View in EDS HTML Full Text PDF Full Text

Do Children Classified with Specific Language Impairment Have a Learning Disability in Writing? A Meta-Analysis

Saved in:
Bibliographic Details
Title: Do Children Classified with Specific Language Impairment Have a Learning Disability in Writing? A Meta-Analysis
Language: English
Authors: Graham, Steve, Hebert, Michael, Fishman, Evan, Ray, Amber B. (ORCID 0000-0002-6527-7102), Rouse, Amy Gillespie
Source: Journal of Learning Disabilities. Jul-Aug 2020 53(4):292-310.
Availability: SAGE Publications and Hammill Institute on Disabilities. 2455 Teller Road, Thousand Oaks, CA 91320. Tel: 800-818-7243; Tel: 805-499-9774; Fax: 800-583-2665; e-mail: journals@sagepub.com; Web site: http://sagepub.com
Peer Reviewed: Y
Page Count: 19
Publication Date: 2020
Document Type: Journal Articles
Reports - Research
Information Analyses
Education Level: Elementary Education
Secondary Education
Descriptors: Writing Difficulties, Language Impairments, Learning Disabilities, Writing Skills, Speech Impairments, Oral Language, Age Differences, Classification, Educational Research, Writing Tests, Test Format, Preschool Children, Adolescents, Elementary School Students, Secondary School Students
DOI: 10.1177/0022219420917338
ISSN: 0022-2194
Abstract: In this meta-analysis, we examined whether children classified with specific language impairment (SLI) experience difficulties with writing. We included studies comparing children with SLI to (a) typically developing peers matched on age (k = 39 studies) and (b) typically developing younger peers with similar language capabilities (k = six studies). Children classified with SLI scored lower on writing measures than their typically developing peers matched on age (g = -0.97) when all writing scores in a study were included in the analysis. This same pattern occurred for specific measures of writing: quality (g = -0.92), output (g = -1.00), grammar (g = -0.68), vocabulary (g = -0.68), and spelling (g = -1.17). A moderator analysis revealed that differences in the writing scores of children classified with SLI and typically developing peers matched on age were not as large, but were still statistically significant, when assessment involved a contrived response format (vs. measured based on students' writing), researcher-created measures (vs. norm-referenced tests), or SLI included just children with a speech disorder (vs. children with a language disorder). Children classified with SLI further scored lower on writing than typically developing peers with similar language capabilities (g = -0.47). We concluded that children with SLI experience difficulties with writing.
Abstractor: As Provided
Entry Date: 2020
Accession Number: EJ1257088
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEdIS0LZFHU0x3k7J-qSV0sAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDF6gs_io7IHWHOOYUAIBEICBmhQBGPH85Tlj-KFeaJxQe25YYktyy3Fw-UdnOe42nQqjUtjlrNpAc9dDNqwbOy4V0fCU12w2_8oEsIf17dmTXWrrMMU1jS3Gh-9nUPFCClYqV9gqt26Q-IaNQ4hVLQN-IbHvfiMbEx4DtJlPS0L5oT8KQpVnexLIGw27N-cxDNX2qiIjoISHwjJQak_9CTOc9IO3ve9Wx4VOtm4=
Text:
  Availability: 1
  Value: <anid>AN0143828403;led01jul.20;2020Jun19.06:20;v2.2.500</anid> <title id="AN0143828403-1">Do Children Classified With Specific Language Impairment Have a Learning Disability in Writing? A Meta-Analysis </title> <p>In this meta-analysis, we examined whether children classified with specific language impairment (SLI) experience difficulties with writing. We included studies comparing children with SLI to (a) typically developing peers matched on age (k = 39 studies) and (b) typically developing younger peers with similar language capabilities (k = six studies). Children classified with SLI scored lower on writing measures than their typically developing peers matched on age (g = −0.97) when all writing scores in a study were included in the analysis. This same pattern occurred for specific measures of writing: quality (g = −0.92), output (g = −1.00), grammar (g = −0.68), vocabulary (g = −0.68), and spelling (g = −1.17). A moderator analysis revealed that differences in the writing scores of children classified with SLI and typically developing peers matched on age were not as large, but were still statistically significant, when assessment involved a contrived response format (vs. measured based on students' writing), researcher-created measures (vs. norm-referenced tests), or SLI included just children with a speech disorder (vs. children with a language disorder). Children classified with SLI further scored lower on writing than typically developing peers with similar language capabilities (g = −0.47). We concluded that children with SLI experience difficulties with writing.</p> <p>Keywords: writing; oral language; speech and language difficulties; learning disabilities; meta-analysis</p> <p>It is estimated that 3% to 10% of children have a developmental disorder commonly referred to as specific language impairment (SLI; [<reflink idref="bib86" id="ref1">86</reflink>]). This developmental disorder includes children who experience difficulty with speech, language, or both. SLI is identified when oral language lags behind other areas of development for no apparent reason ([<reflink idref="bib49" id="ref2">49</reflink>]). Children classified with SLI have adequate hearing and no major handicapping conditions such as acquired brain injury, deafness, blindness, autism spectrum disorders (ASD), or intellectual disabilities ([<reflink idref="bib81" id="ref3">81</reflink>]). Even so, it is commonly assumed that many children classified with SLI experience difficulty learning basic academic skills such as reading ([<reflink idref="bib8" id="ref4">8</reflink>]; [<reflink idref="bib14" id="ref5">14</reflink>]).</p> <p>It is also frequently assumed that many children classified as SLI experience difficulties with writing (e.g., [<reflink idref="bib53" id="ref6">53</reflink>]; [<reflink idref="bib76" id="ref7">76</reflink>]). The current review examined this proposition by conducting a meta-analysis of studies comparing the writing performance of children classified with SLI with the writing performance of their chronological age (CA) and language ability (LA) matched peers. This allowed us to examine whether writing difficulties among children with SLI are prevalent, involve multiple features of writing, reflect different levels of language, differ by type of language difficulty, diverge by type of writing assessments, and vary depending on study quality. To better situate the findings from our meta-analysis, we compared them with effect sizes (ESs) obtained in a previous meta-analysis which compared the writing performance of students classified as having a learning disability (LD) with the writing performance of their CA peers ([<reflink idref="bib36" id="ref8">36</reflink>]).</p> <hd id="AN0143828403-2">Why Are Children With SLI Likely to Experience Difficulties With Writing?</hd> <p>We offer three theoretical explanations for why children classified with SLI are likely to experience difficulties with writing. Two of these explanations focus on the relations between oral language and writing (one from a cognitive perspective and the other from a social cultural view), whereas the third explanation proposes that language disorders are a consequence of problems with basic learning processes which are also manifested in difficulties in writing.</p> <hd id="AN0143828403-3">Shared Knowledge Theory</hd> <p>It is commonly assumed that writing is a language-based skill (e.g., [<reflink idref="bib5" id="ref9">5</reflink>]; [<reflink idref="bib79" id="ref10">79</reflink>]). This view stresses that oral language and writing are similar ([<reflink idref="bib82" id="ref11">82</reflink>]), and writers draw on their oral language capabilities to carry out writing processes. Accordingly, stronger oral language skills provide a better foundation for writing, but difficulties with oral language can manifest themselves in children's writing in adverse ways.</p> <p>The overlap between oral language and writing can be illustrated by comparing an influential model of oral language developed by [<reflink idref="bib50" id="ref12">50</reflink>] to a recent model of writing developed by [<reflink idref="bib35" id="ref13">35</reflink>]. The Levelt model described autonomous processes that allow speakers to go from conceptualizing their spoken intentions to articulating them. The <emph>conceptualizer</emph> is responsible for generating a communicative intent, turning it into a coherent conceptual plan, and monitoring what has been and will be said. The speaker draws upon what has already been said, their general knowledge, as well as their knowledge of the speaking situation to generate intentions and monitor the communication process. The <emph>formulator</emph> gives grammatical and phonological shape to the intended message, transforming the organized preverbal plan into an acceptable linguistic structure that is likely to be understood by the listener. This involves drawing on semantic, phonological, morphological, and syntactic knowledge to create an utterance, phrase, or sentence that successfully conveys the speaker's intended message. The <emph>articulator</emph> executes the motor mechanisms (i.e., respiratory, laryngeal, and supralaryngeal systems) involved in producing the spoken message, leading to <emph>audition</emph> or overt speech. Speakers make use of a <emph>speech-comprehension system</emph> to listen to and process what they are saying (overtly and covertly) as well as what others are saying, which feeds back into the <emph>conceptualizer</emph> which monitors the whole communication process.</p> <p>Analogous production processes are also evident in writing. For instance, [<reflink idref="bib35" id="ref14">35</reflink>] described a series of mental and physical operations writers apply to produce text. This includes <emph>conceptualization</emph>, which involves constructing a mental representation of the writing task. This roadmap can be represented in multiple ways (e.g., remembered goals, written plan, diagrams, pictures, text produced so far); is shaped and constrained by the writers' intentions and knowledge, goals of the writing community, or both; and is open to change through self-monitoring and feedback. The process of <emph>ideation</emph> involves drawing possible writing ideas from memory or external sources. Some ideas may undergo scrutiny by the writer to determine whether they are suitable given how the writing task is conceptualized, whereas other ideas may receive only a cursory examination. <emph>Translation</emph> involves transforming ideas into acceptable language that conveys the author's intended meaning. This includes deciding which words and syntactic structures best convey an intended meaning and draws on semantic, phonological, morphological, and syntactic knowledge. <emph>Transcription</emph> is physically transcribing the phrases and sentences produced by the writer into written text. This includes applying skills such as spelling, handwriting, and typing as well as knowledge about punctuation and capitalization. Finally, <emph>reconceptualization</emph> applies to all production processes described above, as writers can decide to modify plans, intentions, ideas, language, and transcription processes as they monitor the writing process, what they write, and receive feedback from others.</p> <p>Although these two theoretical descriptions of the production process involved in oral language and writing make it clear that they are not identical ([<reflink idref="bib72" id="ref15">72</reflink>]), both communication modes rely on many of the same underlying knowledge resources (e.g., phonological, morphological, semantic, syntactic, and pragmatic knowledge) and apply similar production processes (e.g., generating and monitoring intentions and production, translating messages/ideas into acceptable language, bringing this language to life by articulating it or putting it into written form). Theoretically then, difficulties with oral language skills or processes may manifest themselves in children's writing, as both oral language and writing draw on the same or similar knowledge and processes ([<reflink idref="bib16" id="ref16">16</reflink>]).</p> <p>To illustrate how children's oral language difficulties can become manifest in their writing, we return to the model developed by [<reflink idref="bib50" id="ref17">50</reflink>]. Using this model, [<reflink idref="bib52" id="ref18">52</reflink>] identified a variety of different oral language problems children encounter. This included limited and slow access to vocabulary, which is likely to influence the number of different words children use when writing and how much they write. Oral language problems with pragmatics can potentially impact how children's written messages are received and evaluated by readers. Issues involving phonology and morphology can predictably result in spelling errors, as children use their knowledge of phonology to break words into sounds that are matched with specific letters, whereas morphological knowledge is used to determine specific spellings, segment words into parts, and mark important elements in writing such as tense and number. Oral language problems involving syntax are likely to create grammatical errors in writing and poorly formed or incomplete sentences. Articulation and speech disorders are likely to result in misspelled and misused words, while fluency disorders can possibly impact writing output and cohesion. In combination, these oral language problems can impact multiple aspects of writing. These examples of how oral language impacts writing correspond to the writing outcomes in this review.</p> <p>Evidence that oral language and writing are connected (and by extension oral language difficulties manifest in children's writing) is relatively limited ([<reflink idref="bib79" id="ref19">79</reflink>]), but includes multiple lines of research. For example, children with less proficient verbal working memory produce papers of poorer quality than students with stronger verbal working memory capabilities (e.g., [<reflink idref="bib58" id="ref20">58</reflink>]); morphological errors made when speaking are evident in students' writing (see [<reflink idref="bib13" id="ref21">13</reflink>]); children whose speaking is characterized by weak cohesive harmony demonstrate a similar pattern in their writing (see [<reflink idref="bib30" id="ref22">30</reflink>]); oral language growth over time is associated with better writing (e.g., [<reflink idref="bib60" id="ref23">60</reflink>]); oral phonological awareness instruction enhances students' spelling (ES of 0.69 in 20 experiments; [<reflink idref="bib38" id="ref24">38</reflink>]); and vocabulary instruction improves the quality of students' text (ES of 0.78 in three experiments; [<reflink idref="bib37" id="ref25">37</reflink>]).</p> <hd id="AN0143828403-4">Rhetorical Relations Theory</hd> <p>Another position on how oral language influences writing is based on sociocultural theory, conceptualizing oral language and writing as meaning-making social activities ([<reflink idref="bib82" id="ref26">82</reflink>]). Like speakers with their audience, writers engage in a dialogue with readers. This view blurs the distinction between oral language and writing, as each is inherently conversational ([<reflink idref="bib5" id="ref27">5</reflink>]). Although writers must deal with the problem of an absent reader, participating in oral conversations may provide insights that can be used when writing.</p> <p>To illustrate, children develop pragmatic oral language skills as they engage in conversations with others, allowing them to use oral language to fulfill communicative purposes and regulate social interactions ([<reflink idref="bib3" id="ref28">3</reflink>]). As [<reflink idref="bib88" id="ref29">88</reflink>] noted, pragmatic oral language competence can become evident in children's writing, as they apply what they learned in oral situations to writing. This can influence how writers structure conversations with readers and how they apply various aspects of language (e.g., form and function) while composing. In contrast, a child who is not particularly adept at perspective-taking when speaking may be less likely to consider what readers' believe, know, or want. Similarly, a child who experiences difficulties understanding figurative language when speaking may be less likely to use such language when writing. Although direct evidence on the connections between pragmatic oral language skills and writing is scarce, several studies reported that children with oral language problems experience difficulties using figurative language in their writing (e.g., [<reflink idref="bib66" id="ref30">66</reflink>]).</p> <hd id="AN0143828403-5">Learning Deficit Theory</hd> <p>Although SLI is generally viewed as a deficit/delay specific to the domain of language, an alternative explanation is that oral language difficulties are a result of problems in basic learning processes ([<reflink idref="bib85" id="ref31">85</reflink>]). Multiple explanations for how dysfunctions of nonlinguistic learning processes lead to oral language problems have been championed. This includes claims that general learning deficits, such as processing capacity limitations, are responsible for oral language difficulties. Increasingly though, more localized learning dysfunctions have been proposed. For instance, [<reflink idref="bib89" id="ref32">89</reflink>] proposed that oral language problems occur as a result of dysfunctions in the procedural memory system (PMS). This brain system is engaged in learning new motor and cognitive skills (especially skills involving sequencing) as well as maintaining control over them once they become more habitual.</p> <p>An example of how learning impairments can negatively impact oral language as well as literacy skills was provided by [<reflink idref="bib45" id="ref33">45</reflink>]. They reviewed evidence that supported the proposition that dysfunctions in PMS impact not only oral language but reading as well. When compared with CA peers, children with language and reading difficulties were less adept at consolidating and retaining sequential information during implicit learning. Although such comparisons are not available for oral language and writing, [<reflink idref="bib34" id="ref34">34</reflink>] reported that poor spellers are less adept at implicitly learning new spellings (a skill that involves sequential learning) than good spellers. A PMS dysfunction can even account for why some children with oral language disorders experience handwriting difficulties ([<reflink idref="bib20" id="ref35">20</reflink>]). Handwriting is a motor skill involving sequential learning (letters are created by executing motor movements in particular sequences), and fluency in handwriting is mainly established implicitly as children write ([<reflink idref="bib39" id="ref36">39</reflink>]). This does not mean though that handwriting is solely a motor skill, as [<reflink idref="bib1" id="ref37">1</reflink>] reported that orthographic recoding (representing a word in memory and then accessing specific elements of that representation) was a better predictor of handwriting than fine-motor skills.</p> <hd id="AN0143828403-6">Current Meta-Analysis</hd> <p>This review involved a comprehensive meta-analysis of studies comparing the writing performance of children classified as SLI to the writing performance of peers with no known language difficulties. This included two groups of typically developing peers. One group included children with no known language difficulties match by CA to children in the SLI group in a study. The other group of included younger children with no known language difficulties matched by LA to children in the SLI group in a study.</p> <p>By examining difference in the writing of children classified as SLI and their CA peers, we were able to gauge the direction and magnitude of the writing problems experienced by children classified as SLI when compared with children who had the same amount of time to develop as writers. Based on the three theoretical positions reviewed earlier, we expected that the writing scores of children classified as SLI would be substantially lower than the writing scores of their CA peers when all measures were considered collectively as well as when specific features of writing were considered separately. We further expected that the ESs for writing scores of children classified with SLI would be similar to those of children classified as LD (in [<reflink idref="bib36" id="ref38">36</reflink>]), given the importance of oral language to writing ([<reflink idref="bib76" id="ref39">76</reflink>]).</p> <p>Comparing children classified as SLI to LA peers allowed us to examine whether language difficulties were still related to lower writing scores when oral LAs were comparable. Although children classified as SLI had more time to learn to write, we still predicted that their writing scores would be lower than the scores of their younger LA peers. We anticipated that the writing difficulties of children classified as SLI reflect more than just their level of language (which would be supported if the writing of the two groups of children were comparable). Instead, we expected that their writing problems are atypical, reflecting specific aspects of SLI. This could be due to underlying learning difficulties or the disruptive and potentially lasting impact of oral language difficulties on writing.</p> <hd id="AN0143828403-7">Moderating Factors</hd> <p>A moderator analysis was conducted to determine whether six specific study features (children's age, test type, test format, quality of study, peer-review, and type of language disorder) were statistically associated with variation in obtained effects. We thought it probable that writing differences between children classified as SLI and typically developing peers would vary with age. As children classified as SLI receive services to improve their oral language capabilities, newly strengthened oral language skills should provide a better foundation for writing growth, reducing possible writing differences between these two groups of children. However, it is possible though that once writing problems begin to form, they continue and may worsen even after oral language problems are overcome ([<reflink idref="bib65" id="ref40">65</reflink>]).</p> <p>We also considered it likely that variations in writing differences between children classified with SLI and typically developing peers would be associated with how writing was assessed. It was anticipated that writing scores from norm-referenced tests would differ from scores obtained from researcher-designed tests. Norm-referenced tests are typically designed to assess a construct broadly, whereas researcher-designed tests are often more narrowly construed ([<reflink idref="bib38" id="ref41">38</reflink>]). For instance, spelling measures for norm-referenced tests generally focus on whether students can correctly spell a set of specific words that represent spelling capabilities broadly, whereas researcher-designed measures of spelling words commonly involve items selected to assess particular aspects of spelling (e.g., phonological acceptable and unacceptable spelling of regular and irregular spelling words; [<reflink idref="bib12" id="ref42">12</reflink>]).</p> <p>We further anticipated that writing scores obtained from assessments based on children's writing would differ from scores obtained using a contrived response format (e.g., spelling a list of words, sentences constructed to contain specific grammar errors, or sentences with a missing word that must be supplied). A contrived response format may put children at a disadvantage ([<reflink idref="bib42" id="ref43">42</reflink>]). To illustrate, children may be more likely to use grammatical skills already mastered when writing, but they have no choice about what aspect of grammar is assessed when a measure involves a contrived response format.</p> <p>It was also anticipated that study quality would be related to variation in ESs. Methodological rigor should minimize inflated outcomes or erroneous findings ([<reflink idref="bib69" id="ref44">69</reflink>]). Writing differences between children classified as SLI and their typically developing peers, therefore, should vary based on the overall soundness of studies. A meta-analysis by [<reflink idref="bib38" id="ref45">38</reflink>] examining reading instruction's effects on writing provides partial support for this prediction, as study design was related to variability in effects.</p> <p>We further expected that peer-review would account for variability in ESs. Studies that are undergo peer-review are more likely to evidence larger effects than studies that do not undergo such an analysis, as studies with smaller effects are less likely to be published (see [<reflink idref="bib15" id="ref46">15</reflink>]; [<reflink idref="bib68" id="ref47">68</reflink>] for support for this proposition).</p> <p>Finally, the type of writing difficulties children classified with SLI experience are likely related to differences in their language problems. We expected that the writing of children with just speech problems would be less impacted than the writing of children who had broader oral language difficulties (i.e., expressive language, receptive language, or both). Empirically, this prediction is supported by a prior meta-analysis ([<reflink idref="bib42" id="ref48">42</reflink>]) showing that students classified with SLI are likely to have more serious difficulties with writing (i.e., spelling) when they experience multiple language problems. Because children classified with SLI in the studies reviewed here were mostly identified as having both expressive and receptive language disorders, and separate data for different types of oral language difficulties were rarely presented, we were unable to conduct other oral language comparisons.</p> <hd id="AN0143828403-8">Previous Review</hd> <p>We located one previous meta-analysis that compared the writing of children classified with SLI and their CA and LA peers ([<reflink idref="bib42" id="ref49">42</reflink>]). Unlike our meta-analysis which focused on all measures of writing performance, this previous review focused only on spelling performance. They found that children with SLI evidenced lower spelling performance than CA peers (ES = −1.42), but there was no statistical difference in the spelling performance of children with SLI when compared with LA peers. Heterogeneity among ESs was large, but these differences were not related to children's age, language spoken, or task used to measure spelling.</p> <p>Twenty of the spelling studies included in [<reflink idref="bib42" id="ref50">42</reflink>] were included in our review, but 12 investigations were not suitable as: (a) children were classified with SLI, at least in part, with a writing measure (spelling); (b) CA and LA peers likely experienced some oral language issues as they had been referred for a clinical assessment; or (c) normative test data were used as a proxy for the actual performance of CA or LA peers. We also identified 10 additional spelling studies not included in Joye et al., as we included studies with children from preschool to Grade 12 (the prior review focused on just 5 to 18 years of age) as well as published and unpublished studies (the previous review included only published studies). Furthermore, we systematically assessed the quality of each study; this was not done in the prior review. Finally, we applied a different analytical approach than Joye et al. If a study in Joye et al. had multiple measures, the individual ESs were averaged to obtain an overall single ES for the study. We used a robust variance estimation (RVE) approach, allowing us to include multiple ESs from each study, without having to average ESs or combine groups within a study.</p> <hd id="AN0143828403-9">Research Questions</hd> <p>The primary purpose of this meta-analysis was to determine whether children classified as SLI experience difficulties with writing. To address this purpose, we posed the following questions:</p> <p></p> <ulist> <item> <bold> Research Question 1 (RQ1): </bold> Are there differences in the writing of children classified with SLI and their CA peers?</item> <p></p> <item> <bold> Research Question 2 (RQ2): </bold> Are there differences in the writing of children classified with SLI and LA peers?</item> <p></p> <item> <bold> Research Question 3 (RQ3): </bold> What specific features of writing are problematic for children classified with SLI?</item> <p></p> <item> <bold> Research Question 4 (RQ4): </bold> Are effects moderated by children's age, test features (type and format), quality of study, peer-review, and type of oral language disorder (speech vs. language)?</item> </ulist> <p>As is the case in all systematic reviews, meta-analysts can only analyze what was studied. Across investigations, a wide range of measures were used to assess children's writing (with a few measures also assessing composing rate). For RQ3, we were only able to examine the following five writing outcomes separately: writing quality, output, grammar, vocabulary, and spelling. There were not enough degrees of freedom to recommend doing such analyses for the SLI and CA comparison with other measures (e.g., handwriting, elements) or for any measure involving the SLI and LA comparison (there were only six such studies altogether). Likewise for RQ4, a moderator analysis could not be conducted for the SLI and LA comparison.</p> <hd id="AN0143828403-10">Method</hd> <p></p> <hd id="AN0143828403-11">Inclusion Criteria</hd> <p>Studies in this meta-analysis met all of the following seven criteria: (a) Participants were in preschool to Grade 12. (b) The study included at least one group of children classified as SLI, or a group of children who met the definition for SLI (oral language lags behind other areas of development for no apparent reason; [<reflink idref="bib49" id="ref51">49</reflink>]). (c) The study included at least one group of children in the same grades as children classified with SLI and/or LA children matched on LA. (d) Children's writing was assessed with at least one or more measures. (e) The study included at least eight participants in the SLI, CA, and LA groups each. (f) The study was written in English. (g) The study included the data necessary to calculate a weighted ES (or such information was available from the authors).</p> <p>We excluded studies where children with oral language problems experienced brain injury, deafness, blindness, ASD, or intellectual disabilities. Studies were excluded if the identification of children classified as SLI involved the use of a writing measure (i.e., spelling), as this ensured that students in the SLI group experienced difficulties with this aspect of writing. Studies were also excluded if CA or LA children's writing performance was purposefully restricted, as this artificially inflates or restricts writing differences between children classified with SLI and these two groups. For example, if CA students were purposefully selected as above average writers, this would increase the difference in writing scores between CA and SLI students. If CA students were purposefully selected to be poor writers, this would restrict writing score differences between CA and SLI students.</p> <p>We further excluded studies where CAs or LAs had received or been referred for possible academic or language problems, as this decreases the possibility there would be writing differences between children classified with SLI and CA or LA peers. It also increases the likelihood that CAs or LAs previously experienced oral language difficulties. Finally, we excluded studies where students produced an oral retell of information presented to them or they dictated text. Dictation does not involve students producing text through their own writing and oral retells are a language measure (our focus was on the relation between oral language competence and written language production). These inclusion/exclusion criteria helped to ensure that we obtained an accurate estimation of the writing capabilities of SLI students when compared with typical CA and LA matched peers.</p> <hd id="AN0143828403-12">Search Procedures and Selection of Studies</hd> <p>We applied four strategies to identify studies for this review. (a) Electronic searches (ending May 2017) were conducted in five databases: ERIC, PsychINFO, Education Sources, ProQuest Dissertations and Theses, and World Cat. For each electronic search, the following terms were paired with writing: speech impairment, language impairment, language disorder, developmental language disorder, speech disorder, language delay, specific language impairment, and complex communication needs. These terms were commonly used in prior meta-analyses with this population (e.g., [<reflink idref="bib26" id="ref52">26</reflink>]). (b) We contacted 12 authors who previously conducted studies involving writing and SLI to identify additional studies. (c) When the full-text of a document was obtained, the reference list was searched to identify other pertinent studies. (d) We conducted a hand search of the following journals from their date of first publication to May 2017: <emph>Annals of Dyslexia, Journal of Communication Disorders, Journal of Speech, Language, and Hearing Research, Journal of Writing Research, Language Speech & Hearing Services in the Schools, Reading & Writing, Research in the Teaching of English</emph>, and <emph>Written Communication.</emph> We selected these journals as they commonly publish studies involving students with SLI, focused on writing, or both.</p> <p>Through the electronic searches, we identified 8,413 papers (ERIC = 128, PsychINFO = 1,058, Education Sources = 769, ProQuest Dissertations and Theses = 5,441, and World Cat = 1,017). After eliminating duplicates, 7,874 papers remained. The first author and a graduate student (GA) read the abstract and title of each of these papers to determine whether it met inclusion criteria. Agreement was 97% (all disagreements were resolved by discussion with the first author). This resulted in the identification of 90 possible studies. Correspondences sent to researchers doing work in this area resulted in the identification of an additional nine possible studies; reference lists of obtained studies resulted in the identification of 22 possible studies; and hand searches of journals resulted in the identification of 11 possible papers.</p> <p>The first author and a GA independently read the 132 acquired papers to determine whether each met all inclusion criteria. If a disagreement occurred (again this was rare), a final decision was facilitated by the first author. Of the 132 papers reviewed in full, 47 met our criteria for inclusion. These 47 papers yielded data from 39 unique studies (some studies presented data for the same children over time or included data from a study in more than a single paper).</p> <p>The most common reasons for eliminating a study were: (a) no CA or LA children were included in the study (<emph>k =</emph> 20); (b) no writing measure was administered (<emph>k =</emph> 17); (c) no children classified with SLI were in the study (<emph>k =</emph> 12); (d) a writing measure was used in the selection of children with SLI (<emph>k</emph> = 10); (e) children retold a story and did not create one (<emph>k</emph> = 9); (f) an ES could not be computed (<emph>k =</emph> 9); (g) children dictated and did not write (<emph>k</emph> = 7); and (h) CAs were originally part of a clinic population (<emph>k</emph> = 1).</p> <hd id="AN0143828403-13">Coding of Studies</hd> <p>We coded writing measures in a study as assessing one of the following 12 writing outcomes: writing quality (based on holistic or analytic scales that assess attributes such as ideation, organization of text, vocabulary, voice, sentence fluency, and conventions as well as the overall score on a norm-referenced test of writing); output (based on total writing output, including words, sentence, or <emph>t</emph>-units); grammar (based on measures of grammar, usage, and syntactic complexity); elements (based on number of schematic elements included in text that were relevant to a writing genre, such as the inclusion of characters, setting, and events in narrative writing); vocabulary (based on vocabulary diversity and/or the semantic accuracy of words used in text); spelling (based on correctness of word spelling); handwriting (based on measures of legibility and/or fluency); ideation (based on the clarity, substance, and/or richness of ideas conveyed in a written text); organization (based on the logical progression of ideas in a paper, including the establishment of transitions between paragraphs and sentences); sentences (based on type of sentences); rate (based on words per minute when writing and <emph>t</emph>-units written per minute); and composing time (based on time spent writing).</p> <p>All writing outcome measures were derived from a sample of children's writing or from a contrived response task not involving writing a composition. For example, spelling was assessed as number of correctly spelled words: (a) in a paper or (b) on a spelling test.</p> <p>We coded all studies for the following characteristics: publication type (journal, dissertation, conference, book/book chapter, or other); peer-review (yes or no); author(s); year published; age (grade) of participants; location from which study participants were recruited (school, hospital, other, not specified); language problems of children classified with SLI (receptive, expressive, speech [including articulation, apraxia, and stuttering], other, not specified); comorbid condition(s); sample size, gender and ethnicity of participants; country and community location (urban, suburban, rural, and not specified) of the study; and socioeconomic status (SES) and language of participating children. In most studies, students were identified as having both expressive and receptive delays, and in some studies they also had difficulties with speech. Table 1 provides information for each study, specifying whether the difficulty was singular (e.g., receptive language problem) or a combination of problems (e.g., expressive, receptive, speech).</p> <p>Graph</p> <p>Table 1. Study Characteristics and Average Effect Sizes for Writing Variables for the SLI and CA Comparisons.</p> <p> <ephtml> <table><colgroup><col align="left" /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /></colgroup><thead><tr><th align="center">Study</th><th align="center">Age (Grd.)</th><th align="center">Writing tasks</th><th align="center"><italic>N</italic></th><th align="center">Eng</th><th align="center">RLP</th><th align="center">ELP</th><th align="center">SP</th><th align="center">Other</th><th align="center">Study quality (%)</th><th align="center">Average <italic>g</italic></th></tr></thead><tbody><tr><td><xref ref-type="bibr" rid="bibr2">Anderson (2010)</xref></td><td>6–19</td><td>NR-Sp</td><td>74</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>100</td><td>−1.35</td></tr><tr><td><xref ref-type="bibr" rid="bibr7">Bishop & Clarkson (2003)</xref>*</td><td>7–13</td><td>Narr, NR-Sp</td><td>56</td><td>Yes</td><td>X</td><td>X</td><td>X</td><td /><td>50</td><td>−0.63</td></tr><tr><td><xref ref-type="bibr" rid="bibr6">Bird et al. (1995)</xref>*</td><td>6–8</td><td>NR-Sp</td><td>62</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>50</td><td>−1.51</td></tr><tr><td><xref ref-type="bibr" rid="bibr11">Brizzolara et al. (2011)</xref>*</td><td>14–18</td><td>NR-Sp</td><td>48</td><td>No</td><td>X</td><td>X</td><td /><td /><td>100</td><td>−1.39</td></tr><tr><td><xref ref-type="bibr" rid="bibr12">Broc et al. (2013)</xref>*/<xref ref-type="bibr" rid="bibr27">Favart et al. (2016)</xref>*</td><td>7–18</td><td>Narr, Spelling</td><td>72</td><td>No</td><td /><td>X</td><td /><td /><td>75/75</td><td>−0.62</td></tr><tr><td><xref ref-type="bibr" rid="bibr16">Connelly et al. (2012)</xref>*/<xref ref-type="bibr" rid="bibr20">Critten et al. (2014)</xref>*</td><td>7–11</td><td>NR-Sp, Personal Narr, HW</td><td>66</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>92/75</td><td>−1.35</td></tr><tr><td><xref ref-type="bibr" rid="bibr17">Conti-Ramsden et al. (2012)</xref>*/<xref ref-type="bibr" rid="bibr25">Durkin et al. (2011)</xref>*</td><td>16–17</td><td>NR-Sp, Email</td><td>96</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>75/75</td><td>−0.57</td></tr><tr><td><xref ref-type="bibr" rid="bibr18">Cordewener et al. (2012)</xref>*</td><td>6–7</td><td>NR-Sp, Spelling</td><td>98</td><td>No</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn2">a</xref></td><td>38</td><td>−1.36</td></tr><tr><td><xref ref-type="bibr" rid="bibr21">Davidi & Berman (2014)</xref></td><td>12–13</td><td>Informative</td><td>20</td><td>No</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn3">b</xref></td><td>50</td><td>−1.15</td></tr><tr><td><xref ref-type="bibr" rid="bibr22">de Bree et al. (2009)</xref>*</td><td>8</td><td>Spelling</td><td>38</td><td>No</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn2">a</xref></td><td>100</td><td>−0.98</td></tr><tr><td><xref ref-type="bibr" rid="bibr24">Dockrell & Connelly (2013)</xref>*</td><td>7–10</td><td>NR-Sp, Informative</td><td>46</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>100</td><td>−1.28</td></tr><tr><td><xref ref-type="bibr" rid="bibr28">Ferouhi (2006)</xref></td><td>9–11</td><td>NR-Sp, NR-Narr</td><td>40</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>100</td><td>−0.91</td></tr><tr><td><xref ref-type="bibr" rid="bibr29">Fey et al. (2004)</xref>*/<xref ref-type="bibr" rid="bibr56">McCarthy et al. (2012)</xref>*</td><td>(2–4)</td><td>Narr</td><td>373</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>100/100</td><td>−0.36</td></tr><tr><td><xref ref-type="bibr" rid="bibr33">Goulandris et al. (2000)</xref>*</td><td>14–18</td><td>Spelling</td><td>50</td><td>Yes</td><td>X</td><td>X</td><td>X</td><td /><td>25</td><td>−1.19</td></tr><tr><td><xref ref-type="bibr" rid="bibr40">Hedberg & Fink (1996)</xref>*</td><td>(1–6)</td><td>Narr</td><td>56</td><td>Yes</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn2">a</xref></td><td>100</td><td>−0.73</td></tr><tr><td><xref ref-type="bibr" rid="bibr41">Jacobs (2009)</xref>*</td><td>7–8</td><td>NR-Sp</td><td>84</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>75</td><td>−1.30</td></tr><tr><td>Kim et al.(2015)*</td><td>(1)</td><td>Narr, skills tests</td><td>304</td><td>Yes</td><td>X</td><td>X</td><td>X</td><td /><td>100</td><td>−0.64</td></tr><tr><td><xref ref-type="bibr" rid="bibr44">Koutsoftas & Gray (2011)</xref>*</td><td>(4–5)</td><td>Narr, Exp, skills tests</td><td>56</td><td>Yes</td><td /><td>X</td><td /><td /><td>75</td><td>−0.72</td></tr><tr><td><xref ref-type="bibr" rid="bibr46">Larkin & Snowling (2008)</xref>*</td><td>(4–5)</td><td>Spelling</td><td>46</td><td>Yes</td><td>X</td><td>X</td><td>X</td><td /><td>100</td><td>−1.01</td></tr><tr><td><xref ref-type="bibr" rid="bibr48">Lawrence (2008)</xref></td><td>6–8</td><td>Spelling</td><td>88</td><td>Yes</td><td /><td /><td>X</td><td /><td>100</td><td>−1.19</td></tr><tr><td><xref ref-type="bibr" rid="bibr51">Lewis et al. (1998)</xref>*</td><td>7–14</td><td>NR-Sp, NR-Narr</td><td>51</td><td>Yes</td><td /><td /><td>X</td><td /><td>87</td><td>−0.57</td></tr><tr><td><xref ref-type="bibr" rid="bibr53">Mackie & Dockrell (2004)</xref>*</td><td>6–12</td><td>Narr</td><td>22</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>75</td><td>−1.54</td></tr><tr><td><xref ref-type="bibr" rid="bibr54">Mackie et al. (2013)</xref>*</td><td>7–11**</td><td>Narr</td><td>88</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>75</td><td>−0.89</td></tr><tr><td><xref ref-type="bibr" rid="bibr55">Magnusson & Nauclér (1990)</xref>*/<xref ref-type="bibr" rid="bibr64">Nauclér & Magnusson (1999)</xref>*/<xref ref-type="bibr" rid="bibr63">Nauclér (2004)</xref>*</td><td>(1, 5, 12)</td><td>Spelling</td><td>111</td><td>No</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn4">c</xref></td><td>75/75/75</td><td>−0.71</td></tr><tr><td><xref ref-type="bibr" rid="bibr59">McFadden & Gillam (1996)</xref>*</td><td>7–11</td><td>Narr</td><td>20</td><td>Yes</td><td /><td>X</td><td /><td /><td>100</td><td>−1.19</td></tr><tr><td><xref ref-type="bibr" rid="bibr61">Myklebust (1965)</xref></td><td>7–17</td><td>Narr</td><td>636</td><td>Yes</td><td /><td /><td>X</td><td /><td>50</td><td>−0.92</td></tr><tr><td><xref ref-type="bibr" rid="bibr62">Nathan et al. (2004)</xref>*</td><td>4</td><td>Sp, Narr</td><td>74</td><td>Yes</td><td /><td /><td>X</td><td /><td>25</td><td>−0 56</td></tr><tr><td><xref ref-type="bibr" rid="bibr67">Overby (2008)</xref></td><td>(2)</td><td>NR-Sp</td><td>30</td><td>Yes</td><td /><td /><td>X</td><td /><td>100</td><td>−0.73</td></tr><tr><td><xref ref-type="bibr" rid="bibr71">Puranik & Lonigan (2012)</xref>*</td><td>4–5</td><td>Emergent writing</td><td>219</td><td>Yes</td><td>X</td><td>X</td><td>X</td><td /><td>75</td><td>−1.20</td></tr><tr><td><xref ref-type="bibr" rid="bibr70">Puranik et al. (2014)</xref></td><td>5–6</td><td>NR-Sp, Informative</td><td>219</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>13</td><td>−0.50</td></tr><tr><td><xref ref-type="bibr" rid="bibr73">Reilly et al. (2014)</xref></td><td>7–16</td><td>Narr</td><td>134</td><td>Yes</td><td /><td>X</td><td /><td /><td>75</td><td>−0.50</td></tr><tr><td><xref ref-type="bibr" rid="bibr74">Romaglia (2010)</xref></td><td>(4–5)</td><td>Narr, Exp</td><td>16</td><td>Yes</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn2">a</xref></td><td>50</td><td>−0.93</td></tr><tr><td><xref ref-type="bibr" rid="bibr76">Scott & Windsor (2000)</xref>*/<xref ref-type="bibr" rid="bibr93">Windsor et al. (2000)</xref>*</td><td>7–12</td><td>Narr, Exp</td><td>55</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>25/25</td><td>−0.98</td></tr><tr><td><xref ref-type="bibr" rid="bibr77">Scuccimarra et al. (2008)</xref>*</td><td>(2)</td><td>Sp, HW tests</td><td>45</td><td>No</td><td>X</td><td>X</td><td>X</td><td /><td>50</td><td>−1.88</td></tr><tr><td><xref ref-type="bibr" rid="bibr84">Tattersall (2010)</xref></td><td>6–18</td><td>NR-Sp</td><td>111</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>100</td><td>−0.97</td></tr><tr><td><xref ref-type="bibr" rid="bibr90">Vandewalle et al. (2012)</xref>*</td><td>(2–3)</td><td>Spelling</td><td>32</td><td>No</td><td>X</td><td>X</td><td>X</td><td /><td>50</td><td>−1.40</td></tr><tr><td><xref ref-type="bibr" rid="bibr91">Walz et al. (2000)</xref></td><td>(5)</td><td>State test</td><td>53,242</td><td>Yes</td><td /><td /><td /><td>L<xref ref-type="table-fn" rid="tfn2">a</xref></td><td>100</td><td>−0.78</td></tr><tr><td><xref ref-type="bibr" rid="bibr92">Williams et al. (2013)</xref>*/<xref ref-type="bibr" rid="bibr47">Larkin et al. (2013)</xref>*</td><td>6–10</td><td>Informative, NR-Sp, NR-Grammar</td><td>30</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>75/75</td><td>−1.02</td></tr><tr><td><xref ref-type="bibr" rid="bibr94">Wolter et al. (2011)</xref>*</td><td>10</td><td>NR-Sp</td><td>37</td><td>Yes</td><td>X</td><td>X</td><td /><td /><td>50</td><td>−1.10</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note</emph>. *Peer-reviewed article. SLI = specific language impairment; CA = chronological age; Grd. = grade; <emph>N</emph> = total participants; Eng = English; RLP = receptive language problem; ELP = expressive language problem; SP = speech problem; <emph>g</emph> = effect size; NR-Sp = norm-referenced spelling test; Narr = narrative; Exp = expository; NR-Narr = norm-referenced measure of narrative writing; HW = handwriting; NR-Grammar = norm-referenced test of grammar.</p> <ulist> <item>2 Language tested by school personnel but specific language problems not specified.</item> <item>3 Language learning impairment (specific information about aspects of language impaired not provided).</item> <item>4 Idiopathic (functional) language disorder.</item> </ulist> <p>We coded eight quality indicators for each study. Four of these quality indicators were viewed as critical, as problems with any of the four could undermine the confidence placed in ESs calculated for a study. We refer to these as "primary" quality indicators and they included whether overall attrition was acceptable (less than 10% for all children), unequal attrition within groups was not evident (group attrition between children with SLI, CA, and LA did not differ by more than 5%), reliability of writing measures established (.70 or greater for each measure), and ceiling or floor effects for writing measures were not evident (mean score on a measure for each group was more than one-standard deviation from the lowest and highest score possible for each measure). For each study, we calculated the percentage of "primary" quality indicators met.</p> <p>We coded an additional four quality indicators focusing on analyses and reporting of data. This included whether confidence intervals (CIs) were reported, power analysis was conducted, statistical assumptions were tested and met, and multivariate analyses were applied when there was more than one measure. Although these quality indicators are important, failure to meet them were less likely to undermine the confidence that could be placed in our outcomes, as these quality indicators would not directly influence the calculation of ESs. For all eight quality indicators, the indicator was coded as "not met" if the authors did not provide the information needed to code it.</p> <p>For each writing outcome, we recorded the data needed to calculate an ES (sample size, means, and standard deviations for SLI, CA, and LA groups). If any of this data were missing, other data for calculating ESs were recorded (e.g., <emph>F</emph> and <emph>MSE</emph> statistics for ANOVA).</p> <p>To establish reliability of coding, all papers were independently coded for characteristics, quality indicators, and data by two trained GAs. After a study was coded by both scorers, any differences were resolved through a discussion led by the first author. The mean agreement across both coders across all variables was 94.5%. The percent of agreement for each characteristic, quality indicator, and data ranged from 89.9% to 99.8%.</p> <hd id="AN0143828403-14">Calculation of ESs</hd> <p>We computed ESs for each writing outcome as the difference between the writing performance of children classified with SLI and CA (or children classified with SLI and LA) divided by the pooled standard deviation for the two groups. Using this approach, a negative ES indicated children classified with SLI evidenced a lower writing score than CA or LA children. <emph>Comprehensive Meta-analysis</emph> software–3 (CMA-3; [<reflink idref="bib9" id="ref53">9</reflink>]) was used to calculate ESs using Hedge's <emph>g</emph>. We applied the Effect Size Analysis Program ([<reflink idref="bib78" id="ref54">78</reflink>]), which computes an odds ratio which is converted into <emph>d</emph> (that we then converted into <emph>g</emph>).</p> <p>Calculated ESs and sample sizes were examined to identify outliers. This was done separately for SLI versus CA and SLI versus LA. We applied Tukey's ([<reflink idref="bib87" id="ref55">87</reflink>]) definition of an extreme observation to identify outliers (any ES or sample size falling three interquartile ranges [75th–25th percentile] above or below the mean). Identified outliers were winsorized so they did not exceed these maximum or minimum values. For SLI versus CA, the following were winsorized: (a) the sample size for [<reflink idref="bib91" id="ref56">91</reflink>] as it involved 53,242 students (196 children classified with SLI); (b) ESs for whole word omission errors ([<reflink idref="bib53" id="ref57">53</reflink>]), morpheme omission errors involving "ed" ([<reflink idref="bib54" id="ref58">54</reflink>]), spelling errors ([<reflink idref="bib71" id="ref59">71</reflink>]), <emph>t</emph>-units per minute for both expository and narrative writing ([<reflink idref="bib76" id="ref60">76</reflink>]), nonword spelling errors ([<reflink idref="bib77" id="ref61">77</reflink>]) were reduced to −3.08; and (c) the ES for <emph>t</emph>-unit length was reduced to 1.18 in [<reflink idref="bib7" id="ref62">7</reflink>]. For the SLI and LA comparison, ESs were reduced to −2.08 for spelling (Dockrell & Connelly, 2013) and syntax omission errors (Mackie & Dockrell, 2004).</p> <hd id="AN0143828403-15">Statistical Analyses</hd> <p></p> <hd id="AN0143828403-16">Main effects</hd> <p>We applied RVE to analyze ESs for each research question. This method can be applied no matter the source of dependence among ESs ([<reflink idref="bib75" id="ref63">75</reflink>]; [<reflink idref="bib83" id="ref64">83</reflink>]), allowing the meta-analyst to include multiple individual ESs from a single study, without the need to combine multiple treatment or control groups, combine multiple groups within a treatment, or average ESs within studies (e.g., across groups or measures). Average weighted ES calculations and meta-regression moderator analyses (see below) were conducted in Stata/SE 15.1 using the robumeta package with small sample adjustments. We reported CIs for each average-weighted ES to provide an estimate of the accuracy of the ESs. Next, we estimated and included Tau<sups>2</sups>, which is dependent on the scale of the ES, to provide a measure of the extent of the variation in effects observed between studies ([<reflink idref="bib83" id="ref65">83</reflink>]). We then used the Tau<sups>2</sups> statistic to calculate the 95% prediction interval (PI) to provide an estimate of the dispersion of ESs ([<reflink idref="bib10" id="ref66">10</reflink>]). Finally, we calculated <emph>I</emph><sups>2</sups> to describe the ratio of true heterogeneity to total variance using the robumeta package in R version 3.5.3 (see [<reflink idref="bib31" id="ref67">31</reflink>]), as <emph>I</emph><sups>2</sups> is not provided in the Stata/IC 15.1 package.</p> <p>Because many studies included different measures of writing from the same writing sample (e.g., grammar, output), the RVE approach applied was the "correlated effects case" ([<reflink idref="bib83" id="ref68">83</reflink>]). We specified a random effects model, as it is possible that the distribution of ESs varied from study to study based on children's age, languages, or other characteristics (see Introduction). We therefore specified a common correlation for the within-study ESs of ρ of.80, and conducted sensitivity analyses for different values of ρ.</p> <p>RQ1 only included ESs for the SLI versus CA comparisons, whereas RQ2 involved just ESs from the SLI versus LA comparisons. In addition to the overall average weighted ES for the SLI versus CA comparison, we also employed a shifting unit of analysis approach (RQ3) to examine differences between these two groups on writing quality, output, grammar, vocabulary, and spelling. Because the analysis for RQ3 employed RVE we did not report results for outcomes where there were fewer than four degrees of freedom (e.g., handwriting, elements), as Type I error rates increase and the results of such models cannot be trusted ([<reflink idref="bib31" id="ref69">31</reflink>]). Using these criteria, it was not possible to employ a shifting unit of analysis approach for SLI versus LA.</p> <hd id="AN0143828403-17">Metaregression moderator analysis</hd> <p>Metaregression with RVE was used to examine the potential moderating effects of study characteristics (RQ4). This was only done with the ESs included in RQ1 (SLI vs. CA). There were not enough studies to make such an analysis feasible for SLI versus LA (RQ2), as there were only six SLI versus LA studies. We also elected not to conduct separate moderator analyses for the five writing constructs examined in RQ3 (these ESs involved SLI vs. CA), as the number of moderators included in each model would be more limited and less informative than the moderator analysis conducted for RQ1.</p> <p>For our moderator analysis involving all effects for SLI versus CA, we examined these moderators: age, test type (norm-referenced vs. researcher-created), test format (written vs. contrived response), study quality (the four primary quality indicators), peer-review, and language disorder (speech vs. other language problems). Moderators were dummy coded or centered to estimate the within study effects and improve model interpretability.</p> <hd id="AN0143828403-18">Publication bias</hd> <p>We tested for potential publication bias by examining graphical representations of funnel plots, conducting Egger's test, and conducting Begg's rank correlation test in Stata 15.1. We examined the graphical representation of two funnel plots because (a) many publication bias analyses were developed based on assumptions of a single ES per study and (b) examination of funnel plots with multiple ESs per study may influence publication bias analyses. The first funnel plot included only one ES per study (averaged across measures and groups within the study), and another included all ESs in a study. To aggregate effects in a study, we used techniques from [<reflink idref="bib19" id="ref70">19</reflink>] to account for within- and between-group variance. The funnel plot for RQ2 included all ESs from all studies; we did not include a second funnel plot for RQ2 involving an average ES, as there were only six studies. We also conducted a meta-regression analysis using RVE to examine whether the ESs from the included studies were moderated whether they were published or unpublished (we did this only for the main effect for RQ1, due to the small number of studies included for other research questions). This provided data relevant to publication bias.</p> <hd id="AN0143828403-19">Results</hd> <p></p> <hd id="AN0143828403-20">Description of Studies</hd> <p>Thirty-nine studies published in 47 publications compared children classified with SLI and CA peers on one or more writing outcomes (see Table 1). Six of these studies (in seven publications) also compared children classified with SLI and LA (see Table 2). The 39 SLI versus CA comparisons included 56,945 children (53,242 of them were from a single study; [<reflink idref="bib91" id="ref71">91</reflink>]). The six studies comparing SLI versus LA involved 286 children.</p> <p>Graph</p> <p>Table 2. Study Characteristic and Average Effect Sizes for Writing Variables for SLI and LA Comparisons.</p> <p> <ephtml> <table><colgroup><col align="left" /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /></colgroup><thead><tr><th align="center">Study</th><th align="center">Age (Grd.)</th><th align="center">Writing tasks</th><th align="center"><italic>N</italic></th><th align="center">English</th><th align="center">RLP</th><th align="center">ELP</th><th align="center">SP</th><th align="center">Study quality (%)</th><th align="center">Average <italic>g</italic></th></tr></thead><tbody><tr><td><xref ref-type="bibr" rid="bibr16">Connelly et al. (2012)</xref>*</td><td>7–11</td><td>NR-Sp, Personal Narr, HW</td><td>66</td><td>Yes</td><td>X</td><td>X</td><td align="center">—</td><td>92</td><td>−0.38</td></tr><tr><td><xref ref-type="bibr" rid="bibr24">Dockrell & Connelly (2013)</xref>*</td><td>7–10</td><td>NR-Sp, Informative</td><td>46</td><td>Yes</td><td>X</td><td>X</td><td align="center">—</td><td>100</td><td>−0.31</td></tr><tr><td><xref ref-type="bibr" rid="bibr53">Mackie & Dockrell (2004)</xref>*</td><td>6–12</td><td>Narr</td><td>22</td><td>Yes</td><td>X</td><td>X</td><td align="center">—</td><td>75</td><td>−0.96</td></tr><tr><td><xref ref-type="bibr" rid="bibr54">Mackie et al. (2013)</xref></td><td>7–11</td><td>Narr</td><td>92</td><td>Yes</td><td>X</td><td>X</td><td align="center">—</td><td>75</td><td>−0.55</td></tr><tr><td><xref ref-type="bibr" rid="bibr59">McFadden & Gillam (1996)</xref>*</td><td>7–11</td><td>Narr</td><td>20</td><td>Yes</td><td /><td>X</td><td align="center">—</td><td>100</td><td>−0.32</td></tr><tr><td><xref ref-type="bibr" rid="bibr76">Scott & Windsor (2000)</xref>*/<xref ref-type="bibr" rid="bibr93">Windsor et al. (2000)</xref>*</td><td>7–12</td><td>Narr, Exp</td><td>40</td><td>Yes</td><td>X</td><td>X</td><td align="center">—</td><td>25</td><td>−0.35</td></tr></tbody></table> </ephtml> </p> <p>5 <emph>Note.</emph> SLI = specific language impairment; LA = language ability matched; <emph>N</emph> = total participants; RLP = receptive language problem; ELP = expressive language problem; SP = speech problem; Sp = spelling; <emph>g</emph> = effect size; NR-Sp = norm-referenced spelling test; Narr = narrative; Exp = expository.</p> <p>Of the 47 publications, 35 were journal articles, whereas the remaining 12 publications (see Table 1) included six dissertations, three chapters, one book, one report, and one thesis. Five studies included two papers (see Tables 1 and 2), where the focus of analyses involved different measures ([<reflink idref="bib76" id="ref72">76</reflink>]; [<reflink idref="bib93" id="ref73">93</reflink>]) or a subset of the full sample ([<reflink idref="bib29" id="ref74">29</reflink>]; [<reflink idref="bib56" id="ref75">56</reflink>]). Another study included three papers ([<reflink idref="bib55" id="ref76">55</reflink>]; [<reflink idref="bib63" id="ref77">63</reflink>]; [<reflink idref="bib65" id="ref78">65</reflink>]) in which the same children were followed over time.</p> <p>In the studies that were included in this review, multiple labels were used to refer to children who met our definition of SLI. Most commonly, children were referred to as having SLI (<emph>k</emph> = 13), followed by language impaired (<emph>k</emph> = 11), speech difficulties (<emph>k</emph> = 5; including difficulties with articulation, phonology, or speech), language learning disability (<emph>k</emph> = 4), speech and language difficulties (<emph>k</emph> = 3), oral language difficulty (<emph>k</emph> = 2), and expressive phonology impairment (<emph>k</emph> = 1).</p> <p>Children classified with SLI in the studies reviewed here mostly came from schools (<emph>k =</emph> 28), but in some studies they were recruited via hospitals or clinics (<emph>k</emph> = 4), were part of previous studies (<emph>k</emph> = 3), or located through professional nominations (<emph>k</emph> = 1) and advertisements (<emph>k</emph> = 1). Two studies did not provide information on how these children were located (i.e., [<reflink idref="bib2" id="ref79">2</reflink>]; [<reflink idref="bib41" id="ref80">41</reflink>]). In most of the studies, CA and LA children were recruited through the same avenue as children classified with SLI. Exceptions to this generalization involved studies where children classified with SLI were recruited through hospitals or nominations/advertisements, but CAs were drawn from schools.</p> <p>SLI was established by poor performance on one or more language tests (<emph>k</emph> = 34). Five studies (e.g., [<reflink idref="bib7" id="ref81">7</reflink>]) established such difficulties with a discrepancy model where children's language scores were one or more standard deviations below their scores on an intelligence test. In 28 studies (72%), researchers established the IQ of children classified with SLI was in the normal range (25 studies used a measure of nonverbal intelligence). When identifying children with SLI, 18 studies (46%) specifically indicated that they excluded children with neurological problems (<emph>k</emph> = 8), hearing difficulties (<emph>k</emph> = 8), sensory difficulties (<emph>k</emph> = 6), ASD (<emph>k</emph> = 4), intellectual disabilities (<emph>k</emph> = 3), motor difficulties (<emph>k</emph> = 2), medical problems (<emph>k</emph> = 1), and social/emotional problems (<emph>k</emph> = 1). Children classified with SLI were not identified as having acquired brain injury, deafness, blindness, ASD, or intellectual disabilities in any study reviewed here. In 10 studies, one or more children in the SLI sample had a comorbid condition, which included attention deficit hyperactivity disorder (ADHD; [<reflink idref="bib2" id="ref82">2</reflink>]) and reading difficulties (e.g., [<reflink idref="bib28" id="ref83">28</reflink>]).</p> <p>Twenty-three of the 39 studies (60%) included children who experienced both expressive and receptive language problems (in seven of these studies some children also had speech difficulties; see Tables 1 and 2). In two of seven studies that included children with expressive and receptive language problems as well as speech problems, it was possible to calculate separate ESs for speech problems as well as language problems ([<reflink idref="bib7" id="ref84">7</reflink>]; [<reflink idref="bib70" id="ref85">70</reflink>]). Five studies involved children with just speech problems (e.g., [<reflink idref="bib61" id="ref86">61</reflink>]). Four studies included only children with expressive language problems (e.g., [<reflink idref="bib12" id="ref87">12</reflink>]), one study described children classified with SLI as having functional language difficulties ([<reflink idref="bib21" id="ref88">21</reflink>]), and another study classified them as having language learning difficulties ([<reflink idref="bib55" id="ref89">55</reflink>]). The type of language problem experienced by children classified with SLI was not identified in five studies (e.g., [<reflink idref="bib40" id="ref90">40</reflink>]); in these investigations testing was done by school personnel as part of SLI services.</p> <p>In all studies, CA and LA peers did not have a history of language difficulties. In 24 studies (62%), children classified with SLI and their CA or LA peers were from the same class or school. In addition to matching for CA or LA, 20 studies (51%) matched on IQ (<emph>k</emph> =13), gender (<emph>k</emph> =8), SES (<emph>k</emph> =6), hearing (<emph>k</emph> =5), vision (<emph>k</emph> =1), and ethnicity (<emph>k</emph> =1).</p> <p>Thirty-two of the 39 studies (82%) provided data on gender. In studies where gender was reported, 61% and 51% of children classified with SLI and their CAs peers were boys, respectively. Nine of these 32 studies did not adequately control for gender differences, with all but one of these studies (i.e., [<reflink idref="bib47" id="ref91">47</reflink>]) having proportionally more girls in the CA group than the SLI group (e.g., [<reflink idref="bib11" id="ref92">11</reflink>]). In the six SLI versus LA comparison studies, 68% of children in both groups were boys. There were an equal number of boys and girls in the SLI and LA groups in these studies.</p> <p>In the 12 studies (31%) that provided information on ethnicity, 50% or more of children were White in nine studies (e.g., [<reflink idref="bib48" id="ref93">48</reflink>]). The majority of children were Black in two studies (i.e., [<reflink idref="bib43" id="ref94">43</reflink>]; [<reflink idref="bib70" id="ref95">70</reflink>]). In one study (i.e., [<reflink idref="bib2" id="ref96">2</reflink>]), more than 50% of the children were Black and Hispanic. Only one SLI versus LA study provided information on ethnicity (children were mostly White).</p> <p>The most common aspect of writing addressed in the 39 SLI versus CA studies was spelling (<emph>k</emph> = 32), followed by output (<emph>k</emph> = 16), grammar (<emph>k</emph> = 15), writing quality (<emph>k</emph> = 14), vocabulary (<emph>k</emph> = 10), and handwriting (<emph>k</emph> = 6). Fifteen of these studies only assessed spelling (see Table 1). For SLI versus LA studies, spelling was assessed in five studies, whereas writing quality, output, and grammar were assessed in four studies. Vocabulary was assessed in three studies (see Table 2).</p> <p>Thirty-one of the 39 studies were conducted in English-speaking countries (20 in the United States, 10 in Great Britain, and one in Australia). Four studies were from the Netherlands (e.g., [<reflink idref="bib77" id="ref97">77</reflink>]), whereas one study was from each of the following countries: France (i.e., [<reflink idref="bib12" id="ref98">12</reflink>]; [<reflink idref="bib27" id="ref99">27</reflink>]), Italy (i.e., [<reflink idref="bib11" id="ref100">11</reflink>]), Israel (i.e., [<reflink idref="bib21" id="ref101">21</reflink>]), and Sweden (i.e., [<reflink idref="bib55" id="ref102">55</reflink>]).</p> <hd id="AN0143828403-21">Quality of Studies</hd> <p>We examined eight quality indicators for each study. For our four primary quality indicators, reliability of measures was established in 79% of studies. Attrition overall and attrition by group were not a problem in 74% of studies, whereas ceiling or floor problems were not evident in 67% of studies. When the four primary quality indicators were collectively considered, 36% of studies met all four criteria, 33% of studies met three criteria, 18% of studies met two criteria, and 13% of studies one criterion.</p> <p>For quality indicators that involved reporting data and statistical analyses, 31% of studies conducted multivariate analyses when there was more than one measure, 8% of studies indicated whether statistical assumptions were met, 5% of studies conducted a power analysis, and no studies reported CIs.</p> <hd id="AN0143828403-22">RQ1: Are There Differences in the Writing of Children Classified With SLI and CA Peers?</hd> <p>For RQ1, we calculated 385 ESs from 39 studies. The forest plot as well as the ES and CI for each measure are available from the authors. Table 1 provides an average ES for each study.</p> <p>The average weighted <emph>ES</emph> for the distribution of all ESs for SLI versus CA studies was −0.97 [−1.09, −0.86], indicating that children classified with SLI scored statistically significantly lower than CA peers by almost a full standard deviation (95% of ESs favored CA peers). The Tau<sups>2</sups> of.21 indicated there was some variability in ESs, resulting in 95% PI that ranged from −1.09 to −0.03, which indicates that we should expect ESs from a new study to end up in this range 95% of the time. The <emph>I</emph><sups>2</sups> for the model signified that 73% of the variance in the model was due to between-study factors. Sensitivity analyses indicated no differences in the estimates for different values of ρ, and only a negligible difference of.0012 between the estimates of the between-studies variance component (Tau<sups>2</sups>) for within study correlations ranging from ρ =.01 to ρ =.99.</p> <hd id="AN0143828403-23">RQ2: Are There Differences in the Writing of Children Classified With SLI and LA Peers?</hd> <p>For RQ2, we calculated 90 ESs from six studies. The forest plot as well as the ES and CI for each measure are available from the authors. Table 2 provides an average ES for each study.</p> <p>The average weighted ES for the distribution of all for SLI versus LA studies was −0.47 (95% CI [−0.71, −0.23]), indicating that children classified with SLI scored statistically significantly lower than LA peers by about a half of a standard deviation (83% of ESs favored LA peers). The Tau<sups>2</sups> of.38 indicated that there was variance in the distribution of ESs, and <emph>I</emph><sups>2</sups> for the model signified that 67% of the variance was due to between-study factors. However, we did not calculate a PI for this meta-analysis, due to the small number of studies and unreliability in the prediction ([<reflink idref="bib10" id="ref103">10</reflink>]). Similar to the analysis for the first research question, sensitivity analyses for different levels of ρ indicated negligible differences in the estimates of the model parameters that did not change the findings.</p> <hd id="AN0143828403-24">RQ3: What Specific Features of Writing Are Problematic for Children Classified With SLI?</hd> <p>For RQ3, we applied a shifting units of analysis approach to examine the differences between children classified with SLI and CA peers for five writing constructs. Table 3 includes the average weighted ES for each writing construct, as well as CIs and PIs. Children classified with SLI had statistically significantly lower scores than CA peers for each writing construct, as the CI for the average weighted ES did not cross zero for any of them. Obtained differences between the writing of children classified with SLI and CA peers were substantial: writing quality (<emph>g</emph> = −0.92), output (<emph>g</emph> = −1.00), spelling (<emph>g</emph> = −1.14), grammar (<emph>g</emph> = −0.68), and vocabulary (<emph>g</emph> = −0.68). There was considerable variability between ESs, though, as <emph>I</emph><sups>2</sups> values indicated about 60% to 90% of the variance for these writing constructs could be attributed to between-study variance.</p> <p>Graph</p> <p>Table 3. Average Weighted Effect Sizes for Writing Variables for SLI versus CA Comparison.</p> <p> <ephtml> <table><colgroup><col align="left" /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /></colgroup><thead><tr><th align="center">Writing variables</th><th align="center"><italic>k</italic></th><th align="center"><italic>N</italic></th><th align="center"><italic>g</italic></th><th align="center">95% CI</th><th align="center"><italic>p</italic> value</th><th align="center"><italic>I</italic><sup>2</sup></th><th align="center">Tau<sup>2</sup></th><th align="center">95% PI</th></tr></thead><tbody><tr><td>Writing quality</td><td>13</td><td>24</td><td>−0.92</td><td>[−1.26, −0.59]</td><td><.001</td><td>73.88</td><td>.197</td><td>[−1.96, 0.12]</td></tr><tr><td>Output</td><td>15</td><td>44</td><td>−1.00</td><td>[−1.45, −0.55]</td><td><.001</td><td>89.79</td><td>.644</td><td>[−2.80, 0.80]</td></tr><tr><td>Grammar</td><td>17</td><td>143</td><td>−0.68</td><td>[−0.84, −0.52]</td><td><.001</td><td>60.74</td><td>.143</td><td>[−3.23, 1.87]</td></tr><tr><td>Vocabulary</td><td>11</td><td>30</td><td>−0.68</td><td>[−1.04, −0.32]</td><td>.002</td><td>88.13</td><td>.562</td><td>[−2.43, 1.07]</td></tr><tr><td>Spelling</td><td>30</td><td>104</td><td>−1.17</td><td>[−1.33, −1.01]</td><td><.001</td><td>74.39</td><td>.237</td><td>[−2.18, −0.16]</td></tr></tbody></table> </ephtml> </p> <p>6 <emph>Note.</emph> SLI = specific language impairment; CA = chronological age; <emph>k = n</emph>umber of studies, <emph>n</emph> = number of effect sizes; <emph>g</emph> = effect size; CI = confidence interval; PI = predication interval.</p> <hd id="AN0143828403-25">RQ4: Do Age, Test Features, Study Quality, and Disorder Moderate Effects?</hd> <p>A moderator analysis was conducted to examine covariates of SLI versus CA ESs. We dummy coded and centered six <emph>a priori</emph> selected moderators as follows: <emph>age</emph>: average age of the participants for each study, centered at earliest average age (4.0) for studies in the sample; <emph>test type</emph>: dummy coded to compare norm-referenced tests (0) with researcher-created measures (<reflink idref="bib1" id="ref104">1</reflink>); <emph>test format</emph>: dummy coded to compare contrived response outcomes (0) with written production outcomes (<reflink idref="bib1" id="ref105">1</reflink>); <emph>study quality</emph>: mean centered continuous variable representing the proportion of the four primary study quality indicators met; <emph>language disorder</emph>: dummy coded to compare studies that involved children classified with SLI who just had problems with speech (0) with studies involving children classified with SLI who had problem with receptive language/expressive language (<reflink idref="bib1" id="ref106">1</reflink>).</p> <p>Results of the meta-regression are presented in Table 4. The moderators of <emph>test type, test format, and language disorder</emph> (just speech disorder vs. language disorder) were statistically significant contributors to the model with alpha set to.05. The intercept for the model (−1.44) can be interpreted to represent the average weighted ES difference between 4-year-old English-speaking children with any form of SLI (not just speech) and their CA peers on norm-referenced, contrived response outcomes in studies of average quality.</p> <p>Graph</p> <p>Table 4. Meta-Regression Using Robust Variance Estimation and Small Sample Corrections.</p> <p> <ephtml> <table><colgroup><col align="left" /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /><col align="char" char="." /></colgroup><thead><tr><th align="center">Variable</th><th align="center"><italic>B</italic></th><th align="center"><italic>SE</italic></th><th align="center"><italic>Df</italic></th><th align="center"><italic>p</italic> value</th><th align="center">95% CI</th></tr></thead><tbody><tr><td>Intercept</td><td>−1.44</td><td>0.15</td><td>8.66</td><td><.001</td><td>[−1.77, −1.11]</td></tr><tr><td>Age (centered: 0–4 years old)</td><td>0.01</td><td>0.01</td><td>10.08</td><td>.362</td><td>[−0.01, 0.04]</td></tr><tr><td>Test type (0 = normed, 1 = RC)</td><td>0.24</td><td>0.10</td><td>20.16</td><td>.022</td><td>[0.04, 0.44]</td></tr><tr><td>Test format (0 = contrived, 1 = written)</td><td>0.34</td><td>0.08</td><td>23.96</td><td><.001</td><td>[0.17, 0.51]</td></tr><tr><td>Study quality (mean centered)</td><td>0.00</td><td>0.00</td><td>15.84</td><td>.320</td><td>[−0.00, 0.01]</td></tr><tr><td>Language (0 = English, 1 = other)</td><td>0.04</td><td>0.13</td><td>12.42</td><td>.781</td><td>[−0.25, 0.32]</td></tr><tr><td>Just Speech (0 = not just speech, 1 = only speech)</td><td>0.39</td><td>0.12</td><td>9.83</td><td>.008</td><td>[0.13, 0.65]</td></tr></tbody></table> </ephtml> </p> <p>7 <emph>Note.</emph> The model is based on <emph>k</emph> = 39 studies and <emph>n = 385 effect sizes.</emph> Tau<sups>2</sups> =.172; CI = confidence interval; RC = researcher-designed.</p> <p>For the moderator <emph>test format</emph>, the model indicated a statistically significant expected average weighted ES reduction in the difference in writing scores between SLI and CA of 0.34 for tests where scores were derived from children's writing. There was also a statistically significant reduction in the average weighted ES of 0.24 when researcher-created measures (i.e., <emph>test type</emph>) were used. A statistically significant reduction of 0.39 was evident when children classified with SLI only experienced a speech disorder. In other words, the writing differences between SLI and CA were not as large when assessment involved writing, researcher-created measures, or when SLI included just children with a speech disorder.</p> <hd id="AN0143828403-26">Publication Bias</hd> <p>For RQ1, there were only eight (21%) unpublished reports with unique ESs. Therefore, despite our extensive search, it is possible that some unpublished reports with small study effects were not located. Separate funnel plots (see Figure 1) were created for (a) all 385 ESs extracted from the studies and (b) for the average ES for each study (39 ESs) using the metafunnel command in Stata 15.1. Both funnel plots appeared to be somewhat asymmetrical. Visual analysis indicated the possibility of missing ESs on the lower right-hand side of both funnel plots. Thus, small studies yielding small negative effects may be missing, which if located would slightly lower the overall large negative for RQ1. Results of the Egger's test using all of the ESs from each study also indicated presence of publication bias (<emph>p</emph> <.001), as did the Begg's rank correlation test (<emph>p</emph> <.001). Therefore, we conducted trim and fill analysis using the "metatrim" command with "flip" option due to the negative overall ES in Stata 15.1. This resulted in nine trimmed and filled ESs, with an adjusted estimated average weighted <emph>ES</emph> of −0.62 [−0.68, −0.56]. These results should be interpreted cautiously, as the trim and fill analysis does not use RVE to account for dependencies among <emph>ESs</emph>, yet the results seem to indicate that the presence of publication bias did not impact overall interpretations of the SLI versus CA comparisons. Finally, we conducted a moderator analysis using meta-regression for RQ1 to examine whether the results differed across published and unpublished studies included in the current meta-analysis. The results indicated publication status was not a significant predictor in the model (<emph>p</emph> =.51). In other words, there was not a statistically significant difference in the average weighted ESs obtained from 31 published and eight unpublished studies identified and included in this analysis.</p> <p>Graph: Figure 1. Funnel plots for Research Question 1, with one average ES per study (left) and all of the ESs for each study (right). Note. ES = effect size.</p> <p>The funnel plot for RQ2 (90 ESs for six published studies; see Figure 2) did not evidence asymmetry, and Egger's test was nonsignificant (<emph>p</emph> =.364). However, Begg's rank correlation test showed some correlation between the ESs and their standard errors (<emph>p</emph> =.027) indicating some potential small study bias. Therefore, we conducted a trim and fill analysis for RQ2, which results in a random effects ES of −0.42 [−0.53, −0.32], which is slightly smaller than the average ES of −0.47 [−0.71, −0.23] found for the distribution of ESs for the random effects RVE model, but again did change the interpretations of the SLI versus LA comparison.</p> <p>Graph: Figure 2. Funnel plot for Research Question 2, with all ESs from each study. Note. ES = effect size.</p> <hd id="AN0143828403-27">Discussion</hd> <p>This review examined whether children classified with SLI experience difficulties with writing. We conducted a meta-analysis to determine whether the writing of children classified with SLI differed from their CA and LA peers. In the discussion below, we compare the effects for writing obtained in this study with those from a prior meta-analysis comparing the writing of children classified with LD to their CA peers ([<reflink idref="bib36" id="ref107">36</reflink>]). This previous review included 53 LD versus CA writing comparison studies conducted with children in Grades 1 to 12. Although the studies in this prior review involved a broader range of writing outcomes, it did present average weighted ESs for the primary writing outcomes in the current meta-analysis (writing quality, output, spelling, grammar, and vocabulary). It also calculated a median averaged weighted ES across all writing constructs of −0.87, showing that students classified with LD were poorer writers than CA peers. Although the Graham et al. meta-analysis is not an exact match to the current review, it provides a reasonable benchmark for situating the findings from the current meta-analysis. If the effects obtained in this meta-analysis and previous one focusing on students with LD are similar, then children classified as SLI experience significant problems with writing.</p> <hd id="AN0143828403-28">Children Classified With SLI Experience Substantial Difficulties With Writing</hd> <p>The answer to the basic question driving this review was yes. When considered collectively, the writing scores of children classified with SLI were considerably lower than the writing scores of their CA peers. The average weighted ES for all writing measures in the 39 SLI versus CA studies was −0.97, showing that students classified with SLI scored lower on writing measures than their CA peers. The overall ES of −0.97 is slightly higher than the median average weighted ES of −0.87 for the LD versus CA comparison in [<reflink idref="bib36" id="ref108">36</reflink>]. Although the current review included preschool as well as school-aged children (Graham et al. included only school-aged children) and calculated the overall effect differently (we used RVE to obtain an overall effect whereas Graham et al. calculated a median overall effect across writing constructs), these outcomes from the two meta-analyses are quite similar, indicating that the writing problems of children classified with SLI are substantial.</p> <p>This conclusion is further strengthened when we examine the average weighted ESs obtained for specific writing constructs in the SLI versus CA comparisons. Children with SLI scored lower than CA peers on overall writing quality (<emph>g</emph> = −0.92), output (<emph>g</emph> = −1.00), spelling (<emph>g</emph> = −1.17), grammar (<emph>g</emph> = −0.68), and vocabulary (<emph>g</emph> = −0.68). The ES for spelling performance in this meta-analysis and the earlier one conducted by [<reflink idref="bib42" id="ref109">42</reflink>] were similarly large (<emph>g =</emph> −1.42 in the prior review). In addition, the effects obtained in our meta-analysis are generally comparable to those obtained by [<reflink idref="bib36" id="ref110">36</reflink>] comparing students classified as LD to CA peers: writing quality (<emph>g</emph> = −1.06), output (<emph>g</emph> = −0.87), spelling (<emph>g</emph> = −1.50), grammar (<emph>g</emph> = −0.56), and vocabulary (<emph>g</emph> = −0.89). Like students with LD, the writing problems of children classified with SLI are serious and multi-faceted.</p> <p>Even when language was controlled, by comparing children classified with SLI versus LA peers, difficulties with writing were still evident. The average weighted ES across all measures was −0.47. This finding indicates that the writing problems of children classified with SLI are atypical and not just an artifact of their level of language development. Our finding stands in contrast to [<reflink idref="bib42" id="ref111">42</reflink>] who did not find such a difference when looking at just spelling. This discrepancy may be due to multiple differences in how studies were selected in the two reviews, as we focused on writing broadly, applied different inclusion/exclusion criteria, and used different methods for analyzing effects.</p> <p>It is important to note that there was considerable variability in the obtained ESs in the current study (similar to [<reflink idref="bib42" id="ref112">42</reflink>]). We found that features of how writing was assessed and type of language disorder were statistically significant predictors of this heterogeneity. Differences between the writing of children classified with SLI and CA peers were a third of a standard deviation lower when assessments involved writing versus a contrived response format ([<reflink idref="bib42" id="ref113">42</reflink>] did not find such a difference for spelling measures), one fourth of a standard deviation lower when the measure was researcher-created versus norm-referenced, and four tenths of a standard deviation lower when the SLI involved just a speech disorder. Even with these reductions in effects, children classified with SLI evidenced significant problems with writing.</p> <p>Although additional studies are needed to extend and potentially strengthen our collective findings, this meta-analysis provides compelling evidence that students classified with SLI experience writing difficulties. This does not mean that all children classified with SLI have problems with writing. Nor does it mean that students classified with SLI who find writing challenging experience the same profile of writing difficulties. Additional research is needed to determine the prevalence of writing problems among students classified with SLI, and to pinpoint more directly the causes of these writing difficulties.</p> <hd id="AN0143828403-29">Limitations and Caveats</hd> <p>Interpretation of the findings from this review depends on a number of factors. One issue concerns the quality of studies reviewed. More confidence can be placed in findings from studies that are soundly constructed. To make the quality of studies reviewed transparent, each study was assessed for the following eight metrics: overall attrition, differential attrition, reliability of measures, ceiling and floor effects, CIs reported, power analysis conducted, statistical assumptions tested, and multivariate analyses applied (when appropriate).</p> <p>A variety of assessments were used to measure writing performance across and within studies reviewed. Although the use of RVE allowed us to include the effects from these multiple and overlapping measures in our analyses, we were only able to examine whether heterogeneity of effects was related to two aspects of assessments: type of test (researcher-created and norm-referenced) and test format (written and contrived response). It is possible that other features of assessments (e.g., length of assessment) were also related to variability in effects.</p> <p>Some writing constructs were the subject of more investigations than others (e.g., spelling was assessed in 32 investigations, but handwriting was assessed in only six studies), whereas other aspects were not assessed at all, including voice, writing process, motivation, and knowledge about writing. This must be considered when interpreting our findings.</p> <p>The confidence that can be placed in a meta-analysis depends on locating all relevant studies. Although it is impossible to obtain every study ever conducted, we applied two methods to address the issue of possible missing studies. This included using multiple search methods (electronic searches, hand searches of journals, reference searches, and author contacts), and creating funnel plots to examine the distribution of study outcomes by ES and sample size. Analysis of the funnel plots indicated that some studies with small study effects may not have been located. Visual analysis of the funnel plots, however, did not suggest this was enough of a problem to undermine the findings from this review. Furthermore, we did not find a statistically significant difference in ESs for published and unpublished studies.</p> <p>It is also important to note that our search for relevant studies concluded at the end of May 2017. There are likely new studies that were conducted since then, and new ones that will be conducted in the future. Consequently, there is a need for a future meta-analysis to determine whether the findings from this review hold and can be expanded when these new investigations are included.</p> <p>Although we collected studies from countries where students' native language was not English (although a study had to be presented in English to be in this review), we did not conduct a moderator analysis to determine whether variability in effects was related to students' native language. Most studies involved students who spoke English and there were not enough other studies in a single language to make such an analysis feasible.</p> <p>The types of studies reviewed here were correlational, comparing the writing of different types of children. Although the findings from this review provide important and needed information about the writing difficulties experienced by children classified with SLI, they do not establish a cause and effect relationship. It is important to keep this in mind.</p> <hd id="AN0143828403-30">Future Research</hd> <p>Given that we located 39 studies comparing the writing of children classified with SLI and CA peers, with only six investigations comparing children classified with SLI to younger language-matched peers, there is an obvious need for more research in this area. This future research would benefit from following the recommendations below.</p> <p>(a) Investigators need to more precisely define the language difficulties experienced by children classified with SLI in their studies. The language problems of children in most studies reviewed here were so broadly defined that we could not easily determine how specific language difficulties (e.g., receptive language problems) and specific writing problems were related. (b) Future studies need to control better for external variables that are likely to cloud interpretation of findings. More specifically, children classified with SLI and typically developing children need to be matched by gender, race, SES (including parental education), educational experiences (e.g., drawing children from the same classroom), and cognitive capabilities. (c) Researchers need to establish that typically developing children have not experienced difficulty with language. This should include directly testing the language capabilities of all children included in a study. (d) The writing instruction that all children in an investigation are receiving should be described. When such information is coupled with a description of the language interventions received by children classified with SLI, researchers can better contextualize their findings.</p> <p>Although the studies reviewed in this meta-analysis examined multiple aspects of children's written products, the studies were by no means exhaustive in their assessment of writing, as there was little attention directed at organization of writing, ideation, sentence construction, or voice. Just as importantly, not a single study looked at writing motivations, writing process (other than rate), or children's knowledge of writing. Moreover, writing was mostly treated by investigators as if it was a single thing. There is considerable evidence that how well children write in one genre does not necessarily reflect how well they write in a different one ([<reflink idref="bib4" id="ref114">4</reflink>]). Thus, future research in this area needs to look at writing more broadly (assessing a broader range of writing capabilities) and in a more nuanced fashion (examining whether the writing difficulties of children classified with SLI vary by the type of writing assessed).</p> <p>Our examination of study quality in the research reviewed here made it clear that future investigations need to put into place mechanisms for eliminating/minimizing possible problems involving attrition, measurement (reliability, ceiling, and floor effects), power, and statistical analysis (applying multivariate procedures, testing statistical assumptions). Recommendations for reporting data in all studies include providing mean values, standard deviations, and CIs for all groups by age, gender, race, SES, and so forth (in text or online).</p> <p>To move research in this area forward, investigators need to specify the theoretical rationale(s) underpinning their study. This was not done consistently in the studies we examined. New studies applying different methodologies are also needed. For example, more longitudinal studies comparing children classified with SLI to CA and LA peers would provide a stronger basis for determining how language and writing difficulties develop and evolve. Intervention studies designed to improve specific aspects of the language of children classified with SLI should further assess whether there is a corresponding impact on specific aspects of children's writing.</p> <hd id="AN0143828403-31">Implications for Practice</hd> <p>This meta-analysis demonstrated that children classified with SLI are less adept writers than children who do not experience language difficulties. Therefore, it is important they receive the instruction needed for them to become more skilled writers. Because they do not develop writing competence as quickly as their classmates, an obvious recommendation is that extra time be devoted to teaching them to write. It is equally important that the instructional methods used are effective and efficient. This presents a challenge, as there is very little research directed at how to teach writing to children classified with SLI (see [<reflink idref="bib23" id="ref115">23</reflink>]; [<reflink idref="bib57" id="ref116">57</reflink>]; [<reflink idref="bib80" id="ref117">80</reflink>] for notable exceptions). Until such research takes place, it seems prudent to draw on evidence-based practices shown to be effective with other children who experience difficulties with writing (e.g., see [<reflink idref="bib32" id="ref118">32</reflink>]). Instructional methods need to address multiple aspects of writing, as children classified with SLI produce less text than their classmates, and the text they produce is not as well written, uses fewer different words, and contains more spelling and grammar miscues. If they are to receive the education they deserve, this situation must be rectified.</p> <ref id="AN0143828403-32"> <title> References </title> <blist> <bibl id="bib1" idref="ref37" type="bt">1</bibl> <bibtext> Abbott R., Berninger V. (1993). Structural equation modeling of relationships among developmental skills and writing skills in primary- and intermediate-grade writers. Journal of Educational Psychology, 85, 478–508.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref79" type="bt">2</bibl> <bibtext> *Anderson M. A. (2010). Memory and language: Evidence of relationships from three studies of school-age children [Doctoral dissertation]. Western Michigan University.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref28" type="bt">3</bibl> <bibtext> Bates E. (1976). Language and context: The acquisition of pragmatics. Academic Press.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref114" type="bt">4</bibl> <bibtext> Bazerman C., Applebee A. N., Berninger V. W., Brandt D., Graham S., Matsuda P. K., Murphy S., Rowe D. W., Schleppegrell M. (2017). Taking the long view on writing development. Research in the Teaching of English, 51(3), 351–360.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref9" type="bt">5</bibl> <bibtext> Bereiter C., Scardamalia M. (1982). From conversation to composition. In Glaser R. (Ed.), Advances in instructional psychology (Vol. 2, pp. 1–64). Lawrence Erlbaum.</bibtext> </blist> <blist> <bibl id="bib6" type="bt">6</bibl> <bibtext> *Bird J., Bishop D. V., Freeman N. H. (1995). Phonological awareness and literacy development in children with expressive phonological impairments. Journal of Speech, Language, and Hearing Research, 38, 446–462.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref62" type="bt">7</bibl> <bibtext> *Bishop D. V. M., Clarkson B. (2003). Written language as a window into residual language deficits: A study of children with persistent and residual speech and language impairments. Cortex, 39, 215–237.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref4" type="bt">8</bibl> <bibtext> Bishop D. V. M., Snowling M. (2004). Developmental dyslexia and specific language impairment: Same or different? Psychological Bulletin, 130, 858–886.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref53" type="bt">9</bibl> <bibtext> Borenstein M., Hedges L., Higgins J., Rothstein H. (2014). Comprehensive meta-analysis-3. Biostat.</bibtext> </blist> <blist> <bibtext> Borenstein M. (2019). Heterogeneity in Meta-analysis. In Cooper H., Hedges L. V., Valentine J. C. (Eds.), The handbook of research synthesis and meta-analysis (3rd ed., pp. 453–470). Russell Sage Foundation.</bibtext> </blist> <blist> <bibtext> *Brizzolara D., Gasperini F., Pfanner L., Cristofani P., Casalini C., Chilosi A. M. (2011). Long-term reading and spelling outcome in Italian adolescents with a history of specific language impairment. Cortex, 47, 955–973.</bibtext> </blist> <blist> <bibtext> *Broc L., Bernicot J., Olive T., Favart M., Reilly J., Quémart P., Uzé J. (2013). Lexical spelling in children and adolescents with specific language impairment: Variations with the writing situation. Research in Developmental Disabilities, 34, 3253–3266.</bibtext> </blist> <blist> <bibtext> Carlisle J. (1996). An exploratory study of morphological errors in children's written stories. Reading & Writing: An Interdisciplinary Journal, 8, 61–72.</bibtext> </blist> <blist> <bibtext> Catts H. (1991). Early identification of dyslexia: Evidence from a follow-up study of speech-language impaired children. Annals of Dylexia, 41, 163–177.</bibtext> </blist> <blist> <bibtext> Chow J. C., Ekholm E. (2018). Do published studies yield larger effect sizes than unpublished studies in education and special education? A meta-review. Educational Psychology Review, 30, 727–744.</bibtext> </blist> <blist> <bibtext> *Connelly V., Dockrell J., Barnett A. (2012). Children challenged by writing due to language and motor difficulties. In Berninger V. (Ed.), Past, present, and future contributions of cognitive writing research to cognitive psychology (pp. 217–246). Psychological Press.</bibtext> </blist> <blist> <bibtext> *Conti-Ramsden G., Durkin K., Walker A. J. (2012). The messages they send: Email use by adolescents with and without a history of specific language impairment (SLI). International Journal of Language & Communication Disorders, 47, 217–228.</bibtext> </blist> <blist> <bibtext> *Cordewener K. A. H., Bosman A. M. T., Verhoeven L. (2012). Specific language impairment affects the early spelling process quantitatively but not qualitatively. Research in Developmental Disabilities, 33, 1041–1047.</bibtext> </blist> <blist> <bibtext> Cortina J. M., Nouri H. (2000). Effect size for ANOVA designs (Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-129). SAGE.</bibtext> </blist> <blist> <bibtext> *Critten S., Connelly V., Dockrell J. E., Walter K. (2014). Inflectional and derivational morphological spelling abilities of children with specific language impairment. Frontiers in Psychology, 5, Article 948. https://doi.org/10.3389/fpsyg.2014.00948</bibtext> </blist> <blist> <bibtext> *Davidi O., Berman R. (2014). Writing abilities of pre-adolescents with and without language/learning impairment in restructuring an informative text. In Arte B., Dockrell J., Berninger V. (Eds.), Handbook of writing development and instruction in children with language difficulties (pp. 143–157). Oxford University Press.</bibtext> </blist> <blist> <bibtext> *de Bree E., Wijnen F., Gerrits E. (2010). Non-word repetition and literacy in Dutch children at-risk of dyslexia and children with SLI: Results of the follow-up study. Dyslexia, 16, 36–44.</bibtext> </blist> <blist> <bibtext> De La Paz S. (2001). Teaching writing to students with attention deficit disorders and specific language impairment. Journal of Educational Research, 95, 37–47.</bibtext> </blist> <blist> <bibtext> *Dockrell J. E., Connelly V. (2013). The role of oral language in underpinning the text generation difficulties in children with specific language impairment: Text generation in children with specific language impairments. Journal of Research in Reading, 38, 18–34.</bibtext> </blist> <blist> <bibtext> *Durkin K., Conti-Ramsden G., Walker A. J. (2011). Txt lang: Texting, textism use and literacy abilities in adolescents with and without specific language impairment. Journal of Computer Assisted Learning, 27, 49–57.</bibtext> </blist> <blist> <bibtext> Ebert K., Kohnert K. (2011). Sustained attention in children with primary language impairment: A meta-analysis. Journal of Speech, Language, and Hearing Research, 54, 1372–1384.</bibtext> </blist> <blist> <bibtext> *Favart M., Potock A., Broc L., Quermart P., Bernicot J., Olive T. (2016). The management of cohesion in written narratives in students with specific language impairment: Differences between childhood and adolescence. Research in Developmental Disabilities, 59, 318–327.</bibtext> </blist> <blist> <bibtext> *Ferouhi V. G. (2006). Written language difficulties in language-learning disabled fourth and fifth graders: An evaluative study in a New Jersey school district [Unpublished doctoral dissertation]. Columbia University.</bibtext> </blist> <blist> <bibtext> *Fey M. E., Catts H. W., Proctor-Williams K., Tomblin J. B., Zhang X. (2004). Oral and written story composition skills of children with language impairment. Journal of Speech, Language, and Hearing Research, 47, 1301–1318.</bibtext> </blist> <blist> <bibtext> Fink R. (1986). A comparison of text characteristics in the narrative discourse of normal and learning isabled children. Unpublished doctoral dissertation. University of Colorado, Boulder.</bibtext> </blist> <blist> <bibtext> Fisher Z., Tipton E. (2015). robumeta: An R-package for robust variance estimation in meta-analysis. https://arxiv.org/pdf/1503.02220.pdf</bibtext> </blist> <blist> <bibtext> Gillespie A., Graham S. (2014). A meta-analysis of writing interventions for students with learning disabilities. Exceptional Children, 80, 454–473.</bibtext> </blist> <blist> <bibtext> *Goulandris N., Snowling M., Walker I. (2000). Is dyslexia a form of specific language impairment? A comparison of dyslexia and language impaired children and adolescents. Annals of Dyslexia, 50, 103–120.</bibtext> </blist> <blist> <bibtext> Graham S. (2000). Should the natural learning approach replace traditional spelling instruction. Journal of Educational Psychology, 92, 235–247.</bibtext> </blist> <blist> <bibtext> Graham S. (2018). A revised writer(s)-within-community model of writing. Educational Psychologist, 53, 258–279.</bibtext> </blist> <blist> <bibtext> Graham S., Collins A., Rigby-Wills H. (2017). A meta-analysis examining the writing characteristics of students with learning disabilities and normally achieving peers. Exceptional Children, 83, 199–218.</bibtext> </blist> <blist> <bibtext> Graham S., Harris K. R. (2018). Evidence-based writing practices: A meta-analysis of existing meta-analyses. In Fidalgo R., Harris K. R., Braaksma M. (Eds.), Design principles for teaching effective writing: Theoretical and empirical grounded principles (pp. 13–37). Brill Editions.</bibtext> </blist> <blist> <bibtext> Graham S., Liu K., Bartlett B., Ng C., Harris K. R., Aitken A., Barkel A., Kavanaugh C., Talukdar J. (2018). Reading for writing: A meta-analysis of the impact of reading and reading instruction on writing. Review of Educational Research, 88, 243–284.</bibtext> </blist> <blist> <bibtext> Graham S., Weintraub N. (1996). A review of handwriting research: Progress and prospect from l980 to l993. Educational Psychology Review, 8, 7–87.</bibtext> </blist> <blist> <bibtext> *Hedberg N. L., Fink R. J. (1996). Cohesive harmony in the written stories of elementary children. Reading and Writing: An Interdisciplinary Journal, 8, 73–86.</bibtext> </blist> <blist> <bibtext> *Jacobs D. (2009). The management of cohesion in written narratives in students with specific language impairment: Differences between childhood and adolescence [Unpublished dissertation]. RMIT University.</bibtext> </blist> <blist> <bibtext> Joye N., Broc L., Olive T., Dockrell J. (2019). Spelling performance in children with developmental language disorder: A meta-analysis across European languages. Scientific Studies of Reading, 23, 129–160.</bibtext> </blist> <blist> <bibtext> *Kim Y., Puranick C., Al Otaiba S. (2015). Developmental trajectories of writing skills in first grade: The effects of SES, and speech and language impairments. The Elementary School Journal, 115, 593–613.</bibtext> </blist> <blist> <bibtext> *Koutsoftas D. A., Gray S. (2011). Comparison of narrative and expository writing in students with and without language-learning disabilities. Language, Speech, and Hearing Services in Schools, 43, 395–409.</bibtext> </blist> <blist> <bibtext> Krishnan S., Watkins K., Bishop K. (2016). Neurobiological basis of language learning difficulties. Trends in Cognitive Sciences, 20, 701–714.</bibtext> </blist> <blist> <bibtext> *Larkin R. F., Snowling M. J. (2008). Comparing phonological skills and spelling abilities in children with reading and language impairments. International Journal of Language & Communication Disorders, 43, 111–124.</bibtext> </blist> <blist> <bibtext> *Larkin R. F., Williams G. J., Blaggan S. (2013). Delay or deficit? Spelling processes in children with specific language impairment. Journal of Communication Disorders, 46, 401–412.</bibtext> </blist> <blist> <bibtext> *Lawrence J. M. (2008). Differences in morphological awareness skills between children with phonological impairment and children with typical development [Unpublished dissertation, Florida State University]. ProQuest.</bibtext> </blist> <blist> <bibtext> Leonard L. (1998). Children with specific language impairments. MIT Press.</bibtext> </blist> <blist> <bibtext> Levelt W. (1993). Speaking: From intention to articulation. MIT Press.</bibtext> </blist> <blist> <bibtext> *Lewis B. A., O'Donnell B., Freebairn L. A., Taylor H. G. (1998). Spoken language and written expression: Interplay of delays. American Journal of Speech-Language Pathology, 7, 77–84.</bibtext> </blist> <blist> <bibtext> Loncke F. (2016). Communication disorders. In Kauffman J., Hallahan D., Pullen P. (Eds.), Handbook of special education (Vol. 2, pp. 325–337). Routledge.</bibtext> </blist> <blist> <bibtext> *Mackie C. J., Dockrell J. E. (2004). The nature of written language deficits in children with SLI. Journal of Speech, Language, and Hearing Research, 47, 1469–1483.</bibtext> </blist> <blist> <bibtext> *Mackie C. J., Dockrell J., Lindsay G. (2013). An evaluation of the written texts of children with DLD: The contributions of oral language, reading and phonological short-term memory. Reading and Writing, 26, 865–888.</bibtext> </blist> <blist> <bibtext> *Magnusson E., Nauclér K. (1990). Reading and spelling in language-disordered children-linguistic and metalinguistic prerequisites: Report on a longitudinal study. Clinical Linguistics & Phonetics, 4, 49–61.</bibtext> </blist> <blist> <bibtext> *McCarthy J. H., Hogan T. P., Catts H. W. (2012). Is weak oral language associated with poor spelling in school-age children with specific language impairment, dyslexia or both? Clinical Linguistics & Phonetics, 26, 791–805.</bibtext> </blist> <blist> <bibtext> McCurdy M. (2002). Examining the effects of a multi-component writing program in increasing three writing skills of secondary school students [Unpublished doctoral dissertation]. Mississippi State University.</bibtext> </blist> <blist> <bibtext> McCutchen D. (1996). A capacity theory of writing: Working memory in composition. Educational Psychology Review, 8, 299–325.</bibtext> </blist> <blist> <bibtext> *McFadden T. U., Gillam R. B. (1996). An examination of the quality of narratives produced by children with language disorders. Language, Speech, and Hearing Services in Schools, 27, 48–56.</bibtext> </blist> <blist> <bibtext> Metha P., Foorman B., Branum-Martin L., Taylor P. (2005). Literacy as a unidimensional multilevel construct: Validation, sources of influence, and implications in a longitudinal study in grades 1 to 4. Scientific Studies in Reading, 9, 85–116.</bibtext> </blist> <blist> <bibtext> *Myklebust H. R. (1965). Development and disorders of written language (Vol. 1). Grune & Stratton.</bibtext> </blist> <blist> <bibtext> *Nathan L., Stackhouse J., Goulandris N., Snowling M. J. (2004). Educational consequences of developmental speech disorder: Key Stage 1 National Curriculum assessment results in English and mathematics. British Journal of Educational Psychology, 74, 173–186.</bibtext> </blist> <blist> <bibtext> *Nauclér K. (2004). Spelling development in Swedish children with and without language impairment. Journal of Multilingual Communication Disorders, 2, 207–215.</bibtext> </blist> <blist> <bibtext> *Nauclér K., Magnusson E. (1999). Reading and writing development: A longitudinal study from pre-school to adolescence: Status report. Working Papers, 47, 169–180.</bibtext> </blist> <blist> <bibtext> Nauclér K., Magnusson E. (2002). How do preschool language problems affect language abilities in adolescents. In Windsor F., Kelly M. (Eds.), Investigations in clinical phonetics and linguistics (pp. 99–114). Lawrence Erlbaum.</bibtext> </blist> <blist> <bibtext> Nippold M. (2007). Later language development: School-age children, adolescents, and young adults. Pro-Ed.</bibtext> </blist> <blist> <bibtext> *Overby M. S. (2008). Relationships among speech sound perception, speech sound production, and phonological spelling in second grade children. University of Nebraska–Lincoln.</bibtext> </blist> <blist> <bibtext> Polanin J. R., Tanner-Smith E. E., Hennessy E. A. (2016). Estimating the difference between published and unpublished effect sizes: A meta-review. Review of Educational Research, 86, 207–236.</bibtext> </blist> <blist> <bibtext> Pressley M., Graham S., Harris K. R. (2006). The state of educational intervention research. British Journal of Educational Psychology, 76, 1–19.</bibtext> </blist> <blist> <bibtext> *Puranik C. S., Al Otaiba S., Ye F. (2014). Examining early spelling and writing skills: A comparative analysis of Kindergarteners with speech and oral language impairments and their typically developing peers. In Arte B., Dockrell J., Berninger V. (Eds.), Handbook of writing development and instruction in children with language difficulties (pp. 112–129). Oxford University Press.</bibtext> </blist> <blist> <bibtext> *Puranik C. S., Lonigan C. J. (2012). Early writing deficits in preschoolers with oral language difficulties. Journal of Learning Disabilities, 45, 179–190.</bibtext> </blist> <blist> <bibtext> Rastle K. (2019). EPS mid-career prize lecture 2017: Writing systems, reading, and language. Quarterly Journal of Experi-mental Psychology, 72, 677–692.</bibtext> </blist> <blist> <bibtext> *Reilly J., Bernicot J., Olive T., Uzé J., Wulfeck B., Favart M., Applebaum M. (2014). Written narratives from French and English speaking children with language impairment. In Arte B., Dockrell J., Berninger V. (Eds.), Handbook of writing development and instruction in children with language difficulties (pp. 176–186). Oxford University Press.</bibtext> </blist> <blist> <bibtext> *Romaglia N. (2010). The effect of topic familiarity on expository and narrative writing tasks in children with specific language impairment [Doctoral dissertation]. William Paterson University.</bibtext> </blist> <blist> <bibtext> Scammacca N., Roberts G., Stuebing K. K. (2014). Meta-analysis with complex research designs: Dealing with dependence from multiple measures and multiple group comparisons. Review of Educational Research, 84, 326–384.</bibtext> </blist> <blist> <bibtext> *Scott C. M., Windsor J. (2000). General language performance measures in spoken and written narrative and expository discourse of school-age children with language learning disabilities. Journal of Speech, Language, and Hearing Research, 43, 324–339.</bibtext> </blist> <blist> <bibtext> *Scuccimarra G., Cutolo L., Fiorillo P., Lembo C., Pirone T., Cossu G. (2008). Is there a distinct form of developmental dyslexia in children with specific language impairment? Findings from an orthographically regular language. Cognitive and Behavioral Neurology, 21, 221–226.</bibtext> </blist> <blist> <bibtext> Shadish W., Robinson L., Lu C. (1997). Effect size (ES) analysis program. University of Memphis.</bibtext> </blist> <blist> <bibtext> Shanahan T. (2006). Relations among oral language, reading, and writing development. In MacArthur C. A., Graham S., Fitzgerald J. (Eds.), Handbook of writing research (pp. 171–183). Guilford Press.</bibtext> </blist> <blist> <bibtext> Shen M., Troia G. (2018). Teaching children with language-learning disabilities to plan and revise compare-contrast text. Learning Disability Quarterly, 41, 44–61.</bibtext> </blist> <blist> <bibtext> Snowling M. (2000). Dyslexia (2nd ed.). Blackwell.</bibtext> </blist> <blist> <bibtext> Sperling M. (1996). Revisiting the writing-speaking connection: Challenges for research on writing and writing instruction. Review of Educational Research, 66, 53–86.</bibtext> </blist> <blist> <bibtext> Tanner-Smith E., Tipton E. (2013). Robust variance estimation with dependent effect sizes: Practical considerations including a software tutorial in Stata and SPSS. Research Synthesis Methods, 5, 13–30.</bibtext> </blist> <blist> <bibtext> *Tattersall P. J. (2010). Reading, writing, and repetition: Perfor-mance on nonword measures by students with and without language-learning disabilities [Unpublished doctoral dissertation]. Western Michigan University.</bibtext> </blist> <blist> <bibtext> Thomas M. (2005). Characterizing compensation. Cortex, 41, 434–442.</bibtext> </blist> <blist> <bibtext> Tomblin J., Records N., Buckwalter P., Xhang X., Smith E., O'Brien M. (1997). Prevalence of specific language impairment in kindergarten children. Journal of Speech, Language, and Hearing Research, 40, 1245–1260.</bibtext> </blist> <blist> <bibtext> Tukey J. W. (1977). Exploratory data analysis. Reading, MA: Addison-Wesley.</bibtext> </blist> <blist> <bibtext> Troia G. (2011). How might pragmatics language skills affect written expression of students with language learning disabilities. Topics in Language Disorders, 31, 40–53.</bibtext> </blist> <blist> <bibtext> Ullman M., Peirpont E. (2005). Specific language impairment is not specific to language: The procedural deficit hypothesis. Cortex, 41, 389–433.</bibtext> </blist> <blist> <bibtext> *Vandewalle E., Boets B., Ghesquière P., Zink I. (2012). Auditory processing and speech perception in children with specific language impairment: Relations with oral language and literacy skills. Research in Developmental Disabilities, 33, 635–644.</bibtext> </blist> <blist> <bibtext> *Walz L., Thompson S., Thurlow M., Spicuzza R. (2000). Minnesota's comprehensive assessments: 1998 and 1999 participation and performance of students with disabilities (State Assessment Series Report No. 32). National Center on Educational Outcomes.</bibtext> </blist> <blist> <bibtext> *Williams G. J., Larkin R. F., Blaggan S. (2013). Written language skills in children with specific language impairment. International Journal of Language & Communication Disorders, 48, 160–171.</bibtext> </blist> <blist> <bibtext> *Windsor J., Scott C. M., Street C. K. (2000). Verb and noun morphology in the spoken and written language of children with language learning disabilities. Journal of Speech, Language, and Hearing Research, 43, 1322–1336.</bibtext> </blist> <blist> <bibtext> *Wolter J. A., Self T., Apel K. (2011). Initial mental graphemic representation acquisition and later literacy achievement in children with language impairment: A longitudinal study. Journal of Learning Disabilities, 44, 543–555.</bibtext> </blist> </ref> <ref id="AN0143828403-33"> <title> Footnotes </title> <blist> <bibtext> Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.</bibtext> </blist> <blist> <bibtext> Funding The author(s) received no financial support for the research, authorship, and/or publication of this article.</bibtext> </blist> <blist> <bibtext> ORCID iD Amber B. Ray</bibtext> </blist> <blist> <bibtext>Graph https://orcid.org/0000-0002-6527-7102</bibtext> </blist> </ref> <aug> <p>By Steve Graham; Michael Hebert; Evan Fishman; Amber B. Ray and Amy Gillespie Rouse</p> <p>Reported by Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib86" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib49" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib81" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib14" firstref="ref5"></nolink> <nolink nlid="nl5" bibid="bib53" firstref="ref6"></nolink> <nolink nlid="nl6" bibid="bib76" firstref="ref7"></nolink> <nolink nlid="nl7" bibid="bib36" firstref="ref8"></nolink> <nolink nlid="nl8" bibid="bib79" firstref="ref10"></nolink> <nolink nlid="nl9" bibid="bib82" firstref="ref11"></nolink> <nolink nlid="nl10" bibid="bib50" firstref="ref12"></nolink> <nolink nlid="nl11" bibid="bib35" firstref="ref13"></nolink> <nolink nlid="nl12" bibid="bib72" firstref="ref15"></nolink> <nolink nlid="nl13" bibid="bib16" firstref="ref16"></nolink> <nolink nlid="nl14" bibid="bib52" firstref="ref18"></nolink> <nolink nlid="nl15" bibid="bib58" firstref="ref20"></nolink> <nolink nlid="nl16" bibid="bib13" firstref="ref21"></nolink> <nolink nlid="nl17" bibid="bib30" firstref="ref22"></nolink> <nolink nlid="nl18" bibid="bib60" firstref="ref23"></nolink> <nolink nlid="nl19" bibid="bib38" firstref="ref24"></nolink> <nolink nlid="nl20" bibid="bib37" firstref="ref25"></nolink> <nolink nlid="nl21" bibid="bib88" firstref="ref29"></nolink> <nolink nlid="nl22" bibid="bib66" firstref="ref30"></nolink> <nolink nlid="nl23" bibid="bib85" firstref="ref31"></nolink> <nolink nlid="nl24" bibid="bib89" firstref="ref32"></nolink> <nolink nlid="nl25" bibid="bib45" firstref="ref33"></nolink> <nolink nlid="nl26" bibid="bib34" firstref="ref34"></nolink> <nolink nlid="nl27" bibid="bib20" firstref="ref35"></nolink> <nolink nlid="nl28" bibid="bib39" firstref="ref36"></nolink> <nolink nlid="nl29" bibid="bib65" firstref="ref40"></nolink> <nolink nlid="nl30" bibid="bib12" firstref="ref42"></nolink> <nolink nlid="nl31" bibid="bib42" firstref="ref43"></nolink> <nolink nlid="nl32" bibid="bib69" firstref="ref44"></nolink> <nolink nlid="nl33" bibid="bib15" firstref="ref46"></nolink> <nolink nlid="nl34" bibid="bib68" firstref="ref47"></nolink> <nolink nlid="nl35" bibid="bib26" firstref="ref52"></nolink> <nolink nlid="nl36" bibid="bib78" firstref="ref54"></nolink> <nolink nlid="nl37" bibid="bib87" firstref="ref55"></nolink> <nolink nlid="nl38" bibid="bib91" firstref="ref56"></nolink> <nolink nlid="nl39" bibid="bib54" firstref="ref58"></nolink> <nolink nlid="nl40" bibid="bib71" firstref="ref59"></nolink> <nolink nlid="nl41" bibid="bib77" firstref="ref61"></nolink> <nolink nlid="nl42" bibid="bib75" firstref="ref63"></nolink> <nolink nlid="nl43" bibid="bib83" firstref="ref64"></nolink> <nolink nlid="nl44" bibid="bib10" firstref="ref66"></nolink> <nolink nlid="nl45" bibid="bib31" firstref="ref67"></nolink> <nolink nlid="nl46" bibid="bib19" firstref="ref70"></nolink> <nolink nlid="nl47" bibid="bib93" firstref="ref73"></nolink> <nolink nlid="nl48" bibid="bib29" firstref="ref74"></nolink> <nolink nlid="nl49" bibid="bib56" firstref="ref75"></nolink> <nolink nlid="nl50" bibid="bib55" firstref="ref76"></nolink> <nolink nlid="nl51" bibid="bib63" firstref="ref77"></nolink> <nolink nlid="nl52" bibid="bib41" firstref="ref80"></nolink> <nolink nlid="nl53" bibid="bib28" firstref="ref83"></nolink> <nolink nlid="nl54" bibid="bib70" firstref="ref85"></nolink> <nolink nlid="nl55" bibid="bib61" firstref="ref86"></nolink> <nolink nlid="nl56" bibid="bib21" firstref="ref88"></nolink> <nolink nlid="nl57" bibid="bib40" firstref="ref90"></nolink> <nolink nlid="nl58" bibid="bib47" firstref="ref91"></nolink> <nolink nlid="nl59" bibid="bib11" firstref="ref92"></nolink> <nolink nlid="nl60" bibid="bib48" firstref="ref93"></nolink> <nolink nlid="nl61" bibid="bib43" firstref="ref94"></nolink> <nolink nlid="nl62" bibid="bib27" firstref="ref99"></nolink> <nolink nlid="nl63" bibid="bib23" firstref="ref115"></nolink> <nolink nlid="nl64" bibid="bib57" firstref="ref116"></nolink> <nolink nlid="nl65" bibid="bib80" firstref="ref117"></nolink> <nolink nlid="nl66" bibid="bib32" firstref="ref118"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1257088
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Do Children Classified with Specific Language Impairment Have a Learning Disability in Writing? A Meta-Analysis
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Graham%2C+Steve%22">Graham, Steve</searchLink><br /><searchLink fieldCode="AR" term="%22Hebert%2C+Michael%22">Hebert, Michael</searchLink><br /><searchLink fieldCode="AR" term="%22Fishman%2C+Evan%22">Fishman, Evan</searchLink><br /><searchLink fieldCode="AR" term="%22Ray%2C+Amber+B%2E%22">Ray, Amber B.</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0002-6527-7102">0000-0002-6527-7102</externalLink>)<br /><searchLink fieldCode="AR" term="%22Rouse%2C+Amy+Gillespie%22">Rouse, Amy Gillespie</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Learning+Disabilities%22"><i>Journal of Learning Disabilities</i></searchLink>. Jul-Aug 2020 53(4):292-310.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: SAGE Publications and Hammill Institute on Disabilities. 2455 Teller Road, Thousand Oaks, CA 91320. Tel: 800-818-7243; Tel: 805-499-9774; Fax: 800-583-2665; e-mail: journals@sagepub.com; Web site: http://sagepub.com
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 19
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2020
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research<br />Information Analyses
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Elementary+Education%22">Elementary Education</searchLink><br /><searchLink fieldCode="EL" term="%22Secondary+Education%22">Secondary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Writing+Difficulties%22">Writing Difficulties</searchLink><br /><searchLink fieldCode="DE" term="%22Language+Impairments%22">Language Impairments</searchLink><br /><searchLink fieldCode="DE" term="%22Learning+Disabilities%22">Learning Disabilities</searchLink><br /><searchLink fieldCode="DE" term="%22Writing+Skills%22">Writing Skills</searchLink><br /><searchLink fieldCode="DE" term="%22Speech+Impairments%22">Speech Impairments</searchLink><br /><searchLink fieldCode="DE" term="%22Oral+Language%22">Oral Language</searchLink><br /><searchLink fieldCode="DE" term="%22Age+Differences%22">Age Differences</searchLink><br /><searchLink fieldCode="DE" term="%22Classification%22">Classification</searchLink><br /><searchLink fieldCode="DE" term="%22Educational+Research%22">Educational Research</searchLink><br /><searchLink fieldCode="DE" term="%22Writing+Tests%22">Writing Tests</searchLink><br /><searchLink fieldCode="DE" term="%22Test+Format%22">Test Format</searchLink><br /><searchLink fieldCode="DE" term="%22Preschool+Children%22">Preschool Children</searchLink><br /><searchLink fieldCode="DE" term="%22Adolescents%22">Adolescents</searchLink><br /><searchLink fieldCode="DE" term="%22Elementary+School+Students%22">Elementary School Students</searchLink><br /><searchLink fieldCode="DE" term="%22Secondary+School+Students%22">Secondary School Students</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1177/0022219420917338
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0022-2194
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: In this meta-analysis, we examined whether children classified with specific language impairment (SLI) experience difficulties with writing. We included studies comparing children with SLI to (a) typically developing peers matched on age (k = 39 studies) and (b) typically developing younger peers with similar language capabilities (k = six studies). Children classified with SLI scored lower on writing measures than their typically developing peers matched on age (g = -0.97) when all writing scores in a study were included in the analysis. This same pattern occurred for specific measures of writing: quality (g = -0.92), output (g = -1.00), grammar (g = -0.68), vocabulary (g = -0.68), and spelling (g = -1.17). A moderator analysis revealed that differences in the writing scores of children classified with SLI and typically developing peers matched on age were not as large, but were still statistically significant, when assessment involved a contrived response format (vs. measured based on students' writing), researcher-created measures (vs. norm-referenced tests), or SLI included just children with a speech disorder (vs. children with a language disorder). Children classified with SLI further scored lower on writing than typically developing peers with similar language capabilities (g = -0.47). We concluded that children with SLI experience difficulties with writing.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2020
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1257088
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1257088
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1177/0022219420917338
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 19
        StartPage: 292
    Subjects:
      – SubjectFull: Writing Difficulties
        Type: general
      – SubjectFull: Language Impairments
        Type: general
      – SubjectFull: Learning Disabilities
        Type: general
      – SubjectFull: Writing Skills
        Type: general
      – SubjectFull: Speech Impairments
        Type: general
      – SubjectFull: Oral Language
        Type: general
      – SubjectFull: Age Differences
        Type: general
      – SubjectFull: Classification
        Type: general
      – SubjectFull: Educational Research
        Type: general
      – SubjectFull: Writing Tests
        Type: general
      – SubjectFull: Test Format
        Type: general
      – SubjectFull: Preschool Children
        Type: general
      – SubjectFull: Adolescents
        Type: general
      – SubjectFull: Elementary School Students
        Type: general
      – SubjectFull: Secondary School Students
        Type: general
    Titles:
      – TitleFull: Do Children Classified with Specific Language Impairment Have a Learning Disability in Writing? A Meta-Analysis
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Graham, Steve
      – PersonEntity:
          Name:
            NameFull: Hebert, Michael
      – PersonEntity:
          Name:
            NameFull: Fishman, Evan
      – PersonEntity:
          Name:
            NameFull: Ray, Amber B.
      – PersonEntity:
          Name:
            NameFull: Rouse, Amy Gillespie
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2020
          Identifiers:
            – Type: issn-print
              Value: 0022-2194
          Numbering:
            – Type: volume
              Value: 53
            – Type: issue
              Value: 4
          Titles:
            – TitleFull: Journal of Learning Disabilities
              Type: main
ResultId 1