Friday, July 24, 2015

Controlling Purpose Clauses

This is the first post in quite a while. Again, administrative duties take up way too much time.

Today's post will look at one particular fact from Hallman (2015). I will not mention all the facts about double object constructions that Hallman fails to discuss. Hallman does present one novel fact and proposes an explanation for it. Unfortunately, the explanation does not work, as I will show here.

The fact is that the object of the preposition to with verbs like give can control the null subject of a purpose clause, as in (1) below:

(1) Mary gave a puppy to John [to play with]. (John is the one to do the playing; Hallman 2015, 392, (7a))

In contrast, the object of a preposition with a verb like put cannot control the null subject of a purpose clause:

(2) *Mary put the child on the horse [to carry]. (intended: the horse will carry the child; Hallman 2015, 393, (9a))

With such a verb, the null subject must instead be controlled by the main clause subject, or be interpreted as arbitrary (like one):

(3) John put his Complete Works of Shakespeare on the floor [to sit on]. (John or anyone will sit on the books; Hallman 398, (19a))

Hallman proposes that the object of the preposition with verbs like put does not c-command the purpose clause. In order to control the null subject of an infinitive, an NP must c-command that null subject. So he proposes a structure for put where the PP is lower than the purpose clause (see the tree in his (19b), page 398). With give, the PP is higher (see the tree in his (24) page 402).

This c-command explanation is a common one. For instance, Koizumi (1994) and Hale and Keyser (1996), among others, propose that indirect objects may not control depictive secondary predicates because they do not c-command them. Similarly, objects of prepositions may not control depictive secondary predicates because they do not c-command them. (Note that in Hallman's examples, the object of a preposition does not actually c-command anything outside of the PP. Hallman seems to be assuming that PPs don't count for c-command.)

The problem with this c-command explanation is that standard tests indicate that the NP that cannot control does c-command the adjunct. For instance, Binding Condition C and quantificational binding indicate that an indirect object c-commands a depictive secondary predicate:

(4) I served her a steak rawer than Mary wanted it. (her cannot be Mary)
(5) I served every customer a steak rawer than he wanted it.

Turning back to Hallman's purpose clauses, Binding Condition C and quantificational binding show that the object of the preposition with a verb like put does in fact c-command the purpose clause:

(6) I put a chair beside her [to sit on while reading to Mary]. (her cannot be Mary)
(7) I put a chair next to every invalid [to sit on while reading to him].

Similarly, NPI licensing and the each...the other construction also indicate that the object of the preposition c-commands the purpose clause:

(8) I put a chair next to no invalid [to sit on while reading to anyone].
(9) I put a chair beside each invalid [to sit on while reading to the other].

These phenomena are standardly taken to require c-command. They indicate that the NP that cannot control the subject of the purpose clause nevertheless c-commands it. It therefore cannot be correct that the NP is unable to control the subject of the purpose clause because it fails to c-command it. Some other explanation is necessary.

References

Hale, Ken and Samuel Jay Keyser (2002). Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press.

Hallman, Peter (2015). Syntactic Neutralization in Double Object Constructions. Linguistic Inquiry 46:389-424.

Koizumi, Masatoshi (1994). Secondary Predicates. Journal of East Asian Linguistics 3: 25-79