Aaron White (Rochester) will give our first General Linguistics Seminar talk in the new Schwarzman Centre. 

Characterising uncertainty in lexically triggered inferences

Monday October 13, 2025, 5:15pm. 

Linguistics Large Seminar Room, Schwarzman Centre 30.445.

ABSTRACT: We don’t always know what people we are speaking to intend to say. This uncertainty can have a wide variety of sources. In this talk, I explore inferential uncertainty associated with open class lexical items, focusing in particular on inferences thought to be associated with expressions’ preconditions of use. Classical theories attribute some such inferences to hard contextual constraints imposed by certain lexical items. However, recent studies observe substantial gradience in these inferences, leading some to challenge the classical view and argue that predicates impose no strict contextual constraints.

I argue this dismissal is premature. First, analysis of the distribution of inference judgments collected in formal experiments reveals patterns consistent with discrete inferencing behavior predicted by classical accounts. Second, clustering models applied to large-scale experimental datasets aimed at measuring lexically triggered inferences suggest that gradience arises not from a lack of constraints but from uncertainty about the semantics of specific expressions. These models furthermore reveal classifications aligning closely with classical predicate distinctions.

I argue that these results support a refined classical account where predicates impose contextual constraints, but perceived gradience reflects semantic uncertainty rather than the absence of hard constraints.