Language Development

Project Report

Sun, H. (2020). Animated electronic storybook and children's Mother Tongue development: Tracing the process and the outcome with eye-tracking (Report No. OER 13/16 HS). National Institute of Education (Singapore), Office of Education Research. https://doi.org/10.32658/10497/22664 

Project Publications

Sun, H., Loh, J., & Charles Roberts, A. (2019). Motion and Sound in Animated Storybooks for Preschoolers’ Visual Attention and Mandarin Language Learning: An Eye-Tracking Study With Bilingual Children. AERA Open, 5(2). https://doi.org/10.1177/2332858419848431 (Original work published 2019)  

Sun, H., Roberts, A. C., & Bus, A. (2022). Bilingual children’s visual attention while reading digital picture books and story retelling. Journal of Experimental Child Psychology, 215, 105327. 

Research Brief

Sun, He. (2019). Animated Electronic Storybook and Mandarin Learning (NIE Research Brief Series No. 19-031). Singapore: National Institute of Education. https://singteach.nie.edu.sg/wp-content/uploads/2021/07/NIE_research_brief_19-031.pdf 

Knowledge Resource Bank Article

https://singteach.nie.edu.sg/2021/07/19/e-storybooks/ 

Related Publication

Sun, H. Tan, J., & Feng, L. (2024). Proficiency and Compensatory Strategies in Bilingual Children’s Mandarin Narratives. Reading and Writing. https://doi:10.1007/s11145-024-10541-8  

Collaborators

• Prof Kerry LEE, formerly of NIE
• Dr Roberto SANTIAGO de Roock, formerly of NIE
• Prof CHEUNG Wai Leung Mike, National University of Singapore (NUS)
• A/P CHEONG Siew Ann, Nanyang Technological University (NTU)

Research Assistants

• Ms Yvonne MOH Cui Yun, CRCD, NIE
• Ms LOH Jie Ying, CRCD, NIE

Introduction

Leisure reading refers to reading that children themselves choose to undertake as opposed to reading that is mandated for school or homework tasks (Mellon 1990). Leisure reading has been consistently shown to be closely associated with children's success during school years and beyond (Iyengar & Ball, 2007; Sullivan & Brown, 2015). In light of the proven benefits of leisure reading, language curricula in many education systems, including Singapore, are paying increasing attention to nurture children's love for reading, and large-scale national surveys have been carried out to understand how children practise and perceive reading. However, little research has examined bilingual children's leisure reading in their two languages concurrently. The bulk of extant research fail to take account of the potential heterogeneity of participants' language backgrounds, instead focusing exclusively on English or on the schooling language of the research setting (e.g., Iyengar & Ball, 2007; Loh & Sun, 2018; Rutherford et al., 2018). It is thus of theoretical and practical interest to explore children's reading habits and preferences in bi-/multilingual contexts to better understand how to develop reading curricula, programmes, and instructional methods to promote reading in more than one language.

To fill this gap, an earlier project conducted a bilingual reading survey to explore the reading habits and preferences of 499 English-Chinese bilingual children and 199 English-Malay bilingual children. While the study developed and piloted a bilingual reading survey, it only included two primary schools, failed to include English-Tamil bilingual children and did not include qualitative data, which has been consistently shown to be valuable in further explaining and elaborating quantitative results (Creswell, 2012; Rutherford et al., 2018). To address these issues, the present project seeks to build upon and scale up from the preceding project to better understand reading habits and preferences of bilingual children in Singapore.

 In addition, the project was disrupted by the COViD-19 pandemic during data collection stage in 2020. Given that research on summer reading loss has repeatedly shown that long-term school absence, such as summer school holidays, may result in a lack of reading and a wider reading achievement gap (Alexander et al., 2007; Allington & McGill-Franzen, 2017), the project was adapted to include a section on the changes of children's leisure reading habits and preferences during the lockdown.              

 Research Questions

[RQ1] What are the reading habits and preferences of Singaporean bilingual children in English and in their respective MT languages?

[RQ2] Within each language, what are their reading habits and preferences by age?

[RQ3] Within each language, what are their reading habits and preferences by gender?

[RQ4] Are there any cross-linguistic relationships between children's reading habits and preferences in English and in their respective MT languages?

[RQ5] What are the changes in children's bilingual reading habits and preferences before and during the COVID-19 school closure? 

Participants  

- Quantitative Strand: 4326 Primary Three-Five (P3-P5, aged 9–11) from eight government primary schools in Singapore (31.5% are P3, 32.7% are P4, and 35.8% are P5)

- Qualitative Strand: 36 Primary Four and 36 Primary Five children from 1 primary school. Selection aimed to include a spread of varied reading proficiencies, and gender and Mother Tongue (MT) groups were evenly distributed. FGDs were conducted in school and audio recorded, with each group consisting of four children, and each session lasted around 30-40 minutes."          

Findings

[RQ1] Participants reported enjoying reading books in English more than in their MT languages. Furthermore, results from FGD provided qualitative evidence for the English-MT leisure reading gap. Most of the FGD children shared that they enjoyed reading in English more and read more in English, with only six out of 72 children reading in both languages equally often. Only four children stated that they preferred reading in their MT. The reasons for their reading preferences mostly related to language proficiency. They also had uneven access to English and MT reading materials at home, resulting in more frequent and longer reading sessions in English compared to MT. In addition to having fewer MT books than English books at home, there were qualitative differences in children's access to reading resources in their two languages. Regardless of their preferences for English or MT, the FGD participants showed more familiarity with English books. When asked to list book titles they had recently read, the 72 children mentioned more than 100 different English titles and 25 authors. By contrast, only five different popular Chinese children's books were reported, and Malay and Tamil children's books were restricted to collections of moral stories. Such quantitative and qualitative disparities in accessing reading materials may impede children's MT leisure reading. The participating children preferred reading printed books than e-books/online materials. Children read printed books more often and for longer duration across languages than e-books/online materials. When explaining their preference, FGD participants shared that print books were more convenient to bring to school and read from than a computer (most of them were not allowed to carry devices to school). Children also reported that they suffered eye strain when reading longer texts on devices. The reading preferences of bilingual children participating in the survey showed similarities across languages. The children showed convergence in their preferred reading materials across their two languages, as storybooks and comics were the two most selected choices. Additionally, their preferences for the fiction genres are similar across languages, suggesting that these bilingual children are likely to favour books from their favourite genres in either of their languages.

[RQ2] There was a decline in reading enjoyment and frequency as participants progressed through the upper primary grades. However, Primary Five children read for longer in English, suggesting that reading habits may evolve rather than simply decline. 36 P5 FGD participants were interviewed about the change in reading habits from P3 to P5 and many of them shared that they read in both languages less often than P3 and P4 due to having a busier schedule. Another recurring theme is that some of these children read less in P5 because of their declining interest in reading. FGD results also suggest that children need help in choosing what to read as they grow older. Children repeatedly reported that they would read more books if they knew what to read. There is also evidence that children faced challenges in locating MT reading materials that cater to their MT language proficiency.

[RQ3] While there were some differences in reading habits between boys and girls, the gender difference may not be as strong as portrayed. Boys and girls did not significantly differ in the duration of print or digital reading in English and MT. Interestingly, gender differences are not so clear-cut across the three groups: Chinese children demonstrated more gender differences, whereas gender differences were less evident in Malay and Tamil children. For Chinese children, girls enjoyed reading in their two languages significantly more than boys did. They also read print books and digital materials in English and Chinese more frequently than boys did. Although boys and girls did not differ in their self-rated English proficiency, girls reported higher levels of Chinese proficiency than boys. Intriguingly, boys and girls did not differ in the duration of print reading in English and Chinese, and boys even reported significantly longer duration of digital reading than girls. Boys and girls did not differ in their preferred reading devices and preferred types of reading materials. For fiction books, more girls enjoyed comedy and relationship themes than boys, while boys preferred sport-, war- and spy-related fiction regardless of the language. Both genders expressed converging preferences for the rest types of fiction books. Regarding non-fiction books, children showed gender differences only in sport- and science-related non-fiction books, which boys overwhelmingly preferred. Gender similarities were apparent in children's access to reading resources at home, preferred reading devices, preferred reading materials, and preferred types of fiction and non-fiction books. In the FGD data, the only evident gender difference was that in the book titles FGD participants mentioned, boys named more sport-, war- and spy-related titles.

[RQ4] The study explores the cross-linguistic relationships between children's reading enjoyment and reading habits in English and their respective mother tongue (MT) languages. In consideration of earlier results showing children's preference for print books over e-books/online materials, the correlation analyses focus on print books only. Significant and positive correlations between reading enjoyment, frequency, and duration in bilingual children's two languages were found, suggesting that children's reading habits and preferences can be connected across languages. The finding suggests that it is possible that the changes in syllabus focus positively impacted children's leisure reading in MT. Moreover, the results were analysed at the school level and the same patterns of correlations were found for each school, suggesting that, in general, the participating children who enjoy reading and who read often and for longer durations in one language are likely to do so in the other language. The FGD data further substantiated this interconnection by showing that reading as an enjoyable experience was not restricted to any language, as the children recurrently described reading a good English or MT book as 'enjoyable', 'get addicted', 'I feel like you're inside the book and sometimes I can dream about it', or 'like once I read one book, it's like five minutes, but it's like one hour actually passed'.

 [RQ5] The COVID-19 lockdown affected children's bilingual leisure reading. Reading was considered a more preferable leisure activity during the closure compared to before it. The lockdown had differential effects on children's reading habits and preferences. Children who enjoyed reading before the lockdown reported enjoying reading more and reading more than usual during the lockdown. However, those who did not enjoy reading tended to enjoy reading less and read less than usual. This finding emphasizes the importance of nurturing reading enjoyment in children to sustain their reading habits during prolonged periods spent at home. The lockdown also widened the reading gap between children's stronger (English) and weaker (MT) languages. Children read significantly more in English than in their MT languages, both in print and digital formats. Technology was under-utilized for reading in both languages during the lockdown. Despite having access to devices, spending more time on devices did not necessarily lead to more digital reading. This highlights the need for effective strategies to encourage digital reading habits among children. 

Implications for Research

• Results extend the body of research documenting the important role of leisure reading in monolingual children's reading development by providing a general picture of Singaporean primary school children's reading habits and preferences in English and their MT. 
• The study reports significant correlations between reading enjoyment, frequency, and duration in bilingual children's two languages, suggesting that children's reading habits and preferences can be connected across languages. Children who enjoy reading and who read often and for longer durations in one language are likely to do so in the other language.
  
• The study extends existing research on summer reading loss by documenting pandemic-related lockdown (a similar period of absence from school) effects on bilingual children's leisure reading.
•  Provides baseline data for future longitudinal studies to trace changes in bilingual children's reading habits and preferences.

 Implications for Practice

Providing direct research evidence and promoting research thinking among teachers, Heads of Departments, and principals can impact practice. Teachers showed great interest and found the results very pertinent and useful. More importantly, it was greatly encouraging for teachers to see some of their current practices reflected positively in the results. Consequently, they took the initiative to incorporate findings from the survey in designing their upcoming reading programmes.

To nurture bilingual children's leisure reading in their two languages, universal reading programmes/activities may not work the same way for different languages, grade levels, or genders. More fine-tuned designs are necessary. Some recommendations follow. 

1. Children's interest in reading must be stimulated and reinforced throughout primary school. Reading programmes/instruction must differentiate the needs of lower and upper primary children to ensure a smoother transition from learning to read to reading to learn.
2. Reading enjoyment should remain the central focus of reading programmes/instruction.
3. Children will benefit from a rich literacy environment filled with ample opportunities to interact with different types of reading materials in meaningful and interesting ways.
4. To pique upper primary children's interest in reading, teachers may consider designing reading tasks that are moderately challenging and stimulate curiosity.
5. Teacher support and scaffolding are essential tools for encouraging reluctant readers to read. Research has demonstrated the effectiveness of direct instruction in reading strategy followed by independent and collaborative reading activities in engaging reluctant readers.
6. Both English and MT reading programmes benefit children, though the challenges in each language may differ.     

Potential Future Research

• Longitudinal studies to track changes in bilingual children's reading habits and preferences over time.
• The results revealed by children's survey and FGD data tell us very little about how reading culture in school may influence them to choose certain books in a certain language to read. Understanding leisure reading from teachers' perspective will help schools, policymakers, and teachers better evaluate and design reading instructions to connect bi/multilingual children's leisure reading and bi/multilingual education.
• COVID-19 restrictions prevented FGD with a larger group of children. Future studies should explore bilingual leisure reading on a larger group. 

Co-Principal Investigators

• A/P Loh Chin Ee, English Language & Literature (ELL), NIE
• Dr O’Brien Beth Ann, Head, Early and Middle Childhood Research – CRCD, OER, NIE

 Collaborators

Internal Collaborators Specializing in Leisure Reading:

• A/P Mohd Mukhlis Bin Abu Bakar, Asian Languages & Cultures (ALC), NIE
• A/P Tan Chee Lay, ALC, NIE
• A/P Seetha Lakshmi, ALC, NIE
• A/P Viniti Vaish, ELL, NIE

 Internal Collaborator Specializing in Project Management:

• A/P Rita Elaine Silver, OER, ELL, NIE

 Internal Collaborator Specializing in Survey Design and Data Analysis:

• A/P Nie Youyan, Psychology and Child & Human Development (PCHD), NIE

 External Collaborators:

• Mdm Wee Jane, National Library Board (NLB) Mother Tongue Division
• Mdm Ang Lynn, NLB Mother Tongue Division

 Consultant

• Prof Xiao Lan Curdt-Christiansen, University of Bath, UK

INTRODUCTION/BACKGROUND

This project focuses on improving literacy development for young learners who are struggling with learning to read English by investigating the process of learning grapheme-phoneme correspondences (gpcs). Learning GPC is foundational to learning to read alphabetic languages and is a core problem for struggling readers. In this project, two methods are used in two studies to understand the process of learning English gpcs as the crux of acquiring literacy. First, a machine learning neural network modelling approach is used to study the effect of sound-symbol grain size and consistency and training input on learning progression and outcomes. Second, a behavioural randomized controlled study is conducted to examine the effects of interventions with LSP students focused at different grain sizes. Between these two studies, information about the types of input that may yield most effective learning is corroborated.

STATEMENT OF PROBLEMS

Acquiring grapheme-to-phoneme correspondences (GPC) for reading is a more drawn-out process in opaque writing systems like English because there are many inconsistencies in the symbol-sound mappings. Letters and letter patterns can be pronounced different ways (such as ‘ow in ‘owl’ versus ‘pillow’) and the same phonemic sounds can be spelled in different ways (/m/ as in him, hymn, or thumb). Because there are many such quasi-regularities of the mappings, with some being subtle or context-sensitive, it is difficult to explicitly teach all cases – and so students must pick-up many of the quasi-regular patterns implicitly, through experience. This leads to a key educational question of how to select and organize the best input experiences for learning.

In these studies, the focus is on two key variables known to affect GPC acquisition: grain size and consistency. Grain size relates to the number of letters in a grapheme that corresponds to a single sound, and different psycholinguistic units of sound within words – from syllables to onset-rime patterns, to phonemes. Inconsistent gpcs are often less ambiguous when taken as part of a larger grain size unit (consider ‘au’ in ‘gauge’ or ‘laugh’, compared with the rime unit in ‘caught’ or ‘taught’). This leads to the question of whether it is better to teach these larger, more consistent unit sizes.

These issues are particularly important when considering struggling learners, as they usually lack the amount of input experience that typical or avid readers get. Therefore, to maximize the effectiveness of their learning of GPC patterns, the optimal type of input is required. Secondly, effective ways to increase exposure, so that such learners can pick up gpcs implicitly, have been found in technology-based formats with game-like activities that are engaging without a high cognitive load. This project investigates use of such technology applications along with the question of how to focus input.

PURPOSE OF STUDY

The aim of this project is to examine and identify the types of input that yield effective learning of English sound-symbol correspondence (GPC). This objective is met with two studies using multiple methods: a machine learning approach using artificial neural network (ANN) simulations comparing different input training regimes; and a randomized-controlled behavioural intervention with LSP students contrasting technology-based instruction at different levels of grain size. This aim is met through the following objectives:

• To model the effects of different approaches to teaching English phonics and examine the learning space of input patterns on learning outcomes to inform intervention.
• To extend a successful technology-based approach of teaching GPC to P1 children in the learning support program (LSP), and to extend the technology-based approach from an emphasis on accuracy to fluency for decoding graphemes.
• To examine how individual differences may interact with outcomes of the approach.

PARTICIPANTS

For study 2, there were 148 student participants from seven primary schools (P1-P2, Mean age = 79.91 months, SD = 4.82). All were enrolled into the learning support program (LSP) at their schools. Students were matched in triplets within their LSP class based on their assessed reading ability (British Ability Scales-III). Students within the triplets were randomly assigned to one of three intervention groups for the study (phoneme-, rime, or word-level focused intervention). There were 46 students per intervention group across schools.

METHODOLOGY / DESIGN

Study 1 An artificial neural network (ANN) was built with a similar architecture to that of previous research; a three layer network with orthographic, phonological, hidden, and clean-up units. The model was presented with input in the form of words, first in a ‘pre-literacy’ training phase, mimicking early childhood development of oral language, and then in a ‘literacy’ training phase, where orthography-to-phonology mappings had to be learned during training, via the hidden units layer which adjusted the weightings between these units based on the input during training. Three sets of simulations were run, comparing input with words from low to high grain size and vice versa, and words from high to low consistency and vice versa. The model was tested periodically over and at the end of training for accuracy to read words and to decode nonword (pseudowords).

Study 2 Students engaged with touch-screen applications on their own ipad for 10 minutes of their LSP class 5 days per week over two 7-week periods (in school Terms 2 and 3). The apps were designed to focus learning at different grain size units: phoneme-level (for group I, using See Word reading then grapholearn Phoneme apps), rime-level (for group II, using See Word reading then grapholearn Rime apps); and the word-level (for group III using an in-house developed Words app). All groups were assessed at baseline prior to the intervention on nonverbal cognitive ability, verbal memory, receptive vocabulary (in 2 languages), phonological awareness, rapid symbol naming, orthographic awareness, statistical learning, and basic reading ability. They were assessed in four waves at pre-test, mid-test, post-test, and 3 months’ follow-up on measures of word and pseudoword reading accuracy, word and pseudoword reading fluency, and spelling. Progress over time was analysed with linear mixed model regressions, with intervention group as a predictor, and then adding baseline measures as moderators (as interactions with group).

FINDINGS / RESULTS

Research Question 1

Study 1. Compared to a large-to-small (LS) grain size training regime in simulation 1, a small-tolarge (SL) training regime yielded faster learning and higher accuracy for decoding pseudowords. In simulation 2, a high-to-low consistency training regime yielded faster learning and superior word and pseudoword reading accuracy than a low-to-high consistency training regime. When the number of training sessions was held constant across these same two training regimes (simulation 3), the high-to-low consistency regime still yielded more efficient learning, but the low-to-high regime performed better on low consistency words after training.

Study 2. Comparison between the intervention groups showed that all three improved over time at the same rate for word reading accuracy and fluency, pseudoword fluency, and spelling, whereas the phoneme-level group showed an advantage for decoding accuracy over the rime-level group (with no difference from the word-level group). At follow-up, there were no group differences, and all groups continued to improve in word reading accuracy but declined in word and pseudoword reading fluency relative to peers – meaning they were not keeping up to the progress of typical learners when they entered the next grade level.

Research Question 2

In study 2, children’s baseline scores were examined to see if other skills moderated intervention effects. In particular, phonological awareness and statistical learning showed some moderating effects on learning. Phonological awareness effects are highlighted, as this moderated intervention effects for word reading fluency outcomes. Specifically, there was less of an effect of phonological awareness on outcomes for the rime-level group, and a moderate effect for the phoneme-level group, and more of an effect of phonological awareness on outcomes for the word-level group.

CONTRIBUTIONS / IMPLICATIONS

These findings, together from two studies utilizing different methodologies, suggest that input experience for English literacy development is best when focused at a small grain size initially, especially for the development of decoding skills, which are assessed with pseudoword reading. Simulations from study 1 also showed that an initial focus on consistent words led to more efficient learning, even though the most inconsistent words were not learned as robustly as when they are trained first.

The results have implications for teaching and curriculum, with support for the use of decodable texts before authentic literature with more irregularly spelled, inconsistent words. Further, the moderating effect of individual differences in phonological awareness suggests that children with poor phonological awareness benefitted less from the supplementary intervention, and may require additional preparation to improve this metalinguistic skill.

Further, the intervention study suggests that supplemental, technology-based apps are a promising way to engage LSP students and to give them the additional practice required to learn GPC. The tablet-rendered games could reinforce the gpcs students acquire from direct instruction in LSP lessons, as the set of activities (in See Word Reading and Words apps) were planned and carefully tied to the LSP curriculum.

CONCLUSION

The two studies in this project confirm the benefits of a teaching approach focused on letter-sound correspondence and consistent GPC when students are first learning to read. Several recommendations to support struggling readers include, first, to support their acquisition of GPC by directing their attention to single graphemes in print, and devising word learning lists to begin with those that are most consistently pronounced. Second, to consider individual differences in literacy related skills, like phonological awareness, and include components to improve these skills, while also considering differential intervention approaches for different types of learners. Third, to consider the use of game-based applications for reinforcing GPC learning and decoding skills is a good way to provide more extensive practice of these skills in an engaging format. This could be used during class time when the teacher is conducting 1:1 progress monitoring assessments. Classmates would be able to work independently with the games on tablets without the teacher’s attention.

ACKNOWLEDGEMENTS

This study was funded by Singapore Ministry of Education (MOE) under the Education Research Funding Programme (OER 0417 OBA) and administered by National Institute of Education (NIE), Nanyang Technological University, Singapore. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Singapore MOE and NIE.

Co-principal Investigators

• De Roock Robert Santiago
• Luca Onnis (NTU)
• Tan Chee Soon

Collaborators

• Sally Ann Jones
• TAN Ah Hwee (School of Computer Science and Engineering, NTU)
• Suziana Mohd Ismail (CPDD, MOE)
• Renee Seward (University of Cincinnati, USA)
• Heikki Lyytinen (University of Jyväskylä, Finland)
• Elizabeth Pang (CPDD, MOE)