Case assignment and the linear order of coordinated verbs

Abstract In this paper, we draw on corpus data to show that in Icelandic, verbs that assign distinct cases can be coordinated and share a single object: the verb on the right determines the case that the object bears. However, it turns out that fine-grained details of how case is realized on the object morphologically have an effect on which verb is more likely to come first: the one that assigns accusative or the one that assigns dative. If the object is syncretic for accusative and dative, then there is no preference, and both word orders are equally frequent. If the object is not syncretic, then there is a preference to put the dative-assigning verb last. But this preference is not equal for all objects: when the accusative is realized by a zero affix, the preference is weaker than when the accusative is realized by a non-zero affix. We present an analysis of these facts that is grounded in formal spellout mechanisms and the following two guiding principles: choose the order that expresses the most case features, and choose the order that uses the fewest mechanisms. Non-syncretic objects vary in the strength of the preference in a way that can be connected to these two principles, as long as we make certain specific assumptions about how spellout works, most importantly that zero affixes result from the absence of Vocabulary Insertion, rather than the insertion of a phonologically empty symbol, and Impoverishment exists as a spellout mechanism, one that is distinct from Vocabulary Insertion.