Ellipsis and Anaphora Interpretation

My work on VP-ellipsis, gapping, and pronoun interpretation in the context of coherence establishment is also part of a broader research program concerned with anaphoric processes.

An analysis of strict and sloppy readings in VP-ellipsis and related constructions I have developed accounts for the correct set of interpretations for a wide assortment of benchmark examples [Kehler 1993a]. (Analyses proposed by [Crouch 1995] and [Egg et al. 2001] have also incorporated and extended this approach.) Stuart Shieber and I showed that previously proposed discourse determined approaches to recovering these readings in VP-ellipsis at best reduce the problem to the same problem in another guise, that is, determining such readings for deaccented VPs, to which analyses like mine and others can be generalized to apply [Kehler and Shieber 1997]. Jerry Hobbs and I showed that the empirical advantages of my analysis can be seen to result as a by-product of establishing parallelism among clauses [Hobbs and Kehler 1997]. In addition, a brief squib I recently wrote presents examples that cannot be handled by any syntactic or semantic approach of which I am aware, such as (1), in which neither the syntactic nor semantic material needed for the sloppy reading in the elided clause is made directly available by the antecedent clause [Kehler 2002a].

(1) Mary's boyfriend gave her his school picture, just as all schoolboys do. [= give their girlfriends his school picture]

Gregory Ward and I have published an analysis of so and do so anaphora [Kehler and Ward 1995, Kehler and Ward 1999a, Kehler and Ward 1999b] that demonstrates and accounts for the fact that these forms are neither syntactically controlled (contra [Hankamer and Sag 1976], inter alia) nor can be accommodated in current theories of cognitive status [Prince 1992, Gundel, Hedberg, and Zacharski 1993, inter alia]. In this analysis, the so in do so is shown to be a special case of so anaphora of the sort illustrated in (2), in which the referent is not only nonsyntactic, but must be inferred.

(2) Regarding a possible Elvis Presley stamp, Postmaster General Frank notes that anyone so honored must be ``demonstrably dead'' for 10 years.

Gregory and I have also written a paper for the Handbook of Pragmatics that demonstrates several respects in which theories of information status are unequipped to handle the constraints imposed by a variety of forms of event reference [Kehler and Ward 2004].

I also continue to pursue research program concerned with the interpretation of pronouns. In [Kehler 1997], I showed the untenability of a popular and influential approach to pronoun interpretation using centering theory [Brennan, Friedman, and Pollard 1987], by demonstrating that this approach suffers from the very same drawbacks of other, more semantically-rich methods that motivate a centering-based model in the first place. This result has inspired other researchers to develop different, incremental versions of centering (e.g., [Strube 1998]), but I believe that other problems discussed in my book argue forcefully that any approach to pronoun interpretation based primarily on the superficial cues on which centering relies will be inadequate.

Andy Kehler 2002-07-01