View in EDS HTML Full Text PDF Full Text

Connective Knowledge and Reading Comprehension in Upper Elementary Students: A Growth Analysis

Saved in:
Bibliographic Details
Title: Connective Knowledge and Reading Comprehension in Upper Elementary Students: A Growth Analysis
Language: English
Authors: Bailey Buchanan (ORCID 0009-0009-2574-4835), Paola Uccelli (ORCID 0000-0001-5818-2108)
Source: Reading Research Quarterly. 2026 61(2).
Availability: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed: Y
Page Count: 13
Publication Date: 2026
Sponsoring Agency: Institute of Education Sciences (ED)
Department of Education (ED)
Contract Number: R305A170185
R305A190034
Document Type: Journal Articles
Reports - Research
Education Level: Elementary Education
Descriptors: Elementary School Students, Reading Comprehension, Connected Discourse, Receptive Language, English Learners, Language Proficiency, Reading Achievement
DOI: 10.1002/rrq.70096
ISSN: 0034-0553
1936-2722
Abstract: Connectives--a relatively small, closed set of expressions used to link ideas logically, for example, "therefore," "in contrast"--represent a potentially high-leverage area of focus for literacy interventions due to their prevalence and utility across content-area texts. To inform instruction, however, we first need more research to understand students' development of connective knowledge and test its potential contribution to reading comprehension. In the present study, we used latent growth analysis to examine developmental relations between receptive connective knowledge and reading comprehension in English learners (ELs) and English proficient (EP) students from grade 4 to grade 6 (N = 4100). Three primary findings emerged from our analysis. First, students with greater initial connective knowledge at the start of fourth grade displayed, on average, greater growth in reading comprehension between grades 4 and 6. Second, more rapid growth in connective knowledge across this same timespan predicted, on average, greater growth in reading comprehension. Third, our study finds that the relation between connective knowledge growth and reading comprehension growth was not significantly different for ELs as compared to EP students. To our knowledge, this is the first study to analyze the relation between connective knowledge and reading comprehension in upper elementary students over time and how this relation varies by student language background. These results motivate future intervention-based research to identify possible causal pathways underlying this developmental relation and directs practitioners to consider connective knowledge as a particular instructional area with potential benefits on reading comprehension outcomes for both English learning and English proficient students.
Abstractor: As Provided
IES Funded: Yes
Entry Date: 2026
Accession Number: EJ1503743
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwHAJTIfoC7BIwn4JoQoZlKmAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDDh44106jRi6BxwBvAIBEICBm9lgoS7Ct4STjyO52HrXQQvzdPAx3ohjeYR0nAGojMTwdJQfh6XKQcttmZdVGedcR7m1eW-DqdSbzHN4cXlN8ZygorVfyfkBX4EKkR2b1afa-9zCgtX7kB24Fe12_Umx6_VkUZN_9O45fS6zkKwgSqHpLS8lUtD0DNy04Lq5eWQy2p9u-bnwA_Cyh3kFHbWFRlTOYQSBcPSnKff2
Text:
  Availability: 1
  Value: <anid>AN0193225963;[nrnu]02apr.26;2026Apr27.05:00;v2.2.500</anid> <title id="AN0193225963-1">Connective Knowledge and Reading Comprehension in Upper Elementary Students: A Growth Analysis </title> <p>Connectives—a relatively small, closed set of expressions used to link ideas logically, for example, "therefore," "in contrast"—represent a potentially high‐leverage area of focus for literacy interventions due to their prevalence and utility across content‐area texts. To inform instruction, however, we first need more research to understand students' development of connective knowledge and test its potential contribution to reading comprehension. In the present study, we used latent growth analysis to examine developmental relations between receptive connective knowledge and reading comprehension in English learners (ELs) and English proficient (EP) students from grade 4 to grade 6 (N = 4100). Three primary findings emerged from our analysis. First, students with greater initial connective knowledge at the start of fourth grade displayed, on average, greater growth in reading comprehension between grades 4 and 6. Second, more rapid growth in connective knowledge across this same timespan predicted, on average, greater growth in reading comprehension. Third, our study finds that the relation between connective knowledge growth and reading comprehension growth was not significantly different for ELs as compared to EP students. To our knowledge, this is the first study to analyze the relation between connective knowledge and reading comprehension in upper elementary students over time and how this relation varies by student language background. These results motivate future intervention‐based research to identify possible causal pathways underlying this developmental relation and directs practitioners to consider connective knowledge as a particular instructional area with potential benefits on reading comprehension outcomes for both English learning and English proficient students.</p> <p>In this study, we examined longitudinal relations of connective knowledge and reading comprehension between grades 4 and 6. Students with greater connective knowledge at the start of fourth grade displayed greater growth in reading comprehension through grade 6. In addition, more rapid growth in connective knowledge predicted greater growth in reading comprehension. Finally, the relation between connective knowledge growth and reading comprehension growth was not significantly different for English learners compared to English proficient students.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-toc-0001.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-toc-0001.jpg" title="." /> </p> <p></p> <hd id="AN0193225963-3">Introduction</hd> <p>Among the multiple factors that contribute to skilled text comprehension in the upper elementary years and beyond (such as difficulties with decoding, reading strategies, engagement, or limited relevant background knowledge), knowledge of the language of texts—including the understudied domain of connective knowledge—has emerged as a prominent source of reading comprehension underperformance. Following Andreev and Uccelli ([<reflink idref="bib1" id="ref1">1</reflink>]), we define connectives as linguistic devices that signal logical relations between ideas (e.g., <emph>however</emph>) or organizational relations about textual information (e.g., <emph>first of all</emph>) (Halliday and Matthiessen [<reflink idref="bib16" id="ref2">16</reflink>]; Hyland [<reflink idref="bib18" id="ref3">18</reflink>]). Connectives play an important role in reading comprehension because they facilitate readers' understanding of how ideas in a text relate to one another. For example, reading "<emph>in addition</emph>" signals to a reader that the next idea will be an extension or elaboration on the prior one; reading "<emph>in contrast</emph>," instead, prepares the reader for the next idea to be an opposing one. However, connectives will function as such helpful cues only if readers are familiar with the connectives they encounter in text (Cain and Nash [<reflink idref="bib5" id="ref4">5</reflink>]). As it turns out, some recent evidence suggests that for large proportions of middle‐school readers, connectives may function more often as roadblocks than as signposts (Uccelli, Galloway, et al. [<reflink idref="bib36" id="ref5">36</reflink>]).</p> <p>Starting around grade 4, the shifting language demands of written texts present novel challenges to students, including monolingual, English proficient students (Schleppegrell [<reflink idref="bib33" id="ref6">33</reflink>]). Not surprisingly, the language demands of content‐area texts are particularly challenging for multilingual students who are still in the process of developing grade‐level English proficiency, that is, those bureaucratically designated as "English learners" (ELs) in U.S. schools[<reflink idref="bib1" id="ref7">1</reflink>] (ESSA [<reflink idref="bib9" id="ref8">9</reflink>], 393; Phillips Galloway and Uccelli [<reflink idref="bib12" id="ref9">12</reflink>]). Recent research demonstrates that upper elementary and middle school students—both monolingual and multilingual—exhibit striking individual differences in their knowledge of the language of texts, which in turn help explain their challenges with school reading comprehension, even after taking into account word recognition skills, vocabulary knowledge, and sociodemographic characteristics (Phillips Galloway and Uccelli [<reflink idref="bib11" id="ref10">11</reflink>]; Uccelli, Galloway, et al. [<reflink idref="bib36" id="ref11">36</reflink>]). Prior research reveals, for instance, that while students' knowledge of the language of text tends to increase progressively from grade 4 to grade 8, students' skills vary widely, with individuals spanning a wide range of performance percentiles at any given grade (Uccelli, Barr, et al. [<reflink idref="bib35" id="ref12">35</reflink>]; Barr et al. [<reflink idref="bib3" id="ref13">3</reflink>]).</p> <p>Among the multiple language domains relevant for reading comprehension (e.g., vocabulary knowledge, morphological awareness), in this study we focus on a specific and still understudied language domain: knowledge of connectives. Whereas educationally relevant assessments and interventions focused on connectives have expanded recently (August et al. [<reflink idref="bib2" id="ref14">2</reflink>]; Van Silfhout et al. [<reflink idref="bib37" id="ref15">37</reflink>]), little is known about how connectives and reading comprehension develop over time in upper elementary and middle‐grade students and how that development varies by English proficiency designation (i.e., ELs vs. EPs). Additionally, there is conflicting evidence about how connectives may or may not contribute differently to reading comprehension for English learners compared to English proficient students (Crosson and Lesaux [<reflink idref="bib6" id="ref16">6</reflink>]; Welie et al. [<reflink idref="bib39" id="ref17">39</reflink>]). The present study contributes to this expanding body of work by examining how connective knowledge develops over time in relation to reading comprehension in English learners (ELs) and their English proficient (EP) peers from grades 4 to 6. The research questions driving this study are:</p> <p></p> <ulist> <item> RQ1: Does connective knowledge at the start of fourth grade predict growth in reading comprehension between grade 4 and grade 6?</item> <p></p> <item> RQ2: Does growth in connective knowledge predict growth in reading comprehension between grade 4 and grade 6?</item> <p></p> <item> RQ3: Is the relation between connective knowledge growth and reading comprehension growth in grades 4–6 similar or different for students designated as English learners versus their English proficient peers?</item> </ulist> <p>To examine these questions, we conducted a latent growth analysis on a longitudinal sample of students in grades 4–6 enrolled in U.S. public schools (<emph>N</emph> = 4100). We used a psychometrically robust measure of receptive knowledge of connectives (Uccelli, Barr, et al. [<reflink idref="bib35" id="ref18">35</reflink>]), and an extensively validated measure of reading comprehension (Sabatini et al. [<reflink idref="bib31" id="ref19">31</reflink>]). Results from this study offer novel evidence of the contribution of students' connective knowledge to their reading comprehension growth, revealing an important developmental relation that can inform future literacy intervention studies.</p> <p>In the next section, we review the literature on the relation between connective knowledge and reading comprehension, with a focus on how this association may differ by students' language background. Next, we describe our methods and the results of each research question. We close with a discussion of how our findings contribute novel insights to theory and practice.</p> <hd id="AN0193225963-4">Connectives and Reading Comprehension</hd> <p>Connectives are a set of function words that signal to the reader how ideas are connected across clauses, sentences, or larger units of discourse (Halliday and Matthiessen [<reflink idref="bib16" id="ref20">16</reflink>]; Hyland [<reflink idref="bib18" id="ref21">18</reflink>]). Aligned with the operational definition proposed by Andreev and Uccelli ([<reflink idref="bib1" id="ref22">1</reflink>]), we define connectives as "cohesive devices–single words or phrases–that signal to the reader logical/conceptual relations between ideas (e.g., <emph>consequently</emph>) or organizational relations about the informational flow in a text (e.g., <emph>first of all</emph>)" (p. 175). Although connectives have little semantic meaning, they provide the reader with explicit cues about how ideas or fragments of texts relate to one another, therefore aiding in text processing and comprehension (Van Silfhout et al. [<reflink idref="bib37" id="ref23">37</reflink>]; Cain and Nash [<reflink idref="bib5" id="ref24">5</reflink>]; Townsend et al. [<reflink idref="bib34" id="ref25">34</reflink>]; Fraser et al. [<reflink idref="bib10" id="ref26">10</reflink>]).</p> <p>Even though their frequency and use vary across disciplines (Román et al. [<reflink idref="bib29" id="ref27">29</reflink>]; Ibáñez et al. [<reflink idref="bib19" id="ref28">19</reflink>]), connectives are ubiquitous in school texts read across content areas. Without understanding the connectives commonly used in school texts (e.g., <emph>in contrast, therefore, nevertheless</emph>), students cannot benefit from the signals these markers provide, and may even misinterpret how ideas are related in a text. In such cases, connectives become roadblocks that hinder rather than support comprehension.</p> <p>As connectives are a relatively small and closed set of expressions of high utility in texts across content areas, they represent a potentially high‐leverage area of focus in literacy interventions. To inform instruction, however, we first need more research to understand students' development of connective knowledge and test their hypothesized contribution to reading comprehension. The very few intervention studies on connectives already suggest that explicit instruction can improve students' connective knowledge (August et al. [<reflink idref="bib2" id="ref29">2</reflink>]; Quílez [<reflink idref="bib28" id="ref30">28</reflink>]), but these studies so far have focused only on English learners in the early elementary or high school grades, leaving upper elementary and English proficient students' connectives knowledge underexplored.</p> <p>The limited existing research focused on early and mid‐adolescents already offers evidence of the unique contribution of connective knowledge on reading comprehension, even after controlling for students' vocabulary knowledge, reading fluency, and sociodemographic characteristics (Cain and Nash [<reflink idref="bib5" id="ref31">5</reflink>]; Crosson and Lesaux [<reflink idref="bib6" id="ref32">6</reflink>]; Townsend et al. [<reflink idref="bib34" id="ref33">34</reflink>]). Studies have documented that large proportions of students in the upper elementary and middle grades tend to have low knowledge of connectives. For example, in a sample of over 6000 English‐speaking students in the U.S. found that only 46% of eighth graders understood the meaning of "<emph>in contrast</emph>" (Barr et al. [<reflink idref="bib3" id="ref34">3</reflink>]), and Andreev and Uccelli's ([<reflink idref="bib1" id="ref35">1</reflink>]) writing study revealed that students in grades 5–8 tended to use a very limited set of primarily basic connectives (e.g., <emph>and</emph>, <emph>because</emph>, <emph>or</emph>) in their persuasive essays. Specifically, "84% of all connectives used in students' writing were represented exclusively by 10 connectives, 8 of which were basic connectives" (Andreev and Uccelli [<reflink idref="bib1" id="ref36">1</reflink>], 188). The vast majority of studies on students' connective knowledge, however, has been conducted with cross‐sectional samples. The present study extends prior research by analyzing the longitudinal relation between connective knowledge and reading comprehension in grades 4 through 6.</p> <hd id="AN0193225963-5">Language Learners: Connectives and Reading Comprehension</hd> <p>The National Academy of Sciences, Engineering, and Medicine (NASEM [<reflink idref="bib26" id="ref37">26</reflink>]) identifies fourth grade as a specific year in which "many ELs falter" (p. 300) if they have not acquired sufficient English literacy skills to read informational texts and to understand the analytical language used to convey and discuss abstract ideas in the content areas. In other words, being a proficient decoder or fluent in using English colloquially is not sufficient to successfully engage in content‐area reading and learning. Despite the consensus on the need to support ELs' language learning, there is still a significant gap in the field's understanding of which language resources might best support ELs' literacy development in the upper elementary grades and beyond (NASEM [<reflink idref="bib26" id="ref38">26</reflink>]). We argue here that knowledge of connectives is a promising area that deserves more attention.</p> <p>Whereas cross‐sectional studies show that connective knowledge plays a role in comprehending text, longitudinal research is still scarce. Theoretical and empirical evidence suggests that the magnitude and developmental trajectory of this relation may differ by student language background. For example, Crosson et al. ([<reflink idref="bib7" id="ref39">7</reflink>]) found that vocabulary knowledge and listening comprehension predicted fourth‐grade Spanish‐speaking ELs' connective comprehension, even after controlling for word reading skills. In other words, ELs with stronger vocabulary and listening skills had, on average, not surprisingly, better understanding of connectives. In another study comparing fifth‐grade ELs and English‐only, Crosson and Lesaux ([<reflink idref="bib6" id="ref40">6</reflink>]) found that connective knowledge predicted reading comprehension more strongly for English‐only students. This empirical base is still limited, though, and calls for further research to examine potential differences by English proficiency.</p> <p>Limited research with learners of languages other than English and with adult ELs also suggests that connective knowledge and text comprehension may differ by language background. In their study on the contribution of connectives to narrative text comprehension in ninth‐grade German monolinguals and German‐learning bilinguals, Kohnen and Retelsdorf ([<reflink idref="bib23" id="ref41">23</reflink>]) found that when controlling for vocabulary knowledge, reading fluency, and knowledge of reading strategies, connective knowledge significantly contributed to narrative comprehension only for German monolinguals. In adult English learners, Geva ([<reflink idref="bib13" id="ref42">13</reflink>]) investigated the comprehension of conjunctions and found that the advanced ELs were able to infer the logical relations that connectives indicated better than the intermediate ELs. Taken together, these findings suggest that connectives may not facilitate comprehension as effectively for learners with less proficient language skills when compared to more proficient speakers.</p> <p>Not all studies, however, confirm this trend. For example, Welie et al. ([<reflink idref="bib39" id="ref43">39</reflink>]) found no difference in the contribution of connective knowledge to reading comprehension between bilingual and monolingual Dutch eighth graders in the Netherlands. In their research on contributing factors to connective development, Volodina and Weinert ([<reflink idref="bib38" id="ref44">38</reflink>]) found that students' developmental trajectory of connective growth was predicted more by SES than language background in a sample of monolingual and multilingual primary school students in Germany. The researchers called for further studies on connectives development with larger, more diverse samples of learners representing various socioeconomic and language backgrounds. To advance this research base, in the present study, we examine a large, diverse, and longitudinal sample, and we compare English learner and English proficient students' connective knowledge and reading comprehension over time.</p> <p>First, we describe how receptive connective knowledge at the beginning of fourth grade predicts growth in reading comprehension from grade 4 to grade 6 in the full sample. Then, we examine the relation between connective knowledge growth rate and reading comprehension growth rate from grade 4 to grade 6. Examining each of these relations provides, to our knowledge, novel findings on how growth in receptive connective knowledge and reading comprehension are associated over time. If initial connective knowledge at the start of grade 4 predicts subsequent reading comprehension growth, this finding would offer evidence of the foundational role that connective knowledge plays in literacy development, suggesting that interventions focused on connectives may have a lasting impact. Additionally, if growth in connective knowledge predicts growth in reading comprehension—particularly when controlling for initial connective knowledge—this finding would suggest that interventions designed to teach connective knowledge may support reading development.</p> <p>Finally, we compare how the relation of connective knowledge growth rate to reading comprehension growth rate differs for English learners and English proficient students. Grouping the sample by English proficiency designation, we examine if differences in connective knowledge growth are associated with reading comprehension growth, on average, for one or both groups. To our knowledge, the current study is the first to employ growth modeling to analyze upper elementary students' connective knowledge and reading comprehension and to do so with a large sample of students enrolled in urban public school classrooms.</p> <hd id="AN0193225963-6">Methods</hd> <p></p> <hd id="AN0193225963-7">Sample</hd> <p>A total of 4100 students in grades 4–6 participated in this three‐year longitudinal study. Students came from 25 schools and four districts. The districts represent diverse urban and suburban areas in two states in the Northeastern United States. Among the participants, 50.9% identified as female, 85.5% qualified for free or reduced lunch, and 11.7% were designated as English learners. While we do not have ELs' English proficiency scores in our data, assessment protocols directed that students must be at level 3 or higher on the WIDA as a precondition for assessment. Percentages of student race, according to school records, were as follows: 41.8% Black, 28.6% White, 24% Latine, 2.66% Asian, 1% Native, and 1.9% Mixed (Table 1). The dataset was drawn from the control group of a large study on reading comprehension development and intervention, the <emph>Catalyzing Comprehension Through Discussion and Debate</emph> study (<emph>CCDD</emph>) (Jones et al. [<reflink idref="bib20" id="ref45">20</reflink>]). Data were collected in cohorts with different start dates across the 3 years, as not all schools and districts joined the study in Year 1. Participating students in the control group were included in the present study if they were in grades 4–6 and had complete data for both assessments (the connective knowledge subtest of the <emph>Core Analytic Language Skills</emph> instrument and the <emph>Global Integrated Scenario‐based Assessment</emph> of reading comprehension–GISA) for at least one measured time point.</p> <p>1 TABLE Student characteristics by English learner status.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left" /><th align="center">Full sample, <italic>N</italic> (%)</th><th align="center">English learners, <italic>N</italic> (%)</th><th align="center">English proficient, <italic>N</italic> (%)</th></tr></thead><tbody valign="top"><tr><td align="left">Total</td><td align="center">4100 (100%)</td><td align="center">479 (11.7%)</td><td align="center">3621 (88.3%)</td></tr><tr><td align="left">Cohort</td></tr><tr><td align="left">Cohort 1</td><td align="center">349 (8.5%)</td><td align="center">54 (11.3%)</td><td align="center">295 (8.1%)</td></tr><tr><td align="left">Cohort 2</td><td align="center">2151 (52.5%)</td><td align="center">240 (50.1%)</td><td align="center">1911 (52.8%)</td></tr><tr><td align="left">Cohort 3</td><td align="center">1600 (39.0%)</td><td align="center">185 (38.6%)</td><td align="center">1415 (39.1%)</td></tr><tr><td align="left">District</td></tr><tr><td align="left">District 1</td><td align="center">409 (10.0%)</td><td align="center">131 (27.3%)</td><td align="center">278 (7.7%)</td></tr><tr><td align="left">District 2</td><td align="center">744 (18.1%)</td><td align="center">68 (14.2%)</td><td align="center">676 (18.7%)</td></tr><tr><td align="left">District 3</td><td align="center">1113 (27.1%)</td><td align="center">151 (31.5%)</td><td align="center">962 (26.6%)</td></tr><tr><td align="left">District 4</td><td align="center">1834 (44.7%)</td><td align="center">129 (26.9%)</td><td align="center">1705 (47.1%)</td></tr><tr><td align="left">Demographics</td></tr><tr><td align="left">White</td><td align="center">1174 (28.6%)</td><td align="center">38 (7.9%)</td><td align="center">1136 (31.4%)</td></tr><tr><td align="left">Black</td><td align="center">1713 (41.8%)</td><td align="center">70 (14.6%)</td><td align="center">1643 (45.4%)</td></tr><tr><td align="left">Latino</td><td align="center">986 (24.0%)</td><td align="center">339 (70.8%)</td><td align="center">647 (17.9%)</td></tr><tr><td align="left">Native</td><td align="center">39 (1.0%)</td><td align="center">4 (0.8%)</td><td align="center">35 (1.0%)</td></tr><tr><td align="left">Asian</td><td align="center">109 (2.7%)</td><td align="center">25 (5.2%)</td><td align="center">84 (2.3%)</td></tr><tr><td align="left">Mixed</td><td align="center">79 (1.9%)</td><td align="center">3 (0.6%)</td><td align="center">76 (2.1%)</td></tr><tr><td align="left">SPED</td><td align="center">694 (16.9%)</td><td align="center">91 (19.0%)</td><td align="center">603 (16.7%)</td></tr><tr><td align="left">Low‐SES</td><td align="center">3506 (85.5%)</td><td align="center">461 (96.2%)</td><td align="center">3045 (84.1%)</td></tr><tr><td align="left">Male</td><td align="center">2013 (49.1%)</td><td align="center">257 (53.7%)</td><td align="center">1756 (48.5%)</td></tr><tr><td align="left">Female</td><td align="center">2087 (50.9%)</td><td align="center">222 (46.3%)</td><td align="center">1865 (51.5%)</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note:</emph> Values represent number and percentage of students within each subgroup. Percentages are column percentages, except in the Total row where EL and EP are percentages of the full sample.</p> <hd id="AN0193225963-8">Measures</hd> <p>Students were assessed at six time points, at the beginning and end of each grade: fourth, fifth, and sixth. We analyzed "wave" as the measured time point in a student's developmental trajectory (e.g., wave 1 represents all scores from students at the beginning of fourth grade, regardless of the year they were assessed). To control for cohort effects, we included a "cohort" covariate for the year students entered the study. Trained administrators delivered the Core Analytical Language Skills (CALS) instrument—which included a connectives subtest—and the GISA reading comprehension assessment, both in English, as part of students' regular school day. <emph>z</emph>‐scores were used for both measures in the analysis, though we present raw scores in the descriptives for interpretability. To standardize the results to <emph>z</emph>‐scores, each outcome variable (GISA and CALS scores) at each time point was centered at its mean and scaled by its standard deviation, resulting in scaled variables per time point with a mean of 0 and a standard deviation of 1.</p> <hd id="AN0193225963-9">Connectives Subtest (α = 0.72)</hd> <p>The CALS instrument measures analytical language proficiency in grades 4–8. CALS refer to a constellation of high‐utility language skills that are prevalent in analytical texts across content areas but infrequently found in colloquial discourse. Although the CALS instrument has historically been analyzed unidimensionally, prior psychometric work has demonstrated that a multifactor higher‐order model provides the best fit, as CALS domain‐specific component skills are "distinguishable but interrelated" (Barr et al. [<reflink idref="bib3" id="ref46">3</reflink>], 1006). Since the higher‐order model correlated in excess of 0.94 with a unidimensional Rasch model, the Rasch model was recommended for practical (e.g., ease of scoring) and theoretical reasons, though author and colleagues recommended "that researchers use this higher‐order structure when investigating complex literacy relations, particularly involving relations of the academic language domains with other literacy constructs" (Barr et al. [<reflink idref="bib3" id="ref47">3</reflink>], 1001). While connective knowledge is a related construct to broader analytic language skills, prior work has established its unique role in reading comprehension, beyond other skills such as vocabulary and morphosyntactic knowledge (Cain and Nash [<reflink idref="bib5" id="ref48">5</reflink>]; Crosson and Lesaux [<reflink idref="bib6" id="ref49">6</reflink>]). Taken together, the acceptable internal consistency of the connective knowledge subtest (<emph>α</emph> = 0.72), previous dimensionality analyses, and empirical precedent justify the use of the subtest for an examination of the contribution of receptive connective knowledge to reading comprehension.</p> <p>The Connecting Ideas Logically subtest of the CALS used in this study was a 10‐min, paper‐and‐pencil test with 10 multiple‐choice items which students read and completed independently, each scored 1 for correct and 0 for incorrect or unanswered (see Figure 1 for a sample item). Students' raw scores at each time point ranged from 0 to 10. This subtest measures receptive knowledge of connectives—single words or expressions—frequently used in school texts across content areas, but typically rarely used in colloquial talk among upper elementary and middle school students (e.g., <emph>therefore, nevertheless, in conclusion</emph>) (Halliday [<reflink idref="bib15" id="ref50">15</reflink>]; Hyland [<reflink idref="bib18" id="ref51">18</reflink>]; Heath [<reflink idref="bib17" id="ref52">17</reflink>]). In addition to their prevalence and utility in texts across content areas, connectives were selected to represent: (<reflink idref="bib1" id="ref53">1</reflink>) a range of conceptual relations and discourse functions (e.g., contrastive, causal; transition words, or conclusion markers); and (<reflink idref="bib2" id="ref54">2</reflink>) a range of word frequency levels, in other words, a range of hypothesized ages at which various connectives were expected to be known by students on the basis of prior large word frequency inventories (Biemiller [<reflink idref="bib4" id="ref55">4</reflink>]; Dale and O'Rourke [<reflink idref="bib8" id="ref56">8</reflink>]).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-fig-0001.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-fig-0001.jpg" title="1 Connective knowledge sample item." /> </p> <p></p> <p>Following the general CALS design criteria, items were developed to minimize the impact of decoding skills, vocabulary and background knowledge, and higher‐order inferencing skills by making sure that (a) words around the targeted tested connective were transparently decodable and, to the extent possible, expected to be known by fourth graders (as documented by Biemiller [<reflink idref="bib4" id="ref57">4</reflink>]); (b) content and themes were highly familiar to students by virtue of going to school; and (c) the items did not involve higher‐order cognitive skills but instead focused on comprehension of language at the textbase level (Kintsch [<reflink idref="bib21" id="ref58">21</reflink>]). For this study, we acknowledge the limitation of not having collected participants' word reading or fluency data. While we cannot include such covariates in our present study, prior CALS studies have shown CALS to be predictive of reading comprehension even after controlling for code‐level skills and vocabulary knowledge (Barr et al. [<reflink idref="bib3" id="ref59">3</reflink>]; Uccelli, Galloway, et al. [<reflink idref="bib36" id="ref60">36</reflink>]).</p> <hd id="AN0193225963-11">Reading Comprehension</hd> <p>The Global Integrated Scenario‐based Assessment (GISA) administered to students is a computer‐based assessment developed by the Educational Testing Service (ETS). GISA is a reading comprehension assessment that "includes literal, inferential, and textual evidence integration questions based on informational/explanatory passages. Research on the GISA assessment has yielded adequate psychometric properties (i.e., internal consistency: <emph>α</emph> reliability = 0.89; split‐half reliability = 0.76)" (Sabatini et al. [<reflink idref="bib31" id="ref61">31</reflink>]). During the GISA, students are given a purposeful reading task (e.g., deciding whether a wind farm is good for their community). They then read and integrate multiple related texts on a topic to answer comprehension questions. The GISA consists of various forms that include a range of topics to avoid testing effects. As GISA exists in multiple forms, results are reported as a single score of reading comprehension based on a cross‐form scale constructed and validated by the ETS research team (O'Reilly et al. [<reflink idref="bib27" id="ref62">27</reflink>]).</p> <hd id="AN0193225963-12">Covariates</hd> <p>We included the following student‐level covariates: participants' family sociodemographic status, as indexed by free‐or‐reduced‐lunch eligibility (FRL) (yes = 1, no = 0); English proficiency designation (EL‐yes = 1, EL‐no = 0); gender (female = 1, male = 0); race/ethnicity (White, Black, Latino, Asian, Native, Mixed); special education designation (SPED‐yes = 1, SPED‐no = 0); district; and cohort (the year the student entered the study: 1, 2, or 3). Race/ethnicity, district, and cohort were treated as dummy variables with White as the reference category for race/ethnicity, District 1 as the reference category for district, and Cohort 1 (participant entered the study in year 1) as the reference category for cohort. We used students' covariate data from their first data collection point and treated all covariates as time invariant.</p> <hd id="AN0193225963-13">Analytic Approach</hd> <p>To address the research questions, we conducted parallel process latent growth modeling using the lavaan package in R (Rosseel [<reflink idref="bib30" id="ref63">30</reflink>]). Parallel process latent growth modeling is a statistical technique that examines the interrelatedness of multiple variables over time. The path model guiding the current analysis is presented in Figure 2 which shows the observed variables for connective knowledge and reading comprehension at each time point, latent variables reflecting the intercepts and slopes (i.e., trajectories of growth/change) for these two sets of variables, the covariance between initial levels of both variables (Path A), autoregressive paths between the latent intercept and slope of each variable (Paths B and C), reading comprehension slope regressed on connective slope (Path D), and the associations between the intercepts and growth trajectories across variables (Paths E and F). The previously discussed covariates were each used as exogenous time‐invariant predictors of the latent intercepts and slopes for both connective knowledge and reading comprehension. To account for nesting in our analysis, we include district as a fixed effect. We recognize that students are also nested in schools; however, the relatively small number of school clusters (<emph>n</emph> = 25) limits the stability of multilevel estimation (Maas and Hox [<reflink idref="bib24" id="ref64">24</reflink>]; McNeish and Stapleton [<reflink idref="bib25" id="ref65">25</reflink>]). Thus, we proceeded without estimating school‐level effects, instead relying on student‐level covariates and district as a fixed effect to adjust for group differences.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-fig-0002.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-fig-0002.jpg" title="2 Full sample path model." /> </p> <p></p> <p>Before answering our research questions, we modeled growth in each outcome separately to determine the appropriate growth curve for each variable. All intercept loadings for the observed variables were fixed to 1, following standard estimation procedures. To represent years elapsed since initial measurement at the start of fourth grade, slope terms were fixed to 0, 0.75, 1, 1.75, 2, and 2.75 (in months: 0, 9, 12, 21, 24, and 33) for the six waves of measurement. We considered well‐fitting models to have a comparative fit index (CFI) and Tucker–Lewis index (TLI) above 0.90, and a root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR) below 0.08 (Kline [<reflink idref="bib22" id="ref66">22</reflink>]).</p> <p>Students were present and had valid data on both measures for approximately 2.4 out of 6 time points on average (see Table S1 for details on patterns of missingness for the full sample, by covariate, and by study year). To address missing data, Full Information Maximum Likelihood (FIML) was selected as the estimator for all models. This is based on the assumption that the data are missing at random. Missing data is largely an artifact of study data collection patterns and cohort differences, not selective absenteeism or dropout in particular student groups, and we include cohort as a control in our analysis. Conditioning our model on the covariates described in the previous section helps reduce potential bias in the growth parameter estimates by student‐level factors and strengthens the missing at random assumption. Additionally, we conducted a sensitivity check with 50 iterations of Multiple Imputation to validate the robustness of our paths of interest. Although the full conditional model did not converge under MI, we compared results of the unconditional model under MI and our FIML model across parameters of interest, and the overall magnitude and direction remained stable across approaches. MI produced inflated standard errors and <emph>p</emph>‐values, which is not uncommon in complex models, and we proceeded with FIML as the estimator in our analysis (Graham et al. [<reflink idref="bib14" id="ref67">14</reflink>]). Importantly, we observed that missingness was not significantly different for the ELs versus EP students; students in both EL and EP groups were present, on average, for approximately 2.4 time points, and the distribution of missingness by English proficiency designation across waves was similar as well. There was no missing covariate data, and patterns of missingness were identical by outcome variable.</p> <p>There were two particular paths of interest for the present study. To answer RQ1, we analyzed the association between connective knowledge intercept and reading comprehension slope (Figure 2, Path F). Through this path, we examined if students' connective knowledge at the start of fourth grade predicted their growth in reading comprehension between grades 4 and 6. To answer RQ2, we analyzed the association between connective knowledge slope and reading comprehension slope to gain insight into whether growth in connective knowledge would predict growth in reading comprehension (Figure 2, Path D). First, we analyzed the aforementioned associations for the entire sample. Then, to answer RQ3, we used multigroup analysis to compare the slope association across the EL and EP groups.</p> <hd id="AN0193225963-15">Results</hd> <p></p> <hd id="AN0193225963-16">Exploring Connective Knowledge and Reading Comprehension</hd> <p>Observed data suggested linear growth for both connective knowledge and reading comprehension (Figures 3 and 4). As shown in Tables 2 and 3, means and standard deviations for both measures also indicate a steady growth by grade, on average, despite considerable within‐grade variability. EL and EP students seem to grow steadily over time on both measures, though the mean scores for ELs are about 0.5 standard deviations below EP students at baseline. Given the steady growth curves of observed means, linear models were fit to the data. Since both linear models demonstrated good fit (Connectives: CFI = 0.987, TLI = 0.982, RMSEA = 0.037, SRMR = 0.026; Reading Comprehension: CFI = 0.981, TLI = 0.972, RMSEA = 0.024, SRMR = 0.022), we conducted our parallel process analysis using linear growth modeling. Our parallel growth model had good fit parameters (CFI = 0.987; TLI = 0.982; RMSEA = 0.016; SRMR = 0.028), and we proceeded with our analysis.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-fig-0003.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-fig-0003.jpg" title="3 Connective knowledge scores over time in grades 4–6." /> </p> <p></p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-fig-0004.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-fig-0004.jpg" title="4 Reading comprehension scores over time in grades 4–6." /> </p> <p></p> <p>2 TABLE Raw connective knowledge assessment averages by language group over time.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left">Time</th><th align="center">Connectives raw score: full sample, mean (SD)</th><th align="center">Connectives raw score: English learners, mean (SD)</th><th align="center">Connectives raw score: English proficient, mean (SD)</th></tr></thead><tbody valign="top"><tr><td align="left">Grade 4—BOY</td><td align="center">3.02 (2.06)</td><td align="center">2.13 (1.54)</td><td align="center">3.11 (2.09)</td></tr><tr><td align="left">Grade 4—EOY</td><td align="center">4.31 (2.58)</td><td align="center">3.26 (2.07)</td><td align="center">4.45 (2.61)</td></tr><tr><td align="left">Grade 5—BOY</td><td align="center">4.35 (2.43)</td><td align="center">3.27 (1.89)</td><td align="center">4.50 (2.45)</td></tr><tr><td align="left">Grade 5—EOY</td><td align="center">5.38 (2.61)</td><td align="center">4.10 (2.36)</td><td align="center">5.54 (2.60)</td></tr><tr><td align="left">Grade 6—BOY</td><td align="center">5.57 (2.69)</td><td align="center">4.34 (2.28)</td><td align="center">5.80 (2.70)</td></tr><tr><td align="left">Grade 6—EOY</td><td align="center">6.23 (2.67)</td><td align="center">4.57 (2.49)</td><td align="center">6.59 (2.57)</td></tr></tbody></table> </ephtml> </p> <ulist> <item>2 Abbreviations: BOY = beginning of year, EOY = end of year.</item> <item>3 TABLE Raw reading comprehension assessment averages by language group over time.</item> </ulist> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left">Time</th><th align="center">RC raw score: full sample, mean (SD)</th><th align="center">RC raw score: English learners, mean (SD)</th><th align="center">RC raw score: English proficient, mean (SD)</th></tr></thead><tbody valign="top"><tr><td align="left">Grade 4—BOY</td><td align="center">944.84 (51.80)</td><td align="center">929.18 (42.61)</td><td align="center">946.74 (52.35)</td></tr><tr><td align="left">Grade 4—EOY</td><td align="center">953.77 (57.41)</td><td align="center">933.56 (46.02)</td><td align="center">956.29 (58.22)</td></tr><tr><td align="left">Grade 5—BOY</td><td align="center">967.17 (57.98)</td><td align="center">942.89 (44.10)</td><td align="center">970.31 (58.87)</td></tr><tr><td align="left">Grade 5—EOY</td><td align="center">970.90 (62.63)</td><td align="center">939.56 (49.89)</td><td align="center">974.65 (62.90)</td></tr><tr><td align="left">Grade 6—BOY</td><td align="center">980.22 (79.31)</td><td align="center">946.59 (59.96)</td><td align="center">986.96 (80.65)</td></tr><tr><td align="left">Grade 6—EOY</td><td align="center">1000.06 (82.15)</td><td align="center">961.35 (65.19)</td><td align="center">1009.20 (82.53)</td></tr></tbody></table> </ephtml> </p> <p>3 Abbreviations: BOY = beginning of year, EOY = end of year.</p> <p>Since both connective knowledge and reading comprehension were standardized within wave and, thus, had wave means set to zero, latent intercept and slope means are not interpretable as absolute growth parameters (e.g., the latent intercept reflects the reference group's relative standing compared to the sample mean at baseline, not an interpretable baseline score of connective knowledge). As such, we highlight paths and variances in our analysis as relevant to our research questions and interpretable using <emph>z</emph>‐scores (Table 4). Though not a central research question for this study, Path A indicates a significant covariance between connective knowledge latent intercept and reading comprehension latent intercept (<emph>β</emph> = 0.36, SE = 0.18, <emph>p</emph> < 0.001). As expected based on previous cross‐sectional studies, these constructs are strongly predictive of each other at baseline. Path E shows reading comprehension intercept significantly predicting growth in connective knowledge (<emph>β</emph> = 0.16, SE = 0.05, <emph>p</emph> = 0.001), indicating that students with higher reading comprehension at the start of fourth grade have, on average, higher rates of growth in connective knowledge through grade six. In the sections that follow, we describe the paths that are central to our research questions investigating how connective knowledge relates to reading comprehension in grades 4–6.</p> <p>4 TABLE Parameter estimates for the parallel process latent growth model for connective knowledge and reading comprehension.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left" /><th align="center">Unstandardized coefficient (SE)</th></tr></thead><tbody valign="top"><tr><td align="left">Associations between latent variables</td></tr><tr><td align="left">Path A</td><td align="center">0.36 (0.18)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Path B</td><td align="center">−0.19 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Path C</td><td align="center">−0.13 (0.06)<xref ref-type="fn" rid="tfn5" /></td></tr><tr><td align="left">Path D</td><td align="center">0.30 (0.10)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Path E</td><td align="center">0.16 (0.05)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Path F</td><td align="center">0.12 (0.05)<xref ref-type="fn" rid="tfn5" /></td></tr><tr><td align="left">Variances</td></tr><tr><td align="left">Connective knowledge intercept</td><td align="center">0.50 (0.03)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Connective knowledge slope</td><td align="center">0.03 (0.01)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">RC intercept</td><td align="center">0.46 (0.02)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">RC slope</td><td align="center">0.01 (0.01)<xref ref-type="fn" rid="tfn5" /></td></tr><tr><td align="left">Covariates predicting connective knowledge intercept</td></tr><tr><td align="left">Free/reduced lunch</td><td align="center">−0.30 (0.06)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Black</td><td align="center">−0.29 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Latinx</td><td align="center">−0.18 (0.06)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Native</td><td align="center">−0.13 (0.19)</td></tr><tr><td align="left">Asian</td><td align="center">0.14 (0.11)</td></tr><tr><td align="left">Mixed race</td><td align="center">−0.13 (0.11)</td></tr><tr><td align="left">Gender (female = 1)</td><td align="center">−0.09 (0.03)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">English learner</td><td align="center">−0.48 (0.06)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Special education</td><td align="center">−0.45 (0.04)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">District ID 2</td><td align="center">0.08 (0.08)</td></tr><tr><td align="left">District ID 3</td><td align="center">−0.24 (0.07)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">District ID 4</td><td align="center">−0.25 (0.07)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Cohort 2</td><td align="center">−0.24 (0.08)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Cohort 3</td><td align="center">−0.24 (0.08)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Covariates predicting RC intercept</td></tr><tr><td align="left">Free/reduced lunch</td><td align="center">−0.26 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Black</td><td align="center">−0.37 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Latinx</td><td align="center">−0.19 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Native</td><td align="center">−0.06 (0.17)</td></tr><tr><td align="left">Asian</td><td align="center">0.08 (0.11)</td></tr><tr><td align="left">Mixed race</td><td align="center">−0.36 (0.11)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">Gender (female = 1)</td><td align="center">0.02 (0.03)</td></tr><tr><td align="left">English learner</td><td align="center">−0.47 (0.05)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Special education</td><td align="center">−0.68 (0.04)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">District ID 2</td><td align="center">0.23 (0.07)<xref ref-type="fn" rid="tfn6" /></td></tr><tr><td align="left">District ID 3</td><td align="center">−0.03 (0.07)</td></tr><tr><td align="left">District ID 4</td><td align="center">−0.42 (0.06)<xref ref-type="fn" rid="tfn7" /></td></tr><tr><td align="left">Cohort 2</td><td align="center">−0.08 (0.07)</td></tr><tr><td align="left">Cohort 3</td><td align="center">−0.12 (0.07)</td></tr></tbody></table> </ephtml> </p> <ulist> <item>4 <emph>Note:</emph><emph>n</emph> = 4100. Slope covariates were omitted from the table; only a small number reached significance.</item> <item>5 * <emph>p</emph> < 0.05.</item> <item>6 ** <emph>p</emph> < 0.01.</item> <item>7 *** <emph>p</emph> < 0.001.</item> </ulist> <hd id="AN0193225963-19">Connective Knowledge and Growth in Reading Comprehension</hd> <p>To answer our first research question, we analyzed the path between the latent connective knowledge intercept and latent reading comprehension slope (Figure 2, Path F). To test whether this path significantly improved model fit, we compared nested models with the Satorra‐Bentler scaled chi‐square difference test (Satorra and Bentler [<reflink idref="bib32" id="ref68">32</reflink>]). Removing Path F significantly worsened model fit (chi‐square difference = 43.78, df = 1, <emph>p</emph> < 0.001), indicating that regressing reading comprehension slope on connective intercept meaningfully contributes to the model. Results from Path F indicated that the latent connective knowledge intercept and latent reading comprehension slope had a positive association (<emph>β</emph> = 0.119, SE = 0.046, <emph>p</emph> = 0.010). In other words, students with greater connective knowledge at the start of fourth grade, on average, displayed a higher growth rate in reading comprehension through grade six. Every one standard deviation increase in connective knowledge scores at the start of fourth grade was associated with a 0.12 SD increase in the growth rate of a student's reading comprehension scores between grades 4 and 6.</p> <p>To answer our second research question, we analyzed the path between the latent slopes for connective knowledge and reading comprehension (Figure 2, Path D). As with Path F, we compared nested models with the Satorra‐Bentler scaled chi‐square difference test to determine if this path improved model fit (Satorra and Bentler [<reflink idref="bib32" id="ref69">32</reflink>]). Removing Path D significantly worsened model fit (chi‐square difference = 43.75, df = 1, <emph>p</emph> < 0.001); thus, regressing reading comprehension slope on connective slope meaningfully contributes to the model. Path D displayed a positive predictive association between connective knowledge growth and reading comprehension growth between grades 4 and 6 (<emph>β</emph> = 0.297, SE = 0.100, <emph>p</emph> = 0.003). Based on the regression coefficient, a one SD increase in connective knowledge slope was associated with a 0.30 SD increase in reading comprehension slope. In other words, on average, students whose knowledge of connectives grew at a faster rate also grew faster in their reading comprehension.</p> <hd id="AN0193225963-20">Connective Knowledge and Reading Comprehension by English Proficiency Designation</hd> <p>To answer our third research question, we conducted a multigroup analysis of English learners and English proficient students and compared the parameters from Path D between groups. By dividing the sample into groups, the variability of the latent variables within each group unsurprisingly decreased, particularly for slope factors. As such, we focus on the cross‐domain slope relation in this study—rather than additionally including cross‐domain intercept‐slope paths, as we did in the full model—in order to obtain more stable and interpretable multigroup results. To account for the limited sample size of the English learner group, we made key adjustments to the multigroup model as compared to the full sample analysis to reduce modeling error while maintaining the necessary elements to investigate RQ3 (Figure 5). Covariates were removed from the slope terms to prevent overspecification and model misfit, as in the full group model they were mostly non‐significant. We additionally excluded intercept covariates with less than 3% representation in the sample (Native 1%, mixed 1.9%, and Asian 2.7%) and only included free and reduced lunch eligibility as an intercept covariate for English proficient students, given the limited variability in this covariate for English learners in our sample (96.2% eligible). Cohort membership was collinear with district in the EL student sample, resulting in unstable estimates and convergence issues, so we maintained district fixed effects and removed cohort, since district was a more significant predictor of difference in our full sample model. For the multigroup analysis, we first ran two models: a model in which the regression and intercept parameters were constrained across groups (EL/EP) and another model in which those parameters were free across groups. A chi‐squared difference test validated that the parameters were significantly different (<emph>p</emph> < 0.001) across groups. The final multigroup model demonstrated good fit (CFI = 0.975; TLI = 0.971; RMSEA = 0.024; SRMR = 0.044), and we proceeded with specifically comparing our parameters of interest.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/NRNU/02apr26/rrq70096-fig-0005.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="rrq70096-fig-0005.jpg" title="5 Multigroup path model." /> </p> <p></p> <p>The parameters from Path D, the contribution of the connective knowledge slope to reading comprehension slope, were significant for the EL group and the EP group over grades 4–6 (Table 5). To answer RQ3, we compared this parameter across groups to determine if the magnitude of the relation differed by student language background. To do this, we conducted a Wald test to analyze if the parameters from Path D were significantly different in ELs versus EP students. The coefficient for Path D was stronger for ELs (0.799) than EP students (0.507), though this difference was statistically non‐significant (<emph>p</emph> = 0.310). While the path is significant for both EL (<emph>p</emph> = 0.001) and EP (<emph>p</emph> = 0.003) student groups, we do not have evidence that the coefficients are significantly different between groups for Path D. On average, a higher growth rate in connective knowledge significantly predicted a higher growth rate in reading comprehension for EL and EP student groups alike, and this relation was not significantly different by group in our sample.</p> <p>5 TABLE Parameter estimates for the multigroup parallel process latent growth model for connective knowledge and reading comprehension by English learner designation.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left" /><th align="center">English proficient</th><th align="center">English learner</th></tr></thead><tbody valign="top"><tr><td align="left">Associations between latent variables</td></tr><tr><td align="left">Path A</td><td align="center">0.38 (0.02)<xref ref-type="fn" rid="tfn11" /></td><td align="center">0.15 (0.04)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Path B</td><td align="center">0.02 (0.02)</td><td align="center">0.19 (0.23)</td></tr><tr><td align="left">Path C</td><td align="center">−0.02 (0.01)</td><td align="center">−0.21 (0.07)<xref ref-type="fn" rid="tfn10" /></td></tr><tr><td align="left">Path D</td><td align="center">0.51 (0.17)<xref ref-type="fn" rid="tfn10" /></td><td align="center">0.80 (0.23)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Variances</td></tr><tr><td align="left">Connective knowledge intercept</td><td align="center">0.46 (0.02)<xref ref-type="fn" rid="tfn11" /></td><td align="center">0.15 (0.07)<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left">Connective knowledge slope</td><td align="center">0.02 (0.01)<xref ref-type="fn" rid="tfn10" /></td><td align="center">0.03 (0.02)</td></tr><tr><td align="left">RC intercept</td><td align="center">0.46 (0.02)<xref ref-type="fn" rid="tfn11" /></td><td align="center">0.33 (0.05)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">RC slope</td><td align="center">0.01 (0.00)</td><td align="center">0.02 (0.01)<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left">Covariates predicting connective knowledge intercept</td></tr><tr><td align="left">Free/reduced lunch</td><td align="center">−0.30 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">0.00 (0.00)</td></tr><tr><td align="left">Black</td><td align="center">−0.30 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.28 (0.12)<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left">Latinx</td><td align="center">−0.21 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.22 (0.09)<xref ref-type="fn" rid="tfn9" /></td></tr><tr><td align="left">Gender (female = 1)</td><td align="center">−0.07 (0.03)<xref ref-type="fn" rid="tfn10" /></td><td align="center">0.03 (0.05)</td></tr><tr><td align="left">District ID 2</td><td align="center">0.02 (0.06)</td><td align="center">−0.09 (0.10)</td></tr><tr><td align="left">District ID 3</td><td align="center">−0.28 (0.06)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.30 (0.11)<xref ref-type="fn" rid="tfn10" /></td></tr><tr><td align="left">District ID 4</td><td align="center">−0.45 (0.06)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.33 (0.09)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Special education</td><td align="center">−0.62 (0.03)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.27 (0.08)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Covariates predicting RC intercept</td></tr><tr><td align="left">Free/reduced lunch</td><td align="center">−0.26 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">0.00 (0.00)</td></tr><tr><td align="left">Black</td><td align="center">−0.34 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.58 (0.15)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Latinx</td><td align="center">−0.18 (0.04)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.42 (0.13)<xref ref-type="fn" rid="tfn11" /></td></tr><tr><td align="left">Gender (female = 1)</td><td align="center">0.09 (0.03)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.05 (0.08)</td></tr><tr><td align="left">District ID 2</td><td align="center">0.20 (0.07)<xref ref-type="fn" rid="tfn10" /></td><td align="center">0.25 (0.14)</td></tr><tr><td align="left">District ID 3</td><td align="center">−0.01 (0.06)</td><td align="center">−0.14 (0.12)</td></tr><tr><td align="left">District ID 4</td><td align="center">−0.51 (0.06)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.32 (0.12)<xref ref-type="fn" rid="tfn10" /></td></tr><tr><td align="left">Special education</td><td align="center">−0.67 (0.03)<xref ref-type="fn" rid="tfn11" /></td><td align="center">−0.33 (0.10)<xref ref-type="fn" rid="tfn11" /></td></tr></tbody></table> </ephtml> </p> <ulist> <item>8 Abbreviations: EL = English learner, EP = English proficient.</item> <item>9 * <emph>p</emph> < 0.05.</item> <item>10 ** <emph>p</emph> < 0.01.</item> <item>11 *** <emph>p</emph> < 0.001.</item> </ulist> <hd id="AN0193225963-22">Discussion</hd> <p>In this study, we sought to extend the existing literature on the relation between connective knowledge and reading comprehension by examining their development on a three‐year longitudinal sample of 4100 fourth–sixth graders attending urban public schools in the U.S. Using growth analysis, we examined (<reflink idref="bib1" id="ref70">1</reflink>) if connective knowledge at the start of fourth grade predicted growth in reading comprehension through grade six; and (<reflink idref="bib2" id="ref71">2</reflink>) if growth in connective knowledge predicted growth in reading comprehension between grades 4 and 6. We also examined potential differences in growth rate relations by English proficiency designation, comparing the trajectories of English learners with those of their English proficient peers.</p> <p>Our study has three main findings with implications for language and literacy development research and practice. First, students with greater connective knowledge at the start of fourth grade displayed, on average, greater growth in reading comprehension between grades 4 and 6. Second, greater growth in connective knowledge across this same timespan predicted, on average, greater growth in reading comprehension. Third, our study found that the relation between growth in connective knowledge and growth in reading comprehension, despite slight variations in the magnitude, was not significantly different for English learners as compared to their English proficient peers. Taken together, these findings revealed that between grades 4 and 6 both English proficient students and students designated as English learners displayed variation in their knowledge of connectives that were positively associated with reading comprehension.</p> <p>Our analyses confirm the well‐established finding that connective knowledge contributes to reading comprehension (Cain and Nash [<reflink idref="bib5" id="ref72">5</reflink>]; Crosson and Lesaux [<reflink idref="bib6" id="ref73">6</reflink>]; Townsend et al. [<reflink idref="bib34" id="ref74">34</reflink>]), and we extend the existing scientific literature on connectives and reading comprehension in several ways. Our findings echo and extend Volodina and Weinert's ([<reflink idref="bib38" id="ref75">38</reflink>]) work with German primary school students and Cain and Nash's ([<reflink idref="bib5" id="ref76">5</reflink>]) cross‐sectional data with British English‐speaking 8‐ and 10‐year‐olds: connective knowledge development is an ongoing, incomplete process in young readers. We found that in our U.S. sample of English learners and English proficient students, connective knowledge continues developing through sixth grade (and, we can assume, beyond).</p> <p>Additionally, we found that students' initial levels of connective knowledge (start of grade 4) are predictive of their growth in reading comprehension through grade 6. This is novel evidence that connective knowledge does more for learners than facilitate reading comprehension at a specific time point; it predicts growth in text comprehension longitudinally (in this case, through grade 6). Especially considering the large proportions of upper elementary and middle‐grade students in the U.S. who are still unfamiliar with many connectives that are prevalent in texts (Barr et al. [<reflink idref="bib3" id="ref77">3</reflink>]; Andreev and Uccelli [<reflink idref="bib1" id="ref78">1</reflink>]), this finding encourages future studies to explore the potential impact of developing students' connective knowledge before and into the upper elementary grades, perhaps especially for struggling readers, and how acquiring earlier connective knowledge could contribute to closing gaps in reading comprehension over time.</p> <p>Particularly insightful for its instructional implications is the finding that reveals that higher growth in connective knowledge was significantly associated with higher growth in reading comprehension. Given that prior research has shown that knowledge of connectives is malleable through targeted intervention for young children and older adolescents, these findings suggest that a focus on expanding knowledge of connectives for early and mid‐adolescent students may be a promising complementary approach to other evidence‐based reading comprehension instructional practices. Moreover, given that connectives are a relatively closed class of expressions widely used across content area texts, connective knowledge may be a high‐leverage area of reading comprehension intervention for both English learners and English proficient students in grades 4–6.</p> <p>The relation between connective knowledge growth and reading comprehension growth was replicated in both the EL and EP student groups based on a multigroup analysis by language background. As expected, based on previous studies conducted with samples of second language learners and proficient speakers, our results demonstrate a significant contribution of connective knowledge to reading comprehension for both EL and EP students (Crosson et al. [<reflink idref="bib7" id="ref79">7</reflink>]; Fraser et al. [<reflink idref="bib10" id="ref80">10</reflink>]). Interestingly, however, we did not find evidence that the magnitude of this relation differs by English proficiency designation. These findings contribute to the empirical disagreements about how language background interacts with connective knowledge and reading comprehension. Our results contrast with previous studies showing that connective knowledge contributes less to reading comprehension for language learners than their proficient peers (e.g., Crosson and Lesaux [<reflink idref="bib6" id="ref81">6</reflink>]; Kohnen and Retelsdorf [<reflink idref="bib23" id="ref82">23</reflink>]), and our results align with Welie et al.'s ([<reflink idref="bib39" id="ref83">39</reflink>]) study of Dutch eighth graders, which found no significant difference in the magnitude of the association between connective knowledge to reading comprehension between monolingual or bilingual students.</p> <p>The conflicting evidence on learner language background points to the need for studies that analyze language background on connective knowledge and reading comprehension in greater detail. It is likely that the binary category of "second language learner" does not capture sufficient nuance about the students' language abilities to provide clear findings about how connective knowledge contributes to reading comprehension differently by language background. Future studies should gather large and diverse enough samples to analyze the bilingual/s language learners by proficiency level. Welie and colleagues' sample of bilingual Dutch students, for example, performed much closer to the monolingual students on a vocabulary assessment than did Crosson & Lesaux's language learners, and they hypothesize that this difference may be a contributing factor to the difference of findings. Our own data seem to support this possibility: at baseline, the mean scores on both measures—connectives knowledge and reading comprehension—were only about 0.5 standard deviations lower for ELs than EP students, reflecting much more closely to Welie et al.'s sample than Crosson & Lesaux. Additionally, our connective knowledge measure examines connectives that are frequently encountered in text, while prior studies generally focused on a wider array of connectives. Given our focus on specific connectives that are prevalent in text, students who struggle with reading comprehension (both EL and EP) might perform more similarly on this connective knowledge measure than on measures with less text‐prevalent subsets of connectives.</p> <p>Finally, there may be additional influencing factors to connective knowledge and reading comprehension that are confounded with language background. On one hand, Volodina and Weinert ([<reflink idref="bib38" id="ref84">38</reflink>]) found that SES contributed more to connective knowledge than language background for German primary school students, and they suggest that future work in this area include socioeconomically diverse samples. On the other hand, Kohnen and Retelsdorf ([<reflink idref="bib23" id="ref85">23</reflink>]) replicated Crosson & Lesaux's findings in monolingual and bilingual German ninth graders even after controlling for SES. Our sample is much larger than other studies on connectives and reading comprehension (4100 students compared to just a few hundred in previous work), and we controlled for SES in our full sample analyses; however, the English learner group lacked sufficient socioeconomic diversity to control for SES in our multigroup model, as 96.2% of ELs in our sample were considered low‐SES. In short, future studies should include large samples of English proficient students and second language learners with diverse proficiency levels and socioeconomic backgrounds to more specifically analyze the contributions of each influence on connective knowledge and reading comprehension.</p> <hd id="AN0193225963-23">Limitations and Future Implications</hd> <p>This study offers results based on a large and heterogeneous sample of public‐school students enrolled in grades 4 to 6, yet it is not without limitations. Given that our sample size for English learners is relatively small (<emph>n</emph> = 479) and that we do not have information on ELs' English proficiency levels, additional studies are needed to analyze connective knowledge and reading comprehension with a larger sample of ELs and to examine how this relation may be affected by students' English language development level. There may be a minimal threshold of English language proficiency for connective knowledge to aid students in their comprehension of text (i.e., if ELs are still learning to decode English text or have limited English vocabulary development, understanding connectives is unlikely to improve their comprehension). Larger and more diverse samples across a higher number of districts could employ a multilevel framework to analyze variance by context, such as district‐level variables and demographic makeup. The present study analyzes single measures of receptive connective knowledge and reading comprehension, and future studies should include additional measures of these constructs to further strengthen evidence for the unique contribution of connective knowledge to reading comprehension. Relatedly, future studies should add decoding and fluency measures, as well as vocabulary and syntactic knowledge, to understand the relation between other literacy skills (beyond reading comprehension) and the results presented here.</p> <p>Furthermore, students designated as English proficient represent a variety of language backgrounds (English‐only students, multilingual students fluent in English before school entry, former ELs) that future studies could disaggregate to better understand possible group differences undetected in this study, given that we did not have additional language background information in our sample. Additionally, intervention studies on connectives and reading comprehension are needed to determine causality and directionality, as this study presents only developmental associations but cannot rule out potential time‐varying covariates not captured in our data collection that influence these relations.</p> <p>Our findings reveal connective knowledge as an area of instructional consideration for upper elementary students' reading comprehension that is relevant for future research and reading instruction. More specifically, our results motivate future, intervention‐based research to identify a potential causal link behind this relation. Finally, this study calls practitioners' attention to an area of instructional focus that may move the needle on students' reading comprehension outcomes, a strongly identified area of growth for ELs and EP students alike in grades 4–6.</p> <hd id="AN0193225963-24">Acknowledgments</hd> <p>This research was supported by the Institute of Education Sciences (Grant No. R305A170185; Grant No. R305A190034 awarded to the Harvard Graduate School of Education), U.S. Department of Education. The opinions expressed are those of the authors and do not represent views of the Institute or the U.S. Department of Education.</p> <hd id="AN0193225963-25">Conflicts of Interest</hd> <p>The authors declare no conflicts of interest.</p> <hd id="AN0193225963-26">Data Availability Statement</hd> <p>The authors have nothing to report.</p> <p>GRAPH: Table S1: Data presence by subgroup and timepoint.</p> <ref id="AN0193225963-27"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref1" type="bt">1</bibl> <bibtext> In this manuscript we use the formal designations of "English learner" used by the schools at the time of data collection to refer to multilingual students who are still developing grade‐level English proficiency and "English proficient" to refer to those who are native English speakers, initially fluent English‐speaking multilinguals, and former English learners who have reached grade‐level English proficiency. However, we recognize the limitations of these formal designations, as they have a deficit framing and render students' other language resources invisible.</bibtext> </blist> </ref> <ref id="AN0193225963-28"> <title> References </title> <blist> <bibtext> Andreev, L., and P. Uccelli. 2024. "The Secret Life of Connectives: A Taxonomy to Study Individual Differences in Mid‐Adolescents' Use of Connectives in Writing to Persuade." Reading and Writing37, no. 1: 173–204. https://doi.org/10.1007/s11145‐023‐10425‐3.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref14" type="bt">2</bibl> <bibtext> August, D., P. Uccelli, L. Artzi, C. Barr, and D. J. Francis. 2021. "English Learners' Acquisition of Academic Vocabulary: Instruction Matters, but So Do Word Characteristics." Reading Research Quarterly56, no. 3: 559–582. https://doi.org/10.1002/rrq.323.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref13" type="bt">3</bibl> <bibtext> Barr, C. D., P. Uccelli, and E. Phillips Galloway. 2019. "Specifying the Academic Language Skills That Support Text Understanding in the Middle Grades: The Design and Validation of the Core Academic Language Skills Construct and Instrument." Language Learning69, no. 4: 978–1021. https://doi.org/10.1111/lang.12365.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref55" type="bt">4</bibl> <bibtext> Biemiller, A.2009. Words Worth Teaching. SRA/McGraw‐Hill.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref4" type="bt">5</bibl> <bibtext> Cain, K., and H. M. Nash. 2011. "The Influence of Connectives on Young Readers' Processing and Comprehension of Text." Journal of Educational Psychology103, no. 2: 429–441. https://doi.org/10.1037/a0022824.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref16" type="bt">6</bibl> <bibtext> Crosson, A. C., and N. K. Lesaux. 2013. "Does Knowledge of Connectives Play a Unique Role in the Reading Comprehension of English Learners and English‐Only Students?" Journal of Research in Reading36, no. 3: 241–260. https://doi.org/10.1111/j.1467‐9817.2011.01501.x.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref39" type="bt">7</bibl> <bibtext> Crosson, A. C., N. K. Lesaux, and M. Martiniello. 2008. "Factors That Influence Comprehension of Connectives Among Language Minority Children From Spanish‐Speaking Backgrounds." Applied PsychoLinguistics29, no. 4: 603–625. https://doi.org/10.1017/S0142716408080260.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref56" type="bt">8</bibl> <bibtext> Dale, E., and J. O'Rourke. 1981. The Living Word Vocabulary. World Book–Childcraft International.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref8" type="bt">9</bibl> <bibtext> ESSA. 2015. "Every Student Succeeds Act, 20 U.S.C. § 6301."</bibtext> </blist> <blist> <bibtext> Fraser, C., A. Pasquarella, E. Geva, A. Gottardo, and A. Biemiller. 2021. "English Language Learners' Comprehension of Logical Relationships in Expository Texts: Evidence for the Confluence of General Vocabulary and Text‐Connecting Functions." Language Learning71, no. 3: 872–906. https://doi.org/10.1111/lang.12453.</bibtext> </blist> <blist> <bibtext> Galloway, E. P., and P. Uccelli. 2019. "Examining Developmental Relations Between Core Academic Language Skills and Reading Comprehension for English Learners and Their Peers." Journal of Educational Psychology111, no. 1: 15–31. https://doi.org/10.1037/edu0000276.</bibtext> </blist> <blist> <bibtext> Galloway, E. P., and P. Uccelli. 2024. "Promoting Humanizing, Meaningful, and Just Language Instruction for Multilingual Learners and Their Peers: A Pedagogical Vision Illustrated by Examples From Practice." Linguistics and Education84: 101358. https://doi.org/10.1016/j.linged.2024.101358.</bibtext> </blist> <blist> <bibtext> Geva, E.1986. "Reading Comprehension in a Second Language: The Role of Conjunctions." TESL Canada Journal3: 85–96. https://doi.org/10.18806/tesl.v3i0.996.</bibtext> </blist> <blist> <bibtext> Graham, J. W., A. E. Olchowski, and T. D. Gilreath. 2007. "How Many Imputations Are Really Needed? Some Practical Clarifications of Multiple Imputation Theory." Prevention Science8, no. 3: 206–213. https://doi.org/10.1007/s11121‐007‐0070‐9.</bibtext> </blist> <blist> <bibtext> Halliday, M. A. K.2013. "Literacy and linguistics: A functional perspective." In Collected works of M. A. K. Halliday: Volume 9, edited by J. J. Webster, 339–371. Bloomsbury.</bibtext> </blist> <blist> <bibtext> Halliday, M. A. K., and C. Matthiessen. 2014. Halliday's Introduction to Functional Grammar. 4th ed. Routledge.</bibtext> </blist> <blist> <bibtext> Heath, S. B.2012. Words at Work and Play: Three Decades in Family and Community Life. Cambridge University Press.</bibtext> </blist> <blist> <bibtext> Hyland, K.2005. Metadiscourse: Exploring Interaction in Writing. Continuum.</bibtext> </blist> <blist> <bibtext> Ibáñez, R., F. Moncada, and B. Cárcamo. 2019. "Coherence Relations in Primary School Textbooks: Variation Across School Subjects." Discourse Processes56, no. 8: 764–785. https://doi.org/10.1080/0163853X.2019.1565278.</bibtext> </blist> <blist> <bibtext> Jones, S. M., M. LaRusso, J. Kim, et al. 2019. "Experimental Effects of Word Generation on Vocabulary, Academic Language, Perspective Taking, and Reading Comprehension in High‐Poverty Schools." Journal of Research on Educational Effectiveness12, no. 3: 448–483. https://doi.org/10.1080/19345747.2019.1615155.</bibtext> </blist> <blist> <bibtext> Kintsch, W.2004. "The Construction–Integration Model of Text Comprehension and Its Implications for Instruction." In Theoretical Models and Processes of Reading, edited by R. Ruddell and N. Unrau, 5th ed., 1270–1328. International Reading Association.</bibtext> </blist> <blist> <bibtext> Kline, R. B.2023. Principles and Practice of Structural Equation Modeling. Guilford Publications.</bibtext> </blist> <blist> <bibtext> Kohnen, N., and J. Retelsdorf. 2019. "The Role of Knowledge of Connectives in Comprehension of a German Narrative Text." Journal of Research in Reading42, no. 2: 371–388. https://doi.org/10.1111/1467‐9817.12273.</bibtext> </blist> <blist> <bibtext> Maas, C. J., and J. J. Hox. 2005. "Sufficient Sample Sizes for Multilevel Modeling." Methodology1: 86–92.</bibtext> </blist> <blist> <bibtext> McNeish, D. M., and L. M. Stapleton. 2016. "The Effect of Small Sample Size on Two‐Level Model Estimates: A Review and Illustration." Educational Psychology Review28, no. 2: 295–314. https://doi.org/10.1007/s10648‐014‐9287‐x.</bibtext> </blist> <blist> <bibtext> National Academies of Sciences, Engineering, and Medicine. 2017. "Promoting the educational success of children and youth learning English: Promising futures."https://doi.org/10.17226/24677.</bibtext> </blist> <blist> <bibtext> O'Reilly, T., J. Ieks, J. Sabatini, L. Halderman, and J. Steinberg. 2014. "Designing Reading Comprehension Assessments for Reading Interventions: How a Theoretically Motivated Assessment Can Serve as an Outcome Measure." Educational Psychology Review26: 403–424. https://doi.org/10.1007/s10648‐014‐9269‐z.</bibtext> </blist> <blist> <bibtext> Quílez, J.2021. "Supporting Spanish 11th Grade Students to Make Scientific Writing When Learning Chemistry in English: The Case of Logical Connectives." International Journal of Science Education43, no. 9: 1459–1482. https://doi.org/10.1080/09500693.2021.1918794.</bibtext> </blist> <blist> <bibtext> Román, D. X., A. Briceño, H. Rohde, and S. Hironaka. 2016. "Linguistic Cohesion in Middle‐School Texts: A Comparison of Logical Connectives Usage in Science and Social Studies Textbooks." Electronic Journal for Research in Science & Mathematics Education20, no. 6: 1–19.</bibtext> </blist> <blist> <bibtext> Rosseel, Y.2012. "lavaan: An R Package for Structural Equation Modeling." Journal of Statistical Software48, no. 2: 1–36. https://doi.org/10.18637/jss.v048.i02.</bibtext> </blist> <blist> <bibtext> Sabatini, J., T. O'Reilly, L. Halderman, and K. Bruce. 2014. "Broadening the Scope of Reading Comprehension Using Scenario‐Based Assessments: Preliminary Findings and Challenges." L'Annee Psychologique114, no. 4: 693–723. https://doi.org/10.4074/S0003503314004059.</bibtext> </blist> <blist> <bibtext> Satorra, A., and P. M. Bentler. 2001. "A Scaled Difference Chi‐Square Test Statistic for Moment Structure Analysis." Psychometrika66, no. 4: 507–514. https://doi.org/10.1007/BF02296192.</bibtext> </blist> <blist> <bibtext> Schleppegrell, M. J.2004. The Language of Schooling: A Functional Linguistics Perspective. Routledge.</bibtext> </blist> <blist> <bibtext> Townsend, D., R. Knecht, S. Lupo, L.‐T. Chen, and V. S. Barrios. 2023. "A Mixed‐Methods Investigation of Third and Sixth Graders' Academic Sentence Knowledge." Journal of World Languages9, no. 3: 333–370. https://doi.org/10.1515/jwl‐2023‐0022.</bibtext> </blist> <blist> <bibtext> Uccelli, P., C. D. Barr, C. L. Dobbs, E. P. Galloway, A. Meneses, and E. Sánchez. 2015. "Core Academic Language Skills: An Expanded Operational Construct and a Novel Instrument to Chart School‐Relevant Language Proficiency in Preadolescent and Adolescent Learners." Applied PsychoLinguistics36, no. 5: 1077–1109. https://doi.org/10.1017/S014271641400006X.</bibtext> </blist> <blist> <bibtext> Uccelli, P., E. P. Galloway, C. D. Barr, A. Meneses, and C. L. Dobbs. 2015. "Beyond Vocabulary: Exploring Cross‐Disciplinary Academic‐Language Proficiency and Its Association With Reading Comprehension." Reading Research Quarterly50, no. 3: 337–356. https://doi.org/10.1002/rrq.104.</bibtext> </blist> <blist> <bibtext> Van Silfhout, G., J. Evers‐Vermeul, and T. Sanders. 2015. "Connectives as Processing Signals: How Students Benefit in Processing Narrative and Expository Texts." Discourse Processes52, no. 1: 47–76. https://doi.org/10.1080/0163853X.2014.905237.</bibtext> </blist> <blist> <bibtext> Volodina, A., and S. Weinert. 2020. "Comprehension of Connectives: Development Across Primary School Age and Influencing Factors." Frontiers in Psychology11: 814. https://doi.org/10.3389/fpsyg.2020.00814.</bibtext> </blist> <blist> <bibtext> Welie, C., R. Schoonen, F. Kuiken, and H. Van Den Bergh. 2017. "Expository Text Comprehension in Secondary School: For Which Readers Does Knowledge of Connectives Contribute the Most?" Journal of Research in Reading40, no. S1: S42–S65. https://doi.org/10.1111/1467‐9817.12090.</bibtext> </blist> </ref> <aug> <p>By Bailey Buchanan and Paola Uccelli</p> <p>Reported by Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib16" firstref="ref2"></nolink> <nolink nlid="nl2" bibid="bib18" firstref="ref3"></nolink> <nolink nlid="nl3" bibid="bib36" firstref="ref5"></nolink> <nolink nlid="nl4" bibid="bib33" firstref="ref6"></nolink> <nolink nlid="nl5" bibid="bib12" firstref="ref9"></nolink> <nolink nlid="nl6" bibid="bib11" firstref="ref10"></nolink> <nolink nlid="nl7" bibid="bib35" firstref="ref12"></nolink> <nolink nlid="nl8" bibid="bib37" firstref="ref15"></nolink> <nolink nlid="nl9" bibid="bib39" firstref="ref17"></nolink> <nolink nlid="nl10" bibid="bib31" firstref="ref19"></nolink> <nolink nlid="nl11" bibid="bib34" firstref="ref25"></nolink> <nolink nlid="nl12" bibid="bib10" firstref="ref26"></nolink> <nolink nlid="nl13" bibid="bib29" firstref="ref27"></nolink> <nolink nlid="nl14" bibid="bib19" firstref="ref28"></nolink> <nolink nlid="nl15" bibid="bib28" firstref="ref30"></nolink> <nolink nlid="nl16" bibid="bib26" firstref="ref37"></nolink> <nolink nlid="nl17" bibid="bib23" firstref="ref41"></nolink> <nolink nlid="nl18" bibid="bib13" firstref="ref42"></nolink> <nolink nlid="nl19" bibid="bib38" firstref="ref44"></nolink> <nolink nlid="nl20" bibid="bib20" firstref="ref45"></nolink> <nolink nlid="nl21" bibid="bib15" firstref="ref50"></nolink> <nolink nlid="nl22" bibid="bib17" firstref="ref52"></nolink> <nolink nlid="nl23" bibid="bib21" firstref="ref58"></nolink> <nolink nlid="nl24" bibid="bib27" firstref="ref62"></nolink> <nolink nlid="nl25" bibid="bib30" firstref="ref63"></nolink> <nolink nlid="nl26" bibid="bib24" firstref="ref64"></nolink> <nolink nlid="nl27" bibid="bib25" firstref="ref65"></nolink> <nolink nlid="nl28" bibid="bib22" firstref="ref66"></nolink> <nolink nlid="nl29" bibid="bib14" firstref="ref67"></nolink> <nolink nlid="nl30" bibid="bib32" firstref="ref68"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1503743
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Connective Knowledge and Reading Comprehension in Upper Elementary Students: A Growth Analysis
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Bailey+Buchanan%22">Bailey Buchanan</searchLink> (ORCID <externalLink term="https://orcid.org/0009-0009-2574-4835">0009-0009-2574-4835</externalLink>)<br /><searchLink fieldCode="AR" term="%22Paola+Uccelli%22">Paola Uccelli</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0001-5818-2108">0000-0001-5818-2108</externalLink>)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Reading+Research+Quarterly%22"><i>Reading Research Quarterly</i></searchLink>. 2026 61(2).
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 13
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2026
– Name: SourceSuprt
  Label: Sponsoring Agency
  Group: SrcSuprt
  Data: Institute of Education Sciences (ED)<br />Department of Education (ED)
– Name: NumberContract
  Label: Contract Number
  Group: NumCntrct
  Data: R305A170185<br />R305A190034
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Elementary+Education%22">Elementary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Elementary+School+Students%22">Elementary School Students</searchLink><br /><searchLink fieldCode="DE" term="%22Reading+Comprehension%22">Reading Comprehension</searchLink><br /><searchLink fieldCode="DE" term="%22Connected+Discourse%22">Connected Discourse</searchLink><br /><searchLink fieldCode="DE" term="%22Receptive+Language%22">Receptive Language</searchLink><br /><searchLink fieldCode="DE" term="%22English+Learners%22">English Learners</searchLink><br /><searchLink fieldCode="DE" term="%22Language+Proficiency%22">Language Proficiency</searchLink><br /><searchLink fieldCode="DE" term="%22Reading+Achievement%22">Reading Achievement</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1002/rrq.70096
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0034-0553<br />1936-2722
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Connectives--a relatively small, closed set of expressions used to link ideas logically, for example, "therefore," "in contrast"--represent a potentially high-leverage area of focus for literacy interventions due to their prevalence and utility across content-area texts. To inform instruction, however, we first need more research to understand students' development of connective knowledge and test its potential contribution to reading comprehension. In the present study, we used latent growth analysis to examine developmental relations between receptive connective knowledge and reading comprehension in English learners (ELs) and English proficient (EP) students from grade 4 to grade 6 (N = 4100). Three primary findings emerged from our analysis. First, students with greater initial connective knowledge at the start of fourth grade displayed, on average, greater growth in reading comprehension between grades 4 and 6. Second, more rapid growth in connective knowledge across this same timespan predicted, on average, greater growth in reading comprehension. Third, our study finds that the relation between connective knowledge growth and reading comprehension growth was not significantly different for ELs as compared to EP students. To our knowledge, this is the first study to analyze the relation between connective knowledge and reading comprehension in upper elementary students over time and how this relation varies by student language background. These results motivate future intervention-based research to identify possible causal pathways underlying this developmental relation and directs practitioners to consider connective knowledge as a particular instructional area with potential benefits on reading comprehension outcomes for both English learning and English proficient students.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: CodeSource
  Label: IES Funded
  Group: SrcInfo
  Data: Yes
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2026
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1503743
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1503743
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1002/rrq.70096
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 13
    Subjects:
      – SubjectFull: Elementary School Students
        Type: general
      – SubjectFull: Reading Comprehension
        Type: general
      – SubjectFull: Connected Discourse
        Type: general
      – SubjectFull: Receptive Language
        Type: general
      – SubjectFull: English Learners
        Type: general
      – SubjectFull: Language Proficiency
        Type: general
      – SubjectFull: Reading Achievement
        Type: general
    Titles:
      – TitleFull: Connective Knowledge and Reading Comprehension in Upper Elementary Students: A Growth Analysis
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Bailey Buchanan
      – PersonEntity:
          Name:
            NameFull: Paola Uccelli
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 04
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 0034-0553
            – Type: issn-electronic
              Value: 1936-2722
          Numbering:
            – Type: volume
              Value: 61
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Reading Research Quarterly
              Type: main
ResultId 1