The Influence of Pre-Reading Purpose and Extra-Textual Networks with Summary Writing on Multiple Document Concept Network Integration: A Replication of Wei et al. (2024)
Saved in:
| Title: | The Influence of Pre-Reading Purpose and Extra-Textual Networks with Summary Writing on Multiple Document Concept Network Integration: A Replication of Wei et al. (2024) |
|---|---|
| Language: | English |
| Authors: | Ziqian Wei, Shuting Yin, Roy B. Clariana (ORCID |
| Source: | Educational Technology Research and Development. 2025 73(4):2163-2188. |
| Availability: | Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/ |
| Peer Reviewed: | Y |
| Page Count: | 26 |
| Publication Date: | 2025 |
| Sponsoring Agency: | National Science Foundation (NSF), Directorate for STEM Education (EDU) |
| Contract Number: | 2215807 |
| Document Type: | Journal Articles Reports - Research |
| Descriptors: | Prereading Experience, Expository Writing, Descriptive Writing, Prompting, Networks, Revision (Written Composition), Feedback (Response), Reading Comprehension, Intervention, Network Analysis |
| DOI: | 10.1007/s11423-025-10508-8 |
| ISSN: | 1042-1629 1556-6501 |
| Abstract: | Theories and practices to enhance multiple document comprehension and integration are crucial in both personal and work contexts, especially with the proliferation of printed and online sources. This experimental investigation replicates and extends (Wei et al., Educational Technology Research and Development 72:661-685, 2024) to examine how multiple documents integration is influenced by reading purpose, summary writing, and extra-textual networks (pre-reading, Study 1, and post-writing, Study 2). In Study 1 (N = 102), participants were randomly assigned to a pre-reading purpose set by a prompt (integrative or detailed) and by a network (an integrative or else an intra-text network) and then read three documents about Alzheimer's disease to complete a writing task with revision (but no feedback). Three days later, they completed a delayed writing task and an inference verification test. In Study 2 (N = 90), the same procedure was used except that the network was used as feedback after writing to support revision. Results from the two studies agree with the previous research that the quantity and structural quality of integration can be improved by external cues and by delayed repeated writing. This research further confirms an innovative approach for evaluating different aspects of knowledge integration and contributes to the literature from the concept network perspective as a measure and an intervention of multiple-text reading. |
| Abstractor: | As Provided |
| Entry Date: | 2025 |
| Accession Number: | EJ1483812 |
| Database: | ERIC |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwHdyLUTG6oS-NoXXCSV-CZQAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDPU2DEPvgLIqdFb5rQIBEICBm_Y6ax6paC2ekSpRb05GFAJHsZLo0YPE5I0cCGWY7PM003s1X9T19i9pmNK1w0TfTFbEvdDsnTNueuusQ4sMAPNW9kccwgQcldp2fJ0KOOXsPsoPt25VFJ5JsFksQUjRVA51zUF1SvVErd_y0J4EL88WLH9-7bldcfwn3waM25_uaOVMFgYUlvDccYX0NJ3z50Li_cAcRDRpZ122 Text: Availability: 1 Value: <anid>AN0188049025;etr01aug.25;2025Sep22.01:49;v2.2.500</anid> <title id="AN0188049025-1">The influence of pre-reading purpose and extra-textual networks with summary writing on multiple document concept network integration: a replication of Wei et al. (2024) </title> <p>Theories and practices to enhance multiple document comprehension and integration are crucial in both personal and work contexts, especially with the proliferation of printed and online sources. This experimental investigation replicates and extends (Wei et al., Educational Technology Research and Development 72:661–685, 2024) to examine how multiple documents integration is influenced by reading purpose, summary writing, and extra-textual networks (pre-reading, Study 1, and post-writing, Study 2). In Study 1 (N = 102), participants were randomly assigned to a pre-reading purpose set by a prompt (integrative or detailed) and by a network (an integrative or else an intra-text network) and then read three documents about Alzheimer's disease to complete a writing task with revision (but no feedback). Three days later, they completed a delayed writing task and an inference verification test. In Study 2 (N = 90), the same procedure was used except that the network was used as feedback after writing to support revision. Results from the two studies agree with the previous research that the quantity and structural quality of integration can be improved by external cues and by delayed repeated writing. This research further confirms an innovative approach for evaluating different aspects of knowledge integration and contributes to the literature from the concept network perspective as a measure and an intervention of multiple-text reading.</p> <p>Keywords: Multiple document integration; Writing feedback; Concept network analysis; Psychology and Cognitive Sciences Psychology</p> <hd id="AN0188049025-2">Introduction</hd> <p>Learning from multiple sources is an essential ability in today's world. It is often necessary to use various related documents to learn, solve problems, and make decisions, but understanding within and across texts is often a difficult and complex task. Kintsch's Construction-Integration Model ([<reflink idref="bib44" id="ref1">44</reflink>]) describes single document comprehension at three levels of mental representation: the <emph>surface-level</emph> symbols, the semantic memory of the <emph>text base</emph>, and the <emph>situation model</emph> that includes a representation of the context (e.g., events, people, objects, and their relations). To effectively comprehend multiple complementary documents, readers must construct separate situation models of each text (intra-text integration) plus an <emph>integrated situation model</emph> across texts (inter-text integration) to reach a fuller understanding of the content (Perfetti et al., [<reflink idref="bib67" id="ref2">67</reflink>]).</p> <p>External support can provide cues to improve both intra-text and inter-text models that then support multiple document integration. In particular, a purposeful reading prompt is often effective in enhancing learning by directing attention to relevant information (Ayroles et al., [<reflink idref="bib3" id="ref3">3</reflink>]; Lehmann et al., [<reflink idref="bib50" id="ref4">50</reflink>]), and after-reading essays can be converted into <emph>concept networks</emph> to estimate the quantity and quality of multiple-text learning outcomes, as shown in two recent investigations in this journal (Lehmann et al., [<reflink idref="bib49" id="ref5">49</reflink>]; Wei et al., [<reflink idref="bib89" id="ref6">89</reflink>]). Here, integration quantity is defined as the <emph>Amount,</emph> and quality as both <emph>Accuracy</emph> (content) and <emph>Coherence</emph> of knowledge (structural). This aligns well with the three knowledge dimensions suggested by McCarthy and McNamara ([<reflink idref="bib57" id="ref7">57</reflink>]), while the fourth dimension, <emph>Specificity</emph>, was not evaluated in this investigation, as the complementary text reading and summary writing tasks we used do not involve extra-text information compared with other settings such as argument writing.</p> <p>Both Lehmann et al. ([<reflink idref="bib49" id="ref8">49</reflink>]) and Wei et al. ([<reflink idref="bib89" id="ref9">89</reflink>]) found that integrative reading prompts (vs. control) encourage students to construct more integrative knowledge after reading (i.e., better quantity) by using concept network analysis. However, given the importance of the quality (McNamara et al., [<reflink idref="bib63" id="ref10">63</reflink>]; Parkinson &amp; Dinsmore, [<reflink idref="bib66" id="ref11">66</reflink>]) and the relatively small observed effect of reading prompts on improving quality (at least, in Lehmann and Wei's studies), it is necessary to search for effective interventions to support integration quality. One possible method is to provide an expert network as a learning reference either before or during learning. For example, Leopold et al. ([<reflink idref="bib52" id="ref12">52</reflink>]) reported an advantage for pictorial over verbal interventions in multiple document settings. They note that "the pictorial representation provides structural and spatial information that is related to the referential objects by structural analogy. Therefore, it promotes mental model building and deep understanding" (p. 47).</p> <p>Why may a visualized network be better than textual writing for multiple document integration? Visualized networks have been used to elicit, refine, and re-represent readers' conceptual structure (Chen et al., [<reflink idref="bib17" id="ref13">17</reflink>]; Pirnay-Dummer &amp; Ifenthaler, [<reflink idref="bib68" id="ref14">68</reflink>]), because the greater the re-representation, the higher the level of task performance the second time around (Kentgen et al., [<reflink idref="bib36" id="ref15">36</reflink>]). Summary writing also requires readers to form, refine, and re-represent their conceptual structure, but multiple document summary writing is a very difficult production task. Wei et al. ([<reflink idref="bib89" id="ref16">89</reflink>]) emphasized:"... the importance of updating knowledge to develop a well-constructed mental representation. Such knowledge revision may be performed by providing materials including external expert-level information such as an expert network or outline (Chen et al., [<reflink idref="bib17" id="ref17">17</reflink>]). This should be considered in future studies." (p. 79)</p> <p>Thus, adding immediate feedback as an expert network (i.e., a map-form representation) may help improve the integration quality because students will have a chance to note and repair errors as well as add missing information. This investigation considered how an expert network presented and/or used before reading (Study 1) or as feedback after reading and writing (Study 2) can improve both integrative quality and quantity.</p> <p>Spector et al. ([<reflink idref="bib77" id="ref18">77</reflink>]) called for replication studies to undergird research findings. This investigation responded to the editors' call by conducting these direct replication studies that preserve the original instrumentation and approach with extended experimental manipulations so that the previous findings can be re-tested and relevant interactions and marginal effects can be examined. For this purpose, we used the same population, web-based software tool, materials, and analysis approach as Wei et al. ([<reflink idref="bib89" id="ref19">89</reflink>]).</p> <hd id="AN0188049025-3">Prompts-activated purposeful reading and multiple document integration</hd> <p>Britt et al. ([<reflink idref="bib9" id="ref20">9</reflink>]) noted, "Reading is driven by goals that can influence reading behavior and processing, ...(and) the construct of 'reading goal' may be more complex than previously imagined" (p. 27). This is especially so for multiple texts. As proposed in theories such as the Multiple-Documents Task-based Relevance Assessment and Content Extraction (MD-TRACE) Model (Rouet &amp; Britt, [<reflink idref="bib72" id="ref21">72</reflink>]) and the Reading as Problem Solving (RESOLV) Model (Rouet et al., [<reflink idref="bib73" id="ref22">73</reflink>]), readers are believed to construct a <emph>task model</emph> of their reading purpose. The task model directs and influences how the reader attends to information within and across sources; unimportant and redundant content is ignored (McCrudden &amp; Schraw, [<reflink idref="bib60" id="ref23">60</reflink>]), thereby facilitating task-specific integration.</p> <p>In multiple-text reading, readers have to form an inter-text situation model to capture the links between texts' content (Cerdán &amp; Vidal-Abarca, [<reflink idref="bib15" id="ref24">15</reflink>]; Unsworth &amp; McMillan, [<reflink idref="bib83" id="ref25">83</reflink>]). While individuals are somewhat put upon by the increased task demands, purposeful prompts can support the integration by setting a task model to guide purposeful reading (Ayroles et al., [<reflink idref="bib3" id="ref26">3</reflink>]; Latini et al., [<reflink idref="bib48" id="ref27">48</reflink>]; Mason et al., [<reflink idref="bib56" id="ref28">56</reflink>]). For instance, Ayroles et al. ([<reflink idref="bib3" id="ref29">3</reflink>]) showed that compared with a control condition ("Is the word *** present in this question?"), a task model prompt ("What do you have to find in the text to answer this question?") significantly improves primary school children's performance in multiple document comprehension. The power of relevant information processing activated by focused prompts can be explained as the enhancement of sourcing skills that facilitate inter-text information targeting (Brante &amp; Strømsø, [<reflink idref="bib6" id="ref30">6</reflink>]; Bråten et al., [<reflink idref="bib8" id="ref31">8</reflink>]; Strømsø et al., [<reflink idref="bib78" id="ref32">78</reflink>]).</p> <p>When coding students' qualitative learning performance (e.g., writing, think-aloud, notes), most previous studies focused on either measuring the quantity of integration (e.g., the number of integrative statements in after-reading summaries) (Lehmann et al., [<reflink idref="bib50" id="ref33">50</reflink>]; Linderholm &amp; van den Broek, [<reflink idref="bib53" id="ref34">53</reflink>]; List et al., [<reflink idref="bib55" id="ref35">55</reflink>]) or the sum scores that combine the quantity and the quality of integration (e.g., the number of correct integrative statements) (Cerdán &amp; Vidal-Abarca, [<reflink idref="bib15" id="ref36">15</reflink>]; Cheong et al., [<reflink idref="bib18" id="ref37">18</reflink>]). It is important now to distinguish whether prompt-activated purposeful reading (or other learning supports) separately improves both the quantity and quality, or only quantity. Parkinson and Dinsmore ([<reflink idref="bib66" id="ref38">66</reflink>]) proposed that quality rather than quantity most relates to improved learning outcomes, and this relationship is also observed in multiple document settings (McNamara et al., [<reflink idref="bib63" id="ref39">63</reflink>]). To further examine the "better quality matters" assumption, we utilized an advanced approach called concept network analysis to measure and evaluate students' learning.</p> <hd id="AN0188049025-4">Concept network analysis as measures of integration quality and quantity</hd> <p>Conceptual network analysis (aka knowledge structure analysis) has been used for decades in single document comprehension research (Clariana &amp; Koul, [<reflink idref="bib21" id="ref40">21</reflink>]; Ifenthaler, [<reflink idref="bib32" id="ref41">32</reflink>]) and recent studies have demonstrated its potential to measure different aspects of multiple document reading (Lehmann et al., [<reflink idref="bib49" id="ref42">49</reflink>]; Wei et al., [<reflink idref="bib89" id="ref43">89</reflink>]). By converting students' learning products, such as summary essays, into networks with nodes representing the core concepts and edges (i.e., links) representing semantic relationships between them, researchers can better explore the properties of students' mental representations (Jonassen et al., [<reflink idref="bib33" id="ref44">33</reflink>]; Kintsch, [<reflink idref="bib43" id="ref45">43</reflink>]). Concept links derived from different documents can be used as the indicator of <emph>integrative</emph> knowledge, while links that consist of concepts derived from a single document relate to <emph>intra-text</emph> knowledge (McCrudden et al., [<reflink idref="bib58" id="ref46">58</reflink>]; Primor et al., [<reflink idref="bib70" id="ref47">70</reflink>]).</p> <p>Students differ from domain experts in diverse aspects, such as prior knowledge and reading strategies, when they read texts with unfamiliar topics (Brand‐Gruwel et al., [<reflink idref="bib5" id="ref48">5</reflink>]; McNamara &amp; Kintsch, [<reflink idref="bib62" id="ref49">62</reflink>]; Witherby &amp; Carpenter, [<reflink idref="bib91" id="ref50">91</reflink>]). These gaps will likely limit the accuracy of students' semantic statements even when the quantity is adequate (see Fig. 1). Thus, <emph>the semantic quality of integration</emph> can be measured as the similarity of integrative links between a student's concept network and the expert referent network (Follmer et al., [<reflink idref="bib28" id="ref51">28</reflink>]; Ifenthaler, [<reflink idref="bib32" id="ref52">32</reflink>]).</p> <p>Graph: Fig. 1 An expert network (left side) and four example concept networks (right side) (Wei et al., [<reflink idref="bib89" id="ref53">89</reflink>]). Nodes are concepts in the documents and links are semantic relationships among them. Bold lines indicate links that align with the expert network. All networks have the same quantity of integration (i.e., five identical integrative links) but differ in either the semantic (i.e., Qsemantic) or structural quality (i.e., Qstructural)</p> <p>It is also worthwhile to evaluate <emph>the structural quality of integration</emph>, which reflects how readers form a coherent organization of documents. Structural knowledge is the knowledge of how concepts within a domain are interconnected (Jonassen et al., [<reflink idref="bib33" id="ref54">33</reflink>], p. 4). Such knowledge is especially critical when reading multiple documents, as the structure of cross-document knowledge is usually more ambiguous than in single document comprehension.</p> <p>Freeman's ([<reflink idref="bib29" id="ref55">29</reflink>]) graph centrality (GC) is a holistic measure of network form that indicates how individuals' mental representations are organized, values range from 0 to 1 that can be used to categorize network topology based on the concept-concept links in the network, for example, values &lt; 0.30 are <emph>linear-sequential or chain</emph>, from 0.30 to 0.50 are <emph>hierarchical</emph>, from 0.50 to 0.70 are <emph>network</emph>, and &gt; 0.70 are <emph>hub or star</emph> (Clariana et al., [<reflink idref="bib22" id="ref56">22</reflink>]; Kim &amp; Clariana, [<reflink idref="bib39" id="ref57">39</reflink>]). A relative network form measure from Kim and McCarthy ([<reflink idref="bib42" id="ref58">42</reflink>]) called <emph>graph similarity</emph>, GC<subs>sim</subs> (described in more detail later) is used to compare the student's essay network form to the expert's network form; as with GC above, for GC<subs>sim</subs> larger is better, 1 means the essay GC form is exactly like the expert's GC form while 0 means it is completely unlike the expert.</p> <hd id="AN0188049025-5">Pre-reading expert referent network and after-reading network-based writing feedback</hd> <p>It seems crucial to seek effective methods to improve the semantic and structural quality of integration from reading multiple documents. To bridge the gap between students and domain experts, an explicit approach is to provide additional extra-textual learning support using an expert referent network. Two possible approaches were considered here to explore whether a network reference influences students' integration before reading (Study 1) or after reading (Study 2).</p> <p>In Study 1, participants are given an expert network before reading. These concept-concept network maps are very similar to concept maps but without the linking verbs. Villalon and Calvo ([<reflink idref="bib86" id="ref59">86</reflink>]) noted, "Concept maps have typically been used in reading activities to aid students' comprehension of texts. For instance, ready-made concept maps may be presented as semantic summaries of texts that students need to comprehend" (p.16). The presentation of a pre-reading network can anchor meaningful reading by visually displaying the key objects, events, or ideas in the text materials within one image (Ausubel &amp; Fitzgerald, [<reflink idref="bib1" id="ref60">1</reflink>]; Hawk, [<reflink idref="bib31" id="ref61">31</reflink>]).</p> <p>The influence of graphic organizers on learning has been reported in numerous studies (Casteleyn et al., [<reflink idref="bib14" id="ref62">14</reflink>]; Chang et al., [<reflink idref="bib16" id="ref63">16</reflink>]; Fan &amp; Chen, [<reflink idref="bib27" id="ref64">27</reflink>]; Nesbit &amp; Adesope, [<reflink idref="bib64" id="ref65">64</reflink>]). However, the scale of the whole expert referent network can be large when many texts are used. Since a large and complex network can be confusing and thus limit its effectiveness (Davies, [<reflink idref="bib26" id="ref66">26</reflink>]), a smaller referent network that focuses only on integrative knowledge might be better. Therefore, the full expert referent network used by Wei et al. ([<reflink idref="bib89" id="ref67">89</reflink>]) and in this current investigation was reduced into an intra-text network (i.e., intra-text links within each document) and an integrative network (i.e., inter-text links across the documents).</p> <p>In Study 2, participants were shown the full expert network after reading to compare to their own essays for reflection to support subsequent writing revisions. By utilizing concept network analysis, network-based feedback can be automatically generated from students' essays (Burkhart et al., [<reflink idref="bib10" id="ref68">10</reflink>]; Kim, [<reflink idref="bib38" id="ref69">38</reflink>]; Lachner et al., [<reflink idref="bib46" id="ref70">46</reflink>]; Pirnay-Dummer &amp; Ifenthaler, [<reflink idref="bib69" id="ref71">69</reflink>]; Villalon &amp; Calvo, [<reflink idref="bib86" id="ref72">86</reflink>]). Compared with a pre-reading referent network that is intended to support reading for integration, writing <emph>after reading</emph> is intended to re-represent their mental model of the documents content as personalized writing feedback that is intended to update their mental model to align with the expert referent.</p> <p>Previous studies have reported that network-based feedback after writing enhances learning performance by improving either structural organization or coherence (Kim &amp; McCarthy, [<reflink idref="bib42" id="ref73">42</reflink>]; Lachner &amp; Neuburg, [<reflink idref="bib47" id="ref74">47</reflink>]; Lachner et al., [<reflink idref="bib45" id="ref75">45</reflink>]). For example, Clariana ([<reflink idref="bib19" id="ref76">19</reflink>]) asked Architectural undergraduate students to write summary essays of lesson content for four lessons across the course, including the topics of Building with concrete, Building with wood, Masonry construction, and Sustainable building. Students used the Graphical Interface of Knowledge Structure software (GIKS) that provides immediate structural feedback as an expert referent network alongside a network of the student's essay for comparison and reflection. Results showed that (<reflink idref="bib1" id="ref77">1</reflink>) writing with GIKS compared to the no writing control improved conceptual knowledge structure but at the expense of declarative knowledge and (<reflink idref="bib2" id="ref78">2</reflink>) writing with GIKS compared to writing with a word processor (i.e., no network feedback) improved conceptual knowledge structure but had no effect on declarative knowledge. In this current investigation, GIKS is adapted into Chinese and embedded into the software interface used in Study 2.</p> <hd id="AN0188049025-6">The present investigation</hd> <p>Wei et al. ([<reflink idref="bib89" id="ref79">89</reflink>]) described and applied a concept network analysis approach to explore how pre-reading prompts and post-reading generative learning tasks influence undergraduates' document integration performance. Results showed that integrative prompts relative to detailed prompts enhance the quantity of integration, but not the semantic and structural quality. This current investigation replicates and extends Wei et al. ([<reflink idref="bib89" id="ref80">89</reflink>]) using the same reading materials and pre-reading prompts (integrative prompts vs. detailed prompts), with different but related extra-textual strategies intended to leverage the findings for concept maps, including a pre-reading expert integrative network vs. intra-text network (Study 1) and post-reading network-based writing feedback (Study 2).</p> <hd id="AN0188049025-7">Study 1</hd> <p>A 2 (reading prompts: integrative prompts vs. detailed prompts) × 2 (expert referent network: integrative network vs. intra-text network) × 3 (writing stage: after-reading writing vs. revised writing vs. delayed writing) mixed design was employed to investigate how these factors affect multiple document integration when reading complementary knowledge of the pathology and application of a disorder that is unfamiliar to the public (i.e., Alzheimer's disease). The quantity of integration was expected to be improved by the integrative prompts (H1a), the integrative referent network (H1b) and relevant interactions (H1c) according to the evidence in the literature. Semantic quality was expected to be improved by both the integrative referent network (H2a) and relevant interactions (H2b). The structural quality was expected to be improved by both the integrative referent network (H3a) and relevant interactions (H3b). It should be noted that a revised writing phase would be mainly considered as the baseline for Study 2, in which feedback was given before revision. In this case, the effects of feedback in Study 2 could be discussed by comparing the performance of revised writing in two studies.</p> <hd id="AN0188049025-8">Method</hd> <p></p> <hd id="AN0188049025-9">Participants</hd> <p>One hundred and eight participants were recruited. None majored in psychology or medicine or had engaged in study or work on Alzheimer's disease (i.e., text topic). Participants were randomly assigned to one of four conditions including <emph>integrative prompts</emph> + <emph>integrative referent network</emph>, <emph>integrative prompts</emph> + <emph>intra-text referent network</emph>, <emph>detailed prompts</emph> + <emph>integrative referent network</emph>, and <emph>detailed prompts</emph> + <emph>intra-text referent network</emph>.</p> <p>Considering that the level of prior knowledge would affect the performance of integration (Gil et al., [<reflink idref="bib30" id="ref81">30</reflink>]; Karimi, [<reflink idref="bib35" id="ref82">35</reflink>]; Leon &amp; Perez, [<reflink idref="bib51" id="ref83">51</reflink>]; Tarchi &amp; Mason, [<reflink idref="bib80" id="ref84">80</reflink>]), participants were asked two additional questions at the end of the experiment. First, whether they had read similar materials before. Second, if the former answer was yes, they would be required to rate the similarity of the content on a five-point scale (1 = totally dissimilar, 5 = totally similar). Six participants answered yes to the first question and thus were removed from the statistical analysis (<emph>M</emph><subs>rating of the content similarity</subs> = 3.38, <emph>SD</emph> = 1.06). The remaining 102 participants were included in the statistical analysis (67 females, 35 males, age: <emph>M</emph> = 20.53, <emph>SD</emph> = 1.82).</p> <p>A prior power analysis was conducted via the <emph>Webpower</emph> package in R for sample size estimation (Zhang &amp; Yuan, [<reflink idref="bib92" id="ref85">92</reflink>]). Results showed that the minimum sample size needed should be 64 for the main effect of external referent networks in a 2 × 2 × 3 repeated measures ANOVA with Cohen's <emph>f</emph> = 0.40, α =.05, and power = 0.80. The obtained sample size of 102 was more than adequate to test the hypothesis.</p> <p>The data collection in the current research was reviewed and approved by the local university ethics committee (No. SCNU-PSY-2021–168). All participants provided their written consent before the investigation.</p> <hd id="AN0188049025-10">Materials</hd> <p></p> <hd id="AN0188049025-11">Text materials and referent networks</hd> <p>Three documents from Wei et al. ([<reflink idref="bib89" id="ref86">89</reflink>]) on the topic of Alzheimer's disease were taken from the National Institute of Health website were used as reading materials (note: the text arrangement and the pilot test of the text content coherence and readability can be found in Wei et al., [<reflink idref="bib89" id="ref87">89</reflink>]). Thirty-nine core concepts (verbs, nouns, or nouns modified by adjectives) were selected. Experts then wrote an integrative summary to reflect the gist of the documents. The expert summary consisted of thirteen core concepts. This expert summary was converted to a concept network and is used as the expert network as the referent for calculating network similarity (see the top of Fig. 2).</p> <p>Graph: Fig. 2 The full expert referent network (top) and the corresponding integrative and intra-text networks (bottom). The three document terms are shown as blue, green, and yellow, and the central concept Alzheimer's disease that is common across all three documents is shown in white. These Chinese networks are shown here in English</p> <p>In both Study 1 and 2, this expert network was portioned into two referent subnetworks that were identical in concept terms and spatial layout, except one is an integrative referent network that shows the five most important term-term links across documents and the other is an intra-text referent that shows the five most important term-term links within each document (see the bottom of Fig. 2). In Study 1 one or the other of these subnetworks was displayed as pre-reading learning support to guide reading.</p> <p>The links connecting the common concept <emph>Alzheimer's disease</emph> in all three documents were not included in the two pre-reading networks in order to focus the readers' attention closely on either the within-document terms (intra-text network) or the across-document terms (integrative network). Note that eye-tracking research has shown that in most cases participants pay initial attention to central nodes and links in the network and then notice the spatial clusters of terms from the top left moving leftward around the network (Chen et al., [<reflink idref="bib17" id="ref88">17</reflink>]; Nesbit, Larios, &amp; Adesope, [<reflink idref="bib65" id="ref89">65</reflink>]).</p> <hd id="AN0188049025-12">The reading and writing tasks</hd> <p>The reading and writing tasks were designed and conducted via <emph>PsychoPy-3</emph> (PsychoPy version: 2021.2.3; Python version: 3.6.8). This software interface was modified from that used by Wei et al. ([<reflink idref="bib89" id="ref90">89</reflink>]) to display the network before reading. Notice the integrative/intra-text referent network screen (top of Fig. 3), along with the user interfaces for the reading task (middle) and the writing task (bottom). The concept list was presented in the center column of the writing task. Participants can read the text and write the essay on the left-hand side, and switch to the texts or core concepts in each text by clicking the buttons in the middle to view the instructions on the right.</p> <p>Graph: Fig. 3 The interface of the pre-reading expert reference (top), the reading task (middle), and the writing task (bottom), where 1—Task instruction, 2—Expert network, 3—Continue button, 4—Task title, 5—Switch button, 6—Task duration, 7—Concept list, 8—Input box, 9—Number of concepts in use, 10—Word count</p> <hd id="AN0188049025-13">Prompts for different reading purposes</hd> <p>Prompts in the form of focus questions have been utilized for different reading purposes (Lehmann et al., [<reflink idref="bib50" id="ref91">50</reflink>]; McCrudden &amp; Schraw, [<reflink idref="bib60" id="ref92">60</reflink>]). The integrative prompts used here were four questions that assist integration (e.g., "Can you find concepts related to different documents?"). Similar to Ayroles et al. ([<reflink idref="bib3" id="ref93">3</reflink>]), the detailed prompts were used as a control condition that focused on superficial information (e.g., "How many paragraphs are there in these three documents?"). These prompts were available during the whole session.</p> <hd id="AN0188049025-14">Alzheimer's disease knowledge scale (pretest)</hd> <p>The Chinese version of the Alzheimer's Disease Knowledge Scale (ADKS) was used to measure topic-related prior knowledge (Carpenter et al., [<reflink idref="bib13" id="ref94">13</reflink>]). The final scale contained sixteen items in a verification format (i.e., true/false). Cronbach's α was 0.65.</p> <hd id="AN0188049025-15">Inference verification test (posttest)</hd> <p>An Inference Verification Test (IVT) was used to measure comprehension and performance in recalling and recognizing integrative knowledge (Cerdán &amp; Vidal-Abarca, [<reflink idref="bib15" id="ref95">15</reflink>]; Cheong et al., [<reflink idref="bib18" id="ref96">18</reflink>]; Lehmann et al., [<reflink idref="bib50" id="ref97">50</reflink>]; Royer et al., [<reflink idref="bib74" id="ref98">74</reflink>]). The IVT covered three theory-based cognitive requirements: <emph>collaboration</emph> (Wineburg, [<reflink idref="bib90" id="ref99">90</reflink>]), <emph>integrated situation model</emph>, and <emph>intertext model</emph> (Rouet &amp; Britt, [<reflink idref="bib72" id="ref100">72</reflink>]). All items of the IVT were administered in a verification format (i.e., true/false). The final scale contained eleven items (i.e., six for <emph>collaboration</emph>, three for the <emph>integrated situation model</emph>, and two for the <emph>intertext model</emph>). Cronbach's α was 0.69. The total scores were calculated.</p> <hd id="AN0188049025-16">Procedure</hd> <p>Two sessions were used in Study 1 (see Fig. 4). In the first session, participants were randomly assigned to one of four conditions after signing up for the experiment and completing the ADKS test of prior knowledge. They received either the integrative or detailed prompts and either the integrative or intra-text referent network. Next, participants completed the reading task and the after-reading writing task which includes a writing phase (i.e., a 300-word summary essay to include thirteen core concepts) and a revision phase. In the revision phase, participants used the same thirteen core concepts from their initial writing. The first session took approximately 20 min. Those who finished the task within 1 min in the first session tasks were deemed hasty and could not continue to the second session.</p> <p>Graph: Fig. 4 The procedures of the prior research (Wei et al., [<reflink idref="bib89" id="ref101">89</reflink>]) along with that of Study 1 and 2. The dashed boxes are the different parts between Wei et al. ([<reflink idref="bib89" id="ref102">89</reflink>]) and this research. The red dashed boxes are the different parts between Study 1 and 2</p> <p>The second session delayed measures were collected three days after the first session (Kang et al., [<reflink idref="bib34" id="ref103">34</reflink>]; van Peppen et al., [<reflink idref="bib85" id="ref104">85</reflink>]). All participants wrote delayed essays that reflected the contents of their summary essays in the first session. At this stage, each participant received a specific concept list showing the selected thirteen concepts in the first session. Finally, participants completed the IVT, two additional questions about similar materials reading experience, and a demographic questionnaire. The second session took approximately 15 min.</p> <hd id="AN0188049025-17">Data analysis</hd> <p>Participants' essays from the different stages were converted to concept networks via Analysis of Lexical Aggregates (ALA; (Clariana et al., [<reflink idref="bib24" id="ref105">24</reflink>]) and the Pathfinder network algorithm with <emph>r</emph> = infinity, <emph>q</emph> = <emph>n</emph>—1 (Schvaneveldt et al., [<reflink idref="bib76" id="ref106">76</reflink>], [<reflink idref="bib75" id="ref107">75</reflink>]). In ALA, the co-occurrences of pre-selected core concepts in a linear pass through the essay are used to identify term-term associations, the resulting network edges then consist of links between these pairs of concepts. In other words, conceptual relationships in a student's essay were marked as undirected, uniformly weighted edges in a semantic network, and so their quantity (i.e., the edges) and quality (i.e., network similarity between a student and an expert) can be evaluated. In this investigation, every student used nineteen concepts as required, this allowed us to compare results on the same scale, as edge density can be affected by the number of nodes according to graph theory.</p> <p>The three network features were then estimated as follows. The <emph>quantity of integration</emph> was calculated as the proportion of integrative links. For instance, for a given network containing five integrative links and 20 links in total, the proportion of integrative links is 5/20 = 0.25.</p> <p>The <emph>semantic quality</emph> of the network was measured by calculating the similarity of integrative links (see Eq. 1 from (Tversky, [<reflink idref="bib82" id="ref108">82</reflink>]). In Eq. 1, A and B indicate the subset of integrative links from either a participant's concept network or the expert network. Following the literature of calculating network <emph>percent overlap</emph> (e.g., the intersection divided by the average number of links in the two networks), parameters α and β were set as 0.5 (Clariana et al., [<reflink idref="bib24" id="ref109">24</reflink>]). From 0 to 1, a larger <emph>S</emph><subs><emph>semantic</emph></subs> value indicates greater similarity with the full expert network.</p> <p>1 <ephtml> &lt;math display="block" xmlns="http://www.w3.org/1998/Math/MathML"&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;S&lt;/mi&gt;&lt;mrow&gt;&lt;mi mathvariant="italic"&gt;semantic&lt;/mi&gt;&lt;/mrow&gt;&lt;/msub&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mi&gt;f&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mrow&gt;&lt;mtext&gt;A&lt;/mtext&gt;&lt;mspace width="0.333333em" /&gt;&lt;/mrow&gt;&lt;mo&gt;&amp;#8745;&lt;/mo&gt;&lt;mrow&gt;&lt;mspace width="0.333333em" /&gt;&lt;mtext&gt;B&lt;/mtext&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo stretchy="false"&gt;/&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;mfenced close="]" open="["&gt;&lt;mrow&gt;&lt;mi&gt;&amp;#945;&lt;/mi&gt;&lt;mrow /&gt;&lt;mo&gt;&amp;#8727;&lt;/mo&gt;&lt;mi&gt;f&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mrow&gt;&lt;mtext&gt;A&lt;/mtext&gt;&lt;mspace width="0.333333em" /&gt;&lt;/mrow&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;mrow&gt;&lt;mspace width="0.333333em" /&gt;&lt;mtext&gt;B&lt;/mtext&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;+&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;mi&gt;&amp;#946;&lt;/mi&gt;&lt;mrow /&gt;&lt;mo&gt;&amp;#8727;&lt;/mo&gt;&lt;mi&gt;f&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mrow&gt;&lt;mtext&gt;B&lt;/mtext&gt;&lt;mspace width="0.333333em" /&gt;&lt;/mrow&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;mrow&gt;&lt;mspace width="0.333333em" /&gt;&lt;mtext&gt;A&lt;/mtext&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;+&lt;/mo&gt;&lt;mi&gt;f&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mrow&gt;&lt;mtext&gt;A&lt;/mtext&gt;&lt;mspace width="0.333333em" /&gt;&lt;/mrow&gt;&lt;mo&gt;&amp;#8745;&lt;/mo&gt;&lt;mrow&gt;&lt;mspace width="0.333333em" /&gt;&lt;mtext&gt;B&lt;/mtext&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/math&gt; </ephtml></p> <p>Graph</p> <p>Although these two measures are calculated relative to the full expert network and are correlated to some degree, the quantity measures absolute common links relative to only the student's essay network, while the semantic quality measures common links relative to both the expert network and the student's essay network. This is an important distinction when networks are generated from essays because students rarely include every term in the expert network referent. Leaving out key terms (e.g., central terms with many links in the expert network) has a profound negative impact on semantic quality but not on the quantity of integration.</p> <p>The <emph>structural quality</emph> of the networks was calculated as the similarity of the essay network GC relative to the expert network (i.e., GC<subs>sim</subs>). GC is calculated as the degree centrality of each node in a concept network while accounting for one degree of freedom (see Eq. 2) and then summing these adjusted node degree values into a single value while accounting for two degrees of freedom (see Eq. 3). In Eq. 2 and 3, <emph>v</emph> indicates a vertex (i.e., node), <emph>max</emph> (<emph>C</emph><subs><emph>D</emph></subs>(<emph>v</emph><subs><emph>i</emph></subs>)) indicates edge numbers of the node with the greatest centrality, and <emph>n</emph> indicates the total number of vertices in a concept network. Next, GC<subs>sim</subs> can be calculated by comparing a participant's GC value to the expert network's GC (see Eq. 4) (Kim &amp; McCarthy, [<reflink idref="bib42" id="ref110">42</reflink>]). From 0 to 1, a larger GC<subs>sim</subs> value indicates higher similarity to the expert network.</p> <p>2 <ephtml> &lt;math display="block" xmlns="http://www.w3.org/1998/Math/MathML"&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;mi&gt;v&lt;/mi&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mrow&gt;&lt;mspace width="0.333333em" /&gt;&lt;mtext&gt;deg&lt;/mtext&gt;&lt;/mrow&gt;&lt;mfenced close=")" open="("&gt;&lt;mi&gt;v&lt;/mi&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo stretchy="false"&gt;/&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mi&gt;n&lt;/mi&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/math&gt; </ephtml></p> <p>Graph</p> <p>3 <ephtml> &lt;math display="block" xmlns="http://www.w3.org/1998/Math/MathML"&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;mi&gt;G&lt;/mi&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;msubsup&gt;&lt;mo&gt;&amp;#8721;&lt;/mo&gt;&lt;mrow&gt;&lt;mi&gt;i&lt;/mi&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;/mrow&gt;&lt;mi&gt;v&lt;/mi&gt;&lt;/msubsup&gt;&lt;mrow&gt;&lt;mfenced close="]" open="["&gt;&lt;mrow&gt;&lt;mi&gt;m&lt;/mi&gt;&lt;mi&gt;a&lt;/mi&gt;&lt;mi&gt;x&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;v&lt;/mi&gt;&lt;mi&gt;i&lt;/mi&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;v&lt;/mi&gt;&lt;mi&gt;i&lt;/mi&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo stretchy="false"&gt;/&lt;/mo&gt;&lt;mspace width="0.166667em" /&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;mi&gt;n&lt;/mi&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;mn&gt;2&lt;/mn&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/math&gt; </ephtml></p> <p>Graph</p> <p>4 <ephtml> &lt;math display="block" xmlns="http://www.w3.org/1998/Math/MathML"&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mtext&gt;GC&lt;/mtext&gt;&lt;mtext&gt;sim&lt;/mtext&gt;&lt;/msub&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;mi&gt;a&lt;/mi&gt;&lt;mi&gt;b&lt;/mi&gt;&lt;mi&gt;s&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;G&lt;/mi&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo&gt;-&lt;/mo&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;G&lt;/mi&gt;&lt;mn&gt;2&lt;/mn&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;mspace width="0.166667em" /&gt;&lt;mo stretchy="false"&gt;/&lt;/mo&gt;&lt;mi&gt;m&lt;/mi&gt;&lt;mi&gt;a&lt;/mi&gt;&lt;mi&gt;x&lt;/mi&gt;&lt;mfenced close=")" open="("&gt;&lt;mrow&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;G&lt;/mi&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;mo&gt;,&lt;/mo&gt;&lt;msub&gt;&lt;mi&gt;C&lt;/mi&gt;&lt;mi&gt;D&lt;/mi&gt;&lt;/msub&gt;&lt;mfenced close=")" open="("&gt;&lt;msub&gt;&lt;mi&gt;G&lt;/mi&gt;&lt;mn&gt;2&lt;/mn&gt;&lt;/msub&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/mfenced&gt;&lt;/mrow&gt;&lt;/math&gt; </ephtml></p> <p>Graph</p> <hd id="AN0188049025-18">Results</hd> <p>Welch's correction was used to address the variance heterogeneity in one-way ANOVAs/ANCOVAs. A one-way ANOVA showed that ADKS scores among the four groups were not significantly different, <emph>F</emph><subs>(<reflink idref="bib3" id="ref111">3</reflink>,<reflink idref="bib98" id="ref112">98</reflink>)</subs> = 0.62, <emph>p</emph> = 0.605, η<sups>2</sups> =.019 (<emph>M</emph> = 0.66, <emph>SD</emph> = 0.14). The ANCOVA results incorporating ADKS as a covariate suggested that prior knowledge did not affect the outcomes, which is reasonable since only students who were unfamiliar with this topic were selected for participation. Similar with Wei et al. ([<reflink idref="bib89" id="ref113">89</reflink>]), ADKS was not included in the following analyses.</p> <hd id="AN0188049025-19">Comprehension of integrative knowledge</hd> <p>IVT scores were analyzed with a 2 (reading prompts: integrative <emph>vs.</emph> detailed) × 2 (external referent networks: integrative <emph>vs.</emph> intra-text) two-way ANOVA. Levene's test indicated equal variances among groups, <emph>F</emph><subs>(<reflink idref="bib3" id="ref114">3</reflink>,<reflink idref="bib98" id="ref115">98</reflink>)</subs> = 1.32, <emph>p</emph> = 0.274. Results of the ANOVA showed no significant main effects or interaction, all <emph>p</emph>-values &gt;.05. The observed means and standard deviations are as follows, <emph>integrative prompts</emph> + <emph>integrative network</emph> (<emph>n</emph> = 26): <emph>M</emph> = 0.84, <emph>SD</emph> = 0.15; <emph>integrative prompts</emph> + <emph>intra-text network</emph> (<emph>n</emph> = 24): <emph>M</emph> = 0.82, <emph>SD</emph> = 0.15; <emph>detailed prompts</emph> + <emph>integrative network</emph> (<emph>n</emph> = 26): <emph>M</emph> = 0.84, <emph>SD</emph> = 0.15; <emph>detailed prompts</emph> + <emph>intra-text network</emph> (<emph>n</emph> = 26): <emph>M</emph> = 0.84, <emph>SD</emph> = 0.10. These results indicate that all participants comprehended the texts at a competent level equally (i.e., accuracy is above 0.80).</p> <hd id="AN0188049025-20">Essay network quantity and quality</hd> <p>The proportion of integrative links (quantity of integration), the similarity of integrative links (semantic quality of integration), and GC<subs>sim</subs> (structural quality of network form relative to the expert network) were analyzed with separate 2 (prompts) × 2 (external referent networks) × 3 (writing stages: after-reading writing/after revision writing/delayed writing) repeated measures ANOVAs with writing stages as a within-subject variable. For the dependent variables that Mauchly's test of sphericity was statistically significant, <emph>p</emph>-values were reported after the Greenhouse–Geisser correction to address the variance heterogeneity. The eta-square (η<sups>2</sups>) was reported for the <emph>F</emph>-value's effect size, with.01,.06, and 0.14 indicating small, medium, and large effects, respectively; Cohen's <emph>d</emph> was reported for the <emph>t</emph>-value's effect size, with 0.20, 0.50, and 0.80 indicating small, medium, and large effects, respectively (Cohen, [<reflink idref="bib25" id="ref116">25</reflink>]).</p> <p>First, regarding the proportion of integrative links (quantity of integration), the ANOVA results showed a significant main effect of prompts, <emph>F</emph><subs>(<reflink idref="bib1" id="ref117">1</reflink>,<reflink idref="bib98" id="ref118">98</reflink>)</subs> = 6.14, <emph>p</emph> =.015, η<sups>2</sups> =.051, thereby supporting H1a. Participants in the integrative prompts condition (<emph>M</emph> = 0.22, <emph>SD</emph> = 0.15) constructed more integrative links in their summaries than the detailed prompts condition (<emph>M</emph> = 0.16, <emph>SD</emph> = 0.11) during the whole experiment. However, neither the main effect of external referent networks nor the main effect of writing stages was found, <emph>F</emph><subs>(<reflink idref="bib1" id="ref119">1</reflink>,<reflink idref="bib98" id="ref120">98</reflink>)</subs> = 2.27, <emph>p</emph> = 0.135, η<sups>2</sups> =.019 (did not support H1b); <emph>F</emph><subs>(<reflink idref="bib2" id="ref121">2</reflink>,<reflink idref="bib196" id="ref122">196</reflink>)</subs> = 1.25, <emph>p</emph> = 0.271, η<sups>2</sups> =.001. Regarding the interaction effects, none of the two-way interactions reached significance, all <emph>p</emph>-values &gt;.05. However, the three-way interaction was found, <emph>F</emph><subs>(<reflink idref="bib2" id="ref123">2</reflink>,<reflink idref="bib196" id="ref124">196</reflink>)</subs> = 5.84, <emph>p</emph> =.015, η<sups>2</sups> =.007 (see Fig. 5a and b).</p> <p>Graph: Fig. 5 The proportion of integrative links (quantity of integration) and graph centrality similarity (GCsim, structural quality of network form) in Study 1</p> <p>Further simple effect analysis showed that for the integrative referent network condition, participants receiving the integrative prompts constructed more integrative links than the detailed prompts condition at both after-reading writing (<emph>F</emph><subs>(<reflink idref="bib1" id="ref125">1</reflink>,<reflink idref="bib50" id="ref126">50</reflink>)</subs> = 5.35, <emph>p</emph> =.025, η<sups>2</sups> =.097, 0.22 ± 0.16 vs. 0.13 ± 0.10) and at revised writing (<emph>F</emph><subs>(<reflink idref="bib1" id="ref127">1</reflink>,<reflink idref="bib50" id="ref128">50</reflink>)</subs> = 5.35, <emph>p</emph> =.025, η<sups>2</sups> =.097, 0.22 ± 0.15 vs. 0.14 ± 0.10), but not at delayed writing (<emph>F</emph><subs>(<reflink idref="bib1" id="ref129">1</reflink>,<reflink idref="bib50" id="ref130">50</reflink>)</subs> = 0.50, <emph>p</emph> = 0.483, η<sups>2</sups> =.010, 0.18 ± 0.15 vs. 0.16 ± 0.12). In contrast, for the intra-text referent network condition, participants receiving the integrative prompts only constructed more integrative links than the detailed prompts condition in the delayed writing (<emph>F</emph><subs>(<reflink idref="bib1" id="ref131">1</reflink>,<reflink idref="bib48" id="ref132">48</reflink>)</subs> = 4.77, <emph>p</emph> =.034, η<sups>2</sups> =.090, 0.19 ± 0.15 vs. 0.16 ± 0.11), but not in other stages.</p> <p>Second, regarding the similarity of integrative links (semantic quality), results showed no significant main effects or interactions, all <emph>p</emph>-values &gt;.05. Therefore, H2a and H2b were not supported.</p> <p>Third, regarding GC<subs>sim</subs> (network structural quality relative to the expert), results showed neither the main effect of prompts, <emph>F</emph><subs>(<reflink idref="bib1" id="ref133">1</reflink>,<reflink idref="bib98" id="ref134">98</reflink>)</subs> =.01, <emph>p</emph> = 0.937, η<sups>2</sups> =.001, nor the main effect of external referent network, <emph>F</emph><subs>(<reflink idref="bib1" id="ref135">1</reflink>,<reflink idref="bib98" id="ref136">98</reflink>)</subs> = 1.31, <emph>p</emph> = 0.255, η<sups>2</sups> =.005 (did not support H3a). In addition, results did not show any two-way or three-way interactions, all <emph>p</emph>-values &gt;.05. However, the main effect of writing stages was found, <emph>F</emph><subs>(<reflink idref="bib2" id="ref137">2</reflink>,<reflink idref="bib196" id="ref138">196</reflink>)</subs> = 140.33, <emph>p</emph> &lt;.001, η<sups>2</sups> = 0.371. Results of multiple comparisons under Bonferroni correction showed significantly higher GC<subs>sim</subs> in delayed writing (<emph>M</emph> = 0.48, <emph>SD</emph> = 0.28) than both after-reading writing (<emph>M</emph> = 0.20, <emph>SD</emph> =.08), <emph>t</emph><subs>(<reflink idref="bib98" id="ref139">98</reflink>)</subs> = 11.89, <emph>p</emph> &lt;.001, Cohen's <emph>d</emph> = 1.62, and revised writing (<emph>M</emph> = 0.20, <emph>SD</emph> =.08), <emph>t</emph><subs>(<reflink idref="bib98" id="ref140">98</reflink>)</subs> = 11.82, <emph>p</emph> &lt;.001, Cohen's <emph>d</emph> = 1.61, but a significant difference was not found for GC<subs>sim</subs> network when comparing revised writing and after-reading writing, <emph>t</emph><subs>(<reflink idref="bib98" id="ref141">98</reflink>)</subs> = 1.37, <emph>p</emph> = 0.524, Cohen's <emph>d</emph> =.01 (see Fig. 5c and d). These findings are indicative of increased node degree centrality of important terms (Clariana, [<reflink idref="bib20" id="ref142">20</reflink>]).</p> <hd id="AN0188049025-21">Discussion</hd> <p>Consistent with Wei et al. ([<reflink idref="bib89" id="ref143">89</reflink>]), a beneficial effect of the integrative prompts on the quantity of integration was found. Moreover, the integrative referent network triggered immediate facilitation of the quantity of integration (i.e., for the after-reading writing and revised writing in the first session) in the integrative prompts condition compared with the detailed prompts condition, but this group-level difference was not observed three days later (see Fig. 5a). In contrast, the intra-text referent network triggered a delayed inhibition (i.e., for the delayed writing in the second session) for the detailed prompts group (see Fig. 5b). Although there was an unexpectedly significant boost in the structural quality at the delayed writing stage for all participants (GC<subs>sim</subs> more like the expert, discussed later), no positive effects of expert referent networks were observed on semantic or structural quality of the essays. To consider delayed essay improvement, in Study 2, the expert network as personalized network feedback was provided immediately after the first essay to explore whether it helps students update the quality of final integration.</p> <hd id="AN0188049025-22">Study 2</hd> <p>So, in Study 2, instead of providing a pre-reading expert referent network, post-reading writing feedback was used to investigate how the quality of integration can be influenced. A 2 (prompts) × 2 (feedback: integrative feedback vs. intra-text feedback) × 3 (writing stage) mixed design was employed. The quantity of integration was expected to be improved by both the integrative prompts (H1a) and the integrative feedback (H1 d) and relevant interactions (H1e, especially at the revised and delayed writing). The semantic quality was expected to be improved by the integrative feedback (H4a) and relevant interactions (H4b). The structural quality was expected to improve with integrative feedback (H5a) and relevant interactions (H5b).</p> <hd id="AN0188049025-23">Methods</hd> <p></p> <hd id="AN0188049025-24">Participants</hd> <p>Ninety-four participants were recruited following the same requirements used in Study 1. Four were excluded for reporting a strong topic-related prior knowledge (<emph>M</emph><subs>rating of the content similarity</subs> = 2.50, <emph>SD</emph> = 1.25). The final sample included 90 participants (60 females, 30 males, age: <emph>M</emph> = 20.93, <emph>SD</emph> = 1.95). Consistent with Study 1, the recommended sample size in the prior power analysis was 64, and the obtained sample size of 90 was more than adequate to test the hypothesis.</p> <hd id="AN0188049025-25">Materials</hd> <p>One significant change in Study 2 was the use of automatically generated network-based writing feedback. Following the GIKS tool (Graphical Interface of Knowledge Structure Kim, [<reflink idref="bib38" id="ref144">38</reflink>]; Kim et al., [<reflink idref="bib40" id="ref145">40</reflink>]), after participants completed summary writing, a structural feedback screen (Trumpower &amp; Sarwar, [<reflink idref="bib81" id="ref146">81</reflink>]) was presented showing <emph>the full expert network</emph> on the left side and the participant's essay network on the right for side-by-side comparison (see Fig. 6).</p> <p>Graph: Fig. 6 The interface of the post-writing feedback network, where 1—Task instruction, 2—Expert Network, 3—Feedback network, 4—Continue Button</p> <p>To support this comparison, the spatial positions of the terms that occurred in the participant's essay exactly matched those in the full expert network. For the <emph>integrative feedback network</emph>, existing correct integrative links (green, 5 links possible), incorrect integrative links (red), and missing integrative links (orange) in the individual essay network were highlighted, while intra-text links were shown in gray. The color scheme switched for the <emph>intra-text feedback network,</emph> correct intra-text links (green, 5 links possible), incorrect intra-text links (red), and missing intra-text links (orange) in the individual's essay network were highlighted, while integrative links were shown in gray. The network feedback screen was not available once participants clicked the <emph>continue</emph> button, so they had to retain the key information from the feedback network when revising, similar to what participants must do for the pre-reading referent network learning in Study 1. The other materials and measures were identical to those in Study 1.</p> <hd id="AN0188049025-26">Procedure</hd> <p>The experimental procedure was like that in Study 1 except that the referent networks in Study 1 were used as after-reading writing feedback to support revised writing (refer back to Fig. 4). As a manipulation check, participants accomplished two questions on a five-point scale at the end of the first session. These questions were "How much do you understand the instruction of feedback" (1 = totally no, 5 = totally yes) and "Your tendency to revise your summary after feedback learning" (1 = totally no, 5 = totally yes). Those who did not revise their essays were requested to provide a short statement to explain why they did not.</p> <hd id="AN0188049025-27">Results</hd> <p>As observed in Study 1, no significant difference in ADKS was found among the four groups, <emph>F</emph><subs>(<reflink idref="bib3" id="ref147">3</reflink>,<reflink idref="bib86" id="ref148">86</reflink>)</subs> = 1.42, <emph>p</emph> = 0.242, η<sups>2</sups> =.047 (<emph>M</emph> = 0.65, <emph>SD</emph> = 0.14. The average score of understanding instruction of feedback was 3.73, <emph>SD</emph> = 0.83, and the average tendency of revision was 3.12, <emph>SD</emph> = 0.94. Sixty percent of participants revised their summaries, and those who did not revise provided acceptable explanations, such as being satisfied with their initial writing.</p> <hd id="AN0188049025-28">Comprehension of integrative knowledge</hd> <p>IVT scores were analyzed with a 2 (prompts) × 2 (feedback) two-way ANOVA. Levene's test indicated equal variances among groups, <emph>F</emph><subs>(<reflink idref="bib3" id="ref149">3</reflink>,<reflink idref="bib98" id="ref150">98</reflink>)</subs> = 1.81, <emph>p</emph> = 0.151. Consistent with Study 1, the ANOVA results showed no significant main effects or interaction (all <emph>p</emph>-values &gt;.05) and there was good comprehension for all groups: <emph>integrative prompts</emph> + <emph>integrative network</emph> (<emph>n</emph> = 22): <emph>M</emph> = 0.84, <emph>SD</emph> = 0.10; <emph>integrative prompts</emph> + <emph>intra-text network</emph> (<emph>n</emph> = 23): <emph>M</emph> = 0.81, <emph>SD</emph> = 0.17; <emph>detailed prompts</emph> + <emph>integrative network</emph> (<emph>n</emph> = 22): <emph>M</emph> = 0.82, <emph>SD</emph> = 0.14; <emph>detailed prompts</emph> + <emph>intra-text network</emph> (<emph>n</emph> = 23): <emph>M</emph> = 0.80, <emph>SD</emph> = 0.10.</p> <hd id="AN0188049025-29">Essay network quantity and quality</hd> <p>As in Study 1, the three integration measures (quantity of integration, semantic quality of integration, and structural quality of network form) were analyzed with a 2 (prompts) × 2 (feedback) × 3 (writing stages) repeated measures ANOVA, respectively.</p> <p>First, the proportion of integrative links (quantity of integration) was consistent with Study 1 (see Figs. 7a and b), results of the ANOVA showed a significant main effect of prompts, <emph>F</emph><subs>(<reflink idref="bib1" id="ref151">1</reflink>,<reflink idref="bib86" id="ref152">86</reflink>)</subs> = 5.37, <emph>p</emph> =.023, η<sups>2</sups> =.048 (supporting H1a). Participants in the integrative prompts condition (<emph>M</emph> = 0.23, <emph>SD</emph> = 0.13) showed more integrative links in their writing performance than the detailed prompts group did (<emph>M</emph> = 0.17, <emph>SD</emph> = 0.13). However, neither the main effect of feedback (<emph>F</emph><subs>(<reflink idref="bib1" id="ref153">1</reflink>,<reflink idref="bib86" id="ref154">86</reflink>)</subs> = 3.22, <emph>p</emph> =.076, η<sups>2</sups> =.029) (did not support H1 d) nor the main effect of writing stages was found (<emph>F</emph><subs>(<reflink idref="bib2" id="ref155">2</reflink>,<reflink idref="bib172" id="ref156">172</reflink>)</subs> = 0.90, <emph>p</emph> = 0.391, η<sups>2</sups> =.001). In addition, none of the two-way or three-way interactions were significant, all <emph>p</emph>-values &gt;.05 (did not support H1e).</p> <p>Graph: Fig. 7 The proportion of integrative links (quantity of integration) and structural quality (GCsim network form) in Study 2</p> <p>Second, as in Study 1 above, integrative links (semantic quality of integration) showed no significant main effects or interactions, all <emph>p</emph>-values &gt;.05. Therefore, H4a and H4b were not supported.</p> <p>Third, regarding the GC<subs>sim</subs> quality of network form (see Figs. 7c and d), results showed no significant main effects for prompts (<emph>F</emph><subs>(<reflink idref="bib1" id="ref157">1</reflink>,<reflink idref="bib86" id="ref158">86</reflink>)</subs> = 0.40, <emph>p</emph> = 0.530, η<sups>2</sups> =.001), or for feedback (<emph>F</emph><subs>(<reflink idref="bib1" id="ref159">1</reflink>,<reflink idref="bib86" id="ref160">86</reflink>)</subs> = 0.49, <emph>p</emph> = 0.486, η<sups>2</sups> =.002) (did not support H5a). However, consistent with Study 1, there was a significant main effect of writing stages, <emph>F</emph><subs>(<reflink idref="bib2" id="ref161">2</reflink>,<reflink idref="bib172" id="ref162">172</reflink>)</subs> = 147.22, <emph>p</emph> &lt;.001, η<sups>2</sups> = 0.423 and the two-way interaction between feedback and writing stages was significant, <emph>F</emph><subs>(<reflink idref="bib2" id="ref163">2</reflink>,<reflink idref="bib172" id="ref164">172</reflink>)</subs> = 4.09, <emph>p</emph> =.043, η<sups>2</sups> =.012. No other findings were significant, all <emph>p</emph>-values &gt;.05. Results of multiple comparisons under Bonferroni correction showed significantly higher GC<subs>sim</subs> for delayed writing (<emph>M</emph> = 0.53, <emph>SD</emph> =.03) than both after-reading writing (<emph>M</emph> = 0.21, <emph>SD</emph> =.01), <emph>t</emph><subs>(<reflink idref="bib86" id="ref165">86</reflink>)</subs> = 12.34, <emph>p</emph> &lt;.001, Cohen's <emph>d</emph> = 1.82, and revised writing (<emph>M</emph> = 0.22, <emph>SD</emph> =.01), <emph>t</emph><subs>(<reflink idref="bib86" id="ref166">86</reflink>)</subs> = 12.19, <emph>p</emph> &lt;.001, Cohen's <emph>d</emph> = 1.78, but a significant difference was not found between GC<subs>sim</subs> network form when comparing after-reading writing with revised writing, <emph>t</emph><subs>(<reflink idref="bib86" id="ref167">86</reflink>)</subs> = 1.39, <emph>p</emph> = 0.508, Cohen's <emph>d</emph> =.04.</p> <hd id="AN0188049025-30">Discussion</hd> <p>The direction and magnitude of means in Study 1 and Study 2 are virtually identical (compare Figs. 5 and 7). As in Study 1, in Study 2 integrative prompts were better than detailed prompts, and the quantity and semantic quality of integration were not affected by the network feedback manipulations. As in Study 1, there was a substantial and significant increase in GC<subs>sim</subs> quality of network form from session 1 to session 2 (i.e., delayed writing three days later). Considering the findings of the two studies, the semantic quality of multiple document integration might be more difficult to change than its structural quality, although both measures can be seen as aspects of the semantic quality of integration. This should not be seen as a general conclusion because the difficulty of semantic/structural processing can be altered by text materials, so more empirical evidence is needed. But still, these results show the complexity of the quality of knowledge integration.</p> <hd id="AN0188049025-31">General discussion</hd> <p>This investigation answers the call by Spector et al. ([<reflink idref="bib77" id="ref168">77</reflink>]) for replication studies to support scientific rigor. This investigation replicates and extends an investigation by Wei et al. ([<reflink idref="bib89" id="ref169">89</reflink>]) regarding the effects of prompts-activated purposeful reading, pre-reading expert referent networks (Study 1), and post-reading network-based writing feedback (Study 2) on the quantity and quality of multiple document integration. As expected and consistent with Wei et al. ([<reflink idref="bib89" id="ref170">89</reflink>]), pre-reading integrative prompts improved the quantity of integration in both studies (i.e., η<sups>2</sups> values are around 0.50). The observed effects of pre-reading prompts on the quantity of multiple document integration indicate that reading purpose changes individuals' cognitive strategies when reading to process more purpose-relevant information (McCrudden &amp; Schraw, [<reflink idref="bib61" id="ref171">61</reflink>]; McCrudden et al., [<reflink idref="bib59" id="ref172">59</reflink>]). Also, structural quality relative to the expert network significantly increased in Session 2 three days later (at delayed writing) compared to their Session 1 essays (right after reading) in both Study 1 and 2.</p> <hd id="AN0188049025-32">The integration performance in generated learning outcomes</hd> <p>In this investigation, both reading purpose and network were intended to establish a <emph>task model</emph> that is either integrative or detailed, however, for participants in Study 1, the <emph>integrative prompts</emph> + <emph>intra-text network</emph> group and the <emph>detailed prompts</emph> + <emph>intertext network</emph> group received opposite task cues before reading. This contradictory <emph>task model</emph> could imply that readers should attend to both global and local details. Tal-Savir et al. ([<reflink idref="bib79" id="ref173">79</reflink>]) notes that there is "... a potential trade-off between elaboration [intra-text] and integration [inter-text] processes in multiple document comprehension that may require instructional attention and support" (p. 605). Our results indicate that readers who receive contradictory pre-reading cues primarily follow the textual prompt more than the network prompt. As described in greater detail later, the network prompt did have a different kind of influence three days later in both Studies 1 and 2 on the quality of network form (GC<subs>sim</subs> increased) from Session 1 to Session 2, and also relative to the previous findings from Wei et al. ([<reflink idref="bib89" id="ref174">89</reflink>]).</p> <p>Unexpectedly, there was no main effect of pre-reading expert network or network-based feedback on semantic quality. Previous research on single document comprehension has reported that network-based feedback improves the overall similarity with the expert network from initial writing to revised writing (Kim &amp; McCarthy, [<reflink idref="bib41" id="ref175">41</reflink>], [<reflink idref="bib42" id="ref176">42</reflink>]). Why is there no improvement after revision in this multiple document setting? A possible reason is that the approach we used differs slightly from past studies. Here, the student's essay network was presented with colorful links alongside the full expert reference to indicate similarities/differences, so perhaps this display is too complex to interpret without further instruction. Or perhaps students did grasp their essay network structural similarities/differences with the expert network but were unable or unwilling to immediately revise their essays, given that this experimental setting differs from a real learning context used in those other studies.</p> <p>Future research should continue to seek effective interventions since the results here show that reading prompts only affect the quantity, but that quality can be difficult to influence by prompts or by an expert network reference. Two possible aspects may be valuable. First, students may be <emph>unable</emph> to update their knowledge from network feedback due to strong misconceptions in their minds (Butterfuss &amp; Kendeou, [<reflink idref="bib12" id="ref177">12</reflink>]). Inter-text links in multiple texts are usually implicit and cannot be obtained directly without inference, so students may be highly confident in their misconceptions and resist changing them (Butler et al., [<reflink idref="bib11" id="ref178">11</reflink>]; van Loon et al., [<reflink idref="bib84" id="ref179">84</reflink>]). To improve quality, future research could use a refutation text strategy to correct misconceptions and thus permit knowledge updates (Bråten et al., [<reflink idref="bib7" id="ref180">7</reflink>]; Kim &amp; Kendeou, [<reflink idref="bib37" id="ref181">37</reflink>]).</p> <p>Second, participants in our investigations might feel exhausted after learning multiple texts and thus were <emph>unwilling</emph> to expend extra effort in an experimental setting, which differs from a real learning context that is filled with academic pressure and the need for achievement. Personal factors, such as reading interest, motivation, engagement, and experience with learning technology, should also be incorporated into the investigation to evaluate how integration can be further improved (List &amp; Alexander, [<reflink idref="bib54" id="ref182">54</reflink>]).</p> <hd id="AN0188049025-33">The structural change of integration in knowledge re-generation</hd> <p>In both Study 1 and Study 2 the essay network structural quality of integration for delayed writing three days later (GC<subs>sim</subs> relative to the expert network) was substantially larger in Session 2 (GC<subs>sim</subs> means range from 0.45 to 0.59) than in Session 1 (GC<subs>sim</subs> means range from 0.20 to 0.25). However, this increase was not observed in Wei et al. ([<reflink idref="bib89" id="ref183">89</reflink>]), where the average GC<subs>sim</subs> values for the delayed writing were 0.23 and 0.24, and for the delayed concept mapping, 0.35 and 0.36. Note that participants were not given an expert network in that investigation. This suggests that <emph>the expert network</emph>, whether given before reading or as writing feedback, <emph>influences the conceptual structure of delayed essays</emph>, probably as higher frequencies of key terms (Clariana, [<reflink idref="bib20" id="ref184">20</reflink>]).</p> <p>Inspection of the delayed quality of writing essay network form (GC<subs>sim</subs>) in Study 1 and Study 2 (see Fig. 8) suggests an interaction of network structure (integrative vs. inter-text) and whether the network is presented before reading (Study 1) or after writing (Study 2). When the network was presented before reading (Study 1), the integrative (inter-text) network form was more like the expert than the intra-text network. This pattern was reversed when the network was presented as summary writing feedback (Study 2), the intra-text network form (GC<subs>sim</subs>) was more like the expert than was the integrative (inter-text) network. This pattern is consistent with both detailed prompts and integrative prompts.</p> <p>Graph: Fig. 8 Delayed writing essay network structural quality (GCsim network form) in Study 1 and Study 2</p> <p>Wang et al. ([<reflink idref="bib88" id="ref185">88</reflink>]) noted that, at least for concept map network form, low-knowledge students and non-experts tend to generate hierarchically structured maps (e.g., Graph Centrality, GC, range of 0.30—0.50) that suggests that there is a default conceptual mental network conceptual form, while high-knowledge students and experts draw maps with complex web forms such as radial, cross (two intersecting hierarchical structures) or three-dimensional forms (Ayersman, [<reflink idref="bib2" id="ref186">2</reflink>]; Wang, [<reflink idref="bib87" id="ref187">87</reflink>]). In other words, the content itself has an inherent network form that the expert approximates. Thus, comparisons of students' essay networks to the expert (i.e., measured as GC<subs>sim</subs> here) seem reasonable. More research is needed to better determine how to measure this form difference and to establish whether measures of essay network form add to our theoretical understanding of how knowledge building leads to the progression of network structures as transitions towards expert knowledge.</p> <p>Freeman's graph centrality (GC) values for the three texts we used were 0.37, 0.41, and 0.49 (<emph>hierarchical</emph>) and of the full expert network is 0.60 (<emph>network</emph>), so the delayed writing essay network structural form (i.e., 0.30 in Study 1, and 0.31 in Study 2) has become <emph>more like</emph> the texts read and the expert's network form (compared with the after-reading network from, which is 0.12 in average in both Study 1 and Study 2). Wei et al. ([<reflink idref="bib89" id="ref188">89</reflink>]) stated that....the visuospatial nature of a concept map [sic or network] provides a concise conceptual placeholder for adding information from succeeding documents, while summary writing due to its linear-sequential nature may be better for within-document processing but may be less effective for cross-document processing. (p. 665)</p> <p>Thus, it was necessary for time to pass (three days here) before the content was reorganized and re-generated. Further research is needed to establish whether time for memory consolidation is required for multiple document integration.</p> <hd id="AN0188049025-34">Limitations and generalizability</hd> <p>Freeman's graph centrality measure is a new application of an old metric that is not yet fully validated for this purpose. The findings here for graph centrality are interesting and reasonable and align with and support those of Kim and McCarthy ([<reflink idref="bib42" id="ref189">42</reflink>]), however, there remain unknowns concerning the nature of this type of data.</p> <p>Next, using network measures and comparison approaches from single document studies might not be completely suitable for multiple documents setting. The requirement of the similarity of integrative links might be too stringent for many reasons, for example, due to issues such as using different terms for the same concept across documents. Perhaps other network estimation approaches, such as a sorting task, could be developed and applied (Clariana et al., [<reflink idref="bib23" id="ref190">23</reflink>]).</p> <p>Another limitation of our and many other multiple document studies is the use of self-made comprehension tests designed for specific reading materials. Non-standardized tools often lead to low reliability and a lack of validity, though these tests can provide preliminary and exploratory results of the cognitive mechanism of multiple text reading. Future research should focus more on the psychometric properties of comprehension tests.</p> <p>For theory, this investigation supports the original study by Wei et al. ([<reflink idref="bib89" id="ref191">89</reflink>]) regarding the influence of integrative reading purpose in multiple document comprehension, and that concept network analysis provides another lens for measuring and interpreting learning in multiple document settings as complex, multi-dimensional networks. Clearly, the memory of the network structure of the content and declarative knowledge about the content are internally related but distinct in memory, very much like gist and verbatim memories in Fuzzy Trace Theory (Brainerd &amp; Reyna, [<reflink idref="bib4" id="ref192">4</reflink>]; Reyna, [<reflink idref="bib88" id="ref193">88</reflink>]).</p> <p>For educational practice, teachers and instructors can use integrative prompts to scaffold the quantity of multiple text reading, combined with other approaches such as mapping and repeated, delayed writing that can support quality. Continued research on reading purpose regarding different aspects of text comprehension is promising and warranted.</p> <hd id="AN0188049025-35">Acknowledgements</hd> <p>Grant from Research Center for Brain Cognition and Human Development, Guangdong, China (No. 2024B0303390003); Striving for the First-Class, Improving Weak Links and Highlighting Features (SIH) Key Discipline for Psychology in South China Normal University</p> <hd id="AN0188049025-36">Funding</hd> <p>This work was supported by a grant from Research Center for Brain Cognition and Human Development, Guangdong, China (No. 2024B0303390003). And this material is based upon work supported by the U.S. National Science Foundation under award No. (<reflink idref="bib2" id="ref194">2</reflink>,<reflink idref="bib215" id="ref195">215</reflink>,<reflink idref="bib807" id="ref196">807</reflink>). Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the U.S. National Science Foundation.</p> <hd id="AN0188049025-37">Data availability</hd> <p>The data that support the findings of this study are available from the corresponding author upon reasonable request.</p> <hd id="AN0188049025-38">Publisher's Note</hd> <p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p> <ref id="AN0188049025-39"> <title> References </title> <blist> <bibl id="bib1" idref="ref60" type="bt">1</bibl> <bibtext> Ausubel DP, Fitzgerald D. The role of discriminability in meaningful learning and retention. Journal of Educational Psychology. 1961; 52; 5: 266-274. 10.1037/h0045701</bibtext> </blist> <blist> <bibl id="bib2" idref="ref78" type="bt">2</bibl> <bibtext> Ayersman DJ. Effects of knowledge representation format and hypermedia instruction on metacognitive accuracy. Computers in Human Behavior. 1995; 11; 3–4: 533-555. 10.1016/0747-5632(95)80016-2</bibtext> </blist> <blist> <bibl id="bib3" idref="ref3" type="bt">3</bibl> <bibtext> Ayroles J, Potocki A, Ros C, Raquel Cerdán M, Britt A, Rouet J-F. Do you know what you are reading for? Exploring the effects of a task model enhancement on fifth graders' purposeful reading. Journal of Research in Reading. 2021; 44; 4: 837-858. 10.1111/1467-9817.12374</bibtext> </blist> <blist> <bibl id="bib4" idref="ref192" type="bt">4</bibl> <bibtext> Brainerd CJ, Reyna VF. Memory independence and memory interference in cognitive development. Psychological Review. 1993; 100: 42-67. 10.1037/0033-295X.100.1.42</bibtext> </blist> <blist> <bibl id="bib5" idref="ref48" type="bt">5</bibl> <bibtext> Brand-Gruwel S, Kammerer Y, Van Meeuwen L, Van Gog T. Source evaluation of domain experts and novices during web search. Journal of Computer Assisted Learning. 2017; 33; 3: 234-251. 10.1111/jcal.12162</bibtext> </blist> <blist> <bibl id="bib6" idref="ref30" type="bt">6</bibl> <bibtext> Brante EW, Strømsø HI. Sourcing in text comprehension: A review of interventions targeting sourcing skills. Educational Psychology Review. 2018; 30; 3: 773-799. 10.1007/s10648-017-9421-7</bibtext> </blist> <blist> <bibl id="bib7" idref="ref180" type="bt">7</bibl> <bibtext> Bråten I, Brandmo C, Ferguson LE, Strømsø HI. Epistemic justification in multiple document literacy: A refutation text intervention. Contemporary Educational Psychology. 2022; 71: 102122. 10.1016/j.cedpsych.2022.102122</bibtext> </blist> <blist> <bibl id="bib8" idref="ref31" type="bt">8</bibl> <bibtext> Bråten I, Strømsø HI, Britt MA. Trust matters: Examining the role of source evaluation in students' construction of meaning within and across multiple texts. Reading Research Quarterly. 2009; 44; 1: 6-28. 10.1598/rrq.44.1.1</bibtext> </blist> <blist> <bibl id="bib9" idref="ref20" type="bt">9</bibl> <bibtext> Britt MA, Rouet J-F, Durik A. Literacy beyond text comprehension: A theory of purposeful reading. 2017; Routledge. 10.4324/9781315682860</bibtext> </blist> <blist> <bibtext> Burkhart C, Lachner A, Nuckles M. Assisting students' writing with computer-based concept map feedback: A validation study of the CohViz feedback system. PLoS ONE. 2020; 15; 6: e0235209. 10.1371/journal.pone.0235209</bibtext> </blist> <blist> <bibtext> Butler AC, Fazio LK, Marsh EJ. The hypercorrection effect persists over a week, but high-confidence errors return. Psychonomic Bulletin &amp; Review. 2011; 18; 6: 1238-1244. 10.3758/s13423-011-0173-y</bibtext> </blist> <blist> <bibtext> Butterfuss R, Kendeou P. KReC-MD: Knowledge revision with multiple documents. Educational Psychology Review. 2021; 33; 4: 1475-1497. 10.1007/s10648-021-09603-y</bibtext> </blist> <blist> <bibtext> Carpenter BD, Balsis S, Otilingam PG, Hanson PK, Gatz M. The Alzheimer's disease knowledge scale: Development and psychometric properties. The Gerontologist. 2009; 49; 2: 236-247. 10.1093/geront/gnp023</bibtext> </blist> <blist> <bibtext> Casteleyn J, Mottart A, Valcke M. The impact of graphic organisers on learning from presentations. Technology, Pedagogy and Education. 2013; 22; 3: 283-301. 10.1080/1475939X.2013.784621</bibtext> </blist> <blist> <bibtext> Cerdán R, Vidal-Abarca E. The effects of tasks on integrating information from multiple documents. Journal of Educational Psychology. 2008; 100; 1: 209-222. 10.1037/0022-0663.100.1.209</bibtext> </blist> <blist> <bibtext> Chang KE, Sung YT, Chen ID. The effect of concept mapping to enhance text comprehension and summarization. Journal of Experimental Education. 2002; 71; 1: 5-23. 10.1080/00220970209602054</bibtext> </blist> <blist> <bibtext> Chen X, Wei Z, Li Z, Clariana RB. The influence of the conceptual structure of external representations when relearning history content. Educational Technology Research and Development. 2023; 71; 2: 415-439. 10.1007/s11423-022-10176-y</bibtext> </blist> <blist> <bibtext> Cheong CM, Zhu X, Li GY, Wen H. Effects of intertextual processing on L2 integrated writing. Journal of Second Language Writing. 2019; 44: 63-75. 10.1016/j.jslw.2019.03.004</bibtext> </blist> <blist> <bibtext> Clariana, R. B. (2024). An OER Tool for Writing-to-Learn in Undergraduate STEM Courses ISTELive 24 - Edtech conference</bibtext> </blist> <blist> <bibtext> Clariana RBIsaías P, Ifenthaler D. Framing knowledge as conceptual structure. Computer-based diagnostics and systematic analysis of knowledge. 2025; Springer</bibtext> </blist> <blist> <bibtext> Clariana RB, Koul RCañas AJ, Novak JD, García FMG. A computer-based approach for translating text into concept map-like representations. Concept maps: theory, methodology, technology: Proceedings of the first international conference on concept mapping. 2004; Universidad Pública de Navarra</bibtext> </blist> <blist> <bibtext> Clariana RB, Rysavy MD, Taricani E. Text signals influence team artifacts. Educational Technology Research and Development. 2015; 63; 1: 35-52. 10.1007/s11423-014-9362-5</bibtext> </blist> <blist> <bibtext> Clariana RB, Tang H, Chen X. Corroborating a sorting task measure of individual and of local collective knowledge structure. Educational Technology Research and Development. 2022; 70; 4: 1195-1219. 10.1007/s11423-022-10123-x</bibtext> </blist> <blist> <bibtext> Clariana RB, Wallace PE, Godshalk VM. Deriving and measuring group knowledge structure from essays: The effects of anaphoric reference. Educational Technology Research and Development. 2009; 57; 6: 725-737. 10.1007/s11423-009-9115-z</bibtext> </blist> <blist> <bibtext> Cohen J. Statistical power analysis for the behavioral sciences. 19882; Routledge</bibtext> </blist> <blist> <bibtext> Davies M. Concept mapping, mind mapping and argument mapping: What are the differences and do they matter?. Higher Education. 2011; 62; 3: 279-301. 10.1007/s10734-010-9387-6</bibtext> </blist> <blist> <bibtext> Fan C-Y, Chen G-D. A scaffolding tool to assist learners in argumentative writing. Computer Assisted Language Learning. 2021; 34; 1–2: 159-183. 10.1080/09588221.2019.1660685</bibtext> </blist> <blist> <bibtext> Follmer DJ, Fang S-Y, Clariana RB, Meyer BJF, Li P. What predicts adult readers' understanding of STEM texts?. Reading and Writing. 2018; 31; 1: 185-214. 10.1007/s11145-017-9781-x</bibtext> </blist> <blist> <bibtext> Freeman, L. C. (1977). A set of measures of centrality based on betweenness. Sociometry, 35–41. https://doi.org/10.2307/3033543</bibtext> </blist> <blist> <bibtext> Gil L, Bråten I, Vidal-Abarca E, Strømsø HI. Summary versus argument tasks when working with multiple documents: Which is better for whom?. Contemporary Educational Psychology. 2010; 35; 3: 157-173. 10.1016/j.cedpsych.2009.11.002</bibtext> </blist> <blist> <bibtext> Hawk PP. Using graphic organizers to increase achievement in middle school life science. Science Education. 1986; 70; 1: 81-87. 10.1002/sce.3730700110</bibtext> </blist> <blist> <bibtext> Ifenthaler D. Toward automated computer-based visualization and assessment of team-based performance. Journal of Educational Psychology. 2014; 106; 3: 651-665. 10.1037/a0035505</bibtext> </blist> <blist> <bibtext> Jonassen DH, Beissner K, Yacci M. Structural knowledge: Techniques for representing, conveying, and acquiring structural knowledge. 1993; Lawrence Erlbaum Associates, Inc</bibtext> </blist> <blist> <bibtext> Kang SHK, McDermott KB, Roediger HL. Test format and corrective feedback modify the effect of testing on long-term retention. European Journal of Cognitive Psychology. 2007; 19; 4–5: 528-558. 10.1080/09541440601056620</bibtext> </blist> <blist> <bibtext> Karimi MN. The mediated/unmediated contributions of language proficiency and prior knowledge to L2 multiple-texts comprehension: A structural equation modelling analysis. Applied Linguistics. 2018; 39; 6: 912-932. 10.1093/applin/amw059</bibtext> </blist> <blist> <bibtext> Kentgen L, Allen R, Kose G, Fong R. The effects of rerepresentation on future performance. British Journal of Developmental Psychology. 1998; 16; 4: 505-517. 10.1111/j.2044-835X.1998.tb00768.x</bibtext> </blist> <blist> <bibtext> Kim J, Kendeou P. Knowledge transfer in the context of refutation texts. Contemporary Educational Psychology. 2021; 67: 102002. 10.1016/j.cedpsych.2021.102002</bibtext> </blist> <blist> <bibtext> Kim K. Graphical interface of knowledge structure: A web-based research tool for representing knowledge structure in text. Technology, Knowledge and Learning. 2019; 24; 1: 89-95. 10.1007/s10758-017-9321-4</bibtext> </blist> <blist> <bibtext> Kim K, Clariana RB. Text signals influence second language expository text comprehension: Knowledge structure analysis. Educational Technology Research and Development. 2017; 65; 4: 909-930. 10.1007/s11423-016-9494-x</bibtext> </blist> <blist> <bibtext> Kim K, Clariana RB, Kim Y. Automatic representation of knowledge structure: Enhancing learning through knowledge structure reflection in an online course. Educational Technology Research and Development. 2019; 67; 1: 105-122. 10.1007/s11423-018-9626-6</bibtext> </blist> <blist> <bibtext> Kim M, McCarthy KS. Improving summary writing through formative feedback in a technology-enhanced learning environment. Journal of Computer Assisted Learning. 2020; 37; 3: 684-704. 10.1111/jcal.12516</bibtext> </blist> <blist> <bibtext> Kim M, McCarthy KS. Using graph centrality as a global index to assess students' mental model structure development during summary writing. Educational Technology Research and Development. 2021; 69; 2: 971-1002. 10.1007/s11423-021-09942-1</bibtext> </blist> <blist> <bibtext> Kintsch W. The role of knowledge in discourse comprehension: A construction-integration model. Psychological Review. 1988; 95; 2: 163-182. 10.1037/0033-295X.95.2.163</bibtext> </blist> <blist> <bibtext> Kintsch W. Comprehension: A paradigm for cognition. 1998; Cambridge University Press</bibtext> </blist> <blist> <bibtext> Lachner A, Backfisch I, Nückles M. Does the accuracy matter? Accurate concept map feedback helps students improve the cohesion of their explanations. Educational Technology Research and Development. 2018; 66; 5: 1051-1067. 10.1007/s11423-018-9571-4</bibtext> </blist> <blist> <bibtext> Lachner A, Burkhart C, Nückles M. Mind the gap! Automated concept map feedback supports students in writing cohesive explanations. Journal of Experimental Psychology: Applied. 2017; 23; 1: 29-46. 10.1037/xap0000111</bibtext> </blist> <blist> <bibtext> Lachner A, Neuburg C. Learning by writing explanations: Computer-based feedback about the explanatory cohesion enhances students' transfer. Instructional Science. 2018; 47; 1: 19-37. 10.1007/s11251-018-9470-4</bibtext> </blist> <blist> <bibtext> Latini N, Bråten I, Anmarkrud Ø, Salmerón L. Investigating effects of reading medium and reading purpose on behavioral engagement and textual integration in a multiple text context. Contemporary Educational Psychology. 2019; 59: 101797. 10.1016/j.cedpsych.2019.101797</bibtext> </blist> <blist> <bibtext> Lehmann T, Pirnay-Dummer P, Schmidt-Borcherding F. Fostering integrated mental models of different professional knowledge domains: Instructional approaches and model-based analyses. Educational Technology Research and Development. 2020; 68; 3: 905-927. 10.1007/s11423-019-09704-0</bibtext> </blist> <blist> <bibtext> Lehmann T, Rott B, Schmidt-Borcherding F. Promoting pre-service teachers' integration of professional knowledge: Effects of writing tasks and prompts on learning from multiple documents. Instructional Science. 2019; 47; 1: 99-126. 10.1007/s11251-018-9472-2</bibtext> </blist> <blist> <bibtext> Leon JA, Perez O. The influence of prior knowledge on the time course of clinical diagnosis inferences: A comparison of experts and novices. Discourse Processes. 2001; 31; 2: 187-213. 10.1207/S15326950dp3102_04</bibtext> </blist> <blist> <bibtext> Leopold C, Sumfleth E, Leutner D. Learning with summaries: Effects of representation mode and type of learning activity on comprehension and transfer. Learning and Instruction. 2013; 27: 40-49. 10.1016/j.learninstruc.2013.02.003</bibtext> </blist> <blist> <bibtext> Linderholm T, van den Broek P. The effects of reading purpose and working memory capacity on the processing of expository text. Journal of Educational Psychology. 2002; 94; 4: 778-784. 10.1037/0022-0663.94.4.778</bibtext> </blist> <blist> <bibtext> List A, Alexander PA. Cognitive affective engagement model of multiple source use. Educational Psychologist. 2017; 52; 3: 182-199. 10.1080/00461520.2017.1329014</bibtext> </blist> <blist> <bibtext> List A, Du H, Wang Y, Lee HY. Toward a typology of integration: Examining the documents model framework. Contemporary Educational Psychology. 2019; 58: 228-242. 10.1016/j.cedpsych.2019.03.003</bibtext> </blist> <blist> <bibtext> Mason AE, Braasch JL, Greenberg D, Kessler ED, Allen LK, McNamara DS. Comprehending multiple controversial texts about childhood vaccinations: Topic beliefs and integration instructions. Reading Psychology. 2023; 44; 4: 436-462. 10.1080/02702711.2022.2156952</bibtext> </blist> <blist> <bibtext> McCarthy KS, McNamara DS. The multidimensional knowledge in text comprehension framework. Educational Psychologist. 2021; 56; 3: 196-214. 10.1080/00461520.2021.1872379</bibtext> </blist> <blist> <bibtext> McCrudden MT, Kulikowich JM, Lyu B, Huynh L. Promoting integration and learning from multiple complementary texts. Journal of Educational Psychology, No Pagination Specified-No Pagination Specified. 2022. 10.1037/edu0000746</bibtext> </blist> <blist> <bibtext> McCrudden MT, Magliano JP, Schraw G. Exploring how relevance instructions affect personal reading intentions, reading goals and text processing: A mixed methods study. Contemporary Educational Psychology. 2010; 35; 4: 229-241. 10.1016/j.cedpsych.2009.12.001</bibtext> </blist> <blist> <bibtext> McCrudden MT, Schraw G. Relevance and goal-focusing in text processing. Educational Psychology Review. 2006; 19; 2: 113-139. 10.1007/s10648-006-9010-7</bibtext> </blist> <blist> <bibtext> McCrudden MT, Schraw G. The effects of relevance instructions and verbal ability on text processing. Journal of Experimental Education. 2010; 78; 1: 96-117. 10.1080/00220970903224529</bibtext> </blist> <blist> <bibtext> McNamara DS, Kintsch W. Learning from texts: Effects of prior knowledge and text coherence. Discourse Processes. 1996; 22; 3: 247-288. 10.1080/01638539609544975</bibtext> </blist> <blist> <bibtext> McNamara DS, Watanabe M, Huynh L, McCarthy KS, Allen LK, Magliano JP. Summarizing versus rereading multiple documents. Contemporary Educational Psychology. 2023; 76: 102238. 10.1016/j.cedpsych.2023.102238</bibtext> </blist> <blist> <bibtext> Nesbit JC, Adesope OO. Learning with concept and knowledge maps: A meta-analysis. Review of Educational Research. 2016; 76; 3: 413-448. 10.3102/00346543076003413</bibtext> </blist> <blist> <bibtext> Nesbit, J, Larios, H. &amp; Adesope, O. (2007). How students read concept maps: A study of eye movements. In C. Montgomerie &amp; J. Seale (Eds.), Proceedings of ED-MEDIA 2007--World conference on educational multimedia, hypermedia &amp; telecommunications (pp. 3961–3970). Association for the Advancement of Computing in Education (AACE). Retrieved from https://<ulink href="http://www.learntechlib.org/primary/p/25950/">www.learntechlib.org/primary/p/25950/</ulink></bibtext> </blist> <blist> <bibtext> Parkinson MM, Dinsmore DL. Multiple aspects of high school students' strategic processing on reading outcomes: The role of quantity, quality, and conjunctive strategy use. British Journal of Educational Psychology. 2018; 88; 1: 42-62. 10.1111/bjep.12176</bibtext> </blist> <blist> <bibtext> Perfetti, C. A, Rouet, J.-F, &amp; Britt, M. A. (1999). Toward a theory of documents representation. In H. van Oostendorp &amp; S. R. Goldman (Eds.), The construction of mental representations during reading (pp. 99–122). Lawrence Erlbaum Associates Publishers. Retrieved from: https://psycnet.apa.org/record/1998-06740-004</bibtext> </blist> <blist> <bibtext> Pirnay-Dummer P, Ifenthaler DIfenthaler D, Pirnay-Dummer P, Seel NM. Automated knowledge visualization and assessment. Computer-based diagnostics and systematic analysis of knowledge. 2010; Springer</bibtext> </blist> <blist> <bibtext> Pirnay-Dummer P, Ifenthaler DIfenthaler D, Spector JM, Isaias P, Sampson D. Text-guided automated self assessment. Multiple perspectives on problem solving and learning in the digital age. 2011; Springer: 217-225</bibtext> </blist> <blist> <bibtext> Primor L, Yeari M, Katzir T. Choosing the right question: The effect of different question types on multiple text integration. Reading and Writing. 2021; 34; 6: 1539-1567. 10.1007/s11145-021-10127-8</bibtext> </blist> <blist> <bibtext> Reyna VF. A new intuitionism: Meaning, memory, and development in Fuzzy-Trace Theory. Judgment and Decision Making. 2023; 7; 3: 332-359</bibtext> </blist> <blist> <bibtext> Rouet, J. F, &amp; Britt, M. A. (2011). Relevance processes in multiple document comprehension. In M. T. McCrudden, J. P. Magliano, &amp; G. Schraw (Eds.), Text relevance and learning from text (pp. 19–52). IAP Information Age Publishing. Retrieved from: https://psycnet.apa.org/record/2011-20070-002</bibtext> </blist> <blist> <bibtext> Rouet J-F, Britt MA, Durik AM. RESOLV: Readers' representation of reading contexts and tasks. Educational Psychologist. 2017; 52; 3: 200-215. 10.1080/00461520.2017.1329015</bibtext> </blist> <blist> <bibtext> Royer JM, Hastings CN, Hook C. A sentence verification technique for measuring reading comprehension. Journal of Reading Behavior. 1979; 11; 4: 355-363. 10.1080/10862967909547341</bibtext> </blist> <blist> <bibtext> Schvaneveldt, R. W, Durso, F. T, &amp; Dearholt, D. W. (1989). Network structures in proximity data. In Psychology of learning and motivation (Vol. 24, pp. 249–284). Elsevier. https://doi.org/10.1016/S0079-7421(08)60539-3</bibtext> </blist> <blist> <bibtext> Schvaneveldt RW, Dearholt DW, Durso FT. Graph theoretic foundations of pathfinder networks. Computers &amp; Mathematics with Applications. 1988; 15; 4: 337-345. 10.1016/0898-1221(88)90221-0</bibtext> </blist> <blist> <bibtext> Spector JM, Johnson TE, Young PA. An editorial on replication studies and scaling up efforts. Educational Technology Research and Development. 2015; 63: 1-4. 10.1007/s11423-014-9364-3</bibtext> </blist> <blist> <bibtext> Strømsø HI, Bråten I, Britt MA, Ferguson LE. Spontaneous sourcing among students reading multiple documents. Cognition and Instruction. 2013; 31; 2: 176-203. 10.1080/07370008.2013.769994</bibtext> </blist> <blist> <bibtext> Tal-Savir D, Barzilai S, Abed F, Mor-Hagani S, Zohar ARde Vries E, Hod Y, Ahn J. Scaffolding multiple document comprehension: Students' representations of documents models. ISLS Annual meeting 2021 reflecting the past and embracing the future - 15th international conference of the learning sciences, ICLS 2021. 2021; International Society of the Learning Sciences (ISLS)</bibtext> </blist> <blist> <bibtext> Tarchi C, Mason L. Effects of critical thinking on multiple-document comprehension. European Journal of Psychology of Education. 2020; 35; 2: 289-313. 10.1007/s10212-019-00426-8</bibtext> </blist> <blist> <bibtext> Trumpower DL, Sarwar GS. Effectiveness of structural feedback provided by Pathfinder networks. Journal of Educational Computing Research. 2010; 43; 1: 7-24. 10.2190/EC.43.1.b</bibtext> </blist> <blist> <bibtext> Tversky A. Features of similarity. Psychological Review. 1977; 84; 4: 327-352. 10.1037/0033-295X.84.4.327</bibtext> </blist> <blist> <bibtext> Unsworth N, McMillan BD. Mind wandering and reading comprehension: Examining the roles of working memory capacity, interest, motivation, and topic experience. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2013; 39; 3: 832-842. 10.1037/a0029669</bibtext> </blist> <blist> <bibtext> van Loon MH, Dunlosky J, van Gog T, van Merrienboer JJG, de Bruin ABH. Refutations in science texts lead to hypercorrection of misconceptions held with high confidence. Contemporary Educational Psychology. 2015; 42: 39-48. 10.1016/j.cedpsych.2015.04.003</bibtext> </blist> <blist> <bibtext> van Peppen LM, Verkoeijen P, Heijltjes A, Janssen E, van Gog T. Repeated retrieval practice to foster students' critical thinking skills. Collabra-Psychology. 2021. 10.1525/collabra.28881</bibtext> </blist> <blist> <bibtext> Villalon J, Calvo RA. Concept maps as cognitive visualizations of writing assignments. Journal of Educational Technology &amp; Society. 2011; 14; 3: 16-27</bibtext> </blist> <blist> <bibtext> Wang, P. (1999). An empirical study of knowledge structures of research topics. In L. Woods (Ed.), Proceedings of the 62 annual meeting of the American society for information science and technology ASIS (pp. 557–586), Medford, NJ: Information Today. https://eric.ed.gov/?id=EJ597605</bibtext> </blist> <blist> <bibtext> Wang, P, Bales, S, Rieger, J, &amp; Zhang, Y. (2005). Survey of learners' knowledge structures: rationales, methods and instruments. Proceedings of the 67th annual meeitng of the American society for information science and technology ASIS (pp. 218–228). Information Today. https://doi.org/10.1002/meet.1450410126</bibtext> </blist> <blist> <bibtext> Wei Z, Zhang Y, Clariana RB, Chen X. The effects of reading prompts and of post-reading generative learning tasks on multiple document integration: Evidence from concept network analysis. Educational Technology Research and Development. 2024; 72: 661-685. 10.1007/s11423-023-10326-w</bibtext> </blist> <blist> <bibtext> Wineburg S. Historical problem solving: A study of the cognitive processes used in the evaluation of documentary and pictorial evidence. Journal of Educational Psychology. 1991; 83; 1: 73-87. 10.1037/0022-0663.83.1.73</bibtext> </blist> <blist> <bibtext> Witherby AE, Carpenter SK. The rich-get-richer effect: Prior knowledge predicts new learning of domain-relevant information. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2022; 48; 4: 483. 10.1037/xlm0000996</bibtext> </blist> <blist> <bibtext> Zhang Z, Yuan K-H. Practical statistical power analysis using webpower and R. 2018; ISDSA</bibtext> </blist> </ref> <aug> <p>By Ziqian Wei; Shuting Yin; Roy B. Clariana and Xuqian Chen</p> <p>Reported by Author; Author; Author; Author</p> <p></p> <p>Ziqian Wei Ziqian Wei is a PhD student at the Education University of Hong Kong, Hong Kong SAR, China. He received his master's degree from South China Normal University, China. His research areas include cognitive psychology, psycholinguistics, and learning and memory.</p> <p>Shuting Yin Shuting Yin is a master's candidate in Psychology at the School of Psychology, South China Normal University, China. Her research areas include cognitive psychology, psycholinguistics, and learning and memory.</p> <p>Roy B. Clariana Roy B. Clariana is a professor in the College of Education, Pennsylvania State University, US. His research areas include measures and models of knowledge structure, natural language processing of texts, automated writing evaluation, and writing to learn.</p> <p>Xuqian Chen Xuqian Chen Dr. Phil in Psychology, is an associate professor at the School of Psychology, South China Normal University, China. Her research areas include cognitive psychology, psycholinguistics, and learning and memory.</p> </aug> <nolink nlid="nl1" bibid="bib44" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib67" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib50" firstref="ref4"></nolink> <nolink nlid="nl4" bibid="bib49" firstref="ref5"></nolink> <nolink nlid="nl5" bibid="bib89" firstref="ref6"></nolink> <nolink nlid="nl6" bibid="bib57" firstref="ref7"></nolink> <nolink nlid="nl7" bibid="bib63" firstref="ref10"></nolink> <nolink nlid="nl8" bibid="bib66" firstref="ref11"></nolink> <nolink nlid="nl9" bibid="bib52" firstref="ref12"></nolink> <nolink nlid="nl10" bibid="bib17" firstref="ref13"></nolink> <nolink nlid="nl11" bibid="bib68" firstref="ref14"></nolink> <nolink nlid="nl12" bibid="bib36" firstref="ref15"></nolink> <nolink nlid="nl13" bibid="bib77" firstref="ref18"></nolink> <nolink nlid="nl14" bibid="bib72" firstref="ref21"></nolink> <nolink nlid="nl15" bibid="bib73" firstref="ref22"></nolink> <nolink nlid="nl16" bibid="bib60" firstref="ref23"></nolink> <nolink nlid="nl17" bibid="bib15" firstref="ref24"></nolink> <nolink nlid="nl18" bibid="bib83" firstref="ref25"></nolink> <nolink nlid="nl19" bibid="bib48" firstref="ref27"></nolink> <nolink nlid="nl20" bibid="bib56" firstref="ref28"></nolink> <nolink nlid="nl21" bibid="bib78" firstref="ref32"></nolink> <nolink nlid="nl22" bibid="bib53" firstref="ref34"></nolink> <nolink nlid="nl23" bibid="bib55" firstref="ref35"></nolink> <nolink nlid="nl24" bibid="bib18" firstref="ref37"></nolink> <nolink nlid="nl25" bibid="bib21" firstref="ref40"></nolink> <nolink nlid="nl26" bibid="bib32" firstref="ref41"></nolink> <nolink nlid="nl27" bibid="bib33" firstref="ref44"></nolink> <nolink nlid="nl28" bibid="bib43" firstref="ref45"></nolink> <nolink nlid="nl29" bibid="bib58" firstref="ref46"></nolink> <nolink nlid="nl30" bibid="bib70" firstref="ref47"></nolink> <nolink nlid="nl31" bibid="bib62" firstref="ref49"></nolink> <nolink nlid="nl32" bibid="bib91" firstref="ref50"></nolink> <nolink nlid="nl33" bibid="bib28" firstref="ref51"></nolink> <nolink nlid="nl34" bibid="bib29" firstref="ref55"></nolink> <nolink nlid="nl35" bibid="bib22" firstref="ref56"></nolink> <nolink nlid="nl36" bibid="bib39" firstref="ref57"></nolink> <nolink nlid="nl37" bibid="bib42" firstref="ref58"></nolink> <nolink nlid="nl38" bibid="bib86" firstref="ref59"></nolink> <nolink nlid="nl39" bibid="bib31" firstref="ref61"></nolink> <nolink nlid="nl40" bibid="bib14" firstref="ref62"></nolink> <nolink nlid="nl41" bibid="bib16" firstref="ref63"></nolink> <nolink nlid="nl42" bibid="bib27" firstref="ref64"></nolink> <nolink nlid="nl43" bibid="bib64" firstref="ref65"></nolink> <nolink nlid="nl44" bibid="bib26" firstref="ref66"></nolink> <nolink nlid="nl45" bibid="bib10" firstref="ref68"></nolink> <nolink nlid="nl46" bibid="bib38" firstref="ref69"></nolink> <nolink nlid="nl47" bibid="bib46" firstref="ref70"></nolink> <nolink nlid="nl48" bibid="bib69" firstref="ref71"></nolink> <nolink nlid="nl49" bibid="bib47" firstref="ref74"></nolink> <nolink nlid="nl50" bibid="bib45" firstref="ref75"></nolink> <nolink nlid="nl51" bibid="bib19" firstref="ref76"></nolink> <nolink nlid="nl52" bibid="bib30" firstref="ref81"></nolink> <nolink nlid="nl53" bibid="bib35" firstref="ref82"></nolink> <nolink nlid="nl54" bibid="bib51" firstref="ref83"></nolink> <nolink nlid="nl55" bibid="bib80" firstref="ref84"></nolink> <nolink nlid="nl56" bibid="bib92" firstref="ref85"></nolink> <nolink nlid="nl57" bibid="bib65" firstref="ref89"></nolink> <nolink nlid="nl58" bibid="bib13" firstref="ref94"></nolink> <nolink nlid="nl59" bibid="bib74" firstref="ref98"></nolink> <nolink nlid="nl60" bibid="bib90" firstref="ref99"></nolink> <nolink nlid="nl61" bibid="bib34" firstref="ref103"></nolink> <nolink nlid="nl62" bibid="bib85" firstref="ref104"></nolink> <nolink nlid="nl63" bibid="bib24" firstref="ref105"></nolink> <nolink nlid="nl64" bibid="bib76" firstref="ref106"></nolink> <nolink nlid="nl65" bibid="bib75" firstref="ref107"></nolink> <nolink nlid="nl66" bibid="bib82" firstref="ref108"></nolink> <nolink nlid="nl67" bibid="bib98" firstref="ref112"></nolink> <nolink nlid="nl68" bibid="bib25" firstref="ref116"></nolink> <nolink nlid="nl69" bibid="bib196" firstref="ref122"></nolink> <nolink nlid="nl70" bibid="bib20" firstref="ref142"></nolink> <nolink nlid="nl71" bibid="bib40" firstref="ref145"></nolink> <nolink nlid="nl72" bibid="bib81" firstref="ref146"></nolink> <nolink nlid="nl73" bibid="bib172" firstref="ref156"></nolink> <nolink nlid="nl74" bibid="bib61" firstref="ref171"></nolink> <nolink nlid="nl75" bibid="bib59" firstref="ref172"></nolink> <nolink nlid="nl76" bibid="bib79" firstref="ref173"></nolink> <nolink nlid="nl77" bibid="bib41" firstref="ref175"></nolink> <nolink nlid="nl78" bibid="bib12" firstref="ref177"></nolink> <nolink nlid="nl79" bibid="bib11" firstref="ref178"></nolink> <nolink nlid="nl80" bibid="bib84" firstref="ref179"></nolink> <nolink nlid="nl81" bibid="bib37" firstref="ref181"></nolink> <nolink nlid="nl82" bibid="bib54" firstref="ref182"></nolink> <nolink nlid="nl83" bibid="bib88" firstref="ref185"></nolink> <nolink nlid="nl84" bibid="bib87" firstref="ref187"></nolink> <nolink nlid="nl85" bibid="bib23" firstref="ref190"></nolink> <nolink nlid="nl86" bibid="bib215" firstref="ref195"></nolink> <nolink nlid="nl87" bibid="bib807" firstref="ref196"></nolink> |
|---|---|
| Header | DbId: eric DbLabel: ERIC An: EJ1483812 AccessLevel: 3 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: The Influence of Pre-Reading Purpose and Extra-Textual Networks with Summary Writing on Multiple Document Concept Network Integration: A Replication of Wei et al. (2024) – Name: Language Label: Language Group: Lang Data: English – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Ziqian+Wei%22">Ziqian Wei</searchLink><br /><searchLink fieldCode="AR" term="%22Shuting+Yin%22">Shuting Yin</searchLink><br /><searchLink fieldCode="AR" term="%22Roy+B%2E+Clariana%22">Roy B. Clariana</searchLink> (ORCID <externalLink term="http://orcid.org/0000-0001-9374-0064">0000-0001-9374-0064</externalLink>)<br /><searchLink fieldCode="AR" term="%22Xuqian+Chen%22">Xuqian Chen</searchLink> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="SO" term="%22Educational+Technology+Research+and+Development%22"><i>Educational Technology Research and Development</i></searchLink>. 2025 73(4):2163-2188. – Name: Avail Label: Availability Group: Avail Data: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/ – Name: PeerReviewed Label: Peer Reviewed Group: SrcInfo Data: Y – Name: Pages Label: Page Count Group: Src Data: 26 – Name: DatePubCY Label: Publication Date Group: Date Data: 2025 – Name: SourceSuprt Label: Sponsoring Agency Group: SrcSuprt Data: National Science Foundation (NSF), Directorate for STEM Education (EDU) – Name: NumberContract Label: Contract Number Group: NumCntrct Data: 2215807 – Name: TypeDocument Label: Document Type Group: TypDoc Data: Journal Articles<br />Reports - Research – Name: Subject Label: Descriptors Group: Su Data: <searchLink fieldCode="DE" term="%22Prereading+Experience%22">Prereading Experience</searchLink><br /><searchLink fieldCode="DE" term="%22Expository+Writing%22">Expository Writing</searchLink><br /><searchLink fieldCode="DE" term="%22Descriptive+Writing%22">Descriptive Writing</searchLink><br /><searchLink fieldCode="DE" term="%22Prompting%22">Prompting</searchLink><br /><searchLink fieldCode="DE" term="%22Networks%22">Networks</searchLink><br /><searchLink fieldCode="DE" term="%22Revision+%28Written+Composition%29%22">Revision (Written Composition)</searchLink><br /><searchLink fieldCode="DE" term="%22Feedback+%28Response%29%22">Feedback (Response)</searchLink><br /><searchLink fieldCode="DE" term="%22Reading+Comprehension%22">Reading Comprehension</searchLink><br /><searchLink fieldCode="DE" term="%22Intervention%22">Intervention</searchLink><br /><searchLink fieldCode="DE" term="%22Network+Analysis%22">Network Analysis</searchLink> – Name: DOI Label: DOI Group: ID Data: 10.1007/s11423-025-10508-8 – Name: ISSN Label: ISSN Group: ISSN Data: 1042-1629<br />1556-6501 – Name: Abstract Label: Abstract Group: Ab Data: Theories and practices to enhance multiple document comprehension and integration are crucial in both personal and work contexts, especially with the proliferation of printed and online sources. This experimental investigation replicates and extends (Wei et al., Educational Technology Research and Development 72:661-685, 2024) to examine how multiple documents integration is influenced by reading purpose, summary writing, and extra-textual networks (pre-reading, Study 1, and post-writing, Study 2). In Study 1 (N = 102), participants were randomly assigned to a pre-reading purpose set by a prompt (integrative or detailed) and by a network (an integrative or else an intra-text network) and then read three documents about Alzheimer's disease to complete a writing task with revision (but no feedback). Three days later, they completed a delayed writing task and an inference verification test. In Study 2 (N = 90), the same procedure was used except that the network was used as feedback after writing to support revision. Results from the two studies agree with the previous research that the quantity and structural quality of integration can be improved by external cues and by delayed repeated writing. This research further confirms an innovative approach for evaluating different aspects of knowledge integration and contributes to the literature from the concept network perspective as a measure and an intervention of multiple-text reading. – Name: AbstractInfo Label: Abstractor Group: Ab Data: As Provided – Name: DateEntry Label: Entry Date Group: Date Data: 2025 – Name: AN Label: Accession Number Group: ID Data: EJ1483812 |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1483812 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1007/s11423-025-10508-8 Languages: – Text: English PhysicalDescription: Pagination: PageCount: 26 StartPage: 2163 Subjects: – SubjectFull: Prereading Experience Type: general – SubjectFull: Expository Writing Type: general – SubjectFull: Descriptive Writing Type: general – SubjectFull: Prompting Type: general – SubjectFull: Networks Type: general – SubjectFull: Revision (Written Composition) Type: general – SubjectFull: Feedback (Response) Type: general – SubjectFull: Reading Comprehension Type: general – SubjectFull: Intervention Type: general – SubjectFull: Network Analysis Type: general Titles: – TitleFull: The Influence of Pre-Reading Purpose and Extra-Textual Networks with Summary Writing on Multiple Document Concept Network Integration: A Replication of Wei et al. (2024) Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Ziqian Wei – PersonEntity: Name: NameFull: Shuting Yin – PersonEntity: Name: NameFull: Roy B. Clariana – PersonEntity: Name: NameFull: Xuqian Chen IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 08 Type: published Y: 2025 Identifiers: – Type: issn-print Value: 1042-1629 – Type: issn-electronic Value: 1556-6501 Numbering: – Type: volume Value: 73 – Type: issue Value: 4 Titles: – TitleFull: Educational Technology Research and Development Type: main |
| ResultId | 1 |
