Effects of discourse status on comprehension and production

This thesis explores the effects of discourse context, specifically whether a referent is discourse-given or discourse-new, on the online processing of the dative alternation and the active/passive alternation. Previous research provides support that the discourse status of a referent plays a role in the ordering of constituents in dative alternating sentences (Arnold et al., 2000; Snyder, 2003), but has not addressed the issue from a comprehension standpoint. Data has generally supported a discourse model in which speakers put discourse-given referents earlier in sentences than discourse-new referents. A similar effect has been found for passive sentences as well---listeners find the passive easier to interpret when the logical object is more salient or more given than the logical subject (Olson & Filby, 1972; Gourley & Catlin, 1978). Two eye-tracking experiments and one production experiment were conducted, using the dative alternation (Exp. 1: comprehension, Exp. 2: production) and the active/passive alternation (Exp. 3: comprehension), with a subject population of normal adult English speakers. Results from Experiment 1 were unexpected; listeners hearing a double object sentence (e.g. Show the turkey the bag) with a discourse-given Recipient and a discourse-new Theme launched eye-movements to discourse-given Themes before recovering from the garden-path and locating the correct referent. The results of Experiment 2, a production version of Experiment 1, were more in line with given-then-new expectations, with speakers producing more double object sentences when the Theme was discourse-given and the Recipient was discourse-new and fewer double object sentences when the Theme was discourse-new and the Recipient was discourse-given, both when compared to a condition where both the Theme and Recipient were discourse-given. Experiment 3 results were obscured by a strong "reading" effect; participants surveyed the stimuli left-to-right regardless of experimental condition. Analysis of a portion of the data showed that listeners experienced a strong garden-path effect while hearing a passive sentence when discourse conditions favored an active sentence. Results from all three experiments are discussed in terms of discourse effects on processing.