No i-Sums for Nissim (and Shalom)

Lappin and Francez (1994) present a theory of donkey anaphora which, they claim, captures both their existential and their universal readings, while maintaining a uniform representation of donkey pronouns. This paper shows that their analysis does not in fact assign the correct truth conditions to donkey sentences and so does not account correctly for the distribution of readings. An alternative analysis is proposed which retains LF’s uniform representation for donkey pronouns, but abandons their analysis in terms of i-sums and the corollary derivation of universal readings by means of a maximality constraint. On the proposed analysis, donkey pronouns are uniformly represented with free variables over (Skolemized) choice functions, as in Chierchia’s (1992) E-type analysis. The quantification associated with them is inferred quantification over choice functions. Universal readings arise as in Chierchia (1992) when all possible values for the free variable in the representation of a donkey pronoun are salient. For existential readings, a pragmatic account in the spirit of LF’s analysis in terms of a cardinality constraint is maintained.