From an empirical point of view, ergativity is a marked phenomenon. The pattern occurs in only a quarter of the world’s languages, and even those languages displaying it often apply it restrictively. Former analyses have not paid much attention to this fact, as most of them formulate a macro-parameter whose sole function is to distinguish between ergative and non-ergative languages. This study predicts the marked status of ergativity, deriving the pattern from an independently motivated parameter.

It is argued that Ergative case cannot be structural. Rather, it is like a semantic case in that it occurs on adjunct nouns in clitic-doubling constructions. Nonconfigurational languages like Warlpiri allow for ergative case marking because of the fact that they realize every verbal argument by a pronominal argument (PA). Adjunct nouns may double the PAs. In languages like Kurmanji, Basque, Northwest Caucasian and Mayan, both case and agreement may display ergativity. This is explained by assuming that only the transitive subject is clitic-doubled. Evidence for this explanation is found by comparing verbal inflectional paradigms to independent pronouns and by investigating the referential properties of the supposed adjunct nouns.

Ergativity, then, is linked to a macro-parameter dividing languages between those that do not allow for PAs and those that do. In languages with PAs, ergative patterns may be further restricted to certain values of functional heads such as I, accounting for split ergativity.

The marked status of ergativity is of relevance to both theoretical syntacticians and typologists interested in ergativity, agreement, case, clitic-doubling and nonconfigurationality.