This project investigates alternating predicates in Romanian (Ma enerveaza atitudinea lui, lit. me.ACC irritates attitude.the.NOM his vs. Atitudinea lui ma enerveaza, lit. attitude.the.NOM his me.ACC irritates), and their evolution over time, as a part of a theory-informed empirical study. Alternating predicates are found in several Indo-European (IE) languages. They are a type of oblique subject construction which selects for two distinct and opposed argument structures: DAT-NOM / NOM-DAT and ACC-NOM / NOM-ACC (Barðdal et al. 2019), of which DAT-NOM and ACC-NOM are often analyzed as cases of topicalization (Dobrovie-Sorin 1987). Remarkably, when the order of the arguments is DAT-NOM or ACC-NOM, the dative or the accusative takes on the subject role, while when the order is NOM-DAT, or NOM-ACC, the nominative behaves as a subject. Among IE languages, Romanian has a unique status, showing a high number of accusative subject-like constructions, as is revealed in a pilot study I carried out (cf. also Van Peteghem 2016). The question arises as to how many of the accusative subject constructions are alternating in Romanian, and whether the alternating predicate construction is an innovation, having become productive in the history of this language, or is an IE inheritance. This project investigates the productivity of the construction from a diachronic and synchronic perspective, combining corpus analyses with psycholinguistic experiments.