Abstract: Temporal information plays a very important role in narrative comprehension, but how it is processed has remained unclear. For a long time, the Strong Iconicity Assumption and the Scenario Theory offered competing explanations. We argue that these two theories emphasize different stages of processing time shifts. The Strong Iconicity Assumption concerns the first stage, the updating stage, and the Scenario Theory concerns the last stage, the integrating stage.
We conducted three rating tests and three experiments to test our hypotheses. The first test was designed to explore whether the duration time was significantly different for the three levels of temporal continuity: a moment later, an hour later and a year later, and if the results were significant between any two of them.
The second rating test was used to ensure that the time continuities of a moment and an hour were not beyond the scenario but a year was. Results were consistent with our predictions, with the longest average scenario found to be 4.85 hours and the shortest one 38.72 minutes. In the third rating test, we further evaluated the syntactic and semantic acceptability of the consistency between temporal markers and critical events to ensure that the results of the first rating tests reflected the processing of temporal information.. No significant differences were found in the three temporal conditions for syntactic acceptability, but significant differences were found for semantic acceptability. This indicated that results reflected the processing of the temporal information, which supported our predications.
In Experiments 1 and 2 we used a moving-window technique to verify the Strong Iconicity Assumption and the Scenario Theory, and to provide primary evidences for our hypothesis. Results showed that the response latencies of recognition words, reading times of the critical sentences, and temporal markers were consistent with the Strong Iconicity Assumption. As such, reading times of the critical events and the response latencies of the questions corresponded with the Scenario Theory. In Experiment 3 we used the eye-tracking technique to further investigate the processing mechanism of temporal information. An eye-tracking machine was used to record the eye fixations of the region of interest for the temporal markers and critical events. Reading time and number of the fixations were analyzed. Results showed significant differences for dwell time and total number of the temporal makers. For critical events, the difference of the regression path duration of the temporal makers did not reach statistical significance level but did in the critical events.
These results supported our hypothesis that the Scenario Theory corresponds to the later stage of processing temporal information. However, results the First Run Dwell Time tended to be consistent with the Strong Iconicity Assumption.
In sum, this research addressed the long-lasting debate between the Strong Iconicity Assumption and the Scenario Theory, and further studied the mechanism of processing time shifts in narratives. The results are discussed in terms of a two-phase model.

Key words: the Strong Iconicity Assumption, the Scenario Theory, time shifts, situation model, two-phase model