1 What happens when textual information previously processed appears again or is needed for comprehension? In other words, how information in long term memory is (re)activated for possible integration with information that is currently available in working memory? This issue has been explored in the framework of the memory-based text processing view (Gerrig & McKoon, 1998; Gerrig & O’Brien, 2005; McKoon, Gerrig, & Greene, 1996; McKoon & Ratcliff, 1995) and of the resonance model (Myers & O’Brien, 1998; O’Brien & Myers, 1999). This paper aims to provide a further test of the role of resonance process in the reactivation of goal related information during reading.
2 The memory-based view states that narrative comprehension is primarily influenced by memory retrieval processes that operate automatically. According to this position, concepts and propositions currently processed in working memory (WM) provide retrieval cues for concepts and propositions stored in long term memory (LTM). According to the resonance model (Myers & O’Brien, 1998; O’Brien & Myers, 1999), retrieval of information in LTM is supposed to be accomplished through a “fast-acting, passive resonance process” (Cook, Halleran, & O’Brien, 1998, pp. 110) This model is based on the assumption that concepts from the discourse representation and general world knowledge resonate as a function of the degree of overlap of semantic and contextual features among concepts. According to the resonance model, resonance is a necessary condition for those memory elements to become available. Thus, resonance helps the reader to build of a coherent discourse representation.
3 Three critical aspects of the resonance process can be extracted from the resonance model (Myers & O’Brien, 1998, O’Brien & Myers, 1999, see also Lorch, 1998). First, it is passive, automatic. In other words, the resonance process is rapid, irrepressibly triggered by contextual cues and out of reader’s control. Second, it is “dumb”. Information that resonates sufficiently is returned in the working memory, regardless of whether that information is necessary or relevant to the understanding of the currently processed sentence. In other words, the resonance process does not differentiate information that shares features and is relevant to the currently processed sentence from information that shares features but is not relevant. Finally, the resonance process is unrestricted. Any concept in LTM that shares features with the content of WM would resonate.
4O’Brien and Myers (1999) presented an overview of the empirical studies investigating the reactivation of concepts during reading. Among the variables governing the reactivation, were the degree of featural overlap and the degree of elaboration of the contextual cue in memory (Albrecht and Myers, 1995; 1998).
5 Within the memory-based text processing framework, featural overlap is considered as a major factor influencing the resonance process: without featural overlap, reactivation should not occur. Specifically, Albrecht and Myers (1995, 1998) have demonstrated that the degree of featural overlap can either increase or decrease the probability and the speed with a character’s goal resonates and is reactivated. In Albrecht and Myers (1995) studies, participants read texts in which a character’s goal was satisfied or unsatisfied. Goal information was then backgrounded by several sentences of filler text. The filler section ended with a context reinstatement sentence, which provided either a contextual overlap or no contextual overlap with the goal information from the backgrounded goal context. The overlap consisted in the repetition of a cue (e.g., leather chair) that had first been mentioned in the original goal context. This sentence was followed by two target sentences. When the goal was unsatisfied, these sentences provided an inconsistency. Results indicated that readers noticed inconsistencies only when the texts contained a contextual overlap (i.e., the repetition of a cue that had previously been mentioned in the original goal context) that reactivated the goal information. As a consequence, reading times on the two target sentences were longer in unsatisfied goal conditions than in satisfied goal conditions. However, the reactivation goal information do not systematically influence comprehension of the target sentence (Guéraud & Tapiero, 2001; Long & Chong, 2001)
6 In a set of experiments, Albrecht and Myers (1998) found that the more the contextual cue was elaborated in the goal context, the faster the unsatisfied goal was reactivated (for further evidence that the elaborated antecedent is more quickly retrieved than was unelaborated antecedent, see O’Brien, Albrecht, Hakala, & Rizzella, 1995). The degree of elaboration of the contextual cue depended on whether or not an adjectival modifier (e.g., “leather”) was associated to the cue (e.g., “chair”) in the goal section. Thus, contextual cue was considered as “elaborated” (e.g., “leather chair”) when it was associated to an adjectival modifier and “unelaborated” (e.g., “chair”) when presented alone. Each text included a contextual overlap that consisted in the repetition of an elaborated contextual cue (i.e., adjectival modifier + noun) or an unelaborated cue (i.e., noun) that was previously mentioned in the original goal context. Results demonstrated that readers detect inconsistencies in the elaborated and unelaborated contextual cue conditions. Reading times on target sentences were longer when the goal was unsatisfied than when it was satisfied. Nevertheless, when the contextual cue was unelaborated, this effect appeared only for the second target sentence. Albrecht and Myers (1998) explain this time increase by the reactivation of the unsatisfied goal. The reactivation would be triggered by the processing of contextual cues (elaborated or unelaborated) that would lead readers to detect the inconsistency. According to the results, an unelaborated contextual cue was sufficient to reactivate the unsatisfied goal, but the reactivation was slower. In sum, these findings put in light that resonance process is sensitive to both contextual overlap (for its triggering) and to the degree of elaboration (for its intensity and speed).
7The causal structure of narratives is another important factor in text comprehension. Particularly, it has been demonstrated that causal structure may influence the ease in which information can be accessed. For instance, causal structure affects the processing of causal relations and memory representation of these relations (Caillies, Denhière & Kintsch, 2002; Caillies, Denhière, & Jhean-Larose, 1999; Suh & Trabasso, 1993; Trabasso, van den Broek & Suh, 1989; Trabasso & van den Broek, 1985; van den Broek & Lorch, 1993). The representation of causal relations in narratives can be described either as a linear chain or as a network. The linear chain model assumes that readers attempt to establish causal connections only between events in adjacent sentences in the surface structure of the text (Black & Bower, 1980; Fletcher & Bloom, 1988). According to this model, only these relations would be included in readers’ memory representation of the text. In contrast, the causal network model assumes that multiple connections are possible to a single event. Readers will establish causal connections between events in nonadjacent as well as in adjacent sentences in the surface structure of the text. So, both relations would be represented. Moreover, within the causal network, the retrieval could be function of the presence or not of a direct causal relation between events in the memory representation. In other words, retrieval could be function of the strength of the relation linking events together (Lutz & Radvansky, 1997; Magliano & Radvansky, 2001; van den Broek & Lorch, 1993). As causal structure is known to affect the retrieval of the stored information from long term memory, one can assumed that causal structure, combined with contextual overlap, could facilitate or impede the resonance process.
8In two experiments, we explored the effects of a new contextual overlap manipulation on the resonance process. Specifically, we considered contextual overlap in association with the text’s temporo-causal structure. We constructed 8 narratives on the basis of the ones used by Albrecht and Myers (1995, 1998). First, an introduction presented the character and his/her goal. In the next section, information were given about satisfaction or non satisfaction of this goal. Then, took place a filler section ending with a reinstatement sentence that featured a contextual overlap. The narrative ended with two target sentences (that could be perceived as consistent or inconsistent, depending on the fact that the main goal was satisfied or not (i.e., inconsistency paradigm) and a small conclusion.
9 The main difference with Albrecht and Myers’ texts concerns the fact that we introduced a specific temporo-causal structure (i.e., a fixed goal-subgoals sequence) to subsequently set up a specific manipulation of the contextual overlap. Thus, after the presentation of the character’s goal, three successive subgoals were introduced. These subgoals were performed by the character in order to achieve his/her main goal. We manipulated the contextual overlap in regard to this specific temporo-causal structure. More precisely, the cue used to produce the overlap (i.e., the repetition of words presented earlier in the text) could either be associated with a subgoal close or distant from goal attainment in the causal structure (Note that the terms close and distant must be considered as a function of goal attainment and not of the respective position of the cues in the text).
11 In line with the memory-based processing view (Myers & O’Brien, 1998; O’Brien & Myers, 1999), we assumed an effect of the goal satisfaction: reading times on the target sentences should be longer when a goal had previously been unsatisfied than when it had been satisfied, reflecting the detection of inconsistency. In addition, we assumed a main effect of contextual overlap on target sentences reading times. If, as we hypothesised, the manipulation of contextual overlap in terms of origin of the cue in the temporo-causal structure, leads to a difference in terms of reactivation, reading times on the target sentences should be longer when the contextual cue appeared earlier in a subgoal close rather than distant of the goal attainment. Finally, as a close overlap is assumed to produce a greater goal reactivation, we supposed that the inconsistency effect should be greater in the close overlap condition than in the distant overlap condition.
12 Another purpose was to investigate the effect of the contextual overlap on the reading times of the context reinstatement sentence. Indeed, if the resonance process is automatic, goal information that was associated with the contextual cue in the goal context should immediately resonate in response to the processing of the reinstatement sentence. Accordingly, an effect of contextual overlap could also be obtained on the reinstatement sentences. Reading times on the reinstatement sentence could be shorter when it features a close rather than a distant overlap.
13 The aim of the experiment was to determine to what extent a specific contextual overlap, in association with text’s temporo-causal structure, could affect the resonance process. Exploration of the manipulation of this factor took place in an inconsistency detection paradigm. Texts were constructed in respect with the structure of narratives used by Albrecht and Myers (1998). Stories described a goal that a character wanted to complete and three successive actions corresponding to subgoals performed by the character in order to achieve his/her main goal. As in Albrecht and Myers (1998), target sentences varied in consistency with respect to the goal satisfaction. They were consistent when the character had achieved his/her main goal, but were inconsistent when the character had not satisfied this goal. The goal satisfaction was manipulated so that the main goal could either be satisfied (completed) or unsatisfied (not completed but still needing to be completed). We also manipulated the contextual overlap between the reinstatement sentence and a specific subgoal of the sequence of goal attainment. This factor was manipulated as a function of the distance between the subgoal from which was extracted the contextual cue and the statement of the goal attainment in the temporo-causal structure of the texts. Distance corresponds here to the temporal distance separating a subgoal from the goal attainment. So, the temporal subgoal 3 was close to goal attainment statement but distant from the sentence introducing the goal in the text surface structure. In contrast, the temporal subgoal 1 was distant from goal attainment but close to the goal statement. In line with Albrecht and Myers, our materials included a high contextual overlap, that is the repetition of a contextual cue elaborated earlier in the text. Contextual overlap was manipulated as a function of the “subgoal” from which was extracted the contextual cue used in the context-reinstatement sentence. In the close contextual overlap condition, the context reinstatement sentence introduced an overlap that consisted in the repetition of a contextual cue that previously appeared in a subgoal close to the goal attainment (i.e., Subgoal 3). The distant overlap condition introduced an overlap that consisted in the repetition of a cue that previously appeared in a subgoal distant to the goal attainment (i.e., Subgoal 1).
14 The resonance model proposes that when a contextual overlap occurs, goal information resonates and is reactivated. As the two versions of reinstatement sentence provide an overlap, the goal information should resonate and become accessible for processing during the reading of the first target sentence in both close and distant overlap conditions. In any case, the repetition of a contextual cue that previously appeared in a subgoal should reactivate this subgoal as well as the goal associated with the subgoal. In addition, we assume that the extent in which the goal will resonate will be a function of the goal-subgoals structure and of the strength of the connexions between the activated goal and the contextual cues within this structure. In sum, we assume that such a manipulation of the contextual overlap should have an influence on the extent to which the goal will resonate and become reactivated. Indeed, subgoal 3 – temporally close of goal attainment– should be more strongly connected with the sentence introducing the character’s goal than subgoal 1 – temporally distant of goal attainment. So, goal information should be more easily (re)activated by the processing of contextual cue extracted from subgoal 3 (close overlap) than from subgoal 1 (distant overlap). In both cases, the reactivation process triggered these contextual cues should lead the readers to notice the inconsistency in the target sentences, resulting in an additional processing of these sentences. Nevertheless, according to our hypothesis about contextual overlap, the slowdown in reading of the target sentences following an unsatisfied goal should be greater in the close overlap than in the distant overlap condition.
15 In addition, on the basis of the assumption of a “dumb” resonance process, we hypothesized that the goal should resonate and become reactivated in both satisfied and unsatisfied goal versions of the texts. Indeed, in presence of an overlap, the goal should resonate regardless of whether it has been satisfied or not. However, the satisfied goal (consistent with the target sentences) should be easily integrated with the information already activated in working memory, whereas the unsatisfied goal (inconsistent with the target sentences), should lead the readers to notice the inconsistency, resulting in comprehension difficulties and additional information processing.
16Participants were 44 students of University of Paris 10. Their participation was volunteer and they did not receive course credit for their involvement in the experiment. All of them were native French speakers.
17In respect with the structure of texts used in past investigations (Albrecht and Myers, 1995; 1998), eight narratives were constructed (A translation of one of the text is shown in Appendix). Each passage was composed of five sections: an introduction, a goal section, a filler section, two target sentences, and a conclusion. The introduction presented the main character and the setting. The sixth sentence of the introduction section introduced the goal the character wanted to complete (for example, Nicolas needed to finish to write up the last ten sheet today) and three successive actions corresponding to three subgoals performed by the character in order to try to achieve the desired goal (in the example, subgoal 1: Nicolas opened the blue file; subgoal 2: he underlined the main ideas and subgoal 3: he took index cards out of a metallic locker). This section was followed by the goal section. In this section, the first sentence described an event that occurred before or after the achievement of the main goal. This event reminded the character an alternative goal he/she also had to accomplish. At this point, the goal satisfaction was manipulated. In the unsatisfied goal version, this event occurred before the main goal was achieved and temporarily hindered the goal pursuit. The last sentence of the section stated that the character would resume the pursuit of his/her goal after the alternative goal will be completed. In the satisfied goal version, the event occurred after the main goal achievement, and so, did not interfere with the pursuit of the goal. The following section was a filler section that described actions and events related to the alternative goal pursuit and its achievement. The last sentence of this section, called context reinstatement sentence, introduced a contextual overlap that consisted in the repetition of an elaborated cue appeared earlier either in a subgoal close (i.e., subgoal 3: index cards) or distant (i.e., subgoal 1: blue file) of the main goal attainment. Note that the two context-reinstatement sentence versions made reference to a contextual cue that previously appeared in the text, but never made a direct reference neither to the subgoalnor to the goal. Two target sentences followed the reinstatement sentence. They described the character engaged in actions that were consistent with the satisfied goal but inconsistent with the unsatisfied goal. Finally, the text ended with a two-sentence conclusion.
18We used a 2 (Goal Satisfaction) x 2 (Contextual overlap) within-participants design. Goal satisfaction was combined with contextual overlap to produce four versions of each text: close overlap-satisfied goal, close overlap-unsatisfied goal, distant overlap-satisfied goal and distant overlap-unsatisfied goal. Reading times on the target sentences and the context reinstatement sentence were recorded as dependent measures. Each participant was randomly assigned to the four conditions with two constraints. First, each participant saw two passages in each condition; and second, across participants, each passage occurred in each condition an equal number of times. The order of the passages was the same for all participants.
19 Each participant individually completed a session that lasted approximately 30 min. All materials were presented on a computer screen. Participants were instructed to place their right thumb on the spacebar key, their right index finger on a “Yes” key, and their left index finger on a “No” key. Participants were instructed to read the texts at their own pace. Each passage began with the display of the word ATTENTION! in the middle of the screen. Participants were instructed to press the spacebar when they were ready to read the passage. Once started, each press on the spacebar erased the current line of text and presented the next line. Reading times were recorded for each sentence. After each text, participants answered to two affirmations. These affirmations required participants to assess whether the statement was true or false, based on what they had read in the text. After each response, participants were given a feedback on the validity of their answer. The purpose of this verification task was to check whether or not participants had carefully read the text. The experiment began with two practice texts, to make sure that participants understood well the procedure.
20We conducted three independent 2 (Goal Satisfaction) x 2 (Contextual Overlap) analyses of variance (ANOVAs): one on reading times for the first target sentence, one on reading time for second target sentence and the last one on reading times for the context reinstatement sentence. Outliers were identified using Tukey’s (1977) hinge criterion and replaced in each condition by the mean reading time. This substituted less than 5 % of the reading times. No reliable effects were found on the reading time for the second target sentence. As a consequence, only mean reading times for the target sentence 1 and the context reinstatement sentence are presented in Table 1.
21There was a significant main effect of contextual overlap on reading times, F(1, 43) = 8.61, MSE = 122.34, p < .006. Reading times on the first target sentence were longer in the close overlap condition (M = 4005 ms) than in the distant overlap condition (M = 3622 ms). The main effect of goal satisfaction was not significant, F(1, 43) = 0.28, MSE = 121.54. Planned comparisons also indicated that the difference between these goal conditions in the close overlap condition was not reliable. This was corroborated by an absence of interaction between Satisfaction and Contextual Overlap, F(1, 43) = 0.35, MSE = 151.47.
22We conducted a 2 x 2 analyse of variance on reading times for the reinstatement sentence. Goal Satisfaction (satisfied vs. unsatisfied) and Contextual Overlap (close vs. distant) were treated as within-subjects variables. The analysis yielded a main effect of contextual overlap on reading times, F(1, 43) = 24.42, MSE = 132.79, p < .0001. Context reinstatement sentences including a close overlap were read slower (M = 4059 ms) than sentences that included a distant overlap (M = 3447 ms). Again, reading times did not differ in regard of goal satisfaction F(1, 43) = 0.02, MSE = 129.15. The interaction was not reliable, F(1, 43) = 1.86, MSE = 155.58.
Table 1. Mean reading times (in milliseconds) and standard deviation (under brackets) for the first target sentence and the context reinstatement sentence as a function of goal satisfaction and contextual overlap.
23 Reading times observed on the first target sentence suggest that readers did not notice the inconsistency between the target sentence and the unsatisfied goal in both contextual overlap conditions. In our study, unlike Albrecht and Myers (1995, 1998) findings, goal satisfaction had no impact on the reading time of the first target sentence. Nevertheless, the fact that we did not obtain an inconsistency effect does not mean that the goal information has not been reactivated. Indeed, as demonstrated by Cook, Halleran, and O’Brien (1998) and Long and Chong (2001), reactivation of disconfirmed or irrelevant information do not necessarily influence comprehension of the target sentences. Still, this result is consistent with the description of resonance as an unrestricted, “dumb” process (Myers & O’Brien, 1998).
24 Results also indicated that the contextual overlap had an impact on reading times for both the first target sentence and the context-reinstatement sentence. Specifically, the close overlap condition resulted in longer reading times than in the distant overlap condition. Contrary to our hypothesis, the extent to which the goal resonates and becomes reactivated seems independent of the force of connection between the goal and the subgoals from which the contextual cue were extracted. So, a contextual cue associated with a subgoal close of the main goal attainment (subgoal 3) does not increase the speed of retrieval of the goal and even slows it down. We assumed that, in the situation model, the third subgoal (i.e., the closest of the goal attainment in the temporo-causal structure) was more strongly related with the initial goal information than the first subgoal. Nevertheless, it seems that participants, during the reading of the context reinstatement sentence providing a close overlap, did not establish a direct connection between the third subgoal and the initial goal information. On the contrary, rather than being direct this connection would be indirect (i.e., mediated by subgoal 2 and subgoal 1). Thus, the number of steps in the network to attain the initial goal information would be higher from the reactivation of the close subgoal (3 steps) than the distant one (1 step). Inversely, when the contextual cue appeared earlier in a subgoal distant of the goal attainment, a direct connection between the subgoal information, associated to the cue, and the goal information would have permit a very rapid access to the goal information. In sum, in the close subgoal condition (subgoal 3), goal information reactivation would involve a longer path (3 steps in the temporal goal subgoals structure). Conversely, in the distant subgoal condition, the connection would be direct between the distant subgoal (subgoal 1) and the goal. This explanation is consistent with Albrecht and Myers’ proposal that the reactivation of the goal is not direct, but is “mediated by relations established earlier in the text” (Albrecht and Myers, 1995, pp. 1463).
25 An alternative explanation for the slower reading times in the close overlap condition could be related to a difficulty to integrate the context reinstatement sentence to the immediately preceding filler sentence. If so, reading times variations could be explained by processes involved in achieving local coherence rather than by the resonance process. For instance, local coherence could be weaker between these two sentences in the close overlap condition than in the distant overlap condition. As a consequence, close overlap would lead to longer reading times on the reinstatement sentence. In the experiment 2, we investigated this idea to determine whether the effect of contextual overlap was the result of a local coherence problem or the result of goal reactivation.
26 The goal of Experiment 2 was to determine whether the effect of the specific contextual overlap observed in Experiment 1 was due to a greater difficulty to integrate the reinstatement sentence with the last filler sentence in the close overlap condition (i.e., the local coherence hypothesis) or was due to the amount of time required to retrieve the goal by means of the resonance process (i.e., the retrieval hypothesis). In order to test the local coherence and retrieval hypotheses, filler sentences of the narratives were replaced by a nonverbal material: a mental calculus task. As the filler section, this task should allow a complete backgrounding of goal information. Moreover, as the task is totally unrelated with the preceding textual contents as well as with the reinstatement sentence, there is no reason, at the end of the calculus task, that readers attempt to integrate the reinstatement sentence with the current content of the working memory (i.e., information about calculus operations). Because the reinstatement sentence was not related to the information about calculus operations, it is very unlikely that the close overlap condition leads to further integration than the distant overlap condition. In fact, as information about the calculus task is irrelevant to the content of the narrative, no integration should be required. As a result, there is no reason that reading times on the reinstatement sentence would be longer in the close overlap condition.
27 Predictions concerning the reading times as a function of the contextual overlap differ according to the local coherence and retrieval hypotheses. According to the local coherence hypothesis, the effect of contextual overlap obtained in the experiment 1 (i.e., longer reading times on the reinstatement sentence providing a close overlap) should not occur. Specifically, because the reinstatement sentence was preceded by a same nonverbal material (a mental calculus task), close and distant overlap conditions should be equally unrelated with it. As a result, in the close overlap condition, the completion of the task should have the same influence on reading resumption as in the distant overlap condition. Thus, reading times on the reinstatement sentence should not differ reliably among the overlap conditions. According to the retrieval hypothesis, the variation of reading times on the reinstatement sentences observed in experiment 1 should be replicated. Participants should read the reinstatement sentences more slowly when they provide a close overlap than a distant overlap. If, as we assume, the reading times increase still occurs, then the explanation in terms of integration difficulty could be ruled out.
28Participants were 40 students of University of Paris 10. Their participation was volunteer and they did not receive course credit for their involvement in the experiment. All of them were native French speakers.
29 Materials and procedure were globally similar to those used in Experiment 1. Texts were the same as in Experiment 1. The only difference concerned the filler sentences that were replaced by a mental calculus task. We designed the mental calculus task so that its accomplishment requires a time similar to the reading of the filler section (i.e., 30 seconds). Concretely, after the reading of the last sentence of the goal section, participants were instructed to mentally resolve a calculus operation (e.g., 50 x 28 +12 = ?). The task was 30 seconds long in order to remove all textual content from working memory. After the completion of the task, the reading of the text was resumed. At this moment, the context-reinstatement sentence was displayed. Context reinstatement sentences were the same as in experiment 1 and provided a contextual overlap from a subgoal either close or distant of the goal attainment. Importantly, these sentences never made direct reference to the subgoal nor to the goal. Then target sentences and the conclusion section were displayed. As in the experiment 1, after each text, participants answered two questions about the story.
30 We performed two ANOVAs, the first on reading times for the context reinstatement and the second on reading times for the first target sentence. Goal Satisfaction (satisfied vs. unsatisfied) and Contextual Overlap (close vs. distant) were treated as within-subjects variables. As in experiment 1, outliers were identified using Tukey’s (1977) hinge criterion and replaced in each condition by the mean reading time. This substituted less than 6 % of the reading times.
32The analysis yielded a main effect of the contextual overlap, F(1, 39) = 5.15, MSE = 314.59, p = .028. (see Table 2 ). As in Experiment 1, participants read the reinstatement sentence more slowly when it provided a close overlap (M = 6442 ms) than a distant overlap (M = 5752 ms). There was no main effect of goal satisfaction F(1, 39) = 0.49, MSE = 328.61 and no Goal satisfaction x Contextual Overlap interaction F(1, 39) = 2.92, MSE =. 384.22.
33The main effect of contextual overlap on reading times for the target sentence 1 was significant, F(1, 39) = 4.49, MSE = 224.05, p = .04. Reading times in the close overlap condition were longer (M = 4268 ms) than those in the distant overlap condition (M = 3910 ms). In addition, the main effect of the goal satisfaction did not reach the significance criteria, F(1, 39) = 3.93; MSE = 238.47. Reading times tended to be longer when the goal had not been satisfied (M = 4325 ms) than when it had been satisfied (M = 3853 ms). The interaction of the factors was not significant F(1, 39) = 0.33, MSE =. 258.08.
Table 2: Mean reading times (in milliseconds) and standard deviation (under brackets) for the context reinstatement sentence and the first target sentence as a function of goal satisfaction and contextual overlap.
34 The purpose of this second experiment was to distinguish among two explanations in regard to the increase on reading times of the reinstatement sentence when it featured a close overlap. According to the integration difficulty hypothesis, participants should not have spent more time to read reinstatement sentences introducing a close overlap than a distant overlap. Indeed, in this study, context reinstatement sentences in both overlap conditions were equally unrelated with the content of the working memory (information about calculus operations). As a result, according to this hypothesis, no significant reading times difference between the close and the distant overlap conditions should have been obtained. However, results indicate an increase in reading times for the reinstatement sentence in the close overlap condition. So, we can conclude that the local coherence hypothesis cannot accountfor the observed time increase.
35Finally, the slowdown on reading times in the close overlap condition, observed in both experiments, provides further support to the retrieval hypothesis. So it seems that, when the close overlap was present, it reactivated the character’s goal, and this resulted in longer reading times for the reinstatement sentences. In other words, results suggest that the contextual overlap produce an effect on the resonance process at an early stage: the processing of the context reinstatement sentence.
36 In two studies, we investigated whether the resonance and the reactivation of a goal previously stated in a text is affected by a specific contextual overlap: the repetition of a close vs. a distant subgoal-related contextual cue. In previous research, manipulation of contextual overlap depended on the presence (or not), in the context reinstatement sentence, of a contextual cue that previously appeared in the goal context and/or on whether this cue was fully or partially repeated (Albrecht and Myers, 1995; Albrecht and Myers, 1998). Our experiments extended these researches in manipulating contextual overlap in regard of temporal-causal structure of the narratives. Narratives were composed of a goal and three successive subgoals performed in order to reach the goal and contextual overlap was manipulated as a function of the distance (close vs. distant) between the subgoal from which was extracted the contextual cue and attainment of the goal in the temporal-causal structure. We assumed that, within the temporal-causal structure, subgoal 3 would be more strongly related to the goal than subgoal 1, and that, as a consequence, goal information would be more rapidly reactivated by the processing of a contextual cue from subgoal 3 than from subgoal 1. Given these assumptions, we expected that the nature of this specific overlap, related to the goal-subgoals structure of the narratives, could have a particular effect on the resonance process.
37 In experiment 1, participants spent more time to read the first target sentence when it was preceded by a context reinstatement sentence introducing a close overlap than introducing a distant overlap. This increase of reading times occurred regardless of goal satisfaction. If these results may seem odd in regard of the inconsistency detection hypothesis (i.e., longer reading times when the goal is unsatisfied than satisfied), they are however highly consistent with the characterization of retrieval as an unrestricted, dumb, resonance process (Myers & O’Brien, 1998; O’Brien & Myers, 1999): all backgrounded concepts that share features with information in working memory resonate, independently of their relevance to the content of working memory. In our view, the fact that reading times did not differ reliably as a function of goal satisfaction suggest that the resonance process is triggered by a contextual overlap but also that it operates independently of the goal satisfaction. In other words we think that, in presence of an identical overlap, the goal resonates, regardless of its satisfied or unsatisfied status.
38 Interestingly, we found that participants were slower to read the context reinstatement sentence when it contained a contextual cue that previously appeared in a subgoal that was close from goal attainment (close overlap condition) than distant from goal attainment (distant overlap condition). Thus, it seems that the contextual overlap and, specifically, an overlap linked to the goal-subgoals structure of narratives, may have an impact on the speed of the resonance process. Depending on the type of contextual overlap, goal resonance could be more or less immediate. The interpretation of the effect obtained on the context reinstatement sentence reading times gave rise to two main explanations. The first referred to a difficulty of integration of the context reinstatement sentence (local coherence hypothesis), the second referred to differences in terms of goal information retrieval or resonance speed (retrieval hypothesis).
39Results of experiment 2 provided evidence in favour of the retrieval hypothesis, indicating that the time increase is more likely explained by differences in terms of retrieval. However, even if our data indicate that goal retrieval and reactivation account for the observed time increases, processes need to be further explored and explained.
40 Taken as a whole, results indicated that the reading of the reinstatement sentence reactivates the goal of a character, independently of the nature of the material included between the goal section and the context reinstatement sentence and independently of the goal status. It also indicated that the nature of contextual overlap, relatively to the text structure, seems to influence the re-activation of a goal formerly presented in a text.
41 Finally, results stress the importance of the contextual overlap in the resonance process and highlight the importance of the temporo-causal structure of the text in the manipulation of this factor. In that sense, it provides new information about overlap in the resonance framework and raise new research questions. Particularly, it could be fruitful, for the future research, to explore further the relations between text causal and temporal structure, overlap and resonance.