Does reading comprehension impair the ability to write

Additional information

The theoretical introduction presented here comes from a publication from 2002. It formed the basis for the construction of the reading comprehension test ELFE 1-6. If you are interested in more recent publications and further information, we recommend the book

Lenhard, W. (2013). Reading comprehension and reading skills: Basics - Diagnostics - Promotion. Stuttgart: Kohlhammer.

"The written language is one of the most fascinating achievements of mankind. The process of understanding when reading is very complex and is determined by many influencing factors. In the book, these factors, their development and their interaction are systematically worked out. This opens up a perspective on the question of At which points diagnostics and support can start. This book offers the reader an insight into theories and models and shows current research results and developments in the German-speaking area since the first PISA study. In addition, it illuminates the question of where the special needs are weaker The reader lies, describes diagnostic options and deals with systematic and evidence-based funding options. "



1 Introduction




"A knight sô gelêret what
since he was reading the books
swaz he wrote on it vant;
what Hartman called
dienstman what he ze Ouwe. "

(Hartmann von Aue, 1195,
quoted from Mettke, 1974, p. 9)
Translation:

"A knight was so learned that he read everything that he found written in books. This knight was called Hartmann, and was a servant in Aue." (free translation by the author)

With his work "Der arme Heinrich", the poet Hartmann von Aue created "the classic novella of high feudalism", according to Mettke (1974, p. 1). Interestingly enough, at the beginning of his work, von Aue praised his exceptional erudition. He points out that he even reads the book himself, even though he is a knight.

Such self-praise would be completely unfounded nowadays, because reading and writing are natural cultural skills in industrialized countries. Hardly any adult in this country does not have at least rudimentary forms of these techniques (Klicpera & Gasteiger-Klicpera, 1995, p. 4). In order to better understand Hartmann von Aue's quote, one must be aware of the fact that the mastery of the cultural techniques of reading and writing in the medieval era was exceptional. At that time, writing was considered the epitome of truth. In this sense, the above quote has the following meaning: "What I am about to report is really true. I read it myself!".

Even poets and minstrels were often unable to speak the written language. For example, Walther von der Vogelweide does not have any of his own records, not even written evidence of its existence (Lindner, 1968, p. 47). Hugo von Trimmberg wrote about him around 1300: "Her Walther von der Vogelweide, swer dez vergaeze, I suffered" (Hugo von Trimmberg, quoted from Lindner, 1968, p. 47). All of his works were passed down orally and only later written down by others.

"For a long time reading and writing had been reserved for a small part of the population. It was not until the invention of the printing press that the material basis for wider accessibility of writing was created" (Klicpera & Gasteiger-Klicpera, 1995, p. 4). In our language area, the spread of the written language is closely linked to the Reformation and its reference to the Bible as the source of truth, according to Klicpera and Gasteiger-Klicpera.

Further evidence of the rather young nature of the outstanding importance of reading and writing in our society is the fact that eleven centuries passed from the first testimony of the German language in the 8th century to the creation of a binding set of rules for German orthography in 1901 (Niederhauser, 1995) . Goethe, for example, does not seem to have thought much of a uniform spelling: "I write a word with three different orthographies, and what the bad habits may be, of which I am very well aware and which I only fight against in an extreme emergency overcome me "(Goethe, quoted from Niederhauser, 1995).

Practically everyone in our culture today is confronted with written material. On average, every adult in our culture spends 2œ hours reading, especially in the context of their job. Around 90% of all workplaces require the use of written material, according to Klicpera and Gasteiger-Klicpera (1995, p. 5). First and foremost, with regard to reading, this is not a mechanical handling of letter sequences, but rather meaningful, understanding reading.

During reading socialization, which begins long before school lessons, children go through a development process from the beginning of the written language acquisition to the goal-oriented use of texts, which requires the complex interaction of various sub-skills (Artelt, Stanat, Schneider & Schiefele, 2001). To do this, the child not only has to acquire new, complex skills, but also network them with previously acquired language skills. In the following section, the sub-processes are first examined, the interaction of which enables meaningful reading. The following chapter deals with developmental psychological models of the acquisition of reading comprehension skills.







2. Levels of reading comprehension

According to intuitive everyday ideas, reading is a rather passive reception of information and meaningful content contained in the text (Christmann & Groeben, 1999). According to this view, which also underlies early communication models from the heyday of behaviorism, information is first encoded by the writer and then decoded again by the reader. According to this concept, decoding and encoding are largely mirror-image processes (see Fig. 1).



illustration 1
The general scheme of information transmission by Shannon and Weaver (1949, from: Hoffmann, 1993, p. 16)


In contrast, cognitive psychology and experimental reading research emphasize that reading (or speech reception in general) is a highly active process of dealing with the respective content (Klicpera & Gasteiger-Klicpera, 1995, p. 133; Artelt, Stanat, Schneider & Schiefele, 2001). The cognitive constructivity of language understanding and the interaction between (text) information and the recipient's world knowledge is particularly evident in ambiguous messages, such as the following:

  1. "CDU women demand more weight"
  2. "We are looking for roofers and helpers with immediate effect. Opportunities for promotion available."
  3. "My trouser shop is always open. Of course only from 8 a.m. to 6 p.m."

The process of reconstructing meaningful content not only consists of connecting linguistically conveyed information of individual words with the surrounding information - the other words, parts of sentences, and sentences - but interpreting them on the basis of individual prior knowledge (Christmann & Groeben , 1999). In the following, the processing steps required to understand a text are highlighted, whereby a distinction is made between different levels of complexity: word comprehension, sentence comprehension and text comprehension.

When reading, the focus is initially on factors that influence reading comprehension at the word level. These include the child's vocabulary, the ability to decode individual words, the understanding of the meaning of the word and the knowledge of how this meaning is modified by the respective context (Klicpera & Gasteiger-Klicpera, 1995, p. 133). If the child succeeds in interpreting the meaning of the word on the basis of the context, then it has already established the link to the sentence level. Now the grammatical structure must also be taken into account. Finally, in order to understand a whole text, it is necessary to relate the information from different sentences to one another. At this level, metacognitive skills such as schema-guided text comprehension are becoming increasingly important.






2.1 Models of visual word recognition

Recognizing the meaning of individual words plays a central role in the main goal of reading - understanding the content of the written material. This is also evident in the fact that a word is fixed during the reading process and then a saccade is used to jump to the next word. The reading process is essentially a word for word reading, at least for the experienced reader (Klicpera & Gasteiger-Klicpera, 1995, p. 13).

A fascinating effect, which was already demonstrated towards the end of the 19th century, underscores the importance of the word as a written unit: Cattel observed that letters within words are more quickly perceptible than letters within nonsensical sequences of letters or even single letters - a phenomenon that comes from it was described as the word superiority effect (Klicpera & Gasteiger-Klicpera, 1995, p. 14 f.). Towards the end of the 1960s, Reicher took up this effect again and examined it again using improved research instruments. The explanation of this and other subordinate effects has since been one of the test criteria applied to models of visual word recognition (Jacobs & Grainger, 1994). Another interesting effect in this area is the Stroop effect.

There are numerous such models, which differ significantly in format and complexity. Jacobs and Grainger (1994) report a total of 15 that were created between 1969 and 1994. In view of this diversity, I would like to limit myself to a few essential aspects and therefore single out three different models that have had a particular influence on research and represent different research approaches. These are the Logogen model according to Morton (1969), the activation model according to McClelland and Rumelhart (1981) and the dual route theory according to Coltheart (1978).


2.1.1 The Logogen model

The first process models of reading and writing were developed by Morton, who presented his first work on the word recognition process during reading as early as 1969 (Klicpera & Gasteiger-Klicpera, 1995, p. 97). He developed this design further and in 1979 published his "Logogen-Modell" (Graf, 1994, p. 46), which also includes ideas about the processes that take place when writing (see Fig. 2).


Figure 2
Morton's logogen model in its revised version according to Ellis (1982, quoted from Klicpera & Gasteiger-Klicpera, 1995, p. 98). The abbreviations next to the connecting arrows indicate the type of information passed on (visual, acoustic, grammatical, semantic, phonemic, lexical, as a motor pattern of graphemes and kinesthetic).



Many of the designs that have emerged since then are based on Morton's model or at least have been significantly influenced by it (Graf, 1994, p. 46). Since these models are in the tradition of information processing theory, their rough structure typically follows the scheme "Input -> Processing -> Output". Various input information, such as spoken language or writing, can serve as input, which is further processed in various lexical and non-lexical ways. The output consists of language (when reading) or writing (when writing) (Klicpera & Gasteiger-Klicpera, 1995, p. 97 ff.). Logogenes are units of language that correspond to the mental representations of word meanings. If certain information - e.g. a certain word - reaches the cognitive system, the associated logogen is activated. If the activation exceeds a certain threshold, the word is recognized.

Similar to the dual route theory according to Coltheart (1978), the model not only provides lexical processing options for words but also the conversion of language into writing using the phoneme-grapheme correspondence rules (GPK rules). However, this strategy only plays a subordinate role here, according to Klicpera and Gasteiger-Klicpera (1995, p. 99 f.).


2.1.2 The Activation Model (Interactive Activation Model)

A model which is able to explain the word superiority effect and which overcomes various weaknesses of previous models was proposed by McClelland and Rumelhart (1981). They assume that both basic graphic elements (dots and lines that make up letters) as well as letters and words are represented in the memory and organized in the form of a network. The word recognition process begins with the inclusion of the graphic elements. These features lead to characteristic activation patterns in the neural network: those letters that correspond to the features are activated by means of excitatory connections, the others are inhibited by means of inhibitory connections. Similarly, the activation patterns at the letter level lead to the activation of the associated word when a certain threshold value is exceeded (see Fig. 3).


Figure 3
Word recognition process in the activation model (McClelland & Rumelhart, 1981)



The word superiority effect can be explained according to Christmann and Groeben (1999) on the basis of the activation model as follows: The presentation of a letter embedded in a word leads to activation on all three levels, the presentation of non-words or individual letters, however, only on the property and letter level. The stronger activation, in turn, makes detection easier.

In 1989, Seidenberg and McClelland presented an expanded version of the model, the "parallel-distributed-processing (PDP) connectionist model". It is completely implemented as a network model and is available in the form of a computer simulation (see Fig. 4). The network described in this model consists of 400 orthographic input nodes, 200 so-called "hidden units" and 460 output nodes that describe the phonetic realization of the written word. In contrast to the model by McClelland and Rumelhart (1981), an input node no longer represents a single letter, but a letter triple (Coltheart, Curtis, Atkins & Haller, 1993). If a word is recorded via the visual system, it activates all associated triples of the network, i.e. in the case of the word father the triples "--V", "-Va", "Vat", "ate", "ter", " he "and" r-- ". Since the input nodes are connected to the nodes in the layer of hidden units, the activation patterns first propagate to the layer of hidden units and from there to the layer of phonological output nodes. The individual weighting of the individual connections represents the learning experience of the system. Since the output layer has feedback, the entire system goes through a larger number of cycles until a stable state is finally reached. The phonological realization of the written word is represented in the output layer of the network. Seidenberg and McClelland (1989) let the system learn the phonetic realization of 2,897 monosyllabic words in the English language, whereupon it was able to determine the correct pronunciation for 90% of the non-words. The PDP model also provides for the activation of word meanings by means of the phonological as well as the graphic realization of a word (see Fig. 5). Seidenberg and McClelland (1989) assume that word meanings are organized in a network similar to phonetic realizations of words and that they can be activated via intermediate layers of hidden units. However, it has not yet been possible to empirically confirm these parts of the model.

Figure 4
Sketch of the computer-modeled neural network by Seidenberg & McClelland (1989), from: Coltheart, Curtis, Atkins & Haller (1993).
Figure 5
Overall concept of the "parallel-distributed-processing-model" (after Seidenberg & McClelland, 1989). So far, only the blue-colored parts of the model have been implemented.

2.1.3 The two-way theory

One of the most prominent theories of visual word recognition is certainly the two-way theory or "dual-route theory", which was first proposed by Coltheart (1978) and has since been the subject of numerous publications. It is not only in line with numerous neurological findings from the field of acquired dyslexia, but has also significantly stimulated research into subgroups of reading and spelling disorders (see e.g. Morris et al., 1998; Stanovich, Siegel, Gottardo, Chiappe, & Sidhu , 1997; Fletcher et al. 1997).

The two-way theory describes two possible processing mechanisms in the reading process (Coltheart & Rastle, 1994) - one direct, lexical and one indirect, non-lexical. According to this theory, the first, direct way consists in an orthographic coding of the typeface, via which an entry in the mental lexicon is activated directly. The pronunciation of the word is therefore immediately accessible. In the case of words that are not in the lexicon, however, the second, indirect route is chosen. Since there is no entry in the lexicon here, the word must be read letter by letter using the grapheme / phoneme assignment. The phonological structure of a word is thus serially reconstructed by means of phonological recoding of letters or groups of letters.

Neurologists repeatedly report on patients in whom one of these two routes is not available due to organic brain damage. These studies show that both reading and writing strategies seem essential. If only the direct route is possible, affected people have immense difficulties in reading pseudo-words (Klicpera & Gasteiger-Klicpera, 1995, p. 193). Since such words have no content, they cannot be present in semantic memory. In this case, phonological dyslexia occurs. However, it is theoretically possible to compensate for this deficit in the grapheme / phoneme and phoneme / grapheme assignment with a sufficiently comprehensive semantic memory, since pseudo-words only play a subordinate role in everyday life. In this way, literacy can reach a sufficient level. If, on the other hand, the direct path via semantic memory is impaired in any way, all words must be read or written using the GPK rules (Klicpera & Gasteiger-Klicpera, 1995, p. 196 f.). The result is surface dyslexia, which manifests itself in difficulties with irregular words and homophones. In addition, numerous neologisms are produced in the process. Both types of reading and spelling difficulties have also been observed in children with reading difficulties, even if the question of the division of reading and spelling disorders into subgroups has not yet been adequately clarified (see e.g. Morris et al., 1998; Stanovich, Siegel, Gottardo, Chiappe, & Sidhu, 1997; Fletcher et al. 1997).

For about 10 years, analogous to the activation model (Seidenberg & McClelland, 1989), there has been a computer model with the designation "dual-route cascaded (DRC) model of word recognition", which is based on the dual-route theory (see Fig. 6). The DRC produces the synthesis of the two-way theory and the PDP model. Coltheart, Curtis, Atkins and Haller (1993) created a rule system that successfully integrates all of the GPK rules of the English language and that simulates the indirect route. They chose the PDP model as the module for the direct route (Seidenberg & McClelland, 1989). The DRC thus combines both hierarchical component models with a serial sequence and network models with parallel operation. The individual components of the DRC work bidirectionally and the system runs through so many cycles until a stable optimum is found (Coltheart & Rastle, 1994).


Figure 6
The dual-route cascaded model (DRC) according to Ziegler, Perry & Coltheart (2000). Word recognition can either take place directly with the inclusion of the semantic system, or indirectly by reading out the individual graphemes letter by letter. Arrows symbolize excitatory connections and round endings symbolize inhibitory connections. As in the PDP model (Seidenberg & McClelland, 1989), the inclusion of the semantic system - i.e. the word meanings - has not yet been implemented.



With the DRC it was possible to successfully simulate reading aloud on the computer (Coltheart, Curtis, Atkins & Haller, 1993; Coltheart & Rastle, 1994). In addition, the model has recently been transferred to the German language (Ziegler, Perry & Coltheart, 2000) and also demonstrated its astonishing performance here: It far exceeded the hit rate of the PDP model and cut on monosyllabic German words as well as regular (according to the GPK rules) even slightly better than German-speaking test subjects. In the area of ​​irregularly written foreign words, the error rate of the DRC was 24.2% compared to 19.5% in the test subjects. This achievement is all the more remarkable when you consider that the model only learned the phonetic realization of approx. 1,500 monosyllabic German words.


2.1.4 Summary

The DRC overcomes the weaknesses of the PDP model, which relies on the word to be read - or at least a sufficiently large number of similar words - already being learned and represented in the network. At the same time, it inherits its mental lexicon of word meanings.

However, it must be stated that both models have so far only been valid for the area of ​​reading aloud. At this point in time, words with more than one syllable are also not taken into account. For this reason, both models have not yet been able to break down words into their syllables and analyze them according to their morphemes. Despite these limitations, the DRC in particular has had astonishing success in reading aloud and in applying the model to impaired written language acquisition (see Coltheart, Curtis Atkins & Haller, 1993; Ziegler, Perry & Coltheart, 2000; Coltheart & Rastle, 1994).





2.2 Reading comprehension at word level

2.2.1 Concept formation

The phenomenon of understanding individual words is closely related to the way we mentally organize and represent concepts. Hoffmann (1993, p. 119 ff.) Lists seven different research approaches on the topic of concept formation:

  1. Terms as abstraction of characteristics relevant to classification
    Terms are formed through the abstraction of characteristics relevant to classification. A feature configuration is created in which all features are represented that are invariant and specific for the term. The formation of terms is based on the learning-dependent differentiation of relevant and irrelevant properties. These attributes represent both the sufficient and necessary constitutive properties of a term. They are referred to as "critical characteristics".

  2. Concept formation algorithms
    In the course of the learning process, regularities are abstracted that make it possible to decide whether an object clearly belongs to a term or not. This is usually done by linking binary properties using Boolean operators.

  3. Family resemblance and typicality of natural terms
    As a rule, the properties of natural concepts do not follow the all-or-nothing principle. Objects often belong more or less to a certain concept. For this reason it has been suggested that terms are fuzzy sets of objects. Objects belonging to one concept are more similar than the objects of different concepts. There are no relevant and irrelevant properties, but rather features that differ in their relevance.

  4. Terms as feature sets
    Contrary to what is listed under point 1, the distinction between relevant and irrelevant features is abolished here. Instead of linking characteristics, terms must have a sufficient number of characteristics with sufficient relevance.

  5. Concepts as prototypes
    According to this research approach, terms are stored as prototypes that are abstracted from everyday experience. The family resemblance of objects results from their similarity to the common prototype.

  6. Concepts as a set of experienced examples
    When determining the affiliation of an object to a term, this is first compared with all stored copies. The comparison causes an "echo" in the various specimens and the object is assigned to the specimen that responds with the strongest echo.

  7. Formation of terms in connectionist networks
    As already described in the chapter on the activation model, a connectionist network consists of input and output nodes that are linked to one another by a layer of "hidden units". The learning experience is reflected in the weighting of the links between the nodes with the "hidden units". A special activity configuration in the input node leads to an activity pattern in the output node, which symbolizes the association with a certain term.

In the current "Standard Model of Word Meaning" (Anderson & Nagy, 1991) it is assumed that word meanings can be described as sets of critical - that is, necessary and sufficient - properties. Despite many points of criticism and although practically all theoretical deductions from this theory can be challenged, according to Anderson and Nagy the theory described under 1.) corresponds to the current consensus among researchers:

"[Dissatisfaction] with the standard model is perhaps the most accurate representation of the current consensus. That is, scholars concerned with word meanings concur in their rejection of aspects of the standard model, but not on any model to replace it. ... To some extend, the standard model we define comes close to reflecting a commonsense view of word meaning that forms the implicit basis for much of the thinking that is done about word meanings, especially in applied areas such as reading and vocabulary instruction. " (Anderson & Nagy 1991).

Criticisms can already be made of the fundamental properties of the model. In the vast majority of terms, there is not a single property that is both sufficient and necessary. For example, consider the term robin. This is made up of other, more elementary terms, such as "bird", "can fly", "red neck" and "sings". The robin shares with other birds the characteristic that it can fly. Nevertheless, "can fly" cannot be a sufficient characteristic, as there are very many bird species that, in contrast to the robin, cannot fly, such as penguins or the New Zealand kakapoo. In the same way, we would still recognize a robin as a robin if, for example, it cannot fly due to a broken wing. The property "red neck", on the other hand, distinguishes the robin from ravens, buzzards and sparrows, but not from the strawberry bonnet (a parrot species) or from Oskar Lafontaine. "Sings" is a trait that distinguishes the robin from birds of prey; however, it shares this characteristic with Placido Domingo. And although most robins are certainly able to sing and practice this intensively, especially in spring, we would still be able to classify a robin sitting quietly on a branch as such without any problems. The problem outlined is further exacerbated by the fact that the potential feature lists of a term in turn contain terms that can be broken down in the same way. The creation of lists of critical characteristics is either impossible or very difficult for many terms.

The Standard Model continues to reach its limits when it comes to explaining the variability of meanings depending on the context (Anderson & Nagy, 1991). For example, the term "bank" can have a variety of meanings such as "seating", "financial institution" or "shallow water (sandbank)". The meaning of the respective term depends largely on the meaning of the surrounding words and sentences. In addition, words can also have transferred meanings, e.g. when they appear in the context of idioms. Whether these fixed verbal expressions are to be viewed as a single word from the point of view of concept formation must be highly doubted (Anderson & Nagy, 1991).

Another major point of criticism, in my opinion, is that the model does not explain how a term is associated with the spoken language or graphemic pattern. Anderson and Nagy (1991) marginally postulate that recognizing a word reveals its meaning immediately, but without specifying how this is done. However, they are of the opinion that, despite the weaknesses mentioned, there is no alternative to the standard model. I cannot agree with this opinion, but would instead refer to the point of view of Hoffmann (1991, p. 132). He currently sees networks with hidden units - such as the activation model by McClelland and Rumelhart (1981) - as the most complete representation of natural terms (see also Balota, 1990). Although the computer simulation of the PDP model derived from the activation model (Seidenberg & McClelland, 1989) has so far been limited to pure word recognition, the actual approach provides for linking the output nodes of the word recognition module to the semantic network. The activation could thus spread and immediately activate the content of meaning stored in the semantic network. Due to the interposed layer of hidden units, non-linear links would be possible in this way, which could also take into account the meaning facets and the transferred meanings of a word.

This would imply that if the DRC model (Coltheart, Curtis Atkins & Haller, 1993) is used as a basis, the recognition of a word meaning with the help of the serial reading of the word via the indirect route is not possible or only possible via detours - a fact that many didacticians use is shared (see e.g. Reichen, 1990).

Furthermore, research in the field of subliminal priming questions the strict serial sequence of recognizing a word in its phonetic structure and then determining its meaning (Balota, 1990). For example, if the word “cat” is presented subliminally (including subsequent masking), a test subject can identify the word “dog” more quickly shortly thereafter. If, on the other hand, "pen" is primed first, there is no increase in speed. The person thus activates concepts and associations, although he is neither able to consciously recognize the word nor grasp the phonetic structure. This finding suggests that early, pre-lexical access to the meaning of a written word is possible even if the visual stimulus cannot be consciously analyzed.




2.2.2 The influence of meaning and context on word recognition

The identification of letters within groups of letters presented by tachioscopy is - as described in Chapter 2.1 - much easier if the group of letters results in a meaningful word. Balota (1990) reports on experiments that make use of this word superiority effect and suggest that word meanings influence basic word recognition processes. Various research groups presented their test subjects with pseudo-words or words whose meaning was not known and measured the recognition performance for individual letters in tachioscopic presentation. Following this first experimental phase, one half of the pseudo-words were assigned an arbitrary meaning and the experiment was repeated. The word superiority effect now also appeared with these words.

If one considers the influence of the context on word recognition, an important distinction must first be made with regard to the type of effect (Oakhill & Garnham, 1989, p. 84): On the one hand, the context could be used by the reader to correct or avoid Reading errors are used, and on the other hand, an acceleration of word recognition by pre-activating similar meaningful content would be conceivable. Unfortunately, the empirical evidence in this area is inconclusive. While more experienced readers can use context more easily to predict the next word to come and correct reading errors, they seem to use this skill relatively rarely, according to Oakhill and Garnham. What is certain is that the embedding of words in a congruent context leads to an accelerated word recognition compared to the isolated presentation of words without any context. Incongruent context, on the other hand, presumably only slows down the recognition speed in children, but not in adults.



2.2.3 Summary

In summary, it can be said that the reading process at the word level is not exclusively a bottom-up process and only depends on basic processing mechanisms. Instead, top-down processes do play an important role.







2.3 Reading comprehension at sentence level

Recognizing individual words and their meaning does not guarantee understanding of the meaning of a sentence. The words in a phrase have to be related to each other and integrated into an overall structure.For this purpose, the analysis of both the semantic and the syntactic relations of the individual sentence elements is necessary (Christmann & Groeben, 1999).

Nominal and verb phrases are of particular importance (Klicpera & Gasteiger-Klicpera, 1995, p. 134). In German and English, a noun phrase comprises a noun that can be preceded by an article (determiner) and any number of adjectives (Pinker, 1996, p. 113 ff.). In the example (see Fig. 7) the word combination "the lively fairy" forms the noun phrase. The verb phrase, on the other hand, includes a verb and another noun phrase, i.e. the expression "eats fruit" in the example mentioned.




Figure 7
Syntactic analysis of the sentence "The lively fairy eats fruit" (Pinker, 1996, p. 115)


A noun phrase can appear in different places in a sentence, for example as the subject of the sentence, as an object of a verb phrase, as an object of a preposition (e.g. "for the fairy"), as a possessive phrase (e.g."the fairy's hat") and as an indirect object (e.g." give the fairy a biscuit "). However, the noun phrases must be linked with one another by means of proposition according to the rules of a language. In German, the gender and case must be observed differently than in English (e.g.,"the man", "of the man", "the man", "the man").

The actually given combination of words is called the surface structure, whereas the "superstructure" is called the deep structure of a sentence (Christmann & Groeben, 1999). Misunderstandings, involuntary comedy and word games arise when different deep structures are possible for a given surface structure. The following example should clarify this connection: "I shot the elephant in his pajamas." (Pinker, 1996, p. 117). The reader will interpret this sentence according to his knowledge of the world that "the pajamas" refers to the subject of the sentence. If, on the other hand, you add a second sentence ("How he got into the pajamas will always remain a mystery to me."), The previous sentence receives a different deep structure and therefore has to be reinterpreted.

The analysis of the deep structure of a sentence is referred to in German as in English as "parsing" and follows the so-called "garden path" model - according to Klicpera & Gasteiger-Klicpera (1995, p. 135) this is currently the best developed Model - essentially two principles (Flores d'Arcais, 1990; Ferstl & Flores d'Arcais, 1999):

  1. minimal-attachment:
    the sentence structure to be reconstructed is formed in such a way that it has as few branches as possible

  2. late closure:
    the word just read is appended to the last active phrase if possible

At the moment there is disagreement about the interplay of semantics and syntax in the interpretation of sentences (Traban & McClelland, 1990; see also Christmann & Groeben, 1999), whereby two opposing positions can essentially be distinguished: According to the interactionist syntax theory, the syntactic Analysis of the semantic context, influenced by pragmatics and world knowledge. For example, by means of the meaning of the verbs, the individual words are grouped into noun phrases, which are ultimately assigned their position in the sentence. According to this theory, the syntactic and semantic sub-processes work largely in parallel. Autonomous syntax theory, on the other hand, assumes that syntactic processing precedes semantic processing in time.

The question of which of the two theories describes sentence analysis better was investigated using the "garden path effect" (see also Christmann & Groeben, 1999) (e.g. Mitchell, Corley & Garnham, 1992). In this paradigm, the subjects are presented with sentences whose analysis according to the late closure principle leads to a syntactically incorrect structuring, which must be corrected in a second step. These are sentences such as "Helmut is easy to see". The word "Helmut" is initially perceived as a subject due to its position in the sentence, whereas in the deep structure it is the object of the activity indicated by the verb. Up to the age of seven to eight, children tended to interpret the sentence as meaning that Helmut could see well, according to Klicpera and Gasteiger-Klicpera (1995, p. 135). The situation is similar with the sentence "Hans promises Fritz to come", which younger children understand to mean that Fritz is coming. While children of this age generally have problems recognizing their own mistakes in the syntactic structure of sentences, adults in this case, as described in autonomous syntax theory, analyze the grammatical structure of the sentence in isolation (Christmann & Groeben, 1999). If, on the other hand, no contradictions arise during parsing, interactionist syntax theory seems to describe the adult reading process more adequately.

In addition to the examples given above, children at this age have problems with numerous conjunctors (Oakhill & Garnham, 1988, p. 52). While at the beginning of school primarily temporal conjunctures such as "then" and "after" are preferred, the understanding of causal connections such as "because", "therefore" and "consequently" gradually develops over the course of the first six years of school.

Klicpera and Gasteiger-Klicpera (1995, p. 135) assume that especially children who are still in the language acquisition process have a relative autonomy of syntactic and semantic analysis, while adults have a close interaction between syntax and semantics. According to Klicpera and Gasteiger-Klicpera, an essential factor for the development of reading comprehension is the ability to analyze sentences and / or grammatical competence.







2.4 Understanding the text

The research approaches in the field of text comprehension are subdivided into different areas, each pursuing different objectives: Text-based research is interested in describing text structures as objectively as possible and determining which properties of the text affect reading, understanding and in which way Remembering (Christmann & Groeben, 1999). Scheme-oriented approaches, on the other hand, focus on the characteristics of the recipient (e.g. prior knowledge structures, expectations and objectives) that are required to extract information from a text. Both positions will be discussed below.


2.4.1 Cross-sentence information integration

The process of understanding has only just begun with the syntactic and semantic decoding of individual sentences. Information obtained from the analysis of individual sentences must be integrated and put together to form an overall picture (Klicpera & Gasteiger-Klicpera, 1995, p. 136).

To establish relationships between different sentences or text passages, so-called cohesion means exist in language (Christmann & Groeben, 1999). These are links and references between different sentences, such as back references (anaphors), previous references (cataphors), word repetitions (recurrences) or resumption of sentence sequences using so-called pro-forms ("this", "that", "so") ...). There is a wealth of individual results on the relevance of the various means of cohesion, with the context often determining the importance of the individual stylistic means. In the field of developmental psychology, it was particularly examined when children are able to differentiate old and new information linguistically and to ensure discourse coherence by means of anaphoric references (Grimm, 1998). Obviously, the ability to link two sentences using anaphoric pronouns develops in the age range between five and seven years, i.e. at the time when important phases of the written language acquisition usually take place. Five-year-old children usually do not recognize when combining two sentences with an anaphoric pronoun ("The princess .... you ....") results in pragmatically implausible stories. On the other hand, 7-year-olds are able to establish this reference in comparison to children who are two years younger. Other anaphoric forms, however, also cause problems for older children. These structures include anaphoric references within a single sentence. For example, 10 year old children incorrectly interpret the sentence "Pete's brother ties himself up" in 11% of cases as "Pete is tying himself up" (Oakhill & Garnham, 1988, p. 58).


2.4.2 Models of text comprehension

Different models have been developed to explain the integration of information. The best known approach is proposition analysis (Kintsch & Kintsch, 1986; Kintsch, 1986, McNamara, Miller & Bransford, 1991). According to this model, propositions are the smallest units of knowledge or statements that make up a text. The example "John read an excellent article about the future of psychology on the plane to Frankfurt" (see Fig. 8), which is to be regarded as part of a longer text, can accordingly be analyzed according to Kintsch & Kintsch (1986) as follows: The syntactic analysis determines "las" as the predicate of the sentence on the basis of which the propositions are built. Since the predicate "read" requires a subject, "John" is specified as the doer. "An article" is determined as the object, to which the modifier "excellent" is added in this case. Furthermore, there is the proposition "about the future of psychology", as well as the location information "on the plane to Frankfurt". On the basis of these propositions, the sentence is integrated into the macro position "John's journey to Frankfurt", which in turn can be part of an even more comprehensive macro-structure (eg "Congress for experimental psychology in Frankfurt"), etc. First, the meaning is worked out at the local level which is eventually embedded in a larger macrostructure.



Figure 8
Processing levels in the model of text comprehension (based on Kintsch & Kintsch, 1986)


Christmann & Groeben (1999) note that propositional description models have proven to be successful in basic research, but are afflicted with a number of problems. These include, among other things, that they are extremely uneconomical when applied to longer text passages, which significantly reduces their prognostic value with regard to the retention performance of the recipient. The coherence of a text is also determined exclusively on the basis of referential identity - such as anaphoric references. In fact, causal or functional relationships are not taken into account. It is obvious that even texts whose sentences do not have any references to each other can be interpreted without any problems on the basis of world knowledge, as the following example shows: "The refrigerator is empty. Markus is hungry."

According to Christmann and Groeben (1999), however, there is general evidence that a cross-sentence semantic macrostructure is built up in the reading process. This happens all the easier, the clearer and clearer a text is structured and the more indications are given as to which text information has to be related to one another and which inferences have to be formed.


2.4.3 Scheme-guided text comprehension

The availability of scripts and schemes can make working out the meaning of texts much easier. The term schema goes back to Bartlett and describes holistic structures in which knowledge about typical connections between realms of reality is stored (Christmann & Groeben, 1999; Bergius & Casper, 1998). In this context, so-called scripts and story grammars are important for understanding texts. Scripts describe actions that typically take place in a certain way, such as going to a restaurant (1st entering the restaurant, 2nd waiter guides the guest to the table, 3rd waiter brings the menu ...). Story grammars, on the other hand, contain information about the global arrangement of text elements in narrative texts, such as the obligatory "happy ending" at the end of a children's fairy tale. These typical arrangements of narrative content are familiar to children and they make it easier to understand and remember the specific content (Klicpera & Gasteiger-Klicpera, 1995, p. 140). Story grammars therefore primarily have a cognitive relief function, since the structure of the text does not have to be developed in addition to the content. How important schemas are for us and how much we rely on them becomes clear when test subjects mistakenly change information that is incongruent with the schema in retelling so that what is retold corresponds to their expectations (Oakhill & Garnham, 1988, p. 107).

Even factual texts usually have certain structures that are often derived from rhetorical patterns (Klicpera & Gasteiger-Klicpera, 1995, p. 140). In contrast to factual texts, narrative texts have a concrete reference to the children's life experience. The understanding of factual texts, on the other hand, only matures in the course of school. Those students who know the structure of the text and use it to develop the content can organize the knowledge better and later call up more information.

However, it remains unclear how schemes are applied (McNamara, Miller & Bransford, 1991). Are schemes simply activated and then the missing information added at the designated places? And how is it then possible that readers can also understand schema-incongruent texts? These questions are dealt with below.


2.4.4 Mental models

One problem with propositional models was that they often represented the structure of a text rather than the structure that the reader reconstructed from the text (McNamara, Miller & Bransford, 1991). The psychological research on the topic of schemes, on the other hand, focused one-sidedly on the world knowledge and the experiences of the recipient. So it made sense to combine both approaches. According to this new form of theories, which are called "mental models" (McNamara, Miller & Bransford) or "situation models" (Artelt, Stanat, Schneider & Schiefele, 2001), the reader processes a text on the basis of propositions. At the same time he constructs a mental model in analogy to the structure of the events and situations as they are described in the text.

For example, if the test person has to memorize one of the first two sentences (A1 and A2) from text 1, it will later be difficult for them to decide which of the two they had previously learned (Bransford, Barclay & Franks, 1972, quoted from McNamara, Miller & Bransford, 1991). According to the proposition analysis, this is not surprising, after all, the two sentences differ exclusively in a single proposition. In the case of the last two sentences (B1 and B2) the decision should be just as difficult, after all, here, exactly as in sentences A1 and A2, the same proposition has been changed. However, since, in contrast to the first two sentences, which have exactly the same situation, the last two sentences describe a different situation, it is not difficult for the test subjects to find the right sentence. The different performance of the test subjects in the first and second pair of sentences can therefore be attributed to the fact that in the first case there are identical and in the second case different mental models.

Theorem A1: Three turtles rested on a floating log and a fish swam beneath them.
Sentence A2: Three turtles rested on a floating log and a fish swam beneath it.

Sentence B1: Three turtles rested beside a floating log and a fish swam beneath them.
Sentence B2: Three turtles rested beside a floating log and a fish swam beneath it.


Text 1
The experiment by Bransford, Barclay & Franks (1972, quoted from McNamara, Miller & Bransford, 1991): Well after reading one of the first two sentences (sentence A1 and A2), subjects found it difficult to decide which of the two it was has been. With sentences B1 and B2, the decision is easy, although this pair of sentences also only differs by a preposition. The difference must therefore be in the mental representation of the situation described.


The empirical research on mental models used to a large extent the description of the relations of objects in spaces or of positions on maps (Christmann & Groeben, 1999). It was found that readers are able to organize a text in the form of propositions as well as to reconstruct a mental model. Which processing path a person chooses depends largely on the material presented: congruent information is preferably represented in the form of mental models and incongruent information in propositional form (McNamara, Miller & Bransford, 1991).

An extremely important factor in the creation of mental models is the integration of the recorded information with the previous knowledge (Artelt, Stanat, Schneider & Schiefele, 2001). For example, prior knowledge includes scripts, schemes and specific learning experiences. According to popular belief, knowledge is organized in the form of propositional networks and access to knowledge structures is quick and precise. The activation of prior knowledge structures relieves the process of text understanding.

Another important aspect of mental models is the need to check the reconstructed model for its internal coherence and for consistency with what has been read. Monitoring one's own understanding of the text is a metalinguistic skill that only develops over the years (Oakhill & Garnham, 1988, p. 115 f.). First graders usually fail to discover serious logical errors in texts themselves. Third graders, on the other hand, notice inconsistencies much faster, without being able to verbalize which errors have occurred in the text.


2.4.5 Inferential reading

One skill closely related to mental models is inferential reading. While inference is usually understood to mean drawing logical conclusions (Inference, 1998), when it comes to reading comprehension, this term is used more in the sense of "reading between or behind the lines" (Klicpera & Gasteiger-Klicpera, 1995, p. 138). The point is to abstract information from the literal linguistic realization and to connect different sentences with one another, i.e. to construct a mental model of the text. If you consider the example "Anna tried to make a dress out of the beautiful material. The scissors were blunt and it gave her a lot of trouble." It is immediately clear to the reader that Anna wanted to cut out parts of the material with the scissors. For this to be possible, however, various non-trivial conclusions must be drawn. On the one hand, it is necessary to relate the terms scissors, dress and fabric to one another. On the other hand, the anaphoric pronoun must be related to Anna. These inferences, which are necessary for understanding the text, must be distinguished from further inferences that narrow, supplement or extend what is stated in the text. The sentence "The waiter slipped on the freshly mopped floor and the expensive glass fell to the floor" (Klicpera & Gasteiger-Klicpera, 1995, p. 138) is usually interpreted as meaning that the glass splintered on the floor. After all, expensive glasses are usually made of glass and won't withstand a fall from a height of 1.5 m.

Numerous studies show that good and bad readers differ significantly in the number of inferences drawn (Oakhill & Garnham, 1988, p. 136 f.). These differences are not due to differences in memory capacity or divergences in the knowledge base, according to Oakhill and Garnham. They suspect that the ability to form inferences is ultimately an intelligence achievement.




3. Developmental psychological models of written language acquisition

3.1 The model of written language acquisition according to Frith and Günther

Analogous to the process models of reading and writing, models for the acquisition of the written language were developed in the course of the 1980s (Klicpera & Gasteiger-Klicpera, 1995, p. 46). Stage models of written language acquisition that relate in particular to information processing theories of reading were particularly well received (e.g. Frith, 1985). Frith postulates three phases in the acquisition of the written language:
  1. Logographic phase
    At this first stage, the children identify words based on distinctive features. The words are not recognized on the basis of the identification of letters or the sequence of letters, but only the global visual impression counts. For example, the company logos of popular brands can be recognized and named.
    When the child begins to further differentiate the characteristics used for reading, a significant step has been taken towards the alphabetical phase (Klicpera & Gasteiger-Klicpera, 1995, p. 46).

  2. Alphabetical phase
    In this phase, the child uses knowledge of the identity of the letters and their assignment to phonemes to systematically select words. Words are recognized by letter-by-letter reconstruction of the letter sequences to form whole words. It is noteworthy that at this stage the recognition of words on the basis of outstanding features seems to take a back seat. The words are identified exclusively on the basis of the reconstruction of grapheme-phoneme connections (Klicpera & Gasteiger-Klicpera, 1995, p. 46).

  3. Orthographic phase
    This stage represents, so to speak, the synthesis of the two preceding phases. The phonological recoding takes a back seat and the words are recognized directly (Klicpera & Gasteiger-Klicpera, 1995, p. 46). However, this is done with the help of the sequence of letters and the knowledge already acquired about the structure of the written language. On this basis, words can be identified very quickly and recapitulated from orthographic memory without having to read the words over and over again. This process is probably also economized by dividing the letter sequences into syllables, morphemes or smaller, frequently occurring letter sequences such as st, sp, ng (Klicpera & Gasteiger-Klicpera, 1995, p. 47).

Independently of Uta Frith, Günther (1989) developed his "step model of written language acquisition as a development process" (see Fig. 9). Due to the pronounced parallels between the two models, they are usually combined and referred to as the Frith Günther model. In addition to the phases listed by Frith (1985), Günther (1989) postulates a preliteral-symbolic stage as the forerunner of written language acquisition and thus begins the acquisition of written language at a much earlier point in time: Based on the developmental psychological ideas of Piaget, he assumes that "that symbol awareness in preschool age, due to the dominant forms of thought and activity, initially develops in a general way and thus creates the basis for being applied to these oneself in the actual phase of the written language acquisition." (Günther, 1989).



Figure 9
Stage model of written language acquisition according to Günther (1989). Günther regards the acquisition of written language as a result of strategy changes, which he describes as critical phases (highlighted in color by the author). According to Günther, if a change of strategy fails, the child shows characteristic failures.


In addition, Günther's orthographic phase is followed by the integrative-automated phase, which does not represent a further change in strategy, but symbolizes the long process of automating and consolidating the linguistic rules and dealing with written material (Sassenroth, 2000, p. 46).

Although step models such as the Frith-Günther model enjoy enormous prominence, there are increasing numbers of critical findings that cast doubt in particular on the transferability to the German-speaking area (Schneider, 1998). In particular, the logographic phase seems to play a subordinate role in the German-speaking area. If it can be observed at all, it is only in the first few weeks of school written language acquisition. At the present time, the model has not been adequately confirmed empirically by longitudinal studies.

The one-sided focus of the Frith-Günther model on reading and spelling processes at the word level must also be viewed critically. Meaningful reading, which is already rudimentary in beginners' reading, cannot explain this model.




3.2 The simple view of reading approach

While the focus of the Frith-Günther model is on basic reading strategies, a model was proposed in the course of the 1980s that explicitly deals with the development of reading comprehension and bears the designation "Simple-View-of-Reading" (Gough and Tunmer 1986; Hoover & Gough, 1990; Gough, Hoover & Peterson, 1996; Marx & Jungmann, 2000). According to Gough, Hoover and Peterson (1996), the title is by no means intended to suggest that reading is a trivial skill. Rather, it should be assumed that reading comprehension is essentially determined by only two skills, both of which have the same value: decoding and listening comprehension. The starting point for developing the model was an argument on which some critics of the currently used definition of discrepancy in reading and writing disorders (LRS) rely. The guidelines for diagnosing LRS in German and "developmental dyslexia" in English-speaking countries differ in terms of the choice of methodology used. In Germany, for example, a T-value discrepancy of 12 points between the performance in a standardized reading and spelling test and a procedure for recording non-verbal intelligence is trend-setting (AWMF online, 2000). In the English-speaking world, however, a regression method is preferred (Weber, Marx & Schneider, 2002). Nevertheless, both approaches are based on the idea that in addition to basic reading techniques, the intelligence of a child is decisive for the acquisition of written language. In contrast, Stanovich (1989, 1991, 1994) emphasized that it is not intelligence but rather listening comprehension that is better suited for predicting reading performance. Listening comprehension defines the upper limit of the comprehension achievable while reading.

Gough and Tunmer (1986) also took up this idea and constructed a model in which reading comprehension results from the combination of listening comprehension and basic reading techniques. They refer to these basic reading techniques as "decoding", which for them is the general description for efficient word recognition:
"For the simple view, skilled decoding is simply efficient word recognition: the ability to rapidly derive a representation from printed input that allows access to the appropriate entry in the mental lexicon, and thus, the retrieval of semantic information at the word level." Hoover and Gough (1990)
In the end, it doesn't matter whether a word is recoded letter by letter or recognized in some other way. Hoover and Gough (1990) suspect that recoding dominates reading, especially at the beginning of written language acquisition, but is increasingly being replaced by more efficient forms of word recognition, which they do not specify in any more detail.

Listening comprehension, on the other hand, is the ability to syntactically analyze a sentence, grasp its meaning and relate it to the surrounding sentences (Gough, Hoover & Peterson, 1996).

Despite some differences in the language used (formal versus informal) and the mode of presentation (punctuation for written and intonation for auditory input), Gough, Hoover & Peterson (1996) assume that the process of understanding written and auditory language is essentially the process is identical. According to Marx and Jung (2000), they represent a monistic position that is also advocated by Rost and Hartmann (1992). Supporters of the monistic position assume that processes of understanding are in principle independent of the modality (Rost, 1998). According to this position, differences in modality-dependent comprehension skills are attributed to additional reading-specific factors such as the decoding of letter sequences in the case of written language.

According to Hoover and Gough (1990), when listening comprehension is fully developed, reading comprehension is determined exclusively by the ability to decipher individual words. In reading comprehension there is consequently a multiplicative combination of listening comprehension and decoding, because the meaning of a text can only be grasped when both skills are well developed. They therefore differentiate between four different subgroups of children, three of which are characterized by poor reading comprehension (see Table 1).

Table 1
Possible subgroups of poorly literate or competent children according to Hoover & Gough (1990).
 Decode
-+
Listening comprehension-"Garden Variety Poor Reader"
The child has neither effective word recognition nor listening skills. It is therefore generally poorly read.
"Hyperlexic"
The child is able to decode writing, but has a poor understanding of what is being read.
+"Dyslexic"
There is a reading and spelling disorder because there is a discrepancy between listening and reading comprehension.
good reading comprehension
Both decoding and listening comprehension are well developed. The child can read meaningfully.
The child is only able to read meaningfully when both listening comprehension and the ability to decode are developed.


Various research groups confirm the close connection between listening and reading comprehension (Rost & Hartmann, 1992; Marx & Jungmann, 2000, Rupley & Willson, 1997; Mommers, 1987), with decoding being the limiting factor in reading comprehension, especially for German-speaking novice readers seems. Marx and Jungmann (2000) report a particularly strong increase in reading comprehension compared to listening comprehension between the end of the first and middle of the second grade (see Fig. 10). While listening comprehension increases moderately from grade to grade, there is a particularly rapid development in reading comprehension, which can be attributed to automation in the area of ​​decoding. This influence of decoding on reading comprehension can be demonstrated throughout primary school up to the sixth grade (Rost & Hartmann, 1992; Marx & Jungmann, 2000; Mommers, 1987; Ruply & Willson, 1997). Since, on average, the quantitative level of reading comprehension does not reach listening comprehension, it must be assumed that the presence of hyperlexia is less common than other forms of weak reading comprehension.



Figure 10
Comparisons of mean values ​​between listening and reading comprehension in the various research groups (Marx & Jungmann, 2000). The increase in reading comprehension is most pronounced from the end of the 1st to the middle of the 2nd grade.


The relationship between listening and reading comprehension is substantial from the second grade onwards, and ranges between 0.60 and 0.65 (Marx & Jungmann, 2000, Rost & Hartmann, 1992). In addition, listening comprehension seems to be a better predictor of reading comprehension than linguistic intelligence: while listening comprehension explains about 38% of the variance in reading comprehension, the proportion of linguistic intelligence is much lower at around 28% (see Fig. 11).




Figure 11
common proportions of variance in linguistic intelligence, reading and listening comprehension (Rost & Hartmann, 1992)


Hoover and Gough (1990) tested their model using data from an earlier study. The focus was on clarifying the question of which model could explain more variance in reading comprehension: a model with an additive link between listening comprehension and decoding, or one with a multiplicative link. Although the latter was actually able to explain significantly more variance, the differences were marginal. In addition, a post-hoc model that contained both additive and multiplicative components proved to be superior to the other two approaches. Criticism was sparked, among other things, by the fact that the sample consisted of bilingual children. The data were obtained using English test procedures, i.e. in the children's second language.

For this reason Dreyer and Katz (1992) studied monolingual children with the same experimental design and came to essentially the same results. They were also unable to make a decision for the additive or the multiplicative model, but for them, too, the outstanding importance of the two factors word recognition and linguistic competence for reading comprehension is beyond question.

In the following years, various modifications and extensions for the theory of the simple view of reading were suggested, for example the component model of reading (Joshi & Aaron, 2000) and the rauding theory or simple view of reading II (Carver, 1993; Rupley, Willson & Nichols, 1998). Joshi and Aaron suggested adding the decoding speed into the model, as they achieved an increase in variance explanation of 10% in their sample using this modification (57.76% instead of 47.61%). Rupley, Willson, and Nichols' proposals go in a similar direction. They also recommend considering reading speed as an additional component.




3.3 Level models

Rost, Czeschlik and van der Kooij (1986) analyzed the data of 220 children in the second grade who had passed through all subscales of the diagnostic program "Reading and Understanding" (Kalb, Rabenstein & Rost, 1979). Using a factor analysis yielded a two-factor solution, with the most important factor representing reading comprehension and the second, much less significant factor representing vocabulary. It was therefore not possible to split reading comprehension into distinguishable sub-skills, and so there was no possibility of evaluating these sub-skills differentially.

Andrich and Godfrey (1979) came to a similar conclusion in their factor-analytical study of the sub-tasks of the F.B. Davis's reading comprehension test. They were also able to show that in the case of a one-factor solution, the present scale is quickly scalable. Reading comprehension thus seems to be a uniform construct, which, however, has distinguishable quantitative characteristics in the sense of levels.

More recently, this approach has mainly been followed in the context of international comparative studies.These studies include the research of the International Association for the Evaluation of Educational Achievement (IEA), in which children between the ages of 9 and 14 participated (Elley & Canterbury, 1993; Wagenmaker 1992), the First International Adult Literacy Survey ( OECD, 1995), which examined literacy among adults, and the PISA study (OECD 2001), which recorded, among other things, the reading comprehension of 15-year-old schoolchildren. All these studies were based on the literacy concept, which is widespread in the Anglo-Saxon language area, according to which the command of language and writing is one of the basic cultural techniques and is necessary for a satisfactory lifestyle in personal and economic terms, as well as for participation in social life (Artelt, Stanat, Schneider & Schiefele, 2001).

For example, the authors of the PISA study differentiate between five "competence levels" (OECD, 2001, p. 11). Students who are at level 1 ("elementary level") are accordingly able to extract information from a text, provided that the type of text is familiar and the statements are presented very clearly and, if possible, without competing information. Pupils who reach proficiency level 6 ("expert level"), however, can localize and organize deeply embedded information even with unknown types of text. They are able "to critically evaluate a text with reference to specialized knowledge or to formulate hypotheses about information in the text, even if the relevant concepts contradict expectations" (OECD, 2001, p. 11). The Rasch model was used to form these competence levels (Artelt, Stanat, Schneider & Schiefele, 2001). Students at the lower end of a proficiency level solved 62% of the easy and 42% of the hard tasks for that level. In contrast, 78% of students at the upper end of the level mastered the easy tasks and 68% of the difficult tasks. It is important to note that this conception was designed for the text understanding of 15-year-old schoolchildren and certainly cannot be transferred to novice readers.






4. Bibliography



Artelt, C., Baumert, J., Klieme, E., Neubrand, M., Prenzel, M., Schiefele, U., Schneider, W., Schümer, G., Stanat, P., Tillmann, K.- J., & Weiss, J. (2001). PISA 2000: Summary of Key Findings [Internet]. available at: http://www.mpib-berlin.mpg.de/pisa/results.pdf [07/31/2002].

Artelt, C., Stanat, P., Schneider, W. & Schiefele, U. (2001). Reading skills: test design and results. J. Baumert, E. Klieme, M. Neubrand, M. Prenzel, U. Schiefele, W. Schneider, P. Stanat, K.-J. Tillmann & M. Weiß (eds.), PISA 2000: Basic Competencies of Schoolchildren in International Comparison (69-140). Opladen: Leske + Budrich.

AWMF (2003). Working Group of Scientific Medical Societies: Circumscribed Developmental Disorders of School Skills (F81) [Internet]. Available at: http://www.uni-duesseldorf.de/AWMF/ll/kjpp-017.htm (date of access: May 19, 2004).

Balota, D.A. (1990). The Role of Meaning in Word Recognition. In D. A. Balota, G. B. Flores d'Arcais & K. Rayner (Eds.), Comprehension processes in reading (9 - 32). Hillsdale: Lawrence Erlbaum Associates.

Byrne, B., Freebody, P., & Gates, A. (1992). Longitudinal data on relations of word-reading strategies to comprehension, reading time, and phonemic awareness. Reading Research Quarterly, 27, 140-151.

Christman, U. & Groeben, N. (1999). Psychology of reading. In B. Franzmann & G. Jäger (eds.), Read the manual (145-223). Munich: Saur.

Coltheart, M. & Rastle, K. (1994). Serial Processing in Reading Aloud: Evidence for Dual-Route Models of Reading. Journal of Experimental Psychology: Human Perception and Performance, 20(6), 1197 – 1211.

Ferstl, E. & Flores d'Arcais, G.B. (1999). Reading words and sentences. In A. D. Frederici (ed.), Encyclopedia of Psychology, Series 3. Language, Vol. 2. Speech Reception (203-242). Göttingen: Hogrefe.

Fletcher, J. M., Morris, R., Lyon, G. R., Stuebing, K. K., Shaywitz, S. E., Shankweiler, D. P., Katz, L. & Shaywitz, B. A. (1997). Subtypes of Dyslexia: An Old Problem Revisited. In B. Blachman (Ed.), Foundations of reading acquisition and dyslexia: implications for early intervention (115-141). New Jersey: Lawrence Erlbaum Associates.

Flores d'Arcais, G.B. (1990). Parsing principles and language comprehension during reading. In D. A. Balota, G. B. Flores d'Arcais & K. Rayner (Eds.), Comprehension processes in reading (435-360). Hillsdale: Lawrence Erlbaum Associates.

Gold, A., Mokhlesgerami, J., Rühl, K., Schreblowski, S., & Souvignier, E. (2004). We become text detectives [teacher manual & workbook]. Göttingen: Vandenhoeck & Ruprecht.

Grimm, H. (1998). Language development - viewed from a general theoretical and differential perspective. In R. Oerter & L. Montada (Eds.), Developmental Psychology: A Textbook (4th edition) (705-757). Weinheim: Beltz.

Kalb, G., Rabenstein, R. & Rost, H. (1979). Reading and Understanding - Diagnosis. Braunschweig: Westermann.

Kendall, J. R., & Hood, J. (1979). Investigating the relationship between comprehension and word recognition: Oral reading analysis of children with comprehension or word recognition disabilities. Journal of Reading Behavior, 11 (1), 41-48.

Klicpera, Ch. & Gasteiger-Klicpera, B. (1995). Psychology of reading and spelling difficulties: development, causes, encouragement. Weinheim: Beltz.

Küspert, P. & Schneider, W. (1998). Würzburger Leise Leseprobe (WLLP): A group reading test for elementary school. Göttingen: Hogrefe.

Lenhard, W., & Schneider, W. (2005). ELFE 1-6: A reading comprehension test for first to sixth graders. Göttingen: Hogrefe.

Lienert, G. & Raatz, U. (1998). Test setup and test analysis (6th edition). Weinheim: Psychology Publishing Union.

Marx, H. (1998). Kuspels reading tasks (Knuspel-L). Göttingen: Hogrefe.

Mastropieri, M., & Scruggs, T. (1997). Best Practice in Promoting Reading Comprehension in Students with Learning Disabilities. Remedial and Special Education, 18, 197-213.

McNamara, T.P., Miller, D.L. & Bransford, J.D. (1991). Mental Models and Reading Comprehension. In T. P. McNamara, D. L. Miller, J. D. Bransford & R. Barr (Eds.), Handbook of Research on Reading, Vol 2 (490-511). Hillsdale: Laurence Erlbaum Associates.

Mitchell, D.C., Corley, M., Garnham, A. (1992). Effects of Context in Human Sentence Pasing: Evidence Against a Discourse-Based Mechanism. Journal of Experimental Psychology: Learning, Memory and Cognition. 18th(1), 69 – 88.

Morris, R. D., Stuebing, K. K., Fletcher, J. M., Shaywitz, S. E., Lyon, G. R., Shankweiler, D. P., Katz, L., Francis, D. J. & Shaywitz, B. A. (1998). Subtypes of Reading Disabillity: Variability Around a Phonological Core. Educational Psychology, 90(3), 347 – 373.

Näslund, J.C., & Schneider, W. (1996). Kindergarten letter knowledge, phonological skills, and memory processes: Relative effects on early literacy. Journal of Experimental Child Psychology, 62, 30-59.

Oakhill, J. & Garnham, A. (1988). Becoming a skilled reader. Oxford: Blackwell.

OECD (1995). Literacy, Economy and Society. Results of the first International Adult Literacy Survey. Ontario: OECD.

Palincsar, A.S. & Brown, A.L. (1984). Reciprocal teaching of comprehension-monitoring and comprehension-monitoring activities. Cognition and Instruction, 1, 117-175.

Paris, S.G., Cross, D.R. & Lipson, M.Y. (1984). Informed strategies for learning: A program to improve children’s reading awareness and comprehension. Journal of Educational Psychology, 76 (6), 1239-1252.

Pinker, S. (1998). The language instinct. Munich: Knaur.

Richter, S. (2002). Written vocabulary of elementary school children. Bavarian body. Regensburg: Roderer.

Richter, T. & Groeben, U. (2002). Reading competence: process levels and inter-individual differences. In N. Groeben & B. Hurrelmann (eds.), Reading literacy: conditions, dimensions, functions (25-58). Weinheim: Juventa.

Schaffner, E., Schiefele, U., & Schneider, W. (2004). An expanded understanding of reading skills: the results of the national supplementary test. In U. Schiefele, C. Artelt, W. Schneider & P. ​​Stanat (Eds.), Structure, development and promotion of reading skills (197-243). Wiesbaden: VS-Verlag.

Schneider, W. (2004). Early Development of Reading Skills: The Relevance of Pre-School Language Skills. In U. Schiefele, C. Artelt, W. Schneider, & P. ​​Stanat (Eds.), Structure, development and promotion of reading skills: In-depth analyzes as part of PISA 2000 (13-36). Wiesbaden: VS Verlag.

Schneider, W., & Näslund, J.-C. (1997). The early prediction of reading and spelling: Evidence from the Munich Longitudinal Study on the genesis of individual competencies. In C. K. Leong, R. Joshi (Eds.), Cross-language studies of learning to read and spell (139-159). Dordrecht: Kluwer Academic Publishers.

Schwippert, K., Bos, W., & Lankes, E.-M. (2003). Heterogeneity and equal opportunities at the end of the fourth grade in an international comparison. In W. Bos, E.-M. Lankes, M. Prenzel, K. Schwippert, R. Valtin, & G. Walther (Eds.), First results from IGLU: Student performance at the end of the fourth grade in an international comparison (265-302). Münster: Waxmann.

Stanovich, K., Siegel, L., Gottardo, A., Chiappe, P. & Sidhu, R. (1997). Subtypes of Developmental Dyslexia: Differences in Phonogical and Orthographic Coding. In B. Blachman (Ed.), Foundations of reading acquisition and dyslexia: implications for early intervention (115-141). New Jersey: Lawrence Erlbaum Associates.

Federal Statistical Office (2005). Territory and population [Internet]. available at: http://www.statistik-portal.de/Statistik-Portal/de_jb01_jahrtab1.asp [04/11/2005].

Swanson, H. (1999). Reading research for students with LD: A meta-analysis of intervention outcomes. Journal of Learning Disabilities, 32, 504-532.

Taraban, R. & McClelland, J.L. (1990). Parsing and comprehension: A multiple-contraint view. In D. A. Balota, G. B. Flores d'Arcais & K. Rayner (Eds.), Comprehension processes in reading (231-264). Hillsdale: Lawrence Erlbaum Associates.






5. Tabular overview of reading tests


The following list shows current reading tests that are available or at least used in German-speaking countries. Most of the procedures are available at http://www.testzentrale.de, by direct mail from the test authors or via the Internet.



Surname Short-
denoting
tion
Application edition
dung since
Age
Area
Descriptive
exercise
author Information
for test
normative
tion
A reading comprehension
test for first to sixth graders
ELFE 1-6 from February 2006 End of 1st to end of 6th grade Reading comprehension test, which measures comprehension at word, sentence and text level. Available in computer and paper and pencil versions. W. Lenhard & W. Schneider Class
standards
Reliability
validity
N = 4893
Frankfurt reading comprehension
test for 5th and 6th grades
FLVT 5-6 2008 Grade 5 and 6A narrative and an expository text with 18 multiple chioce questions each E. Souvignier, I. Trenk-Hinterberger, S. Adam-Schwebe and A. Goldr Class
standards (2nd semester each)
internal consistency
validity
N = 2476
Reading speed
Aptitude and comprehension test for grades 6-12
LGVT 6-12 2007 Grade 6 - 12 Determination of reading comprehension and reading speed based on reading a text in the gap. There is a gap in 23 places in the text and the students must mark the word that fits into the gap from a list of 3 words. W. Schneider, M. Schlagmüller and M. Ennemoser Class and school norms
Reliability
validity
N = 2390
Salzburg reading screening for grades 1-4 SLS 1-4 In application
since 2003
End of the 1st to the end of the 4th grade List of very simple correct and incorrect sentences (e.g. bananas are blue), which should be read as quickly as possible and assessed for correctness. If the SLS 1-4 delivers a value in the lower performance range, the problem should be clarified further using the SLRT. H. Mayringer and H. Wimmer Class
standards
Reliability
validity
N = 1867
Salzburg reading screening for grades 5-8 SLS 5-8 In application
since 2005
End of 5th to end of 8th grade
step
analogous to SLS 1-4 M. Auer, G. Gruber, H. Mayringer and H. Wimmer Class
standards
Reliability
validity
N between 714 and 850 per class
step
The Salzburg reading and spelling test SLRT In application
since 1997.
At the end of the 1st to the end of the 4th grade. Method for the differential diagnosis of weaknesses in learning to read and write; Detection of two different components of reading: deficits in automatic, direct word recognition and deficits in synthetic reading aloud. Reading time and reading errors are recorded K. Landerl, H. Wimmer and E. Moser Age
standards
Reliability
validity
N = 2800
Würzburger Leise reading sample WLLP 1st edition 1998 1st to 4th grade Decoding word recognition
speed
P. Küspert
W. Schneider
Age
standards
Reliability
validity
N = 2820
Knuspels reading assignments Knuspel-L 1st edition 1998 End of the 1st to the end of the 4th school year Word-level recoding and decoding
Reading comprehension at sentence level
H. Marx Age
standards
Reliability
validity
Hamburg reading test for 3rd and 4th grades HAMLET 3/4 2nd edition 2006 End of 3rd and end of 4th grade Word test and reading comprehension test (consisting of 10 texts) R.H. Lehmann, R. Peek and J. Poerschke Age
standards
Reliability (limited) Validity (limited)
Reading and Understanding Diagnosis and Training LUV 1st edition 1971 1st and 2nd grade and older children with reading difficulties Diagnosis and training program on the subject of meaningful reading G. Kalb, R. Rabenstein
D. H. Rost
Reliability
Reading test for 2nd grade LT 2 3rd edition 1971 2nd half of the 2nd class to the 1st quarter of the 3rd class Sub-tests thematize the understanding when reading in silence via picture-word, word-word and sentence-sentence assignment -> word comprehension and text comprehension E. Samtleben, F. Biglmaier, K. Ingenkamp Age
standards,
N = 729
Diagnostic reading test for the early diagnosis of reading disorders DLF 12 In application
since 1984
End of 1st to middle of 2nd grade Word reading test for the early detection of reading disorders. Sub-tests include "Save", "Synthesis", "Analysis", "Segmentation" R. Muller Reliability
validity
N = 320
Trip word reading test STOLLE since 2001 (?) 1st to 4th grade List of sentences in which a "disturber" was inserted. The wrong word has to be recognized and crossed out. W. Metze