Processing bound variable “singular they”
Date: Friday March 1, 2019
Time: 11:30 am to 1:00 pm
Location: HH 305
Bio: Keir Moulton is a theoretical and experimental linguist seeking to understand the relation between structure and meaning in natural language. His research connects generative syntax and formal compositional semantics on topics such as clausal complementation and propositional attitudes, relativization, quantification, binding and ellipsis. He conducts experiments on these and other topics at the syntax-semantics interface in the XSyn Lab with Chung-hye Han. Current projects:
Dr. Moulton earned his PhD from the linguistics department at the University of Massachusetts, Amherst, in 2009.
Abstract: Joint work with Chung-hye Han (Simon Fraser University) The experimental, historical and theoretical literatures have identified a range of properties that allow a morpho-syntactically singular noun phrase to antecede “they”. A recurrent intuition expressed in much of this literature, albeit one never tested, is that bound variable “singular they” (ST) as in (1a) is more acceptable than a referential use (1b) (see Bjorkman 2017 and references therein).
(1) a. Every woman said they were the fastest runner.
b. ?The woman said they were the fastest runner.
In the psycholinguistics literature, the focus has been trained on whether ST imposes processing costs when it resolves to a singular antecedent; here the results are mixed (Foertsch and Gernsbacher 1997, Sanford and Filik 2007). It has additionally been found that ST with gender-unknown or gender-irrelevant antecedents (epicene uses) show higher acceptability and faster processing times (Doherty and Conklin 2017, Ackerman 2018).
We show through offline acceptability judgments that bound ST is more acceptable than referential ST, compared to gendered singular alternatives. This preference persists even if the antecedent is gender-known. The online processing record (the results of two self-paced experiments) track the acceptability judgments to some extent: bound variable ST is read as fast or faster than its gendered alternatives—but only when the antecedent is gender-known. Additionally, in line with what Sanford and Filik 2007 found for referential pronouns, we found a slower reading times in the spill-over region for ST across all antecedent types, including quantificational ones.
We take general processing cost of ST to reflect the fact that “they” is underspecified for number (Sauerland et. al 2005, contra Bjorkman 2017 and others) and the slowdown reflects a calculation regarding the fact that a stronger form (“she/he”) was not used (following Maximize Presupposition, Heim 1991). We then propose that the advantage for bound variable ST follows on a Minimal Pronoun approach (Kratzer 2009), wherein underspecification at LF calls off Maximize Presupposition and so does not trigger the epicene implicature. The analysis requires a more nuanced view of the forms that enter into competition for the Gricean reasoning that governs the use of ST.