1
Declaration of authenticity
MA Applied Linguistics - 2018/2019 MA-thesis
Student name: Gerjanne Rozemarije Alkema Student number: 2715139
PLAGIARISM is the presentation by a student of an assignment or piece of work which has in fact been copied in whole, in part, or in paraphrase from another student's work, or from any other source (e.g. published books or periodicals or material from Internet sites), without due acknowledgement in the text.
TEAMWORK: Students are encouraged to work with each other to develop their generic skills and increase their knowledge and understanding of the curriculum. Such teamwork includes general discussion and sharing of ideas on the curriculum. All written work must however (without specific authorization to the contrary) be done by individual students. Students are neither permitted to copy any part of another student’s work nor permitted to allow their own work to be copied by other students.
DECLARATION
• I declare that all work submitted for assessment of this MA-thesis is my own work and does not involve plagiarism or teamwork other than that authorised in the general terms above or that authorised and documented for any particular piece of work.
Signed: ______________________________________________________________________
2
The effects of CLIL education on writing
abilities in the L1
Gerjanne Rozemarije Alkema S 2715139
MA in Applied Linguistics
Faculty of Liberal Arts
University of Groningen
Supervisors:
Dr. R. G. A. Steinkrauss
Dr. M. C. J. Keijzer
Word count: 14 914
(excluding abstract, tables, references and appendices)
3 Acknowledgements
I would like to thank my supervisor, Rasmus Steinkrauss, for his feedback, patience, and support. I can honestly say that I would not have been able to finish this master thesis without him. I would also like to express my gratitude to the teachers and students at the Maartenscollege in Haren, who have helped in setting up the experiments for this thesis. Lastly, I would also like to thank Iris van Dommele, who helped me with testing the students of the Maartenscollege.
4 List of abbreviations
BIA+ Bilingual Interactive Activation + CLI Cross-linguistic influence
CLIL Content and Language Integrated Learning FL Foreign language
havo hoger algemeen voortgezet onderwijs (higher general secondary education) IC Inhibitory control
L1 first language
L2 second language
tto tweetalig onderwijs (bilingual education)
vmbo voorbereidend middelbaar beroepsonderwijs (preparatory secondary vocational education)
5 Table of contents
Abstract 6
1. Introduction 7
2. Background 9
2.1 Content and Language Integrated Learning 9
2.2 Cross-linguistic influence 12
2.3 Theories of the multilingual mind 17
2.4 Statement of purpose 20 2.5 Hypotheses 21 3. Method 23 3.1 Subjects 23 3.2 Materials 24 3.3 Procedures 25 3.4 Measures 26 3.5 Analysis 30 4. Results 32 5. Discussion 40 6. Conclusion 50 References 52 Appendices 55
6 Abstract
Many studies have been dedicated to the effects of Content and Language Integrated Learning (CLIL) education on the acquisition of a second language (L2), but far fewer studies have investigated the effects of CLIL education on the first language (L1) (e.g. Navés, 2009). Previous studies have found no negative effects on the L1 on lexical retrieval (Freije, 2015) and the general level of Dutch as measured by the final exams after secondary school (Admiraal, Westhoff & De Bot, 2006). Since one of the prerequisites for CLIL education in the Netherlands is that the L1 should not be negatively impacted (Nuffic), more research should be conducted into the effects of CLIL on the L1. The aim of this study is to investigate the effects of CLIL on the writing abilities in the L1, and in doing so, to contribute to a growing body of research. The study was conducted at the Maartenscollege in Haren, and 60 students of year 4 and year 5 in secondary education were tested, both CLIL and non-CLIL students. The students were asked to write texts on the same subject, which were then analysed on four different measures: fluency, lexical diversity, lexical sophistication, and whether or not there was an intuitive difference. The results showed no significant effects of CLIL education on L1 written proficiency on any of the measures, although the lexical sophistication shows some small differences between the two groups. These results are similar to what has been found in previous studies, namely that CLIL education does not affect the L1 in a negative way (e.g. Freije, 2015; Navés, 2009).
7 1. Introduction
In the last two decades there has been an increasing interest in bilingual education in the Netherlands (Nuffic). Parents have the option to enrol their children in early-bird programs on primary schools, and teenagers have access to secondary schools where half or all classes are taught in a language that is different from their native language (e.g. Freije, 2015). One of the types of bilingual education that is available is Content and Language Integrated Learning (CLIL). As the name CLIL suggests, the bilingual education is conveyed in a way that language and content are integrated (Pinner, 2013). In the Netherlands this means that student receive about half of their school subjects in the target language and half of the school subjects in the students’ native language. In this way, the students are taught a second language through meaningful interactions as they negotiate for meaning in various school subjects (Pinner, 2013). The most common target language for CLIL education in The Netherlands is English (Nuffic).
The effects that CLIL has on acquiring a second language (L2) are often beneficial. There is a large body of research that states the positive impact of CLIL in learning an L2 (e.g. Navés, 2009; Pinner, 2013). When CLIL students are compared to students who follow language lessons in the monolingual education system, CLIL students often score better on reading comprehension and oral proficiency in the L2 (e.g. Admiraal, Westhoff & De Bot, 2006). Much of this success is attributed to the higher exposure to the L2, as well as the increased need for L2 production, either in spoken or written form (e.g. Navés, 2009; Pinner, 2013). However, previously published studies on the effects of CLIL education have rarely focused on the effects of CLIL education on the students’ first language (L1). Nuffic, the organisation for internationalization of Dutch education states that one of the criteria for a school to offer CLIL education is that the level of Dutch proficiency should not be negatively affected (Nuffic). In light of these criteria set by Nuffic, the lack of research in this area seems controversial. There are few studies (e.g. Freije, 2015) that focus on the effects of CLIL education on the L1, and additional background literature on bilingualism is also reviewed.
8 The aim of this study is to explore the relationship between CLIL education and proficiency in the mother tongue, in particular the proficiency in written Dutch. Written language is an interesting topic in Dutch language education currently, as Dutch secondary school teachers as well as university lecturers have expressed concern with the general level of written Dutch (Elving & Van Den Bergh, 2015). The research questions to guide this paper are as follows: ‘How does written Dutch proficiency of Dutch CLIL students compare to written Dutch proficiency of their non-CLIL, monolingual peers?’ and ‘Is there a change in the fluency and lexical complexity of year 4 compare to that of the year 5 group?’ To answer these questions, this paper will first review a brief overview of the pre-existing body of literature that is related to CLIL education, as well as cross-linguistic influence between languages, and finally a general theory of the multilingual mind. Data was collected at a secondary school. The students were asked to write a short essay and answer some personal questions in a questionnaire. Later these texts were analysed for four different measures: fluency, lexical diversity, lexical sophistication and an intuitive rating of the texts. The findings of these tests were used to answer the research questions.
9 2. Background
2.1 Content and Language Integrated Learning
There are many students worldwide who receive education in a language that is not their first. There are different types of bilingual education, but this study focuses on a specific type of education that is called Content and Language Integrated Learning, also known as CLIL. As the name suggests, CLIL is a method of teaching where language learning and content learning are combined (Cenoz, Genesee & Gorter, 2013; Pinner, 2013). Students who partake in CLIL education are taught the same subject matter as the students in their year group on monolingual schools, but the difference is that for CLIL students a number of subjects are taught in a language that is not their first language (L1). Whereas other teaching methodologies rely on teaching language through tasks, CLIL transforms the target language into the tool through which other information is obtained. In this way, students are required to negotiate for meaning either with other learners or through texts (Pinner, 2013). This aspect of CLIL education is often cited as a more efficient way of language teaching than conventional methods, which lies in the increased exposure to and production of the target language (Cenoz, Genesee & Gorter, 2013; Pinner, 2013; Swain, 2000).
The concept of CLIL started to emerge and develop as a result of immersion education in the 1960s in Canada (Navés, 2009; Pinner, 2013). The aim of this type of education was to make the students fluent in both of the country’s official languages, French and English, whilst simultaneously teaching them their school subjects (Navés, 2009). Similar to the programs nowadays, there were full immersion and partial immersion programs, which start from either elementary school or secondary school. In full immersion programs all of the subjects are taught in the target language, whereas in partial immersion part of the subjects are still taught in the students’ L1 (Navés, 2009; Pinner, 2013;). From the start of the program, the educational track was carefully monitored for scientific documentation as well as to appease concerned parents (Navés, 2009). There were three main questions concerning CLIL. Firstly, whether the content would be acquired in a satisfactory manner,
10 secondly what the effects were of second language (L2) learning, and finally whether the L1 was affected by the L2 in any way (Navés, 2009; Swain, 2000). There were several general outcomes of these studies. With regards to content learning, it was found that the students had to have a certain level of L2 proficiency before they started the CLIL program in order for the content to be learned at an adequate level (Navés, 2009). The total immersion students and the partial immersion students performed at similar levels in content acquisition as the monolingually educated control group, however, the early partial immersion students lagged behind (Navés, 2009). There were also signs that the CLIL students were less proficient in their L1 after the first few years, but this difference disappeared after the third year (Navés, 2009). With regards to the L2 acquisition the early and late immersion students performed equally well and showed a command of the L2 higher than that of the monolingual students (Navés, 2009). As Pinner (2013) points out, nowadays CLIL programmes are definitely as efficient in teaching language and content subjects as non-CLIL programmes.
As a result of the numerous studies on early immersion education in North America, there are many different interpretations of what CLIL education should entail. Regardless of the natural variation in CLIL education between schools and countries, there are a number of general principles. The school should provide at least half of their education in an L2 with no detrimental effects on the uptake of content, whilst simultaneously developing the L1 to the level of the CLIL students’ monolingual peers (e.g. Pinner, 2013; Swain, 2000). These principles are echoed by Nuffic, the organisation for bilingual education in the Netherlands (Nuffic). Nuffic stresses that the exam results for the language subject ‘Dutch’ should not be negatively affected compared to the country’s average (Nuffic). To date a considerable amount of literature has been published on the effects of CLIL on the students’ command of the target language, but Cenoz, Genesee and Gorter (2013) as well as Pérez Cañado (2017) state that there is still a shortage of research on the effects of CLIL on the students’ command of their first language (L1).
11 Since there are a myriad of factors in place in successful education, it is important to look at CLIL in Europe and the Netherlands in particular. The sociolinguistic and sociocultural contexts of immersion are completely different from Europe in North America, where the most crucial difference is that the immersion aspect of CLIL education in Europe is usually limited to the school itself (Cenoz, Genessee, Gorter, 2013; Freije, 2015; Péres Cañado, 2017). Furthermore, CLIL schools in the Netherlands only provide the opportunity for partial immersion, with approximately 50% of all lessons taught in English (Admiraal, Westhoff & De Bot, 2006; Freije, 2015). According to Pérez Cañado (2017) the results of studies regarding CLIL education as it was first introduced in Europe have an overall positive outlook, whereas more recent studies provide a more nuanced view (Pérez Cañado, 2017). In her own research, Pérez Cañado (2017) looks at the effect of CLIL on both the knowledge of the L1 as well as the subject knowledge of the students involved. Results show that there was no negative effect of CLIL on the L1 nor on any of the subjects that was taught in the target language. Furthermore, the CLIL students outperformed the non-CLIL students in content knowledge, although this was likely modulated by the students’ socioeconomic status. The participating CLIL schools in her study are private schools, whereas the participating non-CLIL schools are public schools, and there is a difference in the general performance level between the two types of schools (Pérez Cañado, 2017).
A study by Admiraal, Westhoff and De Bot (2006) looks into the effects of CLIL on content knowledge, as well as the L1 (Dutch) and L2 (English) proficiency of participating students. The results of CLIL students were compared to those of a group of non-CLIL students with a similar level of education. They found no negative effect of CLIL on the content knowledge, and the results of the reading comprehension test and the oral proficiency test in the L2 were performed significantly better by the CLIL group. There was no indication that the CLIL group was less proficient in the L1 than the non-CLIL group. However, Admiraal et al. (2006) state that the results may not be an accurate representation for the results of CLIL education, since there are many extramural factors at play in language development and acquisition. Moreover, the only measure of Dutch proficiency in this study was the final exam grade on the national exams (Admiraal et al. 2006). This measurement gives a
12 limited view of the Dutch proficiency, and cannot pinpoint the results in a specific area of language proficiency. This is especially relevant as the exams in the subject Dutch have been a cause of concern in the Netherlands for a number of years (Elving & Van Den Bergh, 2015; De Glopper, Van Kruiningen & Jansen, 2015), and may therefore also lack a degree of accuracy when taken as a measure for language proficiency. Another study was conducted by Freije (2015), who reviews the results of a vocabulary knowledge test and a picture naming task for Dutch CLIL students with English as their target language. Overall, the results of this study indicate that CLIL students outperformed the non-CLIL students on the English tasks, and this study found no negative effect of non-CLIL on the L1 of the group as a whole. However, there were some individual cases of students who were struggling with lexical retrieval. This is a characteristic that is also often found in bilinguals (e.g. Laufer, 2003; Pavlenko, 2000), which is what will be discussed in the next section.
2.2 Cross-linguistic influence
As was mentioned in the previous section, there is a lack of research of the influence of CLIL on L1 proficiency (e.g. Pérez Cañado, 2017). The following section is included to provide a broader picture of what other studies have found to be the influence of the L2 on the L1. The section starts with what has generally been found, and ends with specifically the influence of the L2 on L1 writing.
Numerous studies have been dedicated to describing the influence of the first language (L1) on learning a second language (L2). These studies include many different types of situations, such as immigrants who move to another country and who have to learn a second language, but also children who may be exposed to a second language after they have already started to learn their first language, as well as students who start to learn a second or a third language when they go to high school or university. In these cases, it has often been observed that language learners use the knowledge of their L1 to fill in the gaps in knowledge in the L2 (Laufer, 2003). If the L1 influences the L2, it is logical to assume that the L2 also influences the L1, at least to some extent. The extent to which the L2 is influenced by the L1 has been studied extensively in attrition studies. Attrition is the partial loss of
13 language structures in the L1, often as a result of living abroad for an extended period of time (e.g. Laufer, 2003; Pavlenko, 2000). Although attrition is unlikely to occur in CLIL students, the knowledge gained from the studies that describe attrition effects could give useful insights into the effects of cross-linguistic influence. Likewise, Laufer (2003) mentions that in the general L2 research area studies have thus far predominantly focused on the ‘non-advanced learners of second languages’ (p. 19), or in other words the learners who study an L2 in a classroom setting for no more than four hours a week. The CLIL students in the Netherlands are not related to the participants in attrition studies, but they are more advanced than the participants in L2 learner studies. They may share characteristics with both groups, and the studies that have been carried out in both of these areas can provide some insight into the influence of the L2 on the L1.
As was mentioned before, a large part of the literature that concerns itself with the cross-linguistic influence of the L2 on the L1 is focused on attrition. Attrition studies typically study immigrants who live in a country where the language spoken is a different language than the immigrant’s L1, and as such this person has learned the other language as their L2 (e.g. Laufer, 2003). Studies have shown that immigrants show signs of L2 influence on the L1 even after short periods of time (Pavlenko, 2000), but in general the effects are noticeable after 5-10 years (e.g. Marian, 2009), as opposed to the CLIL students in the Netherlands who typically follow a four year program (Freije, 2015). As was found in the attrition studies, there are several factors that mitigate the extent to which the L2 influences the L1. These factors include the age of acquisition of the L2, where younger learners are influenced more than the older learners; the fluency in the L2, where in particular the more fluent L2 speakers are affected more in their L1; and the amount of exposure to each language, where increased exposure in the L2 is related to attrition in the L1 (Marian, 2009; Pavlenko, 2000). These factors influence the L1 of the attriters in several areas of language proficiency, such as in word-choice.
Word-choice is the area that is often most easily affected by the presence of the L2 (Odlin, 2009; Pavlenko, 2000). In this area, phenomena such as borrowing, shift and loss can be found as signs
14 of L2 influence, although word loss tends to occur only in serious cases of attrition and is therefore not relevant to the CLIL students in this paper. Borrowing, the process of adding elements of the L2 to the L1, is often observed in speakers who use loanwords and calques, which can in turn become part of the L1 (Pavlenko, 2000). An example of a loanword from English in many other languages is the word ‘computer’, which is copied into many languages in its entirety, whereas a calque constitutes a literal translation of the borrowed word, such as how ‘skyscraper’ is translated as ‘wolkenkrabber’ in Dutch. A shift can be a semantic shift, where the L1 semantic representations are extended to include the semantic representations of the L2 words, which may have a similar form but not the same meaning (e.g. partial cognates). Another type of cross-linguistic influence between the L2 and the L1 can be seen in lexical retrieval, or the difficulty of finding words in general, which is also observed in general bilingualism (e.g. Dijkstra & Van Heuven, 2002; Freije, 2015; Marian, 2009).
The next part of this section will treat the literature that is written specifically about cross-linguistic influence (CLI) in writing. Since there are few studies about CLI on the L1 in writing, some studies on writing in the L2 are also consulted. In a study on free written expression of Russian immigrants in Israel, Laufer (2003) has studied the effects of CLI on collocational knowledge and lexical diversity. The subjects in this study were Russian immigrants who worked in Israel and learnt Hebrew as their L2 while maintaining their Russian by speaking it daily at home. Laufer (2003) found that the L1 lexical diversity as well as the fluency, as measured in the total number of words, decreased as the length of residency in Israel increased. Although signs of CLI are easily detectable when code-switching in speech, Laufer (2003) argues that the effects of the L2 on the L1 are not so obvious when writing. Laufer (2003) suggests that the difference between competent writers and writers with a transitional competence, such as learners or attriters, can be seen in their use of varied vocabulary, and the percentage of non-frequent words. Learners and attriters often tend to avoid risks when writing. As such the percentage of non-frequent words, as well as the lexical diversity and the total number of words tends to be lower for this group of people (Laufer, 2003). Furthermore, the study concluded that the Russian immigrants were less likely to recognize faulty collocations in their L1 as their contact
15 with Hebrew increased. Although these findings may be signs of attrition, Laufer (2003) consciously attributes the CLI effects to Cook’s (1992) framework of multicompetence: it is logical that the L1 is affected somewhat since the Russian immigrants had successfully learnt a second language, which is not a negative accomplishment overall. Cook’s (1992) framework of multicompetence will be discussed in more detail in section 2.3.
Another area in writing that is subject to cross-linguistic influence, is rhetoric (Babaii & Ramazani, 2017; Pavlenko, 2000). Babaii and Ramazani (2017) studied the effects of rhetorical patterns in the L2 on the production of these rhetorical patterns in the L1. There were two groups of subjects in the study: thirty university students who majored in English, and thirty non-English major students. Whereas both groups showed similar organisational levels in their essays, the group of English major students mainly used a deductive rhetorical pattern in the essays written in Persian, which is more typically an English rhetorical pattern than a Persian rhetorical pattern. Babaii and Ramazani (2017) also found a positive effect of lexical and structural complexity in the L1 and the overall proficiency in L2 writing. Babaii and Ramazani (2017) attribute the increased complexity in the Persian essays to a higher proficiency in English, due to the transfer of English rhetorical patterns. Hashemian (2011) and Mehrabi (2014) have conducted similar studies on the influence of the L2 (English) on the L1 (Persian) in writing. Hashemian (2011) has investigated the influence of the L2 on structure and complexity in the L1, and found that proficiency in English was strongly positively correlated with proficiency in Persian, and concludes that the L2 can affect the L1 in a positive way. Hashemian (2011) cites Cook’s (1992) theory of multicompetence as an interpretation of the results. Mehrabi’s (2014) study focuses on a myriad of criteria that are best described as general writing proficiency. Mehrabi (2014) found that the more proficient L2 English speakers were also more proficient in writing in Persian, and attributes this relationship to the increased metalinguistic knowledge that bilinguals are likely to have. However, these studies disregard that the CLI in these cases is probably bidirectional. In other words, the increased complexity in the Persian essays by English majors could also mean that the more
16 advanced L1 writing leads to a higher proficiency in L2 writing, rather than the other way around. Regardless of the interpretation, there is clearly a link between L1 and L2 writing skills.
Similar results have been found in studies on the relationship between L1 and L2 writing skills in secondary schools in the Netherlands. Schoonen, Van Gelderen, De Glopper, Hulstijn, Simis, Snellings & Stevenson (2003) investigate the roles of lexical knowledge, metacognitive knowledge, and the accessibility of the lexical knowledge on the production of both written Dutch (L1) and English (L2). They found that the accessibility of lexical knowledge has a stronger correlation with L2 written proficiency than L1 writing proficiency. In addition, Schoonen et al. (2003) suspect that another relevant factor in L1 written proficiency is the way in which working memory is employed. Furthermore, Schoonen et al. (2003) conclude that general metacognitive knowledge and writing proficiency in the L1 contribute to the L2 writing proficiency, and therefore that L1 proficiency is a strong predictor of L2 proficiency. Schoonen, Van Gelderen, Stoel, Hulstijn and De Glopper (2011) found similar results as Schoonen et al. (2003) in their study on the relationship between the L1 writing proficiency and the proficiency in English of secondary school students. Schoonen et al. (2011) found that the relationship between written proficiency is highly correlated to the metacognitive knowledge, grammatical knowledge and fluency, as their model that used these predictors could account for a large part of the variance. However, Schoonen et al. (2011) offer a more nuanced view on the relationship between Dutch L1 and English L2 writing abilities, as they found no clear indication that there was a relationship in the writing abilities between the two languages in the two years that they tracked the students’ writing abilities. Nevertheless Schoonen et al. (2011) point out that the relationship between the two languages’ writing abilities may still be present, but it may be difficult to notice since the relationship does not change over time and has been present since the beginning. The last study in writing that will be reviewed in this paper, is that of Snellings, Van Gelderen and De Glopper (2004), who have studied the relationship between second language writing and lexical retrieval. Their subjects were secondary school students from grade nine, ages fourteen to fifteen, who were split into two groups. One of the groups was subject to training of specific words, and the other
17 group was not. Snellings et al. (2004) found that the group of trained students used the words that were trained more frequently in the essays written afterwards than from the group of students who were not trained. The trained students produced better overall essays than the students who were not trained. Snellings et al. (2004) attribute the increase in performance of the trained students to an increase in ease of lexical retrieval: the words that occur more frequently and are paid attention to in the environment of the students, are more easily retrieved than the words that are not.
The general finding of studies about writing in a secondary school setting is that the writing skills in the L1 and the L2 are related to each other. Often when a student is good at writing in the L1, they are likely to be good at writing in the L2, too. However, the exact relationship between the L1 and L2 writing skills is still unclear. There may be some underlying skills, such as the skills that comprise metacognitive knowledge, that are shared across languages or may exhibit positive influence between different languages. However, the cross-linguistic influence can be viewed as either positive or negative. An example would be the accessibility of words in either language. According to Dijkstra and Van Heuven (2002), bilinguals may show difficulty in locating specific words, or start code-switching when they cannot find a word in one language. In writing, this could mean that words that are trained or used at a higher frequency in one language (e.g. as in Snellings et al. (2004)) are more easily accessible. In the case of languages that are similar, the lexical cross-linguistic influence can be helpful with words that are similar, or sabotage the retrieval of words that are very different.
2.3 Theories of the multilingual mind
The previous sections considered the findings of some studies related to the influence of an L2 on an L1. As the effects of bilingualism have been found in many different studies, there have been attempts at theories to generalize and understand what happens in the bilingual mind. In the past, the language abilities of bilinguals have often been compared to those of monolinguals. The mistakes that bilingual speakers make were often given as evidence that the other language was interfering in a negative way. Cook (1992) argues that these beliefs were reflected in the language processing models of the time
18 and that bilingual language processing was treated as a side-concern for many of the models, which made them inaccurate. In recent years, the language processing models have started to reflect what was seen in multilinguals. One of the concepts that was introduced is the concept of multicompetence (Cook, 1992). Cook (1992) defines multicompetence as ‘the compound state of a mind with two grammars’ (Cook, 1992, p. 557), as opposed to the monocompetence of monolinguals. This also means that, as there are different levels of proficiency in languages, this multicompetence is a spectrum and no two people will share the same multicompetence. One of the key arguments of multicompetence is that multilinguals are not necessarily bad at their language use, but that their proficiency is different from that of their monolingual peers. This is relevant to the CLIL learners as they are shifting from monolinguals to multilinguals, and although they may or may not show some signs that their L1 is changing as a result of CLIL education, they are still highly proficient in their L1.
Before Cook (1992) proposed his idea of multicompetence, Soares and Grosjean (1984) proposed a theory about bilingual language processing. They proposed that there are different language modes which multilingual speakers may employ in their daily language use: the multilingual mode and the monolingual mode (Soares & Grosjean, 1984). When the context requires a multilingual speaker to employ only one language, they are in a monolingual mode, and when the multilingual speaker is in an environment where others speak the same languages, they may switch to the multilingual mode (Soares & Grosjean, 1984). One of the prerequisites of this theory is that the languages of the multilingual speaker are not separated in the mind, but that there is at least partial overlap of the two or more languages (Soares & Grosjean, 1984). Cook (1992) comes to a similar conclusion, and argues that both the total integration of all the languages in the mind, as well as the total separation of all languages are unlikely. Since bilinguals are able to code-switch at some instances and keep their languages separated at another, it is likely that the languages of a multilingual are part of the same system where some features are interconnected (Cook, 1992).
19 In their study, Soares and Grosjean (1984) test the reaction times of Spanish-English bilinguals in a lexical decision task in both monolingual and multilingual speech modes. They found that the bilinguals in monolingual mode were equally fast at responding to target words as monolinguals, but the bilinguals took longer to respond to non-words than monolinguals (Soares & Grosjean, 1984). This evidence supports the view that the bilingual speakers not only search for a word in the language they are supposed to use, but they also search for the word in the language they are not currently using. Moreover, when the bilinguals were presented with a code-switched word in the bilingual mode, they took longer to respond than to the same-language words in the monolingual mode (Soares & Grosjean, 1984). These results support the view that bilinguals never fully switch off the other language when they are looking for words, but at the same time know what words belong to each particular language (Soares & Grosjean, 1984). These results show that the languages in the mind are interconnected at some level, and could also explain why multilingual people can both keep their languages apart and mix them together on separate occasions.
The next theory that is related to the ability of bilinguals to switch between languages, is Green’s (1998) theory of Inhibitory Control (IC). The IC framework stated that automated responses, such as the activation of both languages in a bilingual speaker, can be controlled. Green (1998) bases his model of inhibitory control on two main assumptions. The first assumption is that language processes are similar to other actions in which control is executed, such as running or washing the dishes. The second assumption is that there are many processes involved in getting from the original thought of speaking to actually speaking a coherent sentence, and therefore control must be executed at many different stages (Green, 1998). If a person wants to speak, multiple options are activated of which the less appropriate options have to be inhibited. For example, if this imaginary person has both a dog and a cat at home, but only wants to talk about their pet cat, the response ‘pet dog’ is inhibited. For bilingual speakers, there is an additional category to inhibit, which is the correct language of the current response. In addition to that, if this person were to talk about their pets in French significantly more often than in English, the French response is more difficult to inhibit. Some responses may be
20 inhibited more than others, depending on the frequency of use, and the similarity of different responses, as well as contextual cues (Green, 1998). In written language there is ample time and there are many additional steps that give people time to think about what it is that they want to write. This means that that the influence of the L2 on the L1 is likely to be small in writing, which is relevant for the current study.
Although IC can be understood as a system of language production on its own, these three aforementioned theories form the groundwork of some of the recent models of bilingual language processing. The Bilingual Interactive Activation + (BIA+) model is one of these models. Whereas the full model is very elaborate and irrelevant to this study, there are a few ways in which it explains language processing as one coherent theory rather than the aforementioned theories on their own. The BIA+ model is a language integrated model, where all the languages of one speaker are interrelated, as Cook (1992) proposed. As a result of the languages being integrated, both languages are active at all times, as proposed by Soares and Grosjean (1984). The languages in the system are separated through inhibitory control (Green 1998), and neighbouring words are activated according to the relative strength of their connection in relation to other words (Dijkstra & Van Heuven, 2002). However, the BIA+ model has not incorporated language learning, but others (e.g. Jacquet & French, 2002; Lieven, 2010) argue that language learning is simply strengthening the connections between words. Through frequent use of words in their related contexts, the representations in the mind are linked and the connections are strengthened (Lieven, 2010). The difference between CLIL and non-CLIL students is that CLIL students have more frequent input in English, and therefore less input in Dutch than the non-CLIL students. The question that remains is whether or not the increased exposure to English is significant enough to interfere with Dutch proficiency.
2.4 Statement of purpose
The specific objective of this study is to explore the relationship between CLIL education and writing proficiency in the mother tongue. Currently, ample research has been done about the effects
21 of CLIL on the target language, as well as the potential differences between monolingual and bilingual minds. As was argued in the literature review, students who partake in CLIL education are somewhere on the spectrum between L2 learners and bilinguals, and are therefore part of neither group exclusively, and there is a lack of research for this group of students specifically. This study provides new insights into the effects of CLIL on the mother tongue.
The specific research questions for this study are as follows: ‘How does written Dutch proficiency of Dutch CLIL students compare to written Dutch proficiency of their non-CLIL, monolingual peers in terms of fluency, lexical diversity, lexical sophistication, and qualitative aspects?’ and ‘How do the fluency and lexical diversity of CLIL students in year 4 compare to year 5?’
The data collected for this study comes from texts written by students from year 4 and year 5 in 2018 of the Maartenscollege in Haren. In both year groups, there is a CLIL group and a non-CLIL group. The students were all asked to write a short essay on the same topic. The essays of year 4 were then analysed on fluency, lexical diversity, lexical sophistication, and qualitative aspects, whereas the essays of year 5 were analysed only on fluency and lexical diversity.
2.5 Hypotheses
First of the aforementioned measures is the fluency test. Laufer (2003) has studied the effects of length of residency in an L2 country on the fluency in the L1. Laufer (2003) measured fluency as the number of words written per minute, and has found that the total number of words written decreases as the length of the residency in the L2 country increases. For this study, the CLIL students have a higher exposure to the L2. Therefore, if they are affected negatively in their L1 by their increased exposure to English, they are expected to write fewer words than the non-CLIL students. However, since the exposure to Dutch is increased in year 5, the hypothesis is that the former CLIL students of year 5 will not be affected and write a similar number of words to the non-CLIL students.
Next is the hypothesis for the lexical diversity part of the research question. As discussed in the literature section, there are different findings of the effects of a second language on the first
22 language with regards to lexical diversity. Laufer (2003) found that the lexical diversity of the L1 decreased when the exposure to the L2 was increased, whereas Babaii and Ramazani (2017) found some evidence, although inconclusive, that the general lexical complexity of texts written in the L1 was increased as a result of increased exposure to the L2. It is therefore difficult to predict what the outcome will be, since the CLIL students could both benefit from their metacognitive skills (e.g. Hashemian, 2011; Schoonen et al., 2003) and be at a disadvantage for lexical diversity due to increased exposure to the L2 (e.g. Laufer, 2003). It is therefore unlikely that the CLIL group will show a large difference to the non-CLIL group in lexical diversity in year 4, although it might be there, and it is hypothesised that there will be no difference in the year 5 group.
The third part is the test for the lexical sophistication between the CLIL and non-CLIL groups of year 4. Although there were no studies that directly addressed lexical sophistication in the literature review, Babaii and Ramazani (2017) found that there is a small indication that the bilingual students have written more complex essays than the monolingual students. The hypothesis is that the CLIL group may show a small difference in word use compared to the non-CLIL group.
The fourth measure of this study is a test to determine whether there is an intuitive difference in any qualitative aspects of the texts by CLIL students. Studies show that there is some form of difference in written language use between monolingual and bilingual people (e.g. Babaii & Ramazani, 2017; Laufer, 2003), and therefore it is possible that CLIL students write differently than their non-CLIL peers. The hypothesis is that the texts written by CLIL students can be differentiated from the texts by non-CLIL students.
23 3. Method
The aim of this study is to explore the relationship between CLIL education and Dutch writing abilities, and to observe whether there is a difference in Dutch writing abilities between CLIL and non-CLIL students. The students were asked to write a short essay on a given topic and answer a short questionnaire with some background information. The essays were then analysed with regards to complexity, accuracy and frequency to determine some differences in language use between the CLIL and non-CLIL students. This section will explain the exact method in more detail.
3.1 Participants
The participants in this study were recruited through the Maartenscollege in Haren, a secondary school that has both a bilingual (CLIL, TTO) and monolingual (VWO) educational track. From here on out the bilingual, CLIL groups will be referred to as TTO 4 and 5, and the monolingual Dutch groups will be referred to as VWO 4 and 5. In total, 60 students participated in the study. However, after removing a number of entries due to various reasons which will be explained later, 45 students remained. Their characteristics are summarized in Table 1. The CLIL students are part of the experimental groups, and the non-CLIL students are part of the control groups. The non-CLIL groups are called ‘VWO 4’ and ‘VWO 5’, whereas the CLIL groups are called ‘TTO 4’ and ‘TTO 5’.
Table 1. Characteristics of participants
Group Number of students Number of females Mean age
TTO 4 13 6 15.7
VWO 4 13 10 15.7
TTO 5 5 5 17.4
VWO 5 14 10 16.8
Table 1 shows the characteristics of the students who were left in the experiment after other students were removed from the experiment due to various reasons.
The Dutch school system is organized into primary and secondary education. Primary education is usually for children ages four through twelve, and secondary school starts upon completion of primary education. There are three different levels in secondary school, called
24
voorbereidend middelbaar beroepsonderwijs (vmbo, literally translated ‘preparatory secondary
vocational education’), hoger algemeen voortgezet onderwijs (havo, ‘higher general secondary education’) and voorbereidend wetenschappelijk onderwijs (vwo, ‘preparatory scientific education’) (‘Secondary education’, 2019). The vmbo track takes four years, and students are prepared for a job in a practical field. The havo track takes five years, and students can follow through to colleges that are both practical and scientific. The vwo track takes six years, and prepares students for university (‘Secondary education’, 2019). The students who are involved in this study are all vwo students of the same school, half of whom follow a partial immersion CLIL track, whereas the other half follows the monolingual track. CLIL education is often synonymous with bilingual education in the Netherlands, or
tweetalig onderwijs, hence the abbreviation tto. The CLIL track ends after four years, and students
follow monolingual education in the last two years to prepare for the national exams that are administered in Dutch.
All students were asked to fill out a questionnaire with some background information, which was used to remove the students whose background was in some way different from the other students in the group. Responses were removed if the students’ mother tongue was not Dutch, if the student was bilingual from birth, or if they lived abroad in an English-speaking country for an extended period of time. Furthermore, some students were eliminated based on their performance on the task: some students had not taken the task seriously and were thus incorrect estimates of their Dutch writing abilities at this point in time (e.g. one student had written the essay in English; another student had only written that they did not want to write the essay). Dyslexic students were not removed, since the time limit was set to be sufficient for all students.
3.2 Materials
During the experiment, all participants were given the same booklet that contained one writing task and three questionnaires. This booklet can be found in Appendix A. The writing task, as well as all questionnaires were filled out on paper. The first part of the booklet was a writing task, accompanied
25 by three pages of writing space. The task instructions were written in Dutch, and translate to: ‘Write an essay wherein you compare the two sides of the Black Pete argument, and give your own opinion in the last paragraph.’ This topic was chosen as it is a topic laden with emotion that is likely to elicit a response from most students in the Netherlands. The experimenter set a time limit of 25 minutes, wherein the remaining time was announced at 15 minutes and 5 minutes before the end. The writing task was the main task of the experiment, and was analysed on a variety of variables, which are elaborated on in section 3.4.
The second part of the booklet contained three short questionnaires. The first questionnaire was about language attitudes and motivation, and consisted of 15 questions. Students were asked to rate statements about three different languages on a five-point Likert scale, 1 being ‘strongly disagree’ and 5 being ‘strongly agree’, where a higher score corresponded to a more positive attitude. The languages in this questionnaire were Dutch, English and a foreign language they were taught at school (i.e. French or German). The second questionnaire was about language use, and again students were asked to rate the statements on a five-point Likert scale, where 1 was ‘not very often/never’ and 5 was ‘very often/always’, with a higher score corresponding to more frequent use of the language. This questionnaire contained 10 questions, and the languages concerned were again Dutch, English and either German or French. Both questionnaires were adapted from Dörnyei and Chan (2013). However, these questionnaires were deemed irrelevant for the analysis of the final results, and will therefore not occur any further in this paper. Finally, the participants were asked to fill out a third questionnaire about their language background and other personal details, such as age and gender. This questionnaire was used to remove participants from the poule who had different backgrounds.
3.3 Procedure
The experiment took place in a classroom at the Maartenscollege in Haren. All students were tested in a group setting at different time-slots, which means that first the TTO 4 group was tested, then the VWO 4 group and finally the year 5 group, which consisted of both the TTO 5 and VWO 5 subgroups.
26 Prior to the experiment, consent forms had been sent out through the school to the parents, where they could object to participation of their child in the study as well as request the results from the study. Anonymity of the participants was ensured.
Participants were first presented with the first part of the booklet, the writing assignment, and instructed not to read the rest of the booklet just yet. They were told that they had 25 minutes to complete the task, and if they had any questions they could raise their hand and the researcher would come to answer this question. They were asked specifically to remain quiet during this part of the experiment, but the VWO 4 group had some instances where the class would become a little loud. Since the time limit was set to allow all students to finish the assignment on time, most students were done around the twenty-minute mark. After 25 minutes had passed, the students were told they could fill out the remainder of the booklet, which consisted of three questionnaires: language attitudes and motivation, language use, and personal details. They had approximately 15 minutes to do this. Again if they had questions, they could ask the researcher about them. Finally, upon handing in the booklet, each participant received a chocolate bar as a small thank-you for their participation.
3.4 Measures
There were four measures in this study that were used to evaluate the language used in the texts. These measures are fluency, lexical complexity, lexical sophistication and an intuitive rating based on qualitative aspects. This section of the paper describes the way in which these measures were operationalized.
Fluency
The way in which fluency was operationalized is by measuring the word count of each individual essay. As the texts had all been written down on paper, the first step was to transcribe the texts on the computer, without correcting any spelling errors yet. As most text editors are quick to correct mistakes, texts were initially typed into Notepad, which does not correct for errors. These files
27 were saved like this for later use (i.e. in the intuitive rating part of this experiment), and for the next steps a copy of the original files was made.
For this study, all of the texts had to be prepared according to CHAT standards (MacWhinney, 2000). This consisted of making sure that each sentence started at a new line; commas, hyphens, brackets and quotation marks were removed; full stops had to be preceded by a space; capital letters were removed except for with proper nouns; and the ID part of the transcript was filled in. Furthermore, for this specific study some additional changes were made: abbreviations were expanded; numbers and year denominations were changed to ‘Jaar’; split verbs were put together, e.g. wij vragen ons af ‘we question’ was changed to wij ons afvragen; lexical and spelling errors were corrected; and the proper noun Zwarte Piet was changed to one word, i.e. ‘Zwartepiet’. These alterations were all carried out to make both the word count as well as the lexical diversity measure as accurate as possible. For any native speaker in this age group at this school it would be safe to assume they know how to use numbers and years correctly, and therefore more numbers in-text should not contribute to the lexical diversity, hence the substitution ‘Jaar’ for either. The split verbs would interfere with an accurate word count as well as with the lexical sophistication later on. Lexical and spelling errors were corrected for a more accurate word count. Finally, as ‘Zwarte Piet’ is only one person or concept, it was changed into ‘Zwartepiet’, as it was one of the most commonly occurring words and would alter the word count as well as the lexical diversity. After all these alterations, the texts were sufficiently altered for the word count to be calculated. There are different ways to calculate this using CLAN, but for this study the command that was used was ‘freq [filename].cha’.
Lexical diversity
To calculate the lexical diversity, the D function of CLAN was used rather than the more traditional token ratio for the lexical diversity, as D corrects for text length whereas the type-token ratio does not. After the texts were all treated for the fluency measure, the morphological tier
28 was added to calculate lexical diversity in CLAN. The Dutch mor package was used and then the mor tier was disambiguated. For this study, the D optimum average value was used.
Lexical sophistication
The lexical sophistication of the texts was operationalized through word frequency in this study. In order to obtain a corpus from the texts, the texts needed to go through CLAN software. Since the texts were already edited for CLAN in a previous part of the data processing, the same text files as were used for the lexical diversity measurement were used in this part. For this part of the data processing, only the two year four groups were used (TTO 4 and VWO 4), since there was such a large difference between the number of participants in TTO 5 (5 students) and VWO 5 (14 students).
For the disambiguation of the mor tiers, the Dutch mor package was used and then completed manually, to make sure the correct stem was tagged. However, a few alterations were made with regards to data processing of the corpus. A small number of words was used as both nouns and adjectives. These included blank(e)/blanken ‘white/white people’ zwart(e)/zwarten ‘black/black people’. These words have all been tagged as adjectives. Other than that, all nouns, verbs, adjectives and adverbs (lexical words) have been tagged correctly. The other word classes were not looked at as extensively, and therefore mistakes in tagging may occur but these are of no further consequence for the analysis. The additional morphosyntactic information was irrelevant for analysis in this paper and was therefore also removed. In short, all words in the mor tier were reduced to their stems.
After the mor tiers were disambiguated, CLAN was used to retrieve the 15 most frequent word stems of the lexical words, which are the nouns, verbs, adjectives and adverbs. After TTO 4 was treated in this way, the texts of VWO 4 were treated the same way. The lexical items had lost their tags when they were moved from the CLAN file to the excel file, so the lexical items (nouns, verbs, adjectives and adverbs) were tagged again. When all the relevant words had been tagged, the number of types and total number of each group was added up, and this information was stored in a table. Next, the 15 most commonly occurring lexical words were taken out per group. This meant that, for both VWO and
29 TTO, the 15 most occurring nouns, verbs, adjectives and adverbs of each group were taken out, to compare the language use of the two groups.
Intuitive rating
For this part of the study, the text files as first typed out for the word count and lexical diversity measures were used. These text files contained only the raw text, with all its spelling errors and paragraphs as they were written down by the students. In theory the original texts could have been used, too, but at this point the errors had already been marked on the written texts, which could cause a bias amongst the raters. The texts were put in a randomized order. The texts of year 5 were left out of this part of the experiment.
There were three raters for this part of the experiment, two women and one man, all 24 years old at the time of rating. They are all university students and have written longer texts in Dutch as part of either their study programme or their work life. They were provided with food and drinks as compensation.
The raters were asked to rate each year 4 text on the following premise: ‘TTO students are often perceived as being worse at Dutch than their monolingual VWO peers. This is likely due to the fact that TTO students receive 50% of their education in English, and are thus ‘missing’ valuable Dutch-speaking time. Can you categorize the texts as TTO and VWO based on this premise?’ There were 26 texts in total, of which 13 were written by the TTO students and 13 by the VWO students. The raters were told that half of the texts were written by VWO students, and the other half by TTO students. The raters were asked to rate however they wanted, whether that was based on a clear reason or a feeling. The rating took a total of approximately two hours, in which the raters were free to take a break whenever they wanted. After all texts had been categorized, the raters were shown how accurate they were. The full table of results of this part of the experiment can be found in Appendix G, and are discussed in sum in the results section.
30 3.5 Analysis
Fluency
After the word count was calculated for each text individually, an ANOVA was used to compare the average word count of the four different groups. For the statistical analysis, R software (R Core team, 2019) was used.
Lexical diversity
The value for lexical diversity (D) was calculated per individual text, and, using R software (R Core team, 2019), an ANOVA was used to compare D across groups, which is summarized in the results section of this paper.
Lexical sophistication
The analysis of the lexical sophistication was comprised of mostly qualitative analysis, when comparing the two groups on their word use and type of lexical words. First, the total number of words used per lexical category were calculated, and added in two tables: one for the VWO group and one for the TTO group. Next, the 15 most commonly occurring words of both the TTO group and the VWO group were put in a table. The words in these lists were then compared between the groups.
Intuitive rating
For the analysis of the intuitive rating part the study, the scores were tallied as correct number of guesses per rater as well as correct number of guesses per text. There are two options: either the text belongs to VWO 4 or TTO 4, and of each group there is an equal number of texts. Per text per rater there is 50% chance of guessing the correct group to which the text belongs. This means that, since there is an added premise that VWO students would be better at Dutch than their TTO peers, the final percentage of correct guesses could show one of three things. If the premise is correct, the raters would be able to more accurately predict which text is written by a TTO student and which text is written by a VWO student. In this case, the percentage of correct guesses would have to be significantly
31 higher than the chance level, which would be >50%. If the premise is incorrect and it is actually the TTO students who are better at Dutch than the VWO students, the percentage of correct guesses would have to be significantly lower than the chance level (<50%). The third option is that the premise is false, and TTO students’ writing cannot be distinguished from VWO students’ writing based on the premise. For this, the percentage of correct guesses would have to be around chance level (50%).
32 4. Results
To compare the difference between the students who followed bilingual CLIL education (TTO students) and the students who followed Dutch monolingual education (VWO students), a number of analytical procedures were carried out. This included the analysis of the fluency of the written texts and the lexical diversity. For the lexical sophistication and the intuitive rating the analysis consisted mostly of qualitative observations.
Fluency
The first part of the analysis concerns itself with the fluency measure. The way fluency was measured in this study was from the word count. These results are summarized in Table 2. As shown in Table 2, the mean length of essays was approximately the same. The range of essay length produced by the TTO groups is smaller than that of the VWO group. However, there was only a limited number of students in TTO 5 (n = 5), compared to the other groups. This should be taken into consideration when interpreting the data. An analysis of variance (ANOVA) was carried out to determine whether there was a significant effect of type of education or year of education on the word count of the essays.
Table 2. Results of fluency measure
Group Number of students Average word count Standard deviation
VWO 4 13 238 67
TTO 4 13 261 57
VWO 5 14 236 111
TTO 5 5 273 64
Table 2 shows the results of the fluency measure, as expressed through the average word count of the texts. The assumptions of normality have been met. There was no significant effect of year and type of education, F(3, 41) = [0.425], p > 0.05. This effect was of an insignificant size (r = .03). This means that the null hypothesis should be accepted (i.e. the year and type of education have no effect on the outcome of fluency).
33 Lexical diversity
The next part of the results section concerns itself with lexical diversity, as measured by CLAN’s D optimum average. As can be seen in Table 3, the VWO 4, TTO 4 and VWO 5 groups show a large range of different values for D, as the standard deviation suggests. TTO 5 consisted of a smaller group, and thus its range is smaller. When comparing the means of the groups, both TTO groups seems to have a small advantage over the VWO groups.
Table 3. Average value lexical diversity per group
Group Number of students Average value for D Standard deviation
VWO 4 13 69.8 14.8
TTO 4 13 83.7 17.0
VWO 5 14 74.5 15.2
TTO 5 5 78.4 6.8
After the assumptions of normalcy had been checked, an ANOVA was carried out to determine the effect of group on D. There was no significant effect of year and type of education, F(3, 41) = [1.060], p = 0.377 i.e. p > 0.05. This effect was of a small size (r = .125). This means that the null hypothesis should be accepted (i.e. the year and type of education have no effect on the outcome of D). However, there is an indication that there is a small effect (r = .125).
Lexical sophistication
The next part of the results is the lexical sophistication measure. The following three tables give some qualitative insights into the lexical sophistication of the texts of the year 4 TTO and VWO groups. The comparison was only carried out for the TTO 4 and VWO 4 groups due to the low number of participants in the TTO 5 group.
34 Table 4. Item types and total number of words of TTO 4 texts
Total item types Percentage of total item types Item types used once Total number of words Percentage of total number of words Verbs 143 21.4 85 612 18.0 Nouns 206 30.8 135 561 16.5 Adjectives 118 17.6 73 322 9.5 Adverbs 61 9.1 29 283 8.3 Other 141 21.1 1619 47.7 Total 669 3397
Table 4 shows the total item types, item types used only once, and the total number of words as they appeared in the corpus of the year 4 TTO students. The category ‘other’ contains the functional words of the texts such as prepositions, pronouns, particles, etc. For easier comparison with the VWO 4 group, the percentages corresponding to the total item types and total number of words have been calculated. Since the aim of this part of the experiment is to compare the two groups, some observations will be written after Table 5 with the VWO 4 lexical sophistication information has been introduced.
Table 5. Item types and total number of words of VWO 4 texts
Total item types Percentage of total item types Item types used once Total number of words Percentage of total number of words Verbs 116 20.8 71 602 20.3 Nouns 132 23.7 78 395 13.3 Adjectives 120 21.5 78 279 9.4 Adverbs 59 10.6 27 194 6.5 Other 131 23.5 1502 50.5 Total 558 2972
Table 5 shows the total item types, item types used only once, and the total number of words as used in the essays by the VWO 4 group. Tables 4 and 5 above show that the TTO group uses a larger total number of both item types and words than the VWO group. The total number of item types for the adjectives, adverbs and other words are relatively equal in numbers between the groups.
In short, for the nouns and verbs there is a noticeable difference in terms of total item types. There is a difference of 27 verbs and 74 nouns between the TTO and VWO groups. The percentage of
35 the verbs seem to be relatively equal between the groups (21.4% for TTO vs. 20.8% for VWO). However, the TTO group uses a larger percentage of nouns than the VWO group, at a 7.1% difference. A large number of these nouns has only been used once (135 of 206, roughly 66%), whereas in the VWO group this number is lower, at 78 of the 132, which is about 59%. The second largest difference between the groups in terms of percentage is the relative difference in the adjectives, the VWO group uses 3.9% more adjectives than the TTO group, although the total number of adjectives is similar in both groups.
Turning now to the total number of words, there two groups are fairly similar. The total number of words for the TTO group is slightly higher than that of the VWO group. However, this difference is insignificant, as was shown earlier in the part dedicated to the fluency measure. The category with the largest difference between the two groups, is the nouns category. At a 3.2% difference, the TTO group has a slightly higher overall percentage of nouns in all its texts combined than the VWO group. The VWO group has a higher percentage of total number of function words (‘other’). The VWO group also has a higher total percentage of verbs, which is a 2.3% difference of the total amount compared to the TTO group.
The next table, Table 6, provides some information about the fifteen words that are used most commonly by both groups. For each group, thirteen texts have been used.
36 Table 6. The fifteen most common words per lexical word class
Verbs Nouns Adjectives Adverbs
TTO 4 VWO 4 TTO 4 VWO 4 TTO 4 VWO 4 TTO 4 VWO 4
Word Q Word Q Word Q Word Q Word Q Word Q Word Q Word Q
ben 139 ben 151 mens 46 mens 42 zwart 25 zwart 23 niet 50 niet 54
word 48 word 36 discussie 25 kind 29 racistisch 22 racistisch 13 er 31 er 32
vind 39 vind 32 traditie 25 slavernij 14 blank 19 ander 12 ook 26 ook 21
heb 27 heb 28 jaar 20 traditie 14 donker 17 gekleurd 11 zo 19 veel 16
moet 25 moet 26 kleur 19 discussie 13 ander 13 gewoon 11 vaak 14 wel 13
ga 17 kan 26 kind 18 kleur 13 groot 9 leuk 10 nog 12 zo 13
kom 13 zal 25 feest 15 slaaf 13 goed 8 heel 8 veel 12 natuurlijk 11
verander 12 kom 18 kant 13 feest 12 meer 8 blank 6 al 10 toch 11
zal 10 voel 16 slavernij 13 jaar 12 alleen 7 meer 6 nu 10 al 10
zie 10 ga 14 mening 11 kinderfeest 12 heel 7 wit 6 weer 10 nog 7
houd 9 denk 13 slaaf 10 schoorsteen 9 lang 6 alleen 5 vooral 7 helemaal 6
maak 9 maak 12 schoorsteen 9 zelf 9 minder 6 donker 5 wel 7 nooit 6
wil 9 doe 11 verhaal 9 racisme 6 erg 5 goed 5 iets 6 nu 6
blijf 8 zie 10 roet 8 volwassene 6 moeilijk 5 slecht 4 steeds 6 weer 6
bedoel 7 bedoel 7 tegenstander 8 kant 5 rood 5 eigenlijk 4 altijd 4 hier 5
kan 7 hoor 7 roet 5 erg 4 net 4 mee 5
sta 7 verhaal 5 nou 5
werk 7 vroeger 5
Table 6 shows the lexical words that have been used most frequently by TTO 4 and VWO 4. The list is in some places longer than fifteen items, since the last items occur equally often. The column ‘Word’ signals the word stem that is concerned, and the column ‘Q’ is the corresponding quantity that this word occurs. The full list can be found in Appendix F.
37 There are several observations that can be made from Table 3. A large number of words that have been presented in this table occur in both the TTO list and the VWO list of the sample. It should be clear that all these observations are simply observations without statistical evidence.
The first category in the table is verbs. Most verbs occur in similar quantities in both groups. The greatest exception to this observation seem to be both kan ‘can’ with 7 items for the TTO group, and 26 in the VWO group; and zal ‘shall’ with 10 counts in the TTO group and 25 in the VWO group. From the TTO list there are seven verbs that don’t occur in the first fifteen of the VWO list: blijf ‘stay’,
discrimineer ‘discriminate’, zeg ‘say’, wil ‘want’, sta ‘stand’, houd ‘hold’ and werk ‘work’. For the VWO
list there are three verbs that are not included in the TTO first-fifteen list: voel ‘feel’, doe ‘do’, denk ‘think’.
The second category in Table 6 is nouns. Most of the fifteen most common nouns occur in both lists. On average, the TTO group uses more tokens of each noun in the list of the most commonly occurring words than the VWO group, with a few exceptions. From Tables 4 and 5 it was clear that the TTO group used more nouns than the VWO group, although a large number of these nouns occurs only once. From Table 6 it is clear that the TTO group also uses a larger number of words of the most commonly occurring words.
Turning now to the adjectives, the first observation is that this is the most varied list of most commonly occurring words. There are seven words on the VWO list that don’t occur on the TTO list, but that do occur lower down the list, in a lesser quantity. These words are gewoon ‘normal’, wit ‘white’, slecht ‘bad’, eigenlijk ‘actual’, leuk ‘fun’ and gekleurd ‘coloured’. Vice versa, there are five words on the TTO list that don’t occur on the VWO list: groot ‘big’, lang ‘long/tall’, minder ‘less’,
moeilijk ‘difficult’ and rood ‘red’.
Adjectives seems to be the word group that has a large variation between the groups, considering they wrote about the same subject. Both groups have a number of words that are used a
