Tests to distinguish control and raising

There are several fairly simple tests to distinguish control and raising matrix predicates.

However, these tests operate only to distinguish subject control predicates from subject-to-subject raising predicates. About object-to-subject-to-subject raising predicates Postal (1971, 27) encouragingly states that “the behavior [of the adjectives involving object-to-subject raising]

seems to involve a myriad of complex and mysterious factors as yet little explored.”

However, since in object-to-subject raising constructions the grammatical subject originates in the object of the lower clause and there is no similar control structure applicable to

4 Object-to-subject raising rule applies to certain adjectives and some NPs (This house isa breeze to clean) (Postal 1971, 28).

5 See 2.4.

adjectival predicates, identification is simpler than between subject control and subject-to-subject raising predicates.

According to Carnie (2002, 262), the best and the most reliable way to distinguish raising predicates from control predicates is to resolve the theta grids. The theta grid of a certain predicate determines the minimal number of arguments that particular predicate takes.

Then we apply the theta criterion which is as follows: Each argument is assigned for one and only one theta (semantic) role, and each theta (semantic) role is assigned to one and only one argument. To decipher the matter, let us take example sentences, both control and raising (Carnie 2002, 259ff.):

(15) a. John is reluctant to leave.

b. John is likely to leave.

Here we have two predicates in both sentences:is reluctant andleave in (15a), andis likely andleave in (15b). Next we have to figure out the arguments they take. First,is reluctant takes two arguments: the one who is reluctant (theta role of experiencer) and the object of reluctance (theta role of proposition). Second,is likely takes only one argument, the occurrence which is likely (the role of proposition). Third,leave needs only one argument:

the one who leaves (theta role of agent) (ibid., ). Thus we have theta grids for all three predicates:

[experiencer]is reluctant [proposition]

is likely [proposition]

[agent]leave

Let us first consider (15a). Foris reluctant we have the experiencerJohn and the proposition to leave. Now all the arguments are assigned for theta roles, but the theta role of agent forto leave is still unassigned and there seems to be no more arguments left. Here we need an understood subject for the non-finite lower clause; an argument called PRO (ibid., 260) (see also 2.3.1. above). The theta role of agent is the assigned to PRO and the theta criterion is fulfilled:

(16) [John]experiencer is reluctant [[PRO] ^agent to leave] proposition

Compare this with the theta roles of the other sentence (15b):

(17) [____] is likely [[John] ^agent to leave ] proposition

The matrix predicate here does not assign an external theta role (ibid., 258), but in the surface structure, the subject is raised from the lower clause to the grammatical subject position in the higher clause. Based on this difference, we can say that if the matrix predicate assigns itself an external theta role (subject) it cannot be a raising predicate.

Langacker (1999, 320) introduces the idiom test. The argument is that in the given frame, only raising matrix predicates form grammatical sentences.

(18) a. Tabs arelikely to be kept on all the radicals.

b. *Tabs arereluctant to be kept on all the radicals.

Heretabs is limited to the idiomkeep tabs on and according to Langacker (1999, 319)tabs must have its origins in the lower clause because the rest of the idiom lies there. The reason for the ungrammaticality of sentence (18b) is thattabs cannot be the subject of a control predicate likebe reluctant, becausetabscan only occur as the object ofkeep in the idiom keep tabs on. Thus, we can derive a frame for distinguishing control and raising:

[Tabs] (predicate) [to be kept on… ]

If the predicate entered results in a grammatical sentence, it is a raising predicate and if not, it is a control predicate.

Carnie (2002, 262) uses a slightly different kind of idiom test, although it is based on the same principle as the test above. This test makes use of idioms, which can have either an idiomatic or a literal meaning. He uses the idiomthe cat is out of the bag to illustrate the point that with raising constructions the idiom has the idiomatic meaning, but with control constructions only the literal interpretation is possible. Examples from Carnie (2002, 263):

(19) a. The cat islikely to be out of the bag.

b. The cat iseagerto be out of the bag.

Sentence (19a) conveys the idiomatic meaning i.e. “there is a good possibility that the secret will be revealed”, whereas sentence (19b) can only be read in the literal way (there is a feline mammal in a bag with an appetite for freedom).

The idiom tests both have the same basic idea. As Carnie (2002, 262) states, “[t]he subject of an idiom must at some point be local to the rest of the idiom”. Only if this condition is fulfilled, the sentence can be grammatical in the first idiom test and have the idiomatic reading in the second idiom test. This is only possible with raising, because in control constructionsnothing moves. In raising constructions the subject is raised from the lower sentence where the rest of the idiom lies.

In addition, Langacker (1999, 319-320) states that “syntactic dummies”it andthere can only form grammatical sentences with raising predicates, not control predicates:

(20) a. It islikely to rain this afternoon b. *It iseager to rain this afternoon.

(21) a. There arelikely to be wombats orbiting Jupiter.

b. *There areeager to be wombats orbiting Jupiter.