Why control of PRO in rationale clauses is not a relation between arguments

Alexander Williams

Abstract:

After (1) or (2) we can use (3), a rationale clause, to mean (4). We then mean that those who acquire a pitcher are the traders of the outfielders.
(1) The team traded away two outfielders
(2) Two outfielders were traded away
(3) PRO to acquire a pitcher.
(4) because then the trader might acquire a pitcher.
On grammatical accounts, this indicates a relation between two arguments that ensures their coreference, for any assignment of values to variables. PRO in (3), for example, would be related in this way to an argument that is linked to the role of trader in either (1) or
(2). Such accounts have good motives, sketched in §3. But in this paper we make two objections, one syntactic and one semantic. The syntactic objection comes from remote control, as in (5). We can use (5) just like (3), again to mean (4) (Higgins 1973, Dowty
1989, Sag & Pollard 1991, Williams 2015). Yet in this case, we will argue in §4, there can be no local binder for PRO, when there isn’t one audible.
(5) The goal was PRO to acquire a pitcher.
The semantic objection comes from the truth-conditions of the rationale clause construction.
These are hyperintensional, and compositionally this is in conflict, we will suggest in §5, with the binding relation that a grammatical account requires. These two objections, if sound, add weight to a non-grammatical account: PRO in a rationale clause acts not as a bound variable, but just as an free expression of type e over a restricted domain (Landau 2000, Williams 2015). We discuss the needed restriction and its implementation in §6 and §7, respectively, before ending with a remark on the broader relevance of our conclusion.