Text

theses

ETD doctoral

Different questions appear to call for different kinds of answers. We can refer to these readings as Mention-Some (MS) and Mention-All (MA), based on their level of exhaustivity (Hintikka 1976, 1978; Karttunen 1977; Asher & Lascarides 1998). For example, (1) is said to require exhaustivity, where Dana knows all of the relevant party-goers. (2) permits non-exhaustivity, where Dana knows some relevant place to find coffee. (3) appear to require non-exhaustivity, where Dana knows at least one way to get to Central Park.

(1) Dana knows who came to the party. hfill #MS/MA
(2) Dana knows where we can find coffee. hfill MS/MA
(3) Dana knows how we can get to Central Park. hfill MS/?MA

MS readings seem to be more tightly constrained than MA readings. However, it has been an open question precisely why this is case. Across the literature, two main hypotheses have emerged. Hypothesis 1: linguistic form constrains MS availability. Three main linguistic form factors have been pinpointed. Ginzburg (1995) and Asher & Lascarides (1998) noted that who-questions favor MA, while others (why, how, and where-questions) favor MS. George (2011), following Heim (1994) argued that the matrix verb know selects for MA. Finally, a number of researchers have pointed out that questions with existential modals/non-finite clauses permit MS (Bhatt 1999; George 2011, Ch 6; Fox 2014; Nicolae 2014; Dayal 2016; Xiang 2016). Hypothesis 2: contextual goals license MS (Groenendijk & Stokhoff 1982, 1984; Ginzburg 1995; Asher & Lascarides 1998; Beck & Rullmann 1999; van Rooij 2003, 2004; George 2011, Ch.2).

Theoretical proposals have taken two different approaches to these observations about MS/MA availability. One strategy posits underlying question ambiguity, housing the variability in the semantics (Beck & Rullmann 1999; George 2011; Nicolae 2014; Fox 2014; Xiang 2016). The second strategy posits a unique semantic representation that is either MA by default (Groenendijk & Stokhof 1982, 1984; Karttunen 1977), MS (Asher & Lascarides 1998; Schulz & van Rooij 2006; Spector 2007; Zimmermann 2010), or semantically underspecified for either (Ginzburg 1995; van Rooij 2003, 2004). No matter which underlying semantics, context then allows for the hearer to resolve (non-)exhaustivity.

This dissertation tests these two hypotheses concerning the sets of factors licensing MS and MA readings, and thereby weighs in on the theoretical debate concerning the baseline representation of question semantics and the role of pragmatics. I provide quantitative empirical evidence that addresses the role of the linguistic factors, but demonstrate that contextual goals can indeed override those interpretational defaults. Furthermore, I demonstrate that not only MS, but MA readings, too, are subject to contextual constraints (see Schulz & van Rooij 2006; Spector 2007; Zimmermann 2010). I argue that baseline interpretations do not reveal underlying semantics, but rather reflect hearer expectations about why a speaker would utter a given question, given that it surface-underspecifies meaning. Under this view, linguistic factors are defeasible cues to speaker goals, which direct the resolution of (non-)exhaustivity: when the context is informative with respect to discourse goals, linguistic factors are neutralized, both in interpretation and production. This finding resonates with a line of psycholinguistic research on communication and audience design (e.g., Brennan & Clark 1996; Clark & Wilkes-Gibbs 1986; Keyser et al 2000; Ferreira & Dell 2000; Ferreira 2019).

Finally, I show that hearer-specific properties drive (non-)exhaustivity resolution in questions, depending on the extent to which a hearer is more 'literal' or more 'pragmatic'. This finding helps us work toward a novel computational model of question-answer dynamics that incorporates aspects of the question, the questioner, and the hearer.

ETD_11265

Ph.D.

Includes bibliographical references

doi:10.7282/t3-bgfk-jt70

The author owns the copyright to this work.