Text Processing and Text Comprehension according to Walter Kintsch

Contents

1. Introduction

2. Important Components of Kintsch’s Model
2.1 Propositions
2.2 The Text Base
2.3 The Situation Model
2.4 Inferences

3. Earlier Models

4. The Construction-Integration Model
4.1 The Construction Process
4.2 The Integration Process
4.3 Evaluation of the Model

5. Experiment
5.1 Description of the Original Experiment
5.2 My Realization of the Experiment
5.3 Results of my Experiment
5.4 Comparison of the Results
5.5 Discussion

6. Conclusion

7. References

1. Introduction

Reading is a part of our daily life. It enables us to get information, for example when we read a newspaper, or it is just for entertainment. Once we have learned to read, we are not able to stop it anymore. If we see a text, we read it automatically and know what it means. But how is it possible that we understand the meaning of a text? What is going on inside our brain while we are reading? And how are we able to remember and recall something from a text?

These are central questions the text processing research concentrates on. In order to find an answer to them, researchers have different approaches. One of them is the construction-integration model by Walter Kintsch, which has its origin in several earlier models of processing.

The main field of application for this model is instruction. The results of research on learning can be used to create new instruction methods, which facilitate the process of learning and advance the ability to remember what has just been learned.

My term paper is going to concentrate on Kintsch’s construction- integration model and its assumptions. It is structured into two parts. The first part gives an overview of the theory. To be able to understand the model, I will initially describe its different components, namely: propositions, the text base, the situation model, and inferences (chapter 2). Then, I will briefly dwell on Kintsch’s earlier models (chapter 3). Afterwards, I will explain the model itself and give a short evaluation of it in chapter 4. The second part of the term paper consists of my imitation of an experiment on the existence of propositions, which was originally carried out by Gail McKoon and Roger Ratcliff (chapter 5).

2. Important Components of Kintsch’s Model

2.1 Propositions

In 1978, Kintsch and van Dijk developed their first processing model, which is based on the assumption that during reading, a text is transformed into several semantic units, called propositions (cf. Grabowski 1990: 29). One clause becomes one proposition. A word meaning is represented by one atomic proposition (cf. van Dijk & Kintsch 1983: 14). A proposition consists usually of “a predicate and one or more arguments” (van Dijk & Kintsch 1983: 113, accentuation taken from the original).

Figure 1

illustration not visible in this excerpt

Source: van Dijk & Kintsch 1983: 14.

The example above shows a sentence (1), and its proposition (2). Each line is an atomic proposition, for instance FASCISTS. The xn replace atomic propositions. Therefore, (ii) could also be written HAVE WON (FASCISTS, ELECTIONS). The whole text is transformed in this manner. Thereby, the transformation is not done arbitrarily but it follows several rules, which have been listed in Construction and use of a propositional text base by A. Turner and E. Greene in 1977.

However, the propositions on their own do not help us to understand the text as a whole. Comprehension is only possible with a certain degree of coherence. According to the construction-integration model, this coherence is established by arranging the propositions into a propositional net to create order (cf. Kintsch 1988: 166).

Such a net consists of several propositions which are interconnected with each other according to their meaning. But I will return to this later on while describing the construction-integration model.

After we know what propositions are, there is still a question to be answered. Why do we need propositions at all? Is it not easier to use normal language? Kintsch explains it this way:

[P]ropositions are designed to capture those semantic relations that are most salient in text comprehension, whereas natural language serves many purposes other than the expression of meaning and hence is often less suited for our purposes than a representation that is focused on meaning (Kintsch 1998: 69).

What Kintsch means by saying “natural language serves many purposes other than the expression of meaning” is that propositions, among other things, do not contain information about “tense, aspect, voice, and the determinacy of the nouns” (Graesser 1997: 169). They concentrate solely on the meaning, which is exactly the information needed for a text analyses.

On the other hand, the transformation of a text into propositions is not always conducted without problems. As proposition building takes place simultaneously to the reading process, and the information for the creation of propositions is not always immediately available, it is possible that false starts or incomplete attempts are made (cf. Kintsch 1988: 166). Instead of going back and correct the wrong propositions, they just disappear in the integration process, which will be explained later.

Although propositions seem to be a useful and appropriate tool for text analyses, they show certain disadvantages, too. “[F]or longer expository texts, there is a problem with the method’s economy” (Grabowski 1992: 22). The construction takes much time, it is a lot of work and the result is even more complex – and mostly longer – than the original text (cf. Christmann 1989: 1). But nevertheless, propositions have proved to be a simple, and successful tool for the analysis of short texts.

2.2 The Text Base

The text base can be described as a linguistic representation of a text in memory. It “consists of those elements and relations that are directly derived from the text itself” (Kintsch 1998: 103).

As I have already mentioned earlier, its construction process functions the following way: In order to achieve a coherent text base, the text has to be transformed into propositions which on their part have to be connected to form a propositional net. Concerning the construction of a coherent text base, it is also important to know that readers try to “establish coherence as soon as possible, without waiting for the rest of the clause or sentence” (van Dijk & Kintsch 1983, 15). They want to understand the text quickly, and therefore, they try to make sense of it right away.

But how are the propositions connected to each other? In a previous model, T. A. van Dijk and Walter Kintsch proposed that propositions are connected according to the principle of argument repetition (cf. van Dijk & Kintsch 1983: 43). That means: If an argument appears again in another proposition, this proposition is added to the previous one. “Coherence is thus reduced to referential ties, which is certainly an oversimplification, but attractive in its simplicity” (van Dijk & Kintsch 1983: 43). Later on, van Dijk and Kintsch realized that this principle is, indeed, too simple and has to be reconsidered.

If we talk about coherence, we have to consider that there are two kinds of coherence: the local and the global coherence. The local coherence is the detailed meaning of the text, whereas the global coherence is more like a summary of the text. According to this division, van Dijk and Kintsch differentiate the micro- and the macrostructure of a text. Whereas the microstructure, which is the text base, is associated with the local coherence, the macrostructure is associated with the global coherence (cf. van Dijk & Kintsch 1983: 52).

The macrostructure has its own propositions, the so-called macropropositions. They are constructed from the micropropositions with the help of certain macrorules leading to different levels of macro- propositions (cf. van Dijk & Kintsch 1983: 52). The highest level would ideally be only one macroproposition. The macrorules, identified by T. A. van Dijk, are deletion, generalization, and construction (cf. van Dijk & Kintsch 1983: 190).

The macropropositions form a network, just like the micro- propositions, and thus create coherence. It is necessary to notice the existence of a macrostructure because this structure helps us to understand a text.

[H]ow easy a text is, how well readers understand it, how well they can remember it, what they learn from it – all this is strongly dependent on a successful macrostructure (Kintsch 1998: 180).

2.3 The Situation Model

So far, we have solely looked at the role of the text during reading. However, in order to comprehend a text, it is not sufficient to construct and process a coherent text base. There is another crucial unit in the text comprehension process, the situation model. It is an addition to the original processing model of van Dijk and Kintsch (1978), and takes into account that our previous knowledge contributes to text comprehension, too. “A situation model is, therefore, a construction that integrates the textbase and relevant aspects of the comprehender’s knowledge” (Kintsch 1998: 107). Hence, there are two units: the text base and the situation model. One might think that these are two entirely different representations in the brain. However, this assumption is wrong. They are parts of only one representation, and the distinction is exclusively made to highlight the fact that information from the text is not enough.

We distinguish between textbases and situation models not because they are somehow different mental objects. On the contrary, there is a single, unitary mental representation of a text (Kintsch 1998: 107).

According to the construction-integration model, the situation model has the form of propositions, just like the text base (cf. Kintsch 1988: 164f). It is saved in the long-term memory of our brain, and is activated during reading. The connection of text propositions and propositions from the situation model is called elaborated propositional net (see Figure 2).

However, not all text processing researchers agree with Kintsch. Johnson-Laird, for instance, does not believe that our knowledge is saved in the form of propositions. He rather has the opinion that we create spatial models, which he calls mental models (cf. Emmott 1999: 43f). In reaction to that, Kintsch admits that spatial models are indeed needed for several texts. Nevertheless, this must not be a contradiction to the assumption of a propositional form. “[I]t is possible to simulate the construction of spatial situation models with a propositional network” (Kintsch 1998: 204).

2.4 Inferences

In the previous sections, I explained that we are able to establish coherence with the help of the text base and the situation model. Nevertheless, it might be the case that these two components are not enough to create coherence. As we will see afterwards, in the integration process of the construction-integration model appropriate propositions are separated from inappropriate ones (cf. Kintsch 1988: 168). During this process, it is possible that certain propositions are sorted out which are, however, appropriate (cf. Kintsch 1988: 168). As a result, the reader is no longer able to establish coherence. There are several possibilities to solve this problem. Primarily, one can make a reinstatement, that is old propositions are again retrieved from long-term memory to fill the coherence gap. Another solution is a re-organization of the propositional net. And the third possibility is the creation of inferences (cf. Grabowski 1991: 50).

[...]