Landau (2000) identifies a split between P(artial) C (ontrol) predicates (e.g., want), which admit a proper superset relation between the controller and PRO, and E(xhaustive) C(ontrol) predicates (e.g., try), which require absolute identity:

(1) a. John_1 wanted [PRO_{1+c} to eat lunch together at noon]. (c = contextually salient others) b. *John_1 tried [PRO_1 to eat lunch together at noon].

Landau (2000); Pearson (2013) both single out tense as a crucial predictor of PC: roughly, predicates whose complements are future- or past-oriented allow PC whereas predicates whose complements are present-oriented disallow PC. However, disagreement over the facts is rampant (see especially Landau 2000; Jackendoff and Culicover 2003; Bowers 2008; Rodrigues 2008; Pearson 2013), and to our knowledge no experimental investigation of PC has yet been undertaken. Consequently, the goal of the work described here is to use experimental techniques to address three questions:

a. Are Landau (2000), Pearson (2013) right that predicates differ in their tolerance for PC? b. If so, do the temporal properties of the predicate reliably predict the predicate’s tolerance for PC, as implied by Landau’s (2000) and Pearson’s (2013) analyses? c. Is there evidence for gradability in the acceptability of PC, and if so, which theory of PC best accounts for it?