Abstract

This research studies the neural systems underlying two integration processes that take place during natural discourse comprehension: consistency evaluation and passive comprehension. Evaluation was operationalized with a consistency judgment task and passive comprehension with a passive listening task. Using fMRI, the experiment examined the integration of incoming sentences with more recent, local context and with more distal, global context in these two tasks. The stimuli were stories in which we manipulated the consistency of the endings with the local context and the relevance of the global context for the integration of the endings. A whole-brain analysis revealed several differences between the two tasks. Two networks previously associated with semantic processing and attention orienting showed more activation during the judgment than the passive listening task. A network previously associated with episodic memory retrieval and construction of mental scenes showed greater activity when global context was relevant, but only during the judgment task. This suggests that evaluation, more than passive listening, triggers the reinstantiation of global context and the construction of a rich mental model for the story. Finally, a network previously linked to fluent updating of a knowledge base showed greater activity for locally consistent endings than inconsistent ones, but only during passive listening, suggesting a mode of comprehension that relies on a local scope approach to language processing. Taken together, these results show that consistency evaluation and passive comprehension weigh differently on distal and local information and are implemented, in part, by different brain networks.

INTRODUCTION

Different situations can induce a person to listen to the same discourse in different ways. It is one thing, for example, to listen to your 18-year-old son's narrative of how he drove your car into a ditch. Another thing is to hear the same story about your neighbor. In both cases, the narratives interest you, but in your son's case, you are more likely to evaluate the consistency of each and every statement against prior context. Given the implications for your insurance record and related expenses, you will likely check your son's utterances for clear consistency. In the case of your neighbor's accident, however, you may be curious, empathic (or amused), and attentive but would not need to evaluate the consistency of each utterance with the rest of the story.

These two routine ways of understanding discourse, consistency evaluation and passive comprehension, differ not so much in the amount of attention devoted to the incoming sentences, but mostly in the way these sentences are integrated with prior context. Intuitively, evaluation is a critical, reflective type of process that assesses the fit of incoming information with prior context. In contrast, passive comprehension is a simpler assimilative process by which incoming information is integrated with prior context and does not necessarily rely on inconsistency monitoring. The goal of this article is to examine how these two processes differ at the neurobiological level.

There is some evidence of neural and cognitive differences between consistency evaluation and passive comprehension. In general, different reading purposes (e.g., for entertainment or study) lead to differences in inference generation and comprehension (e.g., Linderholm & van den Broek, 2002; Narvaez, Van Den Broek, & Ruiz, 1999). More specifically, ERP and behavioral studies have shown that evaluation and passive listening can yield very different results in discourse comprehension (Xiang & Kuperberg, 2014; Egidi & Caramazza, 2013; Egidi & Gerrig, 2006, 2009). For example, in ERP research, consistency evaluation has been shown to lead to stronger differences in N400 than passive comprehension for words that are incoherent with respect to prior discourse context (e.g., “She took the test and aced/failed it. She went home and celebrated”; Xiang & Kuperberg, 2014).

Evidence from behavioral studies on discourse consistency further suggests that evaluative and passive comprehension perspectives engage different types of integration. In two prior behavioral studies (Egidi & Caramazza, 2013) that used the same stimuli also employed in the current fMRI experiment (see Figure 1 for an example), we found that evaluative comprehension engages a more global assessment of discourse context. We examined the processing of consistency with short narratives in which the final sentence was either consistent or inconsistent with the immediately preceding, local context. Independently, the more distal, global context was manipulated to be either relevant or irrelevant to this final sentence. When it was not relevant, its presence had no bearing on the integration of the ending, so that local consistency was the only possibility. When the global context was relevant, however, there was always a possibility for consistency: The ending that was inconsistent with the local context was consistent with the global context, and vice versa for the ending that was consistent locally. By creating a situation of conflict between the demands of the local and global contexts, we were thus able to dissociate their impact on the integration process.

Figure 1. 

Sample story used in the experiments and schematic of the experimental conditions.

Figure 1. 

Sample story used in the experiments and schematic of the experimental conditions.

These texts were presented under evaluative processing or passive comprehension task requirements (in the norming study and Experiment 1, respectively; Egidi & Caramazza, 2013). Evaluative processing promoted global assessments of the entire context as seen in the fact that explicit evaluations of the final sentence reflected the integrated content of both local and global contexts (see Figure 2A). Reading times during passive comprehension, however, always reflected the local consistency status of the target sentence, although with a modulation by the global context (Figure 2B). These results are consistent with other behavioral evidence on discourse comprehension also supporting the hypothesis that evaluation leads to prevalently global integration and passive comprehension leads to prevalently local integration (Egidi & Gerrig, 2006, 2009).

Figure 2. 

Results of behavioral measures obtained with the stimuli used in the current experiment (means and confidence intervals; CI after Loftus & Masson, 1994). The data of A and B are reported in detail in Egidi and Caramazza (2013) and are shown here for comparison purposes only. (A) Consistency ratings for the story endings during evaluative processing. Participants rated how accurately a given ending described what they thought would happen next in the story (on a scale ranging from 1 = definitely not to 9 = definitely yes), following an established procedure that measures participants' evaluation of text consistency (e.g., Rapp & Kendeou, 2007; Egidi & Gerrig, 2006; Rapp & Gerrig, 2006). Participants judged the locally consistent endings as more consistent when the global context was irrelevant. When the global context was relevant, however, the pattern was reversed. (B) Reading latencies for the endings during passive comprehension. Participants read the locally consistent endings faster than the locally inconsistent ones. This effect was stronger when the global context was irrelevant but occurred with both types of global contexts. (C) Consistency ratings as percentage of “yes” responses for the story endings obtained from the judgment task group during the fMRI scanning. Participants indicated whether a given ending described what they thought would happen next in the story (possible answers were “yes” and “no”).

Figure 2. 

Results of behavioral measures obtained with the stimuli used in the current experiment (means and confidence intervals; CI after Loftus & Masson, 1994). The data of A and B are reported in detail in Egidi and Caramazza (2013) and are shown here for comparison purposes only. (A) Consistency ratings for the story endings during evaluative processing. Participants rated how accurately a given ending described what they thought would happen next in the story (on a scale ranging from 1 = definitely not to 9 = definitely yes), following an established procedure that measures participants' evaluation of text consistency (e.g., Rapp & Kendeou, 2007; Egidi & Gerrig, 2006; Rapp & Gerrig, 2006). Participants judged the locally consistent endings as more consistent when the global context was irrelevant. When the global context was relevant, however, the pattern was reversed. (B) Reading latencies for the endings during passive comprehension. Participants read the locally consistent endings faster than the locally inconsistent ones. This effect was stronger when the global context was irrelevant but occurred with both types of global contexts. (C) Consistency ratings as percentage of “yes” responses for the story endings obtained from the judgment task group during the fMRI scanning. Participants indicated whether a given ending described what they thought would happen next in the story (possible answers were “yes” and “no”).

This design shares commonalities with paradigms in the behavioral literature that have used inconsistencies to study coherence at the local or global level (or other aspects of discourse comprehension) in that it also uses inconsistent and consistent information to test for language processing (e.g., Egidi & Gerrig, 2006; Rapp & Gerrig, 2006; Myers, O'Brien, Albrecht, & Mason, 1994; Albrecht & O'Brien, 1993; McKoon & Ratcliff, 1992). However, by creating a situation of conflict between the demands of the local and global contexts, our paradigm dissociates their impact on the integration process in a novel way.

Neuroimaging studies also suggest a dissociation between evaluation and passive comprehension. In general, there is evidence that language processing at the word, sentence, and discourse levels is sensitive to the task employed during comprehension (e.g., words and word pairs: Graves, Binder, Desai, Conant, & Seidenberg, 2010; Ethofer et al., 2009; Kuperberg, Lakshmanan, Greve, & West, 2008; sentence: Huang et al., 2012; Suzuki & Sakai, 2003; discourse: Cooper, Hasson, & Small, 2011; Ferstl & von Cramon, 2002). Interestingly, even different beliefs about the type of content one is processing (e.g., fact vs. fiction) can induce the involvement of different networks (Altmann, Bohrn, Lubrich, Menninghaus, & Jacobs, 2014).

Neuroimaging studies that have compared tasks inducing different types of evaluative processing and passive comprehension have shown that such manipulations influence activity in inferior frontal gyrus (IFG), middle temporal gyrus (MTG), and superior temporal gyrus (STG; Yue, Zhang, Xu, Shu, & Li, 2013; Zhu et al., 2012; Chow, Kaup, Raabe, & Greenlee, 2008; Hasson, Nusbaum, & Small, 2006; Rodd, Davis, & Johnsrude, 2005; Noesselt, Shah, & Jäncke, 2003; but see Rodd et al., 2005, for no difference). These studies have however focused on single words or sentences, thus leaving open the question of how these processes are implemented in discourse. These studies have also operationalized evaluative processing with very different types of judgments, which likely tap into different types of evaluations (e.g., semantic anomaly detection, relatedness judgment between a probe word and a prior sentence, lexical decision). In our study, we examine the most fundamental and usual type of evaluative processing: evaluation of consistency. A basic assumption of all theories of discourse comprehension is that reading or listening to discourse involves creating a coherent representation of the discourse (for a review, see McNamara & Magliano, 2009).

Our fMRI experiment was therefore designed to examine how consistency evaluation and passive comprehension attain local and global consistency at the neural level. By using the same stimuli as in Egidi and Caramazza (2013), we could assess local and global integration separately. To implement evaluative processing in the fMRI study, we asked participants to make a consistency judgment on the story endings (as in, e.g., Rapp & Kendeou, 2007; Rapp & Gerrig, 2006). To implement passive comprehension, we asked participants to listen for comprehension. We conducted a whole-brain analysis on the data obtained from both groups of participants. We expected the two tasks to differ in three respects.

First, because evaluation is a more critical, reflective type of processing than passive comprehension, we expected that regions associated with semantic processing would generally show increased activation for the judgment than the listening task. These regions include IFG, posterior middle frontal gyrus (MFG), middle precentral gyrus (PreCG), dorsomedial pFC (dmPFC), anterior insula (AI), the anterior and posterior parts of STG and STS, the middle and posterior parts of MTG, posterior inferior temporal gyrus (ITG), temporal pole, anterior fusiform gyrus (FUS), and angular gyrus (AG; Mar, 2011; Vigneau et al., 2006, 2011; Binder, Desai, Graves, & Conant, 2009; Hagoort & Indefrey, 2014).

Second, because evaluation promotes more global integration, we expected greater sensitivity to the relevance of the global contexts in the judgment task than in the listening task. Theories of discourse comprehension argue that greater consideration of global context would lead to reinstantiation (or active maintenance) of a larger amount of prior context and, consequently, to the construction of a richer mental model for the story (McNamara & Magliano, 2009). We therefore hypothesized that the regions associated with global integration would be those associated with episodic memory retrieval (Rugg & Vilberg, 2013; Vilberg & Rugg, 2012) and construction of a complex mental scene (Andrews-Hanna, Reidler, Sepulcre, Poulin, & Buckner, 2010; Hassabis & Maguire, 2007). These regions include parahippocampal cortex (PHC), retrosplenial cortex (RSC), and posterior cingulate cortex (PCC) extending to precuneus (PCUN), AG, ventromedial pFC (vmPFC), and MTG. In this network, we expected a stronger differentiation between relevant and irrelevant global contexts in the judgment task than in the passive listening task.

Finally, because passive comprehension prompts a less global approach to language processing that is focused on the most recent context, we also expected greater ease in processing locally consistent endings for the listening task than for the judgment task. This should be seen in a network that prior research has identified as associated with continuous, fluent, monotonic updating of a knowledge base for locally consistent information, a process that we have called “accumulation” (Egidi & Caramazza, 2013, 2014). This process seems to share similarities with processes postulated by text processing models such as the Construction–Integration Model (CI; Kintsch, 1988, 1998) or the Landscape Model (LM; Tzeng, van den Broek, Kendeou, & Lee, 2005). However, although both our notion of accumulation and the CI/LM notion of associative assimilation of words share the common idea of a frequent, cyclical, and dynamic process, in those models, the dynamic activation refers to processes mostly at the word level—of initial activation of words and related concepts, which can be overridden by global considerations during subsequent integration with a wider context. The evidence we have so far for accumulation, however, suggests that this is a process of integration, which occurs between an incoming sentence and its local discourse context (i.e., the preceding two or three sentences). In addition, the brain areas that track coherence with local context are the ones where the textual representation is not overridden by global context (see Egidi & Caramazza, 2013, for evidence of the independence of these networks). In this respect, the process of accumulation may be as resilient as that of integration with global context. Further research is however necessary to fully assess the extent of the similarities between accumulation and the associative assimilation proposed by CI and LM.

From the neurobiological point of view, the accumulation network encompasses, bilaterally, the PreCG (mostly the dorsal part), a posterior part of the superior frontal gyrus (SFG), middle postcentral gyrus (PostCG), supramarginal gyrus (SMG), superior parietal lobule (SPL), central cingulate cortex (CC), anterior cuneus (CUN), and insula (mostly posterior and central insula). Only on the left, this network also includes a portion of vmPFC and posterior parts of MTG, MTS, and ITG. We expected regions of this network to show greater activation for locally consistent than locally inconsistent endings, in line with prior language literature showing that these regions are more strongly activated when linguistic information is easier to understand (e.g., Hasson, Nusbaum, & Small, 2007; Vingerhoets et al., 2003; Ferstl & von Cramon, 2001, 2002; Perani et al., 1996, 1998). For example, regions in this network are more active in bilingual speakers processing their first language as compared with their second language (e.g., CUN, CC; Vingerhoets et al., 2003; Perani et al., 1996, 1998), show greater BOLD signal for sentences coherent with prior one-sentence context (as compared with incoherent ones; e.g., CC, vmPFC; Ferstl & von Cramon, 2001, 2002), and show stronger activity for consistent endings that are consistent with prior discourse (as compared with sentences that require an update of a story mental model; e.g., PreCG, PostCG ; Hasson et al., 2007). In this network, we expected a stronger increase in the activation associated with the locally consistent (vs. locally inconsistent) endings in the listening task than the judgment task.

METHODS

Participants

Twenty-eight native Italian speakers participated in the experiment; 14 completed the judgment task (M age = 27.28 years, SD = 8.13 years; women = 3); and 14, the listening task (M age = 25.29 years, SD = 4.86 years; women = 4). Participants were right-handed; had good or corrected vision, good hearing, and no mood or attention disorders; and did not take psychotropic medications. Before the scan, they underwent a medical interview, which evaluated other exclusion criteria (e.g., claustrophobia, pregnancy). Experimental procedures had the approval of the ethical committee for research involving human participants of the University of Trento. The data of the listening group have been used in Egidi and Caramazza (2013) for different analyses.

Stimuli and Design

We used 40 stories, each 6–10 sentences long, which described simple life events. The critical sentence described a protagonist's action and was the last sentence. The first few sentences of each story constituted the distal context for the endings; they are therefore referred to as the “global context.” The next few sentences, immediately preceding the ending sentence, constituted a proximal context and are therefore referred to as the “local context.” A sample story and a schematic depiction of the experimental conditions are shown in Figure 1.

Each story could have two endings: one consistent and one inconsistent with the local context. The global context could be either irrelevant or relevant for the integration of the ending. This resulted in four experimental conditions. When the global context was irrelevant to the integration of the ending, the coherence of the ending with prior content depended only on the local context. When the global context was relevant for the integration of the ending, a locally consistent ending was inconsistent with the global context, and conversely, a locally inconsistent ending was consistent with global context. A norming procedure ascertained that the stories functioned as designed (see Egidi & Caramazza, 2013, for details).

The stories were in Italian. Because we wanted to compare the BOLD activity associated with the comprehension of the endings, we balanced lexical and semantic overlap across versions of each story. We also made sure that ending sentences had similar syntactic structure and were of similar length (between 12 and 14 syllables). We used a Latin square to assign the stories to different lists in a counterbalanced fashion, so that each participant would be presented with only one version of each story (with one type of global context and one type of ending). This resulted in the construction of four lists in each experiment, each containing 40 experimental stories, which were assigned randomly to each list. In addition, we used 18 additional stories as fillers and practice stories. They were similar in content, length, and structure to the experimental narratives.

All stories were recorded for auditory presentation in the experiment. The speaker was blind to the purpose of the experiment and recorded the endings separately from the stories, so as not to bias intonation in reading the story bodies as a function of the endings. We also selected two video clips of about 8 min each, which we used to keep participants alert during the acquisition of the structural images when they were not required to perform experimental tasks.

Participants either judged the consistency of the ending with the story or listened to stories and endings for comprehension. The design therefore consisted of one between-participant variable, Task (listening, judgment), and two within-participant variables, Global context relevance (irrelevant, relevant) and Ending local consistency (consistent, inconsistent).

Procedure

Procedures for judgment and listening tasks were intended to induce participants to process stories and endings differently in the two task groups. To induce these processing differences, we gave participants different instructions on what to do during the presentation of the endings. Figure 3 shows the procedure for experimental and filler stories for both tasks.

Figure 3. 

Procedure for story presentation in the MR scanner in the judgment and passive listening tasks. The procedure was identical for the presentation of experimental stories in both tasks, and it differed only while participants listened to the story endings: In the judgment task, participants saw a prompt on the screen informing them that they were hearing the part of the story to be evaluated; in the listening task, participants listened to the story endings without engaging in any explicit task. The procedure for the filler stories was identical to that of the listening task. It differed only in that participants answered to a comprehension question after the story had ended. The critical window for both judgment and passive listening tasks was the 4.4 sec between the onset of the ending sentence and the onset of the button press screen.

Figure 3. 

Procedure for story presentation in the MR scanner in the judgment and passive listening tasks. The procedure was identical for the presentation of experimental stories in both tasks, and it differed only while participants listened to the story endings: In the judgment task, participants saw a prompt on the screen informing them that they were hearing the part of the story to be evaluated; in the listening task, participants listened to the story endings without engaging in any explicit task. The procedure for the filler stories was identical to that of the listening task. It differed only in that participants answered to a comprehension question after the story had ended. The critical window for both judgment and passive listening tasks was the 4.4 sec between the onset of the ending sentence and the onset of the button press screen.

Participants were given instructions and practiced the procedure on a block of four filler stories before entering the scanner. The instructions informed participants that each trial would begin with a warning sound and the appearance of a cross in the center of the screen, followed by a voice telling a story. All participants were asked to pay attention to each story as to a friend's voice recounting an anecdote.

For the judgment task, the instructions indicated that, at the end of each story, the cross would disappear from the screen and participants could be asked to do one of two things. In some cases, a question about the content of the story would appear on the screen, and they could answer either “yes” or “no” by pressing one of the two keys on the button box (within 4 sec, as after this delay, the question would disappear and the software would not record the response). Participants were asked to answer this question as accurately as possible.

In other cases, during the presentation of the last sentence of the story, the cross on the screen would turn into a question mark. When this happened, participants had to evaluate whether what the voice said while the question mark was on the screen was what they thought would happen next in the story. This question followed an established procedure that measures participants' evaluation of text consistency (e.g., Rapp & Kendeou, 2007; Egidi & Gerrig, 2006; Rapp & Gerrig, 2006). After this screen disappeared, participants would see a screen that reminded them of the judgment they were asked to make (the screen showed the question: “The sentence described what you thought would happen next?”). Participants were asked to indicate their judgment on this screen with a “yes/no” response by pressing one of the two keys on the button box. This screen would also disappear in 4 sec; participants were therefore asked to respond within this time frame.

Each trial would end with the presentation of a gray screen for several seconds indicating that participants could rest. In this procedure, filler stories were those followed by a comprehension question, and experimental stories were those followed by a judgment.

For the listening task, the instructions were very similar but crucially different on what participants were asked to do while listening to the endings. Participants were again informed that, at the end of each story, the cross would disappear from the screen, and they could be asked to do one of two things. In some cases, they could be asked to answer a question on the content of the story, with the same procedure described above for the filler stories in the judgment task. In other cases, after hearing the story, participants would see two circles on the screen and had to press a key (within 4 sec). Again, trials ended with a gray screen that would be shown for several seconds, during which participants could rest. In this procedure, filler stories were those followed by a comprehension question, and experimental stories were those followed by a key press.

We asked participants to make button presses after experimental stories in the listening task to ensure that they would follow the progress of the experiment and to equate their motor response to that of the participants in the judgment group. We used comprehension questions on filler stories to ensure that participants would pay attention to the content of the stories throughout the experiment. Participants responded using a two-key button box placed under one hand. Half of the participants used the left hand and half used the right hand.

The scanning protocol had the following structure: acquisition of a structural image, during which participants watched a video clip, 20 min for two blocks of stories (10 min each), and then again, acquisition of a structural image, during which participants watched a video clip, and again 20 min for two blocks of stories (10 min each). Participants were asked to pay attention to videos and stories equally and were told that they would be asked questions about both once they finished the part in the scanner.

After the part in the scanner, participants completed a recall test on video clips and stories' endings. They also completed a survey about the experiment and some personal preferences. These data are not relevant for the questions addressed here and will not be discussed further.

fMRI Data Acquisition

We acquired fMRI data in a 4-T scanner with an eight-channel head coil at the Center for Mind/Brain Sciences of the University of Trento. The functional EPI sequences consisted of 37 axial T2-weighted functional images in ascending interleaved order covering the entire brain, slightly tilted parallel to the AC–PC line (repetition time = 2200 msec, isotropic voxels size = 3 mm, gap size between slices = 0.45 mm, echo time = 33 msec, flip angle = 75°, field of view = 192 × 192 mm). Each participant completed four runs of 270 volumes. Each functional run was preceded by an additional brief scan that measured the point spread function of the acquired sequence and allowed correcting for the distortions that the inhomogeneity of the magnetic field might have caused in certain regions (Zaitsev, Hennig, & Speck, 2004). The anatomical images consisted of high-resolution images acquired with a T1-weighted magnetization prepared rapid gradient echo sequence. These were 176 sagittal images (repetition time = 2700 msec, isotropic voxel size = 1 mm, echo time = 4 msec, flip angle = 7°, field of view = 256 × 254 mm). We acquired two anatomical images: one at the beginning of the experiment and one after participants had completed two experimental runs.

fMRI Data Processing, Surface Projection, and First-level Analyses

fMRI data were processed by using AFNI's procedures (Cox, 1996). For each participant's functional data, the initial 11 volumes of each run, which had been acquired to allow for scanner stabilization before the beginning of the experimental tasks, were removed. All runs were then registered to a reference time point in the first run, head motion corrected, and blurred (Gaussian blur, FWHM = 6.0 mm), and their time series were normalized to a percent change value.

A regression model was then applied to the data. Regressors were waveforms with similarity to the hemodynamic response and were generated by convolving a gamma-variant function with the onset time and duration of the endings and with the onset time and duration of the stories. There were five regressors of interest, one for each experimental condition and one that modeled the experimental story bodies and the filler stories. These regressors modeled whether the presentation of each of the elements of interest accounted for variance above and beyond the variance accounted for by the other events occurring in the experiment. Other regressors reflected factors of no interest and included the button presses, the comprehension questions to the filler stories, the six motion parameters estimated during head motion correction, and first- to fourth-order polynomial trends fitted for each run separately to account for instrumentation-induced drifts in the signal. Functional acquisitions associated with strong head movement (>2 mm) were also censored from the regression model, as were acquisitions that presented outlier values in a large number of voxels (>2000). These accounted for 4.6% of the data.

For each participant, the two magnetization prepared rapid gradient echo structural images were co-registered and averaged to increase signal-to-noise ratio. Using warping procedures implemented in FreeSurfer (Fischl, Sereno, & Dale, 1999), left and right hemispheres of each participant were inflated to a surface representation and were aligned to a common template, and a mean cortical surface was created by averaging participants' individual cortical surfaces. All single-participant regression results (i.e., beta weights of interest) were projected to this average cortical representation for purposes of group-level analysis. Each participant's mean anatomical image was aligned to the functional data with AFNI procedures (Saad et al., 2009). The results of the regression analysis (the regressors of theoretical interest) were then projected from the 3-D volumes to the 2-D cortical surfaces.

fMRI Second-level Analysis

We conducted a group analysis that identified brain regions where activity for story endings varied as a function of the three experimental variables and their interaction. Using a whole-brain vertex-wise analysis in the 2-D surface domain, the beta values for all participants (modeled as a random factor) were analyzed using a 2 (Task) × 2 (Global context relevance) × 2 (Ending local consistency) anova for each vertex.

The analysis was controlled for family-wise error using cluster-based constraints: Cluster thresholding was based on Monte Carlo simulation methods (Forman et al., 1995) implemented on the surface domain that controlled for family-wise error rate at p < .05. These simulations take into account the smoothing in the data, the allowed distance between active vertices (2 mm), and vertex thresholding. In cluster-based thresholding, cluster size depends monotonically on the vertex threshold selected: More liberal single-vertex thresholds result in large cluster-size magnitudes. We used two single-vertex thresholds set at an alpha level of p < .05 and p < .005 to identify both larger, distributed clusters and smaller, focal clusters (as in, e.g., Egidi & Caramazza, 2013; Hasson et al., 2007). The results of both clustering procedures are shown in the figures and tables. However, discussion of the results in the article is mainly based on the clustering at vertex thresh old of p < .05, as the more focal clusters tended to be subsets of the larger ones. From the clustering at vertex threshold of p < .005, clusters that do not overlap with those of clustering at vertex threshold of p < .05 are explicitly discussed. Only the reliable effects of interest, that is, those in which the judgment task and the listening task statistically differ, are reported and discussed here.

Data Processing and Statistical Analyses for Judgments and Comprehension Questions

During the fMRI study, participants in the judgment task group responded to each experimental story by indicating whether the given ending was consistent with prior context. On the percentages of “yes” responses obtained for each participant, we conducted a 2 (Global context relevance) × 2 (Ending local consistency) ANOVA.

During the fMRI scanning, participants in both listening and evaluation tasks also answered comprehension questions that followed the presentation of the filler stories. On the percent accuracy obtained for each participant, we conducted an independent t test (listening task vs. judgment task).

RESULTS

Behavioral Results

Judgments of Ending Consistency

Participants in the judgment group provided a consistent/inconsistent judgment for each ending by pressing either of two buttons: one for yes (consistent) and one for no (inconsistent). Figure 2C shows the pattern of results. When the global context was irrelevant, that is, when the stories contained global information that was irrelevant for the evaluation of the endings, participants judged the locally consistent endings as consistent in most of the cases, as would be expected. However, when the global context was relevant, the pattern changed markedly: The locally inconsistent endings—now globally consistent—were judged as more predictable. These response patterns resulted in a statistically significant interaction between global context relevance and ending local consistency (F(1, 13) = 111.95, MSe = .023, p < .001).

The results clearly confirmed our operational definition of the four conditions. These results are also consistent with the results of the norming study of the stories reported in Egidi and Caramazza (2013), with which it can be compared in Figure 2. The convergence of the two patterns of results shows unequivocally that a judgment process promotes integration of global context.

Comprehension Accuracy for Judgment and Listening Tasks

To rule out the possibility that different comprehension tasks would induce one group to pay more attention than the other, we compared the accuracy of the judgment and listening tasks in the comprehension questions that followed the presentation of the filler stories. Participants in both groups responded to the comprehension questions that followed the filler story with very similar accuracy (Mjudge = 88%, Mlisten = 86%, t(26) = −0.3, p = .766).

fMRI Results

Neural Activity Associated with Comprehension during Evaluation and Passive Listening

A whole-brain analysis identified brain regions where BOLD response during comprehension of story endings differed as a function of the task participants performed and the independent variables of the text they listened to (ending local consistency and global context relevance).

We identified three networks: One included areas sensitive to general differences between the two tasks (main effect of task), one included areas sensitive to the differences between the two tasks with respect to global context relevance (interaction between global context relevance and task), and one included regions sensitive to the differences between the two tasks with respect to ending local consistency (interaction between ending local consistency and task). Figures 4 and 5 and Table 1 show these results. We did not find a three-way interaction or a two-way interaction between the local consistency of the endings and the relevance of the global context.1

Figure 4. 

Clusters where BOLD activity varied as a function of the experimental variables in the three effects found in the analysis: main effect of task, interaction between global context relevance and task, and interaction between ending local consistency and task. Clusters are shown on the cortical surface. Clusters in red are those calculated with a voxel threshold of p < .005; the others are calculated with a threshold of p < .05. Clusters are labeled with the abbreviations given in Table 1 for the cluster centers; that is, they do not describe the full extent of the clusters but indicate only their centers. For a full description of the regions involved in each cluster, see text.

Figure 4. 

Clusters where BOLD activity varied as a function of the experimental variables in the three effects found in the analysis: main effect of task, interaction between global context relevance and task, and interaction between ending local consistency and task. Clusters are shown on the cortical surface. Clusters in red are those calculated with a voxel threshold of p < .005; the others are calculated with a threshold of p < .05. Clusters are labeled with the abbreviations given in Table 1 for the cluster centers; that is, they do not describe the full extent of the clusters but indicate only their centers. For a full description of the regions involved in each cluster, see text.

Figure 5. 

Pattern of signal change for each of the found clusters and each experimental condition. Clusters calculated with a voxel threshold of p < .005 are labeled with a “.005” after the region label; the others are calculated with a threshold of p < .05. Only the signal change of the focal clusters that do not overlap with the broader clusters is reported here. Clusters are labeled with the abbreviations used in Table 1 and Figure 4 for the cluster centers; as in Table 1 and Figure 4, they do not describe the full extent of the clusters but indicate only their centers. For a full description of the regions involved in each cluster, see text.

Figure 5. 

Pattern of signal change for each of the found clusters and each experimental condition. Clusters calculated with a voxel threshold of p < .005 are labeled with a “.005” after the region label; the others are calculated with a threshold of p < .05. Only the signal change of the focal clusters that do not overlap with the broader clusters is reported here. Clusters are labeled with the abbreviations used in Table 1 and Figure 4 for the cluster centers; as in Table 1 and Figure 4, they do not describe the full extent of the clusters but indicate only their centers. For a full description of the regions involved in each cluster, see text.

Table 1. 

Talairach Coordinates for Cluster Centers and for Locations of Maximal Effects

EffectSignal ChangeVertex ThresholdCluster CenterCluster Maximum Vertex
RegionAbbrevArea (mm2)xyzxyzt ValueEffect Size (Cohen's d)
Task pos: J > L .05 L intraparietal sulcus IPS 2081 −24 −57 29 −26 −60 48 5.50 2.18 
.05 L lingual gyrus LG 2073 −18 −74 −6 −21 −76 −7 6.59 2.60 
.05 L superior temporal gyrus STG 1529 −50 −27 −61 −31 4.79 1.91 
.05 L inferior frontal gyrus, pars triangularis IFG-TRI 1378 −50 23 20 −47 17 21 6.72 2.65 
.05 L superior frontal gyrus/BA 6 SFG 784 −38 12 46 −40 46 5.69 2.25 
.005 L anterior cingulate cortex ACC 120 −12 42 −1 −12 49 −1 4.43 1.78 
.005 L thalamus THA 38 −12 −22 −3 −2 −24 −11 4.56 1.82 
.05 R lingual gyrus/BA 18 LG 1679 17 −76 −7 29 −78 −11 5.27 2.09 
.05 R middle frontal gyrus MFG 1596 30 10 32 36 16 25 4.32 1.73 
.05 R middle occipital gyrus MOG 682 29 −71 13 35 −76 18 4.25 1.71 
.05 R superior temporal gyrus STG 624 51 −23 45 −33 −5 3.29 1.35 
.005 R inferior parietal lobule IPL 101 35 −67 36 37 −52 40 4.59 1.84 
.005 R inferior frontal gyrus, pars triangularis IFG-TRI 81 47 19 12 47 19 10 5.19 2.06 
.005 R precuneus/BA 7 PCUN 67 −71 38 −68 49 3.85 1.56 
.005 R medial frontal gyrus/BA 32 MedFG 50 41 46 4.69 1.87 
pos: J < L .005 L insula INS 45 −43 −7 12 −33 −4 16 3.80 1.54 
neg: J < L .005 L medial frontal gyrus/BA 32 MedFG 129 −8 10 43 −7 10 54 4.73 1.89 
Global context relevance by task pos: J: RelGlobCont > IrrGlobCont; L: RelGlobCont = IrrGlobCont .05 L superior temporal gyrus/BA 22 STG 1230 −47 −13 −52 −12 4.16 1.67 
.05 L posterior singulate cortex PCC 693 −9 −48 16 −9 −57 13 4.22 1.69 
.05 L parahippocampal gyrus/BA 28 PHG 662 −19 −28 −7 −22 −33 −17 4.21 1.69 
.05 R posterior cingulate cortex PCC 1109 −42 10 16 −30 −7 2.20 0.95 
.05 R superior temporal gyrus STG 765 46 −12 −9 53 −2 −11 4.53 1.81 
.05 R middle temporal gyrus/BA 19 MTG 665 41 −59 17 50 −59 21 3.58 1.46 
.005 R medial frontal gyrus/BA 10 MedFG 52 42 −8 48 −17 4.04 1.63 
Ending local consistency by task pos: J: LocConsEnd = LocIncoEnd; L: LocConsEnd > LocIncoEnd .05 L Postcentral gyrus/BA 43 PostCG 3138 −48 −14 17 −35 −12 −1 4.74 1.89 
.05 L central cingulate cortex CC 676 −6 −1 30 −1 −1 26 4.58 1.83 
.05 R insula INS 2349 37 −9 36 −3 −6 4.64 1.85 
.05 R inferior parietal lobule/BA 40 IPL 743 56 −42 25 64 −38 15 4.93 1.96 
.05 R cuneus CUN 656 19 −70 24 −63 4.01 1.61 
.05 R Postcentral gyrus PostCG 544 45 −33 50 45 −26 53 3.32 1.36 
.05 R central cingulate cortex CC 537 28 −5 26 4.22 1.70 
.005 R central cingulate cortex/BA 23 CC 51 −15 33 12 −25 35 2.21 0.95 
EffectSignal ChangeVertex ThresholdCluster CenterCluster Maximum Vertex
RegionAbbrevArea (mm2)xyzxyzt ValueEffect Size (Cohen's d)
Task pos: J > L .05 L intraparietal sulcus IPS 2081 −24 −57 29 −26 −60 48 5.50 2.18 
.05 L lingual gyrus LG 2073 −18 −74 −6 −21 −76 −7 6.59 2.60 
.05 L superior temporal gyrus STG 1529 −50 −27 −61 −31 4.79 1.91 
.05 L inferior frontal gyrus, pars triangularis IFG-TRI 1378 −50 23 20 −47 17 21 6.72 2.65 
.05 L superior frontal gyrus/BA 6 SFG 784 −38 12 46 −40 46 5.69 2.25 
.005 L anterior cingulate cortex ACC 120 −12 42 −1 −12 49 −1 4.43 1.78 
.005 L thalamus THA 38 −12 −22 −3 −2 −24 −11 4.56 1.82 
.05 R lingual gyrus/BA 18 LG 1679 17 −76 −7 29 −78 −11 5.27 2.09 
.05 R middle frontal gyrus MFG 1596 30 10 32 36 16 25 4.32 1.73 
.05 R middle occipital gyrus MOG 682 29 −71 13 35 −76 18 4.25 1.71 
.05 R superior temporal gyrus STG 624 51 −23 45 −33 −5 3.29 1.35 
.005 R inferior parietal lobule IPL 101 35 −67 36 37 −52 40 4.59 1.84 
.005 R inferior frontal gyrus, pars triangularis IFG-TRI 81 47 19 12 47 19 10 5.19 2.06 
.005 R precuneus/BA 7 PCUN 67 −71 38 −68 49 3.85 1.56 
.005 R medial frontal gyrus/BA 32 MedFG 50 41 46 4.69 1.87 
pos: J < L .005 L insula INS 45 −43 −7 12 −33 −4 16 3.80 1.54 
neg: J < L .005 L medial frontal gyrus/BA 32 MedFG 129 −8 10 43 −7 10 54 4.73 1.89 
Global context relevance by task pos: J: RelGlobCont > IrrGlobCont; L: RelGlobCont = IrrGlobCont .05 L superior temporal gyrus/BA 22 STG 1230 −47 −13 −52 −12 4.16 1.67 
.05 L posterior singulate cortex PCC 693 −9 −48 16 −9 −57 13 4.22 1.69 
.05 L parahippocampal gyrus/BA 28 PHG 662 −19 −28 −7 −22 −33 −17 4.21 1.69 
.05 R posterior cingulate cortex PCC 1109 −42 10 16 −30 −7 2.20 0.95 
.05 R superior temporal gyrus STG 765 46 −12 −9 53 −2 −11 4.53 1.81 
.05 R middle temporal gyrus/BA 19 MTG 665 41 −59 17 50 −59 21 3.58 1.46 
.005 R medial frontal gyrus/BA 10 MedFG 52 42 −8 48 −17 4.04 1.63 
Ending local consistency by task pos: J: LocConsEnd = LocIncoEnd; L: LocConsEnd > LocIncoEnd .05 L Postcentral gyrus/BA 43 PostCG 3138 −48 −14 17 −35 −12 −1 4.74 1.89 
.05 L central cingulate cortex CC 676 −6 −1 30 −1 −1 26 4.58 1.83 
.05 R insula INS 2349 37 −9 36 −3 −6 4.64 1.85 
.05 R inferior parietal lobule/BA 40 IPL 743 56 −42 25 64 −38 15 4.93 1.96 
.05 R cuneus CUN 656 19 −70 24 −63 4.01 1.61 
.05 R Postcentral gyrus PostCG 544 45 −33 50 45 −26 53 3.32 1.36 
.05 R central cingulate cortex CC 537 28 −5 26 4.22 1.70 
.005 R central cingulate cortex/BA 23 CC 51 −15 33 12 −25 35 2.21 0.95 

The table reports clusters where BOLD activity varied as a function of the following: main effect of task, interaction between global context relevance and task, and interaction between ending local consistency and task. For the Signal Change column: “pos” and “neg” indicate whether activation is found above or below zero. Clusters match regions in Figures 4 and 5. J = judgment task; L = listening task; RelGlobCont = relevant global context; IrrGlobCont = irrelevant global context; LocConsEnd = locally consistent ending; LocIncoEnd = locally inconsistent ending.

Main effect of task

Frontal regions sensitive to the differences between the listening task and the judgment task included bilateral IFG, MFG, and the inferior part of the precentral sulcus. The middle and medial parts of SFG were involved on the left and the posterior part on the right. In IFG, pars opercularis and pars triangularis were included on the left, but only the pars triangularis was included on the right. Activation on the left extended to also touch PreCG.

Temporal clusters included the lateral aspect of STG and STS bilaterally. An occipito-temporal cluster included bilaterally the posterior part of FUS, extended to part of the lingual gyrus (LG), and the medial and lateral aspects of the occipito-temporal sulcus. Parieto-occipital clusters included, bilaterally, superior occipital sulcus and the posterior part of the intraparietal sulcus (IPS). On the left, the cluster extended parietally to also include the main part of IPS, a portion of SPL, and a very small part of dorsal PCUN. The pattern of signal change was consistently the same for all clusters: increased activity for the judgment task.

The analysis with a stricter voxel threshold identified additional medial clusters: bilateral clusters in medial frontal gyrus, left anterior medial pFC (amPFC), and right dorsal PCUN. Laterally, this analysis also found clusters in left central insula, thalamus (which we will not discuss further, as the projection of the surface is not very accurate for subcortical regions), right SPL, and an additional, more ventral cluster in IFG pars triangularis. The pattern of signal change for these clusters was the same as for the clusters found with a more liberal threshold: increased activity for the judgment task. The only exceptions were the left amPFC and insular clusters. In left amPFC, the judgment task was associated with greater deactivation, and in the left insular cluster, the judgment task was associated with decreased activation. The pattern of signal change for these two clusters is shown in Figure 5.

Interaction between global context relevance and task

The regions sensitive to the global context as a function of the task included ventral PCUN, PCC, RSC, posterior PHC, the anterior lateral parts of STS, and STG. These regions were found bilaterally. On the left, activation was also found in posterior insula, ventral subcentral gyrus, and transverse temporal gyrus. On the right, AG was also included. The analysis with a stricter voxel threshold also identified a cluster in the right vmPFC. The pattern of activity in these regions showed that the judgment task was more sensitive to the relevance of the global context than the listening task: Only for the judgment task relevant contexts were associated with greater activity than irrelevant contexts.

Interaction between ending local consistency and task

The regions sensitive to the local integration of the story endings as a function of task included bilateral parietal areas such as the middle and inferior sections of the postcentral cortex. A large bilateral cluster included the insula and extended to SMG, although mostly on the right, where it also extended to the posterior part of STG. Medially, clusters included central CC and pericallosal sulcus. On the right, there was an additional cluster that included anterior CUN. These regions showed sensitivity to the local consistency of the endings, but only for the passive listening task. In this task, locally consistent endings were associated with greater signal change than locally inconsistent endings.

DISCUSSION

This project examined how two integration processes, consistency evaluation and passive comprehension, may differ in implementing local and global discourse coherence in the brain. On the basis of prior evidence (see Introduction), we had expected that evaluation, being a more critical and reflective type of processing, would prompt a more global approach to language processing. We expected this to be reflected in increased activation for evaluation (vs. passive comprehension) in regions associated with semantic processing and greater sensitivity to the relevance of global contexts in regions associated with episodic memory and construction of rich mental scenes. Conversely, we expected that passive comprehension would be more sensitive to local context than evaluation and show increased activation in the accumulation network, associated with fluent updating of a knowledge base.2 The results were mostly consistent with the expected pattern.

Evaluation Prompts Increased Semantic Processing and Top–Down Attention

We had expected that greater activation for the judgment tasks would be found in a network previously linked to semantic processing. This network includes IFG, posterior MFG, middle PreCG, dmPFC, AI, anterior and posterior parts of STG and STS, the middle and posterior parts of MTG, posterior ITG, TP, anterior FUS, and AG (Hagoort & Indefrey, 2014; Mar, 2011; Vigneau et al., 2006, 2011; Binder et al., 2009). Our findings for the main effect of task were indeed consistent with the prediction: Greater activation for the judgment task was found in most of the regions of this network (all except ITG, TP, and AG).3 Adjacent regions were also found with the same pattern of activation (e.g., LG, precentral sulcus, PCUN).

We also found this pattern of activity in IPS and SPL. These regions have been associated with top–down orientation of attention (Cabeza, Ciaramelli, Olson, & Moscovitch, 2008; Corbetta & Shulman, 2002). In this case, they may indicate stronger top–down attentional processing in the judgment task. Taken together, these results support the idea that semantic processing and top–down context-driven processing are more prominent in evaluative than passive comprehension.

Two clusters showed relatively greater activity for passive listening. In the left central insula, passive listening was associated with greater above-baseline activation, whereas in the left amPFC, it was associated with reduced below-baseline deactivation. These patterns suggest that passive listening and evaluation differ in other respects beyond those that can be attributed to increased semantic processing or top–down attention. The activation of the insular cluster is difficult to interpret because of the involvement of this region in multiple cognitive and autonomic functions (Craig, 2011; Menon & Uddin, 2010). However, the decreased activity during evaluative listening may be related to suspension of autonomic (parasympathetic) monitoring functions. Similarly, the stronger deactivation of the amPFC during evaluation could indicate a greater decrease in self-relevant, internal cognitive processing during the judgment task (Raichle, 2010).

Some of the regions identified by the main effect of task in our experiment have been previously identified in neuroimaging studies on inferencing and updating of mental models. It is therefore possible to argue that inferencing and mental model updating are more strongly recruited during consistency evaluation than passive comprehension. The neuroimaging literature on inferencing identifies as pivotal regions dorsolateral PFC, IFG, MTG, (Kuperberg, Lakshmanan, Caplan, & Holcomb, 2006; Mason & Just, 2004), and AG (Mason & Just, 2004, 2011) for causal inferences. In addition, IFG, STG (Virtue, Parrish, & Jung-Beeman, 2008; Virtue, Haberman, Clancy, Parrish, & Jung-Beeman, 2006), dmPFC, and MTG (Friese, Rutschmann, Raabe, & Schmalhofer, 2008) have been found to be involved in bridging inferences. Although our results are generally consistent with this literature, it is also possible that the evaluative task recruited other sorts of cognitive processes. It is therefore unclear whether the task effects we found can be truly interpreted as inferencing and updating of situation models, especially because the regions mentioned in this paragraph are not specific to inferencing or updating situation models. They have also been identified in other semantic processes during language comprehension (Hagoort & Indefrey, 2014; Mar, 2011; Vigneau et al., 2006, 2011; Binder et al., 2009).

Evaluation Prompts Increased Retrieval of Global Context and Construction of a Mental Scene

We had predicted that brain networks associated with recollection (Rugg & Vilberg, 2013; Vilberg & Rugg, 2012) and construction of a complex mental scene (Andrews-Hanna et al., 2010; Hassabis, Kumaran, & Maguire, 2007; Hassabis & Maguire, 2007) would show greater sensitivity to the relevance of the global context during the judgment task than the listening task. The regions in question are PHC, RSC, PCC, ventral PCUN, AG, and vmPFC. The network found for the interaction between task and the relevance of the global context mostly coincided with the one expected. The only difference consisted in an upward shift in temporal activity with respect with the regions reported in other studies (in STS/STG rather than MTG, although this could depend on the surface-based analysis we implemented) and in additional activity in posterior insula. Posterior insula has been associated with integration of information (although more commonly for perceptual and somatosensory information; e.g., Craig, 2011). Its pattern of activity is consistent with that of the other regions, suggesting that this region too may be involved in recollection and scene construction network.

Note also that, in discourse studies, activity found in both PCUN and PCC regions has been related to a host of similar global coherence processes. It has been interpreted as linking incoming information with activated knowledge from prior discourse to form a coherent representation (e.g., Martín-Loeches, Casado, Hernández-Tamames, & Álvarez-Linera, 2008; Maguire, Frith, & Morris, 1999; St George, Kutas, Martinez, & Sereno, 1999; Fletcher et al., 1995), with retrieval of relevant context information from long-term memory for global integration (e.g., Ferstl, Rinck, & von Cramon, 2005), and with sensitivity to narrative segmentation into event boundaries (e.g., Whitney et al., 2009; Speer, Zacks, & Reynolds, 2007). Interpretations of the specific function to attribute to these regions vary because their activation is not always found in the same network and is often interpreted post hoc. Our hypothesis is in general consistent with the interpretations from these discourse studies, as they are all based on the crucial contribution of memory processes to language comprehension. Our predictions, however, were based on results obtained with memory studies that have repeatedly and consistently tested the functioning of PCUN and PCC in concomitance with the same set of regions as a (relatively) stable network (but see also Ferstl, Neumann, Bogler, & Von Cramon, 2008, for a meta-analysis indicating some consistency on similar network findings across language studies). From the neurobiological perspective, the importance of the current data is therefore in showing that context reinstatement and construction of a rich mental model for the story are functions that do not only recruit additional regions to those usually involved in more basic language processes but are implemented by a full-fledged memory network. The contribution of the full extent of the network has never been shown (or tested) for discourse comprehension.

Furthermore, the pattern of activity we found in the regions of this network (increased signal change for relevant vs. irrelevant global contexts only for the judgment task) indicates that evaluation boosts sensitivity to the relevance of distal context. It thus supports our hypothesis and that of prior cognitive accounts on which evaluation prompts a more global approach than listening does: When the distal context is relevant for the integration of the ending, evaluation prompts context reinstatement and construction of a rich mental model for the story. This is not to say that, during passive comprehension, distal context is not reinstantiated, which would contradict much of the behavioral literature on the topic and our own prior data. In Egidi and Caramazza (2013), we have reported networks that, during passive listening, were sensitive to the relevance of the global context or to the demands of both the local and global contexts. The data reported here focus on how these processes may differ during evaluative processing or passive comprehension and suggest that evaluation prompts context reinstatement and construction of a rich mental model for the story more so than does passive comprehension.

Passive Comprehension Prompts More Fluent Integration with Local Context

We had also expected an interaction between task and the local consistency of the endings in the network for fluent accumulation of information (Egidi & Caramazza, 2013, 2014). Bilaterally, this network included dorsal PreCG, posterior SFG, middle PostCG, SMG, SPL, CC, anterior CUN, and posterior and central insula. On the left, this network includes vmPFC, posterior MTG, posterior MTS, and posterior ITG. In our reasoning, increased activation should appear for the listening task, the one least sensitive to the global demands of the story.

The results did show increased activation for locally consistent information (vs. locally inconsistent) for the listening task only. This was found in a subset of regions in the accumulation network. The regions excluded were frontal areas bilaterally, temporal areas, and CUN on the left. The involvement of the insula was also greater, as it included AI as well (bilaterally). These results are consistent with the idea that passive listening instantiates a more local approach to integration than evaluation does: When incoming information is consistent with local context, passive listening prompts its fluent integration.

Implications for Neurobiological Theories of Language and Future Language Research

Recent years have seen a rise in attention given to the methods used to study the neurobiology of language. Some have cautioned against using evaluative processing for studying language (Hickok & Poeppel, 2007; Binder, Liebenthal, Possing, Medler, & Ward, 2004; Small & Nusbaum, 2004), arguing that passive comprehension is the only genuine mode of language processing, as it is devoid of other, co-occurring and possibly confounding processes. On the other hand, several studies used tasks that require consistency evaluation to ensure that participants pay close attention to the content of the stimuli (e.g., Ferstl et al., 2005; Ferstl & von Cramon, 2001, 2002). The rationale behind this position is that evaluation guarantees focus on incoming stimuli, thus maximizing the chance that all information is processed carefully. Nonetheless, both positions appear to share the basic idea that evaluation does not introduce a qualitatively different mode of processing: It may produce activity in additional regions, thus confounding measurement, or it may modulate attention, but in any case, it leaves the basic integration process the same. Our data contribute to this debate in showing that performing a consistency evaluation on incoming information is a different operation than listening passively: Not only does the task recruit different brain regions, but it influences the implementation of core integration functions related to the construction of a consistent discourse model: Evaluation reduces the differentiation between locally consistent and inconsistent information, while boosting differentiation between distally relevant or irrelevant information. Perhaps, most importantly, we (and others, e.g., Xiang & Kuperberg, 2014) argue that one should not assume that passive comprehension is a privileged mode of language comprehension: People use different modes of comprehension on different occasions (e.g., Linderholm & van den Broek, 2002), and these do influence how comprehension and the construction of coherence are achieved. It is therefore important to not assume that there is only one type of comprehension in language—a caveat that holds for all domains of language research.

Of course, any listener or reader can choose to approach a discourse in a number of ways, which will have consequences on how the discourse is comprehended. When looking at more artificial ways of understanding language than the two examined here, the literature shows that different processing goals can lead to very different results, as we have mentioned in the Introduction. At the cognitive level, different reading purposes can lead to the generation of different inferences and to overall different comprehension (e.g., Linderholm & van den Broek, 2002; Narvaez et al., 1999). Different reading strategies can also be associated with different activations in some of the language and memory regions that we also found. For example, paraphrasing a text (vs. rereading it) is associated with greater activity in regions such as PPC, PCUN, or IFG (among others; Moss, Schunn, Schneider, McNamara, & VanLehn, 2011). These are extreme ways of processing language, which are very informative of the possibilities that our cognition allows but do not necessarily reflect everyday circumstances. As a matter of fact, theories of discourse comprehension agree that passive comprehension and consistency evaluation are crucial for the most basic type of comprehension (McNamara & Magliano, 2009), and as noted above, some researchers even consider them to be tapping into the same common process.

With respect to issues of attention, our results do not suggest that evaluation and passive comprehension differ on the amount of attention people pay to the incoming stimulus. In the behavioral data, we found no difference in response accuracy to the comprehension questions. In the fMRI data, a difference on this type of bottom–up attention would be reflected in activation differences in more ventral parietal areas than those we found (Corbetta & Shulman, 2002). It is possible, however, that the experiment did not detect bottom–up attentional differences between the two comprehension processes because participants in both groups knew that comprehension questions would appear in some trials (those with the filler stories). This knowledge might have led to an equal engagement of bottom–up attention systems in both tasks. It is noteworthy, however, that even under these conditions, these results identified a difference between evaluation and passive comprehension in regions associated with top–down, orienting attention, which reflects processes of goal-directed selection for stimuli, as demonstrated in studies on memory retrieval and visual attention (e.g., Cabeza et al., 2008; Corbetta & Shulman, 2002). In the case of discourse processing, activity in these regions is taken to reflect preparedness to the most sensible incoming information (Egidi & Caramazza, 2013). Thus, an important difference between evaluation and passive comprehension consists in the degree of context-dependent processing. It is a consequence of the amount of prior context available at the time of integration and of the use that is made of it.

Only recently have discourse studies begun asking whether consistency evaluation and passive comprehension are different processes. On the basis of ERP data, Xiang and Kuperberg (2014) have argued that consistency evaluation promotes the construction of a rich model of prior context and its use to formulate predictions. They also suggest that passive comprehension promotes drawing on general knowledge about relationships between words and concepts within a specific cognitive schema. Our data are generally consistent with this theoretical view and, in addition, make an important point: The two processes make fundamentally different uses of local and global contexts, which is reflected in the activity of different brain networks.

The current results also have implications for the design of experimental paradigms in neuroimaging studies. Prior neuroimaging studies of discourse have mostly studied integration with immediately preceding context or have not distinguished between the information just presented or presented earlier (e.g., Deen & McCarthy, 2010; Yarkoni, Speer, & Zacks, 2008; Crinion & Price, 2005; Ferstl et al., 2005; Xu, Kemeny, Park, Frattali, & Braun, 2005; Crinion, Lambon‐Ralph, Warburton, Howard, & Wise, 2003; Kansaku, Yamaura, & Kitazawa, 2000; Papathanassiou et al., 2000). Thus, future research could, at minimum, consider manipulating the distance between story endings and earlier relevant context to separate effects of local and global integration. Other neuroimaging studies have operationalized global integration by using a paradigm where seemingly unconnected text can only be understood when given prior disambiguating information (Menenti, Petersson, Scheeringa, & Hagoort, 2009; Martín-Loeches et al., 2008; Maguire et al., 1999; St George et al., 1999). Such paradigms, too, do not differentiate between local and global discourse contexts. Their interpretation also requires a contrast between sensible and nonsensical texts, which offers a less sophisticated and clear way to understand the functioning of integration than our paradigm (for more details on how our paradigm differs from those used in discourse comprehension, see Egidi & Caramazza, 2013).

Conclusion

We identified brain networks associated with two comprehension processes that differ in the degree of critical or reflective consideration of incoming information with respect to prior context. Evaluative processing, the more critical of the two processes, promoted a more global view of prior context, thus favoring relevance over proximity for integration with prior context, and was linked with activity in brain networks associated with reinstantiation of prior context and construction of a richer mental model. Passive comprehension, instead, more strongly relied on fluent integration of incoming information with the immediately preceding context, thus favoring proximity over relevance for integration of prior context, and was linked to networks previously associated with greater activity for consistent information. As both types of comprehension processes are ubiquitous in everyday language comprehension, the study of comprehension in conditions implicating both these processes is key for developing a more comprehensive understanding of the neurobiology of language processing.

Acknowledgments

We thank Ketti Mazzocco and Alessia Giovenzana for their help in preparing the stimuli. This research was partially supported by the Fondazione Cassa di Risparmio di Trento e Rovereto.

Reprint requests should be sent to Giovanna Egidi, Center for Mind/Brain Sciences (CIMeC), University of Trento, Via delle Regole 101, 38123 Trento, Italy, or via e-mail: giovanna.egidi@unitn.it.

Notes

1. 

This last result is not surprising, as the two tasks induced participants to engage in different processes. The two-way interaction for passive listening, however, is discussed in depth in Egidi and Caramazza (2013).

2. 

We identify differences by statistical tools, and our statements about qualitative differences rely on the spatial extent of the differences observed.

3. 

The activation found in FUS was posterior, however, and not, as predicted, anterior.

REFERENCES

Albrecht
,
J. E.
, &
O'Brien
,
E. J.
(
1993
).
Updating a mental model: Maintaining both local and global coherence
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
19
,
1061
1070
.
Altmann
,
U.
,
Bohrn
,
I.
,
Lubrich
,
O.
,
Menninghaus
,
W.
, &
Jacobs
,
A.
(
2014
).
Fact vs fiction—How paratextual information shapes our reading processes
.
Social Cognitive and Affective Neuroscience
,
9
,
22
29
.
Andrews-Hanna
,
J.
,
Reidler
,
J.
,
Sepulcre
,
J.
,
Poulin
,
R.
, &
Buckner
,
R.
(
2010
).
Functional-anatomic fractionation of the brain's default network
.
Neuron
,
65
,
550
562
.
Binder
,
J. R.
,
Desai
,
R.
,
Graves
,
W.
, &
Conant
,
L.
(
2009
).
Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies
.
Cerebral Cortex
,
19
,
2767
2796
.
Binder
,
J. R.
,
Liebenthal
,
E.
,
Possing
,
E. T.
,
Medler
,
D. A.
, &
Ward
,
B. D.
(
2004
).
Neural correlates of sensory and decision processes in auditory object identification
.
Nature Neuroscience
,
7
,
295
301
.
Cabeza
,
R.
,
Ciaramelli
,
E.
,
Olson
,
I. R.
, &
Moscovitch
,
M.
(
2008
).
The parietal cortex and episodic memory: An attentional account
.
Nature Reviews Neuroscience
,
9
,
613
625
.
Chow
,
H.
,
Kaup
,
B.
,
Raabe
,
M.
, &
Greenlee
,
M.
(
2008
).
Evidence of fronto-temporal interactions for strategic inference processes during language comprehension
.
Neuroimage
,
40
,
940
954
.
Cooper
,
E. A.
,
Hasson
,
U.
, &
Small
,
S. L.
(
2011
).
Interpretation-mediated changes in neural activity during language comprehension
.
Neuroimage
,
55
,
1314
1323
.
Corbetta
,
M.
, &
Shulman
,
G. L.
(
2002
).
Control of goal-directed and stimulus-driven attention in the brain
.
Nature Reviews Neuroscience
,
3
,
201
215
.
Cox
,
R. W.
(
1996
).
AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages
.
Computers and Biomedical Research
,
29
,
162
173
.
Craig
,
A. D.
(
2011
).
Significance of the insula for the evolution of human awareness of feelings from the body
.
Annals of the New York Academy of Sciences
,
1225
,
72
82
.
Crinion
,
J.
, &
Price
,
C. J.
(
2005
).
Right anterior superior temporal activation predicts auditory sentence comprehension following aphasic stroke
.
Brain
,
128
,
2858
2871
.
Crinion
,
J. T.
,
Lambon‐Ralph
,
M. A.
,
Warburton
,
E. A.
,
Howard
,
D.
, &
Wise
,
R. J. S.
(
2003
).
Temporal lobe regions engaged during normal speech comprehension
.
Brain
,
126
,
1193
1201
.
Deen
,
B.
, &
McCarthy
,
G.
(
2010
).
Reading about the actions of others: Biological motion imagery and action congruency influence brain activity
.
Neuropsychologia
,
48
,
1607
1615
.
Egidi
,
G.
, &
Caramazza
,
A.
(
2013
).
Cortical systems for local and global integration in discourse comprehension
.
Neuroimage
,
71
,
59
74
.
Egidi
,
G.
, &
Caramazza
,
A.
(
2014
).
Mood-dependent integration in discourse comprehension: Happy and sad moods affect consistency processing via different brain networks
.
Neuroimage
,
103
,
20
32
.
Egidi
,
G.
, &
Gerrig
,
R. J.
(
2006
).
Readers' experiences of characters' goals and actions
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
32
,
1322
1329
.
Egidi
,
G.
, &
Gerrig
,
R. J.
(
2009
).
How valence affects language processing: Negativity bias and mood congruence in narrative comprehension
.
Memory & Cognition
,
37
,
547
555
.
Ethofer
,
T.
,
Kreifelts
,
B.
,
Wiethoff
,
S.
,
Wolf
,
J.
,
Grodd
,
W.
,
Vuilleumier
,
P.
, et al
(
2009
).
Differential influences of emotion, task, and novelty on brain regions underlying the processing of speech melody
.
Journal of Cognitive Neuroscience
,
21
,
1255
1268
.
Ferstl
,
E. C.
,
Neumann
,
J.
,
Bogler
,
C.
, &
Von Cramon
,
D. Y.
(
2008
).
The extended language network: A meta-analysis of neuroimaging studies on text comprehension
.
Human Brain Mapping
,
29
,
581
593
.
Ferstl
,
E. C.
,
Rinck
,
M.
, &
von Cramon
,
D. Y.
(
2005
).
Emotional and temporal aspects of situation model processing during text comprehension: An event-related fMRI study
.
Journal of Cognitive Neuroscience
,
17
,
724
739
.
Ferstl
,
E. C.
, &
von Cramon
,
D. Y.
(
2001
).
The role of coherence and cohesion in text comprehension: An event-related fMRI study
.
Cognitive Brain Research
,
11
,
325
340
.
Ferstl
,
E. C.
, &
von Cramon
,
D. Y.
(
2002
).
What does the frontomedian cortex contribute to language processing: Coherence or theory of mind?
Neuroimage
,
17
,
1599
1612
.
Fischl
,
B.
,
Sereno
,
M. I.
, &
Dale
,
A. M.
(
1999
).
Cortical surface-based analysis II: Inflation, flattening, and a surface-based coordinate system
.
Neuroimage
,
9
,
195
207
.
Fletcher
,
P. C.
,
Happe
,
F.
,
Frith
,
U.
,
Baker
,
S. C.
,
Dolan
,
R. J.
,
Frackowiak
,
R. S.
, et al
(
1995
).
Other minds in the brain: A functional imaging study of “theory of mind” in story comprehension
.
Cognition
,
57
,
109
128
.
Forman
,
S. D.
,
Cohen
,
J. D.
,
Fitzgerald
,
M.
,
Eddy
,
W. F.
,
Mintun
,
M. A.
, &
Noll
,
D. C.
(
1995
).
Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): Use of a cluster-size threshold
.
Magnetic Resonance in Medicine
,
33
,
636
647
.
Friese
,
U.
,
Rutschmann
,
R.
,
Raabe
,
M.
, &
Schmalhofer
,
F.
(
2008
).
Neural correlates of inference processes in text comprehension: An event-related functional magnetic resonance imaging study
.
Journal of Cognitive Neuroscience
,
20
,
2110
2124
.
Graves
,
W.
,
Binder
,
J.
,
Desai
,
R.
,
Conant
,
L.
, &
Seidenberg
,
M.
(
2010
).
Neural correlates of implicit and explicit combinatorial semantic processing
.
Neuroimage
,
53
,
638
646
.
Hagoort
,
P.
, &
Indefrey
,
P.
(
2014
).
The neurobiology of language beyond single words
.
Annual Review of Neuroscience
,
37
,
347
362
.
Hassabis
,
D.
,
Kumaran
,
D.
, &
Maguire
,
E.
(
2007
).
Using imagination to understand the neural basis of episodic memory
.
Journal of Neuroscience
,
27
,
14365
14374
.
Hassabis
,
D.
, &
Maguire
,
E.
(
2007
).
Deconstructing episodic memory with construction
.
Trends in Cognitive Sciences
,
11
,
299
306
.
Hasson
,
U.
,
Nusbaum
,
H. C.
, &
Small
,
S. L.
(
2006
).
Repetition suppression for spoken sentences and the effect of task demands
.
Journal of Cognitive Neuroscience
,
18
,
2013
2029
.
Hasson
,
U.
,
Nusbaum
,
H. C.
, &
Small
,
S. L.
(
2007
).
Brain networks subserving the extraction of sentence information and its encoding to memory
.
Cerebral Cortex
,
17
,
2899
2913
.
Hickok
,
G.
, &
Poeppel
,
D.
(
2007
).
The cortical organization of speech processing
.
Nature Reviews Neuroscience
,
8
,
393
402
.
Huang
,
J.
,
Zhu
,
Z.
,
Zhang
,
J.
,
Wu
,
M.
,
Chen
,
H.-C.
, &
Wang
,
S.
(
2012
).
The role of left inferior frontal gyrus in explicit and implicit semantic processing
.
Brain Research
,
1440
,
56
64
.
Kansaku
,
K.
,
Yamaura
,
A.
, &
Kitazawa
,
S.
(
2000
).
Sex differences in lateralization revealed in the posterior language areas
.
Cerebral Cortex
,
10
,
866
872
.
Kintsch
,
W.
(
1988
).
The role of knowledge in discourse comprehension: A construction–integration model
.
Psychological Review
,
95
,
163
182
.
Kintsch
,
W.
(
1998
).
Comprehension: A paradigm for cognition
.
New York
:
Cambridge University Press
.
Kuperberg
,
G.
,
Lakshmanan
,
B.
,
Greve
,
D.
, &
West
,
W.
(
2008
).
Task and semantic relationship influence both the polarity and localization of hemodynamic modulation during lexico-semantic processing
.
Human Brain Mapping
,
29
,
544
561
.
Kuperberg
,
G. R.
,
Lakshmanan
,
B. M.
,
Caplan
,
D. N.
, &
Holcomb
,
P. J.
(
2006
).
Making sense of discourse: An fMRI study of causal inferencing across sentences
.
Neuroimage
,
33
,
343
361
.
Linderholm
,
T.
, &
van den Broek
,
P.
(
2002
).
The effects of reading purpose and working memory capacity on the processing of expository text
.
Journal of Educational Psychology
,
94
,
778
.
Loftus
,
G. R.
, &
Masson
,
M. E.
(
1994
).
Using confidence intervals in within-subject designs
.
Psychonomics Bulletin & Review
,
1
,
476
490
.
Maguire
,
E. A.
,
Frith
,
C. D.
, &
Morris
,
R.
(
1999
).
The functional neuroanatomy of comprehension and memory: The importance of prior knowledge
.
Brain
,
122
,
1839
1850
.
Mar
,
R. A.
(
2011
).
The neural bases of social cognition and story comprehension
.
Annual Review of Psychology
,
62
,
103
134
.
Martín-Loeches
,
M.
,
Casado
,
P.
,
Hernández-Tamames
,
J. A.
, &
Álvarez-Linera
,
J.
(
2008
).
Brain activation in discourse comprehension: A 3T fMRI study
.
Neuroimage
,
41
,
614
622
.
Mason
,
R. A.
, &
Just
,
M. A.
(
2004
).
How the brain processes causal inferences in text
.
Psychological Science
,
15
,
1
7
.
Mason
,
R. A.
, &
Just
,
M. A.
(
2011
).
Differentiable cortical networks for inferences concerning people's intentions versus physical causality
.
Human Brain Mapping
,
32
,
313
329
.
McKoon
,
G.
, &
Ratcliff
,
R.
(
1992
).
Inference during reading
.
Psychological Review
,
99
,
440
466
.
McNamara
,
D. S.
, &
Magliano
,
J.
(
2009
).
Toward a comprehensive model of comprehension
.
Psychology of Learning and Motivation
,
51
,
297
384
.
Menenti
,
L.
,
Petersson
,
K. M.
,
Scheeringa
,
R.
, &
Hagoort
,
P.
(
2009
).
When elephants fly: Differential sensitivity of right and left inferior frontal gyri to discourse and world knowledge
.
Journal of Cognitive Neuroscience
,
21
,
2358
2368
.
Menon
,
V.
, &
Uddin
,
L. Q.
(
2010
).
Saliency, switching, attention and control: A network model of insula function
.
Brain Structure & Function
,
214
,
655
667
.
Moss
,
J.
,
Schunn
,
C. D.
,
Schneider
,
W.
,
McNamara
,
D. S.
, &
VanLehn
,
K.
(
2011
).
The neural correlates of strategic reading comprehension: Cognitive control and discourse comprehension
.
Neuroimage
,
58
,
675
686
.
Myers
,
J. L.
,
O'Brien
,
E. J.
,
Albrecht
,
J. E.
, &
Mason
,
R. A.
(
1994
).
Maintaining global coherence during reading
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
20
,
876
886
.
Narvaez
,
D.
,
Van Den Broek
,
P.
, &
Ruiz
,
A.
(
1999
).
The influence of reading purpose on inference generation and comprehension in reading
.
Journal of Educational Psychology
,
91
,
488
.
Noesselt
,
T.
,
Shah
,
N. J.
, &
Jäncke
,
L.
(
2003
).
Top–down and bottom–up modulation of language related areas—An fMRI study
.
BMC Neuroscience
,
4
,
13
.
Papathanassiou
,
D.
,
Etard
,
O.
,
Mellet
,
E.
,
Zago
,
L.
,
Mazoyer
,
B.
, &
Tzourio-Mazoyer
,
N.
(
2000
).
A common language network for comprehension and production: A contribution to the definition of language epicenters with PET
.
Neuroimage
,
11
,
347
357
.
Perani
,
D.
,
Dehaene
,
S.
,
Grassi
,
F.
,
Cohen
,
L.
,
Cappa
,
S. F.
,
Dupoux
,
E.
, et al
(
1996
).
Brain processing of native and foreign languages
.
NeuroReport
,
7
,
2439
2444
.
Perani
,
D.
,
Paulesu
,
E.
,
Galles
,
N. S.
,
Dupoux
,
E.
,
Dehaene
,
S.
,
Bettinardi
,
V.
, et al
(
1998
).
The bilingual brain. Proficiency and age of acquisition of the second language
.
Brain
,
121
,
1841
.
Raichle
,
M. E.
(
2010
).
Two views of brain function
.
Trends in Cognitive Sciences
,
14
,
180
190
.
Rapp
,
D. N.
, &
Gerrig
,
R. J.
(
2006
).
Predilections for narrative outcomes: The impact of story contexts and reader preferences
.
Journal of Memory and Language
,
54
,
54
67
.
Rapp
,
D. N.
, &
Kendeou
,
P.
(
2007
).
Revising what readers know: Updating text representations during narrative comprehension
.
Memory & Cognition
,
35
,
2019
2032
.
Rodd
,
J. M.
,
Davis
,
M. H.
, &
Johnsrude
,
I. S.
(
2005
).
The neural mechanisms of speech comprehension: fMRI studies of semantic ambiguity
.
Cerebral Cortex
,
15
,
1261
1269
.
Rugg
,
M. D.
, &
Vilberg
,
K. L.
(
2013
).
Brain networks underlying episodic memory retrieval
.
Current Opinion in Neurobiology
,
23
,
255
260
.
Saad
,
Z. S.
,
Glen
,
D. R.
,
Chen
,
G.
,
Beauchamp
,
M. S.
,
Desai
,
R.
, &
Cox
,
R. W.
(
2009
).
A new method for improving functional-to-structural MRI alignment using local Pearson correlation
.
Neuroimage
,
44
,
839
848
.
Small
,
S. L.
, &
Nusbaum
,
H. C.
(
2004
).
On the neurobiological investigation of language understanding in context
.
Brain and Language
,
89
,
300
311
.
Speer
,
N. K.
,
Zacks
,
J. M.
, &
Reynolds
,
J. R.
(
2007
).
Human brain activity time-locked to narrative event boundaries
.
Psychological Science
,
18
,
449
455
.
St George
,
M.
,
Kutas
,
M.
,
Martinez
,
A.
, &
Sereno
,
M.
(
1999
).
Semantic integration in reading: Engagement of the right hemisphere during discourse processing
.
Brain
,
122
,
1317
1325
.
Suzuki
,
K.
, &
Sakai
,
K. L.
(
2003
).
An event-related fMRI study of explicit syntactic processing of normal/anomalous sentences in contrast to implicit syntactic processing
.
Cerebral Cortex
,
13
,
517
526
.
Tzeng
,
Y.
,
van den Broek
,
P.
,
Kendeou
,
P.
, &
Lee
,
C.
(
2005
).
The computational implementation of the landscape model: Modeling inferential processes and memory representations of text comprehension
.
Behavior Research Methods
,
37
,
277
286
.
Vigneau
,
M.
,
Beaucousin
,
V.
,
Herve
,
P.-Y.
,
Duffau
,
H.
,
Crivello
,
F.
,
Houde
,
O.
, et al
(
2006
).
Meta-analyzing left hemisphere language areas: Phonology, semantics, and sentence processing
.
Neuroimage
,
30
,
1414
1432
.
Vigneau
,
M.
,
Beaucousin
,
V.
,
Hervé
,
P.-Y.
,
Jobard
,
G.
,
Petit
,
L.
,
Crivello
,
F.
, et al
(
2011
).
What is right-hemisphere contribution to phonological, lexico-semantic, and sentence processing?: Insights from a meta-analysis
.
Neuroimage
,
54
,
577
593
.
Vilberg
,
K. L.
, &
Rugg
,
M. D.
(
2012
).
The neural correlates of recollection: Transient versus sustained fMRI effects
.
Journal of Neuroscience
,
32
,
15679
15687
.
Vingerhoets
,
G.
,
van Borsel
,
J.
,
Tesink
,
C.
,
van den Noort
,
M.
,
Deblaere
,
K.
,
Seurinck
,
R.
, et al
(
2003
).
Multilingualism: An fMRI study
.
Neuroimage
,
20
,
2181
2196
.
Virtue
,
S.
,
Haberman
,
J.
,
Clancy
,
Z.
,
Parrish
,
T.
, &
Jung-Beeman
,
M.
(
2006
).
Neural activity of inferences during story comprehension
.
Brain Research
,
1084
,
104
114
.
Virtue
,
S.
,
Parrish
,
T.
, &
Jung-Beeman
,
M.
(
2008
).
Inferences during story comprehension: Cortical recruitment affected by predictability of events and working-memory capacity
.
Journal of Cognitive Neuroscience
,
20
,
2274
2284
.
Whitney
,
C.
,
Huber
,
W.
,
Klann
,
J.
,
Weis
,
S.
,
Krach
,
S.
, &
Kircher
,
T.
(
2009
).
Neural correlates of narrative shifts during auditory story comprehension
.
Neuroimage
,
47
,
360
366
.
Xiang
,
M.
, &
Kuperberg
,
G.
(
2014
).
Reversing expectations during discourse comprehension
.
Language, Cognition and Neuroscience
,
30
,
648
672
.
Xu
,
J.
,
Kemeny
,
S.
,
Park
,
G.
,
Frattali
,
C.
, &
Braun
,
A.
(
2005
).
Language in context: Emergent features of word, sentence, and narrative comprehension
.
Neuroimage
,
25
,
1002
1015
.
Yarkoni
,
T.
,
Speer
,
N. K.
, &
Zacks
,
J. M.
(
2008
).
Neural substrates of narrative comprehension and memory
.
Neuroimage
,
41
,
1408
1425
.
Yue
,
Q.
,
Zhang
,
L.
,
Xu
,
G.
,
Shu
,
H.
, &
Li
,
P.
(
2013
).
Task-modulated activation and functional connectivity of the temporal and frontal areas during speech comprehension
.
Neuroscience
,
237
,
87
95
.
Zaitsev
,
M.
,
Hennig
,
J.
, &
Speck
,
O.
(
2004
).
Point spread function mapping with parallel imaging techniques and high acceleration factors: Fast, robust, and flexible method for echo-planar imaging distortion correction
.
Magnetic Resonance in Medicine
,
52
,
1156
1166
.
Zhu
,
Z.
,
Hagoort
,
P.
,
Zhang
,
J.
,
Feng
,
G.
,
Chen
,
H.-C.
,
Bastiaansen
,
M.
, et al
(
2012
).
The anterior left inferior frontal gyrus contributes to semantic unification
.
Neuroimage
,
60
,
2230
2237
.