Séminaire de Recherche en Linguistique

Title:Effect of monotonicity on pronoun interpretation in discourse anaphora

(joint work with Benjamin Spector and Nina Gregorio)

Abstract:

Pronouns may co-vary with indefinites that do not c-command them. The most well-studied example is the donkey sentence, provided in (1) with its different declensions.

It is difficult to explain why such sentences are felicitous in the first place, since standard theories of compositional semantics only expect that a pronoun may co-vary with a quantifier if the latter c-commands the former. Even harder is the question of deriving the correct truth-conditions for these sentences. This issue is often explicitly set aside in many contemporary works.

The difficulty concerns scenarios where one of the farmers owns multiple donkeys. In experimental settings (Foppolo 2008), naive speakers judging (1 a) seem uncertain in such scenarios: participants sometimes just require at least one donkey to be cherished in (existential reading, (2 a)) ; sometimes they require all to be (universal reading in (2 b)). On the other hand, only existential readings are detected in (1 b-c). Since (Kanazawa 1994), the presence or absence of these readings has been linked to the monotonicity properties of the quantifier at play.

In this work, we provide experimental evidence that Kanazawa’s monotonicity generalization extends beyond donkey sentences to cases of discourse anaphora in (5): mirroring the facts for some and every respectively, we find that (5 a) is unambiguously understood as existential, while (5 b) is ambiguous. Because of the literature’s focus on donkey sentences, it has proved tempting to attribute any ambiguity in (1) to the quantifier (every/some/no) itself. Our results show that this strategy is too narrow and will miss the parallels with non-quantified cases in (5).

Our second main finding is that the addition of negation in (5 a), as in (6), reveals an ambiguity that is not detected in the original sentence. While this result suggests a role for monotonicity, they are not compatible with Kanazawa’s generalization as stated. Furthermore, they challenge the received wisdom that sentences like (5 a) must be equivalent to their wide-scope counterparts as in (7), a central assumption in Dynamic Semantics.