Reading acquisition in Dutch and Spanish children

F

S

B

S

Reading Acquisition in Dutch

and Spanish Children

Thesis 2

Yessica Ortega Luna, 5737966@student.uva.nl

Student number 5737966

January 2014

Supervisors: Prof. Dr. Peter F. de Jong

Dr. Cristina M. Rodriguez

2

Abstract

Introduction: Variation in reading acquisition might be due to two language characteristics;

orthographic depth and syllabic complexity of the language, and two underlying skills of the children themselves; phonological awareness (PA) and rapid automatized naming (RAN). The first aim of this study was to examine differences in performance of phonological awareness in Dutch and Spanish children. Based on previous studies it is expected that the experience children have with the structures in their language influences their phonological awareness. The second aim of this study was to analyse whether PA and RAN are related to the reading ability of second Grade Spanish and Dutch children. Based on previous research it was expected that PA would have a bigger influence during the first years of reading instruction than RAN. However, the exact relation remains unclear. Method: 55 Dutch and 55 Spanish children performed several PA, RAN and reading tasks. Results: The PA scores of the Spanish children were lower than the scores from the Dutch children. Complex consonant clusters were the most difficult structure to manipulate, especially for the Spanish children, whereas simple open syllables were shown to be the easiest. It was also shown that RAN non-alphanumeric was related to both word and pseudoword reading in Dutch. But neither PA nor RAN were related to reading in Spanish. Conclusion: Experience with syllable structures seems to influence children’s development of phonological awareness. Furthermore, unexpectedly, PA was not found to be related to reading ability, whereas RAN was.

3

Reading Acquisition in Dutch and Spanish Children

Yessica Ortega Luna, Peter F. de Jong and Cristina M. Rodriguez

Introduction

The acquisition of reading skills in alphabetic orthographies has been the focus of many studies (e.g. Caravolas, Volin, & Hulme, 2005; López-Escribano & Katzir, 2008). Children, in general, learn to read between 6 and 8 years old. However, the speed of acquisition and the factors that affect this acquisition might differ across children learning to read in different languages. This variation in reading acquisition might be due to at least two language factors; the orthographic depth and the syllabic complexity of the language. Two underlying skills of the children themselves might also be of influence on the variation in reading skills; phonological awareness (PA) and rapid automatized naming (RAN). Orthographic depth indicates the degree to which letters map to phonemes, contrasting alphabetic orthographies which approximate a consistent 1:1 mapping with orthographies which contain inconsistencies and complexities (Seymour, Aro, & Erskine, 2003). Most of the studies regarding orthographic depth have focused on the difference between English, which is considered a deep orthography with many inconsistencies and complexities, and other alphabetic European languages, of which most have shallow orthographies with mostly consistent grapheme-phoneme correspondences. These studies have shown that children learning a shallow language make greater gains developing their decoding skills during their first years of schooling than children learning a deep orthography (Caravolas, Volin, & Hulme, 2005; López-Escribano & Katzir, 2008).

Syllabic complexity on the other hand is less studied compared to orthographic depth, but might also play a role in the variation in reading acquisition (Sprenger-Charolles & Siegel, 1997; Taft, Álvarez, & Carreiras, 2007). Syllabic complexity refers to the distinction between

4

languages characterized by a predominance of open consonant-vowel (CV) syllables with few initial or final consonant clusters (e.g., Spanish) and the languages which have several closed consonant-vowel-consonant (CVC) syllables and complex consonant clusters in both onset and coda position (e.g., English) (Seymour et al., 2003). Research has shown that in languages with a more complex syllabic structure it is more difficult to obtain an adequate level of decoding, translating written words into spoken words, than in languages with a more simple syllabic structure (Seymour et al., 2003). This difficulty might be due to the fact that consonant clusters could make it more difficult to understand the separate grapheme-phoneme correspondences.

As mentioned, the child’s reading performance is not only influenced by the orthographic depth and syllabic complexity of a language but also by underlying cognitive processes. Two of these cognitive processes which play an important role are rapid automatized naming (RAN) and phonological awareness (PA). RAN is a task which measures how rapidly individuals can name aloud different stimuli and it can be perceived in two different ways. Firstly, it can be perceived as the ability to match orthographic information with its phonological code. And secondly, it can be perceived as the speed of retrieving information from phonological memory (Wagner & Torgesen, 1987). Phonological awareness is defined as the metacognitive capacity to reflect on and manipulate sounds in spoken words at the level of syllables, onsets and rimes, and phonemes (Caravolas, 2004). Some researchers have suggested that PA is a precondition for developing reading skills (Stahl & Murray, 1994; Durgunoğlu & Öney, 1999). The development of these reading skills, in turn, might improve the child’s PA, because children become more sensitive to the several different phonemes in words and become better in the manipulation of these phonemes. It might be that certain basic levels of phonological awareness precede learning to read, whereas other more advanced levels may be a result of reading instruction (Stahl & Murray, 1994; Durgunoğlu & Öney,

5

1999). Stahl and Murray (1994) stated that children tended to treat consonant clusters as a single unit when performing PA tasks. This might indicate that consonant clusters are more difficult to decode, leading to more decoding problems in orthographies with a more complex syllable structure.

Studies on phonological awareness showed that the growth of phonological awareness is dependent on characteristics of the language being learned (Share & Blum, 2005). In languages with a simple syllabic structure, children reach accurate levels of PA more quickly than in languages with a complex syllabic structure (Caravolas et al., 2005). Also, within the syllables children manipulate phonemes in simple syllable structures more rapidly and more accurately than phonemes in complex cluster units (Bruck & Treiman, 1990; Caravolas & Landerl, 2010). Furthermore, it has been shown that the experience children have with certain aspects of their language influences the development of phonological awareness (Durgunoğlu & Öney, 1999; Caravolas & Landerl, 2010). For example, in the study by Caravolas and Landerl (2010) Czech children had better awareness of onset phonemes than German children, whereas the German children had a better awareness of phonemes in the coda position. These advantages might be due to German children having more experience with complex codas, whereas Czech children have more experience with complex onsets. The same holds for the study by Durgunoğlu and Öney (1999) which showed that Turkish children performed better on codas than their English peers due to their greater experience with complex codas.

In Spanish, phonological awareness has been shown to have an influence on reading acquisition during the first year of reading instruction. It is also shown that PA maintains its influence throughout later years (Jiménez González & Ortiz González, 2000; Bravo-Valdivieso, Villalón, & Orellana, 2006). Also in Dutch, phonological awareness has been shown to play an important role in reading acquisition during the first years of reading

6

instruction. During this phase differences in PA can be found between normal readers and dyslexic children. Later on, these differences in PA were still found although only when the PA task was made more complex. This might indicate that PA maintains its influence in Spanish as well as in Dutch (de Jong & van der Leij, 2003). The influence of RAN on the development of reading skills seems to increase after the first year of reading instruction. But RAN alphanumeric (letters and digits) has been shown to be a better predictor of reading than RAN non-alphanumeric (colours and objects) (Lervåg, Bråten & Hulme, 2009). Some studies have shown that RAN has a stronger relationship with exception word reading than with word reading, which might indicate that RAN is of greater influence when more orthographic knowledge is required (de Jong, 2011). This might also explain why RAN is of greater influence later in de reading development.

This study aims to test for variation in phonological awareness in languages which somewhat vary in orthographic depth but differ considerably in syllabic complexity; Dutch and Spanish. To date, few to no studies have been done comparing the role of phonological awareness between two languages which are similar in transparency level but different in syllabic complexity. The central questions in this study are: (1) Does the syllable complexity of languages influence phonological awareness? (2) Are PA and RAN related to reading ability in second Grade? In neither of the languages the grapheme-phoneme correspondence is perfect; the Spanish alphabet consists of 27 letters representing 29 phonemes, whereas the Dutch alphabet consists of 26 letters representing about 40 phonemes. In Dutch, vowels can be long or short and they can also be represented by digraphs. The vowel ‘e’ is hereby the most inconsistent grapheme (de Jong, 2003). The most important difference between the two languages, however, is their syllabic complexity. The Dutch language has a complex syllabic structure because of its numerous closed syllables with a consonant-vowel-consonant (CVC) structure (e.g. ‘kap-sel’) and its complex consonant clusters in the onset and coda positions

7

(CCV and CVCC) (e.g. ‘grap’ and ‘last’). In Spanish, in contrast, there is a predominance of open CV syllables with few initial or final consonant clusters (e.g. ‘ro-pa’).

CV syllables exist in every language; however they are more prominent in some languages than in others (Sprenger-Charolles & Siegel, 1997). This leads to children in all languages being somewhat familiar to this structure. This is not the case for closed syllabic structures or syllables with consonant clusters. This difference is also visible in the Spanish and Dutch language. The predominance of CV structures in Spanish compared to Dutch might lead Spanish children to have more difficulties when manipulating closed syllables and consonant clusters (CVC and CCV) and making more errors than Dutch children during the PA tasks, because the latter are more familiar with these more complex structures. Spanish children are also expected to have more problems with CVC and CCV words during the RAN tasks. They are expected to read these words less automatized, and thus more slowly than Dutch children. Thus, the experience children have with the syllable structure of their native language might be of importance in developing phonological awareness from an early age on and this, furthermore, might influence the later alphabetic reading skills (Caravolas & Landerl, 2010). Also, studies have shown that single phonemes are easier to manipulate than phonemes in complex clusters (Bruck & Treiman, 1990; Caravolas & Landerl, 2010). This might lead all children, both Spanish and Dutch, to make more errors when manipulating syllables with a CCV structure compared to when they have to manipulate syllables with a CV or CVC structure.

Method

Participants

Participants came from several second grade classes in the Netherlands. Primary schools were approached by letter. Approximately one week after the schools received the letters they were

8

phoned to ask whether they were willing to participate. Two schools decided to participate, leading to a total of 55 Dutch Grade 2 pupils, ranging from 7 to 8 years old, taking part in the study (M = 97.11 months, SD = 3.78). Twenty of them were boys (36.4 %) and 35 were girls (63.6 %). All of the children were born in the Netherlands, however three children sometimes spoke a different language at home and one child previously attended an international school and had education in English. This child however came back to the Netherlands before starting formal reading education. The Spanish participants were selected and tested in several primary schools in Tenerife (Spain) prior to the present study.

To make the comparison between the Spanish and Dutch children more accurate the Spanish participants were matched to the Dutch participants by comparing the word reading level (measured by standardized reading tests) of the Spanish children to the word reading level of the Dutch children. Spanish children were matched to Dutch children who had the same relative word reading level within their language peer group. This lead to 55 Spanish second grade pupils participating in the study (M = 94.44 months, SD = 3.84); 25 boys (45.5 %) and 30 girls (54.5 %). All of these children spoke Spanish both at school as at home. A significant difference was found between the age of the Dutch children and the age of the Spanish children, t(108) = 3.682, p < .000. No difference in gender is found between the Spanish and Dutch children.

Both Dutch and Spanish children in this study had learned to read by phonics instruction; grapheme–phoneme correspondences were taught explicitly in first grade. Through this method children learn simple correspondences first and continue with more complex correspondences and irregularities later on.

9

Instruments

Phonological Awareness

There were four tasks for phonological awareness included in the Dutch version of an in Spain developed computer program (SICOLE-R; Jiménez et al., 2007): Isolation, Deletion, Segmentation and Blending. Due to technical problems, the Dutch blending task was implemented through E-Prime instead of SICOLE-R (Jiménez et al., 2007). The four tasks assess the participant’s ability to detect and manipulate phonemes of spoken words. Each task consisted of 15 items. These 15 items consisted of five items with a CV structure, five with a CVC structure and five with a CCV structure. The matching of the Dutch items to the Spanish items occurred by firstly selecting Dutch words from the CELEX database starting with the same syllable structure as the Spanish corresponding word (CV, CVC or CCV). Secondly, words with the same amount of syllables were selected. Then within each language database, absolute word frequencies were converted to percentile scores. Afterwards Dutch words, which had the same percentile score as the respective Spanish word, used in the Spanish task, were selected. From the remaining Dutch words the items were manually selected.

In the isolation task, the children heard a word (e.g., rugzak [backpack]) and were asked what the first sound of the word was, in this case /r/. After telling what the first sound was, the children had to choose one picture out of three which started with the same sound as the word they heard and click on it with their computer mouse (in this example riem [belt]). Cronbach’s alpha for this task was .98, which shows a good reliability. During the deletion task the children listened to a word (e.g., kano [canoe]) and were asked to delete its first sound. Then they had to say aloud which ‘word’ remained after deleting the first phoneme (in this case ‘ano’ remains after deleting /k/). This task had a Cronbach’s alpha of .97. In the third task, segmentation, the children heard a word (e.g., gebit [teeth]) and then had to say its constituent phonemes (e.g. /g/ /e/ /b/ /i/ /t/). Pronouncing the sounds or saying the names of

10

letters was assessed as a correct response. Cronbach’s alpha for this task was .98. In the

blending task the children listened to a sequence of phonemes (e.g., /f/ /o/ /t/ /o/) and were

supposed to say the word composed by these phonemes (in this example foto [photo]). The reliability was good, Cronbach’s alpha =.94. For full lists of items see Appendix I.

Reading

There were four reading tasks; two of them were included in SICOLE-R (Jiménez et al., 2007) and two were presented on paper. The items in the Dutch tasks were matched to the items presented to the Spanish children. The matching occurred according to the frequency of the words, the length of the words and the syllabic structure. Firstly, Dutch words with the same syllabic structure as the Spanish words which were used in the task were selected from the CELEX-database. From these words, a selection was made with words that were the same length as the Spanish words (i.e. two syllables or three syllables and approximately the same amount of letters). Lastly, within each language database absolute word frequencies were converted to percentiles and Dutch words with the same relative percentile scores as the respective Spanish word were selected (e.g., ‘oven’ for the Spanish word ‘huevo’ and ‘planeet’ for the Spanish word ‘plato’).

The composition of the Dutch pseudowords happened somewhat different. To make sure children would not recognize existing Dutch words in the pseudowords, Dutch words with a low frequency in children’s vocabulary were selected (i.e. frequency 5 – 20; percentile 70 – 85 approximately). The two syllable words had five and six sounds as the Spanish stimuli, and the three syllable words contained seven sounds. To create the pseudowords one letter was changed in the two syllable words with five sounds, and two letters were replaced in the two syllable words with six sounds and in the three syllable words (e.g., ‘imker’ [beekeeper] became ‘isker’, ‘reptiel’ [reptile] became ‘peptief’ and ‘erfenis’ [heritage]

11

became ‘ursenis’). Furthermore, 24 of the pseudowords in the Spanish test battery showed structures which were possible in Dutch and these pseudowords were maintained in the Dutch version (e.g., ‘polton’). In total, there were 32 words and 48 pseudowords (see Appendix II). These items were all administered in one session.

The word and pseudoword reading tasks started with a few practice items. After these practice items the first item appeared, setting in motion the chronometer, which was stopped as soon as the participant started pronouncing the word. After registering the latency time, the second item appeared on the screen. The sequencing in the administration of the items was as follows: blank screen on the computer (200 ms), fixation point (+) in the center of the screen followed after 400 ms by the stimulus word or pseudoword. The speed measurements were calculated from the moment the item appeared on the screen until the moment the vocal key captured the first sound pronounced by the child. For the speed measurements all items were included, whether the child pronounced the item correctly or not. The reliabilities for accuracy and speed of the word task were good, Cronbach’s α = .93 and .99 respectively. The same held for the pseudowords task, Cronbach’s α = .93 and .87 respectively. On the basis of the speed and accuracy measurements a fluency measurement was computed by calculating the amount of words or pseudowords correctly read per second. Since the reliability of both the speed and accuracy measurements were good, it might be assumed that the reliability of the fluency measurements was also good.

A second list with pseudowords was presented to the children, the Klepel (Van den Bos, Lutje Spelberg, Scheepstra & De Vries, 1994). The children were asked to read aloud a list of 116 pseudowords of increasing difficulty during two minutes. They had to do this as quickly as possible while trying to make no errors. The reliability of this instrument was good, Cronbach’s α = .89 - .95. A similar list with words was also presented to the children, the One-Minute Reading Test (EMT; Brus & Voeten, 1979). The children were asked to read

12

aloud a list of 116 words as quickly as they can without making errors within one minute. The reliability of this instrument was good, Cronbach’s α = .82 - .92.

Naming Speed

The naming speed task is also included in SICOLE-R (Jiménez et al., 2007). The participant was required to name, as quickly as possible, four series of symbols (i.e., letters, numbers, colours and objects). Stimuli consisted of five letters, five one-digit numbers, five basic colours and five objects. These five stimuli were repeated 10 times on the screen and distributed in 5 rows and 10 columns. Both accuracy and time for each task were taken as dependent variables.

Covariates (Vocabulary and Non-Verbal Intelligence)

Two tasks were administered to assess two aspects which might influence the performance on the reading tasks; vocabulary and non-verbal intelligence. Vocabulary was assessed through the Revisie Amsterdamse Kinder Intelligentie Test (RAKIT; Bleichrodt, Drenth, Zaal & Resing, 1987). The children were shown four pictures. Afterwards they heard a word and they were asked to pick the picture corresponding with this word. The reliability of this instrument is good, conform the COTAN norms. To measure the non-verbal intelligence Raven’s Progressive Matrices are used (Raven et al., 1986). The children had to complete item 1 to 36, which consisted of 36 different big drawings on a page, with one piece missing, as it were a puzzle. Underneath the drawing six or eight different pieces were depicted. Children were asked to choose the piece which would complete the greater pattern. The reliability of this instrument is sufficient, conform the COTAN norms.

13

Procedure

First, the children had to complete the RAVEN’s test to assess non-verbal intelligence (Raven et al., 1986). This test was administered to the entire class in one session consisting of 45 minutes. Subsequently, the children had to complete a word reading task (EMT; Brus & Voeten, 1979), a vocabulary task (Rakit; Bleichrodt, Drenth, Zaal & Resing, 1987) and a non-word reading task (Klepel; Van den Bos et al., 1994) in this order. These were administered individually during a 15-minute session outside the child’s classroom. When all children finished these tasks, a second individual session started. During this session SICOLE-R (Jiménez et al., 2007) was used to test rapid automatized naming and phonological awareness. The children were allowed to choose in which order they wanted to complete the different tasks in the program. Lastly, the children had to complete a phonological awareness task with E-Prime. The children completed the tasks on the computer in approximately 30 minutes.

Analyses

A Language (Dutch vs Spanish) by Structure (CV vs CVC vs CCV) ANCOVA was performed to study whether syllable complexity influences the performance outcomes of the PA tasks in the Spanish and Dutch children. Language was the between-subjects factor and syllable structure was the within-subject factor.

To answer the question whether PA and RAN are related to reading ability in second Grade several forward regression analyses were performed. The fluency scores of words and pseudowords were entered as dependent variables, while the groups, age, PA and RAN tasks were entered as the independent variables.

14

Results

Preliminary analysis showed that almost all time measurements of the word and pseudoword reading tasks had more than 5% missing data. Little’s MCAR test was non-significant, meaning that the missing values were missing completely at random, thus the missing data were imputed through the expectation-maximization method (EM). Normality was not met for most of the variables. Neither square root transformation nor logarithm transformation of the data led to improvement of normality, thus the original data were used. Furthermore, a ceiling effect was found for the Dutch scores on the isolation, segmentation, and deletion tasks, on the word reading task and on the CV and CVC structures.

Influence of syllable structure on PA performance

A Language by Structure ANCOVA was performed to study whether the performance outcomes of the PA tasks in the Spanish and Dutch children were influenced by the syllable structure of the language. Language was the between-subjects factor and syllable structure was the within-subjects factor. To control for differences in age, this factor was included as a covariate. Table 1 shows the descriptive statistics for the Spanish and Dutch children on phonological awareness across the different tasks and syllable structures.

The analysis showed that there was a significant main effect for language, F(1, 108) = 143.980, p < .001, partial η2 = .571. The Dutch children performed better than the Spanish children on the PA tasks across all the syllable structures. The syllable structures also differed from each other, F(2, 216) = 56.878, p < .001, partial η2 = .345. The CV structure was the easiest structure to manipulate for all the children, whereas CCV structures were the hardest to manipulate. Also, an interaction effect between language and syllable structure was found,

F(2, 216) = 16.648, p < .001, partial η2 = .134. This indicates that the difference between the scores on the separate syllable structures is smaller for the Dutch children than for the Spanish

15

children; CCV structures are much more difficult than CV structures for Spanish children, whereas the difference in difficulty between these two types of structures is small for the Dutch children. Furthermore, it indicates that the difference in scores between the Spanish and the Dutch children is smaller on the CV structures than on the CCV structures, which was expected since Spanish children have less experience with CCV structures than Dutch children do (see Figure 1).

To further study the hypotheses follow-up contrasts were specified. Firstly, it was hypothesized that CV structures were the easiest and CCV structures were the hardest to manipulate. Additionally, it was hypothesized that the difference between CV and CCV structures would be bigger for Spanish children than for Dutch children, since Dutch children have experience with both of the structures whereas the Spanish children hardly have any experience with CCV structures. One contrast compared the CV and CVC structures to the CCV structure. In the other contrast CV structures were compared with CVC structures. Firstly, it was analysed whether there are interactions between structure and language contrasting the two before mentioned contrasts for Dutch children and Spanish children. Both contrasts were significant, F(1, 108) = 20.952, p < .001, partial η2 = .162 and F(1, 108) = 11.868, p = .001, partial η2 = .099. The difference between CV + CVC and CCV was greater for Spanish children than for Dutch children, and so was the difference between CV and CVC (see Figure 1). Because interaction effects were found for both contrasts, separate analyses were conducted in both languages. Contrasting CV and CVC against CCV a significant main effect was found in both Dutch and Spanish children, F(1, 54) = 14.538, p < .001, partial η2 = .212 and F(1, 54) = 85.809, p < .001, partial η2 = .614 respectively. Manipulating words starting with a CCV structure was harder for both Spanish and Dutch children compared to manipulating words which started with a CV or a CVC structure. For the second contrast the analysis showed that in Spanish CV structures were significantly easier to work with than

16

words with CVC structures, F(1, 54) = 24.374, p < .001, partial η2 = .311. In Dutch CV and CVC structures did not differ in difficulty, F(1, 54) = .611, p = .438, partial η2 = .011.

Figure 1 Mean PA Scores on the Syllable Structures for Dutch and Spanish Children

10 11 12 13 14 15 16 17 18 19 CV CVC CCV Dutch Spanish

17

Table 1

Means and Standard Deviations on PA across Syllable Structures and PA Tasks for Dutch and Spanish children

Dutch Spanish CV CVC CCV CV CVC CCV PA Task M SD M SD M SD M SD M SD M SD Isolation 4.58 .96 4.67 .58 4.67 .72 4.31 .98 4.20 1.16 3.47 1.40 Deletion 4.78 .57 4.87 .34 3.62 1.57 4.33 1.26 4.04 1.30 2.80 1.82 Segmentation 4.65 .70 4.64 .80 4.75 .58 4.33 .86 3.89 1.20 3.78 1.15 Blending 4.17 .82 3.78 1.18 4.00 1.10 1.51 1.35 .55 .92 .56 .96 Total 18.11 1.91 17.89 1.96 16.96 2.65 14.47 2.75 12.67 3.08 10.62 3.18

18

Relation PA and RAN with reading ability in second Grade

Table 2 shows the descriptive statistics for the Spanish and Dutch children on PA, RAN, word reading and pseudoword reading. The accuracy measurements for RAN are left out of the table, since hardly any errors were made. The PA scores in the table consist of the sum of the separate PA scores on the different tasks. This sum score for PA is computed to make interpretation more clear, since it is hard to interpret the relation of every separate PA task on reading. This sum score had a non-normal distribution; a square root transformation improved the distribution after converting the scores from negative scores to positive scores (60 – PA score). For further analyses the square root transformation on PA was used. Also, item 12 of the word reading task was eliminated from the analysis due to a difference in number of syllables between the Dutch and Spanish item. Also, when analysing the pseudowords only the first 24 items were taken into account. These items were equal for both Spanish as well as Dutch children, making a better comparison of the scores for children from both countries possible. Furthermore, instead of using the speed measurements in the analyses the fluency measurement was used.

In Table 2 it can be seen that Spanish children scored worse on PA than Dutch children did. No differences were found between Spanish and Dutch children on RAN, except for RAN letters, where Spanish children were slower than Dutch children. Furthermore, Spanish children scored better on word and pseudoword accuracy, but worse on word and pseudoword reading fluency. This means that the Spanish children read less words and pseudowords correctly per minute than the Dutch children did.

19

Table 2

Means and Standard Deviations on PA, RAN, Word Reading and Pseudoword Reading for Dutch and Spanish children

Dutch Spanish

M SD M SD t

PA 52.964 5.378 37.764 7.703 11.999**

RAN Numbers Time (s) 33.711 7.673 34.090 7.560 -.260

RAN Letters Time (s) 34.518 7.478 39.942 11.166 -2.993**

RAN Colours Time (s) 57.978 14.190 56.846 12.903 .438

RAN Objects Time (s) 61.808 13.274 56.898 13.683 1.910

RAN alphanumeric 1.532 .296 1.443 .334 1.479 RAN non-alphanumeric .878 .188 .929 .211 -1.336 Reading Words Accuracy 28.855 2.563 30.018 1.810 -2.750** Words Fluency 1.078 .309 .658 .209 8.354** Pseudowords Accuracy 19.200 3.051 22.146 2.022 -5.967** Pseudowords Fluency .899 .335 .613 .179 5.596**

Note: PA max = 60; Words accuracy max = 31; Pseudowords accuracy max = 24 *p < .05; ** p < .01

The time measurements of RAN are significantly correlated with each other; especially the correlation between letters and numbers (r = .781 for Dutch and r = .705 for Spanish) and the correlation between objects and colours is high (r = .816 for Dutch and r = .614 for Spanish). This led to the decision to make a distinction between alphanumeric RAN (letters and numbers) and non-alphanumeric (colours and objects) RAN and using those scores in the regression analysis instead of using all the separate measurements. This division is used more often in studies involving RAN (e.g., Lervåg, Bråten & Hulme, 2009). Furthermore it was decided to not include the accuracy measurements of RAN since almost all the correlations with the reading tasks were non-significant and hardly any errors were made.

Table 3 shows the correlations for the sum score of phonological awareness, RAN alphanumeric, RAN non-alphanumeric and the (pseudo)word reading tasks. The table shows

20

that in the Dutch language PA is not correlated with word reading nor with pseudoword reading whereas in Spanish PA is significantly correlated to word reading. However, in Dutch all the RAN measurements are significantly correlated to word and pseudoword reading, whereas this is not the case for Spanish children; both RAN alphanumeric and RAN non-alphanumeric are significantly correlated to word reading, but not to pseudoword reading. This might indicate that in Dutch more orthographic knowledge is required, whereas Spanish children can rely on their letter per letter decoding.

Table 3

Correlations of PA, RAN, Word and Pseudoword Reading Fluency for Dutch and Spanish Children

Variables Word Pseudowords PA RAN alpha RAN non-alpha

Word .707** -.230* .337* .288*

Pseudowords .616** -.166 .166 .157

PA -.167 -.009 -.226 -.141

RAN alpha .386** .410** -.130 .599**

RAN non-alpha .480** .486** -.317* .676**

Note: Below the diagonal are the Dutch children; above the diagonal are the Spanish children;

PA = square root (60 – original PA score)

*p < .05; ** p < .01

To assess whether phonological awareness and RAN are significantly related to the variability in reading ability for second Grade Dutch and Spanish children several regression analyses were performed. Table 4 shows the standardized regression coefficients (β) of the final models, and the R2 changes of the word reading models for Dutch and Spanish children

separately as well as Spanish and Dutch children combined. Firstly, word reading fluency for Dutch children was entered as the dependent variable. Age was included in the equation as a covariate. After checking whether age was significant, PA was added, since PA is supposed to

21

have a greater influence on reading acquisition during the first years of reading instruction. Next, RAN alphanumeric was entered. The last variable being entered was RAN non-alphanumeric. To assess whether the unique contribution to the variation being explained by RAN non-alphanumeric is significant without RAN alphanumeric in the equation the last two steps were reversed; thus after PA was entered in the equation, first RAN non-alphanumeric was included and after it was checked whether RAN non-alphanumeric added significantly to the equation RAN alphanumeric was entered. The adjusted R2 value of .183 of the final Dutch model shows that over 18% of the variability in word reading ability is predicted by age, PA, RAN alphanumeric and RAN alphanumeric. However, in the final model only RAN non-alphanumeric had a significant unique contribution in the explanation of the variability in

word reading fluency.

The same steps were repeated for the Spanish children. The adjusted R2 value of .102 of the final Spanish model shows that around 10% of the variability in word reading ability is predicted by age, PA, RAN alphanumeric, and RAN non-alphanumeric. However, in the final model none of the predictors had a significant unique contribution in the explanation of the variability in word reading fluency.

After both single regression analyses were performed a regression analysis was performed for both Dutch and Spanish children combined. Again age was the first variable being entered in the equation, followed by language. The third variable being entered was PA, as in the single analyses. From the single analyses it could be concluded that only RAN non-alphanumeric had a significant unique contribution to the explanation of the variation in word reading in Dutch whereas alphanumeric RAN did not have a unique contribution in neither Spanish nor Dutch. This led to the decision to only include this variable in the combined analysis. RAN non-alphanumeric was thus the fourth variable being entered in the equation. Besides the main effects interaction effects between the predictors and language were also

22

included in the equation. The first interaction being added was PA x language. The last step added the interaction of RAN non-alphanumeric with language to the equation. The adjusted R2 value of .490 of the final combined model shows that almost half of the variability in word reading ability is predicted by age, language, PA, RAN alphanumeric, and RAN alphanumeric x language. However, in the final model only language, RAN non-alphanumeric and RAN non-non-alphanumeric x language had a significant unique contribution in the explanation of the variability in word reading fluency. The RAN non-alphanumeric by language interaction indicated that RAN non-alphanumeric was less related to word reading for Spanish children than for Dutch children.

The same steps were followed for pseudoword reading. Table 5 shows the standardized regression coefficients (β) of the final models, and the R2 changes of the pseudoword reading models for Dutch and Spanish children separately as well as Spanish and Dutch children combined. Pseudoword reading fluency was entered as the dependent variable. First a single regression analysis was performed for Dutch children. The adjusted R2 value of .209 of the final Dutch model shows that 20.9% of the variability in pseudoword reading ability is predicted by age, RAN alphanumeric and RAN non-alphanumeric. However, in the final model only RAN non-alphanumeric had a significant unique contribution in the explanation of the variability in pseudoword reading fluency; age and RAN alphanumeric did not anymore.

Secondly, a single analysis for Spanish children was performed. The adjusted R2 value of .059 of the final model shows that almost 6% of the variability in pseudoword reading ability is predicted by age, PA, RAN alphanumeric and RAN non-alphanumeric. However, in the final model only age had a significant unique contribution in the explanation of the variability.

23

children. The single analyses showed that in Dutch RAN non-alphanumeric maintains its unique contribution whereas in Spanish neither RAN alphanumeric nor RAN non-alphanumeric had a significant unique contribution to the explanation of the variability in pseudoword reading. This lead to the decision to only include RAN non-alphanumeric in the combined analysis, making it the fourth variable in the equation. The adjusted R2 value of .359 shows that over 35% of the variability in pseudoword reading ability is predicted by age, language, PA, RAN non-alphanumeric, PA x language and RAN non-alphanumeric x language. However, in the final model only language, RAN alphanumeric and RAN non-alphanumeric x language had a significant unique contribution in the explanation of the variability in pseudoword reading fluency. Again, the RAN non-alphanumeric by language interaction indicated that RAN non-alphanumeric was less related to pseudoword reading for Spanish children than for the Dutch children.

24

Table 4

Summary of Regression Analysis for Variables Predicting Word Reading Fluency in Second Grade

Dutch Spanish Dutch + Spanish

Steps Variables ∆ R2 β _{∆ R}2 β _{∆ R}2 β 1 Age .012 .076 .009 .140 .081** .085 2 Language .318** -.594** 3 PA .030 -.034 .068 -.197 .026* -.089 4 RAN alpha .128** .114 .081* .217 5 RAN non-alpha .073* .386* .010 .127 .071** .303** 4 RAN non-alpha .194** .386* .062 .127 5 RAN alpha .007 .114 .029 .217 6 PA x Language .000 -.055

7 RAN non-alpha x Language .021* -.153*

R2 (Adjusted R2) .243 (.182) .168 (.102) .518 (.490)

Note: Language, PA and RAN were centered for the interaction effects; ∆ R2 = change in R2 after addition of the variable; -PA = square root (60 – original PA score)-

25

Table 5

Summary of Regression Analysis for Variables Predicting Pseudoword Reading Fluency in Second Grade

Dutch Spanish Dutch + Spanish

Steps Variables ∆ R2 β _{∆ R}2 β _{∆ R}2 β 1 Age .001 -.019 .059 .290* .059** .080 2 Language .174** -.513** 3 PA .000 .151 .053 -.206 .005 .044 4 RAN alpha .169** .128 .014 .075 5 RAN non-alpha .098* .448* .004 .078 .082** .348** 4 RAN non-alpha .258** .448* .014 .078 5 RAN alpha .009 .128 .003 .075 6 PA x Language .005 -.139

7 RAN non-alpha x Language .069** -.273**

R2 (Adjusted R2) .268 (.209) .129 (.059) .394 (.359)

Note: Language, PA and RAN were centered for the interaction effects; ∆ R2 = change in R2 after addition of the variable; PA = square root (60 – original PA score)

26

Discussion

The focus of this study was on the differences in reading acquisition between Spanish and Dutch children. Firstly, it was assessed whether phonological awareness was influenced by the syllabic complexity of a language. All children were expected to have more difficulties with the most difficult CCV structure than with the easier CV and CVC structures. Additionally it was expected that the difference between the CV and CCV structures would be bigger for Spanish children than for Dutch children, since the Dutch children have some experience with all the syllabic structures whereas Spanish children do not have any experience with the difficult CCV structure. The results showed that the Dutch children outperformed the Spanish children on the three different structures being studied (CV, CVC and CCV). The Dutch children performed especially better than the Spanish children on words with complex consonant clusters (CCV). It is possible that the Dutch children performed better because of their greater experience with more complex structures compared to Spanish children, making it easier for them to manipulate words with these complex syllables as well as other simpler structures (Durgunoğlu & Öney, 1999; Caravolas & Landerl, 2010). This effect was visible across all the different phonological manipulations used in this study.

Support was found for the hypotheses, however, a ceiling effect was found for Dutch children, making it hard to interpret the interaction effect found in the analyses. It might be that, if it would have been possible to include more difficult items, the Dutch children would have had higher scores; thereby the interaction effect might disappear and experience would not seem to have an influence since CCV is not clearly more difficult than CV compared to the Dutch children. However, in that case the Spanish children still perform worse than the Dutch children, what might be due to differences in reading instruction; Dutch children somewhat exercise their phonological awareness during reading instruction. In Spain, on the

27

other hand, for as far as we know this is not the case. But, is it possible to make the items difficult enough for the Dutch children to not reach ceiling anymore without harming the measurement of phonological awareness? For example, the items might be more difficult by using spoonerisms; however, it might be that these spoonerisms measure another more abstract aspect instead of phonological awareness.

However, although some of the results should be interpreted with caution, several findings were clear. Firstly, words with a CCV structure were shown to be the most difficult to manipulate for both Spanish as well as Dutch children, and CV structures were the easiest for both groups of children, as was expected according to previous studies (Bruck & Treiman, 1990; Caravolas & Landerl, 2010). This effect of syllabic structure was especially clear for Spanish children; the difference in scores between the Spanish and the Dutch children was smaller on the CV structures than on the CCV structures. This effect of the syllables on phonological awareness was also visible in a previous study by Jiménez González and Ortiz González (2000). Furthermore, this effect was expected since Spanish children have less experience with CCV structures than Dutch children do, and both Spanish and Dutch children have some experience with CV structures (Durgunoğlu & Öney, 1999; Caravolas & Landerl, 2010). Also, results showed that the Spanish children performed clearly worse on phonological awareness than Dutch children across all structures. Lastly, it was found that Spanish children had much more difficulties with the blending task compared to the other three tasks (isolation, deletion, and segmentation). This might be because in Spain children do not learn to blend phonemes to compose a word, whereas in the Netherlands children do blend phonemes in class as a method to read words.

The second aspect of reading acquisition being assessed in this study was whether PA and RAN were related to reading ability in second Grade. Based on former studies it was expected that both PA and RAN would be related to reading, since PA influences reading

28

from the beginning of the reading instruction and the influence of RAN increases after the first year of instruction. Firstly, analysis showed that Spanish children scored better on reading accuracy for words and pseudowords. Surprisingly, however, they scored lower on reading fluency for both words and pseudowords, meaning that Spanish children read less words correctly within one minute compared to Dutch children. Additionally, it was shown that Spanish children scored a lot worse on PA than the Dutch children, which was expected since the previous analysis conducted in this study showed that the Spanish children scored lower on all PA tasks. On RAN, however, Spanish and Dutch children did not differ from each other, meaning that all children performed the task equally fast. Only on the RAN letters task Dutch children were somewhat faster than Spanish children.

Also, analysis showed that only RAN non-alphanumeric was related to word reading in second Grade for Dutch children, whereas in Spanish neither PA nor RAN seemed to be related. When taking all children into account, language, RAN non-alphanumeric and the interaction of RAN non-alphanumeric by language influenced word reading in second Grade. The same results were found for pseudoword reading in second Grade. This was somewhat unexpected since PA has been regularly found to be related to reading during the starting years of reading instruction (de Jong & van der Leij, 2003; Di Filippo et al., 2005; Jiménez González & Ortiz González, 2000; Bravo-Valdivieso, Villalón & Orellana, 2006). On the other hand, it was expected that RAN would be of influence in reading after the first year of reading instruction, when possibly more orthographic knowledge would be needed (de Jong, 2011). Though, it was slightly unexpected that RAN non-alphanumeric was related to reading whereas RAN alphanumeric was not, since reading instruction involves alphanumeric symbols and not objects or colors. Besides, previous studies have shown RAN alphanumeric to be a better predictor of reading than RAN non-alphanumeric (Lervåg, Bråten & Hulme, 2009).

29

In sum, the syllabic structure of a language influences phonological awareness because experience with more difficult syllable structures seems to help children to carry out PA manipulations correctly on more complex structures. Furthermore, it seems as though PA is not related to reading ability in second Grade Spanish and Dutch children, whereas RAN is related. To further investigate the development of the influence of PA and RAN on reading ability a longitudinal study would be advisable. The present study extends previous studies (e.g., Caravolas & Landerl, 2010) by focusing on languages similar in transparency but different in syllabic complexity and forms a good basis to assess the influence of syllabic structures on the reading acquisition of children. Furthermore, three important results should be kept in mind. Firstly, the experience with the structures in their language influences the development of phonological awareness by children. Secondly, Spanish children perform worse on all tasks and across all syllable structures. Lastly, an unexpected result was found; where it was expected that PA and RAN alphanumeric would be related to reading, they did not. RAN non-alphanumeric, nevertheless, did relate to reading, while this was not expected. However, more research should be done across different languages similar on transparency but different on syllabic structure to see whether these results are replicated.

30

References

Bravo-Valdivieso, L., Villalón, M. & Orellana, E. (2006). Predicitibilidad del rendimiento en la lectura: una investigación de seguimiento entre primer y tercer año. Revista

Latinoamericana de Psicología, 38(1), 9 – 20.

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, 156 – 178.

Caravolas, M. (2004). Spelling development in alphabetic writing systems: A cross-linguistic perspective. European Psychologist, 9(1), 3 – 14.

Caravolas, M., Volin, J. & Hulme, C. (2005). Phoneme awareness is a key component of alphabetic literacy skills in consistent and inconsistent orthographies: Evidence from Czech and English children. Journal of Experimental Child Psychology, 92, 107 – 139.

Caravolas, M. & Landerl, K. (2010). The influences of syllable structure and reading ability on the development op phoneme awareness: A longitudinal, cross-linguistic study.

Scientific Studies of Reading, 14(5), 464 – 484.

De Jong, P. F. & Van der Leij, A. (2003). Developmental changes in the manifestation of a phonological deficit in dyslexic children learning to read a regular orthography.

Journal of Educational Psychology, 95(1), 22 – 40.

De Jong, P. F. (2003). Problems in the acquisition of fluent word decoding in Dutch children. In N. Goulandris (Ed.), Dyslexia in different languages: Cross-linguistic comparisons (pp. 33–52). London: Whurr Publishers Ltd

De Jong, P. F. (2011). What discrete and serial rapid automatized naming can reveal about reading. Scientific Studies of Reading, 15(4), 314 – 337.

Durgunoğlu, A. Y. & Öney, B. (1999). A cross-linguistic comparison of phonological awareness and word recognition. Reading and Writing: An Interdisciplinary Journal,

31

Jiménez González, J. E. & Ortiz González, M. R. (2000). Metalinguistic awareness and reading acquisition in the Spanish language. The Spanish Journal of Psychology, 3(1), 37 – 46.

Lervåg, A., Bråten, I. & Hulme, C. (2009). The cognitive and linguistic foundations of early reading development: A Norwegian latent variable longitudinal study. Developmental

Psychology, 45(3), 764 – 781.

López-Escribano, C. & Katzir, T. (2008). Are phonological processes separate from the processes underlying naming speed in a shallow orthography? Electronic Journal of

Research in Educational Psychology, 16(6), 641 – 666.

Seymour, P. H. K., Aro, M. & Erskine, J. M. (2003). Foundation literacy acquisition in European orthographies, British Journal of Psychology, 94, 143 – 174.

Share, D. L. & Blum, P. (2005). Syllable splitting in literate and preliterate Hebrew speakers: onsets and rimes or bodies and codas? Journal of Experimental Child Psychology, 92, 182 – 202.

Sprenger-Charolles, L. & Siegel, L. S. (1997). A longitudinal study of the effects of syllabic structure on the development of reading and spelling skills in French. Applied

Psycholinguistics, 18, 485 – 505.

Stahl, S. A. & Murray, B.A. (1994). Defining phonological awareness and its relationship to early reading. Journal of Educational Psychology, 86(2), 221 – 234.

Taft, M., Álvarez, C. J. & Carreiras, M. (2007). Cross-language differences in the use of internal orthographic structure when reading polysyllabic words. The Mental Lexicon,

2(1), 49 – 63.

Wagner, R. K. & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101(2), 192 – 212.

32

Appendix I

PA Tasks Table AI.1

Items of the Deletion Task

Spanish Dutch Example items lata kano fresa stoffer toro kassa faro tonijn mojo hakker pupa lippen sello mossel marco machtig sexto walvis rastro zondag futbol parfum falda deksel trigo truffel crema greppel fruta brommer claro stoppen flaco knikker Table AI.2

Items of the Isolation Task

Spanish Dutch Example items sofa paleis faro zebra sopa beleg tela kachel pila wekker mono pakket foca puree lanza sandaal susto rugzak lente mixer nalga visboer rostro dokter blusa spannend frito trapper crema slinger frase prettig pluma grappig

33

Table AI.3

Items of the Segmentation Task

Spanish Dutch Examples casa meneer sopa kameel saco monnik mesa koffie tiza mammoet palo gebit foto lepel gorda fontein barba balkon disco potlood mosca circus selva dertien trece stoffig fresa blubber cromo kwekken traje zwemmen clase vliegen Table AI.4

Items of the Blending Task

Spanish Dutch Examples sofá toren foca servies beso visser sopa banaan nido ballon vino koffie torre nagel muslo wasbak tarta balpen palma vergif mundo donker doctor gordijn plano klassiek bruja knuppel frase prettig plato trommel flecha stallen

34

Appendix II Words and Pseudowords

Table AII.1

Items of the Word reading task on SICOLE-R

Spanish Dutch Examples bola tijger bravo kikker arroz alarm boda boter cama kamer comer koken gato koffer ojo agent patio lokaal plato planeet árbol ingang cine zomer fuego vijand grapa* salaris* huevo oven jugar gillen largo rustig leche letter adelante tegenwoordig amarilla ademhalen apellidos meteoriet camiseta overstroming divertida limonade plastilina begonia bolígrafo beschermeling nochebuena rivierwater habitación betekenis ascensor afbeelding descalzo enkelvoud funcionar toveren merienda muzikant servicios zaterdag lágrimas bewoner abecedario antropologe

35

Table AII.2

Items of the Pseudoword reading task on SICOLE-R

Spanish Dutch Examples escani mazetoe tonte tonte redas redas nate sunos proce alnes pona seron esco indos sunos lasda alnes losmo seron ritgo indos vendor delce golmar lasda troros losmo genmor vendor palchos golmar polton noslla tesgro troros brufas genmor jomanto palchos delnico polton protuto ritgo codidas tesgro setudad dulle unsiles brufas inbiles lartia portuto pomacos ravik sucires sumor jomanto alkaar delnico peptief bocueto sapon protuto isker socanos lagon codidas loemer setudad ralop unsiles vrielen inbiles halant portuto tribuur renpertal parlies talgunbros gordaat linsosrial bantor mestruyen tamlier biocamcir jagelaar barcurcaz daloriet

36 puertindor palaren benmacer povenig choflegio komerij berciclas siplota dosglubis ursenis dengelio indoeren