How universal is agent-first? Evidence from symmetrical voice languages

Sonja Riesberg, Kurt Malcher, Nikolaus P Himmelmann

    Research output: Contribution to journalArticle


    Agents have been claimed to be universally more prominent than verbal arguments with other thematic roles. Perhaps the strongest claim in this regard is that agents have a privileged role in language processing, specifically that there is a universal bias for the first unmarked argument in an utterance to be interpreted as an agent. Symmetrical voice languages such as many western Austronesian languages challenge claims about agent prominence in various ways. Inter alia, most of these languages allow for both �agent-first� and �undergoer-first� orders in basic transitive constructions. We argue, however, that they still provide evidence for a universal �agent-first� principle. Inasmuch as these languages allow for word-order variation beyond the basic set of default patterns, such variation will always result in an agent-first order. Variation options in which undergoers are in first position are not attested. The fact that not all transitive constructions are agent-first is due to the fact that there are competing ordering biases, such as the principles dictating that word order follows constituency or the person hierarchy, as also illustrated with Austronesian data.
    Original languageEnglish
    Pages (from-to)523-561
    JournalLanguage (Washington)
    Issue number3
    Publication statusPublished - 2019


    Dive into the research topics of 'How universal is agent-first? Evidence from symmetrical voice languages'. Together they form a unique fingerprint.

    Cite this