How scope and binding depend on order of evaluation
Chris Barker (UCSD)
work with Chung-Chieh Shan (Harvard)
November 21, 2003, 3PM
Abstract:
We propose that three familiar semantic phenomena show sensitivity to the order in which expressions get evaluated. As a descriptive first approximation, we have:
1. Bias towards linear scope
(1) Everyone loves someone.
Quantifiers that precede tend to take wider scope.
2. Weak Crossover
(2) a. Everyone's mother loves him. "everyone" binds "him": ok
b. His mother loves everyone. "everyone" binds "his": bad
If a quantificational NP binds a pronoun, the quantifier must precede the pronoun.
3. Superiority
(3) a. Who (did Tom say) _ bought what? _ precedes "what": ok
b. *What did (Tom say) who bought _? "who" precedes _: bad
A fronted WH-phrase's trace must precede any additional WH-phrases.
Over the past thirty years, people have sought explanations for these phenomena in hierarchical dominance relations rather than in linear order; I will mention some representative accounts, some of their supporting arguments, and some potential counterexamples. The key to reconciling a hierarchical approach with the linear facts, we suggest, is to seek an explanation not in terms of linear order directly, but rather *evaluation order*.
In hopes of arriving at independent validity for our notion of evaluation order, we import unchanged a well-understood theoretical characterization of order-of-evaluation for computer programming languages. This enables us to provide a grammar fragment on which the facts above follow from (a suitably precise and explicit version of) the following assumption:
(4) By default, people process expressions from left to right.
Furthermore, if we reverse the rule in the fragment that embodies (4), all three phenomena reverse direction: the default bias is for inverse scope, quantifiers must follow the pronouns they bind, and additional wh-phrases must intervene between a fronted wh-phrase and its trace.
Two notes on the presentation: the talk is intended for a general linguistic audience, and will emphasize empirical patterns and theoretical motivations and strategies, not technical details. I will give a closely similar talk at Rutgers one week eariler (Friday, 14 November); my plan is to discuss superiority in more detail at Rutgers, and weak crossover more at NYU.
Last Modified: November 5, 2003