6 Raising versus control


In Chapter 5, we derived subject movement from Spec(VP) to Spec(IP) from the structural conditions under which nominative case is assigned in English. In this chapter, we address a related type of movement known as subject raising. Both ordinary subject movement and subject raising affect subjects of sentences, but they differ in how far the subject moves. With ordinary subject movement, the subject remains in its own clause, whereas with subject raising, the subject moves out of its clause into a higher one. The demonstration that English has subject raising relies on the existence of certain special noun phrases, and we therefore begin our discussion of subject raising by introducing them. We then show that certain verbs trigger subject raising, and we present an analysis of it. We conclude by distinguishing raising from a superficially similar phenomenon called control.

Special subjects

English has three types of noun phrases that do not participate in the full range of occurrence of ordinary noun phrases: expletive there (already briefly discussed in Chapter 5), subject idiom chunks (that is, subjects of sentential idioms) and weather it. What makes these noun phrases special is that they are subject to strict locality requirements. In particular, they must be licensed as subjects of an appropriate predicate; what we mean by this is that each of the special noun phrases is ungrammatical unless it is the subject of the appropriate predicate. Here and throughout this section, special subjects are in italics. Predicates that license the special subjects are in green; ones that don't are in red.

Expletive there

As is evident from the contrast between (36) and (37), expletive there is only grammatical when the main verb of its clause belongs to a certain class of verbs---roughly speaking, verbs of being in existence or coming into existence.

(36) a. ok After their military defeat, there arose among the Plains tribes a powerful spiritual movement.
b. ok There is a problem.
c. ok There began a reign of terror.
d. ok In the end, there emerged a new caudillo.
e. ok There ensued a period of unrest and lawlessness.
f. ok There exists an antidote.
g. ok There follows a section on the care of gerbils.
h. ok There has occurred an unfortunate incident.
i. ok There remains a single course of action.
(37) a. * There continued the same problem.
b. * There failed an available time slot.
c. * There proved toxins in the soap.
d. * There rang the mail carrier.
e. * There seems no solution.
f. * There sang an impressive choir from Russia.
g. * There turns out an antidote.

Subject idiom chunks

The second type of special subjects are the subjects of sentential idioms, subject idiom chunks.

(38) a.   The cat has got X's tongue.
b.   The cat is out of the bag.
c. The fur will fly.
d. The jig is up.
e. The pot is calling the kettle black.

What will be crucial for our purposes is that the subjects in (38) have whatever idiomatic meaning they have in connection with the rest of the idiom, but not otherwise. For instance, neither cat in (38b) nor pot in (38e) have a general metaphorical sense of secret or hypocrite, respectively. So although the sentences in (39) are not ungrammatical (unlike the ones in (37)), they have only literal interpretations, which can be semantically or pragmatically anomalous.

(39) a.   The cat is safe with her.
(can't mean: 'The secret is safe with her.')
b. # Fortunately, the pot retracted that piece of duplicitous slander.
(can't mean: 'The hypocrite retracted that piece of duplicitous slander.')

Weather it

The third type of special subject, weather it, is the subject of verbs of precipitation. The relationship between weather it and the predicates that license it is even closer than in the previous two cases. Not only is weather it licensed by verbs of precipitation, but the verbs of precipitation are themselves in turn licensed by weather it. As a result, other semantically conceivable subjects are not possible, as the contrast between (40) and (41) shows. In other words, weather it stands in a mutual licensing relationship with the verbs of precipitation. We indicate this by putting weather it in italics (indicating its status as a special subject that needs to be licensed) as well as in green (indicating its own status as a licenser).

(40)   ok It is { hailing, pouring, raining, sleeting, snowing. }
(41)   * The { air, atmosphere, environment, precipitation, sky, weather } is { hailing, pouring, raining, sleeting, snowing. }

To summarize, all three types of special subjects are special because they must be licensed in a local relation---specifically, as the subject of a particular type of predicate. Expletive there is licensed as the subject of verbs of being in or coming into existence, subject idiom chunks are licensed as subjects of sentential idioms, and weather it stands in a mutual licensing relation with verbs of precipitation.

Subject raising

Data

Given what we have just said about expletive there, consider the contrast between (42) and (43).

(42) a. * There seems to continue the same problem.
b. * There seemed to ring the mail carrier.
(43) a. ok There seems to be a problem.
b. ok There seems to exist an antidote.

The ungrammaticality of (42) is not surprising, since seem does not belong to the class of verbs that license there; recall (37e), repeated here as (44).

(44)   * There seems no solution.

But by the same token, the grammaticality of (43) is unexpected.

A first clue to the answer to this puzzle lies in the fact that each of the examples in (42) and (43) contains two verbs, and hence two clauses, and not just one. In the examples that follow, the lower clause, which is a complement of seem, is enclosed in square brackets. The higher, or matrix, clause contains the bracketed complement clause. Now notice that neither of the complement verbs in (42) licenses there, whereas both of the complement verbs in (43) do. Thus, instead of considering all four sentences in (42) and (43) as parallel to (44), it is more insightful to draw the parallels in (45)-(48). In this view, the fact that seem, the matrix verb in all of the sentences, doesn't license there turns out to be a red herring.

(45) a. * There seems [ to continue the same problem. ]
b. * There continued the same problem.
(46) a. * There seemed [ to ring the mail carrier. ]
b. * There rang the mail carrier.
(47) a. ok There seems [ to be a problem. ]
b. ok There is a problem.
(48) a. ok There seems [ to exist an antidote. ]
b. ok There exists an antidote.

Analogous examples can be constructed for the other two types of special subjects. For instance, the sentences in (49) have an idiomatic interpretation, even though the subject idiom chunk isn't a constituent of the complement clause, along with the rest of the idiom.

(49) a.   The cat seems [ to have got X's tongue. ]
b.   The cat proved [ to be out of the bag. ]
c. The fur began [ to fly. ] ]
d. The jig proved [ to be up. ]
e. The pot seems [ to be calling the kettle black. ]

And finally, the contrast between (50) and (51) is parallel to that between (40) and (41).

(50) a. It continued [ to { hail, pour, rain, sleet, snow. } ]
b. It seems [ to be { hailing, pouring, raining, sleeting, snowing. } ]
(51) a. * The { air, atmosphere, environment, precipitation, sky, weather } continued [ to { hail, pour, rain, sleet, snow. } ]
b. * The { air, atmosphere, environment, precipitation, sky, weather } seems [ to be { hailing, pouring, raining, sleeting, snowing. } ]

Analysis

The upshot of the examples in (42)-(51) is that when the matrix clause contains seem or a similar verb, the subject of the matrix clause is licensed by an element in the complement clause, rather than within the matrix clause itself. Verbs that allow licensing that is ordinarily clause-internal to extend across clause boundaries in this way are called raising verbs.

The term 'raising verb' is potentially confusing since it is not the verb itself that undergoes movement. Rather, it is the complement subject that raises into the matrix clause, as we discuss in detail in a moment. A better term for the verb class in question might therefore be 'raising triggers.' However, we will continue to use the term 'raising verb' because it is standard in the literature.

Raising verbs are semantically deficient in not imposing selectional restrictions on the eventual matrix subject. As the contrasts in (52) show, verbs ordinarily select subjects with particular semantic properties. For instance, breathe, elapse, and write select subjects that are animate beings, periods of time, and literate humans, respectively.

(52) a.   The { cat, #rock } is breathing.
b.   More than a { week, #desk } has elapsed.
c.   The { boy, #cockatoo } wrote an essay.

Raising verbs, on the other hand, impose no such selectional restrictions. In fact, there is no reason to think that they are associated with a subject of their own at all. We will represent this fact by associating raising verbs with elementary trees that have no specifier position, as illustrated for seems in (53).

(53)    

We are now in a position to derive a sentence like (43a). The derivation of the complement clause is given in (54). Notice that in (54a), the eventual matrix subject, expletive there, is licensed by originating in the specifier of a VP headed by licensing verb (in this case, be).

(54) a.       b.  
Local licensing of there Substitute (54a) as complement of to

Substituting the structure in (54b) as the complement of the raising verb yields (55a), which in turn becomes the complement of the matrix I element, yielding (55b).

(55) a.       b.  
Substitute (54b) as complement of raising verb Substitute (55a) as complement of matrix I

As it stands, (55b) is not well-formed, because the complement subject has a nominative case feature that cannot be assigned in the lower clause. We deduce that the case feature is nominative from contrasts in analogous sentences like (56).

(56)     He/*Him seems [ to have a problem. ]

The reason that the case feature cannot be assigned in the lower clause is that neither Spec(VP) nor Spec(IP) in the lower clause stand in the proper structural relation to a head that assigns nominative case. In particular, neither V nor nonfinite I can assign nominative case; recall the ungrammaticality of (28) in Chapter 5. Moreover, even if either of these argument positions were governed by a case-assigning head, that wouldn't help since nominative case isn't assigned under government in English. Finally, the ungrammaticality of (57) forces us to conclude that raising verbs like seem cannot assign oblique case to the complement subject via mediated case assignment.

(57)     Itexpl seems [ him to have a problem. ]

Unable to receive case within its own IP, then, the complement subject is forced to move to the Spec(IP) of the matrix clause, as shown in (58). In this position, it is assigned nominative case under spec-head agreement by the finite I of the matrix clause.

(58)    

The movement of the subject from the complement clause to the matrix clause is called (subject) raising. Notice that the subject first moves within the complement from Spec(VP) to Spec(IP), and only then from the lower Spec(IP) to the higher one. We assume the first step on the grounds that all clauses require a subject in Spec(IP). Notice further that the locality constraint on there (according to which it must occupy the specifier position of a licensing verb) is maintained by its trace.

Be careful not to confuse raising verbs---the verbs that license subject raising---with the verbs that license the various kinds of special subjects. The two classes are distinct from one another, although it is possible for a verb to belong to both classes (just as it is possible for a single person to belong to two distinct clubs).

Further evidence for subject raising

By definition, all raising verbs take nonfinite complements. Some of these verbs, however, are also able to take finite complements headed by the complementizer that, as illustrated in (59).

(59) a.   Itexpl { appears, seems } [ that there is a problem. ]
b.   Itexpl { appears, seems } [ that itweather is raining. ]
c.   Itexpl { appears, seems } [ that the cat is out of the bag. ]

Sentences like those in (59) have two properties that support the analysis of subject raising just presented. First, as we mentioned, raising verbs impose no selectional restrictions on a subject. This is consistent with the fact that the matrix subject in the sentences in (56) is expletive it. In contrast to ordinary subjects, expletive it is semantically empty; that is, it has no properties such as animacy, humanness, and so on. Rather, it satisfies two purely syntactic functions---namely, entering into a case-assigning relationship with the matrix I and providing the matrix clause with a subject. In other words, it fulfills exactly the same syntactic functions as a subject that has undergone raising. Second, according to the analysis of subject raising just presented, the matrix subject originates in the complement clause and is forced to move because its case feature cannot be assigned there. If there were a way for the complement subject's case feature to be assigned in the complement clause, we would therefore expect it to remain within the lower clause, where it is locally licensed. As the sentences in (56) show, this is exactly what happens when the complement of the raising verb is finite. Since the complement subject's case feature can now be assigned by finite I within the complement clause, there is no need for the complement subject to undergo raising.

As we just mentioned, not all subject raising verbs are also able to take finite complements. Tend is one such subject raising verb.
(i) a.   There tend to be traffic jams during rush hour;
itweather tends to rain in Seattle;
the jig tends to be up sooner or later.
b. * Itexpl tends that there are traffic jams during rush hour;
itexpl tends that itweather rains in Seattle;
itexpl tends that the jig is up sooner or later.

Conversely, not all verbs that occur in the construction in (59) are also subject raising verbs.
(ii) a.   Itexpl occurs to me that there is a solution;
itexpl occurs to me that itweather never snows in Lagos;
itexpl occurred to him that the jig was up.
b. * There occurs to me to be a solution;
itweather occurs to me never to snow in Lagos;
the jig occurred to him to be up.

A look-alike: Control

Data

Subject raising verbs are not the only class of English verbs that take nonfinite complements. A second class, the so-called control verbs, is illustrated in (60).

(60)     Larry { attempted, decided, expected, hoped, tried } [ to travel to Bhutan. ]

Superficially, control verbs resemble raising verbs because both types of verbs take nonfinite complement clauses headed by to. However, the two verb classes differ in their syntactic behavior in two important respects. First, control verbs impose selectional restrictions on their subjects. For instance, hope selects a human subject, a requirement that is not met in (61).

(61)   # The chrysalis hoped [ to become a butterfly. ]

Moreover, the restrictions imposed by the control verb are distinct from those imposed by the verb in the lower clause. This is illustrated by (60), where the selectional restrictions of hope are met, but those of elapse are not.

(62)   # Larry hoped [ to elapse. ]

Further evidence for the semantic independence of the matrix and the complement subjects comes from the subset of control verbs that also allow finite complements. In such cases, the matrix and complement subjects need not refer to the same discourse entity.

(63) a.   Larry { decided, expected, hoped } [ that he would travel to Bhutan. ]
b.   Larry { decided, expected, hoped } [ that { we, she } would travel to Bhutan. ]

It is true, however, that when the complement of a control verb is nonfinite, the interpretation of the complement subject is constrained in a way that it isn't when the complement is finite. Thus, the sentences in (60) with counterparts in (63) can only have the interpretation in (63a), never those in (63b). When the complement of a control verb is nonfinite, the matrix subject can be thought of as fixing or 'controlling' the referent of the complement subject; hence the name of the verb class.

Notice that the counterparts of (63) with raising verbs replacing the control verbs in the matrix clause are completely ungrammatical.

(64) a. * Larry { appears, seems } [ that he has traveled to Bhutan. ]
b. * Larry { appears, seems } [ that { we, she } have traveled to Bhutan. ]

The reason is that nonexpletive subjects like Larry need a semantic role like agent, patient, beneficiary, and so on. Such semantic roles are associated with verbs, but not with I, and they are assigned to the argument positions of the heads they are associated with. But since raising verbs have no specifier, and since the complement position in (64) is filled by the that clause, the matrix subjects in (64) end up without a semantic role.

There is a second major difference between raising verbs and control verbs. Since control verbs require their own subject, but do not license any of the special subjects discussed earlier, these special subjects should be ruled out as the subjects of control verbs. This expectation is borne out; the sentences in (65) are either downright ungrammatical or do not allow the relevant idiomatic interpretation.

(65) a. * There { attempted, decided, expected, hoped, tried } [ to be a problem. ]
b. ?* Itweather { attempted, decided, expected, hoped, tried } [ to rain. ]
c. The cat { attempted, decided, expected, hoped, tried } [ to get X's tongue. ]
(only literal interpretation possible)
d. # The jig { attempted, decided, expected, hoped, tried } [ to be up. ]

Analysis

We now present an analysis of control verbs that captures the syntactic differences between them and raising verbs. As we saw in (61), control verbs, unlike raising verbs, impose selectional restrictions on their subjects. We represent this by including a specifier position in the elementary tree for control verbs, as illustrated for decide in (66). Notice further that the complements of control verbs are CPs, not IPs; we return to this point in a moment.

(66)    

Moreover, as we saw in (62), the verb of the complement clause also imposes selectional restrictions. Therefore, this verb too needs a specifier position, which is occupied by a subject that is distinct from the matrix subject (though it will be coreferential with it). The way we represent this is by assuming that English has a silent pronoun, so-called PRO (read as 'big pro'). It can be thought of as comparable to the 'you understood' of imperative sentences like (67a).

(67) a.   ___ be quiet!
b.   You be quiet!

It is this silent pronoun that occupies Spec(VP) of the control verb's nonfinite complement, as shown in (68), much as an overt pronoun occupies that position in a finite complement (recall (63)). The derivation of the complement clause is shown in detail in (68). As just mentioned, we address the issue of the CP in (68c) in a moment.

(68) a.       b.       c.  
PRO as subject of verb of complement clause Substitute (68a) as complement of to Substitute (68b) as complement of silent complementizer

Substituting the structure in (68c) as the complement of the control verb yields (69a), which in turn becomes the complement of the matrix I element, yielding (69b).

(69) a.       b.  
Substitute (68c) as complement of control verb Substitute (69a) as complement of matrix I

Finally, subject movement of the matrix and complement subjects yields (70). Once again, we assume that the complement subject---in this case, PRO---moves from Spec(VP) to Spec(IP) on the grounds that every clause has a subject.

(70)    

Why is the complement of decide assigned the category CP rather than IP? The reason is that decide, along with certain other control verbs, allows indirect questions as complements, as in (71).

(71)     Larry must decide [ whether to travel to Bhutan. ]

Since whether is a complementizer, the indirect question in (71) cannot be an IP, but must be a CP, as shown in (72).

(72)    

For uniformity, we extend this CP analysis to the complements of all control verbs, even to the complements of control verbs such as hope or promise that do not allow indirect question complements.


Notes

Since whether is a complementizer
It has been argued that whether is not itself a complementizer, but that it occupies the specifier position of a silent complementizer. For present purposes, the exact position of whether (head or specifier) is not important, but only that the entire indirect question is a CP.