The macro-event property: The segmentation of motion paths and causal chains

At the center of this presentation is the approach to the crosslinguistic study of event segmentation in language developed by my former colleagues and me within the Event Representation project at MPI Nijmegen. Previous studies have assessed semantic representations of complex events by the amount of lexical information packaged in syntactic units (clauses, verb phrases, etc.; Pawley 1987) or the number of syntactic units packaged between intonation breaks (Givón 1991). The principal limitation of these approaches comes from the lack of a universally valid correlation between syntactic or intonation units and semantic/conceptual event representations. Instead of taking units of syntax or phonology as the starting point, the Nijmegen approach starts from a unit of form-to-meaning mapping, called a ‘macro-event’ (with a term borrowed from Talmy 2000). A construction has the macro-event property to the extent that it packages an event representation in such a way that temporal/aspectual operators cannot access proper sub-events individually, but can only refer to the macro-event as a whole. This definition spells out a number of readily administered semantic tests of the macro-event property and allows for variation in the constructions used in a particular language to encode macro-events. I will introduce the Nijmegen approach drawing on research in the domain of complex motion events, to then move on to an application in a recent pilot study on the segmentation of causal chains in Dutch, Ewe, Japanese, Lao, and Yukatek Maya. In both studies, an unpredicted amount of crosslinguistic variation emerges, driven chiefly by differences in lexicalization. Time permitting, I will also discuss the hypothesis that the core within the 'layered structure of the clause' in Role and Reference Grammar universally has the macro-event property.