The situation model is a hot topic in current narrative comprehension research. The Event-indexing model proposed by Zwaan, Langson, and Graesser (1995) suggests that readers establish a mental representation of events by tracking them through five dimensions: time, space, characters, causality, and protagonist/object. A large number of previous studies have shown that the temporal dimension plays an important role in constructing the situation model. The Scenario Account (Anderson, 1983) argues that scene provides clues for temporal shifts, but the Strong Iconicity Assumption (Zwaan, 1996) argues that readers update the situation model as soon as temporal shifts appear. In this study, we designed two experiments to resolve the disagreement between the Scenario Account and the Strong Iconicity Assumption. We assume that the Scenario Model and the Strong Iconicity Assumption do not contradict each other due to how the updating of a situation model has a variable processing mode in different stages of memory processing. We designed two experiments to test this hypothesis: Experiment 1 examined the effects of temporal shifts on the updating of the situation model in short-term working memory, and Experiment 2 examined this effect in long-term working memory.
In this study, a moving-window technique was used to explore the extent to which temporal shifts (long/short) affect updating of readers’ situation model. Experiment 1a and 1b examined whether long temporal shifts or short temporal shifts affected updating of readers’ situation model in short-term working memory. A single factor within-subjects design (time shift of a moment after or a day later) was used. We predicted the long temporal shifts (Experiment 1a) would not result in the updating of readers’ situation model due to the time limitation and difficulties of processing in short-term memory, but that short temporal shifts (Experiment 1b) would. Experiment 2 further examined the extent to which long temporal shifts affected updating of the situation model in long-term working memory. We predicted that long temporal shifts would cause the updating of the situation model because there was sufficient time for processing and increased memory capacities associated with long-term working memory.
The results confirmed our predictions that latencies of retrieval of removing entities showed no significant differences as a function of long temporal shifts and non-temporal shifts, which suggested that long temporal shifts did not result in situation model updating in Experiment 1a. However, Experiment 1b revealed the latencies of retrieval of removing entities were much slower in the short temporal shifts condition than those in the non-temporal shifts condition. In addition, when the filler sentences were increased in Experiment 2, we also found slower latencies of retrieval of removing entities in the long temporal shifts, which suggested the situation model had been updated both in Experiment 1b and Experiment 2.
In sum, temporal shifts play an important role in the updating of readers’ situation model. The findings showed that longer the temporal shifts were associated with greater difficulty to update the situation model. Therefore, only the short but not long temporal shifts condition resulted in situation model updating in short-term memory, but when information was stored in long-term memory, updating was possible in the long temporal shifts condition. The results collectively demonstrate that temporal situation model updating is dynamic.