In this paper, we provide evidence based on case alternations in passives in favor of the view that dative is a mixed case. Dative, but also other cases, has the property of being either inherent/lexical or structural. We propose an analysis of datives aiming to account for their mixed status within and across languages. Building on Řezáč’s (2008) theory of opacity vs. transparency of theta-related case to Agree, combined with a (modified) theory of case alternations in terms of m(orphological)-case (Marantz 1991), we propose that dative arguments are PPs, unlike accusatives which are DPs. Being complements of the phasal head P, dative DPs are invisible to an outside probe, Voice or T, for Agree. Under certain conditions, however, they become visible: either when a φ-probe is present on P probing and transmitting the features of the DP embedded below it or when P incorporates into the Voice-v complex lifting the phase-hood of the PP. We argue that the actual distribution of m-cases in actives and passives of languages with alternating datives is determined at the PF component, subject to the case-realization disjunctive hierarchy proposed by Marantz (1991). A dative argument entering Agree qualifies as having ‘dependent case’ in the sense of Marantz (1991) and not as having lexically governed case. Finally, we propose that cross-linguistic differences concerning the environments where dative alternations happen (passives vs. middles) depend on the head where the φ-probe entering Agree with dative DPs is located: Voice or v.
- dependent case.