Psych-adjectives and semantic selection

A puzzling generalization, first noted by Faraci (1974), states that (non-causative) psychological adjectives tolerate at most a subject gap in their infinitival complement whereas non-psychological adjectives require exactly one gap (either subject or object). This paper argues that the generalization follows from the fact that the infinitive is a (propositional) argument of a psych adjective but a (predicative) modifier of a non-psych adjective. A series of tests (ellipsis, extraction, extraposition and P-stranding) confirms this asymmetry. A-bar binding is responsible for both subject-gap complements to non-psych adjectives and subject-gap infinitival relatives, explaining their crosslinguistic correlation. This strongly suggests that obligatory control does not fall under operator-abstraction, as argued by predicational treatments of control, but rather involves a different mechanism.