The Macro - Event Property and the Layered Structure of the Clause

1. Introduction Typological research has uncovered a surprising amount of variation in which parts of a complex motion event (Bohnemeyer et al 2007) or causal chain (Bohnemeyer et al 2010) can be described as a single event across the languages of the world. The intuitive property of presenting a scenario as a " single event " has been operationalized in these studies in terms of the Macro Event Property (MEP), a form-to-meaning mapping property that constrains the compatibility of event descriptions with time-positional modifiers. In this article, we investigate the extent to which there is uniformity across languages in the structures that are associated with this property. In other words, we ask whether there is something like a " macro-event phrase " , a uniform level of syntax at which complex scenarios may be described as single events in the sense that time-positional modifiers necessarily locate their parts together in time. Drawing on evidence from English, Ewe, Japanese, and Yucatec Maya, we show that neither clauses nor verb phrases are inherently associated with the MEP. Instead, the MEP appears to be universally associated with constructions that comprise minimally a single lexical event de-scriptor and its syntactic arguments and maximally projections of multiple lexical event de-scriptors that share modifiers and operators (or 'functional categories', i.e., traditional grammatical categories such as tense and aspect) semantically associated with event descriptions, such as modifiers of time and place. We show that the correct generalizations can be captured using two key notions of the theory of inter alia): the 'Layered Structure of the Clause' and the classification of the 'nexus' types that constrain the possible internal structure of each layer. An examination of English event nominalizations, Ewe serial verb constructions, and Japa-nese converb constructions suggests that the putative crosslinguistic " macro-event phrase " may be the verbal (or, in the case of " eventive " nominals, nominal) core. Cores are constituted by expressions of semantic predicates and their syntactic arguments. Across languages, simple cores necessarily have the MEP, whereas complex cores have it if and only if they are integrated by 'cosubordinate' nexus, defined as a symmetric union of two cores that together behave like a single core, sharing operators and modifiers. We furthermore argue that this connection between core cosubordinations and the MEP may help explain a key empirical finding first reported in Foley & Van Valin (1984:261. 304): cosub-ordinate cores seem to always …