs with respect to expletive there.

The explanation behind this data goes as follows. Expletive there must be selected for by a particular predicate. Its appearance in (12a) follows because it has been selected for by the infinitive be in the lower clause. The reason it cannot appear in (12b) is that want cannot select for there.

Scope ambiguity.

This one is a bit more complex, as it presupposes a bit of semantics. This test asks us to consider how quantified expressions are interpreted in Raising and Control structures. A quantified expression is any expression that asserts a “quantity” of a set of individuals.^[1] As example of quantified expressions, consider the sentence Every student met some professor. This sentence is ambiguous with respect to how the quantified expressions every student and some professor interact. On one reading (called the “surface scope” reading), we are talking about the list of students. As long as each student has met a professor, then the sentence is true. On this reading, the professors could be different. Timmy met Prof. Smith. Susie met Prof. Jones. Rachel met Prof. Scott. Etc.

On an alternate reading of this sentence, we are talking about a particular professor. On this reading, the sentence is true if every student met the same professor. So Timmy met Prof. Smith and Susie met Prof. Smith and Rachel met Prof. Smith. The difference between the two readings is the scope of some professor. On the first reading, some professor scope “under” the every student. On the second reading some professor scopes “above” every professor.

Now consider the following sentence with a Raising predicate, be likely.

Notice that reading in (13b) is paraphrased as if someone were in an embedded clause. This is the reading we get when we interpret the quantified expression someone lower than where it’s pronounced. Crucially, this same ambiguity doesn’t seem to exist with want.

The sentence Someone from Florida wants to win the lottery is not ambiguous in the same way. It does not have a reading in which someone is interpreted lower than the position that it is pronounced.

Truth changing passive.

If you passivize the infinitive under a raising verb, it basically means the same as the active version. If you passivize the infinitive under a control structure, it means something very different.

Here again the explanation is that the subject of seem “inherits” whatever thematic role is assigned to the subject of the passivized verb. But the subject of want gets its own thematic role.

The final two tests below pick out only a subset of control verbs. Thus, if they pass, then you can confidently conclude that you have a control predicate. If they fail, you can conclude nothing.

For-CPs

Some control predicates allow both subject positions to be explicitly filled.  The subject of the infinitive appears in front of the complementizer for.

Partial control.

Some control predicates allow a “mismatch” in semantic features. Sometimes the lower subject can refer to “more” than the higher subject. For instance, a verb like gather requires a plural subject.

With certain control predicates, the subject of the higher verb can be singular, even if the infinitive selects for a plural subject. This type of a “mismatch” simply isn’t permitted with any raising verb.

Analyzing raising structures

There are a number of different predicates that share all the raising properties. Note that the term “predicate” includes verbs, adjectives, and even prepositions. The following are standardly recognized raising predicates; this is not an exhaustive list.

• seem (Khufu seems to run fast.)
• appear (Khufu appears to run fast.)
• be likely (Khufu is likely to run fast.)
• be about (Khufu is about to run fast.)
• …

The analysis of raising is fairly straightforward. The subject moves from the lower clause into spec-TP of the higher clause.

Let’s break down the important things to note about this analysis.

1. There is cyclic movement. The subject moves through the specifier of TP in the lower clause. The reason that this movement is necessary is that it satisfies the Subject Condition: Spec-TP must be filled at S-structure.
2. The TP dominating the infinitive doesn’t have a CP. (This is the only time in this class we’ll see a TP without a CP.) The reason that there is no CP here is that we cannot detect a CP here: we cannot use any kind of non-finite complementizer (e.g. for) with raising predicates.
3. There is no VoiceP in the main clause. This is because there is no external argument: seem doesn’t select for a subject.

In a sense, raising verbs present a particularly interesting and complicated study in long-distance dependencies in syntax. Building on the idea in the previous chapter that phrases can move to the subject position, raising verbs illustrate that this movement can cross clauses sometimes.

Analyzing control structures

 It is temping to simply treat control verbs like want analogous to raising verbs, except that the subject starts in the main clause. This is shown in the tree below. (This is the D-structure.)

However, this representation is wrong. It violates two rules of the syntactic theory we’ve built.

• The selectional requirements of run haven’t been met.  (Violates selection.)
• Spec-TP of the lower clause isn’t filled. (Violates the Subject Condition.)

Generally, if we allow trees like that above, then we need to stipulate that sometimes verbs don’t select, and sometimes spec-TP doesn’t have a specifier. This is a bitter pill because it really weakens the overall theory we’ve built.

Instead, to satisfy all of our constraints and keep the theoretical principles we’ve established, we are forced to posit a silent argument. We need something to be the subject of the lower clause. We call this silent argument PRO, pronounced “big pro.” PRO is a pronoun, but it’s a special pronoun. It only appears in particular configurations in order to satisfy grammatical constraints.

PRO solves our problems because it’s a DP, and it is merged in spec-VoiceP and moves to spec-TP, just like any other subject. It’s “special” because it only ever appears in nonfinite clauses.

The idea behind PRO is that it acts kind of like an anaphor. Its reference must be fixed by some c-commanding expression, i.e., it has to be bound. In all the previous examples, PRO is “controlled” by the subject.

Note though that PRO isn’t actually an anaphor. PRO obeys its own rules, and has its own distribution. For instance, PRO can be “arbitrary,” as in the following sentences.

The verb exercise needs a subject, so PRO must be present in these structures. But it gets an “arbitrary” reference, meaning something like “we all” or “someone generic.”

Evidence for PRO

You may be skeptical. Postulating the existence of something that is never pronounced and only appears in very particular circumstances should make you wary. This is good! You are acting like a scientist. PRO (or in general, the phenomenon of control of PRO) is a hypothesis. Our theory predicts that such an element should exist — that is, that there should be something “silent” in the syntax. So we’d like to find empirical evidence for PRO.

In fact, we find a lot of empirical evidence for PRO! This is hypothesis-driven research at its finest. Based on independent principles, our theory forces us to postulate something that seems odd. But then we find evidence for it. Here’s one piece of evidence from English, which utilizes what you already know about Binding Theory.

Binding theory and PRO.

The following sentences contain anaphors. Based on on Binding Theory (Condition A), an anaphor must be bound.

If PRO didn’t exist, how would the anaphor satisfy Condition A?

Where we’re going

The difference between raising and control is widely debated in the current theoretical landscape. The complexities of control in particular continue to be challenge to theoretical models of syntax. (I recommend Syntax II, LING 526, for a deeper investigation into the wonders of PRO.)

There is a larger issue though. Often, we come across structures in languages which we don’t know how to analyze straight-away. Take for instance the sentence in (21).

What is the relationship between John and the predicate get to work? Does John move to its surface position? Or is it generated there? This is a question of raising vs. control—which is really the question of Where do arguments start?  

At this point, we’ve introduced the main issues with one kind of phrasal movement: A-movement. We will continue now with the other kind of phrasal movement: A’-movement. As we’ll see, the two movements “look” different and also “act” different. In my opinion, A-movement is the “trickier” of the two kinds of phrasal movement. A’-movement is quite straightforward to diagnose—usually.

Exercises