On the scope of the referential hierarchy in the typology of grammatical relations
Split alignment in case-marking has often been hypothesized to be affected by the referential hierarchy in that higher-ranking arguments are more likely to be accusatively aligned while lower-ranking arguments are more likely to be ergatively aligned. A survey of typological databases suggests that while there are only very few counterexamples to the hypothesis, the total number of relevant cases is so small as to provide only marginal statistical evidence. In split alignment in agreement morphology and in structures like relativization, the statistical evidence is nil. The only domains where hierarchy effects seem to be common is (a) where they affect optional case-marking in discourse and (b) where they directly define grammatical relations, independently of alignment patterns, such as in hierarchical agreement.