Native and Non-Native Dutch Primary
School Students
Master thesis
Julia Tunderman 6143938 07 July 2017 Klinische Ontwikkelingspsychologie Supervisor: Dr. Patrick SnellingsAbstract
The aim of this study was to gain insight into how non-native primary school students compare to their native peers in their development of several Dutch and English language skills when considering their language background and experience with English. 146 students attending 4th grade participated (Native group N = 96, Non-Native group N = 50). The native group was expected to perform better on all Dutch tests. The non-native group was expected to perform equally well or better on the English tests than the native group. All students
completed parallel Dutch and English tests for vocabulary, spelling, and fluency. A repeated measures analysis showed that there were significant differences between the native and non-native group on the Dutch vocabulary measure, in favor of the non-native group. The difference between the groups for the Dutch spelling and fluency tests was not significant. No significant differences across groups were found for any of the English tests. Experience with English outside of school positively affected the outcomes of the English language tests. This effect was smaller when English students were isolated. These findings provide partial support for the Threshold and Developmental Interdependence Hypothesis and prove the importance of experience with a language for mastering it.
Introduction
Societies are constantly changing, which affects the composition of a country’s
population, as well as the skills necessary to be successful in the ever-globalizing world. In the Netherlands, children need to acquire a sufficient level of language skills in Dutch to complete their academic career successfully, but they will also need to acquire enough English so that they can keep up with the fast-paced globalization. For this reason, good language education, both in Dutch and English, is essential and should start as early as possible. In the Netherlands, the population consists of large groups of people with a different ethnic background than Dutch. Various studies have looked at differences in achievement between native and non-native children for Dutch language abilities, but far less research has been dedicated to looking at differences in English language abilities. The present study will answer the following question: What is the effect of language background and experience with English outside of school on native and non-native Dutch primary school students’ performance on Dutch and English language tests?
For Dutch language abilities, differences were found between native and non-native Dutch children, where the latter group usually scores lower than the former (Blom, 2010; Schoonen et al., 2002; Trapman, van Gelderen, van Steensel, van Schooten, & Hulstijn, 2014; van Gelderen et al., 2003). These lower scores are often notable on vocabulary tests (Calvo &
Bialystok, 2014; Schoonen et al., 2002; Trapman et al., 2014; van Gelderen et al., 2003;
Verhoeven, 2000), which, for non-native students with a smaller command of Dutch, can have a negative effect on their academic career (Appel & Vermeer, 1998; van Gelderen et al., 2003; Verhallen & Schoonen, 1993). The differences in academic career outcomes found between the two groups cannot generally be attributed to intelligence (Dufva & Voeten, 1999), so it might be solely attributable to their Dutch language proficiency.
The difference in Dutch vocabulary knowledge could be explained by the amount of exposure native children have had to Dutch. As opposed to native Dutch children, non-native children likely hear a different language in their home environment (Appel & Vermeer, 1998; Scheele, Leseman, & Mayo, 2009; Verhallen & Schoonen, 1993). By definition, this means that children from non-native parents are less exposed to Dutch, which could negatively affect their command of the language. The number of second-generation immigrants has grown substantially since the aforementioned articles were published (Centraal Bureau voor de Statistiek, 2016). Since the publication of these studies, non-native second-generation families may have incorporated Dutch in their homes, and might now speak a mixture of their native language and Dutch. The amount of exposure to Dutch non-native children have may thus have changed over the last decade. This could mean that the differences in scores on the Dutch
vocabulary tests reported earlier may not differ as much anymore between native and non-native students.
Similar to Dutch language exposure, children need to be exposed to English, too, in order to expand their command of this language. English language education plays a vital part in providing students with basic knowledge. In the Netherlands, English language education is obligatory from the fifth grade onwards, but some schools start teaching English in the first grade. These so-called Early Bird schools believe that early exposure allows children to master the language with little effort (EarlyBird, 2017). The focal point of English education in primary schools is vocabulary acquisition. National regulations have been formulated, which has
resulted in a list of English words and concepts that all students must have mastered by the time they start secondary education at the age of 12 (Thijs, Trimbos, Tuin, Bodde, & De Graaff, 2011). Apart from vocabulary acquisition, attention is also paid to some grammar and spelling rules.
Apart from exposure in schools, children in the Netherlands are exposed to English more and more in everyday life. English is heard frequently in television programs, music, online on YouTube or social media, and in games (Lindgren & Muñoz, 2013). All these channels are available to children in the Netherlands from an early age onwards. This kind of exposure is known to cause children to learn English more easily (Lindgren & Muñoz, 2013). Children whose ethnic background is different than Dutch may experience less exposure to English, because they may be exposed to their native language more than to English. Non-native parents may watch television from their home countries and may encourage their children to watch with them (Peeters & D’Haenens, 2005). This likely results in less exposure to English, which could affect non-native children’s proficiency in English. It seems that more exposure to English outside of school can positively affect a student’s development of English.
Apart from exposure to languages, motivation to learn a specific language can also contribute to a child’s acquired proficiency in said language (Cenoz, 2003; Murphy, 2005; Petscher, 2010). The motivation a child has to study or learn new things has been shown to affect both their overall performance in school, as well as their performance on specific
language tests as measured by school performance in that language (Spinath, Spinath, Harlaar, & Plomin, 2006; Steinmayr & Spinath, 2009). Studies have reported that children from ethnic minorities could experience less motivation to do well in school or to learn a new language such as English (Escamilla, 2009). This might affect their proficiency in English. Another factor that could contribute to less proficiency in a foreign language, is parental encouragement to do well in school or on English language tests (Lindgren & Muñoz, 2013). It seems that quite a few factors can affect a child’s development of a foreign language.
Various studies reporting on language acquisition in bilingual children have suggested that these children may have an advantage over their monolingual peers when it comes to
vocabulary acquisition (Cenoz, 2003; Keshavarz & Astaneh, 2004; van Gelderen et al., 2003). Several researchers have looked into the underlying mechanisms of second language
acquisition. Cummins (1979) suggested that a certain threshold needs to be reached before there can be an advantage for bilingual children who are learning another language. This advantage can occur both for cognitive functioning as well as for learning a new language. In addition, Cummins (1979) formulated the Developmental Interdependence hypothesis, which states that the interaction between a child’s first and second language affects the child’s proficiency in those languages. If a child’s first language is well-developed, this will result in good acquisition of the second language. If, however, a child’s first language is not
well-developed but the child is exposed to a second language, the development of the first language is likely to stagnate. This will negatively affect the acquisition of the second language, which can leave a child with two poorly developed languages (Cummins, 1979).
Later studies have investigated bilingual children in their language acquisition, and these studies rendered mixed findings. Some studies report advantages for bilingual children
(Cárdenas-Hagan, Carlson, & Pollard-Durodola, 2007; Cenoz, 2003; Proctor, Carlo, August, & Snow, 2005), whereas others illustrate the threshold and/or developmental interdependence hypotheses (Cummins, 1979; Sagasta Errasti, 2003). It is still unclear how native Dutch and non-native students compare to each other when it comes to learning English as a Foreign Language (EFL). Research has found that speaking a minority language in various social contexts can foster third language acquisition (Sagasta Errasti, 2003), which could mean that bilingual children in the Netherlands will learn English more easily than native Dutch children. The current study looks at various aspects of English and compares the native group to the non-native group to uncover any significant differences.
The present study measured the children’s levels of vocabulary, fluency, and spelling in both Dutch and English. Vocabulary was included because of the significant differences found between native and non-native children. Little is known about their performance on English vocabulary tests. Vocabulary is one of the most important constituents of language, since without it, reading comprehension would not be possible, which in turn means that the language cannot be used for learning new things (Proctor, Silverman, Harring, & Montecillo, 2012; Royer & Carlo, 1991). Spelling was included because large differences were found between native and non-native children in favor of the native group (Verhoeven, 2000). Knowing the correct way to spell words in a language is related to the language’s orthography, which could significantly affect a person’s spelling skills (Figueredo, 2006). Many children experience difficulties with Dutch spelling (Trapman et al., 2014), but it is still unclear whether similar difficulties occur in English spelling across ethnic groups. Finally, fluency was included because automatized reading is very important for a child’s development since automatized
reading enables a person to focus more on the contents of language than on decoding that language (Crosson & Lesaux, 2010; Cutting & Scarborough, 2006). This could affect a student’s academic career, since reading comprehension is necessary for learning new information (Proctor et al., 2012; Royer & Carlo, 1991). Studies have shown that reading fluency can be transferred from a child’s first language to its second language (Crosson & Lesaux, 2010), but other studies have reported significant differences between native and non-native children for fluency (van Gelderen et al., 2003; Verhoeven, 2000). Non-native children seem to perform worse than their native peers, both on Dutch and English decoding tests used to assess fluency.
This study was part of a larger research project, called ORWELL (ORal and Written English Language Learning). Data used in this study were collected as part of the ORWELL-project. The aim of the present study is to gain insight into how non-native students compare to their native peers in their development of Dutch and English language skills. It was
hypothesized that the Dutch children would perform significantly better on Dutch vocabulary, spelling, and fluency measures. For the English vocabulary, spelling, and fluency tests, however, it was hypothesized that the non-native group would perform equally well or even better than the native group. Furthermore, experience with English was believed to affect the outcomes on the English language test, regardless of whether the students were native or non-native.
Method Participants
For this study, primary school children attending 4th grade (groep 6) were recruited from various primary schools in the Netherlands. Four schools participated, which rendered a sample size of 152 students. Two exclusion criteria were used. The first is whether a student is dyslexic, since mixed findings have been reported regarding their performance on native and foreign language tests. Dyslexic students sometimes perform worse than normal readers on native and foreign language tests (Morfidi, Van Der Leij, De Jong, Scheltinga, & Bekebrede, 2007), but have also been found to score better on foreign language tests if they prefer that language to their native language (Miller-Guron & Lundberg, 2000). Since dyslexic students’ performance seems difficult to predict and seem to differ from normal readers, they were excluded from the sample in the present study. The other exclusion criterion is the age at which students had moved to the Netherlands. They had to have been 6 years old or younger. Children usually start primary school at the age of 4, but formal Dutch language education starts two years later. Most children are 6 years old by then, hence the exclusion criterion. It was expected that scores on Dutch language measures would only be comparable if all children had had the same amount of formal Dutch language education.
A total number of 146 students remained, of which 78 were boys and 68 were girls. Students were around ten years old (M = 9.8 years, SD = 0.46). All children had normal or corrected to normal vision and hearing. Permission for participation was asked through active informed consent, meaning that parents consented to have their child included in the study based on the provided information. Students who completed all tests were given a small reward.
The schools that participated accommodated children from various ethnic backgrounds. Two of the schools were located in the outskirts of the capital, another in a neighboring city, and the fourth school was located more in the north of the province North-Holland. The locations of the schools affected the composition of the sample with regards to ethnicity. Two of the
participating schools were known to teach English before grade 4 (N = 86), whereas the other two schools (N = 60) were known to start teaching English from grade 5 onwards, which is mandatory for all schools in the Netherlands (Thijs et al., 2011). Students from 4th grade were selected for this study in order to gain insight into these children’s performance on English language tests before they had had formal English language education. Selection of these four schools allowed for a comparison between children who had not had English language education and children who had.
The participants were grouped based on the language they reported they were being raised in (De Houwer, 2007; Trapman, 2015). This resulted in a Native (N = 96) and Non-Native (N = 50) group. The latter group was further split up into three groups. One group of children reported speaking English at home (N = 13), amongst whom were several children from Ghana. There was a group who reported speaking Arabic at home (N = 12). Since this is one of the larger minority groups in the Netherlands, it could be interesting to look closer at this group’s performance. Finally, a group was composed of the children who reported speaking
miscellaneous languages, which is why they were categorized in the Other group (N = 25). Operationalization
Schools were recruited if they had two or more classes in 4th grade. If schools agreed to participate, letters of informed consent were sent to the parents, who then had to return a form stating whether their child could participate in the study. Once all consent forms were returned, data collection started. Teachers allowed for the normal curriculum to be interrupted for the duration of the testing session. The tests necessary for this study were completed on various days. The students never worked longer than one hour per session, and never did more than two group sessions or one individual session per day. Sessions consisted of various tests, which were randomized within a session. The order in which the sessions were completed was counterbalanced for the participating schools. Group test sessions were completed by each student on an individual tablet with an external keyboard in the tablet case. Children were given
headphones, which allowed them to work at their own pace. They could decide when they were ready to listen to the next item as provided on their tablet. Before the students started, they had been informed about what kind of tests they would be doing and were asked to take their time so that they would perform the best they could. Tests that required a student to speak were administered in a one-on-one setting in a separate room within the school with a trained test administrator, so that it could be recorded.
Materials
Vocabulary.
The student’s receptive vocabulary knowledge in Dutch and English was measured using the Peabody Picture Vocabulary Test-III-NL (PPVT-III-NL; Dunn, Dunn, & Schlichting, 2005) and Peabody Picture Vocabulary Test-4-EN (PPVT-4-EN; Dunn & Dunn, 2007),
respectively. These instruments are developed to evaluate comprehension of spoken language and are thus believed to measure someone’s achievement in acquiring vocabulary (Dunn & Dunn, 2007). Both tests are to be administered individually and take about 15-20 minutes to complete. The PPVT-III-NL is suitable for children and adults aged between 2 years and 3 months old to 90 years old (Dunn et al., 2005). The test consists of a total number of 204 items divided over 17 sets. Based on the students’ age, five sets of items were selected: sets 7 through 11. The students completed a total number of 60 items for the PPVT-III-NL.
The test is usually administered in a one-on-one setting, in which the student is shown four pictures in a booklet while the test administrator reads out a word. The student must then point to the corresponding picture. For this study, the test has been digitalized, meaning that the students were all given a tablet with headphones and each item was presented on a separate page with a recording of the target word. On the screen, the image from the booklet was shown. Students had to click on one of the four options to indicate which picture corresponded to the word they had just heard. Each correct answer was awarded one point, which amounted to a maximum score of 60. The higher the score, the larger the student’s Dutch vocabulary was believed to be.
The PPVT-4-EN is similar in administration and purposes as the PPVT-III-NL. The English version, however, is suitable for children and adults aged between 2 years, 6 months old and 90 years old (Dunn & Dunn, 2007). The test consists of 228 items in total, divided over 19 sets. Usually a critical range consisting of several sets of items based on the participant’s abilities is established before the test officially starts, but for the purposes of this study, a fixed range of items was selected. The participating students were not believed to have a large vocabulary in English yet, which is why the first seven sets were included in this study. These sets are often used for the screening of English language learners (Dunn & Dunn, 2007). This means that the
students were given 84 items of the PPVT-4-EN. Administration of this test was identical to the PPVT-III-NL described above. For this test, too, one point was awarded for a correct answer, making the maximum score 84. The higher the score, the larger the student’s English vocabulary was believed to be.
Spelling.
The students’ Dutch spelling skills were measured using the Schoolvaardigheidstoets
Spelling (SVT; Braams & De Vos, 2015). This test was designed to measure children’s spelling
skills in Dutch primary education, which has a strong focus on knowing how to apply the rules for correct spelling and grammar (Braams & De Vos, 2015). The test consists of a noun test and a verb test, of which only the noun test was used. Based on the grade the students are in, specific sets were chosen to administer. For this study, sets 4, 5, and 6, all consisting of 15 items, were selected, meaning that the children had to complete 45 items. The standardized administration is one of dictation, in which an entire class listens to the teacher, who reads out a sentence and repeats one word of that sentence that the students must then write down. Like the other tests, the SVT was digitalized. This means that students were all given a tablet with a keyboard and headphones. On their screen, they saw one item at a time. A recording of the target sentence was provided and a space in which the student could type their answer was provided. An example is that the student hears “Het tegenovergestelde van maxi is mini. Schrijf op ‘maxi’” (The opposite of maxi is mini. Write down ‘maxi’). This digitalized administration has been kept as standardized as possible, so that the results would still be valid. Each correct answer was awarded one point, meaning that a student could obtain 45 points. The higher the score, the better a student is at spelling in Dutch.
To assess the students’ English spelling skills, a subtest of the Wide Range Achievement Test (WRAT; Wilkinson & Robertson, 2006) was administered. This large test is designed to measure basic academic skills, such as reading, spelling, and math, and thus consists of several subtests. Of these subtests, only Spelling was used in this study. This test is similar in
administration as the Dutch SVT described above. A test administrator or teacher dictates a word, followed by a sentence carrying that word, after which the target word is repeated once more. Students are thus expected to write down the word that was repeated several times. For example, the student hears “On. It is on the table. On.”, and is expected to write down ‘on’. Similar to the Dutch SVT, students in this study heard the sentences not from their teacher but through a recording that was available to each individual student on their tablet. On the screen, a box was provided in which the students could type their answer. The test consisted of 20 items, including one example, and took about 20 minutes to complete. Each correct answer was
awarded one point, with a maximum of 19. The higher the score, the better the student’s English spelling skills.
Reading fluency.
Dutch reading fluency was measured using the Een-Minuut-Test (EMT; Brus & Voeten, 1979). This test was designed to measure a student’s technical reading skills. Through quick word decoding, a student’s fluency in that language can be determined. An equivalent of the EMT, the One-Minute-Test (OMT; Van Berkel & Philipsen, 2012) was used to assess the students’ English reading fluency. The EMT consists of 116 words presented to the student in four rows on a sheet of paper. The student was asked to read aloud correctly as many words as quickly as possible in one minute. The English equivalent OMT uses the same technique but consists of 114 English words. Every correctly read word within one minute is awarded one point, with a maximum of 116 and 114, respectively. The higher the score, the better the student’s reading fluency is in that language. Administration of these tests was in a one-on-one setting so that the student’s performance could be recorded with a voice recorder. The test administrator took the student to a separate room and had the student read out the EMT and OMT. They were administered on different days. The two tests combined took about 10 minutes to complete per student.
Nonverbal intelligence.
To control for any intelligence influences that may affect the students’ scores on the language tests, a nonverbal intelligence test was administered. This was the Raven Standard Progressive Matrices (SPM; Raven, Court, & Raven, 1996). This test is designed to assess a person’s nonverbal reasoning, no matter their age, education, or nationality. Usually, this test is untimed, meaning that a participant can take as long as they like to complete the test. For this study, a time limit was set due to the combined administration with other tests. Research has shown that timed versions of the Raven Advanced Progressive Matrices can be used as a valid measure of someone’s nonverbal intelligence (Hamel & Schmittmann, 2006). These researchers concluded that the participant’s score after 20 minutes successfully predicted the score of an untimed assessment, if the participant was presented with the usual number of test items. Following that logic, the full version of the Raven SPM was used, but administration was limited to 30 minutes. Students were asked to work at their own pace. After 30 minutes, the students’ progress and answers were saved. Two of the items from Set A were used as examples, meaning that the students had a total number of 58 items to complete. The number of correct answers after 30 minutes was used as an indication of the students’ general nonverbal mental ability. Administration of this test was slightly different than the standardized method, since it was
digitalized like the other tests described above. Students were presented with one item on the screen of their tablet. They had to indicate their answer by selecting the right box by clicking on the piece they believed completed the picture.
Informative questionnaires.
A questionnaire about the students’ ethnic background was included in this study. Each student was interviewed individually about their ethnic background. The questions covered a wide range of possible situations in which students use a language. The questions were about the student’s and his/her parents’ countries of birth, the language they spoke at home, the language they were being raised in, situations in which they might use their home language and the degree to which these children believed themselves to be able to speak their home language. Similar questions were asked about Dutch. Taken together, these questions were believed to render information about the children’s ability to speak Dutch as well as their home language.
Since motivation to learn English can be of major importance to a student’s English language ability (Spinath et al., 2006; Steinmayr & Spinath, 2009), an adaptation of the Attitude and Motivation Test Battery (AMTB; Gardner, 2005) and English Learner’s Questionnaire (ELQ; Dornyei, 2010) was included in this study. The AMTB is an English test designed to measure students’ motivation to learn a foreign language. The ELQ was developed based on the AMTB, but is considerably shorter. For this study, items were selected from both questionnaires. The translated items were divided over seven scales of five items each. This resulted in a
questionnaire consisting of 35 items. Different from the AMTB and ELQ, this questionnaire was rated on a 4-point scale, as opposed to the 6-point scale used in the other two questionnaires. Students had to indicate whether they ‘strongly disagree’, ‘slightly disagree’, ‘slightly agree’, or ‘strongly agree’ with each of the 35 items. Examples are ‘Ik wil graag zo veel mogelijk Engels
leren’ (I plan to learn as much English as possible), or ‘Ik wil graag vrienden die Engels spreken’
(I wish I could have many native speaking English friends). The total score for the seven scales could be 140, with higher scores representing higher levels of motivation to learn English.
Apart from knowing to what extent a student is motivated to learn English, it is also important to know how much experience a student has had with English, since it can have a large effect on their proficiency (Lindgren & Muñoz, 2013). For this reason, a questionnaire was designed to ask the student about their experience with and exposure to English. 16 questions were formulated about the children’s experience with English outside of school. They were rated on a 6-point scale going from ‘Never’ to ‘Always’. The questions covered a wide range of possible situations in which students might use or be exposed to English, such as online chatting, use of English on social media, English books and magazines, and English music and
films. All items taken together make up the total score, adding up to 96 points possible. The higher a score, the more experience a student has had with English.
Data Analysis1
Before data analysis, the students had to be divided over ethnic groups. The division between native and non-native was made based on the language a student reported to be raised in. If the language they are raised in2 was not Dutch, then these students were put in the Non-Native group. All students who reported being raised in Dutch were put in the Non-Native group. The reason this grouping technique was used, was that research has shown that children usually follow their parents in speaking their native language, as opposed to the language of the country, i.e. Dutch (De Houwer, 2007; Trapman, 2015). Apart from their home language, students were also grouped according to their reported Experience with English. Information about a student’s experience with English was rated on a continuous scale. To create groups, the median split technique was used. The median was computed and used as the cut-off point for the two groups. The median was included in the group with high scores. This resulted in a grouped variable with Little and Much Experience with English.
Before any analyses were done, the data had been checked for normality. Outliers were corrected for, so that the data was as normally distributed as possible. The inclusion of any covariates in the main analyses depended on the outcomes of the correlation analyses. These were performed to see which variables correlated enough (at least .20 (Sullivan & Feinn, 2012)) with the various Dutch and English language tests to be included. For each potential covariate, subsequent t-tests were done to check if the observed scores differed significantly across groups. The significant covarying variables with values of .20 or larger were included in the main analyses. Repeated measures were performed with two between-subjects factors. The first was the grouping variable ‘Home Language’, which consisted of the Native and Non-Native group. The second between-subjects factor was the grouped variable that was computed for Experience with English (Little and Much). The analysis had two within-subjects factors. Test-Language was one, consisting of two levels, Dutch and English. The other within-subjects factor was Domain, which was made up of three levels: vocabulary, spelling and fluency.
1 All statistical analyses were done using IBM SPSS Statistics for Windows, version 24.0. 2 ‘Home language’ will be used to indicate the language a student is being raised in.
Results
The participating students were grouped according to their ethnic background, based on the language they reported being raised in. The total of 146 participating students was split over two groups: a Native (N = 96) and Non-Native group (N = 50). Based on their home languages, they were further divided over specific language groups: English (N = 13), Arabic (N = 12), and Other (N = 25). There were 78 boys and 68 girls, who were all around the age of ten (M= 9.8, SD = 0.46). The students were
also split up into two groups based on their reported experience with English. The median of this variable was
computed and used as the cut-off point to form two groups. The median was included in the group reporting much experience. The group with little experience was somewhat smaller than the group with much experience (Little:
N = 67; Much: N = 79). For an overview, see Table 1.
First, correlation analyses were conducted to discover if there were significant
covariates that should be included in the main analyses. Then, a repeated measures ANOVA was done. This had two between-subjects factors: the grouping variable Native or Non-Native3, and the grouped variable with Little or Much experience with English. There were two within-subjects factors: Test-language, consisting of two levels (Dutch and English), and Domain, consisting of three levels (vocabulary, spelling, and fluency). Significant covariates were added in the repeated measures analysis. Exploratory analyses were done using the grouping variable consisting of four groups (Dutch, English, Arabic, and Other).
Correlation Analyses
Before the main analyses could be done, potential significant correlations needed to be tracked down. Since the purpose of the correlation analysis was to identify possible covariates, only those correlations will be described. Some of the outcome measures were not normally distributed, which is why the correlations are reported using bias corrected and accelerated bootstrap (BCa) 95% confidence intervals (CIs; in the square brackets). Since no specific
3 Throughout the following section, the grouping variable will also be referred to as ‘Home language’.
Table 1
Gender, mean age, and Experience with English for Native vs. Non-Native and total
Gender (boys:girls) Mean Age (SD) Experience with English (Little:Much) Native 56:40 9.9 (0.48) 50:46 Non-Native 22:28 9.7 (0.42) 17:33 Total 78:68 9.8 (0.46) 67:79
expectations were formulated for the correlations, the tests were executed using two-tailed correlations.
The first potential covariate was the measure for intelligence. Significant correlations were found between this instrument, the vocabulary tests, and the spelling tests. Intelligence was significantly correlated with the Dutch vocabulary test, r = .45, p < .01 [.312, .580], the English vocabulary test, r = .27, p < .01 [.146, .384], and the Dutch spelling test, r = .21, p = .016 [.027, .373]. The English spelling test correlated borderline significant with the intelligence test:
r = .16, p = .062 [-.007, .326]. This correlation has BCa CIs that cross zero, which makes it likely
that there was no relation between these two measures (Field, 2013). Intelligence did not correlate significantly with the grouping variable Home Language or the grouped variable for experience with English (ps > .05). A subsequent t-test indicated that the difference between the Native and Non-Native group was non-significant: t(144) = 1.55, p = .124. Research has shown that the level of intelligence between monolinguals and bilinguals was not significantly different and that intelligence does not affect the outcomes of foreign language tests (Ganschow, Sparks, Javorsky, Pohlman, & Bishop-Marbury, 1991). In this sample, intelligence does seem to affect the outcomes of several of the language tests. To control for individual variation within the groups, intelligence is included as a covariate in these analyses (based on Mol & Bus (2011)).
The second potential covariate was Motivation to learn English. Significant correlations were found for the Dutch vocabulary test, r = -.18, p = .035, [-.352, -.002], Home Language, r = .22, p = .012, [.036, .387], and Experience with English, r = .29, p < .01, [.142, .434]. This indicates that there is a small relation between a student’s home language and their motivation to learn English. In this case, Non-Native students seem slightly more motivated to learn English than their Native peers. A t-test reveals that the difference between the two groups is
statistically significant: t(142) = -2.90, p < .01, in favor of the Non-Native group (M = 140.69, SD = 16.65 versus M = 133.08, SD = 13.93 for the Native group). This contradicts earlier research (Escamilla, 2009). For the correlation between Motivation and Experience with English, it seems that the more motivated a student is, the more experience they have. A t-test revealed that there was a significant difference between the Much and Little group: t(142) = -3.49, p < .01. The Much-group (M = 139.56, SD = 13.45) was more motivated than the Little-group (M = 130.95, SD = 16.15). It seems, then, that motivation to learn English is at least in part expressed in a
student’s experience with English. Since Motivation to learn English did not correlate
significantly with any of the English outcome measures, this variable was not included in the main analyses.
To check if the outcomes of the English language tests are affected by the school a student attends, the four schools were included in the correlation analysis. Significant
correlations were found with Home Language, r = -.46, p < .01, [-.619, -.276], the subdivision of Home Language (four groups), r = -.43, p < .01, [-.584, -.266], Intelligence, r = .37, p < .01, [.224, .503], Dutch vocabulary, r = .50, p < .01, [.354, .613], and for Experience with English, r = -.19, p = .025, [-.358, -.023]. The negative correlations for the Home Language variables can be attributed to the fact that most of the participating students are native Dutch children. The Dutch children are coded with the lowest value, hence the negative correlation. For Experience with English, it seems that formal language education did not affect the student’s exposure to English. This might in turn be due to the fact that the populations of the schools that offer English from 5th grade onwards are mostly ethnically mixed. These children may have more experience with English in their day-to-day life, even though they do not receive formal English language education. Subsequent t-tests for the significant correlations with Intelligence and Dutch vocabulary reveal that the scores are significantly better for the group of children who attend schools that offer English before the 5th grade. These surprising results could be due to the characteristics of the four schools and the differences between them. The correlation between Intelligence and the schools supports the decision to include Intelligence as a
covariate. By doing so, any individual differences with regards to intelligence can be eliminated. For Dutch vocabulary, the significant difference could be the result of the generally more mixed population in the schools that start English education in 5th grade. These schools were coded with the lowest value and coincidentally accommodated an ethnically mixed population. As a whole, however, the characteristics of the schools did not affect the English outcome measures, which is why it will not be included in further analyses.
Main Analyses
After establishing that intelligence covaried with most of the outcome measures, the main repeated measures analysis was done. All the language tests were scored on a different scale, which resulted in different maximum scores. To compare the scores of all the tests, the outcome variables were standardized into z-scores. The mean scores for all the variables are reported in Table 2. Below the results are given for the repeated measures ANOVAs that were done.
Table 2
Mean scores of outcome variables for Native vs. Non-Native groups and total
Group DU Vocab EN Vocab DU Spelling EN Spelling DU Fluency EN Fluency Native 39.59 (6.02) 56.98 (7.92) 15.60 (5.59) 4.73 (2.61) 66.28 (12.90) 41.49 (15.92) Non-Native 34.06 (5.43) 56.45 (10.35) 16.48 (6.06) 5.33 (4.49) 68.20 (14.12) 43.71 (22.24) Total 37.70 (6.37) 56.81 (8.73) 15.90 (5.75) 4.92 (3.34) 66.93 (13.31) 42.26 (18.30)
Note. Standard deviations are provided in parentheses. Only the scores for DU Vocab differed significantly across groups (p < .01). DU Vocab was measured using the PPVT-NL, EN Vocab using the PPVT-EN, DU Spelling using the SVT, EN Spelling using the WRAT, DU Fluency using the EMT, and EN Fluency using the OMT.
The assumption for sphericity was met for Domain (p = .073), but not for the interaction between Test-Language and Domain, χ²(2) = 9.30, p = .009. The degrees of freedom were corrected using the Greenhouse-Geisser estimates of sphericity (ε = .934). Significant main effects were found for Intelligence (F(1, 129) = 14.82, p < .01, r = .32), and for Experience with English (F(1, 129) = 8.35, p < .01, r = .25). It thus seems that the higher the intelligence or the more experience, the better the performance on the tests. No significant effects were found for Home Language or Home Language combined with Experience with English (ps > .05).
A main effect was found for Domain, F(2, 258) = 6.72, p < .01, r = .16. Further contrasts revealed that the outcomes for vocabulary and spelling differed significantly, F(1, 129) = 4.20, p = .042, r = .18 in favor of spelling (standardized mean score for vocabulary -.094 +/- .065 and for spelling -.034 +/- .076). The difference between vocabulary and fluency was also significant,
F(1, 129) = 11.40, p < .01, r = .29 (standardized mean score for vocabulary -.094 +/- .065 and for
fluency -.020 +/- .079). The difference between spelling and fluency was not significant. The standardized scores reveal that the vocabulary tests rendered the lowest outcomes compared to the other two domains. Overall, the fluency tests rendered the highest standardized results.
A significant interaction effect was found between Domain and Intelligence, F(2, 258) = 6.14, p < .01, r = .15. The contrasts reveal a similar trend as with the main effect. There was a borderline significant difference between vocabulary and spelling, F(1, 129) = 3.65, p = .58, r = .17 (standardized mean score for vocabulary -.094 +/- .065 and for spelling -.034 +/- .076), and a significant difference between vocabulary and fluency, F(1,129) = 10.44, p < .01, r = .27 (standardized mean score for vocabulary -.094 +/- .065 and for fluency -.020 +/- .079). No significant difference was found between spelling and fluency. It seems that intelligence affects the scores on the outcome measures for specific language aspects. As was found in the initial correlation analyses, the correlation is strongest between intelligence and the two vocabulary measures. This suggests that intelligence contributes to a student’s vocabulary knowledge in a particular language. In this study, it seems that a higher score on the intelligence measure leads to a higher score on the vocabulary measures. Fluency is an automatized process that develops quickly in normal readers. This might be why intelligence does not seem to affect the outcomes on the fluency measures, as was found in the correlation analysis, and why the difference with vocabulary is significant (Aarnoutse, van Leewe, Voeten, & Oud, 2001).
Another significant interaction was found between Domain and Home Language,
F(2,258) = 5.12, p < .01, r = .14. Here, too, the contrasts indicate that the significant differences
vocabulary and fluency, F(1, 129) = 6.06, p = .015, r = .21. The difference between spelling and fluency was not significant. The Native group performs best on both Dutch and English
vocabulary measures compared to the Non-Native group (standardized mean score for the Native group on vocabulary tests .085 +/- .071 and for the Non-Native group -.273 +/- .110). For spelling and fluency, however, the Non-Native group outperforms the Native group in both Dutch and English (standardized mean score for the Native group on spelling tests -.079 .083 and for the Non-Native group .011 +/- .129; standardized mean score for the Native group on fluency tests -.049 +/- .086 and for the Non-Native group .009 +/- .133). These findings suggest that Home Language affects the outcomes of the tests in a specific language, as well.
The third significant interaction effect was found between Test-Language, Domain, and Home Language, F(1.87, 241.07) = 5.52, p < .01, r = .15. This interaction reveals that all the interaction effects described above are moderated by Test-Language. Much like in the earlier interaction effects with Domain, significant contrasts were found between vocabulary and spelling, F(1, 129) = 5.79, p = .018, r = .21, and between vocabulary and fluency, F(1, 129) = 9.99,
p < .01, r = .27, but not between spelling and fluency. Looking at the mean scores per group and
for each domain (see Table 2), it becomes clear that the largest difference is found on the Dutch vocabulary test. The Native group performs significantly better on this test than the Non-Native group. Error! Reference source not found. illustrates the results found in this interaction.
Finally, a significant interaction effect was found between Test-Language and
Experience with English, F(1, 129) = 13.42, p < .01, r = .31. The main effect for Experience with English reported earlier suggests that there is a significant difference between the Little and Much group. Mean scores for these groups (see Table 3) show that the significant differences between these groups are found on the English outcome measures (English vocabulary group M = 54.30, SD = 7.56 versus Much-group M = 58.89, SD = 9.12; English spelling group M = 4.06, SD = 2.53 versus Much-group M = 5.66, SD = 3.76; and English fluency Little-group M = 34.21, SD = 14.74 versus Much-Little-group M = 49.25, SD = 18.30). Experience with English does not seem to affect the outcomes on the Dutch tests, but the results on the English tests are significantly higher for the group reporting much experience with English. These results indicate that for the performance on English language tests, it matters greatly how much experience a student has with English.
Overall, it can be concluded that intelligence affects the outcomes of tests that focus on a particular domain of language. Nonverbal intelligence seems to affect a student’s vocabulary knowledge in a specific language. Furthermore, the outcomes of specific domains are affected by a student’s home language. Especially for vocabulary, it appears to matter which language a student hears at home. Combined with one another, it seems that the outcomes depend on the language of the test, the aspect of language that is being tested and which language a student is exposed to at home. Moreover, Experience with English significantly affects the students’ performance on the various English tests, indicating that this factor is important for a student’s performance. Compared to the Native group, a larger proportion of the Non-Native group have reported higher scores for Experience with English (48% vs. 66%, respectively). Combined with the fact that the Non-Native group consists of several English-speaking students, it is interesting to investigate if similar results are found when the Non-Native group is divided further. An exploratory analysis will be done to see if this changes the results.
Table 3
Mean scores of outcome variables for Little vs. Much Experience with English
Group DU Vocab EN Vocab DU Spelling EN Spelling DU Fluency EN Fluency Little 38.10 (6.17) 54.30 (7.56) 15.90 (5.03) 4.06 (2.53) 65.15 (12.51) 34.21 (14.74) Much 37.35 (6.56) 58.89 (9.12) 15.91 (6.33) 5.66 (3.76) 68.46 (13.86) 49.25 (18.30)
Note. Standard deviations are provided in parentheses. Only the scores on the English measures differed significantly across groups (p < .01). DU Vocab was measured using the PPVT-NL, EN Vocab using the PPVT-EN, DU Spelling using the SVT, EN Spelling using the WRAT, DU Fluency using the EMT, and EN Fluency using the OMT.
Exploratory Analysis
Seeing as the previous analysis revealed significant differences between the Native and the Non-Native group, it is interesting to explore which effects are found for the subgroups within the sample. As was described above, a division between Native and Non-Native was made, and within the Non-Native group, students were grouped according to the language they reported being raised in. Home Language thus consisted of four groups: the Dutch group (N = 96), an English-speaking group (N = 13), an Arabic-speaking group (N = 12), and a group consisting of miscellaneous languages, called Other (N = 25). Mean scores per group are reported in Table 4. The same repeated measures analysis as described above was conducted.
Mauchly’s test indicated that the assumption of sphericity had been violated for the interaction effect of Test-Language and Domain, χ²(2) = 9.01, p = .011. Therefore, the degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = .94). The
assumption for sphericity was met for Domain (p = .083). A significant main effect was found for Intelligence, F(1, 125) = 14.42, p < .01, r = .32. The interaction between Home Language and Experience with English was borderline significant, F(3, 125) = 2.27, p = .083, r = .13.
A significant main effect was found for Domain, F(2, 250) = 7.90, p < .01, r = .18.
Contrasts revealed significant differences between vocabulary and spelling, F(1, 125) = 5.62, p = .19, r = .21 (standardized mean score for vocabulary -.055 .109 and for spelling .166 .129), and between vocabulary and fluency, F(1, 125) = 13.27, p < .01, r = .31 (standardized mean score for vocabulary -.055 +/- .109 and for fluency .130 +/- .134). The difference between spelling and fluency was not significant. This time, the results of the spelling tests were highest. Vocabulary scores were still lowest of the three domains. This result is similar to the main analysis above.
A significant interaction effect was found between Domain and Intelligence, F(2, 250) = 6.56, p < .01, r = .16. Contrasts revealed that this difference was significant between vocabulary and fluency, F(1, 125) = 11.15, p < .01, r = .29 (standardized mean score for vocabulary -.055 .109 and for fluency .130 +/- .134). There were borderline significant differences between
Table 4
Mean scores of outcome variables for groups
Group DU Vocab EN Vocab DU Spelling EN Spelling DU Fluency EN Fluency Dutch 39.59 (6.02) 56.98 (7.92) 15.60 (5.59) 4.73 (2.61) 66.28 (12.90) 41.49 (15.92) English 35.69 (6.03) 61.08 (9.88) 15.00 (8.23) 7.08 (4.81) 70.15 (20.58) 56.15 (20.88) Arabic 33.83 (5.25) 53.80 (7.50) 18.58 (6.36) 3.18 (3.66) 69.75 (13.13) 37.08 (20.06) Other 33.32 (5.22) 55.14 (11.26) 16.24 (4.36) 5.43 (4.43) 66.38 (10.27) 40.29 (22.13)
Note. Standard deviations are provided in parentheses. The scores for DU Vocab differed significantly across groups in favor of the Dutch group (p < .01). For the English measures, the English group performed significantly best compared to the other three groups. DU Vocab was measured using the PPVT-NL, EN Vocab using the PPVT-EN, DU Spelling using the SVT, EN Spelling using the WRAT, DU Fluency using the EMT, and EN Fluency using the OMT.
vocabulary and spelling, F(1, 125) = 3.48, p = .064, r = .17 (standardized mean score for
vocabulary -.055 +/- .109 and for spelling .166 +/- .129), and between spelling and fluency, F(1, 125) = 3.74, p = .055, r = .17 (standardized mean score for spelling .166 +/- .129 and for
fluency .130 +/- .134). This, too, is largely in line with the results from the main analysis.
Surprisingly, an interaction was found between Test-Language and Home Language, F(3, 125) = 3.35, p = .021, r = .16. This interaction was not significant in the main analysis, but the subgroups within Home Language seem to affect the outcomes of the various outcome
measures. As the means per group illustrate (see Table 4), the English group scores higher on all the English tests compared to the other groups. It thus seems that the English group benefits from their home language when doing English language tests, much like Dutch students seem to benefit from their home language when doing a Dutch vocabulary test.
Figure 2 Standardized mean scores for groups per variable
Another interaction was found between Test-Language, Domain, and Home Language,
F(5.61, 233.64) = 2.36, p = .034, r = .10. There were significant differences between vocabulary
3.48, p = .018, r = .17. The difference between spelling and fluency was not significant. Mean scores for Dutch vocabulary differed significantly between Dutch and Arabic students, and between Dutch and Other students, both in favor of the Dutch group. The difference between the Dutch and English group was not significant for Dutch vocabulary. For English vocabulary, there was a significant difference between the Dutch and English students, in favor of the latter group. The English group also outperformed the Dutch, Arabic, and Other group on English spelling and English fluency tests.
A final interaction effect was found borderline significant, but considering the outcomes of the main analysis, it is important to report. The interaction between Domain, Experience with English, and Home Language, F(6, 250) = 2.03, p = .062, r = .09, is the only interaction in which Experience with English seems to affect the outcomes on the tests. In the main analysis,
Experience with English was a significant main effect, but this is not the case in this exploratory analysis. The interaction between Domain, Experience with English, and Home language
suggests that home language affects the outcomes of the various tests and that it affects the amount of experience with English a child has. Arguably, the interaction effect between Test-Language and Home Test-Language reported earlier has taken away the effect of experience with English. Table 5 illustrates that all the children in the English group reported having much experience with English, whereas the other groups are roughly equally divided over the Little and Much group. Students who report having much experience with English do better on the various tests (standardized mean score for Little Experience Dutch group -.092 +/- .087 and for Much Experience Dutch group .064 +/- .091; Little Experience Arabic group -.546 +/- .268 versus Much Experience Arabic group .254 .270; Little Experience Other group -.391 .213 versus Much Experience Other group .140 +/- .167; information about the English group is not representative due to the single student in the Little Experience group). The groups perform best on the spelling tests compared to vocabulary (standardized mean score for spelling .166 +/- .129 and for vocabulary -.055 +/- .109), and compared to fluency (standardized mean score for spelling .166 +/- .129 and for fluency .130 +/- .134). The English group outperforms the other three groups on all English measures.
Table 5
Degree of Experience with English per group.
Little Much Total
Dutch 50 46 96
English 1a 12 13
Arabic 6 6 12
Other 10 15 25
Note. a This student had a score of 33 on the Experience scale, just below the
As was to be expected, the English children performed better on all the English language tests compared to the other participating students. This can be explained by the fact that these children will have had more exposure to the language. The Dutch students have performed best on the Dutch vocabulary test, which supports the idea that exposure to a particular language affects a student’s performance on tests in that same language. The other Dutch language tests were not significantly better for the Dutch group compared to the other three groups. The fact that the Dutch students seem to have performed worst on all outcome measures except the Dutch vocabulary test is interesting. This difference could be attributable to the small number of students in the three subgroups within the Non-Native. The Dutch group is by far the largest and can thus be seen as a fairly representative outcome for all Dutch students. The higher scores from the Non-Native groups could be the result of a small sample of high performing students.
Discussion Main Findings and Implications
The aim of this study was to gain insight into how non-native primary school students compare to their native peers in their development of several Dutch and English language skills. It was hypothesized that the native children would perform best on all Dutch language tests. For the English tests, it was hypothesized that the non-native group would perform equally well or better than the native group. Finally, how much experience a student has had with English was hypothesized to affect their scores on the English language tests, regardless of their language background. Taken together, these hypotheses were expected to provide insight into what the effect of language background and experience with English is on native and non-native primary school students’ development of Dutch and English language skills.
For Dutch language skills, the only test that the native group did better on than the non-native group was the vocabulary test. Native children had a larger command of Dutch than the non-native children. The exploratory analysis revealed that the Dutch group outperformed the Arabic group and the children in the Other group on the vocabulary test. The difference with the English group was not substantial enough to be significant, although the Dutch group did
perform better than the English group. The outcomes of the Dutch spelling and fluency tests did not differ significantly across groups. For the English language tests, no differences were found across the two groups for all three domains. Further exploratory analyses revealed that the English group did better on all the English language tests than the other three groups. Finally, differences between the group with little and much experience with English were found for all English language tests. Students who reported more experience with English outside of school performed better than students with little experience. In the exploratory analysis, this effect had
become smaller. It seemed that the previously found effect was largely due to the level of experience of the English students.
The findings regarding Dutch vocabulary are in line with previously reported findings from other studies. Non-native children have frequently been reported to have a smaller
vocabulary than native children (Calvo & Bialystok, 2014; Trapman et al., 2014; van Gelderen et al., 2003; Verhoeven, 2000). Although it was suggested that second-generation non-native families might have incorporated Dutch in their home language, this does not appear to be the case. Language is a major aspect of someone’s identity and many families in the Netherlands speak their native language at home to preserve their native identity. Often children are exposed to their native language through media as well (Peeters & D’Haenens, 2005). It seems, then, that non-native children are minimally exposed to Dutch in their home settings, which limits their exposure to Dutch to school and other social settings. This could explain the lower scores of the non-native group for Dutch vocabulary.
Interestingly enough, however, no differences were found between the native and non-native group for Dutch spelling and fluency. These results can be attributed to the smaller sample size of the non-native groups. These groups had some quite extreme scores, both on the high end as the low end, and due to the small sample, these data could have affected the mean scores. Notwithstanding, these findings contradict earlier studies reporting differences on these domains of language. Usually, the non-native group does worse on these tests than the Dutch children included in the study (van Gelderen et al., 2003; Verhoeven, 2000). This might, however, be due to the ages of the participating students. The students in Van Gelderen et al.’s study (2003) were between 13 and 14 years old, whereas Verhoeven’s study (2000) tested children who were younger than 7 years old. The students who participated in the current study were around 10 years old. In 4th grade, a lot of attention is paid to the correct way to spell words and fast, accurate reading. The children in the current study may thus benefit from the fact that they are still in the middle of formal Dutch language education. Students in the aforementioned studies might not have been exposed to formal education long enough (seven year olds have only had one year of formal Dutch language education), or were at an age at which accurate spelling and reading is expected to be fully developed (Bonset & Hoogeveen, 2009, 2012).
The English language tests did not show any significant differences between the native and non-native groups. This contradicts earlier studies that reported advantages for bilingual students acquiring vocabulary in a third language (Cenoz, 2003; Keshavarz & Astaneh, 2004; van Gelderen et al., 2003). The exploratory analysis revealed that there were significant
differences on the English vocabulary, spelling, and fluency tests between the English group and the three remaining groups. Compared to the English group, the Arabic and Other groups performed significantly worse. This suggests that these groups do not (yet) experience
advantages from their bilingualism. According to the threshold hypothesis, bilingual children must achieve a certain level of competence in their first language before they can benefit from their knowledge and learn a second language more easily (Cummins, 1979). It might be that bilingual children in the Netherlands are at the lower threshold, also known as semilingualism, which could negatively affect their cognitive abilities resulting in poorer academic achievements (Cummins, 1979). This has been found in several studies (Appel & Vermeer, 1998; van Gelderen et al., 2003; Verhallen & Schoonen, 1993). The situation in the Netherlands is such that bilingual children usually do not receive formal education in their native language but are immersed in Dutch language education from the age of six. More recent research has found that in order for bilingual children to benefit from their bilingualism, they must use their native language in more than one setting (Sagasta Errasti, 2003). It is likely that native language use of children in the Netherlands is limited to their family or home setting. This might be the reason for the lower scores of the Arabic and Other groups on the English language tests.
In this sample, Experience with English outside of school affected the outcomes of the English language tests in the main analysis. It seemed that the more experience a student had had with English, the better they scored on the English language tests. This effect had become much smaller in the exploratory analysis. The English group performed significantly best on all English tests and seemed to have affected the overall effect in the main analysis. Several other studies have reported that more exposure to a language, be it through print or other media, has a positive effect on a child’s proficiency in that language (Lindgren & Muñoz, 2013; Mol & Bus, 2011; Olson, Keenan, Byrne, & Samuelsson, 2014; Scheele et al., 2009; Sénéchal & LeFevre, 2014). The current study not only shows that more experience with English renders higher scores on the English language tests, but also shows that more experience with Dutch renders higher scores on the Dutch vocabulary test. The experience with English the students reported seemed unrelated to their experience with English in schools. It seems that students learn much from external exposure to English (i.e. outside of school) and that formal education is less effective. This finding is socially relevant and can be used in schools. As opposed to formal instruction, schools could expose their students to educational English media that allows the students to learn the language implicitly. This will give them a good basis on which they can expand their knowledge of English later in school, when formal instruction starts (e.g. in 5th grade). Parents should be encouraged to expose their children to as much stimulating English material as possible.
Intelligence was added in the analysis because an initial correlation analysis revealed that it correlated significantly with several outcome measures. There were, however, no differences in intelligence between the native or non-native group. Usually, studies report influences from verbal intelligence on foreign language learning (Sparks, Patton, Ganschow,
Humbach, & Javorsky, 2006), but the current study used a nonverbal measure of intelligence. As the correlation analysis also revealed, intelligence correlated significantly with the schools that participated in the sample. Students from schools that offer English from the first grade onward seem to have done better on the intelligence test and the Dutch vocabulary test. The difference in Dutch vocabulary can be explained by the population of these schools. The population of the schools that offered English from the first grade was predominantly native, whereas the other two schools were ethnically mixed. For intelligence, however, it is unclear how these differences came to be. It may be due to the relation between intelligence and Dutch vocabulary found in the first correlation analysis. As was mentioned above, the differences in Dutch vocabulary could be attributed to the schools’ populations. The correlation between Dutch vocabulary and intelligence suggests that the difference across schools can somehow be the result of the schools’ populations. Nothing conclusive can be said about this, however, and should be researched in future studies.
Motivation to learn English was not included in the main analysis because it did not correlate significantly with any of the English outcome measures. It did, however, correlate significantly with Experience with English, which suggests that a student’s experience with English makes them more motivated to learn that language. It could also be, however, that experience with English is the result of a student’s motivation to learn English. Motivation thus seems to contribute to a student’s experience with English. Whether this interaction is linear or circular is still unclear. The correlation revealed more motivation for the non-native group than for the native group, which contradicts earlier findings (Escamilla, 2009). The correlation analysis also showed no significant relation between the schools the students attended and the English language tests. As with motivation, the schools did correlate significantly with
Experience with English, which suggests that the exposure the schools have offered could have been translated into experience with English. It might, however, also be due to the ethnically mixed population of the schools that reported more experience with English. These children might experience more exposure to English in their day-to-day life and may therefore have more experience with English.
Returning to the significant findings from the main analysis, it should be kept in mind that the effect sizes were small to medium, at best. Despite their humble size, the results did point to specific potential interventions for education. It seems that the non-native students’ vocabulary development lags behind the process of their native peers, which seems to be due to the degree of exposure they have had. In an ideal world, specific interventions focused on increasing the students’ Dutch vocabulary could be proposed for primary schools. This way, the differences in vocabulary might diminish across groups. However, such interventions are not realistic in the Dutch educational system, simply because most teachers do not have time to
compose individualized material for each student. Research has shown that bilingual children tend to split their word knowledge over the two languages they know (Bialystok, Luk, & Kwan, 2005). Therefore, they may learn some words later than their monolingual peers. Bilingual children might already know a word in their native language, but not yet in the majority language. This suggests that the difference in vocabulary knowledge between monolingual and bilingual children may diminish by itself in due time.
Overall, then, it can be concluded that exposure to a language can positively affect a person’s proficiency in that language. In this sample, students who have had the most
experience with a language, either native Dutch children or native English children, performed best on the respective vocabulary tests. The subgroups of students who speak neither Dutch or English at home seem to be at a disadvantage. The isolated Arabic group performed worst on the Dutch and English vocabulary tests, which could be explained by the fact that their home language is very different from Dutch and English with respect to phonology and orthography. This means that they might be unable to benefit from their native language in learning the new languages (Figueredo, 2006). Experience with a language seems to be one of the most important influences on a student’s command of language, which is something that can be used both in schools as in the home settings of the students.
Alternative Explanations
The findings from this study are only partially in line with expectations. A possible explanation for the results found for this sample could be the formation of the groups. Children were grouped according to the language they reported to be raised in, but numerous other factors could have contributed to the students’ performances on the tests in this study. By focusing on the language a child is raised in, factors such as birth countries and mother tongues of the parents, the number of siblings and the language a child speaks with them, how often they use their mother tongue and how often they use Dutch, and which language they think they have mastered better were all ignored, while they could all contribute to their performance on Dutch and English language tasks (Hoff, 2006; Scheele et al., 2009). This made it difficult to prove or disprove Cummins’ Threshold hypothesis or Developmental Interdependence
hypothesis (Cummins, 1979). Moreover, some children who reported being raised in Dutch may not have native Dutch parents. Their parents may choose to use the majority language at home, while they themselves are not fluent. Dutch language input will, by definition, be different from language input from native Dutch parents. Future research should pay more attention to the creation of the groups and should incorporate as many different aspects of a child’s home language as possible.
