Construction integration model

from Wikipedia, the free encyclopedia

The construction-integration model (English construction integration model ) is a psychological process model of text comprehension, the Walter Kintsch and Teun van Dijk back. The model can therefore be classified at the intersection of psycholinguistics and cognitive science .

All of the following statements refer to reading and understanding a written text as well as listening to and understanding a discourse.

Propositional meaning representation

A basis of the model is the concept of the proposition as a representation of meaning . Propositions consist of a predicate and one or more arguments that are related by the predicate. Propositions are noted as follows:

  • P1: LOVE (HANS, INGE)

In front of the colon it is indicated which proposition it is, so that it can be referred to more easily in the later analysis. The predicate (here: LOVE) relates the arguments HANS and INGE to each other. P1 thus expresses the following sentence in natural language: S1: “Hans loves Inge.” It should be noted that predicates and arguments are concepts, i.e. meanings, referents or denotations of words. For this reason, they are usually written in capital letters or small caps.

Besides concepts, other propositions can also be arguments. The natural language sentence: "Because Peter has a flat tire, he has to wait." Is thus phrased propositionally as follows:

  • P2: BECAUSE (P3, P4)
  • P3: HAVE (PETER, FLAT TIRED)
  • P4: MUST (PETER, P5)
  • P5: WAIT (PETER)

or

  • P6: BECAUSE (HAVE (PETER, FLAT TIRED), HAVE TO (PETER, WAIT (PETER)))

P2 here links the proposition P3 and P4 causally and P5 is embedded in P4. So there is an argument overlap here. P6 is just a more compact and confusing way of displaying P2 to P5.

The Construction Integration Model of Text Understanding

It is assumed that the reader, when reading a text , translates it into a propositional representation . A distinction is made between the construction phase and the integration phase.

Construction phase

In the construction phase, the meaning of the text is extracted in propositional form. A propositional network is created in which the propositions are connected to one another via argument overlap (see above). This network does not have to be fully connected to begin with. This means that there can be several small, independent networks . The reader will now try to create a coherent , i.e. a uniform, fully connected network. In most cases, this requires a large number of associations and inferences (conclusions) from the knowledge of the reader. The following example should clarify this:

The mini text

  • “Peter had a flat tire. After he had changed the bike, he could continue. "

explicitly contains the following propositions:

  • P7: HAVE (PETER, FLAT TIRED)
  • P8: AFTER (P9, P10)
  • P9: CHANGE (PETER, RAD)
  • P10: CAN (PETER, P11)
  • P11: DRIVE ON (PETER)

However, these propositions would not suffice to create a coherent representation of this text. For this purpose, inferences and associations would have to be generated such as: "It is not Peter who has a flat tire, but the vehicle in which Peter is sitting." And: "A vehicle can no longer drive if it has a flat tire."

The associations and inferences that are generated during the construction phase are unguided, in part chaotic and idiosyncratic , so only understandable for the understanding person. The design process is therefore a classic bottom-up process.

After these inferences have all been drawn and the corresponding propositions have also been integrated into the propositional network, the integration process takes over.

Integration process

The integration process takes over from the construction phase, to put it bluntly, a chaotic network of propositions in which important, unimportant and sometimes even contradicting things are connected. Now a constraint satisfaction process begins , which adjusts the activation values ​​for the various nodes in the network, i.e. the propositions, so that all constraints (restrictions, contradictions, etc.) are taken into account - the network is integrated. Unimportant propositions can be omitted, contradicting information can be corrected and contextual influences (one of several top-down components in the model) can come into play. The Constraint Satisfaction process runs over a plurality of cycles in which the principle of the activation spreading the activation values are adjusted for the nodes (spreading activation).

annotation

Construction and integration do not take place one after the other, but rather simultaneously. That is, while a word is being processed, the meaning representation that has already been built up to this point is integrated. The new word is built into the network and the integration process continues while the next word is already being processed.

Text base and situation model

In terms of the representation structures, a theoretical distinction can be made between the text base and the situation model. It should be noted that this is a theoretical distinction as these constructs are not independent structures. However, for the purpose of analysis they can be separated.

The text base represents all propositions explicitly mentioned in the text. This text base can therefore almost never be completely coherent, since a text in which really all information is explicitly mentioned is inconceivable (see also the mini-text above). Associations and knowledge labs will always be needed to create a coherent representation of a text in the mind of the understanding one.

Every enrichment of the text base with knowledge elements of the reader leads to a situation model. This theoretical level was included in the Construction-Integration-Model in 1983, which - as Kintsch explicitly writes - is due to the great influence of the theory of the mental models of Johnson-Laird .

Situation models therefore consist of the text base as well as the enrichment of the reader's knowledge. It should be noted that situation models no longer have to be represented purely propositionally. Pictorial or process-like situation models are also conceivable, this depends, among other things, on the aim of the reader. For example, if the reader wants to solve a word problem, he will also produce a so-called problem model that represents the arithmetic operations that can lead to the solution of the problem. A word task can thus be understood in terms of content (on the level of the text base), the reader can also have an idea of ​​what is described in the task (situation model). But that does not mean that he also knows what has to be done arithmetically in order to arrive at a solution to the problem (problem model).

Micro and macro structure

One distinction that is made orthogonally to that between the text base and the situation model is that between micro- and macro-structure. The microstructure consists of the micropropositions. These are the propositions that can be extracted from the text and the associations that come from the knowledge of the reader. These are in no way organized hierarchically.

This is exactly what happens in the macrostructure, which gives the microstructure a hierarchical structure (or organization). For this purpose, so-called macro propositions are constructed using three macro rules:

  • Selection / deletion : Any proposition that is not an interpretation condition for another proposition can be deleted
  • Generalization : If every proposition P1, P2, P3, ... of a set of propositions can be expressed by another proposition MP1, P1, P2, P3, ... can be replaced by MP1.
Example:
    • P12: LIKE (PETER, DALMATIANS)
    • P13: LIKE (PETER, BERHARDINER)
    • P14: LIKE (PETER, LONG-HAIRED DACKEL)
    • P15: LIKE (PETER, SHEEPDOGS)
P12, P13, P14 and P15 can each be interchanged with
    • MP1: LIKE (PETER, DOGS)
This is a generalization based on the knowledge that Dalmatians, St. Bernard dogs, long-haired dachshunds and shepherds are all dogs and so different that Peter probably actually likes all dogs. As can be seen here, macro propositions do not have to be formally and logically correct.
  • Construction : If the sequence of the propositions P1, P2, P3, ... can be expressed by another proposition MP1, the sequence P1, P2, P3, ... can be replaced by MP1.
Example:
    • P16: EXPAND (HEART MUSCLES)
    • P17: CONTRACT (HEART MUSCLES)
P16 and P17 can both be replaced by:
    • MP2: PUMPS (HEART)
This construction of a completely new proposition (more precisely: a macro proposition) takes place against the background of the knowledge of the reading person.

Which macro propositions and which macrostructure are / will be formed and when, is highly context-dependent. It depends on the nature of the text, the reader's expectations and goals, the external circumstances, and of course the reader's knowledge, which hierarchical organization assumes the representation of the meaning of the text in the mind of the reader.

However, there can be clear indications of the development of a very specific macrostructure in the text. This includes subheadings, structuring notes, statements highlighted as particularly important, etc.

However, a single reader can disregard these as well. Let us imagine the following scenario: We are reading a text about the economy, the government, the environment and the population structure of Brazil and Argentina. The text is structured in such a way that, for each of the individual areas, Brazil is discussed first and then Argentina. The text suggests the following macrostructure:

Macro structure1.gif

But if we are very interested in Brazil and not at all in Argentina, we will develop a different macrostructure, even though we are reading the same text. This could possibly look like this:

Macro structure2.gif

In addition, if we are more interested in Brazil, we can assume that the left half of the graph has a much richer representation (i.e. more associations and elaborations) than the right half.

Empirical studies

References to empirical studies on the construction integration model are given below with a brief comment. These represent only a small selection.

  • W. Kintsch, J. Keenan: Reading Rate and Retention as a Function of the Number of Propositions in the Base Structure of Sentences . In: Cognitive Psychology , 5, 1973, pp. 257-274.
    A classic empirical study of reading times and retention performance in relation to the number of propositions in a sentence. This study must be seen as a forerunner of the construction integration model.
  • W. Kintsch: Learning from Text . In: Cognition and Instruction , 3 (2), 1986, pp. 87-108.
    A study with two interesting studies on the situation model and the text base, which in Experiment 2 even raises an interesting gender aspect.
  • Till et al .: Time course of priming for associate and inference words in a discourse context . In: Memory & Cognition , 16, 1988, pp. 283-298.
    Very interesting study on the timing of the construction integration process, here the so-called activation-selection-elaboration approach is presented.
  • F. Schmalhofer, MA McDaniel, D. Keefe: A Unified Model for Predictive and Bridging Inferences . In: Discourse Processes , 33 (2), 2002, pp. 105-132.
    An interesting study of inferences from the knowledge of the reader.
  • S. Mehl: Fiction and Identity in the Esra Case. Multi-disciplinary processing of legal proceedings . LIT Verlag, Berlin u. a. 2014, chap. 4.
    A comprehensive empirical application of the model to the Esra case     (Roman) .

See also

literature

  • I. Barshi: Message length and misunderstandings in aviation communication: Linguistic properties and cognitive constraints. University of Colorado, Boulder 1997.
  • W. Kintsch: The representation of meaning in memory. Erlbaum, Hillsdale 1974.
  • W. Kintsch: The Role of Knowledge in Discourse Comprehension: A Construction-Integration Model . In: Psychological Review , 95 (2), 1988, pp 163-182.
  • W. Kintsch: Comprehension: A Paradigm for Cognition . Cambridge University Press, Cambridge 1998.
  • W. Kintsch: An Overview of Top-Down and Bottom-Up Effects in Comprehension: The CI Perspective . In: Discourse Processes , 39 (2 & 3), 2005, pp. 125-128, doi: 10.1080 / 0163853X.2005.9651676
  • W. Kintsch, JG Greeno: Understanding and solving word arithmetic problems . In: Psychological Review , 92 (1), 1985, pp. 109-129.
  • W. Kintsch, VL Patel, KA Ericsson: The role of long-term working memory in text comprehension . In: Psychologia , 42, 1999, pp. 186-198.
  • W. Kintsch, T. Van Dijk: Toward a model of text comprehension and production . In: Psychological Review , 85, 1978, pp. 363-394.
  • W. Kintsch, JJ Yarborough: Role of rhetorical structure in text comprehension . In: Journal of Educational Psychology , 74, 1982, pp. 828-834.
  • T. Van Dijk, W. Kintsch: Strategies of discourse comprehension. Academic Press, New York 1983.
  • Ilka Hanne Unsöld: The formation of inferences in the cognitive processing of media texts. A study on children and adults . Publishing house Dr. Kovac, Hamburg 2008, ISBN 978-3-8300-3852-8 .