On the adpositional nature of ergative subjects

This paper argues that ergative ‘case’ is not a case morpheme, but a morphologically dependent theta-assigning adposition. The proposal explains agreement variation within ergative languages (e.g., ergative agreement, as in Circassian, accusative agreement, as in Warlpiri, and the absence of agreement with the ergative subject, as in Hindi). Specifically, we propose that ergative agreement arises when the verb agrees with the adpositional features and the φ-features of the ergative subject, while accusative agreement arises when the verb agrees solely with the φ-features of the subject. Ergative subject fails to trigger agreement when the ergative adposition acts as an agreement blocker. Whether or not the ergative adposition blocks agreement depends on the manner in which it combines with the embedded NP, a parameter along which ergative language vary. In addition to accounting for agreement variation in ergative languages, our proposal also accounts for the numerous parallels between ergative subjects and nominal possessors. The second part of the paper explores the roots of these parallels.11The following abbreviations are used: “3sg.subj/obj” = “3rg singular subject or object”; the period “.” separates the different meaning components of a single morpheme, for example “je” is glossed as “3sg.subj” to indicate that the same morpheme means both 3rd singular and subject; the dash “-” separates different morphemes; “m”/“f” refer to masculine and feminine respectively; numbers as in “3” in Chichewa example (17) refers to classes of nouns.