Language use in school books

Language use in school books

SUMMARY

TABLE OF CONTENT

Introduction 4

Purpose of the present study 6

Chapter 1: Reading 7

1.1 Learning to read 7

1.2 The process of reading comprehension 10

1.3 Problems with reading comprehension 16

1.4 The reading process and the present study 17

Chapter 2: Text Analysis 17

2.1 Types of text analysis 17

2.2 Coherency analysis 20

2.3 Readability formulas 23

2.4 Text analysis and the present study 24

Chapter 3: Design and Method 27

3.1 Outline 27

3.2 Material 27

3.3 Procedure 28

3.4 Analysis 31

Chapter 4: Results 33

4.1 Differences in language use in texts 33

4.2 Appropriate language use in texts 35

4.3 Coherency analysis 38

Conclusion and Discussion 44

Reflection 47

Bibliography

List of references 48

INTRODUCTION

Every year, thousands of Dutch secondary school students buy new school books or receive them from their school. When using these books, most of them will not realize how much of an effort it takes to make them. The publisher has to develop a plan for the content and the lay out, find good writers and manage the process of writing and rewriting. Fortunately, for the first two years of secondary school, publishers in the Netherlands can make three of four quite equal books in once.

In the Netherlands, there are three main levels of secondary school: vmbo (lower general secondary education), havo (higher general secondary education) and vwo (pre-university education). vmbo is divided into four sublevels, namely basisberoepsgerichte leerweg (practice-oriented stream), kaderberoepsgerichte leerweg (basic stream), gemengde leerweg (mixed stream) and theoretische leerweg (theoretical stream). These four sublevels are often indicated by their abbreviations: bb, kb, gl and tl. In the first en second class of secondary school, the differences between these levels and sublevels are small, but already too big to let the students all use the same school book. Therefore, publishers develop different versions of one school book. But due to the relatively small differences, most publishers make one book for students of two or three different school levels. Common divisions are: bb/kb, gl/tl and havo/vwo or bb, kb/gl/tl and havo/vwo. These different versions of a book have a different content, because students of different school levels do not have to learn the same knowledge. However, the language use should also be different in these books, because the reading abilities of vwo-, havo- and vmbo-students are very different (Hacquebord, Linthorst, Stellingwerf & De Zeeuw ,2004).

abilities above or below average. When tested on reading abilities, they score higher or lower than a student with a DA of 61 should. This score is indicated by the DAE. A student who develops fast can have a DA of 61, but a DAE of 65. A student who develops slowly can have a DA of 61, but a DAE of 57. On average, a group of students should have an equal DA and DAE. This is true for the students of vmbo-tl and havo, but above-average students (vwo) are expected to have a DAE of 71 in the first month of secondary school. Below-average students (vmbo-bb/kb/gl) are expected to have a DAE of 51. The expected standard deviation is 5. The results of Hacquebord et al. (2004) are given in the table below.

Level Average DAE Standard deviation N

Vmbo-bb/kb 51,9 10,8 1001

Vmbo-tl/havo 60,7 10,9 551

havo/vwo 72,2 9,8 1052

Total 62,0 13,8 2604

Table 1: Showing the average DAE’s and the standard deviations per secondary school level in the first month of the first class.

The average DAE’s are close to the expected DAE’s, but the standard deviations (SD’s) are much higher than expected: between 9.8 and 13.8. This indicates that within each group a large number of students develops faster or slower than expected.

Level - 1 SD (%) - 1.5 SD’s (%) - 2 SD’s (%) Total (%)

Vmbo 24.2 - - 24.2

Vmbo-tl/havo 7.7 8.7 0.2 16.6

havo/vwo 9.3 4.1 4.7 18.1

Table 2: Showing per level the percentages of the students who score 1, 1.5 or 2 SD’s below average.

The study of Hacquebord et al. (2004) makes clear that there are not only big differences in reading abilities between the different school levels; there are also big differences within each group. On each level, approximately 20% of the students have developed their reading abilities slower than expected.

Purpose of the present study

The purpose of the present study is to find out whether the mentioned differences in reading abilities are taken into account during the production of school books for different school levels by adjusting the language use. School books for vmbo, havo and vwo will be analyzed to find out if there are differences in language use. These results will be studied to find out whether the language use in the different school books is appropriate for the school level of the readers.

CHAPTER 1: READING

It seems very simple to answer the question: what is reading? But although many people can do it, what we really do while we are reading a text is still subject of study. There is no clear theoretical model of the process of reading and therefore, no clear answer to what seems a simple question (Van Elsäcker, 2002). Terms that are often used to describe reading and reading comprehension are ‘to take in’, ‘to understand’ and ‘to attribute an interpretation’ (Bernhardt, 1991). For the purpose of this study, the definition of reading comprehension of the RAND report (2002) will be used: ‘the process of simultaneously extracting and constructing meaning through interaction and involvement with written language’ (p. 11).

This chapter is about reading. First, the process of learning to read in the first and second language will be discussed. Then one of the most popular theories of reading comprehension will be reviewed and finally, problems that may occur during reading are addressed.

1.1 Learning to read

The process of learning how to read is not a natural one. (Elsäcker, 2002) For both learning to read in a first and in a second language instruction is needed to acquire the skill. These learning processes will both be described.

Learning to read in L1

Learning how to read contains a number of different steps. The first, and perhaps most important step, is that children need to learn that words have certain form features (Thomassen, Noordman & Eling, 1991). Rozin, Bressman and Taft (1976) showed that young children are not aware of these form features of words. They showed young, Dutch children two cards with the words vlieg (‘fly’) written on one card and the word

vliegmachine (‘airplane’) written on the other card. The children were not able to say

Secondly, children must learn to link specific word forms to certain sounds. This is difficult, because children need to learn that not each character is a specific sound, but that groups or combinations of characters represent certain sounds. These sounds are called phonemes and are not a natural boundary for children within a word. Moreover, children first learn to distinguish syllables, before they learn to distinguish phonemes (Thomassen, Noordman & Eling, 1991).

Marsh, Friedman, Welch and Desberg (1981) describe four phases within the process of learning to read a first language:

• Guessing by context: word meanings are guessed based on the context of the

words. The contexts can be illustrations or pictures. Words that a child recognizes in this phase are its own name or the name of the favorite candy.

• Guessing by arguments: The context is still very important by guessing the

meaning of an unfamiliar word, but the child starts to pay more attention to the similarities in form between known and unknown words.

• Sequential voicing: the decoding of words start. The child learns that there is a

relation between a written character or combination of characters and a sound. These characters and combinations of characters are always read in the same way in every word or sentence.

• Hierarchical decoding: in the final phase, the child learns that the pronunciation

of a character or combination of characters is depended on the context. The child can now read also words with a more difficult pronunciation correctly.

Learning to read in L2

The process described above explains how a child learns to read a first language (L1). But how do people learn to read a second language (L2)? Elly and Mangubhai (1983) classified three crucial differences between L1 and L2 learning. The first is strength of motivation. L1 learners need to learn to read, because they need it for communication. They are intrinsically motivated. Often L2 learners do not need to learn to read the language to communicate with friends or family. Although there are intrinsic motivated L2 learners, many L2 learners need to be extrinsically motivated to learn to read. The second difference is the amount of exposure to the language. People do often have books and magazines in their first language and receive letters and emails in their first language. However, the amount of books, magazines, letters and emails written in L2 someone has or receives is generally lower. The third difference is the type of exposure to the language. The exposure to the first language is usually natural, but exposure to the second language is often “planned, restricted, gradual, and largely artificial.” (p. 55.).

L2 learners are less intrinsically motivated, have less written input and the type of input is not natural. Besides these different conditions, Bernhardt (1991) adds that learning to read in L2 involves two types of processes different from learning to read in L1. The first process is to learn the linguistic rules of the new language. The reader must learn what the L2 phonemes are to read aloud correctly. To understand the meaning of the texts the reader has to learn the meaning of the words, but also the grammar rules to determine the relations within and between sentences. Sometimes the rules of the first language can be used, sometimes new rules have to be learned.

1.2 The process of reading comprehension

So far is described how a person learns to read in a first and second language. However, no attention has yet been paid to the question that was addressed in the first place: what is reading? There is no clear answer to this question, because – as stated before – there is no clear theoretical model of the process of reading (Van Elsäcker, 2002), but a number of models will be discussed to focus in the end on the model that is the most widely accepted (Paris & Hamilton, 2009).

Theories of reading comprehension

Perhaps the oldest models of reading comprehension are the bottom-up models. A bottom-up model assumes that the processing of a word goes from character recognition, via word comprehension, to sentence comprehension and in the end to text comprehension. This process is linear and goes in only one direction. From characters to text, from bottom to top (Thomassen, Noordman & Eling, 1991; Van Elsäcker, 2002). An example of a top-down model is the serial stage model of Gough (1972). The focus of this model is on the lower cognitive processes word encoding and lexical access. Only when children have mastered the principles of word decoding they will be able to read sentences and texts.

The opposite of bottom-up models are the top down-models. Top-down processing of a text indicates that comprehension of a higher level stimulates the comprehension at a lower level. For example, comprehension of the sentence structure stimulated word recognition and word comprehension stimulates the recognition of the separate characters of the word (Thomassen, Noordman & Eling, 1991; Van Elsäcker, 2002).

amount of attention is available. This attention has to be shared by all four ‘machines’. By automation of the decoding process, more attention can be paid to the other processes, which will improve reading comprehension.

The final type of reading comprehension models are the interactive comprehension models. These models assume that both bottom-up and top-down processes are used at the same time during reading and that they interact with each other (Van Elsäcker, 2002). The most popular model of this type is the construction-integration model of Van Dijk and Kintsch (1983) (Paris & Hamilton, 2009). This model will be discussed detailed.

The construction-integration model

According to Van Dijk and Kintsch (1983), a reader of a text creates a mental representation of the text. This mental representation exists of a text base and a situation model. The text base is ‘the semantic representation of the discourse input in episodic memory’ (p. 11). To construct the text base, the information in a sentence is changed into one or more propositions and added to the propositions of the previous text that are already in the text base. The situation model is ‘the cognitive representation of the events, actions, persons and in general the situation, a text is about’ (p. 11-12). To construct the situation model, previous experiences and general knowledge of the reader are added to the text base.

The written discourse is the input of the model. This input is processed on different levels. These levels are not distinguished based on linguistic features (an orthographic level, a semantic level, a syntactic level), but on the complexity of the input: first the words, then clauses, complete sentences, sequences of sentences and in the end the complete text. The model is an interactive model: it assumes that higher and lower level processes influence each other.

formulated hypothesis about the meaning of a word, clause or sentence is based on not only the discourse input, but also on the knowledge of the reader.

Using strategies, the meaning of the words in the input is established. The words are put into propositions: the predicate-argument schema of the meaning of one sentence clause. This proposition is based on the meaning of the words and the syntactic structure of the clause. There are two types of proposition. The atomic proposition, written as PREDICATE [ARGUMENT, ARGUMENT,…] and the complex predicate, written as:

Category (action, event or state)

> Predicate:

> Arguments (agents, objects, source, goal, …): >> Modifiers:

> Circumstances >> Time: >> Place:

The sentence Yesterday, Mary gave Fred the old book in the library can thus be written as:

GIVE [MARY, BOOK, FRED] OLD [BOOK] YESTERDAY [GIVE[.]] IN-LIBRAY [GIVE[.]] Or as: Predicate: GIVE Arguments: Agent: MARY Object: BOOK Modifier: OLD Goal: FRED Circumstances Time: YESTERDAY Place: LIBRARY

One type is indirect coherence: ‘the meaning units are part of the same episode. (…) they share a time, place, or argument.’ (p. 39). A second type of local coherence is direct

coherence. The relation is the same as indirect coherence, but is marked by for example

signal words, connectives or sentence structure. The final type of local coherence is

subordination: a complete proposition is part of another proposition. These connected

propositions together form the text base.

The process of constructing text base and situation model is supervised by the control system. The control system ‘will supervise processing in short-term memory, activate and actualize needed episodic and more general semantic knowledge, provide the higher order information into which lower order information must fit, coordinate the various strategies, decide which information from short-term memory should be moved to episodic memory, activate the relevant situation models in episodic memory, guide effective search of relevant information in long-term memory and so on’ (p. 12).

The blueprint of the reader

Figure 1: the ´blueprint´ of the reader (Perfetti, 1999)

This ‘blueprint’ of the reader ‘represents the information sources that a reader would be expected to use in gaining comprehension of written language’ (Perfetti, 1999, p. 168). Pre-reading processes are used to identify the characters within the lines, curves and dots of the written or typed text. These characters become the visual input for the reading process.

The specific characters activate corresponding grapheme units: characters, character groups and entire words. But not only grapheme units are activated, also phonological units that correspond to the input. These activated phonological units increase the activation of corresponding grapheme units. This is called phonological

mediation and accelerates the word identification process. With the activation of

possible meanings and linguistic information that might be needed for sentence comprehension.

After word recognition, the correct semantic meaning has to be chosen based on the context of the word. Linguists do not agree about how this selecting of the correct meaning takes place. According to the selective access model (Glucksberg, Kreuz & Rho, 1986) word meaning activation is strongly influenced by context and only meanings that can be correct in the given context are activated. The opposite view is the multiple access model, which states that, before the correct meaning is chosen based on the given context, all possible meaning of the word are automatically and quickly activated. Than the word meaning is selected, which corresponds best with the information from the context (Seidenberg et al., 1982; Kintsch & Mross, 1985; Onifer & Swinney, 1981). In his overview of research on meaning selection, Simpson (1994) states that the overall pattern of results supports the multiple access model, but that not all meaning are activated equally: it depends on the frequency of the meaning and of the context of the word.

For comprehension at sentence level, the word meanings have to be combined with other sources of information to build a mental representation. In this process, the selected words go through a syntactic parser. This parser puts the words into a syntactic structure. However, inferential processes and general linguistic knowledge influence this syntactic structure. It is an open question whether the parser builds the syntactic structure first or is immediately influenced by the other sources. This part of the model of Perfetti is not in line with the CI-model of Van Dijk and Kintsch (1983). They state that the clauses of a sentence are changed into propositions, but Perfetti puts them into a syntactic tree. Still, Perfetti calls this the text base, probably because it only contains the information from the text and the different parts of the texts are syntactically and semantically linked to each other. This text base changes into a situational model. Therefore, inferences with general knowledge have to be made.

A) North of the highway and east of the river is a gas station. B) On your left just after you cross the river you see a gas station.

Readers who read the text with sentences like A, the overview-text, could easier draw a map of the town. Reader of the other text remembered the literal sentences of the text better. This difference can be explained by the text base and the situation model. The readers of the overview-text were able to build a situation model of the town, but the readers of the other text could not, or less accurate, and used the text base to remember the text. According to Zwaan, Langston and Graesser (1995), time, space, protagonist, causality and intentionality are best represented in the situational model, because readers build situation models along these five dimensions. Linguistic information is better remembered by the text base.

1.3 Problems with reading comprehension

Besides a description of the blueprint of the reader, Perfetti also indicated which problems can arise in the model during reading. He distinguishes five different types of problems:

• Decoding problems: the reader is not able to process the written input due to

visual problems or a lack of vocabulary knowledge. The reader knows not enough words or not enough meanings of a word.

• Capacity problems: the working-memory can not handle the load of information

during reading.

• Phonological problems: the phonological processes of word identification and

word memory do not work properly.

• Syntactic problems: the reader has difficulties with the morphology and the

syntactic structure of sentences.

• Inference problems: the reader is not able to make the correct within-text and

without-text inferences.

called lower class process. Meso comprehension is comprehension by combining the meaning of different sentences and parts of texts. This process is an interaction of lower and higher class processes. Macro comprehension is comprehension of the entire text. This process is a higher class process. Hacquebord et al. (2004) linked problems with one or more types of comprehension to a certain reader. They distinguish three types of readers: the compensating reader, the school reader and the problem reader. The compensating reader has problems at the micro level. He finds it difficult to understand the meaning of separate words. Because of his good meso and macro abilities, he is still able to understand the general meaning of a text. The school reader is the opposite of the compensating reader. He has good micro abilities and knows the meaning of separate words and sentences. However, because of problems on meso and/or macro level, he is not able to understand the general meaning of a complete text. The problem reader has problems on micro, meso and macro level. He is likely to avoid reading tasks, which makes his arrear even bigger.

These different combinations of problems on micro, meso and macro level can be explained by the reading comprehension theory of Paris and Hamilton (2009). Their main assumption is that comprehension only occurs when the needed skills have reached a certain level: a threshold. Paris and Hamilton make two important notions about these thresholds:

1. the value of the thresholds depends on the text.

2. after a threshold is reached, the skill has not been mastered completely. The threshold is the minimum level for reading comprehension and the skill can be developed much deeper.

Thresholds explain how it is possible that the compensating reader has good meso and macro abilities, but problems on the micro level. He has reached the thresholds on the higher levels, but not on the micro levels. A problem of this theory - Paris and Hamilton mention it themselves - is how to determine the levels of the different thresholds.

1.4 The reading process and the present study

language, the reader must also understand what he reads. To reach this understanding he builds a text base and a situation model of the text. Together these two elements form a mental picture of the event described in the text. However, in the process of constructing this mental picture a lot of problems can occur, which causes problems with micro, meso or macro comprehension. Problems on the lower level can occur without problems on the higher level if the reader has reached the threshold for the higher order processes already. These thresholds are the minimum abilities a reader must have of a certain reading skill and are dependent of the text.

Hacquebord et al. (2006) found that vmbo-students develop language comprehension slower than havo/vwo students. This probably indicates that vmbo-students experience more reading problems on the micro, meso and/or macro level than havo/vwo students. They need more time and perhaps more help to reach the thresholds of different reading skills. The purpose of this chapter was to emphasize the importance of appropriate language use in school books by making the reading problems clear. The next chapter will focus on text analyses as a way to indicate the language use in school books.

CHAPTER 2: TEXT ANALYSIS

Discourse analysis studies the use of language within its cultural and social context and aims to find patterns across texts. The term discourse analysis was used first by Zellig Harris in 1952 as a way to analyze the connection between parts of written texts or speech (Paltridge, 2006). Discourse refers to both written and spoken language (Renkema, 2004). However, there are a number of differences between these two types of discourse, which influences the way of analyzing (Paltridge, 2006). A few of these differences are:

• Clauses in spoken discourse have more complex relations and a greater distance to each other than in written discourse.

• Written discourse has a higher percentage of content words than spoken language. • The final difference is grammatical metaphor (Halliday, 1989). This means that

parts of discourse are not put in the grammatical class that one should expect. This is the case with nominalizations, for example. Written texts have more grammatical metaphors than spoken texts (Paltridge, 2006)

Within the field of discourse analysis, the analysis of written discourse has its own place. The goal of studying written discourse is ‘to find how texts are structured and how texts can be made useful’ (Schellens & Steehouder, 2008, p.1). By analyzing what the different elements of a text are and how these elements are connected, knowledge about the text arises. This knowledge can than be used to indicate the quality of a text (Pander-Maat, 2002; Renkema, 2004).

2.1 Types of text analyses

Studying written discourse can be done in many different ways and there are many ways to order all these types of analyses.

The second ordering is from Schellens and Maes (2000). They distinguish three different types of analyses based on the person who performs the analyses: a text-focused analysis is an analysis in which the writer of the text analyses his own text and gives a judgment. The most common example of text-based analysis are readability formulas. In an expert-focused analysis an expert gives a judgment about the text. This expert can both be a communication expert or an expert regarding the content of the text, like a lawyer or a doctor. In a reader-focused analysis is the text given to a group of the target readers of the texts to find out how well the text is adjusted to this target group (Schellens & Maes, 2000). Within the reader-focused analysis, different subtypes can be distinguished. There is a distinction between on-line and off-line methods. On-line methods are methods that measure processes during the reading of a text. Off-line methods measure the result of certain processes after reading of the text (Renkema, 2004).

The third and final division is from Schellens and Steehouder (2008), who distinguish different types of text analyses by the language aspect of the text where the analysis focuses on. They distinguish the functional analysis, the coherency analysis, the rhetoric analysis, the argumentation analysis and the genre analysis.

It is important to understand that a study is always a combination of subtypes of the different divisions. For example, a study can be a qualitative, expert-focused, function analyses. In the present study, two types of text analyses will be used to indicate the language use of texts in school books: coherence analysis and measurements from readability formulas. These types of analyses will be discussed in more detail.

2.2 Coherency analysis

relational coherence (Hustinx & Karreman, 2008). Referential coherence is coherence caused by referring back or forth to other text elements (Pander-Maat, 2002). Relational coherence is coherence, caused by a meaningful relation between sentences or parts of sentences (Hustinx & Karreman, 2008).

Research into coherency has revealed that the use of elements of coherence influences the difficulty of a text. Verhoeven (1991) found that referring to persons, time and space are the easiest types of referential coherence, referring to complete sentences turned out to be more difficult and referring to actions most difficult. Further, referential coherence becomes more difficult when the distance between the referring element and the text elements it refers to becomes bigger or when the referring element is ambiguous (Bos-Aanen, Sanders & Lentz, 2001). A study of Sanders and Noordman (2000) into relational coherence revealed that a text can be processed faster when it contains coherence markers and that the type of coherence relation between sentences influences processing speed and recall.

Coherency analysis in school books

Coherency analysis can be done on all kinds of discourse. As the focus of the present study is on language use in school books, two studies about coherence in school books will be discussed in more detail.

Jentina Land (2009) studied the effect of coherence markers and text lay-out on text comprehension of student from vmbo-bb, vmbo-kb and vmbo-tl. She compared two hypotheses that exclude each other. The first hypothesis is the hypothesis of maximal

coherency. This hypothesis states that a text is the easiest to understand when the

relations between texts elements are marked. This will make it easier for a reader to understand the relations within a text and improve text comprehension. This should especially be true for readers with little knowledge about the topic of the text. The other hypothesis is the hypothesis of minimal cognitive load. This hypothesis states that short sentences will take less cognitive energy to process. There will be more cognitive space left to analyze the meaning of the sentences. Coherence markers make sentences longer. Therefore, coherence markers make texts more difficult to understand.

Land adjusted eight texts from history books. From each text, she made an integrated version with many coherence markers and with the sentences presented in paragraphs. She made also a fragmented version with only a few coherence markers and with each sentence placed on a new line. She presented these texts to vmbo students and asked them questions about the text after they read the texts. Land found that students from all school levels answered the question better when they had read an integrated version of the text than when they had read a fragmented version. This effect was most clear for the vmbo-bb students. She concluded that the integrated text is easier to understand, thought to the number of coherence markers and the presentation of the text in paragraphs. This supports the hypothesis of maximal coherency.

than in kb-books and tl-books. These results indicated that publishers write their books corresponding to the hypothesis of minimal cognitive load.

2.3 Readability formulas

Readability formulas are a way of language use research that focuses on text properties that are measurable, like average sentence length and average word length (Noordman & Maes, 2000). A readability score indicates how easy a text is to understand or to comprehend due to the style of writing (Klare, 1963).

Readability formulas are conducted by collecting a number of texts with different levels of difficulties. These texts are analyzed on a number of measurable characteristics like average sentence length. Statistical processes are used to find correlations between the appearance of the characteristics and the difficulty of a text and to determine which characteristics predict the language use of the texts best. These characteristics are put together in a readability formula (Renkema, 2004).

Although widely used they have received much criticism, because the formulas only indicate the difficulty of a text, but do not explain why it is difficult. B. Lively and S.L. Pressey developed the first readability formula in 1923 to measure the vocabulary burden of science books. Since than, many formulas have been developed, but the most widely used and one of the most reliable is the formula of Rudolf Flesh (1948). His formula for the Flesch Reading Ease score is:

Score = 206.835 – (1.015 x ASL) – (84.6 x ASW) Where:

ASL = average sentence length

ASW = average number of syllables per word (DuBay, 2004)

Also for the Dutch language readability formulas are developed. One of the newest is the CLIB formula of the CITO (Staphorsius & Krom, 2008). This formula is:

CLIB = 46 - 6,603GWL + 0,474PFREQ - 0,365PTYPES + 1,425PZW Where:

PTYPES: percentage types (percentage types)

PZW: percentage zinnen per woord (percentage sentences per word, the opposite of average sentence length)

Another measurement that indicates the readability of a text is percentage common words of a text (Hacquebord & Andringa, 2000). Hacquebord and Andringa (2000) counted per text, the number of words that are part of the list of the 2000 most common or most elementary Dutch words (Kleijn & Nieuwborg, 1999) and divided that number by the number of total words of the texts. They tested the difficulty of the texts and found that the percentage of common words predicted the difficulty.

Diataal

Part from readability formulas can not only be used to indicate the difficulty of a certain text, but also to find criteria for texts. Hacquebord, Andringa, Linthorst and Pulles (2006) analyzed many texts for different school levels for average word length, average sentence length and percentage of common words. They were able to conduct recommended values for these language aspects for readers with different reading abilities. They use this criteria for the Diataal, a set of tests to measure the reading abilities of children in the age of 10 till 15 year.

2.4 Text analysis and the present study

Students of different school levels have different reading abilities (Hacquebord et al., 2004). Therefore, the language use in books for different school levels should be adjusted to the readers to make sure that the content is comprehensible. Land (2009) made clear that an integrated text is easier to understand, especially for students with lower reading abilities. Readability formulas and Diataal (Hacquebord et al., 2006) made clear that the text aspects average word length, average sentence length and percentage of common words do indicate the difficulty of a text. Therefore, these language elements will be studied to find out if there are differences in language use in school books.

CHAPTER 3: DESIGN AND METHOD

3.1 Outline

A corpus of texts from science school books was analyzed in a quantitative way to study difference in language use. Language use in texts was operationalised into four measurements: average word length, average sentence length, percentage coverage of common words and number of coherence markers per expression.

The results were compared with criteria for language use to judge the appropriateness of the language use. The texts were also analyzed in a qualitative way to explain the results of the coherence analysis.

The research questions of this study are:

• Are there differences in language use in texts between science books for different levels of Dutch secondary education?

• Is the language use in texts in science books for different levels of Dutch secondary education appropriate for the school level of the reader?

3.2 Material

The school books

The texts were taken from two methods that incorporate biology, science and physics. The methods are developed for the first two years of secondary school. The first method is called Explora and is published by Noordhoff Uitgevers. The second method is called

Vita and is published by Malmberg. Malmberg and Noordhoff Uitgevers are two of the

three largest publishing houses for school books in the Netherlands. Vita and Explora are the only two methods in the Netherlands that combine the three mentioned subjects. Both methods work with both text books and assignment books. The analyses will only focus on the texts from the textbooks and not on the assignment books.

also contains 8 sections. From each book, there is a different version for bb, vmbo-kb/gt/tl (vmbo-kgt) and havo/vwo.

Text selection

The texts were only selected from the textbooks for the first year, because Noordhoff Uitgevers is working on a new edition of Explora and the books for the second year were not available yet.

The size of the corpus is based the corpus of Land (2009). For her analyses of school books, she selected on average 15 texts per method per school level and each text had an average length of 15 sentences. Per method, approximately the same amount of texts and sentences per text were selected

For Explora, from each of the six textbooks, two sections were selected. From each section, the first text and the final text were selected. The introduction text was skipped; titles and references to tasks were left out. This was done first for the vmbo-kgt book. From the vmbo-bb book and the havo/vwo book texts with corresponding subjects were selected. Most of the time, these were also the first and final text of the section. In total, 24 texts per school level were selected with an average number of 12 sentences.

For Vita, from each of the six textbooks, two sections were selected. From each section, only one text was selected, because the texts were in general much longer than the texts in Explora. Titles were left out and there were no introduction texts or references to tasks. Texts were selected from vmbo-kgt. Texts with corresponding subjects were selected from the vmbo-kb/bb and havo/vwo books Attention was paid to select both texts from the beginning of a section and from the end of a section. In total, 16 texts per school level were selected with an average number of 23 sentences. The selected texts can be found in Appendix A.

3.3 Procedure

Differences in language use in texts

common words. On meso level, the texts were analyzed on number of coherence markers per expression.

These four measurements are defined as follows:

Average word length (AWL). All characters were counted per text. This number was

divided by the number of words per text. Titles, subtitles and formulas were left out of this analysis.

Average sentence length (ASL). The number of sentences per text was divided by the

number of words per text. Titles, subtitles and formulas were left out of this analysis.

Percentage coverage of common words (PCCW). Per text, it was counted how many of

the words in the text were part of the 2000 most common or most elementary Dutch words (Kleijn & Nieuwborg, 1999). This was calculated as a percentage of all the words in the text. Titles and subtitles were included. Formulas were left out of the analysis.

For this analysis, the computer program TxtSCR was used. TxtSCR was used for the analysis of texts in the Diataal project (Hacquebord et al., 2006). The texts had to be adjusted for this program, because certain words that are not on the common words-list should still not be marked as ‘not common’. Those words are known by the reader, because it is explained in the text or because it is the name of a person or object. Names are not listed on the common word list, because there are too many names. Therefore, the following words were replaced by a common word before the analysis with TxtSCR:

• Names of persons and lands;

• Words that were explained in the text;

• scientific substances or materials which were referred to very often; • body parts, when mentioned more than onces in a text;

• references to illustrations, tables and other parts of the book or method; • Scientific units.

Further, abbreviations were replaced by the original word.

Number of coherence markers per expression (CMperE). The texts were divided into

markers were tagged. The coherence markers were all connectives and phrases that explicitly link the current phrase with an earlier phrase. For example, ‘the reason for this decision is…’. All coherence markers were counted and divided by the number of expressions.

Appropriate language use in texts

The outcomes of the first analysis were compared with criteria for appropriate language use. Average word length, average sentence length and coverage of common words were compared with the criteria from the Diataal project (Hacquebord et al., 2006). The criteria are given in table 3:

Average word length Average sentence length Percentage coverage common words Vmbo-bb/kb 4,8 11,5 86 Vmbo-tl/havo 5,0 13 83,5 Havo/vwo 5,2 14,5 81

Table 3: The criteria form the Diataal project for average word length, average sentence length and percentage coverage of common words (Hacquebord, 2008).

This data contains the recommended average word length, average sentence length and coverage of common words for the first class of vmbo-bb/kb, vmbo-tl/havo and havo-vwo.

Coherency analysis

Four texts from the science books were compared with a high coherent text from Land to find out whether differences in content could influence the number of coherent markers per expression. The average result for number of coherence markers per expression for the high coherent texts from Land was 0.79. A text from Land with a result close to this average (0.73) was analyzed. This text was compared with four texts from the vmbo-kgt science books. Two texts with high coherence and two texts with low coherence were analyzed. Each couple of texts existed of one text from Vita and one text from Explora. To indicate the relations between expressions, the method of Pander-Maat (2002) was used.

Pander Maat (2002) states that there are five kinds of coherence relations, each with a different number of subtypes. The first group is the causal relations-group. These relations are ‘expressions about a situation or an event which causes another situation or event’ (Pander Maat, 2002, p. 78). The second group is the characteristic relations-group. These are ‘expressions which provide information about the characteristics of an object, situation or event from a previous expression’ (p. 78). The third group is the group of the

explanation relations. These relations explain or illustrate a previous expression or part

of a previous expression. The fourth group is the reasoning relations-group. These are the relations between opinions and arguments or between a statement and the explanation for that statement. The final group is the nuances relations-group. This group contains expressions that contradicts, strengthens or weakens the previous expression. Each group contains four up to sixteen different subtypes.

of all subtypes of relations can be found with the corresponding prototypical questions. The new formulated questions are marked with an *.

The classification of Pander Maat (2002) is rather complete. However, he seems to assume that there is always a specific relation between two expressions and in the short texts he analyses there is. However, in longer texts sometimes a topic shift occurs or an expression provides just extra information about the topic without a clear relation with previous expressions. Therefore, two types of relations are added: expression – addition and topic shift. Expression – addition is also used when a sequence of expressions form an enumeration of causes, reasons, goals, and etcetera. These kinds of relations could also be marked as for example effect – cause, but the second cause does not only have a relation with the effect, but also with the first cause. The decision to mark these type of relations as expression – addition was also motivated by the fact that these expressions often contain coherence markers for addition, like ‘and’ and ‘also’. To distinguish them from the other kind of expression – addition relation, the type of expression that is added is placed between brackets. For example, expression – addition (cause). One final change is made to the division of Pander Maat: the texts in the science books sometimes refer to images, tables or other parts of the book. If there is no clear relation between such an expression and previous expressions, the expression is marked as reference to outside text.

3.4 Analysis

Differences in language use in texts

one-way ANOVA’s were run in SPSS 16.0 to find whether the differences in AWL, ASL, PCCW and CMperE between school levels in general and between school levels in each methods are significant. Alpha was set at 0.05.

Appropriate language use in texts

Two one-sample T-tests were run in SPSS 16.0 to compare the number of CMperE from the analyzed texts with the average number of CMperE of the two sets of texts from Land (2009).

Coherency analysis

CHAPTER 4: RESULTS

In the following section the results of the described analyses will be presented. First, the results for differences between school levels are presented; secondly, the results of appropriateness of language use and finally the results of the qualitative analysis of coherence.

4.1 Differences in language use in texts

In the following section the results of the described analysis will be presented. The results between school levels are presented first, the results within methods second.

Results between school levels

The results from the analysis are presented in table 1.

Vmbo-bb Vmbo-kgt Havo/vwo

AWL 4.76 (0.45) 4.86 (0.38) 4.97 (0.37)

ASL* 8.92 (1.94) 9.98 (1.31) 11.19 (1.96)

PCCW 85.15 (5.29) 85.15 (5.05) 84.38 (5.88)

CMperE 0.35 (0.19) 0.41 (0.14) 0.43 (0.15)

Table 4: mean results per school level for AWL, ASL, PCCW and CMperE, with the standard deviation between brackets. The * indicates a significant difference.

The differences between the school levels on average sentence length were significant, F(2) = 16.699, p< 0.01 (ANOVA). A Post Hoc analysis revealed that the differences between all school levels were significant. For the differences between vmbo-bb(kb) and vmbo-kgt and the difference between vmbo-kgt and havo/vwo p<0.05. For the difference between havo/vwo and vmbo-bb(kb) p<0.01. There were no significant differences between the school levels on average word length, coverage of common words and coherence markers per expression per school level.

Results within methods

First, the results for Vita are presented and secondly, the results for Explora. The mean results and standard deviations are the same as the numbers in table 2, 3 and 4, but now organized by method instead of by school level.

Vita

Vmbo-bb/kb Vmbo-kgt Havo/vwo

AWL 4.92 (0.49) 4.95 (0.39) 5.06 (0.41)

ASL* 7.70 (1.06) 10.22 (0.89) 12.09 (1.87)

PCCW 86.36 (4.49) 85.29 (4.49) 85.03 (4.29)

CMperE 0.34 (0.19) 0.43 (0.15) 0.44 (0.14)

Table 5: mean results for Vita per school level for AWL, ASL, PCCW and CMperE, with the standard deviation between brackets. The * indicates a significant difference between the different school levels.

The difference between the school levels on average sentence length was significant, F(2) = 43.25, p< 0.01 (ANOVA). A Post Hoc analysis revealed that the differences between all school levels were significant with p>0.01. There were no significant differences between the school levels on average word length, coverage of common words and coherence markers per expression per school level.

In Vita, the fluctuation of average word length, average sentence length and percentage coverage of common words per school level is low, but the fluctuation of coherence markers per text is very high.

Explora

Vmbo-bb Vmbo-kgt Havo/vwo

AWL 4.66 (0.39) 4.80 (0.37) 4.90 (0.34)

ASL 9.73 (1.98) 9.83 (1.54) 10.60 (1.83)

PCCW 84.34 (5.71) 85.05 (5.38) 83.95 (6.79)

CMperE 0.36 (0.21) 0.39 (0.13) 0.43 (0.16)

The tests (ANOVA) revealed no significant differences between the school levels on average word length, average sentence length, coverage of common words and coherence markers per expression.

Also in Explora, the fluctuation of average word length, average sentence length and ipercentage coverage of common words per school level is low, but the fluctuation of coherence markers per text is very high.

4.2 Appropriate language use in texts

The results of the described analyses are presented below. The order of presentation is average word length, average sentence length, percentage coverage of common words and number of coherence markers per expression.

Average word length

Level and criteria Vmbo-bb(kb) Vmbo-kgt Havo/vwo

Vmbo-bb/kb 4.8 4.76 4.86 4.97*

Vmbo-tl/havo 5.0 4.76* 4.86* 4.97

Havo/vwo 5.2 4.76* 4.86* 4.97*

Table 7: mean results of average word length per school level compared with the recommended average word length of Diataal. The * indicates a significant difference with the criteria from Diataal (Hacquebord, 2008).

The tests revealed that the average word length for vmbo-bb(kb) has no significant difference with the criteria from Diataal for vmbo-bb/kb, but is significant lower than the criteria from Diataal for vmbo-tl/havo (t(39)=-3.292, P.0.05).

The average word length for vmbo-kgt is significant lower than the vmbo-tl/havo criteria from Diataal. (t(39)=-2.301, p>0.05). There is no significant difference with the criteria for vmbo-bb/kb.

Average sentence length

Level and criteria Vmbo-bb(kb) Vmbo-kgt Havo/vwo

Vmbo-bb/kb 11.5 8.92* 9.98* 11.19

Vmbo-tl/havo 13.0 8.92* 9.98* 11.19*

Havo/vwo 14.5 8.92* 9.98* 11.19*

Table 8: mean results of average sentence length per school level compared with the recommended average word length of the Diataal project. The * indicates a significant difference with the criteria from Diataal (Hacquebord, 2008).

The average sentence length for vmbo-bb(kb) and vmbo-kgt score both significant lower than the vmbo-bb/kb criteria from Diataal. For vmbo-bb(kb), (t(39)=-8.423 with p>0.01. For vmbo-kgt, t(39)=-7.307 with p>0.01. The results for havo/vwo were not significant different from the criteria for vmbo-bb/kb, but significant lower than vmbo-tl/havo (t(39)=-4.004, p>0.01).

Percentage coverage of common words

Level and criteria Vmbo-bb(kb) Vmbo-kgt Havo/vwo

Vmbo-bb/kb 86.0 85.15 85.15 84.38

Vmbo-tl/havo 83.5 85.15 85.15* 84.38

Havo/vwo 81.0 85.15* 85.15* 84.38*

Table 9: mean results of coverage of common words per school level compared with the recommended percentage coverage of common words of the Diataal project. The * indicates a significant difference with the criteria from Diataal (Hacquebord, 2008).

Number of coherence markers per expression

Land

– low coherence

Result per school level Land

– high coherence

Vmbo-bb(kb) 0.07* 0.35 0.79*

Vmbo-kgt 0.07* 0.41 0.79*

Havo/vwo 0.07* 0.43 0.79*

Table 10: mean results of number of coherence markers per expression per school level compared with the low coherence result and high coherence result of the texts of Land (2009). The * indicates a significant difference with the criteria from Land (2009).

4.3 Coherency analysis

Seven texts were analyzed. First, the text of Land is presented, than the high coherent texts and finally the low coherent texts. The analyzed texts can be found in appendix C.

Text from Land (2009)

Coherence markers per expression: 0,73

Expressions Type of relation Group Coherence

marker 1 – 2 2a – 2b 2 – 3 3a – 3b 4a – 4b 3 – 4 4b – 5 5 – 6 7 7 – 8 8 –9 10 10 – 11 11 – 12 Expression - addition opinion – argument reason – action

expression – addition (action) reason – action expression - addition situation – judgement cause – effect topic shift situation – judgement reason – action topic shift reason – action time – situation - Reasoning Causal Causal Causal - Characteristic Causal - Characteristic Causal - Causal Characteristic No Yes Yes Yes Yes No Yes Yes No Yes Yes No Yes Yes

Table 11: The type of relation between expressions, the group of the relation and whether or not the relation was marked with a coherence marker

Text with high coherence

Vita – vmbo-kgt - module 2 Subject: how to furnish a room?

Coherence markers per expression: 0,63

Expressions Type of relation Group Coherence marker

1a – 1b 1 – 2 2 – 3 3 – 4 2 – 5 5 – 6 6a - 6b 6 – 7 6,7 – 8 8 – 9 9a – 9b 9b – 9c 9 –10 10 – 11 11 – 12 Goal – way

Expression – addition (way) Process – course

Term – definition Process – course Expression - addition Goal – way

Expression – addition (way) Explanation – statement Expression - addition Condition – situation Expression – completion Reason – action Condition – situation Term – definition Causal Causal Characteristic Explanation Characteristic - Causal Causal Reasoning - Causal Explanation Causal Causal Explanation Yes Yes Yes No No Yes Yes Yes Yes No Yes Yes Yes Yes No

Table 12: The type of relation between expressions, the group of the relation and whether or not the relation was marked with a coherence marker

Explora – vmob-kgt - module 6 Subject: Water

Coherence markers per expression: 0,73

Expressions Type of relation Group Coherence

marker 1 – 2 2 – 3 3a – 3b 3b 4 4 – 5 5a – 5b 5 – 6 4 – 7 7a – 7b 4-7 – 8 4 – 9 10 10 – 11 11 – 12 12a – 12b 13 13 –14 14 – 15 15 – 16 13,16 – 17 17 – 18 18 – 19 17 – 20 20 – 21 21a – 21b Explanation – statement Expression – addition Expression – example Reference to outside text

Topic shift Cause – effect Object – properties Cause – effect

Expression – addition (cause) Object – properties Expression – example Expression – weakening Topic shift Cause – effect Expression – addition Condition – situation Topic shift Expression – addition Condition – situation Expression – addition Explanation – statement

Expression – addition (statement) Statement – explanation

Expression – addition (statement) Statement – explanation Cause – effect Explanation - Explanation - - Causal Characteristics Causal Causal Characteristics Explanation Nuance - Causal - Causal - - Causal - Reasoning Reasoning Reasoning Reasoning Reasoning Causal Yes No Yes No No Yes Yes Yes Yes Yes Yes Yes No Yes No Yes No Yes Yes No Yes Yes No Yes No Yes

Table 13: The type of relation between expressions, the group of the relation and whether or not the relation was marked with a coherence marker

text also 1 time. Almost all relations are marked, with an exception for topic shift, reference to outside text and expression – addition, which are never marked and for reasoning, which is marked sometimes. In a sequence of five time reasoning, the relation is marked three times.

Texts with low coherence

Vita – vmbo-kgt - module 4

Subject: how do you smell with your nose? Coherence markers per expression: 0,18

Expressions Type of relation Group Coherence

marker 1a – 1b 1 – 2 2 – 3 3a – 3b 3 – 4 4 – 5 5a – 5b 6 6 – 7 7 – 8 8 – 9 9 – 10 10 – 11 11 – 12 Object – place Expression – addition Object – properties Effect – cause Object – properties Object – properties Situation – condition Topic shift Object – properties Expression – addition Object – similarity Process – course Process – course Process – course Characteristic - Characteristic Causal Characteristic Characteristic Causal - Characteristic - Characteristic Characteristic Characteristic Characteristic No No No Yes No No Yes No No No Yes No No No

Table 14: The type of relation between expressions, the group of the relation and whether or not the relation was marked with a coherence marker

Explora – module 2 Subject: senses

Coherence markers per expression: 0,16

Expressions Type of relation Group Coherence

marker 1a – 1b 1 – 2 3 3a – 3b 1,2,3 – 4 4 – 5 5 – 6 7 7 – 8 8 – 9 9 – 10 9 – 11 11 – 12 9 – 13 9 – 14 15 9-14 – 16 16 – 17 Result – activity Effect – cause Topic shift Condition – situation Example - term Term - definition Expression - addition Topic shift Expression - addition Expression - addition Group – members/parts Group – members/parts Expression – paraphrase Group – members/parts Group – members/parts Reference to outside text

Sequence of expressions – recapitulation Expression - paraphrase Causal Causal - Causal Explanation Explanation - - - - Characteristics Characteristics Explanation Characteristics Characteristics - Explanation Explanation Yes No No Yes No No No No Yes No No No No No No No No

Table 15: The type of relation between expressions, the group of the relation and whether or not the relation was marked with a coherence marker

The text contains 18 relations between expressions. There are twelve different subtypes of relations. The explanation relation occurs five times, the characteristic relation four times. The causal relation and expression – addition relation occur both three times. There are two topic shifts and there is one reference outside text. The causal relations and expression – addition relation are not always marked. The other relations are never marked.

Overview of differences and results

CONCLUSION AND DISCUSSION

The first research question was: are there differences in language use in texts between

science books for different school levels of Dutch secondary education? The results make

clear that there are no differences in average word length, percentage coverage of common words and number of coherent markers per expression. The only difference is in average sentence length. In the school books for vmbo-bb(kb) the shortest sentences are used, in the books for havo-vwo the longest sentences. However, the within methods-tests revealed that this is not a result that can be generalized to all science methods, because in Vita were the differences in average sentence length significant, but not in

Explora. Another noticeable result was the high fluctuation in coherence markers per

texts. This was found for Vita, Explora and in general.

This leads to the conclusion that there are no or only small differences in language use in texts per school level in science books, depending on the method one uses. Further, average word length, average sentence length and percentage coverage of common words are rather stable values, but number of coherence markers per expression fluctuated very much in all methods and on all school levels

The second research question was: Is the language use in science books for

different levels of Dutch secondary education appropriate for the school level of the reader? The results of the tests indicated that the answer of first research question need

some nuance. Although there were no significant differences between average word length, percentage common words and coherence markers per expression, the small not-significant differences were enough to cause differences in appropriateness. The results are discussed per measurement:

• Average word length: Despite there was no significant difference between the

• Average sentence length. Because there were significant differences between the

average sentence length, one would expect that the three results would all be on a different level. However, this can not be concluded, because the results for vmbo-bb(kb) and vmbo-kgt are both significant shorter than the lowest recommend average sentence length from Diataal. The average sentence length for havo/vwo is on the level of vmbo-bb/kb. This leads to the conclusion that the sentences all are shorter than recommended by Diataal. It is the question whether this makes the sentences easier or more difficult, because short sentences can indicate that a text is fragmented and has few coherence markers. This will be discussed in more detail later.

• Percentage coverage of common words. On percentage coverage of common

words, the results of vmbo-bb(kb) and vmbo-kgt were almost equal. The mean is the same, but the standard deviation is different. Surprisingly, this small difference causes different results for appropriateness. The PCCW of vmbo-bb is on the level of tl/havo and the PCCW of kgt is on the level of bb/kb. The percentage coverage of common words of havovwo is also on vmbo-tl/havo level. It is questionable whether the small difference between vmbo-bb(kb) en vmbo-kgt really causes a difference in appropriateness. What the results do indicate is that the texts for havo/vwo could contain more difficult words.

• Number of coherence markers per expression. This question is more difficult to

coherence markers while a high number of explanation and characteristic relations reduce the number of coherence markers. Science texts are about processes and concepts. Processes can be described with causal relations, but concepts needs characteristic relations. Further, both will need explanation relations. Finally, information in a science book is presented as objective as possible to the reader. Therefore, reasoning relations will not be used very often. This indicates that it may be possible to increase the number of coherence marker in science books, but that it is probably impossible to reach an average result of 0.79.

The purpose of the present study was to answer two questions about language use in science books for different school levels. The model of Van Dijk and Kintsch (1983) made clear that the reading process is a complex process and children need to acquire many strategies to become fluent readers. Perfetti (1999) pointed out that problems with acquiring these strategies can lead to decoding, capacity, phonological, syntactic and inference problems. Hacquebord et al. (2004) assigned certain kind of problems to certain kind of problem readers. The compensating reader has problems at the micro level. The school reader has problems on meso and/or macro level the problem reader has problems on micro, meso and macro level. These differences between readers were explained by the theory of Paris and Hamilton (2009), who stated that a certain threshold has to be reached before a certain reading ability is mastered on a basic level. These thresholds are dependent of the text.

Text analysis into coherence revealed that coherence markers make a text easier to read. Land (2009) found that school book texts with many coherence markers made the vmbo students also understand the texts better. She also found that the use of coherence markers in school books for vmbo seems very random. Prenger (2005) found that students have difficulties with the language use in mathematical assignments.

text more comprehensible may not always be correct as Land (2009) showed that short texts with few coherence markers are more difficult to understand than longer texts with coherence markers. Fortunately, they use quite a lot of coherence markers to indicate the relations between the short sentences, but this number fluctuates very much between different texts. This could indicate that publishers are not aware of the importance of coherence markers and do not instruct their writers on how often they must use them. This study made clear that publishers try to make comprehensible texts for their students, but that they underestimate the language abilities of especially the students of havo/vwo level. The language use in science texts books could be more adjusted when a different version of a school book is created and more attentions should be paid to the use of coherence markers.

Reflection

Although it was possible to formulate answers to both research questions, the current studies had some problems. In studies like Land (2009), Land et al. (2002) or Prenger (2005), texts analyses are always done with a group of reviewers. In the present study, there were no extra reviewers to discuss and control the texts.

Further, only two methods were compared which became a problem with the average sentence length, because Vita had a significant difference, but Explora had not. Now it was not possible to indicate which method is most representative for other science books. If more methods were used, a better answer could have been formulated about average sentence length.

Some insight has been created about coherence markers in science books. However, further research is needed. Land (2009) found the highest and lowest value for coherency, but the current study revealed that these values may be different for different texts.

BIBLIOGRAPHY

List of references

Andringa S. & Hacquebord, H.I. (2000). De moeilijkheidsgraad van schoolboekteksten als grondslag voor het vaststellen van tekstbegripvaardigheid. Toegepaste Taalwetenschap in Artikelen, 64, 83-94.

Bernhardt, E.B. (1991). Reading Development in a Second Language. Theoretical,

Empirical & Classroom Perspectives. Norwoord NJ: Ablex Publishing Corporation.

Bos-Aanen, J, T. Sanders & L Lentz (2001). Tekst, begrip en waardering. Amsterdam: Stichting Lezen.

Chall, J.S. (1996). Stages of reading development. Fort Worth TX: Harcourt Brace.

Dijk, T.A. van, & Kintsch, W. (1983). Strategies of Discourse Comprehension. New York: Academic Press.

DuBay, W.H. (2004). The Principles of Readability. Costa Mesa: Impact Information Plain Language Services.

Elly, W.B. & Mangubhai, F. (1983). The impact of reading on second language learning. Reading research Quarterly, 19 (1), 53-67.

Elsäcker, van W. (2002). Development of reading comprehension: The

engagement perspective. Dissertation. Nijmegen: Katholieke universiteit Nijmegen

Firestone, W.A. (1987). Meaning in Method: The Rethoric of Quantitative and Qualitative Research. Educational Researchers, 16, 7, 16-21.

Flesch, R. (1948). A new readability yardstick. Journal of Applied Psychology 32, 221-233.

Glucksberg, S., R.J. Kreuz & S. Rho (1986). Context can constrain lexical access: Implications for models of language comprehension. Journal of Experimental Psychology:

Learning, Memory, and Cognition, 12, pp. 323-335.

Gough, P.B. (1972). One second of reading. In J.F. Kavanagh & I.G. Mattingly (Eds.), Language by ear and eye. Cambridge, MA: MIT Press

Hacquebord, H.I. (1989). Tekstbegrip van Turkse en Nederlandse leerlingen in