Abstract: Introduction. Readers continually integrate incoming information with evolving discourse representation during reading to maintain a fully updated situation model (O’ Brien et al., 1998). There are two different views about the process of updating -- the memory-based text processing view and the constructionist theory. According to the memory-based text processing view, every new piece of linguistic information is mapped into the information stored in working memory. Resonance of the ideas of the text is sufficient and necessary to make comprehension possible (McKoon et al., 1996; Albrecht & O’ Brien, 1995). On the other hand, according to the constructionist theory, readers pursue coherent relations throughout the text and attempt to explain why the actions, events, and states are mentioned in the text (Trabasso et al., 1989; Graesser et al., 1994). That is, related information in long-term memory is activated without the resonant process. Recently, there is convergence between the two views (Cook & Gueraud, 2005; van den Broek & Rapp, 2005). Within the constructionist theory, however, there are disagreements concerning what kind of goal information is to be reactivated. Some research has shown that unachieved goals are more available than achieved goals (Suh & Trabasso, 1993; Lutz & Radvansky, 1997; Magliano & Radvansky, 2001), but other research suggests the contrary (Richards & Singer, 2001; Singer & Richards, 2005). In the present study two kinds of goal-information integration were proposed, namely the reinstatement integration and the facilitation integration. When reading texts with goals containing an inclusive relation, goal information in long-term memory becomes reactivated by the achievement of sub-goals. On the other hand, when reading texts with goals containing a parallel relation, the failed goals facilitate the integration of goal information. Richards & Singer (2001) suggested that the signal of goals could activate the goal information in long-term memory, thus demonstrating reinstatement-integration. However, they did not explore whether the activation is primed by the signal or just represents spontaneous processing. The present study was to investigate this critical issue.
Method. A moving-window display technique was used. Ninety-four participants read 18 paragraphs, in which two characters attempted to accomplish two independent sub-goals to achieve the joint main goal. A 3 (success vs. failure vs. control) x 3 (strong signals vs. weak signals vs. control) experimental design was adopted.
Results. Repeated-measures ANOVA was performed to compare the recognition time of the probe words. The results indicated that there was a significant main effect of the success variable: success condition < failure condition < control. There was also a significant main effect of the magnitude of the signals: strong signals < weak signals < control. A significant interaction between the two independent variables was found: For the success condition, recognition time for strong signals was equal to weak signals, but lower than the control signals. For the failure condition, recognition time for strong signals was lower than weak signals, which was equal to the control signals. For the control condition, recognition time was the same for the three kinds of signals.
Conclusion. The present study shows the following: (1) The reactivation and integration of goal information when comprehending text with an inclusive relation of goals belong to reinstatement integration. (2) The signal of goals activates the achieved goal in long-term memory. (3) The activation is not a spontaneous processing, but is primed by the closing signal. (4) If the signal is strong, then the failed goal in long-term memory will also be reactivated

Key words: the inclusive relation of goals, constructionist theory, reinstatement-integration of goal