Symmetric and Interacting Alternatives for Implicature and Accommodation∗

This paper uses the proviso problem to argue that accommodation is computed by the system responsible for scalar implicature. Standard procedures for computing implicatures involve: (i) Generating a candidate set of potential implicatures based on formal principles, (ii) Employing an algorithm over this set of candidates to determine which become actual. I argue that a very similar computation takes place in accommodation: A candidate set of accommodations is generated, from which an algorithm selects a subset as the actual accommodations. The parallel runs extremely close. The candidate sets for both implicature and accommodation will be argued to derive from a single source, namely, the scalar alternatives of the asserted sentence. Moreover, the algorithm for converting potential enrichments into actual enrichments will be argued to be the same. Specifically, the algorithm will take the union of the two sets as input, and will compute implicatures and accommodations together. This move will allow potential implicatures and potential accommodations to cancel each other out, in a sense to be made precise. This will be argued to be the key to solving the proviso problem, in addition to some problems that I’ll identify in the theory of implicature. The general architecture has obvious roots in the work of Gazdar [12], Soames [45], and Heim [22]. Some of the specific technology employed depends on recent developments in the theory of implicature, specifically, Katzir’s [27] procedure for generating scalar alternatives, and Fox’s [8] algorithm for converting potential implicatures into actual ones.