the week of october 1st

pragmatic strengthening via world knowledge

This week's big topic was 'pragmatic strengthening'. We approached this topic by taking a close look at the little word or. The logic textbook interpretation of or says that a disjunction is true just in case at least one of its disjuncts is. Everyday disjunctions often seem to have stronger interpretations. We often interpret them exclusively, rather than inclusively. On its exclusive interpretation, a disjunction is true just in case exactly one of its disjuncts is. There is a broad consensus among philosophers, cognitive psychologists, and linguists today that the inclusive interpretation of disjunctions is basic and is delivered by the semantic interpretation component. The exclusive interpretation is pragmatically derived. The disagreement begins when we ask how. What are the pragmatic processes that take us from the inclusive interpretation of a disjunction to the exclusive interpretation we observe? A first possibility we considered was that pragmatic strengthening of a sentence is brought about via default assumptions about the situations described by the sentence. When I tell you that I entered South College at 9:00 AM, for example, you automatically assume that I entered through the door, rather than the window, and that I walked into the building and didn't fly. Combining your semantic knowledge with your knowledge about the way our world works allowed you to pragmatically strengthen what I literally said. We looked at many examples of exclusive interpretations of disjunctions where the strengthened interpretations we perceived seemed to be the result of our semantic knowledge interacting with world knowledge about the situations described. According to W. V. Quine, this represents the normal case for disjunctions. In his Methods of Logic (chapter 1) Quine remarked that most uses of or in everyday language are either obviously inclusive or of the type "x < y or x = y", which provides no evidence for positing an exclusive interpretation for or . We found that it takes some work to find examples where an instance of or all by itself might be held responsible for a perceived exclusive interpretation of a disjunction. A variation of an example by Mandy Simons' 1998 dissertation (p. 103; big file) and ascribed to Barbara Partee seems to have the right properties: "Jane is working or she is in the library". Suppose this sentence is uttered in a context where it's clear that Jane is a UMass student and the library we are talking about is the Dubois Library. We perceive an exclusive interpretation in this example, in spite of the fact that it is rather likely that a UMass student would be working in the UMass library. There is something about or itself, then, that can trigger an exclusive interpretation under circumstances that still have to be explored. Here are the slides with the examples that formed the background for our discussion.

pragmatic strengthening via scalar implicatures

On Wednesday we looked at attempts to derive the exclusive interpretation of or as a scalar implicature. We set aside orthodox Gricean accounts from the very beginning, and started our discussion with neo-Gricean accounts that rely on competitor sets generated by Horn scales. What the right competitor set is is fairly obvious for disjunctions with just two disjuncts, but once we look at disjunctions with more disjuncts, we encounter a difficulty that requires attention. Look at Luis Alonso-Ovalle's squib to appreciate the difficulty and explore a possible solution. In class, we only talked about disjunctions with two disjuncts and experienced a slight disappointment that you had already been prepared for by the Gamut chapter: Standard Gricean reasoning does not deliver the exclusive interpretation of disjunctions. All we get are inferences to the effect that the speaker did not have sufficient evidence for the claim that both disjuncts are true. Should we leave it at that? Gennaro Chierchia says "no" and proposes that the human language faculty includes a completely compositional and purely mechanical device for computing strengthened meanings via comparison sets. We can think of this device as the outcome of hard-wiring Gricean principles in the course of evolution. We do no longer have to worry about the exact Gricean or neo-Gricean steps in deriving scalar implicatures, then. An extended argument against a fully grammaticized derivation of scalar implicatures is Benjamin Russell's "Against Grammatical Computation of Scalar Implicatures." See also Levinson's book and the Grice biography on the bookshelf.Here is a slightly augmented version of the slides we looked at.

technical tutorial

On Friday, we had a tutorial on two technical issues. The first had to do with the formalization of sentences like If a woman is in Amherst, she is not in Boston. We observed that we couldn't formalize this sentence as ∃x [ [woman(x) & in-Amherst(x)] → ¬ in-Boston(x) ]. Under this formalization, the mere existence of something that is not a woman would make the sentence true. We get the correct formalization by replacing the existential quantifier with the corresponding universal one. This means that the formal language we are using seems to force us to posit an ambiguity for English indefinite DPs like a woman. Cases like these have pushed linguists to look for a unified interpretation of indefinites. Two possible ways of achieving such a unified interpretation of indefinites are proposed in Irene Heim's 1982 UMass dissertation.

The second technical issue we discussed had to do with the perceived quantifier ambiguity of sentences like Every child saw a movie. Suppose we formalized the two perceived readings as in (a) and (b).

a. ∀x [child(x) → ∃y [movie(y) & saw(x)(y) ]]

b. ∃y [movie(y) & ∀x [child(x) → saw(x)(y) ]]

The problem is that (b) logically implies (a). This means that the possible situations described by (b) are a subset of the possible situations described by (a), and consequently, we do not have a case for ambiguity here. Situations where all the children saw the same movie are already covered by formalization (a). The alleged two readings are not logically independent.