Sense and simplicity

I show that two different motivations drive the differential case marking of direct objects cross-linguistically. On the one hand, direct objects can be marked to signal their markedness with respect to certain semantic features (local distinguishability). Opposed to this we find systems where overt object marking is dependent on global distinguishability and is only applied in cases of actual ambiguity or comparison between subject and object features. In some DOM systems we even find both strategies at work. I argue that these two strategies correspond to different modes of optimization. Whereas local DOM systems can be modeled by referring only to productive optimization, global systems require a model in which interpretive optimization plays a role as well. I introduce an asymmetric model of bidirectional optimization in which the outcome of production is constrained by interpretation. This model will be shown to provide a straightforward analysis of different DOM patterns, in contrast to existing models of bidirectional optimization.