I've never heard of a natural language that has ergative-absolutive marking alone. From what I've read, languages with said marking also have nominative-accusative marking, with the choice or "split" between the two types of marking conditioned by some semantic or syntactic factor. For example, in some languages we see the following: Clauses whose subjects are most likely to be agents have nominative-accusative marking. So we have "I-nom. ate the pork-acc.." But clauses whose subjects are less likely to be agents have ergative-absolutive marking. So we have "The rock-erg. hit the car-abs."
But from what I've read, said split can be conditioned by factors other than the agency, including aspect! Though I know that aspect does condition such a split, but why does it do so? What's the semantic or cognitive logic behind this?