This study argues that split-ergative alignment in Dutch nominalisations are functionally motivated by discourse, drawing on the concept of Preferred Argument Structure (Du Bois, 1987) and the contrast between fore- and backgrounded discourse (Hopper & Thompson, 1980). Presented corpus evidence confirms the claims. Finally, the analysis is molded in a Functional Discourse Grammar model (Hengeveld & Mackenzie 2008), covering its so far insufficiently outlined Contextual Component.