Underspecification in anaphoric reference to structured entities

There is evidence that discourse representations may vary in detail, and may sometimes be underspecified (e.g., Ferreira, Ferraro, & Bailey, 2002; Sanford & Sturt, 2002). Poesio, Reyle, and Stevenson (2003) proposed that underspecification in anaphoric reference may occur if (a) the potential antecedents are part of a single mereological structure, and (b) the structure makes the interpretations equivalent. For example, in (1), if the potential antecedents (engine and boxcar) are 'hooked up' to form a single structure (train), then using 'it' to refer to any of these entities (engine, boxcar, or train) is equivalent, as hooking them up means that they are all going to the central station.

We investigated this proposal by assessing disruptions in processing using eye-tracking. Sentences (e.g., 1) were either mereology constructing (hooked up) or neutral (saw), and anaphoric reference was singular (it) or plural (them). First-pass regressions from Region 4, and regression path times for Region 6 suggested that singular reference was easier for mereology constructing than neutral sentences. This is consistent with singular reference being underspecified for mereology constructing sentences, but there being a conjunction cost (a processing cost found when two potential antecedents of a singular pronoun occur in a coordinated NP) for neutral sentences. Plural reference was equally difficult to process for mereology constructing and neutral sentences. Results were supported by findings from a text change detection study using materials such as (1) in which 'them', in the first display of the text, was changed to 'it' in the second display. Participants noticed significantly more changes in neutral than mereology constructing sentences. Results from both studies offer support for Poesio et al.'s hypothesis that anaphoric reference can be underspecified when potential antecedents are part of a single mereological structure and the structure makes the interpretations equivalent. Implications for how structured objects may be mentally represented will be discussed.

References

Ferreira, F., Ferraro, V., & Bailey, K. G. D. (2002). Good-enough representations in language comprehension. Current Directions in Psychological Science, 11, 11-15.

Poesio, M., Reyle, U., & Stevenson (2003). Underspecification in reference. In H. Bunt and R. Muskens (eds.), Computing Meaning 3, Kluwer.

Sanford, A. J. & Sturt, P. (2002). Depth of processing in language comprehension: not noticing the evidence. Trends in Cognitive Sciences, 6(9), 382-386.