Modeling progress: causal models and the imperfective paradox

Prerna Nadathur, Ohio State University

Joint work with Elitzur Bar-Asher Siegal (Hebrew University of Jerusalem)

Abstract

Under progressive marking, telic predicates (e.g., write a novel, build a house) can describe eventualities that fail to reach culmination. Prominent accounts of these imperfective paradox effects tie the truth of telic progressives to the accessibility of culmination (Dowty 1979, Asher 1992, Landman 1992, a.o.), intensionalizing the progressive operator (PROG), so that it instantiates qualifying (culminated) eventualities in modal alternatives to the evaluation world. This approach faces empirical challenges from acceptable progressives of unlikely-to-succeed events (e.g., cross a minefield) and progressives in 'out of reach' contexts, where culmination is precluded by reference-time facts. 

We propose that telic progressives are instead sensitive to structure inherited from an event type introduced by (telic) predicate P.   An event type constitutes a formal causal model (e.g., Pearl 2000) in which P's culmination condition C occurs as a dependent (caused) variable. The model provides a set of causal pathways for realizing C, each of which comprises a set of jointly sufficient conditions (events and/or properties) for C, and establishes (sets of) conditions which preclude C. On this approach, the progress of an actual (token) P-eventuality can be measured with respect to the event type. A reference time situation s satisfies PROG(P) just in case it is a plausible cross-section of an incomplete causal pathway in P

This account delivers improved judgements for challenging paradox data, by severing the truth of telic progressives from the local accessibility of culmination and assigning the intensional element of 'paradox' effects to the structure of telic predicates themselves.  Looking ahead, it suggests a new approach to the denotation of eventuality predicates, on which familiar aspectual class properties can be derived from features of (language-independent) causal models which capture common-sense intutions and idealizations about how the world works.