The
traditional theory of PRO does not provide a general
theory of control Depending on the nature of the verb
PRO can be controlled by either the subject or the object
and verbs must be marked for subject or object control.
Sometimes normal control properties are reversed (Chomsky;
1981b: 76). Sometimes the choice of the controller is
determined by theta roles on other semantic properties
of the verb (Chomsky; 1981b : 76). Chomsky observes
that in the purposive infinitival construction, control
may be determined by the "properties of the matrix
verb" (Chomsky; 1981b ; 77), although he does not
specify what this property is. He further observes that
c-command is not required for control and that the antecedent
of PRO may be contained within an NP that is a possible
controller. There is no explanation as to why it is
indeed so and why although there is a general theory
of antecedents for anaphors there is none for PRO. PRO
is categorized as a pronominal anaphor and this has
consequences on the position where PRO can occur and
where it cannot occur but it is not clear why this has
no bearing on the choice of antecedent of PRO. |
We do not propose a special theory of control here,
special in the sense that it does not operate on considerations
different from those that are used for the determination
of antecedents to other dependent arguments in the grammar.
PRO being as a pronominal anaphor, its pronominal and
anaphoric properties together determine its antecedent
in conjunction with some other considerations we mention
below. As a pronominal it is antecedent - free in P,
outside P it searches for an antecedent as does a pronoun
and an anaphor. As an anaphor, it chooses the mukhya
as its antecedent and as pronoun it chooses the non-mukhya
as its antecedent when two possible (animate etc.) arguments
exist outside P. Therefore, one would expect PRO to
have (i.e. more than one) split antecedents when there
are two arguments which can function as its antecedents.
When there is a single such argument, there is no problem
because as a pronominal, PRO does not have to have an
antecedent within the sentence. Otherwise if PRO does
not have split antecedents, then it is because one link
between PRO and its antecedent is rendered anomalous
by semantic or pragmatic properties of the relevant
predicate. When the predicate, of which PRO is an argument
requires plural antecedent, then PRO would have split
antecedents, this being the only way PRO can satisfy
this requirement. Consider: |
