Professor Ezra Keshet recently gave a presentation at the joint 1st International Pragmatics Conference of the Americas and the 5th International Conference on Intercultural Pragmatics, held in Charlotte, North Carolina. This was in inaugural meeting of the recently formed  American Pragmatics Association (AMPRA). This association comprises the countries of North, Central, and South America, and the Caribbean, and supports interaction and scholarly debate with particular regard to the linguistic, cognitive, social, inter-cultural and inter-language paradigms of pragmatics. 

In his presentation, Ezra explored one of the classic problems of formal semantics, so-called "donkey anaphora", from a new angle. The title and abstract of his presentation follows below.

Coherence and Donkey Anaphora

Coherence relations can hold between two clauses in a quantified sentence:  

(1)   Every time I get an `A' one day, I get a `B' the next day. [Parallel]

(2)   No one who drinks Pomegranate juice regularly dies young. [Result]

These suggest local resolution of a coherence relation holding over a range of related situations. Coherence also drives pronoun resolution, as demonstrated by Kehler et al. (2008):

(3)  Samuel threatened Justin with a knife, and ...
      a. Erin blindfolded him. [Parallel]
      b. Erin stopped him. [Result]
      c. he blindfolded Erin. [Parallel]
      d. he alerted security. [Result]

Experimental participants overwhelmingly chose the pronoun referent most compatible with the coherence relation in effect (Samuel in b&c, Justin in a&d). I observe similar effects in quantified sentences containing so-.-called donkey pronouns:

(4)   Whenever a man threatens another man with a knife, …
       a. An accomplice blindfolds him.
       b. Security stops him.

Even without well-.-defined referents for the pronouns, the same biases result (first man in b, second in a). I propose that coherence resolution affects the pronoun assignment function. This is simple in (3), where the related clauses are full sentences, but more dynamic in (4). For instance, he in (4b) resolves to whomever security is most likely to stop when someone threatens someone else with a knife. This analysis explains the tight connection between coherence and pronoun resolution better than E-.-type, DRT, and Dynamic Semantics analyses. For instance, consider the Indistinguishable Participants problem (Heim 1991):

(5)   When a bishop meets a/nother bishop on the road, he blesses him.

Although the meeting relation is symmetrical, he seems to refer to the subject-position bishop, while him refers to the object. My analysis predicts this resolution as necessary to satisfy a Parallel relation between the two clauses.