26.01.2013 Views

Presuppositions in Spoken Discourse

Presuppositions in Spoken Discourse

Presuppositions in Spoken Discourse

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 4<br />

Here I will first give some background <strong>in</strong>formation on global discourse<br />

structure and how it is meant to affect the availability of antecedents for anaphoric<br />

expressions. Two well-known proposals use tree structures to represent discourse<br />

structure, Polanyi’s (1988) L<strong>in</strong>guistic <strong>Discourse</strong> Model and Webber’s model (1991).<br />

Lascarides & Asher’s (1993) model also uses tree structures to organize rhetorical<br />

relationships. The attentional model of Grosz & Sidner (1986) represents<br />

hierarchical structure as stacks of focus spaces.<br />

One of the types of examples that have been used to support the need for a<br />

hierarchical discourse structure are so-called ‘return pops.’ However, Walker (1996)<br />

argues that the evidence from return pops can be expla<strong>in</strong>ed by other factors<br />

without the need to refer to a hierarchical discourse structure. I’ll present an<br />

example of a return pop with an anaphoric expression and then one from my own<br />

data with a bound presupposition. Then I’ll discuss whether or not we need to<br />

hypothesize a hierarchical discourse structure to expla<strong>in</strong> these examples. The<br />

answer seems to be no, and I’ll discuss the issue <strong>in</strong> relation to the rest of the corpus<br />

data.<br />

In a global discourse structure, <strong>in</strong>formation is organized <strong>in</strong>to units, called<br />

discourse segments, which are related to each other by coherence or rhetorical<br />

relationships which may be coord<strong>in</strong>at<strong>in</strong>g or subord<strong>in</strong>at<strong>in</strong>g, creat<strong>in</strong>g a hierarchical<br />

structure. This structure then supports the use of underspecified forms efficiently<br />

because their resolution can be guided by it. The ma<strong>in</strong> dist<strong>in</strong>ction <strong>in</strong> discourse<br />

structure is between what is <strong>in</strong> focus, attentionally available or related to the current<br />

topic from what is not, as determ<strong>in</strong>ed by identify<strong>in</strong>g the relevant “focus space”<br />

(Grosz & Sidner, 1986), “context” (Fox 1987) or part of the hierarchical structure<br />

(usually the right frontier, <strong>in</strong> e.g. Polanyi 1988, Webber 1991, and Lascarides &<br />

Asher 1993).<br />

The search for antecedents of anaphoric expressions will be most efficient if<br />

it beg<strong>in</strong>s by exam<strong>in</strong><strong>in</strong>g salient or activated referents, and these should be the<br />

referents found <strong>in</strong> a pre-def<strong>in</strong>ed discourse structural position. The relationship that<br />

holds between discourse structure and an anaphor may differ depend<strong>in</strong>g on the<br />

type of anaphor. Generally, abstract object anaphoric reference has been argued to<br />

be more dependent on discourse structure than anaphoric reference to concrete<br />

<strong>in</strong>dividuals <strong>in</strong> the discourse, perhaps <strong>in</strong> part because abstract object anaphors can<br />

sometimes refer to the <strong>in</strong>formation derived from one particular discourse segment.<br />

This is discussed <strong>in</strong> Webber (1991), who has argued that some abstract object<br />

anaphors can only access discourse segments on the right frontier, because those<br />

are the ones that are <strong>in</strong> focus or salient.<br />

In one of the first computational models of discourse structure, presented <strong>in</strong><br />

Grosz & Sidner (1986), contexts represent<strong>in</strong>g topics with discourse referents and<br />

other <strong>in</strong>formation are stored <strong>in</strong> so-called focus spaces, which are “pushed” onto a<br />

stack when <strong>in</strong>troduced. When a subtopic is completed, it is “popped” from the<br />

stack <strong>in</strong> order to get to the previous focus space and return attention to the earlier<br />

(superord<strong>in</strong>ate) discourse topic. The search for antecedents of anaphora <strong>in</strong> this<br />

model is restricted to the focus space on top of the stack. If there is no compatible<br />

90

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!