Why agent prominence persists even under challenging conditions

The agent of an event – the one who is performing the action – plays a central role in human cognition and in linguistic structure. Critically, the privileged role of the agent is argued to be a general phenomenon, relevant for all languages. However, in this paper, we zoom in on typological patterns that deviate from the typologically prevalent way of coding agent prominence. We focus on languages in which agents may not be marked as default and on languages that do not exhibit a general preference for placing the agent argument in sentence-initial position, namely Tima (a split ergative language) and Totoli (a language with a symmetrical voice system). Totoli also does not have a preference for linking agents to subject functions. Here we shed new light on how agent prominence is reflected in these typologically diverse languages. Furthermore, by bringing together typological studies, corpus work, and elicitation data, as well as evidence from psycholinguistic and neurophysiological studies, we conclude that agents maintain a privileged status across languages, even if typological features seem to suggest otherwise. More generally, we propose that cross-linguistic comparison – especially considering data from highly diverse languages – offers key insights into which aspects of agent prominence interact with language-specific properties and how a concept of a general agent prominence still remains universally applicable.