Macrostructure Representation Research Paper

Discourse comprehension may be thought of as a process of constructing a memory representation. In the case of text comprehension, a reader is credited with good comprehension to the extent that the reader constructs a coherent representation that is consistent with the message intended by the author. It is particularly critical that the reader accurately represent the text’s main ideas and their organization, or macrostructure. This research paper addresses the processes involved in the construction and use of macrostructure representations as a critical component of discourse comprehension.

1. Two Levels of Discourse Representation

A text has both a horizontal organization, or microstructure, and a vertical organization, or macrostructure (Kintsch and van Dijk 1978, van Dijk 1980, van Dijk and Kintsch 1983). A full understanding of a text requires that a reader represent both the microstructure and the macrostructure.

A text’s horizontal organization derives from the fact that most sentences can be interpreted in the context of the immediately preceding sentence. A microstructure representation captures these local coherence relations. A formal analysis of a text’s microstructure requires that the sentences of the text be analyzed into their underlying ideas, or propositions and their relations. Propositions are related, for example, if they share referents or if they denote events that are causally related. The resulting analysis of a text’s microstructure can be summarized in a network representation with propositions represented as nodes and relations represented as connections between the nodes.

A text’s vertical organization derives from the fact that expressed and implied propositions are hierarchically related. The macrostructure represents these hierarchical relationships and thus captures the basis of the global coherence of a text. Formally, a macrostructure representation can be derived from a microstructure representation by the application of three macrorules. One rule deletes micropropositions that are not interpretation conditions of other propositions. A second rule creates new macropropositions by generalization from more specific micropropositions in the microstructure. Finally, a third rule constructs new macropropositions by replacing sequences of micropropositions that denote more complex events. The three macrorules are applied recursively to create new macropropositions that dominate increasingly larger sections of the text. Intuitively, it is useful to think of successively higher levels of a macrostructure as corresponding to increasingly concise summaries of the text.

2. Constructing a Macrostructure Representation While Reading

2.1 Some Cognitive Obstacles to Macroprocessing

A macrostructure represents the ‘gist’ of a text so an adequate representation of a text’s macrostructure is essential to text comprehension. However, even in circumstances where a text is well written and the reader has the requisite vocabulary and background knowledge to understand it, there are several impediments to processing a text’s macrostructure. One is that a written text is an incomplete representation of the author’s intended communication. Authors omit from the written text information that they assume their audience can compute from the information provided. Sometimes, they will explicitly state a macroproposition; other times, they will leave it to the reader to infer (generalize or construct) the relevant macroproposition from the more specific information that is explicitly provided.

A second obstacle to macroprocessing is that related macropropositions are often widely separated in a text. This is because the written text is linearly organized whereas the underlying macrostructure is hierarchically organized. In many models of discourse comprehension (e.g., Kintsch and van Dijk 1978; see also Text Comprehension: Models in Psychology), this fact creates problems for identifying the relationship between text topics because of readers’ memory limitations.

Finally, the identification or construction of a macroproposition can involve the reduction of large amounts of information. For example, a page-long description of a sequence of events might ultimately be summarized by a single macroproposition (e.g., John ate at an Italian restaurant). Again, construction of the appropriate macroproposition may place a heavy burden on the memory abilities of the reader.

2.2 Monitoring the Macrostructure During Reading

How do readers cope with the cognitive demands of macroprocessing? There are several empirical demonstrations that readers are sensitive to topic changes and pay more attention to sentences that communicate macropropositions. These important junctures in the underlying macrostructure are signaled in the written text in a variety of redundant ways.

An author can use a variety of explicit signals to alert readers to macrorelevant information (Lorch 1989). Headings, overviews, and other devices emphasize the main topics and their organization within a text and have consistent benefits on reader’s memory for a text’s macrostructure. In addition, function indicators (e.g., ‘It is important to note …’; ‘To sum up …’) can be used to indicate the relevance of specific statements within a text (van Dijk 1980) with the consequence that readers attend more carefully to the cued information. An author can also be very explicit about a change of topics (e.g., ‘In contrast to …’; ‘Consider another issue’).

The underlying macrostructure is also reflected in the distribution of information within a text. A topic is introduced and elaborated, then a new topic is introduced. The identity of the currently relevant discourse topic is usually reflected in the topic-comment structure of the sentences that elaborate the topic (Kieras 1981). Thus, repeated sentence topics provide a basis for identifying the relevant discourse topic. When a change of topic occurs, several writing conventions aid the process of recognizing the change and identifying the new topic. A change of topic often corresponds with a break in local text coherence which, in turn, is often indicated by paragraph structure. The identification of topic transitions may be further facilitated by the use of marked syntactic constructions that shift the reader’s focus to the new topic. Finally, after a transition has been made, the identification of a new topic is facilitated by the convention that the initial sentence of a new paragraph often corresponds to a macroproposition (Kieras 1981).

2.3 Representing Macropropositions and their Relationships

Although authors occasionally denote the macrorelevance of a statement explicitly, macropropositions are frequently left implicit. In that event, readers are expected to use their prior knowledge and their understanding of information presented in the text to infer relevant macropropositions. Even when a macroproposition is explicitly communicated, readers usually must determine its macrorelevance. Syntactic cues, initial position in a paragraph, and other cues may serve as heuristics in identifying the possible macrorelevance of a statement. However, the macrorelevance of a proposition must be erified by determining that it occupies a superordinate position with respect to other propositions in the text (i.e., it is elaborated by more specific propositions).

When a macroproposition is identified or inferred, its place in the macrostructure must be determined. This involves not only determining its semantic relationships to subordinate micropropositions, but also determining its relationships to previously identified macropropositions. Again, the author sometimes explicitly communicates these relationships; however, readers are often left to infer how a newly identified macroproposition is related. This involves first identifying potentially related information, then evaluating its relationship to the new macroproposition.

There is very little definitive research on how readers infer implicit macropropositions and how they determine unstated relationships between macropropositions. However, such macroprocessing surely entails interactions between (a) cognitive processes that operate automatically in response to information presented in the text and (b) processing strategies that are invoked by readers to meet their particular reading goals. Successful macroprocessing requires that a reader be able to access relevant background knowledge and information in the text representation. There is substantial empirical evidence of powerful, automatic memory retrieval processes that serve this purpose (Kintsch 1988, Myers and O’Brien 1998; see also Knowledge Acti ation in Text Comprehension and Problem Sol ing, Psychology of). The involvement of strategic processes is demonstrated by evidence that the reading task modulates readers’ macroprocessing: readers are sensitive to macrorelevant information even when the reading task renders such information irrelevant, but they pay even greater attention to the macrostructure when it is relevant to their reading goal (e.g., summarization).

3. Using a Macrostructure Representation

3.1 Influences on Comprehension Processes During Reading

The macrostructure representation a reader constructs during reading has a privileged status with respect to processing of subsequent text. Macropropositions are more accessible than micropropositions so they are particularly likely to serve as the context for the interpretation of new information during reading (McKoon and Ratcliff 1992). This is true in two respects. First, because a recently identified macroproposition is likely to be the most available information when a new sentence is read, it will be the first context consulted as the reader attempts to establish the local coherence of the new information (Kintsch and van Dijk 1978, Kintsch 1998). Second, because macropropositions are typically elaborated in a text, macropropositions should generally be relatively accessible from the text representation even when they were not recently processed. Thus, macropropositions should also be relatively available as a context for establishing global coherence relations (Myers and O’Brien 1998). Indeed, there is substantial empirical evidence that macrorelevant information (e.g., titles) can greatly influence how readers interpret statements and organize their text representations.

3.2 Influences on Memory for Text

A macrostructure representation can also play a prominent role in memory for a text after it has been read. For the same reasons that individual macropropositions tend to be relatively accessible during reading, they are also memorable after reading. As elaborated information, there are many retrieval routes to macropropositions in the text representation. Further, in tasks that emphasize reconstructive processes (e.g., free recall), readers often use their macrostructure representations strategically as a retrieval plan to cue their memories for text content.

4. Future Directions

There are several directions in which future research should develop. First, there is a need for new methods of studying macroprocessing as they occur on-line (i.e., during reading). Although there are many empirical studies demonstrating that readers are sensitive to macrorelevant information as they read, very few studies address the nature of the cognitive operations involved in verifying macropropositions, inferring macropropositions, or determining the relationships among macropropositions. Most on-line methods require great precision in specifying the text conditions eliciting a given process. Because macroprocessing often involves the integration of several propositions and depends heavily on a reader’s background knowledge, such theoretical precision may not be possible. Thus, methods are required that are sensitive to on-line processes, but which do not require pinpointing a particular text event as the stimulus eliciting some process.

In addition to the need for new research methods, theoretical advances are needed in several areas. More detailed models are needed of the cognitive processes that underlie the construction of a macrostructure. The models should identify the text conditions that trigger the processes. More detailed models are also needed of how the availability of a macrostructure representation influences subsequent processing. Models are also needed to explain how the construction of a macrostructure interacts with a reader’s knowledge of conventional text structures (e.g., narrative; empirical research report) or what van Dijk and Kintsch (1983) call the superstructure of a text. Finally, theoretical advancement on all fronts depends on the development of ways to model the involvement of readers’ background knowledge in all aspects of macroprocessing. Current theory ascribes an extensive role to background knowledge, but promising methods to model its role have only recently begun to take shape (Kintsch 1998).

Bibliography:

1. Kieras D E 1981 Component processes in the comprehension of simple prose. Journal of Verbal Learning Behavior 20: 1–23
2. Kintsch W 1988 The role of knowledge in discourse processing: A construction–integration model. Psychological Review 95: 163–82
3. Kintsch W 1998 Comprehension: A Paradigm for Cognition. Cambridge University Press, New York
4. Kintsch W, van Dijk T A 1978 Towards a model of discourse comprehension and production. Psychological Review 85: 363–94
5. Lorch R F Jr 1989 Text-signaling devices and their effects on reading and memory processes. Educational Psychological Review 1: 209–34
6. McKoon G, Ratcliff R 1992 Inference during reading. Psycho- logical Review 99: 440–66
7. Myers J L, O’Brien E J 1998 Accessing the discourse representation during reading. Discourse Processes 26: 131–57
8. van Dijk T A 1980 Macrostructures: An Interdisciplinary Study of Global Structures in Discourse, Interaction, and Cognition. Lawrence Erlbaum, Hillsdale, NJ
9. van Dijk T A, Kintsch W 1983 Strategies of Discourse Comprehension. Academic Press, New York