{eventive verb + VP}

I propose that the behaviour of an eventive verb with a VP complement can be summarised as follows:

• Raising Interpretation:
  • The VP complement must be an uncontrolled event.
  • The VP complement cannot be an achievement, but can be of any other eventive type.
• Control Interpretation:
  • The VP complement must be an event controlled by the subject NP.
  • The VP complement can be of any eventive type.

The distinction between the complements on the raising and control interpretations, in the latter case that the event must be controlled by the subject NP and in the former that the event must be uncontrolled, follows simply from the definition of raising and control verbs. I postulate two distinct lexical entries for the general form {eventive verb + VP}, corresponding to the raising and control interpretations. This is supported by co-predication data, as shown by beg71-beg72 in constrast to beg94.

 

*John began to read the book and find coffee stains on it. John began to read the book. He began to find coffee stains on it.

 

*John began to dance and to feel ill. John began to dance. He began to feel ill.

  John began to sing and dance/laugh and cry/jump and scream.

These examples show that the raising and control uses of begin correspond to distinct lexical entries: beg71-beg72 evidence that it is impossible for a compound VP in which each of the conjunct VPs demand a different interpretation of begin to appear with a single instance of begin, in contrast to beg94 in which each of the conjunct VPs requires the same logical interpretation of begin (and can therefore co-predicate of the same lexical entry). Therefore it would not be enough to claim that begin is ambiguous between the control and raising interpretations on a single use.

The other claims, that the VP complement of begin on a control interpretation is unconstrained while it is constrained on a raising interpretation not to be an achievement, are evidenced by the data in beg60. The judgement of controlled vs. uncontrolled here is made on the basis of Perlmutter's test frame Mary forced John to X - if this is an infelicitous sentence, e.g. *Mary forced John to feel ill, then the event described by X is uncontrolled, otherwise, e.g. Mary forced John to leave the party, the event is controlled.

 

The acid began corroding the marble. [uncontrolled process] The guests began arriving. [uncontrolled process] John began to find fleas on his dog. [uncontrolled process] John began to feel ill. [uncontrolled state] *John began arriving. [uncontrolled achievement] *John began to find a flea on his dog. [uncontrolled achievement] John began to leave the party. [controlled achievement] John began to dance. [controlled process] John began to read the book. [controlled accomplishment]

The aspectual restrictions proposed, and supported by the data, thus conflict with the Pustejovsky & Bouillon claim that begin must have a complement of a left-headed transition on a control use and can have a complement of any aspectual type on a raising use. This proposal reflects the data more accurately and remedies the problems raised by the Pustejovsky & Bouillon analysis of sentences such as beg57 in Section 5.4.1 above.