Position as a behavioral property of subjects

Abstract A subject analysis of oblique subject-like arguments remains controversial even across modern languages where the available data are not finite: while such arguments are considered syntactic subjects in Icelandic, they have more often been analyzed as objects in Lithuanian, for example. This issue has been left relatively neglected for the ancient Indo-European languages outside of Sanskrit (Hock 1990), Gothic (Barðdal & Eythórsson 2012), and Ancient Greek (Danesi 2015). In this article, we address the status of oblique subject-like arguments in Old Irish, whose strict word-order enables us to compare the position (relative to the verb and other arguments) of nominative subject arguments of the canonical type to oblique subject-like arguments. We first establish a baseline for neutral word-order of nominative subjects and accusative objects and then compare their distribution to that of oblique subject-like arguments under two conditions: i) on a subject analysis and ii) on an object analysis. The word-order distribution differs significantly across the two contexts when the oblique arguments are analyzed as syntactic objects, but not when they are analyzed as syntactic subjects. These findings add to the growing evidence that oblique subject-like arguments should be analyzed as syntactic subjects, although their coding properties are non-canonical.