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 6<br />

different from example (21) where the brakes <strong>in</strong> some way are predictable from the<br />

car.<br />

(22) I walked <strong>in</strong>to the room. The chandeliers sparkled brightly.<br />

I am not sure if examples like (22) should be subsumed under the notion of<br />

bridg<strong>in</strong>g at all. However, the relationship room-chandelier does fulfill the three<br />

requirements for the subgroup of bridg<strong>in</strong>g NPs I identified. This is unfortunate,<br />

because it shows that these criteria are not precise enough to make a dist<strong>in</strong>ction<br />

between examples that seem to be more new than bridged. Inducible parts are less<br />

related to the context than e.g. necessary parts and bear a strong resemblance to the<br />

examples found <strong>in</strong> the group that annotators generally identify as new(N).<br />

However, the characterization of bridg<strong>in</strong>g as licensed by the context <strong>in</strong> which<br />

it appears, and not by a unique discourse <strong>in</strong>dividual does help dist<strong>in</strong>guish this<br />

example from other more prototypical examples of bridg<strong>in</strong>g. I would argue that we<br />

cannot consider the chandelier to be licensed by the context because this context is<br />

too poor. A richer context would probably add <strong>in</strong>formation about the room<br />

mak<strong>in</strong>g it more likely and more clear that it is the type of room that is likely to have<br />

a chandelier.<br />

Perhaps the number of anchors that can be identified for a particular bridg<strong>in</strong>g<br />

NP dist<strong>in</strong>guishes those that are bridg<strong>in</strong>g from those that are actually<br />

accommodated. Those bridg<strong>in</strong>g NPs that are only loosely related to their context<br />

may be those perceived as new rather than related. This is a question that future<br />

work should address.<br />

6.5 CONCLUSIONS<br />

Summariz<strong>in</strong>g, naturally produced examples of bridg<strong>in</strong>g NPs are often related to<br />

several anchors <strong>in</strong> the discourse context rather than to one unique anchor as is<br />

usually assumed. These def<strong>in</strong>ite NPs should more appropriately be considered to<br />

be licensed by the context <strong>in</strong> which they are used rather than by one unique<br />

<strong>in</strong>dividual. The understand<strong>in</strong>g of bridg<strong>in</strong>g as be<strong>in</strong>g related to only one <strong>in</strong>dividual <strong>in</strong><br />

the context is an effect caused by the practice of us<strong>in</strong>g examples with a poor or<br />

non-existent context to illustrate bridg<strong>in</strong>g. This new analysis accounts for the<br />

difficulty found <strong>in</strong> gett<strong>in</strong>g a high degree of agreement <strong>in</strong> def<strong>in</strong>ite NP annotation<br />

tasks.<br />

Additionally there are good reasons to reflect on what types of relationships<br />

should be considered bridg<strong>in</strong>g. The orig<strong>in</strong>al taxonomy seems to encompass an<br />

extremely heterogeneous group, many of which are currently treated by different<br />

methods. There seem to be three characteristics are associated with prototypical<br />

bridg<strong>in</strong>g examples: They are 1) related semantically to the context <strong>in</strong> which they are<br />

used, 2) require an <strong>in</strong>ference for their resolution and 3) are marked as<br />

presuppositional. On the basis of these characteristics some subgroups considered<br />

180

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

Saved successfully!

Ooh no, something went wrong!