The effect of narrative time shifts on the representation of goal-related information William H. Levine, Joel A. Hagaman, Cari Anne Bogulski, Rebecca R. Green, & Dorthie S. Ortigo University of Arkansas Background Most theories of language comprehension assume that readers construct a mental representation of the situation described by a narrative. This representation has been labeled a situation model (van Dijk & Kintsch, 1983). Most investigations of situation model construction have been directed at understanding how readers comprehend information about single dimensions, such as space (e.g., Morrow, Greenspan, & Bower, 1987) and time (e.g., Zwaan, 1996). However, despite that naturalistic narratives are typically complex and multidimensional, there has been relatively little attention directed toward comprehension of narratives in which multiple dimensions have been experimentally manipulated. The experiments presented here examined comprehension of narratives in which multiple dimensions were manipulated. According to the event-indexing model (Zwaan, Langston, & Graesser, 1995), readers monitor the dimensions of a narrative that may be important. These dimensions include time, location, causation, protagonists, and the protagonists’ intentions as they seek to satisfy their goals. On each processing cycle, the situation model is updated to reflect new information. New information is incorporated into the existing model to the extent that it shares dimensional indices. When new information does not share indices with the existing model, a different type of updating occurs. The new information is still incorporated into the model, but new indices are created in the model to reflect the situational discontinuity. For example, after processing the sentence, John woke up in his motel room and then went to Denny's for breakfast, the situation model will contain separate nodes for the events waking up and eating breakfast, and each node will be marked with an index specifying the locations of the events. Links between the nodes will exist for those indices that are shared across the locations (i.e., there is a shared protagonist index), to represent the continuity in the situation. As narratives get more complex, readers may have to choose which dimensions to monitor for updating because of limited working memory capacity. The choice of which dimensions to monitor may be strategic based in part on the goals of the reader, implicit based on general biases to consider some dimensions as more important than others, or relatively automatic because basic memory processes have made information about some dimensions easily accessible. Those dimensions that are monitored are said to be foregrounded. Foregrounded information is more accessible in memory, and thus information related to foregrounded dimensions is more easily updated. The kinds of information that are foregrounded is a matter of ongoing debate, and the current research was designed to examine how changes in one dimension (i.e., time) affect the foregrounding of information in another dimension (i.e., goals/intentions). Two hypotheses were considered. The situational presence hypothesis states that the presence of an entity, action, or event in the "narrative now" dictates whether it is foregrounded (cf. Zwaan, Madden, & Whitten, 2000). Thus, a long time shift should reduce the foregrounding of a satisfied goal, as it is no longer “present” after the shift. For an unsatisfied goal, time shifts should have no effect. By contrast, the goal urgency hypothesis states that when goals remain unsatisfied, as narrative time passes, readers may focus an increased amount of attention on them, due perhaps to a desire to see if protagonists will achieve their goals (cf. Rapp & Gerrig, 2002). Experiment 1 Participants read short narratives in which protagonists were trying to achieve relatively urgent goals. Critically, the goals were such that if they were not satisfied within a short period of time, it was still possible to satisfy them. Equally important, the goals "expired" in about a day. Following manipulation of goal satisfaction, a time shift manipulation was implemented. The long time shifts moved the narrative close to the expiration time. The accessibility of information about the goals was measured by a recognition probe. Sample narrative for Experiment 1 Don was hosting a dinner party in a few days. His apartment was a mess and he wanted to clean up. He didn't know if he'd get to it, though, because of all the other preparations that had to be made. On the day before the party, he spent a lot of time preparing food before he started cleaning, and he didn't get to clean up the mess. (unsatisfied goal) and he also got to clean up the mess. (satisfied goal) A few minutes later, Don picked up the phone. (short time shift) The next morning, Don picked up the phone. (long time shift) PROBE WORD: MESS The short conclusion for each story differed depending on the status of the goal, and most stories were followed by a comprehension question. Fillers passages were interspersed among experimental passages. Experiment 1 Results 950 900 short RT long 850 800 unsatisfied satisfied goal status For unsatisfied goals, reaction times were significantly shorter in the long time shift condition. For satisfied goals, reaction times were non-significantly shorter in the long time shift condition. The effect size for time shifts for unsatisfied goals was d = .40, whereas for satisfied goals it was only d = .11; however, the time shift- by-goal status interaction did not reach significance. These results are consistent with the goal urgency hypothesis. The pattern of results in the satisfied goal conditions is surprising, but fits the goal urgency hypothesis given the assumption that the goals remained relevant through the conclusion of the story, even after the primary action needed to satisfy the goal was completed. Experiment 2 Because of the surprising results of Experiment 1, Experiment 2 was a replication and refinement. Here are the critical changes that were made: The timing of the goal expiration was made more explicit Prior to the time-shift sentence, narrative time was “marked” at a specific moment so that temporal adverbials like, a few minutes later, made better sense More explicit reasons for the goal not being satisfied were provided The second mention of each probe word was removed from each passage Sample narrative for Experiment 2 Don hung up the phone and ended his fourth call from his mother that day. He had just moved her into a nursing home and she was not happy to be there. In addition to moving his mother, visiting her, and the telephone calls, Don was hosting a dinner party on Sunday night. It was already Saturday morning and he still needed to clean his house. He didn't know if he'd get to it, though, because of all the other preparations that had to be made. He spent all day Saturday shopping, and he wasn't able to straighten up. (unsatisfied goal) but he was still able to straighten up. (satisfied goal) At 9 o'clock, Don sat down exhausted. (time “marked”) A few minutes later, Don was on the phone with his mother again. (short time shift) Early the next morning, Don was on the phone with his mother again. (long time shift) PROBE WORD: CLEAN Experiment 2 Results 1350 1300 short RT long 1250 1200 unsatisfied satisfied goal status Conclusions Time shifts increase the accessibility of unsatisfied goal information, and sometimes, unexpectedly, satisfied goal information. Satisfied goals in these experiments were more accessible than unsatisfied goals, although the reasons for this are not clear. One possibility is their relative urgency, another is their centrality to the stories. Theories of situation model updating need to take into account not only the continuity / discontinuity in situations, but also how situational changes influence the moment-to-moment importance of information. Goal (and causal) information appears to interact with changes in at least one other dimension, time.