Psychology Of Narrative Comprehension Research Paper

1. What Does It Mean To Comprehend A Narrative?

Narratives relate how a series of events, fictitious or real, unfold. In doing so, they provide more than a simple list of the individual events: they also indicate how these events are interrelated and interdependent (etymologically, the roots of the term ‘narrative’ and its cognate ‘story’ are ancient forms of ‘knowing /understanding,’ indicating the interpretative and explanatory nature of narratives). Analogously, the goal of narrative comprehension is to construct a memory representation in which the various individual parts of the narrative are connected in meaningful ways (e.g., see Mental Representations, Psychology of ). Psychological examination of narrative comprehension includes consideration of the nature of this representation, of the processes that take place during its construction, and of the connection between these processes and the resulting representation.

2. Comprehension Of Narrative Texts

Narrative comprehension is investigated most extensively in the domain of reading (van den Broek 1994, van Dijk and Kintsch 1983; see also Memory for Text). There are several reasons for this. Reading is one of the most uniquely human and complex of all cognitive activities, with enormous social and educational importance. Moreover, texts allow researchers considerable experimental control over the materials they present to their participants.

2.1 Memory Representation Of Narratives: Inferences And Coherence

When reading is successful, the result is a coherent and usable mental representation of the text. This representation resembles a network, with nodes that depict the individual text elements (e.g., facts, events, settings) and connections that depict the meaningful relations between the elements. These relations are what make a representation coherent and usable. Many possible relations exist, but two types have been found to be particularly important for establishing coherence: referential and causal/logical. Referential relations enable readers to keep track of elements such as objects, people and events mentioned in several places within the text (Kintsch and van Dijk 1978; see also Propositional Representations in Psychology). Causal and logical relations enable readers to identify how different events or facts depend on each other (Schank and Abelson 1977, Graesser and Clark 1985, Trabasso and van den Broek 1985). For example, when reading ‘The lady gave the waiter 10 Euros. He returned to give her the change,’ referential coherence is established by recognizing that ‘he’ in the second sentence refers to the waiter and ‘her’ to the lady. Causal/logical coherence is established by inferring that the waiter is returning change because the lady consumed something in an establishment where the waiter works, and because the lady had given him 10 Euros for a purchase of a lesser amount.

Several properties of coherent representation of narrative texts are to be noted. First, the identification of relations requires readers to make an inference, that is, to integrate the textual information with their semantic knowledge (or background knowledge). In the above example the requisite semantic knowledge (e.g., that ‘she’ tends to refer to a female protagonist) is likely to be readily available to most readers, but in other situations it is not. For example, many readers have difficulty recognizing the causal relation between the following two sentences: ‘The presence of the moon exerted a strong gravitational pull on the earth. Thus the moon contributed to the emergence of life on earth’; even though the referential relation is direct (i.e., ‘moon’) and the causal relation is explicitly marked by linguistic means (‘thus’). The reason is that most readers lack the semantic knowledge necessary for the inference: The moon’s gravitational pull is instrumental in creating the earth’s magnetic field which, in turn, shields the earth’s surface from lethal cosmic radiation. The semantic knowledge that readers bring to bear includes both individual facts (as in the above example) and generalized assemblies of knowledge (e.g., the sequence of events in ‘going to a restaurant’; see Schemas, Frames, and Scripts in Cognitive Psychology).

Second, many of the referential and causal logical relations that readers must infer are much more complex than the simple ones in the ‘lady and the waiter’ example. Relations may extend over long distances in the text, may be difficult to interpret, or may require coordination of numerous pieces of information. Third, the text’s linguistic structure may aid the reader by explicitly marking a relation, thereby making the inference more obvious. Linguistic markers take various forms, ranging from inflections to explicit connectives (as the word ‘thus’ in the above example). Fourth, the number of relational inferences required for a coherent representation usually is fairly large, even if many are generated without effort or awareness by skilled readers.

Considerable empirical evidence supports the key role that referential and causal logical networks play in successful comprehension. The evidence is too extensive to be reviewed in its entirety (for a review, see van den Broek 1994), but some examples may be helpful: (a) the more relations events have, the more frequently they are recalled and the more important they are judged by readers; (b) answers to questions about the text tend to follow the connections captured by the networks; (c) when reminded, after reading, of an event or fact from the text, readers spontaneously and automatically bring to mind other elements directly related to it but not unrelated ones from the same text; and (d) the time it takes to read a sentence is proportional to the number of inferences needed for comprehension. Thus, referential and causal logical relations form the backbone of the mental representation of narrative texts.

Although referential and causal logical relations are the most prevalent in narratives, other types of relations such as spatial and temporal ones can also be generated. These relations are less central to comprehension, however, and generally are only included in the memory representation if they are functional in establishing referential or causal logical coherence.

The network of nodes and semantic relations constitutes the fundamental representation of a narrative text. It also forms the basis for further organization within the narrative. This is the case because, within a network, events are clustered into episodes, each of which revolves around a protagonist’s goal. The nature of episodes was first explored in the 1970s, with the development of story grammars (e.g., Mandler and Johnson 1977). Following anthropological work on folktales, story grammarians noted that story statements can be categorized into a small number of content types and, furthermore, that the different types are sequenced according to a small number of grammatical rules. These story grammars differ in detail and use different labels to designate categories, but they show substantial overlap. In general, stories are described as consisting of a number of Setting statements and one or more episodes. The Setting statements provide the backdrop for the remainder of the narrative by describing main characters, objects, temporal and geographic information, and so on. Episodes describe the events and actions of the narrative proper. Each episode is organized around a goal and its outcome. Within an episode, events differ in function. In the simplest episode, an Initiating Event describes the occurrence of an event that sets the train of action in motion. It elicits an Internal Response, describing the reactions of the protagonist, which in turn causes the implicit or explicit establishment of a Goal on the part of the protagonist. The goal motivates one or more Actions, which lead to the Outcome of the episode. The outcome results in a Reaction, describing the protagonist’s reaction to the success or failure of the outcome. Most narratives have multiple instantiations of each category, with the grammatical rules dictating their order. Although story grammars now generally are considered derivative from the various semantic relations that exist between narrative elements, they serve an important function in drawing attention to the semantic structure of narratives.

The episodes in a narrative are themselves interrelated, forming more complicated macrostructures. A particularly important episode structure results when episodes arrange into a hierarchy, with one goal (or outcome) causing the establishment of another goal. This happens, for example, when one goal is pursued in order to attain another: the former is subordinate to the latter, its superordinate. Goals higher in the hierarchy form the main ideas of the narrative, with the most superordinate goal representing the narrative’s theme. Other macrostructures exist: linear (e.g., streams of consciousness), intertwined (e.g., maintaining different story lines), competing strands (e.g., some genres of detective stories), and so on.

2.2 The Comprehension Process: Attention And Inferences

The memory representation of a text is the result of inferential processes that take place during reading. These processes allow the reader to establish coherence as he or she proceeds through the text. In general, the more relations a reader detects, the more coherent the text representation is and the better comprehension will be. Limitations in the reader’s information processing system restrict the generation of large numbers of inferences, however. Extensive inferential processes would overwhelm the reader’s attentional capacity or slow down reading to a snail’s pace. Instead, only a subset of all possible inferences is generated.

Which subset is generated depends on a number of factors. The primary of these factors concerns the function of a possible inference in establishing coherence. At each reading cycle (i.e., each new sentence), the comprehender attempts to make sense of the information explicitly stated in the current sentence by inferring connections to his or her background knowledge and/or statements in the prior text. Thus, the most important types of inferences are those that provide coherence to the text, in particular those that resolve referential ambiguity and provide causal and logical explanations. These are required for comprehension. Results from a large number of studies indicate that these coherence-building inferences are routinely made during reading. For example, when readers are interrupted during reading by the presentation of a word, they recognize or speak it aloud more quickly if the word corresponds to one of these inferences than if it does not. Thus, the concept represented by this word apparently already was activated (i.e., inferred) before the interruption. Similar evidence indicates that other types of inferences (e.g., spatial, temporal, instrumental) usually are made only when they also provide referential or causal coherence or when the reader has special goals in reading.

Other factors determining whether a particular inference is made are the extent to which the semantic information in the text and the linguistic cues provide sufficient constraints for the potential inferences, the reader’s background knowledge relevant to the inference, the reader’s attentional resources, and any special objectives in reading the text.

2.3 The Construction Of A Memory Representation

The processes that take place during comprehension ultimately lead to a memory representation of the text. Several theoretical models capture the mechanisms by which this is accomplished (see van Oostendorp and Goldman 1999 for descriptions of several prominent models). These models differ in their details, but they are remarkably similar in their general description of the process by which a memory representation emerges from the processes during reading.

Central is the notion that reading involves a continuous and dynamic fluctuation of activations. This is a direct consequence of a reader’s limited attentional capacity and of the inferential processes described above. At each reading cycle, new concepts are (re)activated, some old ones are retained, and others are removed from the focus of attention. Sources of activations are (a) the currently read text, (b) carry-over from the preceding cycle, (c) the reader’s semantic knowledge, and (d) the reader’s episodic memory representation of the text read so far. Carryover from the preceding cycle is limited, and primarily includes theme or goal information. Background knowledge is accessed either through an automatic activation of concepts associated with the ideas in the current sentence or through a more strategic search for relevant information. Finally, information from preceding sentences can be reactivated by retrieval (‘reinstatement’) from the episodic memory representation of the prior text or by consultation of the actual text (‘looking back’).

With each cycle, the memory representation of the narrative is updated. New text elements are added and existing ones increased in strength as a function of their activation. In general, the more often and the more strongly an element is activated over the course of reading, the more strongly it will feature in the final representation of the narrative. Furthermore, when two text elements are activated simultaneously, meaningful relations between them can be identified. The more often and more strongly two text elements are coactivated, the stronger their relation in the memory representation will be.

In this fashion, the memory representation gradually evolves out of the processes that take place during reading. The relation between process and product is reciprocal: at each cycle, the activation of concepts is partly determined by the memory representation that resulted from the preceding cycles; the new activations, in turn, update this representation.

2.4 Variations And Veracity In Narrative Comprehension

Individual readers differ in attentional capacity, cognitive skills (e.g., coherence-building strategies), personal and cultural background knowledge, and so on. These variations alter the comprehension processes and, hence, the representation of the narrative. Moreover, readers also differ in terms of their standards for coherence. The present account has focused on the standards of referential and causal logical coherence, which are basic and shared by most fluent readers. However, other—stricter, more lenient, or simply different—standards may be employed. For example, different representations result when people are asked to read a narrative taking place in a house with the perspective of a potential homebuyer or that of a burglar (Bransford and Johnson 1972). Moreover, standards vary as readers have different goals for reading the narrative (e.g., for a test, for entertainment, to learn about the spatial layout). Similar variations on the above description of narrative processing and representation occur within individuals from one reading situation to the next.

Further, a representation that is internally coherent need not be identical with that intended by the author of the narrative. Indeed, the overlap between intended and obtained representations is likely to be partial at best. In some situations (e.g., educational contexts) comprehension may require that the overlap is substantial, whereas in others (e.g., entertainment) it may not.

3. Comprehension Of Narratives In Other Contexts

Narratives exist in many contexts outside reading. Examples are television and movie viewing, oral communication, historical accounts, and so on. The comprehension of narratives in these contexts has received far less attention than that during reading. Moreover, most extant research has focused on the product of comprehension (the representation) rather than on the process. Nevertheless, the available evidence suggests that the processes involved in the comprehension and representation of narratives are remarkably similar across contexts. For example, causal relations have been found to play a central role in the processing of television and movie narratives, of orally presented narratives, and even in the interpretation of arguments in courts.

4. Future Directions

The psychological investigation of narrative comprehension is likely to proceed in several directions. One direction is the further development of the computational models of reading comprehension described above. Aided by increasingly powerful computers, these models are likely to address hitherto underexplored issues such as the mechanism by which semantic knowledge is recruited during comprehension, how semantic knowledge itself is altered by the comprehension of individual texts, and so on. A second direction is the investigation of the neurological bases for these and other comprehension processes. Such investigations will be focused not only on where in the brain certain processes take place, but also on the theoretical implications of such determination.

Third, the affective component of narrative comprehension will be investigated and the results integrated with those of the cognitively oriented models developed so far. Clearly, narratives elicit strong emotions on the part of the comprehender; indeed, such emotions are part of what makes narratives interesting. A final direction will be the extension of the detailed models developed for narrative comprehension to comprehension in other domains, both in reading and in other contexts. Thus, the study of narrative comprehension truly becomes a study of the human mind in general.

Bibliography:

1. 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: 717–26
2. Graesser A C, Clark L F 1985 Structures and Procedures of Implicit Knowledge. Ablex, Norwood, NJ
3. Kintsch W, van Dijk T A 1978 Towards a model of text comprehension and production. Psychological Review 85: 363–94
4. Mandler J, Johnson N S 1977 Remembrance of things parsed: story structure and recall. Cognitive Psychology 9: 111–51
5. Schank R C, Abelson R 1977 Scripts, Plans, Goals, and Understanding. Erlbaum, Hillsdale, NJ
6. Trabasso T, van den Broek P 1985 Causal thinking and the representation of narrative events. Journal of Memory and Language 24: 612–30
7. van den Broek P 1994 Comprehension and memory of narrative texts: inferences and coherence. In: Gernsbacher M A (ed.) Handbook of Psycholinguistics. Academic Press, San Diego, pp. 539–88
8. van Dijk T A, Kintsch W 1983 Strategies of Discourse Comprehension. Academic Press, New York
9. van Oostendorp H, Goldman S R (eds.) 1999 The Construction of Mental Representations During Reading. Erlbaum, Hillsdale, NJ