Writing Development from Grade 3 to 6 in Students with a History of Reading Difficulties

Abstract

In this study, we explore writing development from Grade 3 to 6 in four groups of students with varying reading abilities measured in a screening in Grade 2 as follows: students with (1) poor word reading, (2) poor reading comprehension, (3) mixed difficulties, and (4) typical reading. The written texts were collected from the narrative writing assignments in the National Assessment Tests in Swedish (L1) in Grades 3 and 6. The texts were analysed in relation to vocabulary diversity (VocD), narrative text quality (Narrative Scoring Scheme), and text length (total number of words). A mixed ANOVA was used to explore between-group differences and within-group changes in writing measures from Grades 3 to 6. A significant main effect for time was found for vocabulary diversity, narrative text quality, and text length, indicating a growth in writing performance for all groups from Grade 3 to 6. There were no significant interaction effects between group and time. Furthermore, there was a significant main effect for group on all three written text measures. Post-hoc analyses revealed significant differences between the groups with typical reading and mixed difficulties on all written text measures. The results will be discussed in relation to the varying reading abilities among the four reading profiles and in relation to possible educational implications.

1. Introduction

Reading and writing are a central part of the curriculum in language subjects and also important prerequisites for learning in all school subjects. In early primary school, the main focus is on learning to read and write, while in later primary and secondary school, there is a shift towards reading and writing to learn (Fasting et al., 2009). Previous studies have shown that students with reading difficulties are at risk of lower educational attainment in primary and secondary school (Holopainen et al., 2017; Levlin et al., 2024; Ricketts et al., 2014), and that students with reading difficulties often experience writing difficulties both in early and later school years (Graham et al., 2020a). However, there is a lack of longitudinal studies examining to what extent early identified reading difficulties may have long lasting effects on written text production throughout primary school. Further, previous studies on writing development in primary school are mainly cross-sectional, exploring short time periods, students within the same grade or across adjacent grades. In this study, we explore the characteristics of written text production (text length, vocabulary diversity, and narrative text quality) in L1 Swedish longitudinally from Grade 3 to 6 in a group of students identified with varying reading profiles in Grade 2, that is, students with typical reading, poor word reading, poor reading comprehension, or mixed difficulties (Tunmer & Greaney, 2010). The study adds new knowledge on how different reading profiles in Grade 2 may affect the characteristics of written text production and writing development from Grades 3 to 6 in primary school.

1.1. Previous Research

1.1.1. Typical Writing Development

Typical writing development across schooling has been described from different perspectives. Bereiter and Scardamalia (1987) described writing development as a progression from a novice stage of knowledge-telling, via a stage of knowledge-transforming, to an expert stage of knowledge-crafting (cf. Kellogg, 2008). Myhill (2009) identified three developmental trajectories in writing, which partly overlap with Bereiter and Scardamalia’s (1987) stages. The trajectories include transitions from spoken to written language, from simple to elaborate and clear statements, and from conveying ideas through sentences (cf. knowledge-telling) to transforming ideas into sentences with rhetorical impact (cf. knowledge-transformation).

A few older studies have also been concerned with how various linguistic features change during writing development (e.g., Harpin, 1976; Hunt, 1965; Loban, 1976; Perera, 1984). While Hunt (1965) and Loban (1976) were longitudinally covering the age range from the earliest writing to writing at the end of schooling (Grade 12), Harpin (1976) and Perera (1984) focused on the age range 7–12 years, which is the age in focus in the current study. Harpin (1976) studied a range of linguistic features such as sentence length and use of pronouns as well as simple and complex sentences among 7–11-year-olds and found that personal pronouns decreased, whereas sentence length and subordinate clauses increased with age. Perera (1984) studied writing development from 8 to 12 years of age, focusing specifically on how grammatical complexity changed with increasing age. She found that the younger children relied heavily on coordination to create more complex sentences, primarily used simple nominal phrases and active lexical verbs, and overused personal pronouns in sentence-initial position. With increasing age, children used more subordinate clauses, longer and structurally more complex nominal phrases, passive and modal verbs, and connectives expressing causal and adversative meanings.

Building on the insights from previous research, Christie (2010) defines four phases in writing development, with phases 1 and 2 being of interest for the current study. In phase 1 (first years in primary school, 6–8 years), children develop motor skills for forming letters and coding skills for spelling (letter knowledge, sound-letter knowledge), as well as basic knowledge of grammar. In phase 2 (the last years of primary and the beginning of secondary schooling, 9–13 years), children gradually enhance all their linguistic resources, enabling them to master the grammar of written language and sentence formation (cf. Abbott et al., 2010; Witte & Faigley, 1981). Thematic progression, which in early narrative writing often builds on chronology, develops from a reliance on using nouns and pronouns in sentence initial position to establish a theme and create reference in phase 1, to a more varied use of theme choices, for example, nominal and prepositional adverbials, to build cohesion and thematic progression in phase 2 (Christie, 2010).

Another important feature in narrative writing is the description of characters and their mental states, the setting/environments, and the development of events (conflict and resolution). To our knowledge, no previous studies have explored these aspects of narrative writing longitudinally. In an older cross-sectional study by Roth and Spekman (1986; referred to in Nippold, 2006), students in three age groups (8–9 years, 10–11 years, and 12–13 years) were asked to orally tell an imaginative story. The number of complete episodes was analysed across age groups with complete episodes defined as including at least three key elements as follows: an initiating event that prompts a character to take action/introduction of a problem, an attempt to address the problem, and finally a consequence or resolution to the problem. The proportion of complete episodes increased from 50% for the youngest children to 69% for the oldest children. Furthermore, there was an increase in embedded episodes across age groups indicating that the older children were more likely to use a more complex story plot in their narratives.

However, when considering writing development, ideas and narrative episodes need not only to be translated into language, but they also need to be transcribed into text, putting demands on transcription level skills (such as typing, handwriting, and spelling). In the Simple View of Writing (Berninger & Amtmann, 2003), the importance of transcription level skills for early writing development was highlighted, as well as their possible restraining effect on text production, planning, and revising. In a meta-review by Kent and Wanzek (2016), it was found that spelling correlated moderately with writing quality in both early and later primary school, an outcome confirming the role of spelling in early writing development in English. However, to what extent spelling explains individual variation in writing quality may depend on the degree of transparency in orthography. For example, Pinto et al. (2015) found that spelling in Italian, which in contrast to English has a transparent orthography with a systematic mapping between phonemes and graphemes, was found to no longer influence written text production in Grade 2 when the participants had learned to master spelling. Thus, there is a need to examine the role of spelling in writing development in languages with various degrees of orthographic transparency.

1.1.2. Connections Between Reading and Writing

Several previous longitudinal studies have found reading and writing to be correlated (e.g., Abbott et al., 2010; Kim et al., 2018). In Abbott et al. (2010), significant bidirectional relations between word reading and spelling were identified across Grades 2 to 7. Reading comprehension predicted variation in written text quality from Grade 2 to 6, while written text quality predicted variation in reading comprehension across Grades 3 to 6. In Kim et al. (2018), approximately 300 students were followed from Grade 3 to 6. Word reading and spelling were strongly related across grades, while reading comprehension and written composition were only weakly related. Despite these indications of connections between reading and writing, previous models have mostly been concerned with reading (e.g., Simple View of Reading, Construction Integration Model of reading comprehension) or writing separately (e.g., Simple View of Writing).

Kim (2020) addresses this lack in previous theory when introducing the Interactive Dynamic Literacy Model (IDL). IDL demonstrates how reading and writing development is a complex and dynamic process in which multiple levels of skills interact and reinforce one another. The model emphasises that literacy is not a linear process but is dynamically shaped by cognitive, linguistic, and social factors. Core components include linguistic skills (e.g., vocabulary, grammar, and discourse comprehension), cognitive skills (e.g., working memory and inferencing), and motivation and self-regulation. These interact with individual experiences and the surrounding environment, including instruction and sociocultural context.

In the IDL, Kim (2020) describes four central hypotheses. According to the first hierarchical hypothesis, the development of reading and writing skills can be understood as hierarchically organised, where skills at different levels interact both directly and indirectly. Discourse-level competencies, such as reading comprehension and writing ability, rest upon word- and spelling-level skills as well as oral language abilities and higher cognitive functions such as inferencing and self-regulation. General cognitive abilities, subject knowledge, and socio-emotional factors also contribute to discourse-level processes in reading and writing. Word reading and spelling, in turn, are grounded in early orthographic, phonological, and semantic skills. Thus, development proceeds hierarchically, though not strictly linearly, as different skills evolve in parallel and in reciprocal interaction.

In the second hypothesis of interactive relations, reading and writing are supposed to develop in close interaction and influence each other reciprocally. At the word- and spelling-level, the connections are strongest, whereas at the sentence- and text-level, the relationships tend to be weaker. Reading often has a more decisive impact on writing development than the reverse, but both skills rely on shared processes such as language comprehension, vocabulary, and knowledge of text structure. Furthermore, reading and writing development contributes to strengthening oral language abilities, subject knowledge, and higher-order cognition. Socio-emotional factors, particularly motivation and self-efficacy, play a central role in this development.

The third Hypothesis of Co-occurrence of Difficulties builds on the assumption that reading and writing largely share underlying skills, which means it is common for difficulties to appear in both areas simultaneously. Most students therefore tend to demonstrate either strong skills in both reading and writing or difficulties in both. The connections are particularly evident at the word- and spelling-level, while a greater variety of profiles emerge at the text-level. It is considerably more common for weak reading skills to negatively affect writing ability than for strong writing skills to develop in combination with weak reading skills.

In the fourth Hypothesis of Dynamic Relations, the relationships among the different components are supposed to change over time and are influenced by individual differences as well as by how reading and writing abilities are assessed. In the early stages, word decoding and spelling are most critical, whereas language comprehension and higher cognitive functions gain importance in later grades, as texts become more complex. Writing also develops in parallel, evolving from simple reproductions to more reflective and knowledge-generating productions (cf. with knowledge-telling and knowledge-transforming in Bereiter & Scardamalia, 1987). Factors such as linguistic background, the presence of learning difficulties and the choice of assessment methods influence developmental patterns in different ways.

1.1.3. Reading Difficulties and Writing Development

Considering the close relationship between reading and writing, it is not surprising that reading difficulties often co-occur with writing difficulties. In the model Simple View of Reading/SVR (Tunmer & Greaney, 2010), reading development depends on two key dimensions: decoding and oral language comprehension. Both components are necessary for the development of reading comprehension (Lervåg & Melby-Lervåg, 2022). Depending on individual variations in decoding (word reading accuracy and fluency) and oral language/listening comprehension (vocabulary, grammar, inferencing, and discourse level processing), there will be three different profiles with reading difficulties according to the SVR model (Tunmer & Greaney, 2010) as follows: (1) poor decoding/word reading (with typical listening comprehension); (2) poor listening and reading comprehension (with typical word reading); and (3) mixed reading difficulties (poor word reading and comprehension).

Several previous studies have found that written text production seems to be extra challenging for students with reading difficulties (Berninger et al., 2008; Costa et al., 2016; Graham et al., 2020a; Puranik et al., 2007; Sehlström et al., 2023). Poor word reading often co-occurs with poor spelling due to these skills’ mutual dependency on well-developed phonological processing (Kim, 2020; Melby-Lervåg et al., 2012). This may explain the frequent challenges with transcription level skills in written text production among students with word reading difficulties/dyslexia (Berninger et al., 2008; Graham et al., 2020a; Sumner et al., 2013, 2016). When it comes to higher level aspects of written text production (e.g., linguistic complexity, text organisation, and overall text quality), previous studies show varying results. For example, students with word reading difficulties have been found to perform below peers in linguistic complexity (vocabulary diversity and grammar) and overall text quality in some studies (Berninger et al., 2008; Wengelin, 2007; Sumner et al., 2016), but not in others (Puranik et al., 2007). In a meta-review by Graham et al. (2020a), students with reading difficulties (57% of studies were conducted in the age range 9 to 12) performed significantly below age-matched peers on all written text measures. Spelling was particularly challenging (1.5 SD below peers), as well as writing quality, vocabulary, and syntax (approx. 1 SD below peers). Handwriting and text organisation were relative strengths (approx. 0.75 SD below peers). Difficulties in linguistic complexity and overall text quality have mainly been interpreted as spillover effects from the challenges with spelling and reading, taxing this group of students’ working memory capacity and their possibilities to put effort into higher-level planning and revising when writing (Berninger et al., 2008; Sumner et al., 2016; Wengelin, 2007).

Studies examining written text production in students with poor oral language and/or reading comprehension all point to substantial challenges in higher-level aspects of text production, for example, lower linguistic complexity with restricted vocabulary diversity and a simpler syntax with more syntactic errors (Carretti et al., 2013, 2016; Puranik et al., 2007), as well as poorer idea development and text organisation (Carretti et al., 2013, 2016; Cragg & Nation, 2006) compared to peers with typical comprehension and oral language. In a study by Cragg and Nation (2006), including 9–11-year-old students with poor reading comprehension and typical decoding skills, it was found that the participants produced equally long texts with the same level of complexity in grammar and spelling as their peers. However, their texts contained significantly fewer details and events as well as poorer text organisation compared to their peers with age-typical reading comprehension. The same pattern was found in studies by Carretti et al. (2013, 2016) comparing oral and written narrative production between a group of students (ages 8 to 10) with poor reading comprehension and a group of age-matched peers with typical reading development. In both studies, idea development, coherence, and structure were less well developed in the group with poor reading comprehension. When describing a narrative picture sequence, the students tended to describe each picture without connecting the events to a coherent storyline. Furthermore, lexical and syntactic complexity were found to be below the performance of their age-matched peers in both studies, and the use of connectives was restricted. The challenges were the same in both oral and written production. An interesting finding in Carretti et al. (2013) was that the participants performed on par with their peers in a descriptive task. The authors suggest that it is the demands on cohesion and causality in the narrative genre that pose the main challenges for students with poor comprehension. Furthermore, in both studies, text length was the same as for their peers, and spelling was within age norms. A meta-analysis by Graham et al. (2020b) further confirms that writing is a challenge for students with oral language/comprehension difficulties (in this case, students with Developmental Language Disorder/DLD). Students with DLD performed significantly below peers in overall writing quality, text length/output, grammar, vocabulary, and spelling. However, the meta-analysis did not control for the extent to which the participants had a concurrent difficulty in word reading. Hence, it is not known if the participants were mostly in the poor comprehension only or in the mixed difficulty profile.

To our knowledge, few previous studies have explored reading and writing difficulties longitudinally. For example, in a study by Costa et al. (2016), 137 students’ reading and writing development was followed from Grade 1 to 4. In Grade 1, 54 students were found to have a writing difficulty, of whom 30% had a combination of reading and writing difficulties. There was a high degree of instability across grades with movements between the groups with typical development, writing difficulties only, and the combination of reading and writing difficulties. However, the number of students with a combination of reading and writing difficulties was constant across years. This instability over time regarding group membership may reflect the change from assessing transcription level and pre-reading skills in Grades 1 and 2 to written text production and word reading in Grades 3 and 4 in Costa et al. (2016), indicating that students struggling with transcription-level skills in early school years may not be the same students that struggle with text production later.

1.2. The Current Study

The current study was conducted in a Swedish school context in primary school. A group of students identified with reading difficulties in Grade 2, and their writing development was followed longitudinally from Grade 3 to 6. In Grade 2, all students word reading, spelling, and reading comprehension were assessed as part of the participating schools regular screening procedure. In Grades 3 and 6, all students took part in the compulsory national assessment tests in Swedish, and for the present study we collected the students’ narrative writing assignments from Grades 3 and 6. When this study was conducted, the knowledge requirements for the narrative writing assignment in the national test in Grade 3 were as follows (Swedish National Agency for Education, 1994): the student is expected to write a narrative text with a distinct and clear story plot, be able to spell words that frequently occur in age-appropriate texts, and use basic rules for interpunction. In Grade 6, the knowledge requirements for the narrative writing assignment were as follows (Swedish National Agency for Education, 2011): The student is expected to write different types of texts with functioning text structure (the content should be understandable to the reader) and with some variation in language use, implement basic rules for spelling, interpunction, and grammar, and use simple descriptions of characters and include basic elements in a story plot. In the current study, we did not use the teachers’ examinations of the writing assignments; instead, we conducted separate assessments by using written text measures common in research (text length, vocabulary diversity, and narrative quality). Considering that previous research on reading difficulties and writing mainly consists of cross-sectional studies (Berninger et al., 2008; Cragg & Nation, 2006; Carretti et al., 2013, 2016; Puranik et al., 2007; Sumner et al., 2013, 2016), this study may contribute additional knowledge regarding the impact of early identified reading difficulties on writing development across primary school.

Furthermore, the study is conducted in Swedish, a language with a semi-transparent orthography. Most previous studies have been conducted in English with an opaque orthography, i.e., a high incidence of orthographic inconsistencies like multi-letter graphemes, context dependent rules, irregularities, and morphological influence on spelling rules. In contrast, Swedish has a fairly regular mapping between phonemes and graphemes. With only a few multi-letter graphemes and some irregularities (e.g., silent letters, archaic spelling, loan words), as well as some morphological influence on spelling, making it potentially easier to learn to read and spell in Swedish compared to English (Caravolas, 2004; Seymour et al., 2003). Thus, the study may add new insights into the specific challenges students with reading difficulties may or may not experience when learning to write when the phoneme-grapheme connections are more ambiguous than in a transparent orthography, but less ambiguous than in a non-transparent orthography, such as English.

In the current study, we specifically examined the following research questions:

• What is the effect of the reading profile and Grade (3 vs. 6) on written text production?
• What characterises written text production in Grades 3 and 6 in a group of students identified with reading difficulties in Grade 2?

2. Method

2.1. Participants

Ninety-eight students were recruited from 14 schools in two municipalities. For 64 of these students, the writing assignments for both Grade 3 (age 9 to 10) and 6 (age 11 to 12) were available, leading to a final number of 64 participants (39 girls and 25 boys) in the current study. All participants have Swedish as their first language.

The participants were allocated to different reading profiles based on reading measures collected through the municipality’s regular screening programme for all students in Grade 2. The definition of reading profiles followed the SVR (Tunmer & Greaney, 2010). Raw scores in reading comprehension and word reading were converted to z-scores using the tests’ standard reference data (Jacobson, 2001; Järpsten, 1999). The following criteria were used when allocating participants to the four reading profiles:

(1)

Typical reading (TR): word reading and reading comprehension z ≥ −0.5.

(2)

Poor word reading (PWR): word reading z ≤ −0.7, reading comprehension z ≥ −0.5.

(3)

Poor reading comprehension (PRC): reading comprehension z ≤ −0.7, word reading z ≥ −0.5.

(4)

Mixed difficulties (mixed): reading comprehension and word reading z ≤ −0.7.

Table 1. Descriptives and one-way ANOVA for reading and spelling measures in Grade 2.

Measures	TR n = 18	PWR n = 16	PRC n = 16	Mixed n = 11	F (3, 57)	η2	Post-Hoc
	M (SD)	M (SD)	M (SD)	M (SD)
Grade 2
Word reading (z)	0.47 (0.75)	−1.0 (0.31)	0.03 (0.54)	−1.1 (0.24)	32.02 ***	0.63	TR = PRC > PWR *** TR = PRC > Mixed ***
Reading comprehension (z)	0.66 (0.63)	0.39 (0.69)	−1.16 (0.32)	−1.49 (0.52)	54.24 ***	0.74	TR = PWR > PRC *** TR = PWR > Mixed ***
Spelling (z)	0.20 (1.11)	−0.06 (0.59)	0.24 (0.86)	−0.79 (0.64)	3.91 *	0.17	TR = PWR = PRC TR = PRC > Mixed *

Note. TR = Typical reading; PWR = poor word reading; PRC = poor reading comprehension; Mixed = mixed reading difficulties. * p < 0.05; *** p < 0.001.

shows the descriptives on reading measures in Grade 2 for each reading profile. A one-way between-groups analysis of variance was conducted to explore the differences between the four reading profiles in word reading and reading comprehension. Spelling was also included, considering its possible effects on written text production. As expected, considering the selection criteria, there was a significant difference for the four reading profiles in word reading: F (3, 57) = 32.02, p < 0.001, and in reading comprehension: F (3, 57) = 54.24, p < 0.001, with large effect sizes (calculated using eta-squared) for both measures (see Table 1). Post-hoc comparisons using Bonferroni indicated that the groups with TR and PRC performed significantly higher mean scores in word reading compared to the groups with PWR (p < 0.001) and mixed difficulties (p < 0.001). In reading comprehension, the groups with TR and PWR performed significantly higher mean scores compared to the groups with PRC (p < 0.001) and mixed difficulties (p < 0.001). In spelling, the groups with TR and PRC performed significantly higher mean scores than the group with mixed difficulties (p < 0.05). The outcome indicates that the group with mixed difficulties struggled with spelling to a larger extent than the other two groups with reading difficulties.

2.2. Materials and Measures

2.2.1. Reading and Spelling Measures in Grade 2

Word reading. Word-chains (Jacobson, 2001). Students read chains of semantically unrelated words and marked the word boundaries by drawing a line. The score was the total number of correctly marked words in two minutes. Reported test-retest correlations in the manual were 0.89 for Grade 2 (Jacobson, 2001).

Reading comprehension. Diagnostic reading and writing (Järpsten, 1999). The students silently read short paragraphs of texts and then answered multiple-choice questions. The score was the total number of correct answers within 30 min. Reported internal validity in the manual was 0.62 for Grade 2 (Järpsten, 1999).

Spelling. Diagnostic reading and writing (Järpsten, 1999). The students listened to a story with each sentence containing a target word. The target word was orally repeated by the teacher and thereafter written by the students. Most target words had a regular spelling and a word length of 3–7 letters. Half of the words included consonant clusters, and a few words contained irregular spelling patterns. Reported internal validity in the manual was 0.73 for Grade 2 (Järpsten, 1999).

2.2.2. Writing Assignments in Grade 3 and 6

The National Assessment tests in Sweden are compulsory for all students in Grades 3 and 6. The tests are administered by the teacher in the students’ classrooms during the spring semester.

Narrative assignment in the National Assessment Test in Grade 3 (Swedish National Agency for Education, 1994). The students were instructed to write a story about a self-experienced or fictive adventure. Three supportive questions were given in written and oral form as part of the introduction to the task: How does the story start? What happens next? How does it end? The writing session was in total 60 min, including teacher instructions. The students wrote by hand.

Narrative assignment in the National Assessment Test in Grade 6 (Swedish National Agency for Education, 2011). The students were instructed to write a story about their first show as a circus artist. Instructions included supportive questions like: What are you good at? How did you become good at that? How did your first show go? The writing session was in total 60 min, including teacher instructions. A majority of the students wrote by hand, but some students with learning difficulties (n = 16) wrote on the computer, but without digital tools such as spelling correction.

2.2.3. Written Text Measures in Grade 3 and 6

Text length. Text length was measured by calculating the number of words per text.

Vocabulary diversity. Vocabulary diversity was calculated using the CLAN programme VocD (MacWhinney, 2000; McKee et al., 2000). VocD has been shown to be a valid measure of lexical diversity in texts (McCarthy & Jarvis, 2010). It is a measure of Type Token Ratio (TTR) that reduces the impact of text length by randomly selecting 50 words on three occasions from the same text while calculating TTR and controlling for how much TTR usually decreases with increasing text length. Errors in spelling or morphology in the included texts were corrected before calculating VocD. Texts shorter than 50 words were excluded from these analyses (n = 4 in Grade 3, and n = 0 in Grade 6).

Narrative text quality. The Narrative Scoring Scheme (NSS) is a tool developed for assessing both oral and written narrative production in primary school (Heilmann et al., 2010; Miller et al., 2015). NSS is an analytic scoring scheme including the following seven categories that are key elements in narratives: introduction (with description of both environment and characters), character development, mental states of characters, referencing, conflicts and resolutions, cohesion, and conclusion. Each category is assessed separately by using scorings between 0 and 5 points, where 1 = immature, 3 = emergent, and 5 = proficient (see Heilmann et al., 2010, for a detailed description). A composite score of overall narrative quality based on the seven categories was used in the analyses in this study, giving a maximum score of 35 points. The internal consistency for the composite score was 0.72 in Grade 3 and 0.75 in Grade 6. See the Section 2.2.4 for more detailed information on specific adaptations of NSS made for this study.

2.2.4. Procedure

This study was assessed and approved by the Regional Ethical Review Board in Umeå, Sweden (dnr. 09-220 Ö 2009-1426-31). The project complied with the Swedish Ethical Review Act on research involving humans (SFS 2003:460, 2003) and followed the ethical guidelines of the Swedish Research Council (Stafström, 2017).

Caregivers and participants gave their written consent and admitted access to reading measures from the municipalities general screening programme in Grade 2 and the writing assignments from the National Assessment tests in Grades 3 and 6. At the end, 64 students writing assignments from Grades 3 and 6 were possible to locate in the municipality’s archives.

Reading and spelling scores in Grade 2 were collected as a part of the municipality’s regular screening procedure. The assessments were implemented by the teacher or special needs teacher in the student’s classroom from October to February.

Narrative writing assignments in Grades 3 and 6 were part of the National Assessment Test in Swedish. The National Assessment Tests were implemented according to the national guidelines during the spring semester in Grade 3 and Grade 6.

Prior to scoring the collected texts with the Narrative Scoring Scheme, NSS (Heilmann et al., 2010), an initial translation to Swedish and revision of the original version of the Narrative Scoring Scheme were conducted. This version was further revised throughout the course of a pilot study in Grade 3.

In the original version of the NSS, several of the criteria are not applicable to written narratives or when the story is unfamiliar to the reader. These criteria were either reworded/reformulated when possible or excluded from the scoring scheme. This applied to categories such as character development, referencing, conflict/resolution, and conclusion. For example, a criterion in character development that assesses the use of a narrator’s voice was excluded, as this cannot be evaluated in written text. Similarly, the criterion in conflict/resolution addressing omission of key conflicts that drive the story forward was disregarded, as it is not applicable when the story is unknown to the reader. In the conclusion category, the criterion for a score of 1 point was revised to better suit written narratives, as the original version refers to the child ceasing to speak. During the pilot study in Grade 3, challenges also emerged when assigning a single score for both setting and character in the Introduction category. These were scored separately and then calculated as an average score. Further, some of the scoring points in the original version of the NSS were not explicitly defined. When possible, such scoring points were specified in the revised version used in the current study. In the remaining categories, the scores were assigned based on professional judgement, as described in the original version (Heilmann et al., 2010).

The handwritten texts were digitalized and analysed by the authors and two research assistants in the project. The assistants received training in using NSS. Before the assessment of the texts in Grade 3, a two phased pilot study was conducted to ensure that the texts were analysed in an equivalent way and with the same approach by the two assistants and the first and second authors. In the first phase, three texts were randomly selected, and then analysed. In the second phase, 10 randomly selected texts were independently analysed to establish interrater reliability. Interrater reliability was acceptable with r = 0.77 for the composite score.

For Grade 6, interrater reliability was established based on a sample of 18 randomly selected texts that were assessed by one of the assistants and the first author. Interrater reliability was acceptable with r = 0.79 for the composite score.

2.3. Statistical Analyses

IBM’s Statistical Package for the Social Sciences (SPSS 25) was used for descriptive and inferential statistical analyses. Analyses of skewness revealed values between −0.15 and 0.79 for the dependent written text measures in Grades 3 and 6, apart from text length in Grade 6 with a slightly higher value of 1.26. No extreme outliers were identified in the boxplots, apart from one extreme positive value (+4 SD from the mean) for text length in Grade 6. Considering the small sample size, this value was changed to a less extreme value within 3 SD from the mean to not distort the statistical analyses.

A Pearson product-moment correlation was initially conducted to explore the relationships between word reading, spelling, and reading comprehension in Grade 2 and the written text measures in Grades 3 and 6 (see Appendix A).

A mixed design analysis of variance (ANOVA) was conducted to answer RQ 1 regarding the effect of reading profile and grade on the development of written text production from Grade 3 to 6, with reading profile as a between-subject effect and time period (Grade 3 and 6) as within-subject effects.

To answer RQ 2 regarding the characteristics of written text production for the four reading profiles in Grades 3 and 6, a one-way between-groups ANOVA was used with reading profile as an independent variable and text length, vocabulary diversity, and narrative text quality as dependent variables. Levene’s test for the dependent written text measures were non-significant, all p-values > 0.12, indicating equal variance across groups for all the dependent measures. Post-hoc comparisons were made using Bonferroni. Since a Pearson product-moment correlation (see Appendix A) showed no significant correlations between spelling in Grade 2 and the written text production measures in Grades 3 and 6, we decided to not include spelling as a covariate.

The significance value was set at p < 0.05 for all comparisons. Effect sizes for the ANOVA are reported as partial eta-squared, η_p²; small effect = 0.01, medium effect = 0.06, and large effect = 0.138 (Cohen, 1988).

3. Results

See Appendix A for information on correlations between reading and spelling measures in Grade 2 and written text measures in Grades 3 and 6. In addition to the abovementioned lack of significant correlations between spelling in Grade 2 and all written text measures in Grades 3 and 6, Appendix A shows that word reading and reading comprehension in Grade 2 correlated significantly with all written text measures (apart from text length in Grade 6), varying from medium to strong relationships for all measures except vocabulary diversity in Grade 3 (showing a weak significant correlation).

3.1. The Effect of Reading Profile and Grade on the Development of Text Length, Lexical Diversity, and Narrative Quality

A mixed between-within subjects’ ANOVA was conducted to assess the impact of reading profile (typical reading, poor word reading, poor reading comprehension, mixed reading difficulties) and Grade (3 vs. 6) on the students written text measures (see

Table 2. A mixed ANOVA exploring the effect of reading profile on the development of text length, lexical diversity, and narrative quality from Grade 3 to 6.

Text Measures	TR n = 18 M (SD)		PWR n = 16 M (SD)		PRC n = 16 M (SD)		Mixed n = 11 M (SD)		F (ηp2)
	Gr 3	Gr6	Gr 3	Gr 6	Gr 3	Gr 6	Gr 3	Gr 6	Reading Profile	Grade	Inter- Action
Total number of words	231 (102)	409 (199)	167 (101)	327 (157)	135 (78)	392 (171)	117 (40)	247 (102)	3.78 * (0.17)	80.62 *** (0.59)	1.57 (0.08)
Vocabulary diversity	70.3 (14.6)	77.0 (13.7)	60.7 (23.4)	70.6 (14.5)	57.3 (20.2)	66.3 (16.8)	46.8 (11.3)	62.8 (9.5)	4.68 ** (0.21)	18.07 *** (0.26)	0.72 (0.04)
Narrative quality (NSS)	25.9 (4.8)	29.2 (3.8)	23.7 (4.5)	25.3 (5.6)	22.4 (5.2)	26.1 (4.2)	22.6 (2.9)	23.2 (3.8)	4.48 ** (0.19)	11.13 ** (0.16)	1.07 (0.05)

Note. TR = Typical reading; PWR = poor word reading; PRC = poor reading comprehension; Mixed = mixed reading difficulties. * p < 0.05; ** p < 0.01; *** p < 0.001.

).

There were no significant interaction effects between reading profile and Grade on any of the written text measures. There was a significant main effect for reading profile on text length, F (3, 55) = 3.78, p = 0.016, η_p² = 0.17, on lexical diversity, F (3, 52) = 4.68, p = 0.006, η_p² = 0.21, and on narrative quality, F (3, 57) = 4.48, p = 0.007, η_p² = 0.19. The outcome indicates that reading profile explained from 16 to 21% of the variation in the three written text measures.

There was a significant main effect for Grade on text length, Wilks’ Lambda = 0.41, F (1, 55) = 80.62, p < 0.001, η_p² = 0.59; on vocabulary diversity, Wilks’ Lambda = 0.74, F (1, 52) = 18.07, p < 0.001, η_p² = 0.26; and on narrative quality, Wilks’ Lambda = 0.84, F (1, 57) = 11.13, p = 0.002, η_p² = 0.16.

In sum, the results indicate a positive development from Grade 3 to 6 in all written text measures for all four groups. However, reading profiles do explain a large amount of the individual variations in students writing performance, considering the large effect sizes.

3.2. The Characteristics of Written Text Production in Grades 3 and 6

A one-way between-groups ANOVA was conducted to explore the impact of reading profile on written text production, measured as text length, vocabulary diversity, and narrative quality in Grades 3 and 6 (see

Table 3. Descriptives and one-way ANOVA for written text production in Grades 3 and 6.

Written Text Measures	TR n = 18	PWR n = 16	PRC n = 16	Mixed n = 11	F (3, 55)	η2	Post-Hoc
	M (SD)	M (SD)	M (SD)	M (SD)
Grade 3
Total number of words	231 (102)	167 (101)	135 (78)	117 (40)	4.86 **	0.21	TR > Mixed ** TR > PRC *
Vocabulary diversity	70.3 (14.6)	60.7 (23.4)	57.3 (20.2)	46.8 (11.3)	3.83 *	0.18	TR > Mixed **
Narrative quality (NSS)	25.9 (4.8)	23.7 (4.5)	22.4 (5.2)	22.6 (2.9)	2.05	0.10
Grade 6
Total number of words	409 (199)	327 (157)	392 (171)	247 (102)	2.56	0.12	TR > Mixed (p = 0.07)
Vocabulary diversity	77.0 (13.7)	70.6 (14.5)	66.3 (16.8)	62.8 (9.5)	2.77	0.13	TR > Mixed (p = 0.055)
Narrative quality (NSS)	29.2 (3.8)	25.3 (5.6)	26.1 (4.2)	23.2 (3.8)	4.71 **	0.20	TR > Mixed ** TR < PWR (p = 0.06)

Note. TR = Typical reading; PWR = poor word reading; PRC = poor reading comprehension; Mixed = mixed reading difficulties. * p < 0.05; ** p < 0.01.

).

In Grade 3, there was a significant difference for the four reading profiles in text length, F (3, 55) = 4.86, p = 0.005, and in vocabulary diversity, F (3, 55) = 3.83, p = 0.015, with large effect sizes for both measures. No significant differences between reading profiles were detected regarding narrative quality, F (3, 55) = 2.05, p = 0.12; however, the students with typical reading performed a higher mean score than the three profiles with poor word reading and/or poor reading comprehension, and the calculated eta-square showed a moderate effect size (η² = 0.10) for the difference between mean scores. Post-hoc comparisons using Bonferroni indicated that students with typical reading produced significantly longer texts compared to the students with poor reading comprehension (p = 0.016) and mixed difficulties (p = 0.008). Furthermore, students with typical reading produced texts with significantly higher vocabulary diversity than students with mixed difficulties (p = 0.008).

In Grade 6, there was a significant difference for the four reading profiles only for narrative quality, F (3, 55) = 4.71, p = 0.005, and a tendency towards a significant difference for vocabulary diversity, F (3, 55) = 2.77, p = 0.05. The difference between mean scores is confirmed by the medium to large effect sizes for both measures. No significant difference between reading profiles was detected regarding text length, F (3, 55) = 2.56, p = 0.064; however, the effect size for the actual difference between mean scores was moderate. Post-hoc comparisons using Bonferroni indicated that students with typical reading produced texts with significantly better narrative quality compared to the groups with mixed difficulties (p = 0.004), and for vocabulary diversity, there was a tendency towards significant difference, with students with typical reading producing texts with higher lexical diversity than students with mixed difficulties (p = 0.055).

In sum, the results indicate that students with mixed reading difficulties in Grade 2 are more likely to perform the lowest mean scores on all three measures of written text production in both Grades 3 and 6. Furthermore, when looking at mean scores and effect sizes, text length seems to be a greater challenge in Grade 3 than in Grade 6 for students with poor reading abilities (regardless of reading profile), while narrative quality becomes a greater challenge for students with poor reading abilities in Grade 6.

4. Discussion

4.1. The Effect of Reading Profile and Grade (3 vs. 6) on Written Text Production

A mixed between-within subjects’ analysis of variance was conducted to assess the impact of reading profile in Grade 2 (i.e., typical reading, poor word reading, poor reading comprehension, and mixed reading difficulties) on the development of written text production from Grade 3 to 6. There was a significant main effect for reading profile on text length, vocabulary diversity, and narrative text quality, indicating that early reading difficulties identified in Grade 2 had an impact on writing performance in Grades 3 and 6, at least in this specific context when examining the narrative writing assignments from the compulsory National Assessment test in Swedish. Also, a main effect was found for grade (time), showing that the growth in text length, vocabulary diversity, and narrative quality from Grade 3 to 6 for all groups was significant. No significant interaction effects were found between reading profile and grade.

To our knowledge, no previous study has examined early reading difficulties and their effect on writing development through primary school. In Costa et al. (2016), writing difficulties and the overlap with reading difficulties were explored from Grade 1 to 4. Writing and reading difficulties were found to co-occur across Grades 1 to 4 for a subgroup of participants even though group membership was quite unstable across grades between typical writers, poor writers, and poor writers + poor readers. In the current study, we did not explore to what extent group membership changed over time. Instead, this study contributes new knowledge indicating that early reading difficulties identified in Grade 2 pose a risk for persistent writing difficulties through primary school, also in later primary school (Grade 6). However, depending on the specific reading profile, the developmental patterns may vary across grades.

Regarding text length, all reading profiles made significant progress from Grade 3 to 6, but an interesting finding was the substantial catch-up in number of words from Grade 3 to 6 among students with poor reading comprehension. In Grade 3, both profiles with poor reading comprehension performed very short texts, but in Grade 6, the students with poor comprehension performed on par with students with typical reading. This indicates that early on in writing development, producing longer texts may be a challenge for this group of students, but not anymore at the end of primary schooling. This is in line with the outcome in, for example, Cragg and Nation (2006), showing that text length was not an issue in a group of 9- to 11-year-old students with poor reading comprehension. The current study indicates that this may vary depending on the age and stage of writing development, at least in this group of students with poor comprehension. However, for the students with mixed difficulties, producing longer texts continued to be a challenge throughout primary school. This is probably due to their combined difficulties on both the transcription level with word reading and spelling (Berninger et al., 2008; Sumner et al., 2013, 2016) as well as with comprehension and higher-level aspects of writing such as organisation, translating ideas into words and sentences, and developing the genre elements in a narrative assignment (Carretti et al., 2013, 2016).

Regarding vocabulary diversity, there was a significant main effect for grade (time). However, raw scores reveal that it was mainly the three profiles with poor reading abilities that made the most substantial progress from Grade 3 to 6, with the students with poor word reading closing most of the gap to the students with typical reading. This is in line with the assumption that students with poor word reading and typical reading comprehension have relative strengths in oral language, for example, in vocabulary, making it easier to translate ideas into language when writing (Sumner et al., 2013, 2016). However, contrary to the studies by Sumner and colleagues, spelling (assessed in a dictation task in Grade 2) did not correlate with vocabulary diversity in either Grade 3 or 6 (see Appendix A). This may be a result of the semi-transparent Swedish orthography not interfering with lexical choices in the writing process to the same extent as in English, at least not this early in writing development. See, e.g., a study by Wengelin (2007) showing that spelling does interfere with lexical choices for Swedish students with dyslexia later in writing development in secondary school.

The developmental patterns for narrative quality indicate positive progress from Grade 3 to 6 for all reading profiles with a significant main effect for grade (time). However, looking at the raw scores, the profiles with poor word reading and mixed difficulties made very modest progress in raw scores from Grade 3 to 6 (1 to 1.5 points), while the profiles with typical reading and poor reading comprehension made more substantial progress (approx. 4 points). Again, the students with poor reading comprehension showed a more positive developmental pattern than the students with poor word reading and the students with mixed difficulties. However, all three profiles with poor reading are still judged as having emergent skills in narrative writing in Grade 6, indicating that narrative writing continues to be challenging for all profiles with poor reading, especially so for the students with mixed difficulties who had the lowest mean score.

4.2. The Characteristics of Written Text Production in Grades 3 and 6

A one-way between-groups analysis of variance was conducted to explore the characteristics of written text production and possible differences between reading profiles in Grades 3 and 6 in more detail.

In Grade 3, there was a significant difference between the reading profiles in text length and vocabulary diversity, but not in narrative quality. Students with typical reading wrote significantly longer texts than the students with poor reading comprehension and mixed difficulties. They also wrote texts with a higher vocabulary diversity than students with mixed difficulties, while students with poor word reading performed on par with the students with typical reading.

The results indicate that early in writing development, students with poor word reading perform on the same level as typical readers in text length, vocabulary diversity, and narrative quality, while for students with poor reading comprehension, it is challenging to produce longer texts and use a more varied vocabulary, and especially so for the students with mixed difficulties.

The positive outcome for the students with poor word reading is not in line with previous studies of the English orthography. For example, in studies by Sumner with colleagues (Sumner et al., 2013, 2016), it was found that spelling difficulties may interfere with the writing process (leading to frequent pauses) as well as affecting vocabulary diversity and overall text quality for students with dyslexia. In the current study, there were no significant correlations between spelling ability in Grade 2 and vocabulary diversity in writing in Grades 3 and 6; this was also true for text length and narrative quality (see Appendix A). This finding may be due to the Swedish semi-transparent orthography not posing the same challenges at the transcription level as is the case in an opaque orthography like English. In a study by Pinto et al. (2015) in the transparent Italian orthography, spelling was found to have only an indirect mediating effect on the relationship between oral narrative skills in kindergarten and written narrative skills in Grade 1. The continuity from oral narrative production to written narrative production was only shortly disrupted in Grade 1, but no longer in Grade 2 when the participants had learned to master spelling. The authors suggest that the reason may be that children master spelling earlier in a transparent orthography, and hence spelling will only have a brief taxing effect on working memory capacity and thereby also a brief influence on overall structure, cohesion, and coherence when writing narratives. This study confirms the pattern that spelling does not seem to explain variation in written text quality in a semi-transparent orthography in early writing development.

However, the poor performance of students with poor reading comprehension regarding text length and vocabulary diversity in Grade 3 is in line with previous studies. This finding confirms that poor reading comprehension is probably related to underlying difficulties in oral language and vocabulary (Carretti et al., 2013, 2016; see also Graham et al., 2020b) and often leads to poor linguistic complexity in writing (Carretti et al., 2013, 2016; see also Graham et al., 2020b).

The lack of significant differences between the different profiles in narrative quality in Grade 3 is maybe surprising but seems to be related to the narrative writing task being equally challenging for the students with typical reading development. The composite score of overall narrative quality based on the seven categories showed that the narrative quality was “emerging” (on a scale from minimal/immature (1)—emerging (3)—proficient (5)) for all four reading profiles. In Grade 3, students are still in the phase of learning to write. In the curriculum at the time of the study (Swedish National Agency for Education, 1994), the main focus was on mechanics (writing letters, using basic rules for interpunction, and spelling high frequency words), and as regards narrative writing, on being able to write a story with a clear introduction, simple storyline, and a clear ending. This description is in line with the criteria for emerging narrative writing skills in the NSS. Therefore, all reading profiles seem to perform very much in line with the requirements of the curriculum, and at this stage of writing instruction, the students with poor reading abilities are performing as expected in Grade 3.

In Grade 6, the only significant difference between reading profiles was found in narrative quality. The students with typical reading had a significantly higher mean score in narrative quality than the students with mixed difficulties. However, all three profiles with poor reading abilities had mean scores indicating “emerging narrative writing skills”, while the students with typical reading approached “proficient narrative writing skills”. This indicates that writing proficiently in the narrative genre is a challenge over time for students with poor reading abilities when the requirements are increasing. This finding is also in line with previous studies indicating that students with poor word reading may experience writing difficulties also in higher-level aspects of written text quality (see the meta-review in Graham et al., 2020a), even though this is probably a secondary effect from challenges at the transcription level (Sumner et al., 2013, 2016).

For the students with poor reading comprehension, the challenges with written narrative quality are probably a primary effect from underlying difficulties in oral language (such as vocabulary and grammar) and in higher-level processes (such as inferencing and comprehension monitoring), which makes it challenging to translate ideas into language, plan and revise a text on a discourse-level. Such difficulties have been identified in several previous studies among students with poor reading comprehension in various age groups (Carretti et al., 2013, 2016; Cragg & Nation, 2006).

No significant differences between reading profiles were found for text length and vocabulary diversity, however the post-hoc comparisons indicated a tendency towards significant differences between the profiles with typical reading and mixed difficulties in both text length and vocabulary diversity.

Overall, the students with mixed difficulties performed the lowest scores on all three written text measures while the students with typical reading performed the highest scores in both Grade 3 and 6, indicating that the group with mixed difficulties is the most vulnerable one of the three profiles with poor reading abilities.

4.3. Study Limitations

There are some limitations in the study design that need to be considered when interpreting the findings. We only had access to one writing assignment for each grade. Several writing assignments per grade would have provided a more solid empirical base for drawing conclusions about the students’ writing performance across different tasks. In future studies, it would also be valuable to include a similar oral narrative assignment to be able to evaluate if the difficulties with vocabulary diversity and narrative quality are restricted to the written modality or evident also in the oral modality. To fully understand the participants with poor reading abilities and their possible challenges when writing, it would have been valuable to include information on their handwriting abilities, oral language comprehension, working memory capacity, and if they had any other learning difficulties apart from reading. The limited number of participants in each reading profile also highlights a need to interpret the outcome with caution. In future studies, it would also be valuable to know more about how much and what kind of writing instruction the participants received during primary school.

5. Conclusions and Implications

Students with mixed reading difficulties seem to be the most vulnerable group, as they perform significantly below typical readers on almost all written text measures in both Grades 3 and 6. This indicates that this group of students should be monitored extra carefully regarding their writing development. They seem to need extra support when it comes to translating ideas into language, e.g., by receiving support in vocabulary selection and grammar, as well as in developing the narrative elements when writing a story. Furthermore, this group of students seems to have more severe spelling difficulties (see Table 1), indicating a need for extra support regarding transcription level skills.

Students with poor reading comprehension make substantial progress in raw scores from Grade 3 to 6 in text length and narrative quality compared to the two other groups with poor reading abilities. However, they still struggle with narrative text quality in Grade 6, which confirms that these students need support with discourse-level aspects in writing, such as text structure and the development of ideas, content, and narrative genre elements.

Students with poor word reading in Grade 2 perform on par with typical readers on all written text measures in Grade 3, but in Grade 6 they do struggle with narrative quality, indicating a need for support in developing narrative elements in their writing.

Spelling does not seem to interfere with writing development in the current sample. However, this should not be interpreted as a lack of need to include spelling in early writing instruction in semi-transparent orthographies. More studies are needed also in older age groups and with a larger number of participants to expand the knowledge about the role of spelling in writing development in (semi-)transparent orthographies.