Forests and Trees: the Formal Semantics of Collective Categorization (ROCKY)


8 February 2018
15:30 - 17:00
Trans 10, room 0.19 (A.W. de Grootkamer)

Lelia Glass: An empirical investigation of distributivity, lexical semantics, and world knowledge

We are happy to announce that on Thursday, February 8th, Lelia Glass (Stanford University) will give a LUSH talk in Utrecht. We hope to see you all there!

Date: Thursday, February 8th, 2018
Time: 15:30 – 17:00
Location: Utrecht, Trans 10, room 0.19 (A.W. de Grootkamer)
SpeakerLelia Glass
Title: An empirical investigation of distributivity, lexical semantics, and world knowledge

A predicate is understood distributively (1) if it is inferred to be individually true of each member of a plural subject;nondistributively if not (2). Some predicates can be understood in both ways (3) [Scha 1981, Link 1983, Roberts 1987…]

  1. Alice and Bob smiled …. CONVEYS that they each smiled [distributive]
  2. Alice and Bob met …. DOES NOT CONVEY that they each met [nondistributive]
  3. Alice and Bob lifted the table …. COULD CONVEY they each lifted it [distributive]; COULD CONVEY they did so by working together [nondistributive]

In this talk, I address the open question: Which predicates are understood in which ways, and why? Which ones act like smilemeet, or lift the table? Of course, a predicate’s potential for distributivity depends on what we know about the event it describes: people have their own faces, so can only smileindividually; but it is left open which other predicates behave in which ways.

To make progress on this question, I motivate and test two large-scale patterns in the distributivity potential of Verb Phrases:

  • Causatives (open the door) can be understood nondistributively, because as a general fact about causation, multiple individuals’ contributions might be jointly sufficient, individually insufficient to bring about a result.
  • Predicates with objects construed as incremental (eat the pizza; Tenny 1987, Krifka 1989) can be understood nondistributively, because multiple individuals might each affect different parts of the object (eat different parts of the pizza), only jointly affecting (eating) the whole thing.

I find evidence consistent with these predictions in a large-scale dataset of quantitative ratings for the distributivity potential of over 2600 Verb Phrases, supported by corroborating experiments.

When the pragmatics of distributivity is concretely specified in this way, I suggest that it can take on some of the explanatory burden that other theories attribute to the semantics, shifting the division of semantics and pragmatics in the analysis of distributivity.