Investigating the unique predictive values of several variables on
English spelling in Dutch ESL-learners in their final year of
primary school: A multiple regression analysis
Julia Weening 116551434 University of Amsterdam
June 2020
2 Abstract
Introduction: In this study, the unique predictive values of Dutch and English phonological awareness (PA), Dutch and English vocabulary and Dutch spelling on English spelling were investigated in 10-12 year old Dutch students. It was hypothesized that English variables would have better predictive values than Dutch variables, but that this difference would be larger for vocabulary than for PA. It was hypothesized that Dutch spelling and PA would not have unique predictive values because there is no phoneme-grapheme relation in English. Methods: 276 students were tested on the above mentioned skills. Correlations were calculated and a stepwise multiple regression analysis was done.
Results: The difference in correlation with English spelling between the English and Dutch variable was not higher for vocabulary than for PA. Dutch PA did have a higher correlation than English PA. English vocabulary had the highest unique predicting value, followed by Dutch spelling and Dutch PA, respectively. English PA and Dutch vocabulary did not have a unique predicting value.
Discussion: There was no difference between vocabulary and PA with regard to the difference in simple relation of the Dutch and English variables and English spelling. Possibly, the difference between Dutch and English PA was large because the expected cross-linguistic transfer of PA might not have been completed yet. Or the difference between Dutch and English vocabulary was small because underlying skills may have interfered with both. Dutch spelling and PA did have a significant predicting value, despite the absence of direct
phoneme-grapheme relation. The students in this research might have used lexical strategies to spell words including grapheme-phoneme relations that are not present in their first language (L1) and sub-lexical strategies to spell words with similar phoneme-grapheme relations in their L1. Findings from this research can be implemented to prepare students better for English in higher education.
Keywords: Spelling, Vocabulary, Phonological awareness, ESL-learners, phoneme-grapheme relation
3 Table of Contents Abstract ... 2 Introduction ... 4 Orthographic Depth ... 5 Cross-linguistic transfer ... 6
Phonological awareness and L2 spelling ... 6
Vocabulary and L2 spelling ... 7
Research design and hypothesis ... 8
Materials and Methods ... 10
Sample ... 10 Procedure ... 10 Design ... 11 Statistical analysis ... 15 Results ... 18 Discussion ... 22 Conclusion ... 22 Interpretation of results ... 22
Validity and limitations ... 25
Implications ... 26
4 Introduction
Over the last decades, the English language has become more and more important. After the second world war, English became the main language of many international organizations (Van der Sijs, 1996). It is now a generally accepted language, both in
worldwide society as in science. Especially in the Netherlands, the use of English increased rapidly in the 20th and 21st centuries. Several linguists have concluded that English in the Netherlands is slowly changing from a foreign language to a language that plays a
considerable part in society, culture, commercials, and education (Booij, 2001; Gerritsen, Van Merus, Planken, & Korzilius., 2016). The number of hours of English instruction in secondary education in the Netherlands is very high compared to other European countries (Edelenbos & De Jong, 2004). In 2019, there were 317 English bachelor studies (Sandström & Neghina, 2017) and the proportion of English in Dutch master studies was even higher (Van
Oostendorp, 2012). Because English plays such a large part in higher education, it is incredibly important that children become literate in English already in primary school. Learning how to spell is an important aspect of literacy (McCardle, Chhabra, & Kapinus, 2008). Moreover, spelling is an area in which students who are learning English as a second or additional language (ESL-learners) are known to experience difficulties with (James, Scholfield, Garrett, & Griffiths, 1993; Kibel & Miles, 1994). However, according to the Dutch national centre of expertise for curriculum development, most primary schools mainly focus on oral skills like listening and speaking (Rose, 2016). The reason for this is that the end goals for English in primary education, set by the Dutch national program of English in
primary education (EIBO), are centred around those skills.
Acquisition of spelling (both first language and second language) has received less attention in research than reading (Treiman, 1997). A small increase in spelling research in the past years has occurred due to the significance of spelling in literacy acquisition (Conrad,
5 2008), yet several studies that will be discussed in the subsequent paragraphs have only investigated reading acquisition and not spelling acquisition. Therefore, it is important that more research will be done into the acquisition of English spelling and which first and second language skills might have an influence on this acquisition.
Orthographic Depth
An important term that is often used in research on the acquisition of a language is ‘orthographic depth’. Orthographic depth indicates how often the same combination of letters have the same pronunciation in a certain language (Leufkens, 2015; Sipra, 2016). A
difference is made between shallow orthographies (transparent) and deep orthographies (opaque) (Besner & Smith, 1992). Shallow orthographies have a direct letter-sound correspondence, which causes the spelling of words to be very consistent. This can be reflected in good spelling performance (Borzone de Manrique & Signori, 1994). Besides Spanish, the languages Finnish, Italian, and Turkish turned out to be very transparent in writing (Seymour, Aro, & Erksine, 2003; Marjou, 2019). According to Seymour et al. (2003), Dutch was also below the threshold of orthographic complexity. English, on the contrary, scored extremely low in phonemic transparency, suggesting that English has a deep
orthography. Here, the relation between graphemes (letters) and phonemes (speech sounds) is less direct and the student must learn the unusual pronunciation and spelling of words.
Children who have a deep orthography as their first language usually develop reading skills at a slower rate (Seymour et al., 2003), and this delay was also found for spelling skills
(Caravolas, 2006). Learning to read and spell in a deeper orthography involves the
implementation of a logographic (word reading) and alphabetic (letter reading) foundation, the so-called ‘dual foundation model’, whereas learning to read and spell in a more shallow orthography only needs an alphabetic foundation (Seymour et al., 2003).
6 Cross-linguistic transfer
When a new language is being acquired, first-language transfer takes place
(Figueredo, 2006). First-language transfer refers to the effect of knowledge of a first language (L1) on the acquisition of second language (L2) skills, in this case, English spelling. This cross-linguistic transfer (a transfer between languages) also seems to be affected by the difference in the orthographic depth of L1 and L2.
Conflicting findings have been observed about the influence of L1 spelling on L2 spelling. Sparks, Patton, Ganschow, Humach, and Javorsky (2008), who examined students learning an L2 with a more shallow orthography than their L1, found that early L1 spelling predicted 47% of the variance in L2 spelling. Wade-Woolley and Siegel (1997), who studied students learning an L2 with a deeper orthography than their L1 did not find a relation
between L1 spelling performance and L2 spelling performance. Learning to spell in L2 might be more dependent on L1 spelling skills when a more shallow orthographic language is being learned because students can then rely on the direct grapheme-phoneme relation (Dixon, Zhao & Joshi, 2010). Since English has a deep orthography, L1 spelling skills might not have a large influence on L2 spelling.
Phonological awareness and L2 spelling
In general, one of the most important contributors to spelling is phonological
awareness. Phonological awareness is a cognitive skill that shows the ability to identify and manipulate spoken parts of words (Stanovich, 1994). Several studies have found significant medium to high correlations between L1 phonological awareness and L1 spelling (Tornéus, 1984; Bruck & Treiman, 1990; de Manrique & Signori, 1994). In the study of de Manrique and Signori (1994), in which Spanish was the L1 of the students, the relative shallow
7 An enhanced phonological awareness also seems to be transferred from L1 to L2 (Yeung, Siegel, & Chan, 2012; De Sousa, Greenop & Fry, 2010). Durgunoğlu, Nagy, and Hancin-Bhatt (1993) showed that enhanced L1 (Spanish) phonological awareness not only resulted in better L2 (English) phonological awareness but also better L2 word reading. Besides, Zeguers, Van den Boer, Snellings, and De Jong (2018) found that word reading in deep orthographic French was slightly more influenced by cognitive skills like phonological awareness than by L1 reading skills. Nonetheless, this effect might be different for spelling. In general, spelling is more unpredictable than word reading. This effect could be even greater for deep orthographies like English because of the lack of direct grapheme-phoneme relation (Caravolas, Hulme, & Snowling, 2001). Therefore, the influence of L1 phonological
awareness on L2 spelling might be diminished by the orthographic depth of the English language.
Vocabulary and L2 spelling
Van Gelderen et al. (2004) found that L2 (English) vocabulary is very important for L2 reading. L2 linguistic knowledge like vocabulary seems to have even more impact on L2 reading than L1 (Dutch) reading does. It would make sense that the same effect occurs for English spelling. In English, where there is no direct grapheme-phoneme relation, students can often not rely on their L1 spelling knowledge. Instead, they might rely more on their memory of the graphic representation of L2 words.
L1 vocabulary is not likely to have a unique influence on L2 spelling skills. It can be speculated that a good L1 vocabulary will result in a good L2 vocabulary when similar underlying skills like memory are involved. Nonetheless, no evidence has been found of a direct cross-linguistic transfer of vocabulary. Tabors, Paéz, and López (2003) even found moderate negative correlations between English and Spanish vocabulary for bilingual preschool children. Furthermore, San Francisco, Mo and Carlo (2006) investigated the
8 influence of English and Spanish vocabulary on English spelling and found that Spanish (L1) vocabulary is a positive predictor for Spanish-influenced (L2) spelling, but a slightly negative predictor for deep orthographic English (L2) spelling. English vocabulary (L2) was indeed a positive predictor for English (pseudo-)spelling. This would suggest that Dutch vocabulary is not a unique predictor for English spelling, but English vocabulary could be. Since Van Gelderen et al. (2004) also did not find a significant unique influence of Dutch vocabulary on English reading, it would be interesting to investigate if this effect also occurs for Dutch vocabulary on English spelling.
Research design and hypothesis
The goal of this study was to investigate if and how much Dutch and English
phonological awareness, Dutch and English vocabulary, and Dutch spelling predict English spelling in Dutch ESL-learners of 9-12 years old. The results of this study can help improve the English curriculum in primary education, which will help students perform better in higher education.
The research question 'Which Dutch and English skills have the highest unique predictive value for English spelling skills?' was explored. Predictive value is a term used to indicate how much a certain variable predicts the variance in English spelling. Both the simple relationship as well as the additional predictive value of each variable were assessed. Simple relationships were assessed with the association between a variable and English spelling. The additional, unique predictive values were taken into account to find out how much a specific variable adds to the predicting model when corrected for other variables.
To answer this research question, 276 children were tested on several skills. English spelling was tested with the EIBO method, which was based on a list of words that primary school students are expected to know before entering high school. In addition, English and Dutch phonological awareness, English and Dutch vocabulary, and Dutch spelling were
9 measured. Correlations between each variable were calculated to test the assumptions for the multiple regression analysis and to determine the simple relationships. Then, to investigate the additional, unique predictive value of each variable on English spelling, a stepwise multiple regression analysis was done. Dutch variables were entered first to test the additional
predictive value of English variables. Furthermore, cognitive skills (phonological awareness) were entered first to test the additional predictive value of linguistic skills (vocabulary and spelling). The initial predictive values of the variables were compared to the additional predictive values in the multiple regression analysis to investigate whether the observed predictive value in the correlation analysis is unique and still present when corrected for other variables.
Expected was that English variables would have larger predictive values of English spelling than Dutch variables, but that this difference would be larger for linguistic skills like vocabulary than for cognitive skills like phonological awareness. Two findings from previous research support this hypothesis and it was hypothesized that these findings would also be observed in this study. Firstly, it was hypothesized that no large difference would be observed between the predictive values of Dutch and English phonological awareness. Phonological awareness seems to transfer from one language to another, but this effect has not been found for vocabulary. Secondly, it was hypothesized that a large difference would be observed between the predictive values of Dutch and English vocabulary because San Francisco et al. (2006) found that L1 vocabulary had a small negative predicting power of L2 spelling, while L2 vocabulary had a large positive predicting power of L2 spelling. Furthermore, it was expected that L1 phonological awareness and spelling might not even have such a large influence on L2 spelling. Students can often not rely on their L1 spelling or phonological awareness knowledge because there is no direct grapheme-phoneme relation in English.
10 Materials and Methods
Sample
The sample of this study consisted of 276 primary school students from 7 different primary schools. Age varied from 10-13 years, with a mean age of 11 years and 9 months. The male/female ratio was 52/48. All students were in their final grade of primary school (MYP 1 in International Baccalaureate Grade Level, or grade 6 in US Grade Level). The 7 different primary schools were all located in the Netherlands (6 in North-Holland and 1 in South-Holland). The vast majority of the students (89%) were born in the Netherlands. Of the students, 62% only spoke Dutch at home, 35% spoke Dutch and another language at home (including Moroccan, Chinese, German, Ghanaian, Spanish, Turkish). The remaining 3% did not speak Dutch at home. Besides being in MYP 1, no further exclusion criteria were set.
Procedure
For this study, data of the ORWELL-project from the Rudolf Berlin Center has been used. In this large project, several Dutch skills, English skills, and cognitive skills were tested for three years in a row. The first tests were taken in spring 2017 when the students were in Primary 5 (IB Grade Level). The second tests were taken in spring 2018 when the students were in Primary 6 (IB Grade Level). For this study, the data of the final round of experiments in spring 2019 were used. The primary schools were selected carefully and differences in Social Economical Statuses (SES) were taken into account. For the SES-indication, the status scores of the Dutch Social Cultural Research Institute (SCP) in 2016, based on the mean income of the district of the school, were used. The schools were approached by employees from the ORWELL-project and the parents of the students provided informed consent. Tests for Dutch and English vocabulary and spelling were taken in class. Those tests took place in the classroom the students were already familiar with. The desks of the students were placed in rows and the students all had assigned spots. All students were given a tablet,
11 headphones, and a specific participation code. For privacy reasons, only the participation code was used in documentation of the results and not the name of the student. The participation code had to be entered by a test supervisor first and then by the student. After that, the test session started. First, Dutch vocabulary (15 min) and spelling (30 min) were tested. Second, English spelling (25 min) and vocabulary (20 min) were tested. Tests for Dutch and English phonological awareness were taken individually. First, a group explanation was given by one of the supervisors. After that, each student was called into a separate room where the tests took place. A supervisor was sitting across from the student with a laptop and the student was not allowed to look at the screen. No substantive feedback was given during the test exercises. After each session, the student was given a little present such as a sticker. The answers were recorded with a voice recorder and tracked in Qualtrics or the programme used to test Dutch phonological awareness. Dutch phonological awareness (15 min) was tested first and English phonological awareness (10 min) second.
All schools have received a feedback report describing the outcomes of all the tests and the correlations between skills.
Design
Dutch phonological awareness
Dutch phonological awareness was tested with the phonemic analysis test (FAT-R) (De Groot, Van den Bos, & Van der Meulen, 2014). FAT-R contained two subtests, phoneme deletion, and phoneme exchange. In this project, only the subtest phoneme exchange was used. Students heard two words and were asked to interchange the initial letters of the words and pronounce the newly formed words. The subtest contained three practice items and twelve test items. Both the number of correct answers and the reaction time were tracked. Dutch phonological awareness was not tested in 2019. Therefore the data of 2018 was used.
12 Figure 1. Test design used to test Dutch phonological awareness
Dutch vocabulary
Dutch vocabulary was tested with the Dutch Peabody Picture Vocabulary Test (PPVT-NL), the Dutch version of a standardized international vocabulary test (Dunn & Dunn, 2005). Students heard one word and had to pick one of the four presented pictures that matched the word best. Originally, PPVT-NL contained 17 sets of 12 items. However, in the Orwell-project session of 2019, only sets 9-13 were used, containing a total of 60 items.
Figure 2. Test design used to test Dutch vocabulary Dutch spelling
Dutch spelling was tested with the Dutch Schoolvaardigheidstoets (SVT) (De Vos & Braams, 2015). This skill test contained two components, spelling of nouns and spelling of verbs. Students first heard a sentence and after the sentence, the word the students had to write down was mentioned. The test of the spelling of nouns originally contained 6 blocks of each 15 words. However, the test used in this project contained a total of 30 words, divided into 2 blocks, because the first blocks were assumed to be too easy for the students. The test
13 of spelling of verbs originally contained 3 blocks of each 18 words, but the test used in this project contained a total of 36 words, also divided into 2 blocks. All the blocks both had increasing difficulty.
Figure 3. Test design used to test Dutch spelling English phonological awareness
English phonological awareness was tested with the Comprehensive Test of
Phonological Processing (CTOPP) (Wagner, Torgesen, Rashorte, & Pearson, 1999). CTOPP measured phonological processing in English and contained several subtests. Phonological memory and Rapid naming were also tested with a subtest of CTOPP. To test phonological awareness, the subtest Blending Words was used. Students were asked to put sounds together and make an existing word out of it. Before the test started, the students were allowed to practice with 6 items. The real test contained 20 items. When three consecutive mistakes were made, the test was terminated.
Figure 4. Test design used to test English phonological awareness
14 English vocabulary
English vocabulary was tested with the Peabody Picture Vocabulary Test (PPVT-EN) (Dunn & Dunn, 1997). Students heard an English word, voiced by a native British speaker, and had to pick one of the four presented pictures that matched the word best. In this project, sets 3-9 were used, containing a total of 84 items.
Figure 5. Test design used to test English vocabulary English spelling
English spelling was tested with two different tests, the Wide Range Achievement Test (WRAT) and the English in Primary Education (EIBO)-spelling test. The EIBO spelling test was designed for the ORWELL-project. A list of words that students are expected to master before entering secondary education had been made. The EIBO spelling test was based on these words. Since this research focuses on how children acquire English spelling in school and with the purpose of performing better in higher education, only the results of the EIBO spelling test were included in this research. Students heard a full English sentence and saw the same sentence on their screen but with one word missing. They had to write down the missing word. The test contained 30 items. All items, in both tests, were voiced by a native British speaker.
15 Figure 6. Test design used to test English spelling
Statistical analysis
All statistical analysis was done in R-studio Version 1.0.153. First, correlations were calculated. Then, a stepwise multiple regression was done. In this study, the effect of several Dutch and English variables on English spelling was investigated. Therefore, English spelling (EN_spelling) was the dependent variable in the stepwise multiple regression. Dutch spelling (NL_spelling), Dutch phonological awareness (NL_PA), English phonological awareness (EN_PA), Dutch vocabulary (NL_voca), and English vocabulary (EN_voca) were the independent variables. All variables were continuous.
Figure 7. Overview of dependent variable English spelling (EN_spelling) and
independent variables Dutch phonological awareness (NL_PA), Dutch vocabulary (NL_voca), Dutch spelling (NL_spelling), English phonological awareness (EN_PA) and English vocabulary (EN_voca).
16 In EN_PA, one student did not participate, so the results of that student in the other tasks were excluded. NL_PA was not measured in 2019, so for NL_PA data of 2018 was used. In 2018, 285 students participated, while in 2019 only 277 did. Therefore, the data of students who did not participate in 2019 and the student who did not participate in EN_PA were excluded. In total, data of 276 students were used.
All variables were recalculated into a percentage of correct answers. The total amount of correct answers was divided by the total amount of items and multiplied with 100. For NL_spelling, the scores for the nouns test and the verbs test were combined. For NL_PA, a standardized t-score was calculated, bases on the number of correct answers and reaction time. This t-score was also recalculated into a percentage.
First, correlations between all variables were calculated with the Pearson Correlation test to test the assumptions for the regression analysis and to determine the simple relationship of each variable with the dependent measure English spelling. The assumption of normality was tested with the Shapiro-Wilk test. The assumption of homoscedasticity was tested with the Levene test. Table 2 shows the correlations between all variables. Variables with significant correlations with EN_Spelling were selected. It was checked if intercorrelations between variables higher than 0.8 were present. No variables were found to have an
intercorrelation higher than 0. So all variables were put into a multiple regression analysis.
Second, a stepwise multiple regression analysis was done by adding variables one by one to determine the additional unique predictive value of a variable, when corrected for other variables. This was done to test how much the predictive value of the model changes by adding a variable that is expected to have an influence on EN_spelling and thus change the predictive value. Dutch variables were entered first to test the additional predictive value of English variables. Furthermore, cognitive skills (phonological awareness) were entered first to test the additional predictive value of linguistic skills (vocabulary and spelling). Therefore,
17 NL_PA were entered first, followed by NL_voca and NL_spelling. EN_PA and EN_voca were entered last.
Table 3 shows the adjusted R2 of the prediction model in each step. For each step, the additional variance (ΔR2) by assessing the difference in R2 of the new model and the previous
model. The best prediction variables were chosen by looking at the additional variance and the β-coefficient of the variable in the final model.
The additional predictive value of each variable was compared to the correlation between that variable and English spelling to investigate whether the observed predictive value in the correlation is unique and still present when corrected for other variables.
18 Results
The descriptive statistics of the raw scores of all variables were summarized in table 1. EN_spelling had a remarkably wide range of scored scores (1-30).
Table 1. Descriptive statistics of the results of all 276 participants, showing the
maximum possible score (Max score), Mean, Standard deviation (SD) and the range between minimum and maximum scored scores (Range).
The Shapiro-Wilk test showed that the variables EN_PA (W=.905; p=.100) and NL_PA (W=.937; p=.051) were normally distributed. EN_spelling (W=0.881; p=0.003), NL_spelling (W=0.896; p=.001), EN_voca (W=.931; p=.009) and NL_voca (W=.890; p=.001) were, despite transformations, not normally distributed. For those variables, the non-parametric Spearman method was used to calculate correlations.
The Levene test showed that the results of EN_spelling and EN_PA were not equally distributed (F=1.976; p=.017). All the other variables were equally distributed.
Calculations of correlations between each variable showed that all correlations were significant (see table 2). No intercorrelations higher than 0.8 were found, meaning that all variables could be included in the stepwise multiple regression analysis. EN_spelling and NL_spelling had the highest correlation (ρ=.602; p<2.2e-16).
EN_spelling - EN_voca correlation (ρ=0.542 ) and EN_spelling – NL_voca correlation (ρ=0.33) differed 0.21 from each other. EN_spelling – EN_PA correlation (ρ=0.34) and EN_spelling – NL_PA correlation (ρ=0.51) differed 0.17 from each other. For
Variable name Max score Mean (SD) Range
English spelling 30 20.138 (6.046) 1-30
Dutch spelling 66 46.717 (9.922) 12-64
English phonological awareness 20 13.660 (2.932) 4-19
Dutch phonological awareness 60 49.486 (8.188) 20-60
English vocabulary 84 62.975 (10.577) 23-84
19 phonological awareness, the Dutch variable had a higher correlation with EN_spelling. For vocabulary, the English variable had a higher correlation with EN_spelling.
Table 2. Correlations between variables English spelling (EN_spelling), Dutch spelling
(NL_spelling), English vocabulary (EN_voca), Dutch vocabulary (NL_voca), English phonological awareness (EN_PA) and Dutch phonological awareness (NL_PA).
1. p < .05, ** p < .01, *** p < .001
2. Each correlation was found to be significant. No intercorrelations higher than 0.8 were found.
3. A Spearman correlation coefficient (ρ) was calculated for correlations that involved variables that were not normally distributed.
4. A Pearson correlation coefficient (r) was calculated for correlations that involved variables that were normally distributed.
5. EN_spelling and NL_spelling had the highest correlation (ρ=0.60; p<2.2e-16).
1 2 EN_spelling NL_spelling EN_voca NL_voca EN_PA NL_PA
EN_spelling x - - - - - NL_spelling p=0.60 *** 3, 5 x - - - - EN_voca ρ=0.54 *** ρ=0.19 ** x - - - NL_voca ρ=0.33 *** ρ=0.42 *** ρ=0.38 *** x - - EN_PA ρ=0.34 ***3 ρ=0.26*** ρ=0.22*** ρ=0.28 *** x - NL_PA ρ=0.51 *** ρ=0.47 *** ρ=0.15 * ρ=0.30 *** r=0.403 *** x
20 Stepwise multiple regression (table 3) showed that in model 1 (F(1,274) = 96.85; R2=.259), NL_PA accounted for 26% of the variation in English spelling. In model 2
(F(2,273) = 55.66, ΔR2=.026), NL_voca was added to the model and explained an additional 2,6%3% of variance. In model 3 (F(3, 272) = 60.14, ΔR2 = .107), addition of NL_spelling explained an additional 10.7% of variance. Simultaneously, NL_voca (β = .09, t(271) =1.06, p = .288) lost its significance after the introduction of NL_spelling.In model 4 (F(4,271) = 45.70, ΔR2 = .002), addition of EN_PA did not explain significant differences. In model 5 (F(5,270)=64.19, ΔR2=.103), addition of EN_voca explained an additional 14.1% of the variance in EN_spelling.
When all 5 independent variables were included, only NL_PA, EN_voca, and
NL_spelling remained to have significant predictive values of EN_spelling. EN_voca (β =.65, t(270) =.9.1 p<2.2e-16) had the highest β-coefficient and explained the highest additional variance (ΔR2 = .141). NL_spelling had the second-highest β-coefficient in the final predicting model (β =.52, t(270) =7.75, p=1.43e-13) explaining an additional variance of 10.3% when it was added to the model in step 3. NL_PA had the third-highest β-coefficient(β =.43, t(270) =5.84, p=1.47e-8) and explained an initial variance of 25.9% in the first step . Together, all 5 variables accounted for 53.5% of the variance in EN_spelling.
The additional variance and β-coefficient of each variable were compared to the correlation. NL_spelling had the highest correlation while EN_voca had the highest additional variance and β-coefficient in the multiple regression analysis. EN_PA did not have a
significant additional variance and β-coefficient but did have a significant medium correlation. NL_voca did not have a significant β-coefficient in model 3, 4, and 5, but did have a significant medium correlation.
21 Table 3. Hierarchical stepwise multiple regression analysis showing the additional
predictive value of Dutch phonological awareness (NL_PA), Dutch vocabulary
(NL_voca), English phonological awareness (EN_PA), English vocabulary (EN_voca) and Dutch spelling (NL_spelling) on English spelling. (N=276).
1. In each step, a variable was added to the model. For each step, the adjusted total
variance (R2), additional variance (ΔR2), and F-statistics of the predicting model were calculated.
2. For each variable in a model, the β-coefficient, t-score, and p-value were calculated.
3. * p < .05, ** p < .01, *** p < .001.
4. In model 1 (F(1,274) = 96.85; R2=.259), NL_PA accounted for 26% of the variation in English spelling.
5. In model 2 (F(2,273) = 55.66, ΔR2=.026), NL_voca was added to the model and explained an additional 3% of variance.
6. In model 3 (F(3, 272) = 60.14, ΔR2 = .107), addition of NL_spelling explained an additional 10.7% of variance. Simultaneously, NL_voca (β = .09, t(271) =1.06, p = .288) lost its significance after the introduction of NL_spelling.
7. In model 4 (F(4,271) = 45.70, ΔR2 = .002), addition of EN_PA did not explain significant differences.
8. In model 5 (F(5,270)=64.19, ΔR2=.103), addition of EN_voca explained an additional 14.1% of the variance in EN_spelling.
Model 1 R2 ΔR2 F-statistics Independent
Variable2
β-coefficient t-score p-value3
Model 1 (NL_PA)4 .259 - 96.85*** NL_PA .76 9.84 <2.2e-16 *** Model 2 (+ NL_voca)5 .285 .026 55.66*** NL_PA .68 8.58 7.38e-16 *** NL_voca .29 3.31 .001** Model 3 (+ NL_spelling)6 .392 .107 60.14*** NL_PA .45 5.73 2.61e-8 *** NL_voca .09 1.06 0.287 NL_spelling .53 7.03 1.7e-11 *** Model 4 (+ EN_PA)7 .394 .002 45.7*** NL_PA .42 5.03 8.99e-7 *** NL_voca .07 .83 .409 NL_spelling .52 7.00 2.06e-11 *** EN_PA .01 1.35 .179 Model 5 (+ EN_voca)8 .535 .141 64.19*** NL_PA .43 5.84 1.47e-8 *** NL_voca -.15 -1.92 .056 NL_spelling .52 7.79 1.43e-13 *** EN_PA .03 .43 .670 EN_voca .65 9.11 <2.2e-16 ***
22 Discussion
Conclusion
In this paper, the research question 'Which Dutch and English skills have the highest unique additional predictive value for English Spelling skills?' was explored. Based on correlations, all variables had a significant medium or large simple relationship with English spelling. There was no difference between vocabulary and phonological awareness with regard to the difference in simple relation of the Dutch and English variables and English spelling. English vocabulary had a higher correlation with English spelling than Dutch vocabulary, but Dutch phonological awareness had a higher correlation than English
phonological awareness. Based on the multiple regression analysis, English vocabulary turned out to have the highest additional predictive value of English spelling. It was added last, correcting for all other variables, but it still had the highest additional variance and the highest β-coefficient in the final predictive model. After that, Dutch spelling had the best additional predictive value, followed by Dutch phonological awareness. English phonological awareness did not have a significant contribution to the predictive model. Dutch vocabulary had a
significant correlation, coefficient, and additional variance, but it lost its significance in β-coefficient in the predicting model after Dutch spelling was entered into the model. In sum, English vocabulary and Dutch spelling and phonological awareness had the highest unique additional predictive value.
Interpretation of results
It was expected that English skills would have better predictive values of English spelling than Dutch skills, but that this difference would be larger for vocabulary than for phonological awareness. The first part of this hypothesis was found not to be true because, for phonological awareness, the Dutch variant had a better simple relationship with English spelling and a better unique predictive value than the English variant did. The second part of
23 the hypothesis was also not observed. There was no difference between vocabulary and phonological awareness with regard to the difference in simple relation of the Dutch and English variables and English spelling. Maybe the difference in simple relationship with English spelling between Dutch and English phonological awareness was larger than expected because the expected transfer of phonological awareness was not fully completed yet. The threshold hypothesis states that a minimum threshold in language proficiency (both L1 and L2) must be passed for L1 skills to be transferred to L2 (Goodman, 1971; Cummins, 1979; Alderson, 1984). Given that most students in this research were just starting to learn English and that languages with a deep orthography are usually acquired at a slower rate (Seymour et al., 2003), their L2 language proficiency might not have been developed well enough for a successful transfer. It would be interesting to investigate if English phonological awareness has a better unique additional predictive value and if the difference between the predictive value of English and Dutch phonological awareness is smaller when students are a little bit older and their English proficiency is higher.
Another reason for the fact that the second part of the hypothesis was not confirmed by the results could be that the difference in the simple relation with English spelling between Dutch and English vocabulary was smaller than expected. Dutch vocabulary did have a significant relationship with English spelling, but no significant unique predictive value. Possibly, underlying skills like long term memory that are involved in both Dutch and English vocabulary were responsible for the significant correlation of Dutch vocabulary and English spelling. Long term memory plays an important role in vocabulary because vocabulary connects the graphic form of words to the knowledge about the meaning of that word and both are stored in the long term memory (Malovr & Benati, 2018). It is believed that
bilinguals have a common storage system for word representations in both their L1 as well as their L2 (Macaro, 2003). A well-developed word representation system in the long term
24 memory is thus likely responsible for the relationship between Dutch vocabulary and English spelling. However, knowledge of L1 words does not always directly facilitate L2 spelling. San Fransisco et al. (2006) found that Spanish (L1) vocabulary only predicted Spanish-influenced L2 spelling. In that case, the memory of the graphic representation of L1 words was used to facilitate L2 spelling. In this research, no difference was made between Dutch-influenced and non-Dutch-Dutch-influenced words, but according to Williams (1986), less than 1% of all English words are Dutch-influenced. Therefore, Dutch vocabulary was not likely to facilitate English words. This could explain why Dutch vocabulary did not have a significant unique predictive value.
Moreover, the expectation that Dutch spelling and phonological awareness might not have significant predictive values because of the absence of direct grapheme-phoneme
relation, was also not supported by the results. Dutch spelling and phonological awareness did not have as much influence as English vocabulary, but they still had high, significant
additional predictive values. This suggests that there is indeed a cross-linguistic transfer of phonological awareness and spelling, which is in line with Figueredo (2006), despite the differences in the orthographic depth of Dutch and English. An explanation for the significant influence of Dutch phonological awareness and spelling could lie in the distinction between lexical and sub-lexical strategies. According to Coltheart, Rastle, Perry, Langdon, and Ziegler (2001), there are two separate routes for processing words, lexical and sub-lexical. The lexical strategy works at the whole-word level. If a word has been previously memorized, it does not matter if the word is regularly or irregularly spelled. The sub-lexical route relies on grapheme-phoneme relations. This route facilitates the spelling of regular words and non-words, but not irregular words (Joubert & Lecoures, 2000). It has been assumed that spelling in transparent languages only requires sub-lexical strategies (De Sousa, Greenop & Fry, 2011). However, lexical systems seem to be present in transparent languages too (Fisher-Baunn & Rapp, 2015).
25 Individuals with damage in their sub-lexical systems, are not impaired in their spelling of real words, because they can use lexical strategies. The students in this research might have used their lexical strategies acquired in L1 spelling to spell words including grapheme-phoneme relations that are not present in their L1. Likewise, they might have used their sub-lexical strategies to spell words with similar phoneme-grapheme relations in their L1. As mentioned before, vocabulary probably facilitates English spelling through the memorization of the graphic form of words. Therefore, vocabulary is likely connected to the lexical spelling strategy. In further research, it would be interesting to make a distinction between words including a phoneme with the same grapheme-phoneme relation in English and words including grapheme-phoneme relation that does not exist in English. Expected is that skills like phonological awareness are correlated with words including grapheme-phoneme relations that exist in both languages, allowing the sub-lexical strategy to be used. Van Berkel (1987) found that Dutch ESL-learners have little trouble using ‘ou’ to represent /au/ in Dutch ‘mouw’ as well as in English ‘house’. For words including a grapheme-phoneme that does not exist in their L1, an inappropriate application of the sub-lexical strategy might be used, and the lexical strategy might be necessary to correctly spell the English words. Therefore it is expected that English spelling of words including a grapheme-phoneme relation that does not exist in their L1 will be correlated with skills that involve the long-term memory, necessary for the lexical strategy.
Validity and limitations
This research has measured correctly what needed to be measured. All of the used tests were standardized and reliable tests and often used internationally. The test for English
spelling was a test based on the EIBO-list, a list of words students are supposed to know before entering higher education. Since this paper was focused on English spelling to prepare students better for higher education, the EIBO test was very suitable.
26 Tests were taken in a classroom setting, similar to the setting in which students learn a foreign language at school. Besides, the tests were taken in seven different Dutch elementary schools with very different SES-scores and education systems. Therefore, the conclusions are generalizable to almost all other Dutch students in their final years of primary school.
However, since the results are likely influenced by the orthographic depth of the first and second language, it is not automatically applicable to students in other countries with a different first language or students learning a different language.
Dutch phonological awareness was not measured in 2019. Therefore data of one year earlier was used, meaning that the students were also one year younger. Their phonological awareness could have been further developed in 2019, which could have resulted in a better transfer of phonological awareness. It would be interesting to investigate whether English phonological awareness does have a significant influence if both Dutch and English phonological awareness are further developed.
Implications
In this research, it was found that English vocabulary as well as Dutch spelling and phonological awareness influence English spelling. This knowledge needs to be implemented in ESL education in Dutch primary schools to improve the English spelling skills of students before they enter high school. The focus of the English curriculum in primary schools is currently mainly on oral skills (Rose, 2016). Vocabulary already plays are a large part of the curriculum, but the students often only process the words auditory and do not see them visually written out. Because the spelling of irregular words relies on the memory of the spelling of words, it is important to familiarize students with the visual representation of English words. Another good idea would be to make students more familiar with English grapheme-phoneme relations, so they can apply those and start using their sub-lexical strategy more. Improving the general English language proficiency of the students will likely also
27 result in a better cross-linguistic transfer of phonological awareness and spelling. All in all, the implementation of these findings in the adaptation of primary school curricula will enable students to be better prepared for learning and using English in higher education.
28 References
Alderson, J. C. (1984). Reading in a foreign language: A reading problem or a language problem? In J. C. Alderson & A. H. Urquhart (Eds.) Reading in a foreign language (pp. 1-27). New York, United States: Longman.
Besner, D. & Smit, M.C. (1992). Basic processes in reading: Is the orthographic depth hypothesis sinking? In R. Frost & M. Katz (Eds.) Advances in psychology: orthography, phonology, morphology and meaning (pp. 45-66). Amsterdam, The Netherlands: North Holland Elsevier.
Booij, G. (2001). English as the lingua franca of Europe: A Dutch perspective. Lingua e Stile, 36(2): 347-357.
Borzone de Manrique, A. M., & Signorini, A. (1994). Phonological awareness, spelling and reading abilities in Spanish-speaking children. British Journal of Educational
Psychology, 64(3), 429–439. https://doi.org/10.1111/j.2044-8279.1994.tb01114.x
Bruck, M., & Treiman, R. (1990). Phonological awareness and spelling in normal children and dyslexics: The case of initial consonant clusters. Journal of Experimental Child Psychology, 50(1), 156–178. https://doi.org/10.1016/0022-0965(90)90037-9
Caravolas, M. (2006). Learning to spell in different languages: How orthographic variables might affect early literacy. In R.M. Joshi & P.G. Aaron (Eds.), Handbook of orthography and literacy, (pp. 497–511). Mahwah, New Jersey: Erlbaum.
Caravolas, M., Hulme, C., & Snowling, M. J. (2001). The foundations of spelling ability: Evidence from a 3-year longitudinal study. Journal of Memory and Language, 45(4), 751-774. https://doi.org/10.1006/jmla.2000.2785
Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual-cascade model of visual word recognition and reading aloud. Psychological Review, 108, 204–56. https://doi.org/10.1037/0033-295X.108.1.204
29 Conrad, N. J. (2008). From reading to spelling and spelling to reading: Transfer goes both ways. Journal of Educational Psychology, 100(4), 869–878. https://doi.org/10.1037/a0012544
Cummins, J. (1976). The Influence of Bilingualism on Cognitive Growth: A Synthesis of Research Findings and Explanatory Hypotheses (Working Papers on Bilingualism 9).
Toronto, Canada: The Ontario Institute for Studies in Education, Bilingual Education Project.
De Groot, B.J.A., Van den Bos, K. P., & Van der Meulen, B. F. (2014). Fonemische Analyse Test Revised (FAT-R). Amsterdam, The Netherlands: Pearson.
De Sousa, D. S., Greenop, K., & Fry, J. (2010). The effects of phonological awareness of Zulu-speaking children learning to spell in English: A study of cross-language
transfer. British Journal of Educational Psychology, 80(4), 517–533. https://doi.org/10.1348/000709910x496429
De Sousa, D. S., Greenop, K., & Fry, J. (2011). Cross-language transfer of spelling strategies in English and Afrikaans Grade 3 children. International Journal of Bilingual Education and Bilingualism, 14(1), 49–67. https://doi.org/10.1080/13670051003657959
De Vos, T., & Braams, T. (2015). Schoolvaardigheidstoets Spelling. Amsterdam, The Netherlands: Boom Test Uitgevers.
Dixon, L. Q., Zhao, J., & Joshi, R. M. (2010). Influence of L1 Orthography on Spelling English words by bilingual children: A natural experiment comparing syllabic, phonological, and morphosyllabic first languages. Learning Disability Quarterly, 33(3), 211–221.
https://doi.org/10.1177/073194871003300309
Dunn, L. M., & Dunn, L. M. (1997). Peabody picture vocabulary test-III. Circle Pines, Minnesota, United States: American Guidance Service.
Dunn, Lloyd M. & Dunn, Leota M. (2005). Peabody Picture Vocabulary Test-III-NL, Dutch version by Liesbeth Schlichting. Amsterdam, The Netherlands: Harcourt Assessment B.V.
30 Durgunoğlu, A. Y., Nagy, W. E., & Hancin-Bhatt, B. J. (1993). Cross-language transfer of phonological awareness. Journal of Educational Psychology, 85(3), 453–465.
https://doi.org/10.1037/0022-0663.85.3.453
Edelenbos, & de Jong (2004). Vreemdetalenonderwijs in Nederland. Retrieved from
http://www.gerardwesthoff.nl/downloads/nabmvt-Vreemdetalenonderwijs%20in%20Nederland.pdf
Figueredo, L. (2006). Using the known to chart the unknown: A review of first-language influence on the development of English-as-a-second-language spelling skill. Reading and Writing, 19(8), 873–905. https://doi.org/10.1007/s11145-006-9014-1
Fisher-Baum, S., & Rapp, M. (2015). The Handbook of Adult Language Disorders. Amsterdam, the Netherlands: Amsterdam University Press.
Gerritsen, M., Van Merus, F., Planken, B., & Korzilius, H. (2016). A reconsideration of the status of English in the Netherlands within the Kachruvian Three Circles model. World Englishes, 35(3), 457–474. https://doi.org/10.1111/weng.12206
Goodman, K. S. (1971). Psycholinguistic universals in the reading process. In P. Pimsleur & T. Quinn (Eds.), The psychology of second language learning (pp. 135-142). Cambridge, United Kingdom: Cambridge University Press
James, C., Scholfield, P., Garrett, P. & Griffiths, Y. (1993). ‘Welsh bilinguals’ spelling: an Error Analysis.’ Multilingual and Multicultural Dev. 14(4), 287-306.
Joubert, S.A. & Lecours, A.R. (2000). The role of sub-lexical graphemic processing in reading. Brain and Language, 72(1): 1–13. https://doi.org/10.1006/brln.1999.2279
Kibel, M. & Miles, T. R. (1994). Phonological errors in the spelling of taught dyslexic
children. In C. Hulme & M. Snowlings (Eds), Reading development and dyslexia: a challenge for educators (pp. 105-127). London, United Kingdom: David Fulton Publishers
31 Leufkens, S. (2015). Transparency in language. Retrieved from
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.832.8251&rep=rep1&type=pdf
Macaro, E. (2003). Teaching and Learning a Second Language. London, United Kingdom: Continuum.
Malovrh, P. A., & Benati, A. G. (2018). The Handbook of Advanced Proficiency in Second Language Acquisition. Hoboken, New Jersey, United States: Wiley.
Marjou, X. (2019). OTEANN: Estimating the Transparency of Orthographies with an Artificial Neural Network. ArXiv, abs/1912.13321.
McCardle, P. D., Chhabra, V., & Kapinus, B. A. (2008). Reading Research in Action. Zaltbommel, The Netherlands: Van Haren Publishing.
San Francisco, A. R., Mo, E., Carlo, M., August, D., & Snow, C. (2006). The influences of language of literacy instruction and vocabulary on the spelling of Spanish–English bilinguals, Reading and Writing, 19(6), 627–642. https://doi.org/10.1007/s11145-006-9012-3
Sandström, A., & Neghina, C. (2017). English taught bachelor’s programme. Retrieved from https://www.studyportals.com/wp-content/uploads/2017/09/EAIE-StudyPortals-English-taught-bachelor-programmes-Europe.pdf
Seymour, P. H. K., Aro, M., & Erskine, J. M. (2003). Foundation literacy acquisition in European orthographies. British Journal of Psychology, 94(2), 143–174.
https://doi.org/10.1348/000712603321661859
Sipra, M. A. (2013). Impact of English Orthography on L2 Acquisition. English Language Teaching, 6(3), 116–124. https://doi.org/10.5539/elt.v6n3p116
Rose, P. (2016). Engels in het basisonderwijs: Domeinbeschrijving ten behoeve van peilingsonderzoek. Retrieved from https://slo.nl/thema/vakspecifieke-thema/mvt/engels/
32 Sparks, R. L., Patton, J., Ganschow, L., Humbach, N., & Javorsky, J. (2008). Early first-language reading and spelling skills predict later second-first-language reading and spelling skills. Journal of Educational Psychology, 100(1), 162–174. https://doi.org/10.1037/0022-0663.100.1.162
Stanovich, K. (1994). Phonological Awareness. Retrieved April 16, 2020, from http://dyslexiahelp.umich.edu/professionals/dyslexia-school/phonological-awareness
Tabors, P. O., Páez, M. M., & López, L. M. (2003). Dual language abilities of four-year-olds: Initial findings from the Early Childhood Study of Language and Literacy Development of Spanish-Speaking Children. NABE Journal of Research and Practice, 1, 70–91.
Tornéus, M. (1984). Phonological awareness and reading: A chicken and egg problem? Journal of Educational Psychology, 76(6), 1346–
1358. https://doi.org/10.1037/0022-0663.76.6.1346
Treiman, R. (1997). Introduction to special issue on spelling. Reading and Writing, 9, 315– 319 https://doi.org/10.1023/A:1007990526828
Van Berkel, A. (1987). Serious spelling problems and foreign language learning, a weak link in the training of teachers. European Journal of Teacher Education, 10(2), 195-211.
https://doi.org/10.1080/0261976870100208
Van der Sijs, N. (1996). Leenwoordenboek (1st edition). Den Haag, The Netherlands: SDU.
Van Gelderen, A., Schoonen, R., de Glopper, K., Hulstijn, J., Simis, A., Snellings, P., & Stevenson, M. (2004). Linguistic Knowledge, Processing Speed, and Metacognitive Knowledge in First- and Second-Language Reading Comprehension: A Componential Analysis. Journal of Educational Psychology, 96(1), 19–30. https://doi.org/10.1037/0022-0663.96.1.19
Van Oostendorp, M. (2012). Bilingualism versus multilingualism in the Netherlands. Language Problems and Language Planning, 36(3), 252–272.
33 Wade-Woolley, L., & Siegel, L. S. (1997). The spelling performance of ESL and native speakers of English as a function of reading skill. Reading and Writing, 9(5/6), 387– 406. https://doi.org/10.1023/A:1007994928645
Wagner, R. K., Torgesen, J. K., Rashotte, C. A., & Pearson, N. A. (1999). Comprehensive test of phonological processing: CTOPP. Austin, Texas, United States: Pro-ed.
Williams, J. M. (1986). Origins of the English Language. Amsterdam, The Netherlands: Amsterdam University Press.
Yeung, S. S. S., Siegel, L. S., & Chan, C. K. K. (2012). Effects of a phonological awareness program on English reading and spelling among Hong Kong Chinese ESL children. Reading and Writing, 26(5), 681–704. https://doi.org/10.1007/s11145-012-9383-6
Zeguers, M. H. T., Van den Boer, M., Snellings, P., & De Jong, P. F. (2018). Universal and language-specific predictors of early word reading in a foreign language: An analysis of the skills that underlie reading acquisition in three different orthographies. Developmental Psychology, 54(12), 2274–2290. https://doi.org/10.1037/dev0000606
