RESEARCH DOSSIER
Discourse and Narrative Comprehension
Abstract. Discourse and narrative comprehension transform sequences of words, sentences, and events into coherent representations of meaning. This fundamental research monograph examines the history of text-comprehension theories, propositional coherence, construction–integration and structure-building models, situation-model construction, causal and referential inferences, event indexing and boundaries, memory retrieval, narrative mentalizing, multimodal stories, neural temporal integration, and developmental variation. Original experimental paradigms are evaluated alongside contemporary neuroimaging and event-memory studies, with attention to alternative explanations, measurement validity, and generalizability. The analysis distinguishes online inference from post-reading reconstruction, situation-model representation from literal recall, and neural correlates from causal mechanisms. The study is a critical narrative synthesis of published research, with no original experiments or claims of a newly validated theory.
Keywords: discourse comprehension; narrative comprehension; situation models; text comprehension; construction–integration; event indexing; causal inference; referential coherence; textbase; mental models; event boundaries; reading comprehension; narrative memory; event segmentation; working memory; narrative neural systems; narrative transportation; development of discourse comprehension; computational discourse models; Artificial Era
Introduction: Connected Language as a Psychological Object
Discourse comprehension is the construction and maintenance of meaning across connected utterances; narrative comprehension is its event-centered form, in which an interpreter organizes agents, settings, actions, consequences, and temporal change into an intelligible account. The central empirical problem is how separately encountered words and sentences yield an integrated representation that supports inferences, explanations, recollection, and anticipation. Comprehension includes retrieving prior information, reconciling newly expressed material with what is already represented, and constructing a model of what the discourse concerns. A coherent interpretation can preserve relations that were never stated as a complete sentence. This makes extended-language comprehension a distinct research object from recognizing a word or computing the syntax of an isolated clause. Its experimental signatures include selective reading times, delayed recognition, referential accessibility, event segmentation, and the organization of subsequent memory Kintsch & van Dijk (1978) Zwaan & Radvansky (1998) Graesser et al. (1994).
The inquiry covers spoken and written discourse, ordinary stories, connected explanatory prose where its mechanisms illuminate narratives, and the relations between sentence-level signals and event-level representations. It examines what a text explicitly specifies, what background knowledge contributes, which inferences become accessible during reading, and how current goals change the depth of representation. The same narrative may permit a surface-form memory, a propositional account of stated relations, and a situation model tracking events in an imagined or described world. These products can dissociate experimentally, so the ability to repeat words is an insufficient index of understanding. The field accordingly treats comprehension as a sequence of measurable transformations rather than as one undifferentiated verbal score Kintsch (1988) Morrow et al. (1987) Glenberg et al. (1987).
This monograph is a critical, source-based synthesis of published psychological and cognitive-neuroscientific research. It draws on original comprehension experiments, model-building articles, reviews, neuroimaging investigations, and historically important monographs, while retaining distinctions between theory, measurement, association, and causal evidence. It does not report newly conducted experiments or claim to be a protocol-driven systematic review. Its organizing questions concern the representational contents of discourse, the dynamics of coherence and inference, the role of event boundaries, the time scales of neural processing, individual and communicative variation, and the limits of current explanatory models. The discussion is grounded in human evidence; computational and artificial systems are considered as testable comparators where their behavior can be evaluated on matched materials Gernsbacher (1990) Ferstl et al. (2008) Baldassano et al. (2017).
Research Scope, Operational Definitions, and Levels of Representation
Discourse denotes connected language functioning within a context of production and interpretation. A narrative is discourse that characteristically describes events and their relationships, frequently through temporally ordered actions, changing circumstances, agents, and purposes. Local coherence concerns bridges between nearby statements; global coherence concerns an intelligible organization of the entire passage. Referential continuity is the preservation of an entity across names, pronouns, descriptions, and shifts of focus. Causal coherence concerns explanatory relations between events, while temporal coherence concerns their ordering and duration. A situation model is an integrated representation of the state of affairs described, constructed through linguistic input and background knowledge. These terms identify different empirical questions and are not interchangeable performance measures Zwaan & Radvansky (1998) Graesser et al. (1994) Morrow et al. (1987).
Three analytical levels are useful. Surface representation preserves relatively literal linguistic form; a textbase codes propositions and their explicit or inferred relations; a situation model represents events, agents, spatial relations, goals, and other features of the described domain. This distinction is theoretical and operational: different recognition probes and memory tasks may disproportionately sample different representational resources, and individual tasks are seldom pure measures of one level. A reader can remember a description while losing the scene configuration, or retain a causal storyline while forgetting sentence wording. Evidence for a situation model arises when performance depends on the narrated situation even after lexical or propositional overlap is controlled Kintsch & van Dijk (1978) Glenberg et al. (1987) Bower & Morrow (1990).
An experimental contrast should isolate the hypothesized mechanism. A pronoun probe can test the accessibility of an antecedent, but response time also reflects lexical frequency and task strategy. A reading-time increase after a temporal discontinuity can indicate model updating, but also variation in sentence length, unpredictability, or semantic complexity. An fMRI pattern at an event boundary indexes a measured hemodynamic correlate under a particular analysis, not a complete mechanism of comprehension. Strong conclusions therefore triangulate across text manipulations, behavioral signatures, delayed memory, and neural results while reporting what each method can actually identify. Measurement validity depends on distinguishing an online process from a later decision about the passage Zwaan et al. (1995) Speer et al. (2007) Baldassano et al. (2017).
Historical Development: From Text Memory to Situation Representation
Experimental studies of connected language helped establish that context organizes what can be understood and later recalled. Bransford and Johnson's 1972 experiments manipulated whether clarifying contextual information was supplied before or after difficult prose passages. Providing appropriate context before comprehension improved participants' judgments and recall more effectively than providing it after exposure. The result demonstrated an experimentally manipulable contribution of prior interpretive organization, while leaving open which knowledge structures were activated and how they interacted with incoming propositions. It also warned against evaluating reading in isolation from a person's prior knowledge. Subsequent accounts sought mechanisms that could explain why materially similar sentences yield different memories when their situational settings change Bransford & Johnson (1972) Kintsch & van Dijk (1978).
Kintsch and van Dijk's 1978 model offered a systematic treatment of local propositional coherence, processing cycles, working-memory limitations, and macrostructures that condense a text into an organized gist. Kintsch's 1988 construction–integration model then framed comprehension as initial activation of multiple candidate meanings and knowledge relations followed by integration in a network in which compatible relations gain strength. The two contributions addressed distinct problems: how meaning-bearing units can be organized under resource constraints, and how knowledge can become integrated without requiring a perfect contextual prediction at every step. These formulations became experimentally influential because they specify intervening representational states rather than equating comprehension with a final test score Kintsch & van Dijk (1978) Kintsch (1988).
Work on mental and situation models extended the empirical target from textual propositions to described environments. Morrow, Greenspan, and Bower examined the changing accessibility of objects while participants read about movement through a memorized building. Glenberg, Meyer, and Lindem manipulated whether a narrated actor remained spatially associated with an object while keeping the textual propositions comparable. Such studies indicated that readers update accessibility in ways sensitive to described scenes. Later historical developments connected these results to event indexing, inference theories, psychological structure building, and cognitive neuroscience. The resulting research tradition is cumulative but plural: different models identify different mechanisms and measurement regimes, and their claims remain open to experimental discrimination Morrow et al. (1987) Glenberg et al. (1987) Zwaan et al. (1995).
From Sentences to a Coherent Textbase
Understanding a passage requires recognizing how one statement contributes to the preceding discourse. A sentence introduces predications about entities, relations, or events; later material may preserve an argument, elaborate a circumstance, or change the topic. Propositional overlap can support connections when the same entity or relation recurs, but overlap alone does not guarantee that readers understand the narrative's causal and temporal organization. Conversely, situational continuity can remain strong across sentences using different lexical expressions. Experiments using naturalistic stories have measured reading-time consequences of changes in time, space, and cause, showing that comprehension monitors more than simple repetition of words Kintsch & van Dijk (1978) Zwaan et al. (1995).
Macrostructure organizes information over longer spans. Readers may distinguish central developments from incidental description, maintain an overarching objective, and summarize a sequence according to its explanatory importance. The operations of selection, generalization, and integration can be modeled as transformations of local material into a higher-level representation, but the chosen gist depends partly on the reader's task and knowledge. Retellings and summaries reveal this organization through what survives compression. However, a good summary may draw on prior knowledge or inference, making it necessary to separate faithful representation of the source from a plausible reconstruction. The original text-comprehension model explicitly linked comprehension to the generation of recall and summarization protocols Kintsch & van Dijk (1978) Kintsch (1988).
The construction–integration framework permits several initially accessible interpretations and evaluates their fit within a network of propositions and knowledge. A local sentence may support a transient association that is subsequently suppressed when broader discourse makes it inappropriate. Structure-building theory likewise treats comprehension as founding an initial mental structure, mapping compatible input onto it, and shifting to additional structures when mapping fails. These accounts generate different emphases on activation, suppression, shifting, and memory access. Their most informative tests manipulate the availability of competing material and measure processing over time, rather than asking participants only for a final true–false judgment Kintsch (1988) Gernsbacher (1990) Gernsbacher et al. (1990).
Situational Continuity in Time and Space
Readers treat described time and space as organized dimensions of a continuing situation. In the event-indexing framework, temporal and spatial relations join protagonist, causality, and intentionality in representing events across a story. A temporal jump can require readers to reposition the developing model; a spatial transition can change which locations or objects are accessible. Zwaan, Magliano, and Graesser manipulated continuity in naturalistic narratives and observed reading-time patterns associated especially with temporal and causal changes. These results support ongoing multidimensional monitoring, although a discontinuity's processing cost can also reflect surprise, linguistic elaboration, or unequal prior familiarity. The theoretical claim is strongest when discontinuity effects persist after relevant sentence-level differences have been controlled Zwaan et al. (1995) Zwaan et al. (1995) Zwaan & Radvansky (1998).
Spatial accessibility experiments provide a more concrete test of represented rather than merely stated relations. Morrow, Greenspan, and Bower asked readers to follow a character moving through a known building and probed spatial information during the narrative. The availability of information varied with the person's current position and the structure of the described environment. Glenberg, Meyer, and Lindem contrasted stories where an actor remained near or moved away from an object, using recognition and reading-time measures. Both paradigms indicate that a reader can keep track of objects as parts of a simulated scenario. Nevertheless, these studies often employ unusually clear spatial settings, and generalization to abstract, metaphorical, or socially complex narratives requires separate evidence Morrow et al. (1987) Glenberg et al. (1987) Bower & Morrow (1990).
Temporal structure is influenced by grammar and world knowledge. Magliano and Schleich compared events portrayed as ongoing with those marked as completed; in their experiments, activities described with imperfective aspect were more likely to remain active across subsequent sentences than completed activities. Their findings demonstrate that tense–aspect information contributes to the maintenance of event representations, interacting with knowledge about how long an activity normally lasts. This supplies a bridge between local morphosyntactic cues and extended situation models. It also shows why comprehension cannot be inferred from discourse continuity alone: seemingly minor linguistic choices influence how an event is situated across a narrative's timeline. Future work must separate habitual timing conventions from task-dependent simulation Magliano & Schleich (2000) Zwaan & Radvansky (1998).
Causal Coherence, Goals, and Explanation
Causal comprehension connects actions and changes through explanations. A narrative may state that a character opened an umbrella and later remained dry; readers can integrate these propositions by recruiting knowledge about rainfall and protective objects. Such integration is affected by the strength of world knowledge, the number of intervening events, and whether an unstated bridge is necessary for local coherence. Van den Broek's work on causal inferences sought to specify the relationships supporting a coherent causal account. Constructionist approaches likewise emphasize efforts to explain why a described event occurred and how it relates to a protagonist's aims. These are models of inference generation, not evidence that every possible causal relation is immediately represented van den Broek (1990) Graesser et al. (1994).
Goals give narrative actions functional organization. The protagonist's movements often become intelligible only because the reader infers an intention, considers alternative routes, or updates an expectation when an attempted action fails. In event-indexing theory intentionality is a distinct dimension along which represented events can be linked. Studies of mental models also show that readers maintain relevant objects and locations as characters pursue goals. A cause, a reason, and a goal must nevertheless be kept analytically separate: physical causation, social explanation, and deliberate planning can support different predictions about what will follow. Evaluating intentional coherence requires text conditions in which motive information is independently varied rather than merely correlated with temporal adjacency Zwaan et al. (1995) Morrow et al. (1987) Fletcher et al. (1995).
Coherence can be maintained by integrating propositions in memory, elaborating the situation model, or revising an earlier interpretation. These processes may occur at different times and demand different amounts of attention. In experiments, a predictable action may be processed quickly because relevant relations are already active, while a surprising action can produce a longer reading time and stronger subsequent recall. Such a cost could index successful updating as well as difficulty; performance must therefore be interpreted against a memory or inference measure. A single narrative does not have one inevitable causal parse: the reader's goals, cultural knowledge, and beliefs about agency influence which links are sought and whether ambiguity remains. An adequate theory must account for both coherent understanding and plausible alternative readings Graesser et al. (1994) McKoon & Ratcliff (1992) Kintsch (1988).
Reference Tracking, Accessibility, and Discourse Focus
Extended discourse depends on identifying which expressions refer to the same individual, object, place, or idea. Names, definite descriptions, pronouns, and ellipses carry different degrees of explicitness and require the reader to retrieve information from an evolving context. A referent can remain mentally prominent because it has been recently mentioned, occupies the focus of action, is spatially near the protagonist, or explains the current event. Retrieval accessibility is consequently a joint product of language and the constructed situation. In a spatial narrative, the most easily recognized object need not be the most recently named object; modeled distance and action context can also influence availability Morrow et al. (1987) Glenberg et al. (1987) Zwaan & Radvansky (1998).
Reference resolution is a critical point of contact between linguistic processing and memory. An ambiguous pronoun can recruit multiple candidates; a subsequent phrase may disambiguate and force a revision of the reader's representation. A story with several protagonists imposes additional demands on maintaining distinct people, perspectives, and relations. Experimental probes that present a proper name or object label after a narrative segment can reveal accessibility, but probe recognition combines comprehension with retrieval and response selection. A strong design varies discourse position, reference expression, and situational focus separately, and uses control probes to estimate a general speed of recognition. These controls are indispensable when comparing individuals or text conditions with different baseline difficulties Gernsbacher et al. (1990) Morrow et al. (1987) Kintsch (1988).
Conversational reference adds a history of jointly established descriptions. Brennan and Clark's experiments on repeated reference support partner-specific conceptual agreements that are negotiated and then reused. This finding helps explain how a listener can recover a referent from expressions that depend on prior interaction, yet such collaborative common ground is a neighboring mechanism rather than the whole of narrative comprehension. A written novel can also represent multiple perspectives without real-time negotiation with its reader. The present dossier incorporates referential coordination insofar as it illuminates the interpretation of connected discourse, while the larger science of conversational pragmatics retains its own research questions about speaker beliefs and communicative choices Brennan & Clark (1996) Graesser et al. (1994).
Inference During Reading: Automaticity and Reader Goals
Inference is the establishment of information that follows from, connects, or enriches expressed material under available knowledge and assumptions. Bridging inferences connect adjacent clauses; elaborative inferences add plausible but unnecessary detail; predictive inferences project an event beyond what has already been stated. These classes differ in how strongly they are licensed and when they become accessible. The 1992 minimalist hypothesis by McKoon and Ratcliff proposes that automatic inferences draw primarily on easily available information or restore local coherence, while more elaborate interpretation can depend on reader purposes. Graesser, Singer, and Trabasso's 1994 constructionist theory emphasizes search after meaning and explanation of narrative events. These are substantive rival accounts of the circumstances in which readers generate knowledge-based inferences McKoon & Ratcliff (1992) Graesser et al. (1994).
Apparent disagreements between inference theories can partly reflect experimental tasks. An immediate recognition probe following a sentence may sample different information than a retrospective question asked after the entire story. Instructions to read for enjoyment, evaluate a protagonist, identify errors, or remember causal relations change which inferences are useful. Moreover, a fast response to a probe is not conclusive evidence that an inference was uniquely encoded during the original sentence; retrieval from general knowledge at test can produce similar outcomes. Experimental interpretation therefore depends on carefully timed measures, control probes, and manipulation of causal necessity and knowledge availability. A rigorous synthesis retains the distinction between the possibility of an inference and demonstrable online construction McKoon & Ratcliff (1992) Graesser et al. (1994) Kintsch (1988).
An empirically useful integrated account anticipates graded inference accessibility. Strong local bridges are likely to be recruited across ordinary reading circumstances; distal causal explanations vary with personal knowledge and task goals; richly imagined scenes recruit further elaborations when the narrative supports sustained attention. These are predictions about conditions, not proof of a single universal mechanism. A factorial experiment could vary required coherence, inference distance, and reading purpose while measuring immediate processing and delayed memory. Support for one theory would require a distinctive pattern that alternative accounts cannot explain through retrieval or strategic decision-making. The relevant evidence favors treating inference generation as a family of processes with multiple temporal scales and constraints Graesser et al. (1994) McKoon & Ratcliff (1992) Gernsbacher et al. (1990).
Narrative Event Models and Event Indexing
Situation models develop through a sequence of events that become connected in memory. Zwaan, Langston, and Graesser proposed that narrative events are indexed along time, space, protagonist, causality, and intentionality. The event-indexing proposal provides a structured prediction: discontinuity along a dimension should alter the activation and integration of incoming information, with effects depending on the current representation and available cues. The original 1995 report tested clustering of narrative verbs and interpreted the results as support for multidimensional event representation. Related reading-time studies provide a complementary measure, especially for temporal and causal continuity. The theory has explanatory strength because it specifies meaningful dimensions of change; the relative importance of each dimension remains dependent on genre, task, and the realism of the described situation Zwaan et al. (1995) Zwaan et al. (1995) Zwaan & Radvansky (1998).
Events may be nested. An entire journey is one episode, a visit to a station is a constituent episode, and buying a ticket contains a series of actions. Nested organization helps explain why a story can be summarized at several levels without becoming arbitrary. The reader can monitor a current fine-grained action while retaining a higher-level goal and the broader episode's trajectory. Event segmentation research in perception and memory reports hierarchical segmentation over multiple temporal scales; narrative comprehension adds linguistic instructions and referential relationships to this problem. Whether an event boundary is primarily triggered by a shift in location, time, goal, prediction, or their combination is an empirical question Kurby & Zacks (2008) Zacks et al. (2007) Baldassano et al. (2017).
Event models also generate selective remembering. When a new episode begins, some prior information becomes less immediately accessible because the active model changes; other information may be stabilized through distinctive event organization. The Event Horizon perspective treats boundaries as consequential for encoding, accessibility, and later retrieval, and cautions that benefits and costs depend on the materials and test. A boundary is therefore neither a general memory eraser nor a guaranteed mnemonic benefit. Studies must specify whether they measure immediate access to a previous object, recognition of within-event associations, order memory, or freely reconstructed event sequences. These outcomes can diverge without logical contradiction Radvansky (2012) Kurby & Zacks (2008) Baldassano et al. (2017).
Prediction, Updating, and Event Boundaries
At an event boundary the comprehender encounters cues that the current representation has become insufficient for the developing situation. Event segmentation theory proposes that ongoing predictions about near-future activity are monitored and that increases in prediction error can signal transitions. This interpretation does not require a reader to forecast every word; the relevant expectations can concern event continuations and action structure. In narrative discourse an unexpectedly changed goal, entrance into a new room, or unexplained temporal transition may produce the conditions for model updating. The prediction-error framework is useful because it links boundary detection to measurable patterns of transition, but specific sources of prediction error must be experimentally identified Zacks et al. (2007) Reynolds et al. (2007).
Speer, Zacks, and Reynolds recorded fMRI while participants read everyday-activity narratives and later segmented those stories into events. Brain responses during the initial reading were related to the boundaries participants subsequently identified; changes in described circumstances, including goals, helped account for activation. The study connects a behavioral event-segmentation judgment with an online neural measurement, strengthening the case that event structure is represented during comprehension. It does not demonstrate that every region activated at a boundary computes the same prediction error or that event segmentation is wholly automatic in every task. Hemodynamic timing, the retrospective segmentation procedure, and the limited set of stories are relevant interpretive constraints Speer et al. (2007) Zacks et al. (2007).
Reynolds, Zacks, and Braver developed a computational account of segmentation based on perceptual prediction. Its importance is methodological as well as theoretical: a proposal about event boundaries can be expressed in terms of simulated processing states and tested against the distribution of behavioral boundary choices. To distinguish such a model from alternative interpretations, researchers must vary predictive uncertainty without simply altering semantic difficulty or the number of scene changes. Natural stories offer ecological richness, whereas controlled stimuli can isolate which changes produce a boundary. A productive evidence program compares both and reports parameter sensitivity rather than fitting an explanation to each observed transition after the fact Reynolds et al. (2007) Kurby & Zacks (2008) Speer et al. (2007).
Working Memory, Retrieval, and the Changing Narrative Focus
Comprehension entails a changing balance between currently available content and material that must be recovered from longer-term memory. Working-memory constraints matter when a reader needs to connect a new sentence to a character introduced many paragraphs earlier or reinterpret an event after a delayed explanation. The 1978 text-processing model explicitly incorporated limited working-memory resources and repeated cycles of proposition processing. The construction–integration model elaborated a possible mechanism through activation and settling of relevant networks. These theories make no requirement that an entire narrative be held simultaneously as fully activated propositions. Instead, a compact representation can guide retrieval of details when subsequent language makes them relevant Kintsch & van Dijk (1978) Kintsch (1988).
Structure-building research predicts that establishing too many poorly integrated mental substructures can compromise comprehension. Gernsbacher, Varner, and Faust investigated general comprehension skill using written, spoken, and picture stories as well as memory-access and suppression tasks. Their results link skill differences to accessibility and the management of irrelevant information, while the correlational and task-specific features of such measures leave questions about mechanisms and causal direction. The same participant can perform differently with unfamiliar material, complex referential shifts, or different reading purposes. A complete account should distinguish a stable individual tendency from momentary mismatch between text demands and available knowledge Gernsbacher et al. (1990) Gernsbacher (1990).
Event boundaries help organize the transition between active maintenance and retrieval. An unfolding episode maintains immediate focus on actions and persons; a new episode may cause some prior information to recede while preserving an organized account of what has occurred. Retrieval can reactivate a formerly prominent person or event when it becomes causally necessary. The distribution of attention over time thus determines which details appear available in online probes, but delayed memory depends additionally on consolidation, cue quality, interference, and reconstruction. Without separate time points and retrieval formats, the claim that a particular event was forgotten is stronger than most experiments justify. This distinction is central when evaluating the merits of continuous versus segmented discourse Radvansky (2012) Kurby & Zacks (2008) Morrow et al. (1987).
Discourse Comprehension across Modalities and Formats
Written sentences, spoken stories, illustrated sequences, and audiovisual narratives supply different perceptual input while sharing the problem of relating events across time. Linguistic discourse specifies many relations explicitly, whereas pictures and films may require viewers to infer action transitions and motives from perceptual continuity. A shared situation-model account predicts partial convergence in event interpretation across these formats, but modality introduces distinct sampling demands. A spoken story unfolds at a speaker-selected pace, a written story allows rereading, and a film can externalize spatial relations that prose requires the reader to infer. Consequently, equivalent narrative content does not imply equivalent comprehension difficulty Gernsbacher et al. (1990) Kurby & Zacks (2008) Zwaan & Radvansky (1998).
Gernsbacher and colleagues' investigation of written, spoken, and picture-story comprehension found correlated performance across these modalities in their sampled adult participants. The result favors research into domain-general processes such as constructing and maintaining coherent structure. It does not erase the importance of literacy, auditory perception, vocabulary, or visual attention, each of which can impose independent limits. A reader who repeatedly rereads a sentence may compensate for difficulty that a listener experiencing transient speech cannot. Comparative experiments should match narrative event relations, comprehension questions, and prior familiarity, while acknowledging that item-by-item equivalence may be unattainable Gernsbacher et al. (1990) Gernsbacher (1990).
Neuroimaging research using narrated stories suggests that cortical systems integrate information over a hierarchy of temporal windows, with relatively brief windows in sensory regions and longer integrations in higher-order association regions. These findings are relevant to comprehension of extended speech because a story requires relationships across words, sentences, and episodes. Yet a temporal receptive window is defined through neural response stability or dependence on earlier context, not through direct proof that a person understood every causal link. Behavioral comprehension measures remain necessary to connect the neural signature to representational success. Modalities can recruit partially overlapping temporal organization while maintaining their own sensory and control demands Lerner et al. (2011) Hasson et al. (2015) Baldassano et al. (2017).
Narrative Comprehension and the Social Mind
Stories frequently require readers to represent what a character knows, believes, wants, or misunderstands. Social information becomes relevant when it explains actions and allows readers to anticipate how different characters may respond to the same situation. This form of comprehension connects event models to mental-state attribution, especially where a protagonist acts on a false belief or pursues a goal that another character does not share. A physically coherent plot can still be difficult if the distribution of knowledge across characters has not been represented. Experimental designs therefore contrast stories requiring mental-state inferences with stories requiring comparable physical reasoning, and assess both the interpretation of a scene and the attribution of a particular belief Fletcher et al. (1995) Mar (2011).
Fletcher and colleagues used positron emission tomography to compare story comprehension requiring mental-state attribution, stories emphasizing physical events, and unrelated sentences. Their reported activation differences offered an early neural test of mentalizing in connected discourse. Mar's later synthesis reviewed overlaps between social-cognitive and story-comprehension networks, motivating further investigation of how narrative representations recruit knowledge of people. These findings do not establish that all story comprehension is mentalizing, or that activity in any one brain region uniquely identifies a specific social inference. Contrasts can differ in linguistic difficulty, imagery, and familiarity as well as mental-state demand; these alternative explanations belong in the design and interpretation Fletcher et al. (1995) Mar (2011) Ferstl et al. (2008).
Human goals, emotional responses, and reader knowledge interact with interpretation, but the consequences are heterogeneous. Understanding why a character acted need not entail approving the action, experiencing the character's emotion, or accepting the story's worldview. In a well-designed experiment, comprehension questions, character evaluations, and attitudinal outcomes should be measured separately. For example, readers can infer deception accurately while maintaining a negative view of the deceiver. Narrative perspective can change attention to goals and the accessibility of earlier information without forcing agreement with a character's beliefs. These distinctions are especially valuable when discourse comprehension is investigated alongside empathy, persuasion, or education Green & Brock (2000) Gerrig (1993) Graesser et al. (1994).
Engagement, Imagery, and Narrative Transportation
Narrative engagement includes attention to unfolding events, imagery of described situations, and emotional or evaluative participation. Transportation is a related construct referring to absorption into a story; it has been studied as a possible influence on beliefs and judgments. Green and Brock developed and examined a transportation measure and reported experimental relations between absorption and story-consistent evaluations under their materials and instructions. Their findings address how readers experience and respond to narratives, but transportation should not be used as an interchangeable name for accurate comprehension. A person might be strongly absorbed in a story yet misremember an event, or reconstruct its causal structure while remaining emotionally detached Green & Brock (2000) Gerrig (1993).
Gerrig's account of experiencing narrative worlds brought together comprehension, simulation, emotion, inference, and the consequences of fiction and nonfiction for judgment. The concept is useful because it situates the reader's processing in an experience that unfolds across a whole text. However, phenomenological descriptions require complementary operational measures, and statements about the persuasiveness of stories require controls for selection, prior beliefs, and familiarity. The distinction between being able to answer questions about a fictional event and taking the event as evidence about the real world is crucial. Labeling a work fiction does not automatically eliminate its impact on attitudes, but such effects are contingent and should be analyzed with specific outcome measures Gerrig (1993) Green & Brock (2000).
Engagement also changes the allocation of processing resources. Reading for pleasure, composing an analysis, and searching for factual discrepancies encourage different attentional priorities and thereby alter which details are encoded. Emotional salience can strengthen the memorability of some scenes while narrowing attention away from less salient contextual information. These are plausible interacting pathways, and a causal demonstration requires independent manipulation of story features, goals, and affect. Measures should avoid circularity: defining excellent comprehension by high immersion and then using immersion to explain comprehension supplies no independent test. The appropriate question is which aspects of comprehension remain improved when engagement, prior knowledge, and reading ability are measured separately McKoon & Ratcliff (1992) Green & Brock (2000) Gernsbacher et al. (1990).
Neural Systems and Extended Time Scales
Neural studies extend the research program from behavioral access and recall to measurements of distributed activity while narratives unfold. A meta-analysis by Ferstl and colleagues examined neuroimaging research on text comprehension, emphasizing the involvement of an extended language network in context-sensitive interpretation. The localization and magnitude of reported activations vary with task contrast and analysis choices; individual regions participate in multiple cognitive activities. The value of the synthesis is that coherence and discourse processing require neural resources beyond a narrowly defined word-recognition pathway. It also demonstrates why a region associated with discourse cannot be treated as a complete anatomical seat of narrative understanding Ferstl et al. (2008) Mar (2011).
Lerner, Honey, Silbert, and Hasson used a spoken narrative and scrambled versions preserving different amounts of temporal structure to estimate temporal receptive windows across cortex. Their results support a hierarchy of integration scales. Hasson, Chen, and Honey placed this work within a hierarchical process-memory framework in which processing at longer time scales depends on information retained from prior input. The approach makes a clear prediction: altering temporal order at a particular scale should disrupt response consistency in systems relying on that scale. Nevertheless, response consistency is distinct from comprehension quality, and shared neural responses can emerge from stimulus similarity. Behavioral tests of inference and narrative reconstruction are needed to interpret processing windows Lerner et al. (2011) Hasson et al. (2015).
Baldassano and colleagues analyzed continuous narrative perception and later recall to identify changes between relatively stable cortical event patterns. Their modeling described nested event structure and linked higher-level event boundaries to hippocampal activity and subsequent reinstatement. This provides evidence that neural event structure and memory retrieval are connected during complex narratives. The inference is supported by the reported experimental comparisons and modeling assumptions; it does not show that hippocampal activity is the unique or sufficient cause of an understood storyline. Cross-dataset generalization, annotation reliability, model selection, and controlled event-boundary manipulations are important checks when bringing such findings into a general theory Baldassano et al. (2017) Kurby & Zacks (2008).
Narrative Memory, Retelling, and Interpersonal Transmission
Remembering a narrative is a constructive process shaped by the event structure available at retrieval. A retelling can preserve the central causal arc while omitting descriptive details, and a remembered detail may acquire a new explanation as later knowledge is integrated. The distinction between textbase recall and situation-model retrieval helps explain this selective reconstruction. Models of comprehension have long linked encoding to summarization and retelling, but a narrative reconstruction must be assessed for source fidelity, inferred additions, and relational consistency. A coherent account of a story is therefore not proof of literal storage, and high word overlap is not proof that a listener represented the described situation Kintsch & van Dijk (1978) Zwaan & Radvansky (1998) Gerrig (1993).
Zadbood and colleagues studied communication of remembered events and examined the similarity of neural representations across individuals, providing a naturalistic framework for how information may pass from one person to another. Their findings support the possibility of partially shared neural structures related to communicated memories. The interpretation requires restraint: a shared multivariate response pattern is not an identity of memories, and successful communication cannot be inferred from brain correspondence alone. The verbal account produced by one participant influences what a listener can know, and differences in description, prior knowledge, and listener interpretation remain causally important. The design encourages future research that evaluates neural similarity alongside independently scored event understanding Zadbood et al. (2017) Baldassano et al. (2017).
Transmission adds a social filter to individual comprehension. A speaker selects events, specifies some causal links, leaves others implicit, and adjusts references according to an audience; the listener in turn builds an event model from this selected evidence. Repeated retellings can preserve a broad narrative trajectory while shifting emphasis, perspective, and confidence. Experiments should distinguish the speaker's memory of an event from the produced story and the listener's reconstruction of that story. An independent ground-truth record, when available, allows fidelity to be evaluated across stages. Computational measures of semantic similarity may help quantify retellings, but should be validated against human annotations of event order, goals, and causal relations Zadbood et al. (2017) Brennan & Clark (1996) Kintsch & van Dijk (1978).
Development, Learning, and Individual Differences
Narrative comprehension develops through interaction among language knowledge, memory, inference, and experience with stories. Developmental changes in vocabulary and syntactic comprehension enlarge the range of events that a child can reconstruct from spoken or written descriptions, while familiarity with causal explanations and narrative conventions can aid the organization of a story. This makes comprehension development a multi-component problem: reproducing a sequence, identifying a protagonist's goal, and explaining an unstated causal relation may follow different trajectories. Researchers should therefore specify the age group, modality, genre, assessment demands, and amount of prior practice before treating a single score as comprehensive narrative ability Kim (2023) Gernsbacher et al. (1990).
Kim's 2023 study examined oral narrative and expository comprehension and retelling among 529 English-speaking second-grade children. Confirmatory analyses supported four related dimensions rather than an undifferentiated oral-discourse score. The results underscore a methodological distinction between understanding information and retelling it, and between narrative and explanatory discourse. Retelling draws additionally on production, memory, and organization, so treating it as a pure proxy for comprehension can conceal meaningful differences. The findings characterize a particular age group and language context; claims about developmental trajectories, cross-language equivalence, or causal influences of instruction require data designed for those questions Kim (2023) Kintsch & van Dijk (1978).
Individual differences are also visible in adult research. Gernsbacher, Varner, and Faust reported relations among written, spoken, and pictorial comprehension together with findings concerning accessibility and suppression. More recent work on visual narrative processing examines how proficiency and experience with a visual storytelling convention affect neural dynamics. These approaches direct attention to the knowledge and processing routines that support comprehension across media. Their designs should be interpreted in light of selection, task reliability, and the potential for shared variance arising from general cognitive skills. A strong developmental program combines longitudinal measurements, experimental teaching manipulations, and item-level analyses instead of assigning all variation to one fixed capacity Gernsbacher et al. (1990) Coderre & Cohn (2024).
Narrative Conventions, Culture, and Language Experience
Readers bring conventions about what normally follows what: restaurants have characteristic activities, journeys involve changes of place, and interpersonal encounters can follow recognizable social sequences. Such schematic knowledge supports anticipation and supplies plausible explanations for omitted actions. It can also mislead when a narrative deliberately violates expectations or when the reader's culturally learned assumptions differ from those of the author. Experimental work must separate knowledge of an event script from skill in comprehending the language that expresses it. The same story can be grammatically simple yet conceptually demanding because its underlying practices are unfamiliar Bransford & Johnson (1972) Graesser et al. (1994) De Soares et al. (2024).
De Soares and colleagues presented narratives combining place-related and social event scripts, examining both neural and behavioral event boundaries. Directing attention toward a particular script changed the alignment of perceived boundaries and relevant neural response structure in their study. The result is significant because it identifies an experimentally manipulated goal that changes how the same material is organized. It supports the view that segmentation partly reflects top-down attention and schematic knowledge, alongside input-driven changes. This evidence does not quantify differences among national cultures or establish a universal event-script inventory. Cross-cultural research requires stimuli developed and validated with participants from the populations under study De Soares et al. (2024) Zacks et al. (2007).
Language experience affects access to cohesive signals, and stories may use indirect reference, temporal markers, connectives, and culturally specific ways of presenting motives. These features can alter comprehension even when the broad event sequence is preserved in translation. A study comparing languages must establish that questions, connective meanings, narrative conventions, and familiarity have been appropriately matched. It should also distinguish linguistic proficiency from content knowledge rather than attributing every discrepancy to language structure. Cross-cultural generalization remains an open area because many established laboratory studies use university samples and researcher-authored English passages. The evidence base is therefore better described as strong for selected mechanisms under tested conditions than as exhaustive across linguistic communities Kim (2023) Zwaan & Radvansky (1998) Kintsch (1988).
Recent Evidence on Event Encoding and Sequence Reconstruction
Recent studies sharpen the distinction between preserving event content and recovering a coherent temporal order. Hahamy, Dubossarsky, and Behrens reported selective reactivation of relevant past information at boundaries in naturalistic narratives, with replication across two datasets. Their results suggest a candidate process for reconnecting distant events during comprehension. This is compatible with, but more specific than, a general claim that stories are stored as ordered propositions. A neural replay-like pattern is an observed signature whose functional role must be tested with interventions or converging behavioral manipulations; neural similarity alone cannot decide whether reactivation is sufficient to construct a correct causal narrative Hahamy et al. (2023) Baldassano et al. (2017).
Barnett and colleagues investigated functional interactions between hippocampus and posterior medial cortical systems during encoding of naturalistic movie events. They reported associations between connectivity near event offsets and subsequent recall, including later retention of detail. The experimental design extends boundary research into the relationship between encoding conditions and later retrieval over an interval. However, connectivity estimated from fMRI establishes statistical dependence between measured time series and does not identify direct synaptic communication or the direction of causal influence. The most useful next step is to link event-boundary manipulations to both comprehension and memory outcomes while controlling for salience and surprise Barnett et al. (2024) Speer et al. (2007).
Park, Song, and Shim's 2025 research used a scrambled movie and later narrative reconstruction to distinguish the accuracy of event content from the ability to infer temporal sequence. Their analysis associated hippocampal–posterior medial connectivity with content encoding and hippocampal–ventromedial prefrontal connectivity with ordering-related performance, under the reported tasks and modeling procedures. The study incorporated language-model-derived estimates to locate candidate moments of integration; these computations serve as analytic instruments rather than independent neural measurements. Together with prior work, these results make temporal sequencing a measurable research target and suggest partially differentiable neural correlates, while leaving open how the same relations operate in written prose Park et al. (2025) Hahamy et al. (2023) Barnett et al. (2024).
Methods, Measurement, and the Validity of Comprehension Claims
Behavioral methods include self-paced reading, eye tracking, immediate verification, antecedent probes, recognition, event-order judgments, causal explanation, free recall, and structured retelling. Each measures a different combination of comprehension and task performance. A slowed reading segment can reveal a processing consequence without uniquely identifying its cause. Correct recognition can reflect recent repetition or deliberate retrieval, while an accurate explanation may require knowledge not present at the time of reading. Multiple measurements are most informative when their expected relations follow from an explicit hypothesis and when competing accounts predict different patterns. Pretesting linguistic difficulty, counterbalancing stimuli, and testing item effects are basic safeguards against artifactual findings Zwaan et al. (1995) McKoon & Ratcliff (1992) Kim (2023).
Naturalistic paradigms increase relevance to extended narrative experience by retaining story arcs, characters, and event boundaries that isolated sentences rarely supply. Their interpretive challenge is that temporal position, emotion, linguistic complexity, scene changes, and model surprise are often correlated. Statistical modeling may account for selected covariates, but unmeasured differences can survive. The strongest naturalistic studies complement realistic stories with independent annotations, controlled disruptions, held-out material, and behavioral checks of actual understanding. Neural segmentation and multivariate-pattern analyses also depend on model-selection assumptions, temporal alignment, and researcher decisions about what counts as a boundary Baldassano et al. (2017) Brich et al. (2024) De Soares et al. (2024).
Reliability and construct validity require particular attention in individual-difference research. A comprehension test built around factual multiple-choice questions may underestimate causal integration; one based on verbal retelling may confound understanding with expressive production. Narrative and expository texts can measure related but dissociable abilities, as demonstrated in recent dimensional analyses. Researchers should report scoring rubrics, inter-rater agreement for open-ended answers, alternate forms, text difficulty, and the degree of similarity between training and test materials. For neural studies, preregistered regions, cross-validated predictive accuracy, and appropriately separated datasets help constrain flexible analyses. A coherent explanation of behavior becomes stronger when a manipulated mechanism changes the predicted specific outcome Kim (2023) Ferstl et al. (2008) Park et al. (2025).
Alternative Explanations, Model Comparison, and Reproducibility
The interpretation of an online coherence effect depends on whether plausible alternatives have been excluded. A longer reading time after a scene change may indicate construction of a new situation model, but it can also reflect a less predictable sentence, a low-frequency expression, an unfamiliar setting, or a change in attentional strategy. Similarly, an accurate answer to an inference question may arise from a representation formed during reading or from a deduction made when the question is presented. Explicit model comparison begins by specifying which result each explanation predicts under independently manipulated conditions. A study that varies causal necessity while holding sentence plausibility, length, and lexical overlap as constant as practical can make an inference claim more discriminating. This principle applies equally to classical self-paced reading and contemporary neural-event decoding McKoon & Ratcliff (1992) Graesser et al. (1994) Brich et al. (2024).
Brich and colleagues' 2024 study examined whether signatures of visual narrative processing are better characterized by initial construction or updating of event representations. Its framing is valuable because an observed boundary response can be compatible with both accounts, and theories need tests that specify what changes at an event transition. The choice between mechanisms requires sufficient temporal resolution and experimental variation rather than assigning distinct process labels solely from a response peak. In naturalistic materials, event boundaries co-occur with changes in depicted action, camera perspective, or semantic content; independent annotations and analyses of stimulus features can improve interpretation. The same methodological logic extends to temporal-window fMRI, where neural response reliability cannot alone establish a reader's grasp of causality or character goals Brich et al. (2024) Lerner et al. (2011) Baldassano et al. (2017).
Reproducibility encompasses reliable stimuli, scoring, analyses, and interpretation across readers, languages, and laboratories. Narrative texts provide rich but often unique materials, so apparent replications may use different story worlds and operational measures. Item-level variation belongs in the statistical model, and conclusions should report how effects vary by story, genre, participant knowledge, and task demands. The use of preregistered manipulations, independent test stories, transparent coding of event boundaries, and held-out model comparisons can reduce opportunities to select the result that best suits a preferred explanation. Publicly available stimulus sets and multiple definitions of comprehension are particularly valuable. The goal is to identify which mechanisms generalize, which are context dependent, and which apparent effects depend on a specific measurement procedure Kim (2023) Coderre & Cohn (2024) Ferstl et al. (2008).
Computational Accounts and Artificial Narrative Understanding
Computational models can make discourse theories more precise by representing entities, goals, causal relations, memory activation, and transitions between events in explicit states and algorithms. The construction–integration model is historically important because it describes how activated propositions and knowledge relations settle into an organized representation. Event-segmentation models similarly formalize predictions about incoming information and the conditions under which the representation changes. Such models are useful when they predict experimentally discriminating patterns of reading times, reference accessibility, boundary judgments, or recall. Model fit should be compared with a simpler baseline and with competing explanatory architectures, and a successful fit to one dataset should be tested on genuinely new materials Kintsch (1988) Reynolds et al. (2007) Zwaan & Radvansky (1998).
Artificial language systems create a further research opportunity: they can be given narratives that require preservation of protagonist identities, detection of temporal changes, explanation of unstated causal links, and reconstruction of event order. Performance on these tasks can be measured through carefully defined outputs, controlled perturbations, and novel narratives kept separate from training and evaluation materials. The relevant psychological question is functional: which aspects of extended discourse can be reproduced under which task conditions, and what interventions change the outcomes? A language model may produce an acceptable summary while losing a crucial causal relation or silently importing general knowledge into a fictional account. Behavioral success therefore needs to be scored at the level of referents, events, temporal order, and source faithfulness, rather than inferred from fluent prose alone Kintsch & van Dijk (1978) Graesser et al. (1994) Park et al. (2025).
The Artificial Era expands the comparative scope of discourse research by placing biological readers and constructed language-processing systems in the same experimental field, while requiring explicit descriptions of different training histories, sensory channels, memory arrangements, and interaction opportunities. A useful benchmark presents both populations with unfamiliar narratives and evaluates questions about anaphora, nested goals, contradiction, temporal reordering, and justified inference, using common outcome definitions alongside system-specific controls. Its theoretical value rests on testable performance patterns, not on assigning an unmeasured human experience to an artificial system. This approach can contribute to human–artificial communication, educational interfaces, and error analysis while retaining empirical distinctions between psychological claims about human minds and computational hypotheses about artificial architectures Gernsbacher et al. (1990) Hasson et al. (2015) Brich et al. (2024).
Integrated Scientific Synthesis and Future Research Questions
Across six decades of research, the most defensible synthesis is that comprehension of connected language depends on interacting levels of representation that vary in accessibility over time. Words and sentences constrain a developing textbase; prior knowledge and event-specific cues help organize a situation model; the resulting representation supports further inference, anticipation, and retrieval. Local bridges and broader causal explanations have distinguishable triggers. Spatial changes, temporal shifts, protagonist continuity, intentions, and causal relations jointly contribute to updating, though their weights vary by task and material. A single explanatory framework must therefore connect sentence-level linguistic evidence to evolving event-level relations without assuming that a reader stores every proposition in a fully active state Kintsch & van Dijk (1978) Zwaan & Radvansky (1998) Graesser et al. (1994).
The evidence also supports a multi-scale architecture of narrative processing. Behavioral event-segmentation studies indicate that discontinuities change processing and subsequent accessibility, while neuroimaging reports nested temporal integration and event-related changes in distributed cortical and hippocampal systems. Recent investigations of contextual reactivation, boundary-associated connectivity, and event-sequence reconstruction refine hypotheses about which memory processes operate near transitions. These results are convergent in their relevance to event organization, but the available measurements do not isolate a single neural implementation of comprehension. Experimental manipulation of narrative structure, sensitive behavioral indices, and independently validated neural predictions are needed to connect representational theory with causal neurocognitive explanation Speer et al. (2007) Baldassano et al. (2017) Hahamy et al. (2023) Barnett et al. (2024) Park et al. (2025).
The immediate research agenda includes separating online inference from inference at test, identifying the conditions under which boundaries disrupt or strengthen memory, and determining how causal versus temporal relations are encoded and retrieved. Additional priorities are longitudinal measurement of narrative development, validation of comprehension across spoken, printed, pictured, and filmed narratives, and tests with wider linguistic and cultural samples. Computational benchmarks should contain adversarial changes to entities, goals, and event order, along with calibrated measures of source fidelity. Each program benefits from a shared principle: specify the exact representation and operation at issue, manipulate its proposed cause, and measure outcomes that discriminate alternatives. Such an agenda makes discourse and narrative comprehension an integrated but experimentally tractable field Kim (2023) De Soares et al. (2024) McKoon & Ratcliff (1992) Reynolds et al. (2007).
Conclusion
Discourse and narrative comprehension organize connected language into representations that support understanding beyond the literal sequence of sentences. The foundational psychological result is that readers and listeners track situations, entities, temporal progress, causal dependencies, and intentions while interpreting new information. Classical models of propositional coherence and construction–integration explain key relations between textual material and knowledge; situation-model and event-indexing theories specify how an evolving account of described events is maintained. Experiments involving spatial accessibility, causal inference, reading-time discontinuities, and narrative recall establish that these processes can be investigated through distinct observable measures. Their interpretation improves when immediate access, later retrieval, and comprehension accuracy remain analytically separate Kintsch (1988) Morrow et al. (1987) Zwaan et al. (1995) Graesser et al. (1994).
The field's contemporary scientific strength lies in convergence among controlled comprehension experiments, naturalistic story paradigms, computational accounts, and measurements of distributed brain responses. Event segmentation, hierarchical temporal integration, neural reactivation, and sequence reconstruction expand the investigation of narrative beyond isolated sentences. These advances also create methodological obligations: neural correspondence must be distinguished from causal implementation, model fit from theoretical uniqueness, and vivid engagement from accurate understanding. Variation among populations, languages, modalities, and story structures is a substantive part of the phenomenon and a condition for evaluating the generality of any account Ferstl et al. (2008) Baldassano et al. (2017) Hahamy et al. (2023) Park et al. (2025).
The resulting scientific picture is of a dynamic, knowledge-guided and experimentally constrained organization of connected meaning. Narrative comprehension is simultaneously linguistic, cognitive, mnemonic, and frequently social because it preserves relationships among statements, events, and agents across extended time. Its unresolved problems are concrete: which inferences are generated during reading, how and when event models are revised, why some boundaries facilitate later memory while others impair access, and which mechanisms generalize across modalities and agents. The next stage is defined by precise operational contrasts, transparent evidence assessment, and independent testing of theoretical predictions. This synthesis establishes the research boundaries and continuing questions of Discourse and Narrative Comprehension Zwaan & Radvansky (1998) Kurby & Zacks (2008) Kim (2023) Brich et al. (2024).
References
- Baldassano, C., Chen, J., Zadbood, A., Pillow, J. W., Hasson, U., & Norman, K. A. (2017). Discovering event structure in continuous narrative perception and memory. Neuron, 95(3), 709–721.e5. https://doi.org/10.1016/j.neuron.2017.06.041
- Barnett, A. J., Nguyen, M., Spargo, J., Yadav, R., Cohn-Sheehy, B. I., & Ranganath, C. (2024). Hippocampal-cortical interactions during event boundaries support retention of complex narrative events. Neuron, 112(2), 319–330.e7. https://doi.org/10.1016/j.neuron.2023.10.010
- Bower, G. H., & Morrow, D. G. (1990). Mental models in narrative comprehension. Science, 247(4938), 44–48. https://doi.org/10.1126/science.2403694
- Bransford, J. D., & Johnson, M. K. (1972). Contextual prerequisites for understanding: Some investigations of comprehension and recall. Journal of Verbal Learning and Verbal Behavior, 11(6), 717–726. https://doi.org/10.1016/S0022-5371(72)80006-9
- Brennan, S. E., & Clark, H. H. (1996). Conceptual pacts and lexical choice in conversation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22(6), 1482–1493. https://doi.org/10.1037/0278-7393.22.6.1482
- Brich, I. R., Papenmeier, F., Huff, M., & Merkt, M. (2024). Construction or updating? Event model processes during visual narrative comprehension. Psychonomic Bulletin & Review, 31, 2092–2101. https://doi.org/10.3758/s13423-023-02424-w
- Coderre, E. L., & Cohn, N. (2024). Individual differences in the neural dynamics of visual narrative comprehension: The effects of proficiency and age of acquisition. Psychonomic Bulletin & Review, 31, 89–103. https://doi.org/10.3758/s13423-023-02334-x
- De Soares, A., Kim, T., Mugisho, F., Zhu, E., Lin, A., Zheng, C., & Baldassano, C. (2024). Top-down attention shifts behavioral and neural event boundaries in narratives with overlapping event scripts. Current Biology, 34(20), 4729–4742.e5. https://doi.org/10.1016/j.cub.2024.09.013
- 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(5), 581–593. https://doi.org/10.1002/hbm.20422
- Fletcher, P. C., Happé, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak, R. S. J., & Frith, C. D. (1995). Other minds in the brain: A functional imaging study of “theory of mind” in story comprehension. Cognition, 57(2), 109–128. https://doi.org/10.1016/0010-0277(95)00692-R
- Gernsbacher, M. A. (1990). Language comprehension as structure building. Lawrence Erlbaum Associates. https://www.routledge.com/Language-Comprehension-As-Structure-Building/Gernsbacher/p/book/9780805806762
- Gernsbacher, M. A., Varner, K. R., & Faust, M. E. (1990). Investigating differences in general comprehension skill. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16(3), 430–445. https://doi.org/10.1037/0278-7393.16.3.430
- Gerrig, R. J. (1993). Experiencing narrative worlds: On the psychological activities of reading. Yale University Press. https://yalebooks.yale.edu/book/9780300054347/experiencing-narrative-worlds/
- Glenberg, A. M., Meyer, M., & Lindem, K. (1987). Mental models contribute to foregrounding during text comprehension. Journal of Memory and Language, 26(1), 69–83. https://doi.org/10.1016/0749-596X(87)90063-5
- Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text comprehension. Psychological Review, 101(3), 371–395. https://doi.org/10.1037/0033-295X.101.3.371
- Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701–721. https://doi.org/10.1037/0022-3514.79.5.701
- Hahamy, A., Dubossarsky, H., & Behrens, T. E. J. (2023). The human brain reactivates context-specific past information at event boundaries of naturalistic experiences. Nature Neuroscience, 26, 1080–1089. https://doi.org/10.1038/s41593-023-01331-6
- Hasson, U., Chen, J., & Honey, C. J. (2015). Hierarchical process memory: Memory as an integral component of information processing. Trends in Cognitive Sciences, 19(6), 304–313. https://doi.org/10.1016/j.tics.2015.04.006
- Kim, Y.-S. G. (2023). Oral discourse skills: Dimensionality of comprehension and retell of narrative and expository texts, and the relations of language and cognitive skills to identified dimensions. Child Development, 94(5), e246–e263. https://doi.org/10.1111/cdev.13935
- Kintsch, W. (1988). The role of knowledge in discourse comprehension: A construction-integration model. Psychological Review, 95(2), 163–182. https://doi.org/10.1037/0033-295X.95.2.163
- Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85(5), 363–394. https://doi.org/10.1037/0033-295X.85.5.363
- Kurby, C. A., & Zacks, J. M. (2008). Segmentation in the perception and memory of events. Trends in Cognitive Sciences, 12(2), 72–79. https://doi.org/10.1016/j.tics.2007.11.004
- Lerner, Y., Honey, C. J., Silbert, L. J., & Hasson, U. (2011). Topographic mapping of a hierarchy of temporal receptive windows using a narrated story. The Journal of Neuroscience, 31(8), 2906–2915. https://doi.org/10.1523/JNEUROSCI.3684-10.2011
- Magliano, J. P., & Schleich, M. C. (2000). Verb aspect and situation models. Discourse Processes, 29(2), 83–112. https://doi.org/10.1207/S15326950dp2902_1
- Mar, R. A. (2011). The neural bases of social cognition and story comprehension. Annual Review of Psychology, 62, 103–134. https://doi.org/10.1146/annurev-psych-120709-145406
- McKoon, G., & Ratcliff, R. (1992). Inference during reading. Psychological Review, 99(3), 440–466. https://doi.org/10.1037/0033-295X.99.3.440
- Morrow, D. G., Greenspan, S. L., & Bower, G. H. (1987). Accessibility and situation models in narrative comprehension. Journal of Memory and Language, 26(2), 165–187. https://doi.org/10.1016/0749-596X(87)90122-7
- Park, J., Song, H., & Shim, W. M. (2025). Hippocampal systems for event encoding and sequencing during ongoing narrative comprehension. Communications Biology, 8, Article 954. https://doi.org/10.1038/s42003-025-08377-1
- Radvansky, G. A. (2012). Across the event horizon. Current Directions in Psychological Science, 21(4), 269–272. https://doi.org/10.1177/0963721412451274
- Reynolds, J. R., Zacks, J. M., & Braver, T. S. (2007). A computational model of event segmentation from perceptual prediction. Cognitive Science, 31(4), 613–643. https://doi.org/10.1080/15326900701399913
- Speer, N. K., Zacks, J. M., & Reynolds, J. R. (2007). Human brain activity time-locked to narrative event boundaries. Psychological Science, 18(5), 449–455. https://doi.org/10.1111/j.1467-9280.2007.01920.x
- van den Broek, P. (1990). Causal inferences and the comprehension of narrative texts. Psychology of Learning and Motivation, 25, 175–196. https://doi.org/10.1016/S0079-7421(08)60255-8
- Zacks, J. M., Speer, N. K., Swallow, K. M., Braver, T. S., & Reynolds, J. R. (2007). Event perception: A mind–brain perspective. Psychological Bulletin, 133(2), 273–293. https://doi.org/10.1037/0033-2909.133.2.273
- Zadbood, A., Chen, J., Leong, Y. C., Norman, K. A., & Hasson, U. (2017). How we transmit memories to other brains: Constructing shared neural representations via communication. Cerebral Cortex, 27(10), 4988–5000. https://doi.org/10.1093/cercor/bhx202
- Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in language comprehension and memory. Psychological Bulletin, 123(2), 162–185. https://doi.org/10.1037/0033-2909.123.2.162
- Zwaan, R. A., Langston, M. C., & Graesser, A. C. (1995). The construction of situation models in narrative comprehension: An event-indexing model. Psychological Science, 6(5), 292–297. https://doi.org/10.1111/j.1467-9280.1995.tb00513.x
- Zwaan, R. A., Magliano, J. P., & Graesser, A. C. (1995). Dimensions of situation model construction in narrative comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(2), 386–397. https://doi.org/10.1037/0278-7393.21.2.386