Writing development models: Individual skills in early text production

Given the significance of acquiring writing skills, it is imperative to explore current research approaches for understanding writing development. Developmental models of writing, such as the simple view of writing [9] and its extension, Berninger’s not-so-simple view of writing (NSVW) [10,11], suggest that written text production is influenced by three closely connected component skills: transcription skills, oral language skills, and executive functions. These components play a crucial role in shaping the process of written text production. These skills interact dynamically with working memory during the writing process, shaping its progression [12]. In the model NSVW, transcription skills refer to the mechanical process of converting ideas into written symbols, such as spelling and handwriting. Oral language proficiency plays a crucial role in the text generation phase, facilitating the conversion of ideas into linguistic forms stored in working memory. Executive functions oversee and regulate the writing process, facilitating attention, planning, and revision. Thus, the early stages of writing require transcription skills, executive functions and oral language skills [13,14].

As mentioned above, transcripton skills consist of handwriting and spelling. Handwriting involves motor movements, while spelling entails symbol retrieval and assembly [15,16]. These skills are vital for quality and productivity in writing development as early as in kindergarten [17], aiding language-to-text transformation [10,18,19]. The development of transcription skills continues throughout schooling, contributing to enhanced writing proficiency [7,20,21]. Although both skills positively correlate with writing performance, their impact varies across educational levels [20]. In primary school, they directly influence writing, while in middle school, their influence is indirect through planning and self-efficacy [22].

Previous research has produced mixed findings about the relationships among handwriting, spelling, and composition. Although comprehensive analyses by Santangelo and Graham [23] revealed notable improvements in writing quality, length, and fluency following handwriting fluency training, the longitudinal connections between handwriting and composition have been somewhat limited [24]. For example, Graham and Santangelo [25] observed a minimal impact of spelling instruction on overall writing performance, in addition to spelling accuracy, which contrasts with the findings of Abbott et al.’s [24] longitudinal investigation where spelling emerged as a robust predictor of composition quality. As noted by Vieira et al. [4], these contrasting findings may arise from methodological disparities across studies, encompassing participant characteristics, evaluation metrics, study design, linguistic factors, cultural influences, and pedagogical methodologies.

According to the not-so-simple view of writing, both text generation and transcription skills are core processes in writing. Oral language skills, which are essential to the text generation process, serve as significant predictors of written expression [26–31]. In addition to their direct impact on text generation, oral language skills also exert an indirect impact by reducing the processing demands on working memory when fluency in idea translation is achieved [32,33]. Evidence supports the contribution of oral language skills to writing in children across various grades, with syntax instruction shown to enhance writing performance [34,35].

Notably, the findings have been mixed across different grades in opaque orthographies. While some studies have reported this relationship in first-grade children [28], the majority have found it from the 2nd or 3rd grade onward [1,18,26,36]. In a classic study conducted by Juel et al. [28], ideation, as assessed by a storytelling task, contributed significantly to writing in Grades 1 and 2. However, Abbott and Berninger [18] and Berninger and Abbott [26] found that oral language skills significantly contributed to English written composition in grades 2 to 3 and grades 6 to 7 but not in grade 1 or intermediate grades. Yeung et al. [1] observed similar results in Chinese written composition, with Grade 1 oral language skills predicting Grade 2 composition but not Grade 1. In their meta-analysis, Kent and Wanzek [36] examined the correlation between various proposed component skills, including handwriting fluency, spelling, reading, and oral language, and writing outcomes. They evaluated both the quality of students’ compositions and the quantity of their written output. The results indicated that oral language exhibited the weakest relationship with writing quality, explaining slightly more than 10% of the variance. Nevertheless, longitudinal evidence regarding the contribution of oral language skills to writing is limited [16]. While some studies suggest that oral language skills assessed in the early grades may predict narrative writing growth over time [35], others find a relationship between oral vocabulary proficiency and writing quality and quantity from the outset [37]. These findings underscore the complex interplay between oral language skills and written composition, highlighting the need for further research in this area.

While much research has focused on transcription and oral language skills, there is a need for further exploration of the role of executive functions in writing development [16]. Although several studies have been conducted in this area, much less is known about the impact of executive functions on writing outcomes compared to that of transcription and oral language skills [13,14,16,38]. Executive functions (EFs), including working memory, inhibitory control, and cognitive flexibility, are essential cognitive processes for goal-directed behavior and problem solving [39]. EF encompasses both higher-level cognitive functions involved in planning, translating, reviewing, and revising to regulate the writing process, as well as foundational skills supporting these processes [40,41]. Foundational EFs are considered cognitive subcomponents of a single supervisory attentional mechanism, typically involving inhibition, working memory updating, and shifting [42]. Inhibition involves selectively attending to specific stimuli while suppressing attention to others (selective attention), maintaining task focus despite distractions (sustained attention), and inhibiting prepotent responses. Working memory updating entails retaining pertinent information, such as representational structures, mentally manipulating it, and taking actions based on it. Finally, shifting, or cognitive flexibility, involves adapting swiftly and flexibly to changing situations, such as tasks or mental sets [43].

These functions are critical for writing and influence various aspects of composition, including transcription and text quality [44–46]. Working memory facilitates the storage and manipulation of information during composition, aiding in planning, language generation, and review [45]. Inhibitory control helps writers suppress irrelevant ideas and select appropriate words and structures [47], while cognitive flexibility allows for switching between perspectives [16].

However, studies examining the relationship between EFs and writing in primary school children have presented mixed findings [13,16,35,38,48–50]. For instance, WM measured in Grade 2 predicted text quality six months later, but its direct relation to writing was not observed in Grade 1 [13,50]. Inhibitory control is longitudinally related to syntactic complexity in narratives [16], while inhibitory control and cognitive flexibility contribute to spelling and written expression [48]. Moreover, attentional processes, as emphasized in studies by Kent et al. [8] and Hooper et al. [35], are pivotal in early writing, facilitating students’ focus on relevant tasks, retention of prior content for future composition, and filtration of irrelevant information. Valcan et al. [38] analyzed the contribution of EF to handwriting composition in Year 2 children using the HTKS, a task that integrates multiple EF components (i.e., working memory, inhibition, and attention shifting). They found that handwriting automaticity and spelling mediated the relationship between children’s EF and handwriting composition (i.e., fluency and quality). Kent et al. [49] demonstrated that attention predicted both composition quality and fluency in the first grade, while Hooper et al. [35] identified attention/executive function as the primary predictor of first-grade writing. These findings emphasize the significance of integrating attentional mechanisms into our understanding and prediction of early writing development, justifying its inclusion as an EF latent factor in our study. In contrast, in recent studies with primary and middle graders, both inhibitory control and cognitive flexibility failed to predict text quality [13]. Moreover, findings regarding the most predictive EF vary across studies, indicating a lack of consensus in the field [35,51].

One reason for the failure to establish a clear link between EF and the quality of written production may stem from the initial unitary development of EF, which gradually differentiates into distinct components [16,35,38]. EF functions as a unified construct during the ages of 3 to 8 [52]. It evolves into three interrelated but distinct components as children grow older [42]. This developmental trajectory suggests that analyzing EF as a unified construct may be more advisable, especially when dealing with kindergarten students, as in the present study.

Examining transcription, oral language, and executive function skills on dimensions of early writing production in a shallow orthography

First, it has been proposed that the relationships among handwriting fluency, spelling proficiency, and narrative writing may be stronger in opaque orthographies such as English [7,9,24,53–55] than in shallow orthographies like Spanish. Opaque orthographies exhibit less predictable correspondences between letters and sounds, potentially increasing the demand for spelling during text production. Conversely, in shallow orthographies, where the correspondences between letters and sounds are more consistent, the demand for spelling during text production is reduced (e.g., in Dutch, [16,55,56]; in Finnish, [57]; in Italian, [3,5,58]; in Icelandic, [59]; in Norwegian, [21]; in Portuguese, [4,22]; in Spanish, [60–63]; and Turkish, [30,32]). As noted by Longobardi et al. [3], studies conducted in shallow orthographies generally do not support the role of the transcription process in writing. Taken together, this body of research suggests that children acquiring languages with more uniform orthographies exhibit decreased dependence on transcription skills, specifically word-spelling abilities, compared to those learning languages with irregular orthographic patterns.

Oral language skills play a critical role in the development of writing [35], which requires children to translate ideas into linguistic forms, organize thoughts coherently, and employ an extensive vocabulary [54,64]. In languages with shallow orthographies, the demands on spelling during text production are reduced due to consistent correspondences between letters and sounds [5,65]. Studies conducted with Italian children, such as the research by Arfé et al. [5], demonstrate that oral language skills contribute to microstructural productivity, complexity, accuracy, and overall text quality among Italian second and third graders. Receptive grammar skills have a more significant impact on the macrostructural level of written composition than spelling. Similarly, oral narrative competence directly affects the written narrative production of second graders [58]. In a longitudinal study with Icelandic children, Oddsdóttir et al. [59] assessed text generation skills in first grade and narrative and informational text composition in second and fourth grades. They found that children activated their text generation skills earlier in narrative writing than in information text writing. Given that our study followed children from kindergarten to grade 1, we chose to focus on narrative writing.

Research has shown that while the role or influence of oral language skills remains consistent across different languages [30,35], differences emerge in the timing of this influence, with an earlier onset observed in shallow orthographic systems due to children’s precocious automation of transcription skills [66]. In opaque orthographies, the scenario is different. For instance, Hooper et al. [35] reported that oral language skills assessed prior to kindergarten entry predicted the developmental trajectory of narrative writing from third to fifth grade, suggesting that the predictive power of oral language skills emerges in later grades. In a longitudinal study that followed New Zealand children from kindergarten to grade 1, McDonald et al. [66] found that the link between oral language skills and writing proficiency emerged later once transcription skills were more consolidated. Similarly, in the studies cited above (e.g., [1,18,26]), the lack of significance of oral language skills in Grade 1 was attributed to the need to first overcome the constraints imposed by transcription skills before oral language skills could exert influence.

Although there is less emphasis on spelling accuracy in languages with shallow orthographies, young writers still face challenges arising from the intricacies of grammar and morphology, as highlighted by previous studies [5,67,68]. Moreover, research suggests that children learning to write in languages with shallow orthographies may rely more heavily on executive functions to manage the cognitive demands of text production, particularly in navigating the complexities of grammar and morphology [4,13,16]. For instance, Drijbooms et al. [16] observed a longitudinal relationship between EFs, specifically inhibition and planning, and syntactic complexity in written narratives among Dutch students in grades 4 and 6. Similarly, Cordeiro et al. [13] reported that among Portuguese Grade 2 students, working memory and planning at the beginning of the school year made a significant contribution to text quality six months later, surpassing the influence of transcription skills. Vieira et al. [4] reported that EFs were associated with text quality in Grade 3, pointing to the central role of EFs in written expression, especially among older children.

Several previous studies conducted in Spanish have examined how transcription and oral language skills contribute to the productivity and quality of written compositions in young children [60–62]. For instance, in a study that involved Spanish students in grades 1 and 2, Jiménez and Hernández-Cabrera [60] engaged children in a conversation about their interests, favorite games, hobbies, among other topics, and after the child described one or two situations, they were asked to compose two sentences about the conversation. The total number of words written correctly within one minute during these tasks served as an index of writing fluency. The results showed that handwriting fluency and spelling directly contributed to writing fluency in sentence composition. Notably, handwriting fluency contributed less variance than spelling to writing fluency in this study. Spelling contributed to both narrative writing and writing fluency. Jiménez and Barrientos [61] studied the impact of transcription skills, assessed through graphonomic measures, on text generation proficiency in primary school students in Spain. They used a longitudinal design with 278 students across three cohorts (cohort 1: 1st-2nd-4th grade; cohort 2: 2nd-3rd-5th grade; and cohort 3: 3rd-4th-6th grade). Four multigroup structural equation models were used to examine direct pathways from graphonomic measures (e.g., pressure, speed, pauses, and road length) to text generation (e.g., length, fluency, planning, revision, and organization). The results showed that handwriting significantly influenced text production in Cohort 1 (early grades) but not in Cohort 2 (intermediate grades) or Cohort 3 (upper grades). Recently, Rodríguez et al. [62] explored the relationships among transcription skills, oral language abilities, and writing quality and productivity in Spanish-speaking kindergarten children. The results showed that transcription skills strongly predicted both writing quality and productivity, while oral language only predicted writing quality. However, the limitations of the previous studies on Spanish include the sole focus on transcription skills without assessing handwriting [62], reliance on observed variables instead of latent variables [62], and the absence of oral language and executive function assessments (e.g., [60,61]). The present study aimed to address these limitations by incorporating handwriting assessment while considering oral language skills and executive functions in a longitudinal design.

The present study

This study examined the contribution of transcription skills, oral language abilities, and executive functions in kindergarten to written production in grade 1 among Spanish children. The Spanish language exhibits a relatively shallow orthographic system compared to other languages (such as French and English), where morphology is more extensively represented [64]. Despite respecting some aspects of morphology, such as the use of ’h’ in verb forms such as "haber," Spanish orthography is relatively limited in its representation of morphology. On the other hand, Spanish orthography is characterized by its consistent and efficient representation of phonological information, making it relatively easier to learn to read and spell compared to languages with more complex orthographic systems [69].

Previous studies in languages with shallow orthographies have shed light on the unique challenges posed by these writing systems. While there is less emphasis on spelling accuracy in shallow orthographies, young writers still encounter difficulties stemming from the complexities of grammar and morphology [5,67,68]. Similarly, the role of transcription skills, oral language skills, and executive functions has also been studied in the development of writing in shallow orthographies. However, comprehensive investigations encompassing all these facets within a single study, especially targeting kindergarten-aged children, are notably lacking. Therefore, the present study not only extends previous research to a younger age group but also provides new insights by examining how these three sets of skills influence different dimensions of written production (quality, productivity and syntactic complexity) in a shallow orthography. We expected that in a shallow orthography such as Spanish, transcription skills may be less predictive of textual dimensions such as productivity, quality and syntactic complexity than oral language skills and executive functions. In other words, considering the nature of the orthography, oral language skills and executive functions are expected to play more significant roles in shaping written production.

Participants

A total of 198 Spanish students participated in the study. Participants were recruited from seven schools in the Canary Islands during their final year of kindergarten. Children with special educational needs, such as sensory, motor, or intellectual impairments, as well as those with established diagnoses such as Down syndrome or autism spectrum disorder, were excluded from the analysis. The final sample comprised 191 children (96 boys and 95 girls; kindergarten age M = 5.40; SD = 0.28; 1st grade age M = 6.92; SD = 0.28). The sample included students from both state and nonstate schools, ensuring socioeconomic diversity. Approximately 50.2% of the families in the sample provided responses regarding their educational and socioeconomic levels. For mothers, 27.1% held a secondary education (Bachillerato) diploma, 29.2% held a university degree, and 13.5% completed vocational training. For fathers, 30.2% held a secondary education (Bachillerato) diploma, 18.8% held a university degree, and 14.6% completed vocational training. With respect to socioeconomic status, the majority of the mothers (56.3%) reported middle-income levels, while 24% fell into the low-income category. A smal percentage, 6.3%, reported having a high income, and 9.4% of the mothers reported having no income. Similarly, for fathers, the largest group (61.5%) also reported middle income levels. Fathers in the low-income category constituted 20.8%, while those in the high-income category made up 4.2%. Additionally, 1% of fathers reported having no income. All participants were native speakers of Spanish, and due to the cultural and linguistic context of the Canary Islands, the sample consisted exclusively of individuals who spoke Spanish as their first language.

Materials

Assessments of executive functions, transcription, and oral language were conducted in kindergarten. Subsequently, when the students reached the the first grade, assessments of narrative writing were conducted.

Procedure

This study followed the guidelines of the Research Ethics Committee at Universidad de La Laguna, which included the submission of an informed consent model to the parents of participating students. Prior to commencing the study, approval was obtained from the schools and families involved. The principal investigator provided the informed consent model, detailing the research objectives and ensuring anonymity and confidentiality, with the provision for participants to withdraw at any time. Written consent was obtained from the directors of the participating schools, who facilitated the distribution of the informed consent forms to families and collected them upon completion. Data collection in kindergarten commenced in early May 2022 and concluded by mid-month. For first grade, data collection took place in early May 2023 and also concluded by mid-month. Eighteen examiners were trained in task administration via an app designed to assess transcription skills, oral language skills, and executive functions among kindergarten students. The software was developed in Unity 2019.4.20f1, utilizing the Android platform support package. It is compatible with.NET 4.x API and requires at least Android 4.4 KitKat API for operation on mobile devices or tablets. Writing tasks in the software necessitate a Samsung tablet equipped with a stylus. Furthermore, the application requires microphone access for oral tasks. Optionally, the program supports data transmission to a server for result organization and download, developed using Django web framework version 4.2. The assessments conducted in kindergarten spanned three separate sessions across different days. Task order was counterbalanced among participants, who were allocated into three groups. Training encompassed task administration and assessment of both oral and written narrative texts. Recorded oral narratives were captured using the app and transcribed for subsequent analysis. An Olympus DM-620 digital voice recorder served as a backup in case of recording issues. A second session focused on evaluating and scoring recorded oral and written narratives, with guidance provided on entering assessment data into an Excel template. Training included using writing protocols as examples. Each examiner received a manual with correction criteria to ensure objectivity. Consistency among examiners was ensured by using the same correction protocols; results were reviewed collectively. Disagreements prompted revisiting correction criteria, iteratively applied to multiple examples. Tests were administered in spring, coinciding with the end of the school year, during school hours, with students informed of study objectives and confidentiality. Tasks assessing written narration at the end of first grade were administered using paper and pencil.

Data analysis

Data were analyzed using structural equation modelling (SEM) with the "Lavaan" package [97] in the R statistical software [98], assisted by ULLRToolbox [99]. Latent variables were constructed to represent exogenous constructs such as transcription, oral language, and executive functions, while productivity, quality, and syntactic complexity were treated as endogenous variables. The maximum likelihood robust (MLR) method was employed for all estimated models. To address the research objectives, SEM models were developed and evaluated using multiple indices, including chi-square statistics, the comparative fit index (CFI), the Tucker‒Lewis index (TLI), the root mean square error of approximation (RMSEA), and standardized root mean square residuals (SRMRs). The acceptable model fit criteria included RMSEA values below .08, CFI and TLI values equal to or greater than .95, and SRMR values equal to or less than .05, following recommendations by Hu and Bentler [100]. Additionally, CFI and TLI values exceeding .90 were considered acceptable [101].

Model 1: Text productivity

The estimated model showed good internal consistency, with an Omega coefficient of 0.809. For executive functions, its indicators attention (0.53), digit spam backward (0.63), oral cloze task (0.44), inhibitory control (0.56), and cognitive flexibility (0.20) all exhibited significant standardized factor loadings. For transcription skills, the indicators phoneme isolation (0.78), letter copying (0.39), name writing (0.57), picture word writing (0.61), and phoneme segmentation (0.60) showed significant standardized factor loadings. With respect to narrative competence, narrative structure (0.51), unique word written (0.96), T-units (0.91), and total number of written words (1.00) were also robust indicators of the latent construct. Moreover, the latent variable productivity demonstrated significant paths to its indicators unique words written (0.95), correctly written sequences (0.96), written word fluency (0.90), and total number of written words (0.70). As shown in Table 3, these significant factor loadings suggest substantial connections between observable and latent variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Unstandardized and standardized path coefficients for transcription skills, narrative competence, executive functions, and productivity.

https://doi.org/10.1371/journal.pone.0315748.t003

The model estimates indicate a good fit to the observed data. Notably, the chi-square statistic was significant (χ² = 202.164, p < 0.001), suggesting a discrepancy between the proposed model and the observed data. However, this statistic tends to be sensitive to sample size, so it is important to consider other fit indices [102]. The NFI showed a value of 0.908, indicating that the model explained 90.8% of the variance and covariance of the observed data. Similarly, the NNFI had a value of 0.958, while the CFI and TLI had values of 0.964. These indices surpassed the recommended threshold of 0.95, suggesting a good fit of the model. The RMSEA was 0.054, with a 90% confidence interval between 0.040 and 0.069, indicating an acceptable model fit. Additionally, a scaled version of the chi-square test (scaled) demonstrated a good fit (χ² = 212.998, p < 0.001; RMSEA = 0.057. In summary, these results suggest that the proposed model fit well with the observed data, explaining substantial variance and covariance in the latent and observed variables.

In terms of structural paths, transcription skills significantly predicted text productivity with a standardized path coefficient of 0.381 (p < 0.01), indicating a positive influence of transcription on productivity. Similarly, EF also significantly predicted text productivity with a standardized path coefficient of 0.271 (p < 0.05), suggesting a positive influence of executive functions on productivity. However, the path from the narrative competence to the text productivity (β = .009) (p = .90) was not statistically significant (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. This graph shows a structural model of the relationship between executive functions, transcription skills, narrative competence, and productivity (Model 1).

Circles represent factors (i.e., latent variables), rectangles represent indicators (i.e., observed variables), black arrows represent direct paths, and gray arrows represent covariances. EF = Executive functions; CF = cognitive flexibility; IC = inhibitory control; OCT = oral cloze task; DSB = digit spam backward; AT = attention; TR = transcription skills; PS = phoneme segmentation, WPI = picture word writing; NW = name writing; LC = letter copying; PI = phoneme isolation; NC = narrative competence; TNW = total number of words; TU = T-units; UW = unique words written; NS = narrative structure; PRO = productivity; TNWW = total number of written words; WWF = written word fluency; CWS = correctly written sequences; UWW = unique words written.

https://doi.org/10.1371/journal.pone.0315748.g002

Model 2: Text quality

As shown in Table 4, the measurement model produced significant standardized factor loadings for all latent variables and their respective indicators. A total omega coefficient of 0.887 was observed, indicating good overall reliability of the model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Unstandardized and standardized path coefficients for transcription skills, narrative competence, executive functions, and quality.

https://doi.org/10.1371/journal.pone.0315748.t004

For narrative competence, strong factor loadings were observed for its indicators, namely, narrative structure (0.51), unique words (0.96), T-units (0.92), and total number of words (1.00). For transcription, significant factor loadings were found for phoneme isolation (0.78), letter copying (0.43), naming writing (0.56), picture word writing (0.59), and phoneme segmentation (0.61). Likewise, for EF, significant factor loadings were observed for attention (0.52), digit spam backward (0.60), oral cloze task (0.43), inhibitory control (0.57), and cognitive flexibility (0.21). Notably, the factor loading for cognitive flexibility was lower than other indicators, though it was still statistically significant. Finally, for text quality, strong factor loadings were evident for written narrative structure (0.79), written sentence fluency (0.17), percentage of correctly written words (0.35), and number of written causal connectors (0.45). The significant factor loadings suggest robust relationships between the latent variables and their indicators. The model fit results obtained through the MLR method indicate a satisfactory overall fit. The chi-square statistic was significant (χ² = 184.655, df = 129, p < 0.001), suggesting some deviation between the proposed model and the observed data. However, given the sensitivity of the chi-square statistic to sample size, additional fit indices were considered. The NFI was 0.871, while the NNFI was 0.948. Furthermore, the CFI and TLI were 0.956 and 0.948, respectively. Despite the TLI slightly below the recommended threshold of 0.95, both indices indicate a good model fit. The MFI was 0.862. Additionally, the RMSEA was 0.048, with a 90% confidence interval between 0.031 and 0.063, suggesting an acceptable model fit. A scaled version of the chi-square statistic yielded a value of 195.723 (df = 129, p < 0.001), with an RMSEA of 0.051 and a 90% confidence interval between 0.036 and 0.065. Regarding the structural model, the relationships between text quality and both transcription (β = 0.344, p < 0.05) and narrative competence (β = 0.184, p < 0.05) were found to be statistically significant. However, the association between text quality and EF did not reach statistical significance (β = 0.291, p = 0.156) (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. This graph shows a structural model of the relationship between executive functions, transcription skills, narrative competence, and quality (Model 2).

Circles represent factors (i.e., latent variables), rectangles represent indicators (i.e., observed variables), black arrows represent direct paths, and gray arrows represent covariances. EF = Executive functions; CF = cognitive flexibility; IC = inhibitory control; OCT = oral cloze task; DSB = digit spam backward; AT = attention; TR = transcription skills; PS = phoneme segmentation, WPI = picture word writing; NW = name writing; LC = letter copying; PI = phoneme isolation; NC = narrative competence; TNW = total number of words; TU = T-units; UW = unique words written; NS = narrative structure; QUA = quality; NWC = number of written casual connectors; PCW = percentage of correctly written words; WSF = written sentence fluency; WNS = written narrative structure.

https://doi.org/10.1371/journal.pone.0315748.g003

Model 3: Syntactic complexity

Table 5 shows the unstandardized and standardized path coefficients for transcription skills, narrative competence, executive functions, and syntactic complexity. The Omega total coefficient was 0.833, further supporting the internal consistency of the measured variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Unstandardized and standardized path coefficients for transcription skills, narrative competence, executive functions, and syntactic complexity.

https://doi.org/10.1371/journal.pone.0315748.t005

The measurement model produced significant standardized factor loadings for all latent variables and their respective indicators. For narrative competence, significant standardized factor loadings were observed for narrative structure (0.52), unique words (0.96), T-units (0.91), and total number of words (1.00), indicating strong associations between these indicators and the latent variable. Transcription skills demonstrated robust standardized factor loadings for phoneme isolation (0.79), letter copying (0.38), name writing (0.54), picture word writing (0.61), and phoneme segmentation (0.61), highlighting their substantial contributions to the transcription skills construct. Executive function exhibited significant standardized factor loadings for attention (0.55), digit span backward (0.58), oral cloze task (0.46), inhibitory control (0.56), and cognitive flexibility (0.20), indicating their influence on the executive function latent variable. Syntactic complexity was strongly influenced by grammatical structure (0.71) and written T-units (0.64), underscoring their importance as indicators of syntactic complexity in the model.

The model fit statistics obtained through the MLR method indicate a satisfactory overall fit. The chi-square statistic was significant (χ² = 139.668, df = 98, p < 0.004), indicating some deviation between the proposed model and the observed data. However, considering the sensitivity of the chi-square statistic to sample size, additional fit indices were evaluated. The NFI had a value of 0.897, while the NNFI was 0.959. Furthermore, the CFI and TLI were 0.966 and 0.959, respectively, surpassing the recommended threshold of 0.95, indicating a good model fit. The MFI was 0.893. Additionally, the RMSEA was 0.048, with a 90% confidence interval between 0.028 and 0.065, suggesting an acceptable model fit. A scaled version of the chi-square statistic yielded a value of 145.112 (df = 98, p < 0.001), with an RMSEA of 0.050 and a 90% confidence interval between 0.032 and 0.067. These findings indicate that the proposed model aligned well with the observed data, explaining substantial variance and covariance among the latent and observed variables. The structural model revealed several significant pathways linking latent variables. Specifically, the standardized estimates indicated that narrative competence significantly predicted syntactic complexity (β = 0.191, p < .05), while transcription had a beta value of 0.202, but this effect was not statistically significant (p = 0.261). However, executive functioning demonstrated a stronger and more significant association with syntactic complexity (β = 0.460, p < .01), suggesting a robust influence of executve functioning on syntactic complexity (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Structural model of the relationship between executive functions, transcription skills, narrative competence, and syntactic complexity (Model 3).

Circles represent factors (i.e., latent variables), rectangles represent indicators (i.e., observed variables), black arrows represent direct paths, and gray arrows represent covariances. EF = Executive functions; CF = cognitive flexibility; IC = inhibitory control; OCT = oral cloze task; DSB = digit spam backward; AT = attention; TR = transcription skills; PS = phoneme segmentation, WPI = picture word writing; NW = name writing; LC = letter copying; PI = phoneme isolation; NC = narrative competence; TNW = total number of words; TU = T-units; UW = unique words written; NS = narrative structure; SC = syntactic complexity; GS = grammatical structure; WTU = written T-units.

https://doi.org/10.1371/journal.pone.0315748.g004

For all three models, we observed robust covariance between the latent variables of transcription skills and executive function constructs, indicating that elevated scores in executive functions often align with higher scores in transcription.

Educational implications

The findings of our study, which examined Spanish, a shallow orthography, have significant educational implications. Our results underscore the importance of transcription skills, oral language abilities, and executive functions in early writing development. While transcription skills play a crucial role in generating written text, oral language skills and executive functions also have a substantial influence on different dimensions of writing. As a result, writing instruction programs should address not only writing practice but also oral and executive skills to enhance students’ writing quality and syntactic complexity.

Our findings support the need to integrate teaching practices that foster the development of transcription skills from the early stages of learning to write. Since transcription skills are closely related to productivity and writing quality, educators should incorporate writing activities that encourage handwriting fluency and spelling accuracy, which in turn increase text production and quality. Furthermore, the significant influence of oral language skills on writing quality and syntactic complexity highlights the need of promoting oral language development in the classroom. Educational programs should include activities that promote vocabulary development, grammatical understanding, and oral storytelling, as these skills enrich students’ writing and enhance their ability to express ideas coherently and sophisticatedly. Finally, educators can implement activities that specifically target executive functions, such as storytelling activities that require students to sequence events in order, problem-solving tasks, and activities to improve attention and self-regulation. By providing a variety of engaging activities that target transcription skills, oral language skills, and executive functions, educators can create a learning environment that promotes the development of proficient writing skills.

Limitations and future research

The findings of our study must be considered in the context of its limitations. One notable constraint was the relatively small sample size, which may have hindered our ability to detect smaller effects and reduced the generalizability of the findings. Additionally, the study was conducted in a specific geographical region (the Canary Islands), which may limit the generalizability of the results to other regions of Spain or beyond. Future research should aim to include samples from different regions to obtain more robust conclusions and account for potential regional differences in educational practices or student characteristics.

Another limitation is the absence of a non-verbal intelligence measure, which is often included in research related to executive functions. We also did not assess other linguistic skills such as orthographic knowledge and morphological awareness. These measures are particularly relevant in the context of a shallow orthography, as they may play a crucial role in the development of writing abilities. Acknowledging these limitations, we suggest that future research incorporate these variables to provide a more comprehensive understanding of the cognitive and linguistic factors influencing writing performance.

Furthermore, there is evidence suggesting that foundational skills (i.e., transcription, text generation, and executive functions) affect narrative and descriptive texts differently, as children activate their text generation skills for narrative writing earlier than for informational texts [59]. Therefore, our findings cannot be generalized to other text genres. Future studies examining the role of these skills in writing informational texts will allow us to understand whether the relationships identified in our study are universal or limited to a specific genre [4]. Furthermore, the study did not explore the potential influence of teacher characteristics or instructional practices on students’ writing development. Research in this area will provide valuable insights into the role of the educational environment in shaping writing proficiency. Moving forward, future longitudinal research should address these limitations by employing larger and more diverse samples and exploring the influence of contextual factors such as teacher characteristics, instructional practices, and technology integration on students’ writing development.