Abstract
In this work we analyze objects with non-canonical case markers in Turkish. We show that semantic factors give rise to non-canonical case marking and try to dertermine criteria to decide on the grammatical function status of non-canonical objects. We conclude that ablative partitive and dative objects participating in an affectedness alternation should be analyzed as straightforward objects (OBJ), but that lexically specified non-canonically case marked object should be treated as OBJ-TH. Based on our analysis, we provide an LFG implementation as part of the Turkish ParGram grammar and show how this analysis provides just the right results for the data observed with respect to causativization, passivization, and raising.