Mainland Scandinavian object shift and the puzzling ergative pattern in Aleut

Eskimo-Aleut languages turn out to have the same two types of object shift that \citet{holmberg1986word} describes for Scandinavian. Specific objects move out of the VP in Inuit \citep{bittner1996ergativity} and I argue that object shift also occurs in Aleut, but it is limited to pronouns as in Mainland Scandinavian. Aleut differs from Mainland Scandinavian in that, for independent reasons, only third pronouns successfully undergo object shift. Shifting first and second person pronouns is blocked by PCC-like constraints on the portmanteau agreement that occurs in object shift constructions. Shifting reflexives is also blocked, because it would incur a violation of the Anaphor Agreement Effect.
The surface pattern in Aleut has been described as one where ergative case marks the subject only when another argument in the clause is null. I argue that there is no direct cause and effect relationship between these. The key is the fact that pronouns that agree are not spelled out. Agreement correlates with ergative case because, as in Inuit, ergative case marks the subject in Aleut only when the object moves out of the VP, and in this situation, again as in Inuit, there is portmanteau agreement with the ergative subject and nominative object in object shift constructions. Like Inuit, Aleut has possessor raising/stranding so that the possessor of an object can undergo object shift, trigger agreement, and thus pro drop. From the English translations of Aleut sentences, it initially appears that null objects of prepositions also correlate with ergative subjects, but Aleut, like Inuit, has possessed relational nouns which function like prepositions if they take locative case. These also allow possessor raising and object shift, with the same consequences described above.