Writing Fluency: Its Relations With Language, Cognitive, and Transcription Skills, and Writing Quality Using Longitudinal Data From Kindergarten to Grade 2

Abstract

Theoretical models hold that written products (e.g., writing quality) are the outcome of the writing process (e.g., translation, transcription, revision), and skills and knowledge on which the writing process draws (e.g., language, transcription, cognitive skills). In the present study, we examined the relations among writing quality; the writing production process measured by writing fluency; and language, cognitive, and transcription skills, using longitudinal data from English-speaking beginning writers who were followed from kindergarten to Grade 2 (N = 261). Children’s working memory, attentional control, spelling, handwriting fluency, vocabulary, grammatical knowledge, and writing quality were measured in kindergarten, Grade 1, and Grade 2. Writing fluency was measured in Grade 2 in two ways, using a process-based measure, writing burst length (chunks of text produced between pauses during writing), and a product-based measure, the number of words produced per total writing time. Results from structural equation modeling showed that Grade 2 writing fluency was moderately related to Grade 2 writing quality (.40) and completely mediated the relations of Grade 1 spelling and handwriting fluency skills to Grade 2 writing quality. In contrast, Grade 1 vocabulary was directly related to Grade 2 writing quality over and above Grade 2 writing fluency. Kindergarten working memory was indirectly related to Grade 2 writing fluency via Grade 1 spelling skill. These findings indicate that the writing production process measured by writing fluency mediates the relations of transcription skills to writing quality, and executive function is indirectly related to writing fluency via transcription skills for beginning writers.

Producing written texts involves a complex, nonlinear, and iterative process of generating, translating, transcribing, evaluating, and revising ideas. High-quality written products (e.g., quality of written compositions) are outcomes of the writing process (Hayes & Flower, 1980) and the skills and knowledge that undergird and support the writing process (Berninger & Winn, 2006; Kim & Graham, 2022). Although central foci in theoretical models of writing differ, they all assume the relations among contributing and component skills and knowledge, the writing process, and the written product. In the present study, we investigated the nature of the assumed relations among skills and knowledge (e.g., vocabulary, spelling, handwriting fluency), the writing process as measured by writing fluency, and the written product as measured by writing quality for beginning writers. Specifically, we examined the relations of executive function (working memory and attentional control), transcription skills (spelling and handwriting fluency), and oral language skills (vocabulary and grammatical knowledge) to writing fluency, and the relations of writing fluency and executive function, language, and transcription skills to writing quality, using longitudinal data from English-speaking children in primary grades.

Theoretical Models of Writing

Several theoretical models of writing focus primarily on the processes involved in writing. For example, the cognitive model of writing by Hayes and Flower (1980, 1986) elaborated on planning (goal setting, idea generation, organizing), text and sentence generation, and revision/evaluation processes, which were later expanded to the processes of proposer, translator, transcriber, and evaluator/revisor (Chenoweth & Hayes, 2003; also see the functional dynamic approach, van den Bergh et al., 2016). The knowledge-telling and knowledge-transforming model detailed the text generation process (Bereiter & Scardamalia, 1987; Scardamalia & Bereiter, 1987). Other models focused on the skills and knowledge that are employed during the writing process (e.g., oral language, transcription skills, background knowledge; Berninger & Winn, 2006; Kim, 2020; Kim & Graham, 2022). For example, the not-so-simple view of writing states that writing draws on transcription skills, text generation skills, and executive function defined as supervisory attention, goal setting, and planning as well as working memory (Berninger & Winn, 2006). According to the direct and indirect effects model of writing (DIEW; Kim & Graham, 2022; Kim & Park, 2019), the following component skills and knowledge make contributions to written products: executive function (e.g., working memory, inhibitory and attentional control, shifting), oral language skills (e.g., vocabulary, grammatical knowledge, oral discourse), transcription skills, code-related emergent literacy skills (e.g., phonological awareness, orthographic knowledge and awareness, morphological awareness), higher order cognitions and regulation (e.g., inference, perspective taking, monitoring), background knowledge (world, topic, and discourse knowledge), reading skills (e.g., word reading, text reading fluency, reading comprehension), and social-emotional aspects (e.g., motivation, self-concept). Furthermore, DIEW explicitly mapped skills and knowledge with various writing processes (e.g., the transcription process draws on transcription skills and associated subskills such as emergent literacy skills; the evaluation/revision process draws on a number of skills including higher order cognitive skills, reading skills, and background knowledge; see Kim & Graham, 2022). Despite variation in their central foci and in their degree of specificity about the writing process and the skills that contribute to it, all theoretical models implicitly (Berninger & Winn, 2006; Kim & Park, 2019) and explicitly (Graham, 2018; Hayes, 1996, 2012b; Kim & Graham, 2022) assume that written products are a consequence of the writing process and the writing process draws on resources and component skills such as language, cognitive, and transcription skills and knowledge.

Writing Fluency

Fluency refers to automaticity and effortlessness in information processing (LaBerge & Samuels, 1974). The construct of fluency has been applied to oral language (e.g., speech; Lennon, 1990), reading (text reading fluency, word reading fluency; Kim et al., 2021; Wolf & Katzir-Cohen, 2001), and writing skills. When applied to writing, (text) writing fluency refers to “being able to write quickly, with few pauses or hesitations” (Chenoweth & Hayes, 2001, p. 80) or “the ease with which an individual produces written text” (Ritchey et al., 2015, p. 27; italics in original). Writing fluency reflects the writing production process (Chenoweth & Hayes, 2001, 2003; Van Waes & Leijten, 2015) and has been examined and operationalized in multiple ways, which can be broadly categorized as (a) product-based measures and (b) process-based measures. Product-based measures include, for example, text length or number of words in written composition during a specified time (e.g., 1 min) or divided by total writing time, and have been widely used (e.g., Chenoweth & Hayes, 2001; Kellogg, 1990, 1996, 2004; Leijten et al., 2010; Mackie & Dockrell, 2004; Olive et al., 2009; Van Waes & Leijten, 2015).

The process-based measures are collected during the writing production process, considering pauses and transition times between words in keystroke logging studies (e.g., Baaijen et al., 2012; Leijten & Van Waes, 2013; Van Waes & Leijten, 2015; Wengelin, 2006) and writing burst studies (Alves et al., 2008; Chenoweth & Hayes, 2001; Hayes & Chenoweth, 2006; Kim, 2022; Limpo & Alves, 2017). A writing burst, a measure employed in the present study, is “a group of words produced by a writer for inclusion in a text that is bounded by breaks in the production process” and “a break is defined either as a pause of at least 2 seconds or a grammatical discontinuity indicating that the prior language was being revised” (Chenoweth & Hayes, 2003, p. 103; see Wengelin, 2006, for a discussion on the 2-second threshold). As individuals engage in writing tasks, their writing involves an ebb and flow of moments of writing or producing texts (e.g., words, phrases, sentences) followed by pauses. Bursts of text production followed by pauses were examined initially in oral language contexts (e.g., Chenoweth & Hayes, 2001; Kaufer et al., 1986). For example, Chenoweth and Hayes (2001) examined oral language burst segments that ended in pauses (P-bursts), which were posited to tap the translation process, and segments that were primarily revised or repeated protocol (R-bursts), which were posited to tap the revision process. Work on oral language bursts was extended to written texts, writing bursts or bursts of written language—chunks of text produced between pauses during the writing process. Studies found that skilled writers have longer writing bursts or produce a greater number of linguistic units such as words, phrases, and sentences per writing burst than less skilled writers (e.g., Friedlander, 1989; Limpo & Alves, 2017), and writing burst length is related to written products such as writing quality (Almond et al., 2012; Kim, 2022; Limpo & Alves, 2017). Studies also found that young developing writers rarely employ revision, and therefore, it is generally not meaningful to distinguish P-bursts from R-bursts for young developing writers (Kim, 2022; Limpo & Alves, 2017).

Writing fluency is posited to tap into the writing process noted above such as idea generation, translation, transcription, and reviewing and revising (Baaijen et al., 2012; Chenoweth & Hayes, 2001, 2003; Galbraith & Baaijen, 2019; Hayes, 2012a; Kim et al., 2018). For example, writing fluency as measured by writing bursts were observed in a task that involved idea generation (Hayes & Chenoweth, 2007). Writing bursts have also been shown to tap into the translation process (Chenoweth & Hayes, 2001; Connelly et al., 2012). The translation process draws on oral language skills (Berninger et al., 2002; Kim & Graham, 2022) as writers use their vocabulary knowledge to choose words and use their morphosyntactic and syntactic knowledge to construct sentences that accurately represent their ideas. Therefore, if writing bursts tap the translation process, oral language skills would relate to writing bursts. Empirical evidence supports this hypothesis. For adults learning a second language, the amount of language exposure in second language learning was related to writing bursts (Chenoweth & Hayes, 2001)., Writing burst length was shorter for children with developmental language disorder than that for children without developmental language disorder, and oral language skills were related to burst length (Connelly et al., 2012).

Furthermore, writing bursts were shown to tap into the transcription process—transcribing translated ideas into print—particularly for developing writers or individuals with transcription difficulties (Alves & Limpo, 2015; Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017). When not automated, the transcription skills spelling and handwriting fluency severely constrain the transcription process and consequent written products such as writing quality (Berninger & Winn, 2006; Graham, 2018; Kim & Graham, 2022). Spelling is the ability to encode sounds to written words that adhere to the orthographic system of a language. Handwriting fluency is the skill in forming and writing letters accurately with speed. Difficulty with spelling and laborious and slow handwriting slow down the transcription process, which hampers producing an intended amount of text, which, in turn, influences writing quality because expressing ideas with quality requires a certain amount of text. In addition, difficulty with spelling and slow handwriting constrain limited resources such as working memory and attention as they tax the writer’s working memory and attention, which interferes with competing processes that the writer has to attend to while writing such as ideation and translation (e.g., the writer may forget or lose ideas held in working memory; Berninger & Winn, 2016; Graham, 2018). Evidence supports the hypothesis that writing bursts tap the transcription process. For example, individuals with dyslexia have persistent difficulties in spelling in addition to word reading (e.g., Berninger et al., 2008; Graham et al., 2021), and adults with dyslexia had more frequent pauses than age-matched controls (Wengelin & Strömqvist, 2000). Studies also showed that transcription skills, handwriting fluency and spelling, consistently predict burst length for developing writers (Alves & Limpo, 2015; Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017). Spelling and handwriting fluency were independently related to writing burst length after accounting for working memory and nonverbal skills for children with developmental language disorder and their peers matched for age and language skill (Connelly et al., 2012). Spelling and handwriting fluency also predicted writing burst length for Portuguese-speaking children in Grades 2–7 (Alves & Limpo, 2015). Moreover, targeted instruction on transcription skills, especially handwriting skill, resulted in longer writing bursts for Portuguese-speaking second graders (Alves et al., 2016; also see Limpo & Alves, 2018).

The writing process also relies on domain-general cognitive resources or executive function such as working memory (Berninger & Winn, 2006; Chenoweth & Hayes, 2001; Hayes, 1996; Kim & Graham, 2022) and attentional control (Berninger & Winn, 2006; Kim & Graham, 2022; Perverly et al., 2014). Working memory refers to the ability to hold and process information temporarily and simultaneously (Baddeley, 2003; Daneman & Carpenter, 1980), and attentional control is the ability to alert, orient, and selectively sustain attention on task-relevant information (Scerif, 2010). Translating ideas into linguistic forms and transcribing those into print require searching for and retrieving relevant information (e.g., lexical items, orthographic information); formulating phrases, clauses, and sentences; and formulating and executing motor plans for handwriting or keystrokes while at the same time, continuing to plan, maintain, monitor, and update ideas/texts in the mind. Studies have shown the role of working memory and attentional control in the writing process (e.g., Kellogg, 2001, 2004) and their relations to writing bursts (Kim, 2022).

Relations Among Component Skills, Writing Fluency, and Writing Quality

If component skills and knowledge such as oral language, spelling, and handwriting skills are employed during the writing process (e.g., translation, transcription) which is tapped by writing fluency, writing fluency should mediate the relations of component skills to the written product—writing quality—at least partially. In other words, the relations of component skills to writing quality should be at least partially explained by writing fluency. This is suggested by two lines of work. First, a large body of literature indicates that oral language skills, transcription skills, and domain-general cognitions or executive function are related to writing fluency and writing quality. As reviewed above, previous studies have shown the relations of oral language, transcription, working memory, and attentional control to writing fluency (Chenoweth & Hayes, 2001; Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017). In addition, a large body of literature has shown the relations of working memory and attentional control (Bourdin & Fayol, 1994; Hayes & Chenoweth, 2007; Kim & Schatschneider, 2017; Olive, 2004), oral language skills (Babayiğit, 2014; Coker, 2006; Kim et al., 2011, 2015), and transcription skills (Graham & Santangelo, 2014; Santangelo & Graham, 2016) to writing quality. Second, as reviewed above, studies have also demonstrated the relation of writing fluency to writing quality (Almond et al., 2012; Friedlander, 1989; Kim, 2022; Limpo & Alves, 2017).

Taken together, theoretical models and the literature reviewed above suggest a mediating role of writing fluency, where domain-general cognitions, and transcription and oral language skills relate to writing fluency (Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017), which, in turn, relates to writing quality (Almond et al., 2012; Kim, 2022; Limpo & Alves, 2017). However, studies that simultaneously examined the relations among component skills, writing fluency, and writing quality are scarce. In fact, to our knowledge, there are only two studies that explicitly examined mediated relations among component skills, writing fluency, and written products. Limpo and Alves (2017) used data from Portuguese-speaking Grade 2 students, and examined the relations of spelling and handwriting fluency to burst length and pauses, and writing quality. Their results showed that writing burst length completely mediated the relation of spelling to writing quality whereas it partially mediated the relation of handwriting fluency to writing quality—that is, handwriting fluency was directly related to writing quality over and above writing bursts, but spelling was not. Kim (2022) used data from English-speaking Grade 2 students, and found that writing burst length partially mediated the relation of spelling to writing quality while it completely mediated the relation of handwriting fluency. Furthermore, burst length completely mediated the relation of attentional control to writing quality whereas it partially mediated the relation of working memory to writing quality. These studies overall suggest that writing bursts mediate the relations of transcription skills and executive function to writing quality for developing writers although the nature of mediation, partial or complete mediation, was different for spelling versus handwriting fluency for Portuguese-speaking children and English-speaking children.

Present Study

The goal in the present study was to extend our understanding of the relations among writing component skills (working memory, attentional control, transcription, and oral language skills), the writing process measured by writing fluency, and writing product measured by writing quality for young developing writers, using longitudinal data from primary grade students. The present study extends previous work in three important ways. First, we examined a theoretically guided mediation of writing fluency. Despite the assumed relations among writing component skills, writing fluency, and writing quality, studies that explicitly examined the mediating role of writing fluency are highly limited, and exceptions include the two studies reviewed above (Kim, 2022; Limpo & Alves, 2017). Second, we used longitudinal data from kindergarten to Grade 2 to examine mediation, whereas the two previous studies used cross-sectional data. Although theory-guided mediational models using cross-sectional data are informative and important, longitudinal data are preferred because a time sequence among predictors, mediators, and outcomes is important to establishing causal mediation (see Selig & Preacher, 2009, for details). Third, writing fluency was measured in two ways—using a process-based measure, writing bursts, and a product-based measure, the number of words written in a composition divided by total writing time—and modeled as a latent variable in data analysis. Both the multi-faceted approach to measuring the writing fluency construct and the employment of a latent variable approach allow a more comprehensive and precise examination of the writing fluency construct.

Specific research questions that guided the present study were as follows: Does Grade 2 writing fluency mediate the relations of kindergarten executive function and Grade 1 transcription and oral language skills to Grade 2 writing quality? Do Grade 1 transcription and language skills mediate the relations of kindergarten working memory and attentional control to Grade 2 writing fluency and writing quality?

Based on previous studies and theories, we hypothesized that writing fluency would predict writing quality (e.g., Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017). We also hypothesized that writing fluency would, in general, mediate the relations of working memory, attentional control, handwriting fluency, spelling, vocabulary, and grammatical knowledge to writing quality. However, we did not have specific hypotheses regarding the extent of mediation (complete or partial) due to mixed previous findings (e.g., Connelly et al., 2012; Kim, 2022; Limpo & Alves, 2017). With regard to the mediating role of writing fluency in the relations of oral language skills to writing quality, we expected partial mediation. Although oral language skills such as sentence formulation skill were directly related to writing fluency (writing bursts; Connelly et al., 2012), we expected that writing fluency would be more strongly influenced by transcription skills than oral language skills for beginning writers in primary grades and therefore, writing fluency in the beginning phase of writing development may not fully capture the influence of oral language skills on writing quality.

Method

Participants

Data for the present study were from a longitudinal study on elementary grade children’s literacy development. Data in Grade 2 were reported in a previous study (Kim & Graham, 2022). Children in kindergarten (N = 261, 55% male) from seven schools in the southeastern part of the US were followed until Grade 2 (M_age = 5.32 and SD = .44 in kindergarten, M_age = 6.17 and SD = .32 in Grade 1, and M_age = 7.19 and SD = .34 in Grade 2). Approximately 53% were Caucasian children, 33% were African American children, 3% were Hispanic children, 5% were mixed race, and .4% were Asian American children. According to school district record, approximately 9% were receiving speech services, and .8% were identified as having language impairment. Approximately 1% of children were classified as English learners. Two thirds (68%) of the children were eligible for the free and reduced lunch program, an indicator of poverty in the US. The school district did not report having any officially adopted commercial writing curriculum. Instead, teachers in the participating schools reported using the writers’ workshop approach. Study protocols were approved by the Internal Review Board at the Florida State University (HSC No. 2016.17872).

Measures

Writing quality, working memory, attentional control, vocabulary, grammatical knowledge, spelling, and handwriting were measured in kindergarten, Grade 1, and Grade 2. Writing fluency was measured in only a single grade (Grade 2) because a pilot study revealed that children in kindergarten and Grade 1 did not find comfortable the Livescribe Pulse pen, which was used to capture writing bursts. Unless otherwise noted, items were scored dichotomously (1 = correct; 0 = incorrect). Table 1 shows reliability estimates for all the measures.

Table 1.

Descriptive Statistics

Variable	Reliability+	N	Mean	SD	Min-Max	Skewness	Kurtosis
Writing Quality
K TEWL	.81–.96*	246	1.21	0.90	0–3	−0.36	−1.55
K One Day	.81–.96*	248	0.89	1.18	0–4	0.79	−1.02
K Pets	.81–.96*	248	1.69	1.00	0–4	−0.72	−0.53
G1 TEWL	.81–.96*	228	2.08	1.31	0–6	0.19	−0.10
G1 One Day	.81–.96*	228	2. 58	1.45	0–6	−0.14	−0.32
G1 Pets	.81–.96*	227	1.89	1.13	0–5	0.47	−0.03
G2 TEWL	.81–.96*	173	3.25	1.50	0–9	0.12	1.26
G2 WIAT	.81–.96*	172	2.44	1.26	0–6	0.15	−0.06
G2 Beaver	.81–.96*	173	2.12	1.23	0–5	−0.57	−0.74
Writing Burst Length
G2 TEWL Burst Length	.95	168	1.77	1.07	0.29–5.97	1.23	1.63
G2 WIAT Burst Length	.95	150	2.02	1.48	0.14–9.78	2.34	7.20
G2 Beaver Burst Length	.95	169	1.13	0.71	0.00–4.17	1.59	3.17
Words per Time
G2 TEWL Words per Time	NA	164	15.43	14.55	.59–114.55	3.55	16.77
G2 WIAT Words per Time	NA	143	15.44	12.35	3.46–70.66	2.38	6.11
G2 Beaver Words per Time	NA	167	9.40	6.05	2.06–49.47	2.80	13.23
Transcription Skills
K Spelling	.78	247	4.64	2.57	0–12	0.06	−.20
G1 Spelling	.89	228	5.31	4.44	0–18	0.89	0.01
G2 Spelling	.92	173	10.31	6.03	0–22	−0.06	−1.03
K Copy Task 1	.90–.98**	247	16.04	5.66	0–35	0.26	.98
K Copy Task 2	.90–.98**	248	14.23	6.88	0–41	0.15	.51
K Copy Task 3	.90–.98**	248	18.46	7.92	0–43	−0.02	.15
G1 Copy Task 1	.90–.98**	227	25.39	9.89	0–56	0.17	0.19
G1 Copy Task 2	.90–.98**	228	26.54	11.13	0–60	−0.06	−0.18
G1 Copy Task 3	.90–.98**	228	31.3	12.72	0–67	−0.05	−0.15
G2 Copy Task 1	.90–.98**	173	39.8	17.45	3–91	0.26	−0.31
G2 Copy Task 2	.90–.98**	172	40.41	17.01	0–94	0.14	−0.17
G3 Copy Task 3	.90–.98**	172	48.22	18.17	0–99	0.05	−0.03
Oral Language Skills
K WJ Picture Vocabulary	.69	261	16.61	2.79	6–26	−0.20	1.30
K WJ Picture Vocabulary SS	NA	261	99.29	9.47	58–129	−.40	2.12
K Grammaticality	.94	259	11.09	9.51	0–43	1.06	.52
G1 WJ Picture Vocabulary	.73	233	19.25	2.92	8–26	−0.64	1.25
G1 WJ Picture Vocabulary SS	NA	232	101.71	10.31	58–126	−0.78	1.77
G1 Grammaticality	.95	232	19.98	12.06	0–53	0.33	−0.44
G2 WJ Picture Vocabulary	.70	179	19.75	2.95	11–28	−.00	.02
G2 WJ Picture Vocabulary SS	NA	179	96.48	9.91	68–121	−.17	−.01
G2 Grammaticality	.95	179	29.24	12.99	1–59	0.08	−.34
G2 Grammaticality SS	NA	179	94.77	12.40	62–122	−.32	.05
Executive Function
K Working Memory	.89	259	6.87	5.94	0–23	0.50	−0.84
K Attentional Control	.99	253	121.41	31.17	46–206	0.34	0.19
G1 Working Memory	.61	226	12.00	6.55	0–25	−0.33	−0.84
G1 Attentional Control	.99	228	127.5	37.16	38–210	0.66	−0.06
G2 Working Memory	.77	179	6.94	4.21	0–18	0.21	−0.34
G2 Attentional Control	.98	176	121.22	30.87	40–210	0.39	0.41

Note. K = Kindergarten, G1 = Grade 1, G2 = Grade 2, Copy = Copying, WJ = Woodcock Johnson, SS = Standard Score, TEWL = Test of Early Written Language, WIAT = Weschler Individual Achievement Test, Beaver = Beaver prompt.

⁺Unless otherwise noted, estimates are Cronbach’s alpha.

^*Cohen’s Kappa, using 40 written compositions for each task.

^**Interrater agreement, exact agreement, using 40 samples for each sentence-copying task.

Writing Quality in Kindergarten, Grade 1, and Grade 2.

The following three identical tasks were used in kindergarten and Grade 1: TEWL-Third Edition (Hresko et al., 2012), a researcher-developed narrative task called One Day (Kim et al., 2015), and a researcher-developed opinion task called Class Pets (Wagner et al., 2011). In the TEWL task (skateboard prompt), the child was presented with a series of four illustrations and was asked to write a story that goes with the illustrations. In the One Day task, the child was asked to write about any interesting events that occurred responding to the prompt “One day when I got home from school…” In the Class Pets task, the child was asked to write about a class pet they would like and three reasons for it. In Grade 2, the following three tasks were used: TEWL-Third Edition (Hresko et al., 2012), the Essay subtest of the Wechsler Individual Achievement Test-Third Edition (WIAT; Wechsler, 2009), and a researcher-developed task called Beaver. The TEWL task had an identical procedure, but used a different illustration (kitchen prompt) from those used in kindergarten and Grade 1. In the WIAT Essay task, the child was asked to write about their favorite game and three reasons for it. In the Beaver task, the child was provided with a passage about beavers (297 words; adapted from the Qualitative Reading Inventory [QRI]; Leslie & Caldwell, 2011) and was asked to write about what beavers do and how they do it. The source text described beavers’ habitat and the process of building a dam. Although the original source QRI text did not include illustrations, three publicly available images were included in the study to aid children’s comprehension of the text. The child was given access to the text while composing. The three writing tasks were administered in different sessions, and children were given 15 minutes for each task in line with previous studies (e.g., Graham et al., 2002; Kim et al., 2015; Olinghouse, 2008).

Children’s essays were coded for writing quality, which was operationalized as the extent to which ideas were developed and presented in an organized manner on a rating scale of 1 to 10 (e.g., Graham et al., 2007; Hooper et al., 2002). Compositions with clearer ideas, greater relevant rich details, and logical arrangement of ideas (organization, including macro- and micro-organization) were given higher scores. Compositions with an incomprehensible random series of letters or consisting of a verbatim copy of the source materials for the Beaver task were scored a zero. These general guidelines were applied to all the tasks but the rubric was slightly modified for different genres (see Appendix A for the rubric for the different tasks). This way, the operationalization of quality as defined by clarity of ideas, relevant rich details, and logical arrangement of ideas (organization, including macro- and micro-organization) is maintained across genres while recognizing genre-specific aspects. Raters included a PhD student in school psychology and an individual with a master’s degree in literacy education, and both had extensive experience in coding children’s language and writing samples in a previous large-scale project. They were trained in a series of meetings and practice sessions. Rubric and anchor essays were introduced first. Then, raters practiced coding written compositions and anchor essays were further refined, followed by discussions of their scores resolving any discrepancies.

Writing Fluency in Grade 2.

Writing fluency was measured in two ways, using a product-based measure and a process-based measure. The product-based measure was the number of words written divided by total writing time (words per time hereafter) from the three writing tasks in Grade 2 described above. Words were characterized as real words recognizable in the context of composition despite some spelling errors (e.g., Espin et al., 2000). If children attempted to write a word and then rewrote it (e.g., “th” followed by “the”), only the subsequent attempt was counted as a word, whereas the initial attempt was not counted.

The second, process-based measure of writing fluency was writing bursts. Writing burst data were collected while children were composing in response to the above-noted composition tasks using a digital pen, a Livescribe Pulse, and paper sheets, which had a microdotted pattern printed on it. The Livescribe Pulse has an infrared camera at its nib, which logs handwriting data. Children’s handwritten compositions were uploaded to the HandSpy web application, which displays students’ real-time writing process (see Limpo & Alves, 2017). Following previous work, a writing burst was defined as graphomotor activity between two consecutive pauses longer than 2 seconds. We recognize that any interval (e.g., 2, 3, or 5 seconds) would be arbitrary but the 2 seconds were used following previous studies with adults (e.g., Chenoweth & Hayes, 2003) and young children (Alves et al., 2016; Connelly et al., 2012; Limpo & Alves, 2017). A pause was defined as the moment between when the child stopped writing and when the child resumed writing. Punctuation was not counted as a graphomotor activity, and the transition time for line-breaks was not taken into consideration for pause (Limpo & Alves, 2017). P-bursts and R-bursts were not distinguished because there were few R-bursts in the sample. Indeed, there were no R-bursts characterized by revisions of content in the observed data. Instead, we observed instances of rewriting individual letters (e.g., rewriting the letter “a” in the word “make” or “th” followed by “the”) or crossing out and rewriting entire words (e.g., replacing “allso” with “also”). This type of correction was relatively infrequent, accounting for approximately 4.8%, 4.6%, and 5.3% of the total words attempted in the TEWL, WIAT, and Beaver tasks, respectively. For the assessment of writing bursts, only the rewritten words (irrespective of accuracy of spelling) were considered, while the initial attempts were excluded.

When a word was split between two bursts (pause longer than 2 seconds within a word), the full word was included in the burst where the greater part of the word was written. Two scorers (one individual had a master’s degree in literacy education and the other was a master’s student in school psychology) were trained on capturing writing bursts in several sessions in which the scoring manual was introduced and practice sessions and discussion followed; training continued until a minimum of 95% exact agreement was reached using 40 written samples. Scoring was straightforward but there was some room for disagreement (e.g., reading/deciphering the child’s sloppy handwriting) and therefore, reliability was established. Dr. Teresa Limpo was available for any questions related to the HandSpy application and supported troubleshooting. More information on HandSpy can be found online (http://daar.up.pt/index.php/handspy).

It is important to clarify that in our study, we did not consider text length (i.e., the number of words written) as an indicator of writing fluency, despite its use in some previous studies (e.g., Berninger et al., 1994; Limpo & Alves, 2017). The reason for this exclusion is for theoretical precision because text length does not take into account time, which is a key feature of the construct “fluency.” Writing fluency is not solely measured by the amount of text produced; rather, it involves the time component as well. When text length was included in the analysis, not surprisingly it was extremely strongly related with writing burst length and words per time, and the structural results were essentially the same as reported here (Figure 2).

Figure 2.

Standardized Coefficients of the Results

Note. Only the paths involving the variables of primary interest are shown. Solid lines indicate statistically significant paths and dashed lines indicate statistically nonsignificant paths. Control variables are grayed out in order to reduce clutter, but were included in the model. Covariances between executive function, language, and transcription skills in each grade were allowed, but not shown. Results for control variables and covariances are found in Online Supplemental Materials. K = Kindergarten, G1 = Grade 1, G2 = Grade 2, Task = Copying, TEWL = Test of Early Written Language, WIAT = Weschler Individual Achievement Test, Beaver = Beaver prompt, Burst = Burst length, W/T = Number of Words per Time.

Transcription Skills.

Children’s spelling skill was measured by a researcher-developed experimental dictation task. There were 13 items in kindergarten, 18 items in Grade 1, and 22 items in Grade 2. While there were some common items across grades, some words were changed and more words of increasing difficulty were added in each consecutive grade considering developmental phase. Spelling words are found in Appendix B.

Children’s handwriting fluency was measured by three identical sentence-copying tasks across the grades (e.g., Wagner et al., 2011). The sentences included a pangram, “The quick brown fox jumps over the lazy dog,” and two researcher-developed sentences, “My dog jumps and runs when I tell him to jump and run,” and “My mom put the lid on the pan to cook the food.” The student was presented with each sentence and was asked to copy it as many times as possible in 1 minute. The number of correctly copied letters was their score.

Oral Language Skills.

Children’s vocabulary was measured by the Picture Vocabulary subtest of the Woodcock Johnson-3^rd Edition (Woodcock et al., 2001). The child was asked to identify pictured objects, which were ordered by increasing level of difficulty (e.g., car), and test administration discontinued after six consecutive incorrect items.

Children’s grammatical knowledge was measured using the Grammaticality Judgment task of the Comprehensive Assessment of Spoken Language (Carrow-Woolfolk, 1999). The child was asked whether a heard sentence was grammatically correct (e.g., The child were running). This task was normed for individuals age 7 and older. Therefore, in kindergarten and Grade 1, eight easy experimental items modeled after the Grammaticality Judgment task were developed and pilot tested before the longitudinal study. These eight items were added as the first items before administering the items from the normed task. For items in later positions, the child was asked to also correct the sentence if grammatically incorrect. For these items, the total possible score for each item was 2 (1 for correctly identifying grammaticality of the given sentence, and 1 for accurately correcting the sentence). Test administration discontinued after five consecutive incorrect items.

Executive Function.

Working memory was measured by a listening span task (Cain et al., 2004; Daneman & Merikle, 1996). The child was presented with three-word sentences involving common knowledge familiar to children (e.g., Do frogs jump? Are apples blue?) and was asked to identify whether the heard sentences were correct or not (Yes/No). Then the child was asked to identify the last word of each sentence. Considering the developmental phase of the children, the number of items and complexity differed across the grades. There were 14 items in kindergarten (four one-sentence items; seven two-sentence items; three three-sentence items), 15 items in Grade 1 (four one-sentence items; eight two-sentence items; three three-sentence items), and 13 items in Grade 2 (five two-sentence items; four three-sentence items; three four-sentence items; and one five-sentence item). Children’s responses regarding the veracity of the statements (yes/no) were not scored, but their responses on the last words were given a score of 0 to 2: correct last words in correct order were given 2 points, correct last words in incorrect order were given 1 point, and incorrect last words were given 0 point. There were four practice items and 13–15 experimental items with a total possible score of 26–30 points, depending on the grade level. Testing discontinued after three incorrect responses.

Children’s attentional control was measured by the Strengths and Weaknesses of ADHD (Attention-Deficit Hyperactivity Disorder) Symptoms and Normal Behavior Scale (SWAN; Swanson et al., 2006) that was completed by the child’s teacher. SWAN is a behavioral checklist that includes 30 items (e.g., Sustain attention on tasks or play activities) that are rated on a 7-point scale ranging from 1 (far below average) to 7 (far above average). Previous studies have shown validity of SWAN for children in elementary grades (Arnett et al., 2013; Sáez et al., 2012).

Procedures

Children were assessed in a quiet space in schools. Working memory, vocabulary, and grammatical knowledge tasks were individually administered whereas spelling, handwriting, and writing tasks were group administered (small groups of 4 to 8 children).

Transparency and Openness

We report our sample size, missing data, and measures in the study, following JARS (Kazak, 2018). Data used in the present article will be available upon reasonable request. Data were analyzed using Mplus 8.6 (Muthén & Muthén, 2017). This study’s design and analysis were not pre-registered.

Data Analytic Strategies

Confirmatory factor analysis and structural equation modeling using full information maximum likelihood estimation were the primary data analytic strategies. Latent variables were created for writing quality in kindergarten, Grade 1, and Grade 2; writing fluency measured by the words per time variables and writing burst variables from each of the three writing tasks in Grade 2; and handwriting fluency in kindergarten, Grade 1, and Grade 2 as these variables were measured with multiple tasks. To address the research questions, the structural equation model in Figure 1 was fitted, where kindergarten working memory and attentional control predicted Grade 1 language and transcription skills and Grade 2 writing bursts and writing quality, and Grade 1 language and transcription skills predicted Grade 2 writing bursts and writing quality. In addition, autoregressors were included as control variables: that is, transcription, oral language skills, and writing quality in kindergarten; executive function and writing quality in Grade 1; and executive function, transcription, and oral language skills in Grade 2 (see gray part of Figure 1). Given that the writing fluency and writing quality variables were obtained from the same tasks (prompts), residual variances for the same prompts were allowed (see Figure 2). Detailed paths for the control variables are not shown in Figure 1 to minimize clutter, but the Online Supplemental Materials show full details. Model fit was evaluated using multiple indices: chi-square statistic, comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residuals (SRMR).

Figure 1.

The Relations of Kindergarten Working Memory and Attentional Control, Grade 1 Transcription Skills (Handwriting Fluency and Spelling) and Oral Language Skills (Vocabulary and Grammatical Knowledge), and Grade 2 Writing Fluency and Writing Quality, After Controlling for Autoregressors

Note. Control variables are grayed out in order to reduce clutter, but were included in the model. Covariances between executive function, language, and transcription skills in each grade were allowed, but not shown—see Online Supplemental Materials for details. Residual covariances among the same composition tasks in Grade 2 were allowed but are not shown. K = Kindergarten, G1 = Grade 1, G2 = Grade 2, Task = Copying, TEWL = Test of Early Written Language, WIAT = Weschler Individual Achievement Test, Beaver = Beaver prompt, Burst = Burst length, W/T = Number of Words per Time.

Results

Descriptive Statistics and Preliminary Analysis

Sample sizes for each variable and descriptive statistics are shown in Table 1. In kindergarten, missingness ranged from 6.5% in TEWL Writing Quality to .8% in Picture Vocabulary. Little’s test revealed that missing completely at random (MCAR) could not be rejected in kindergarten (χ² = 41.16, df = 37, p = .29). In Grade 1, there was approximately 10% attrition, and total missingness, including attrition, varied from 13.7% in the working memory task to 11.4% in Picture Vocabulary. Little’s test revealed that missing completely at random (MCAR) could not be rejected in Grade 1 (χ² = 29.27, df = 33, p = .65). In Grade 2, there was 14% further attrition such that the sample size for writing quality variables ranged from 172 to 173. Writing burst data had further reduction due to technical difficulties and malfunction, and this was particularly the case for the WIAT task (22 cases in the WIAT task, four cases in the TEWL task, and three cases in the Beaver task). Perhaps not surprisingly, Little’s test revealed that missing completely at random (MCAR) was rejected in Grade 2 (χ² = 216.13, df = 47, p < .001). Missing data were handled by using the full information maximum likelihood estimator.

As shown in Table 1, mean burst length in Grade 2 was small, ranging from 1.13 to 2.02 words. Distributional properties of the variables were adequate (e.g., see skewness and kurtosis) for the vast majority of variables. Exceptions included the burst length of the WIAT writing task and words per time on the TEWL, WIAT, and Beaver tasks, for which the kurtosis values were greater than ideal. For these variables, outliers were identified according to the definition of three times the interquartile range and visual inspection of the distribution. After winsorizing outliers, skewness and kurtosis values were acceptable as follows: .77 for skewness and −.39 for kurtosis for the WIAT writing burst length, 1.75 for skewness and 2.84 for kurtosis for the TEWL words per time, 2.11 for skewness and 4.23 for kurtosis for the WIAT words per time, and 1.53 for skewness and 3.15 for kurtosis for the Beaver words per time. Winsorized data were used in subsequent analysis. Children’s mean performances on the normed tasks, the Picture Vocabulary task in all the grades and the Grammaticality Judgment task in Grade 2 (when normative information was available), were in the solid average range. Subsequent analysis was conducted using raw scores. Bivariate correlations are found in Online Supplemental Materials due to the large table size (see Table S1). Grade 2 burst lengths were moderately to strongly related with words per time (.40 ≤ rs ≤ .83). Working memory and attentional control were weakly to moderately correlated with transcription and language skills (.10 ≤ rs ≤ .50) and with writing burst length in Grade 2 (.19 ≤ rs ≤ .36). Vocabulary and grammatical knowledge had zero to moderate correlations (|.02| ≤ rs ≤ .40) with writing burst length, and transcription skills had mostly moderate correlations with writing burst length (.33 ≤ rs ≤ .63) except for the kindergarten copy task 2 (.26 ≤ rs ≤ .28). Executive function, language, and transcription skills had weak to moderate correlations with writing quality (.25 ≤ rs ≤ .59) in each grade.

Confirmatory factor analysis was conducted for writing fluency in Grade 2 and for writing quality and handwriting fluency in kindergarten, Grade 1, and Grade 2. For the writing fluency construct, we explored whether words per time variables and writing burst length variables can be considered as two related but dissociable latent variables. However, results showed that they were perfectly related (r = 1.00), and therefore, a single latent variable of writing fluency was fitted using the six variables (three words-per-time variables and three burst-length variables). Loadings for the writing fluency latent variable were fairly strong to very strong, ranging from .55 (WIAT words per time, p < .001) to .83 (Beaver words per time, p < .001). Loadings for Grade 2 writing quality were also fairly strong, ranging from .60 to .65 (ps < .001), and loadings for Grade 1 handwriting fluency were strong, ranging from .82 to .88 (ps < .001).

The Relations of Kindergarten Executive Function and Grade 1 Transcription and Oral Language Skills to Grade 2 Writing Fluency and Writing Quality

The model shown in Figure 1 was fit to the data, and model fit was good: χ² (579) = 829.90, p < .001, RMSEA = .041 [90% CI = .034, .047], CFI = .95, and SRMR = .087. Figure 2 shows standardized coefficients. Grade 2 writing fluency was moderately and independently related to writing quality (.40, p < .001) after accounting for all the other variables in the model. In addition, Grade 1 vocabulary (.17, p = .048) was directly related to Grade 2 writing quality. Grade 1 handwriting fluency (.39, p < .001) and spelling (.26, p = .005) independently predicted Grade 2 writing fluency, but they were not directly related to Grade 2 writing quality after accounting for Grade 2 writing fluency (ps ≥ .14). There was also a statistically significant suppression effect of Grade 1 vocabulary on Grade 2 writing fluency (−.25, p < .001). In other words, in the bivariate correlations (see Online Supplemental Materials), Grade 1 vocabulary knowledge was not related to Grade 2 writing burst length and words per time (|.02| ≤ rs ≤ |.12|), but it had a statistically significant negative relation to Grade 2 writing fluency when accounting for the other variables in the model. The suppression effect appears to be mostly due to the moderate relations among spelling, grammatical knowledge, writing quality, and vocabulary (see Online Supplemental Materials) such that once spelling, grammatical knowledge, and writing quality were no longer in the model together, the suppression effect of vocabulary knowledge was no longer present. Kindergarten working memory (ps ≥ .32) and attentional control (ps ≥ .08) were not directly related to Grade 2 writing fluency or writing quality after accounting for Grade 1 language, cognitive, and transcription skills and other control variables. Kindergarten working memory was independently related to Grade 1 spelling (.15, p = .008) and grammatical knowledge (.30, p < .001) while kindergarten attentional control was independently related to Grade 1 grammatical knowledge (.12, p = .03). Online Supplemental Materials show results for control variables. The included variables explained 81% of total variance in Grade 2 writing quality and 50% of total variance in Grade 2 writing fluency.

Discussion

According to theoretical models of writing, language, cognitive, and transcription skills and knowledge contribute to and are employed during the writing process, which, in turn, result in written products. Therefore, written products such as writing quality are an outcome of the writing process and component skills. A large body of literature has shown that component skills such as oral language skills, transcription skills, cognitive skills, and knowledge are important to writing quality (e.g., see Graham, 2018 and Kim & Park, 2019 for reviews). In addition, a growing body of literature has examined the writing process measured by writing fluency (e.g., Almond et al., 2012; Alves et al., 2016; Chenoweth & Hayes, 2001; Kim, 2022; Limpo & Alves, 2017; Xu, 2018) and its relation to writing outcomes such as writing quality (Almond et al., 2012; Kim, 2022; Limpo & Alves, 2017). In the present study, we extend these previous studies in three ways. First, we simultaneously examined component skills, the writing process (measured by writing fluency), and a writing outcome (writing quality) for beginning writers. Specifically, we examined the nature of mediated relations among cognitive skills (working memory and attentional control), transcription skills (handwriting fluency and spelling), oral language skills (vocabulary and grammatical knowledge), writing fluency, and writing quality. Second, the present study used longitudinal data such that we examined the relations of working memory and attentional control in kindergarten to transcription and oral language skills in Grade 1 and writing texts—writing fluency and writing quality—in Grade 2, after controlling for autoregressors. While theory-based mediation can be examined using cross-sectional data, longitudinal data allow stronger inferences on causal mediation because mediation assumes temporal sequence of the variables in hypothesized mediated relations (Selig & Preacher, 2009). Lastly, writing fluency was measured using both a process-based measure (writing bursts) and a product-based measure (words per time), and modeled as a latent variable in statistical analysis.

The mean length of writing bursts—the number of words produced between pauses during writing —was small, ranging from 1.13 to 2.02. These findings are in line with previous research involving Grade 2 Portuguese-speaking children (Alves & Limpo, 2015; Limpo & Alves, 2017, 2018), suggesting that the participants were at an early stage of their writing development. It is important to note that the interpretation of these results should take into account the method used to measure writing bursts in this study. While previous literature has distinguished between P-bursts and R-bursts in studies involving adults (e.g., Chenoweth & Hayes, 2001, 2003; Van Waes & Leijten, 2015), our sample of beginning writers showed a rare occurrence of R-bursts (also see Limpo & Alves, 2017). As a result, our focus in this study was primarily on P-bursts. Notably, we observed that children engaged in correcting letters or rewriting words within their writing bursts, albeit relatively infrequently. The counting of words per burst in this study excluded the child’s initial attempt and considered only subsequent attempts (regardless of spelling accuracy) as words. The potential impact of this approach on our findings remains unclear. Future research could address this question by adopting an alternative approach, such as counting initial attempts as words per burst.

While not the primary focus of the study, it is notable that both measures of writing fluency—the length of writing bursts and the number of words written per total writing time—were found to represent a single latent variable rather than two related but dissociable variables. This result, although not unexpected, offers empirical evidence that both approaches capture the writing production process.

One of the key findings of the present study is that writing fluency completely mediated the relations of transcription skills to writing quality. As a measure of writing process, writing fluency was posited to draw on oral language and transcription skills, and executive function. Overall the present findings are in line with theoretical models of writing, which implicitly or explicitly presume that written products such as writing quality are the outcome of the writing process (e.g., Graham, 2018; Hayes & Flower, 1980; Kim & Graham, 2022; van den Bergh et al., 2016), which draws on language, cognitive, and transcription skills (see above). However, the complete mediation by writing fluency in the relation of transcription skills to writing quality is divergent with previous studies that found partial mediation (Kim, 2022; Limpo & Alves, 2017). There might be at least two reasons for discrepant findings. First, the previous studies used cross-sectional data whereas the present study used longitudinal data. Second, there were differences in what was accounted for in statistical models: Limpo and Alves (2017) did not include oral language skills or executive function, and Kim (2022) included oral discourse language as measured by children’s retell and production whereas in the present study vocabulary and grammatical knowledge were used. Although patterns of mediation from previous studies and the present study differ to some extent, the results overall indicate that writing fluency largely taps the transcription process for young beginning writers and is directly influenced and constrained by transcription skills for beginning writers in primary grades (Alves et al., 2016; Limpo & Alves, 2018).

Oral language skills showed a different pattern of relations than transcription skills. Oral language skills were hypothesized to relate to writing bursts based on prior evidence (Chenoweth & Hayes, 2001; Connelly et al., 2012). However, vocabulary had a suppression effect and grammatical knowledge was not independently related to writing fluency after accounting for transcription skills, working memory, and attentional control. In bivariate correlations, vocabulary was not related to writing bursts and words per time, and grammatical knowledge was weakly but positively related to writing bursts and words per time (see Online Supplemental Materials). As noted above, the suppression effect of vocabulary appears to be due to the interrelations among vocabulary, grammatical knowledge, spelling, and writing quality. The lack of the relations of oral language skills to writing fluency is divergent from a previous study which showed that oral language skill measured by the sentence formulation task was positively related to writing bursts (Connelly et al., 2012). In conjunction with the results on transcription skills, these results suggest two possibilities. First, writing fluency is largely influenced by transcription skills such that it does not reflect the translation process (hence no relation of oral language skills) for young beginning writers. This possibility is in line with the DIEW’s dynamic relations hypothesis as a function of development (Kim & Graham, 2022), which states that the contributions of component skills and knowledge to writing vary as a function of development. Specifically, in the beginning phase of writing development, transcription skills place large constraints on the writing process and therefore, transcription skills act as a primary determinant of writing process and outcomes. The influences of other concurrent processes such as translation and revision/evaluation are limited inasmuch as transcription is not automated. As transcription skills develop and the transcription process becomes more automated, the constraint of transcription skills on the other writing processes becomes increasingly reduced, allowing other processes to be tapped. Therefore, the nature of what writing fluency taps may change as a function of development: In the beginning phase of writing development, writing fluency largely taps the transcription process, whereas in a later phase, it additionally and increasingly taps other processes (e.g., translation, revision) and associated skills (e.g., oral language, background knowledge). The second possibility is that although theoretically vocabulary and grammatical knowledge are foundational for sentence skills (e.g., sentence construction or formulation), and vocabulary and grammatical knowledge are strongly related to sentence skills (Carrow-Woolfolk, 2017), sentence skills such as sentence construction might be more directly relevant for writing fluency than vocabulary and grammatical knowledge per se; in other words, sentence skills such as sentence construction, rather than vocabulary and grammatical knowledge, may be directly related to writing fluency if writing fluency taps the translation process for young writers. Future work is warranted to examine and shed light on these speculations.

Results also showed that working memory and attentional control were not directly related to writing fluency and writing quality. Instead, the relation of working memory to writing fluency was mediated by spelling skill, and its relation to writing quality were mediated by transcription and writing fluency. Surprisingly, attentional control was not related to directly or indirectly to writing fluency or writing quality. The indirect relation of working memory to writing quality is in line with recent evidence (Kim, 2020; Kim & Schatschneider, 2017). However, these results are discrepant from a recent study that examined similar mediated relations (Kim, 2022) and found a direct relation of attentional control to writing burst length, as well as a direct relation of working memory to writing quality over and above spelling, handwriting fluency, and discourse oral language. Reasons for the discrepancies are not clear. One important difference between Kim (2022) and the present study is the nature of data—the former used cross-sectional data whereas the latter used longitudinal data. By using longitudinal data, the present study overcomes some issues associated with cross-sectional data for examining causal mediation and revealed that the contribution of working memory to writing fluency and writing quality is indirect via spelling skill. Overall the present findings confirm the role of working memory in the writing process and the writing product, writing quality, which is widely recognized in theoretical models (Berninger & Winn, 2006; Graham, 2018; Hayes, 1996; Hayes & Flower, 1980; Kim & Graham, 2022; Kim & Park, 2019).

Inferences on practical implications are limited due to the correlational nature of the present study. However, together with causal evidence on the importance of language, cognitive, and transcription skills on writing quality (Graham et al., 2012; Graham & Perin, 2007; Graham & Santangelo, 2014; Santangelo & Graham, 2016), the findings suggest that maximally effective writing instruction needs to systematically and explicitly attend to these skills that support the writing process and consequent writing products. Specifically, high-quality instruction that supports development of children’s transcription skills (e.g., see Santangelo & Graham, 2016, for a meta-analysis) would help automate the transcription process, which allows young beginning writers to achieve quality writing (e.g., Alves et al., 2016; Limpo & Alves, 2018). In addition, the independent contribution of vocabulary to writing quality suggests the importance of quality instruction on vocabulary and oral language. Furthermore, strategies for attentional control and working memory (e.g., planning using graphic organizers) can support children’s writing process and its products (e.g., Graham et al., 2012).

Regarding the implications for assessment, the present findings suggest that, for young beginning writers, the product-based measure (i.e., the number of words written per writing time) is adequate as a measure of writing fluency. This conclusion is supported by the finding that both product-based and process-based measures are best described as measures of a single construct. Measuring writing bursts requires more complex procedures and, as a result, may not be practical in classroom and clinical settings. Therefore, utilizing the number of words written per writing time can be a more feasible and reliable method to assess writing fluency in this population.

Limitations, Future Directions, and Conclusion

One limitation of the present study is that some of the skills—working memory, attentional control, spelling, vocabulary, and grammatical knowledge—were measured using a single task whereas others were measured using multiple tasks and latent variables in data analysis. Latent variables are preferred because they have reduced measurement error. One example is working memory. In the present study, working memory was measured using a listening span task. This approach captures the ability to hold and process information simultaneously and has been widely used (e.g., Daneman & Merikle, 1996; Gaulin & Campbell, 1994; Nouwens et al., 2021). However, although children’s responses on the veracity of the statements were not scored and only their recall of final words was scored, the linguistic nature of the task might have influenced the present results. This possibility was examined in a previous meta-analysis, which found that verbal working memory (e.g., listening span task) was more strongly related with language comprehension (r = .41) than was nonverbal working memory (e.g., numerical or arithmetic span task, r = .30; Daneman & Merikle, 1996). Future studies using multiple measures of varying nature of working memory are warranted.

As noted above, the present findings reflect how writing bursts were measured. Although the present approach is aligned with previous research (e.g., Alves & Limpo, 2015; Alves et al., 2016; Limpo & Alves, 2017, 2018), future work can examine alternative approaches, such as counting the initial attempt of a word’s spelling as well as subsequent attempts. Additionally, due to the extremely limited occurrence of R-bursts (revision of content) in our sample, we chose not to distinguish between R-bursts and P-bursts. In future research, it would be valuable to explore whether it is beneficial to differentiate between these types of bursts for young beginning writers.

Another future direction is replication with children at a different developmental phase and learning to write in languages other than English to examine generalizability. A replication with individuals at a more advanced phase of writing development (e.g., upper elementary and secondary schools) would allow examination of P-bursts and R-bursts. It would also allow examination of the developmental hypothesis described above—whether what is tapped in writing bursts changes with development such that for writers in an advanced phase, writing bursts tap the translation and revision processes and associated skills and knowledge (e.g., oral language, background knowledge, writing strategies) in addition to the transcription process. In the present study, we used data from novice writers in primary grades in the US context, and future work with children at a more advanced phase of writing development and with children of various linguistic backgrounds is needed to reveal whether the nature of relations among component skills, the writing process, and written products varies as a function of writing development and writing systems.

The writing process draws on language, cognitive, and transcription skills, and knowledge, and the writing process and component skills together give rise to written products. Therefore, it is imperative to understand the relations among component skills, the writing process, and written products. The present study was an effort to further our understanding of their relations, and suggests a need for future work to expand our understanding.

Educational Impact And Implications Statement.

This study advances our understanding of how a written product, writing quality, is predicted by the writing production process measured by writing fluency, and language, cognitive, and transcription skills. We measured young beginning writers’ writing fluency by a process-based measure, writing bursts (chunks of text produced between pauses while writing) using a digital pen with an infrared camera at its nib, as well as a product-based measure, the number of words produced per total writing time. Results showed that children who had higher writing fluency—those who produced longer writing bursts (who can write more words and letters per burst) and wrote a greater number of words per total writing time—had higher quality writing, and this was explained directly by transcription skills and indirectly by executive functions such as working memory.

Appendix A: Writing Quality Rubric

Rubric for narrative tasks (TEWL and One Day)

1	2	3	4	5
Main idea is not clear; Contains at least 1 word that is relevant to the prompt; No details are provided; Ideas are so few that there is no room for or evidence of macro-organization (beginning, middle, & end), micro-organization (e.g., characters, setting, events, problem, resolution), or sequencing of ideas.	Main idea is not relevant to the prompt OR the piece is so short that only the most simplistic topic is emerging; Details are largely missing; Writing lacks sense of direction or macro-organization; content is strung together in loose, random fashion. There is no evidence of micro-organization (e.g., characters, setting, events, problem, resolution).	There are ideas that may be potentially considered as main ideas, but they are conveyed in a very general way; there is little elaboration of details or stated details are mostly extraneous; Some macro-organization is emerging (e.g., beginning), but it is rudimentary. There is little evidence of micro-organization (e.g., characters, setting, events, problem, resolution). Ideas have very little evidence of logical sequencing.	Main idea is emerging and developing, but they tend to be vaguely conveyed; Some details are provided, but mostly reads like a list; Some macro-organization is evident but may be confusing or ineffective. One section may be more developed than others. There is evidence of micro-organization (e.g., characters, setting, events, problem, resolution), but it is very basic. Logical sequencing of ideas begins to emerge.	Main ideas and themes are more evident although not fully developed. In the case of a narrative task, a sense of story emerges and begins to develop. Sometimes sequence of ideas is choppy; Basic relevant details are included, but some details may not be focused or are extraneous or confusing; Macro-organization of beginning, middle, and end is more clearly present but still under developed in some areas. An aspect of micro-organization (e.g., characters, setting, events, problem, resolution) might be mentioned, but there is no strong sense of micro-organization. Logical sequencing of ideas is developing but there are mistakes.
6	7	8	9	10
Main ideas tend to be coherent, but ideas could be more focused. Key details are included but not as developed as possible; some key information may be missing; Demonstrates clear macro-organization, but some parts may be less clear than others. There is a sense of micro-organization (e.g., characters, setting, events, problem, resolution) with key information (i.e., characters and problems).	Main ideas are reaching coherence towards a single theme, but there is room for greater clarity and focus. Key details provide relevant information; Clear and effective macro-organization of beginning, middle, and end is present; Micro-organization (e.g., characters, setting, events, problem, resolution) is developing with key information (i.e., characters and problems) but could be clearer (e.g., clear identification of characters without vague pronouns that have to be inferred). Sequencing of ideas tends to be effective (e.g., appropriate transition with or without using transition or linking words) but there are places where flow could be improved.	Main ideas are reaching coherence with focus, but may not be particularly original or unique. Key relevant details move reader through the composition; Clear and effective macro-organization is present. Micro-organization (e.g., characters, setting, events, problem, resolution) is more clearly present without a need for readers to make inferences. Sequence of ideas is tight and effective in general with an occasional mistake.	Main ideas are clearly coherent toward a single theme. Ideas and focal theme are original and unique. Key supporting details are interesting, accurate, developed, and elaborated; Clear and effective macro-organization is present and micro-organization (e.g., characters, setting, events, problem, resolution) supports the story’s theme and coherence. Sequence of ideas is tight and effective.	Main ideas are clearly coherent on a single theme. Ideas and focal theme are original and unique, precise and engaging. Author presents new ways of thinking about topic based on personal knowledge/experience. Key supporting details are interesting, accurate, developed, elaborated, thoughtful, and go beyond the obvious; Clear and effective macro-organization is present and micro-organization (e.g., characters, setting, events, problem, resolution) is fully and effectively structured. Sequence of ideas is tight and effective, and precise, using transition words accurately.

Rubric for the WIAT task

1	2	3	4	5
Responds to the prompt in a very simplistic manner; At least 1 key word is present (‘game’ by itself does not count); Details are completely missing. Ideas are so few that there is no room for a sense of organization.	May include only a single sentence identifying favorite game and a simplistic/general reason (e.g., “I like this game because it’s fun.”); Might have sentences that are irrelevant, not on topic, or detract from the overall clarity of the piece; little details are provided. With little information, organization does not emerge or is not relevant.	The main idea is conveyed in a single sentence or two and main idea is skeletal without any necessary details about the game; may include one basic reason explaining why they like the game beyond a simplistic/general reason, but details tend to be still basic without elaboration. Organization is vague or disorganized or confusing.	The main idea is conveyed in a general way but is elaborated to some extent (e.g., nature of a game) and has one or two reasons that are relevant to the prompt; reasons are not developed or elaborated and feels listy; a sense of organizational structure is emerging.	The main idea is conveyed clearly but may lack some important details about the game; has some reasons (not all) elaborated with some supporting and descriptive details although some relevant details may be still missing; organizational structure is clearer with the main idea followed by reasons.
6	7	8	9	10
The main idea is conveyed clearly and important details about the game are elaborated; has some reasons (not all) elaborated with supporting and descriptive details, but some aspects may include extraneous information or a lapse in logic or be less developed; organizational structure is clear with the main idea followed by reasons, but some parts (introduction, middle, and end) may be missing or not fully developed (e.g., no ending or very simplistic ending).	The main idea is conveyed clearly and important details about the game are elaborated; has some reasons elaborated with supporting and descriptive details in a convincing manner without extraneous or detracting points; organizational structure is clear with the main idea followed by reasons, but final thoughts (conclusion) are explicitly stated but incomplete or not fully developed.	The main idea is conveyed very clearly and coherently with all relevant information on the game and reasons; sequence of ideas is logical, accurate, and precise; reasons are elaborated with supporting and descriptive details in a convincing manner without extraneous or detracting points; details go beyond the obvious; organizational structure is clear with the main idea followed by reasons, and final thoughts (conclusion) are explicitly stated and developed.	The main idea is conveyed very clearly and coherently with all relevant information on the game and reasons; reasons are elaborated with supporting and descriptive details in a convincing manner without extraneous or detracting points; details go beyond the obvious and keep readers engaged; sequence of ideas is logical, accurate, and precise; organizational structure clearly supports expression of ideas with engaging introduction (hook), clearly articulated body (middle), and well-developed conclusion.	The main idea is conveyed very clearly and coherently with all relevant information on the game and reasons; sequence of ideas is logical, accurate, and precise; reasons are elaborated with supporting and descriptive details in a convincing manner without extraneous or detracting points; details go beyond the obvious and keep readers engaged; organizational structure clearly supports expression of ideas with engaging introduction (hook), clearly articulated body (middle), and well-developed conclusion. The difference between 9 and 10 is unique aspects and perspectives that are presented in an outstanding manner.

Rubric for the Beaver task

1	2	3	4	5
Responds to the prompt in a very simplistic manner; At least 1 key word is present; Missing or no details. With little information, organization does not emerge or is not relevant.	May include only a single sentence identifying the topic; Might have sentences that are irrelevant, not on topic, or detract from the overall clarity of the piece; little details are provided. With little information, organization does not emerge or is not relevant.	The main idea is conveyed in a single sentence or two and main idea is skeletal; may include one detail, but it tends to be still basic without elaboration. Organization is vague or disorganized or confusing.	The main idea is conveyed in a general way but is elaborated to some extent; details supporting the main idea are not developed or elaborated and feels listy; a sense of organizational structure is emerging.	The main idea is conveyed clearly but may lack some important details; some relevant details (not all) are elaborated with description; organizational structure is clearer with the main idea followed by reasons.
6	7	8	9	10
The main idea is conveyed clearly and important details about the topic are elaborated and descriptive but some aspects may include extraneous information or a lapse in logic or be less developed; organizational structure is clear with the main idea followed by reasons, but some parts (introduction, middle, and end) may be missing or not fully developed (e.g., no ending or very simplistic ending).	The main idea is conveyed clearly and important details about the topic are elaborated with descriptive and supporting information in a convincing manner without extraneous or detracting points; organizational structure is clear with the main idea followed by reasons, but final thoughts (conclusion) are explicitly stated but incomplete or not fully developed.	The main idea is conveyed very clearly and coherently with all relevant information on the topic; sequence of ideas is logical, accurate, and precise; reasons are elaborated with supporting and descriptive details in a convincing manner without extraneous or detracting points; details go beyond the obvious; organizational structure is clear with the main idea followed by reasons, and final thoughts (conclusion) are explicitly stated and developed.	The main idea is conveyed very clearly and coherently with all relevant information on supporting details; details are described in a convincing manner without extraneous or detracting points; details go beyond the obvious and keep readers engaged; sequence of ideas is logical, accurate, and precise; organizational structure clearly supports expression of ideas with engaging introduction (hook), clearly articulated body (middle), and well-developed conclusion.	The main idea is conveyed very clearly and coherently with all relevant information on the topic; details are elaborated with descriptive information in a convincing manner without extraneous or detracting points; details go beyond the obvious and keep readers engaged; sequence of ideas is logical, accurate, and precise; organizational structure clearly supports expression of ideas with engaging introduction (hook), clearly articulated body (middle), and well-developed conclusion. The difference between 9 and 10 is unique aspects and perspectives that are presented in an outstanding manner.

Appendix B: Words in the Spelling Task

Item Number	Kindergarten	Grade 1	Grade 2
1	is	hop	ship
2	the	ship	bump
3	you	bump	hang
4	fan	hang	pool
5	bed	pool	cute
6	hop	cute	know
7	ship	know	girl
8	bump	girl	drive
9	hang	drive	hurt
10	pool	hurt	catch
11	cute	catch	leaf
12	know	leaf	bright
13	girl	bright	shopping
14		shopping	sign
15		sign	phone
16		phone	bridge
17		bridge	making
18		making	island
19			clapped
20			sailor
21			scratches
22			marched