Influence of the Historical Discourse Record on Language Processing in Dialogue

Abstract

Speakers typically design definite referring expressions to uniquely identify the intended referent with respect to the alternatives in the referential context, and addressees interpret these expressions with respect to the contextual alternatives. Although it is clear that the relevant context includes entities in the immediate context, less clear is how the historical discourse context affects language interpretation in the moment. This article presents the results of two experiments that examine interpretation of definite referring expressions in cases where the local context has recently changed. In Experiment 1 eye-tracked participants followed an experimenter's instructions to click on objects multiple times each. When expressions were over-modified for the immediate context (e.g., the striped shirt, when the shirt would suffice), referential interpretation was significantly facilitated when that expression had been previously used in a supporting context. By contrast, interpretation of nonmodified expressions that were appropriate for the local context (e.g., the shirt) was somewhat impaired when a modified term had been used previously. Experiment 2 replicated these findings with a more sensitive, cohort-competitor design. These findings demonstrate that both the local and the historical context influence the online interpretation of referring expressions.

Funding

This material is based on work supported by the National Science Foundation under grant no. NSF BCS 12-57029 to S. Brown-Schmidt.

Notes

¹ A value of .5 was added both the numerator and the denominator as the calculation is problematic when the numerator or denominator is zero; see Barr (2008) for a similar approach.

² Note that a slightly different calculation for the dependent measure was used in Experiment 1 (the elogit of target fixations vs. all other objects combined), because that experiment did not contain a temporary ambiguity between potential referents. The measure we use in Experiment 2—the ratio of target to competitor fixations—capitalizes on the cohort-competitor design by providing a more focused measure of the competition between the nonmodified noun and the modifier.