Unaccusativity and the active/stative split in Guarani

Guarani is known as one of the canonical active/stative languages in which intransitive verbs vary in whether they take so-called active (looks like subject) or so-called stative (looks like object) agreement. However, there are two problems with this view: i) lots of exceptions are attested and ii) it begs the question of how verbal agreement can be sensitive to the semantic content of the root. In this talk, I reframe the active/stative split as syntactic and not strictly semantic. I propose that actives are unergatives and statives are unaccusatives, based on novel language-internal diagnostics ( i) passivization, ii) additional argument introduction, and iii) imperatives). The question still remains of how verbal agreement is sensitive to unaccusativity. I model this using Deal’s (2015, 2022) Interaction and Satisfaction model, with an adaptation from Bejar (2003). In short, probes become less picky after failed Agree, allowing us to model instances of non-fatal failed Agree.