Examining the impact of the number of sources on students' writing performance and self-efficacy beliefs

Source-based writing tasks, which require selecting, organizing, and connecting information from one or more texts, are increasingly used in academic contexts but are cognitively demanding. Prior research has examined task variables such as prompts, topic familiarity, information type, cognitive demands, and genre, but the effect of the number of sources remains underexplored. The present study investigates whether the number of sources (single vs. multiple) affects first-year university students' source-based writing performance and their self-efficacy beliefs. The study focuses on broader writing performance aspects—selecting key ideas, integrating ideas through paraphrasing and summarizing, organizing ideas coherently, and accurate language use—and examines changes in self-efficacy before and after the task. Research questions: (1) What is the impact of writing from multiple sources compared to writing from a single source on students' writing performance? (2) Is writing from multiple sources more effective than writing from a single source in improving students' self-efficacy in writing?

The study is grounded in the knowledge-telling vs. knowledge-transforming model (Scardamalia and Bereiter, 1987) and self-efficacy theory (Bandura, 1977; 1997). The model suggests that single-source writing may elicit knowledge telling (retrieving and presenting information), whereas multiple-source writing may promote knowledge transforming (integrating and synthesizing across texts). Self-efficacy derives from mastery experiences, vicarious experiences, feedback, and physiological states, with mastery experience being most influential; complex tasks successfully completed can enhance self-efficacy. Empirical work on number of sources is limited. Studies comparing single vs. multiple sources (e.g., Wiley & Voss, 1999; Britt & Aglinskas, 2002; Stadtler et al., 2013) found that multiple sources can increase transformation, integration, and handling of conflicting information, though many focused on comprehension rather than comprehensive writing performance. Golparvar and Rashidi (2021) reported increased syntactic complexity, lexical diversity, and cohesion (with decreased accuracy) in multiple-text summaries, but did not rate broader rhetorical features (content, organization, selection). Research on self-efficacy shows robust associations with writing performance (e.g., Pajares, 2003; Multon et al., 1991), and in source-based contexts, components of writing self-efficacy (linguistic, self-regulatory, performance) predict aspects of source-based writing (Golparvar & Khafi, 2021). However, it is unclear whether task characteristics like number of sources influence self-efficacy itself. The present study addresses these gaps by comparing single vs. multiple source tasks on broader writing performance features and self-efficacy changes.

Design: Between-subjects experiment with two conditions: single-source vs. multiple-source source-based writing. Participants: 57 Year 1 college students (General English course), two intact sections randomly assigned to conditions (single-source n=29; multiple-source n=28). Predominantly male (55 males, 2 females); IELTS 5.5–6.0; groups comparable in demographics and proficiency. Proficiency checks: LexTALE showed no significant group differences (t(55)=6.676, p=0.502). Baseline writing ability: pre-writing IELTS-type opinion essay (250 words, 40 min) showed no significant difference (t(55)=1.722, p=0.091). Baseline integrated writing (summary of short passage) also showed no difference (t(55)=1.865, p=0.068). Materials and tasks: Topic: distracted driving in the UAE. Both tasks had equal length (705 words) and similar readability (Flesch-Kincaid 9–11). Multiple-source task included four sources (newspaper article, FAQ, graph, TED Talk comments); single-source task provided a consolidated article containing overlapping information. Experts rated difficulty to ensure comparability; the sole manipulated variable was the number of sources. Writing prompt for both groups: Write an opinion essay on whether the UAE distracted driving law should be changed, using ideas from the text(s). Time: 45 minutes total to read, plan, and type. Measures: Writing performance scored with a rubric on four criteria—content, source use, organization, language use—each rated 1 (Poor) to 4 (Very Good). Two trained PhD student raters independently scored essays and reached consensus; inter-rater reliability (Cronbach’s alpha) was 0.81 (single-source) and 0.87 (multiple-source). Source integration attempts were coded as exact copies, borrowed words/strings, or paraphrases (Keck, 2006). Self-efficacy survey: Initially 22 items adapted from validated instruments (Bruning et al., 2013; Teng et al., 2018; Zhou et al., 2022), 5-point Likert scale. After piloting and PCA (KMO=0.67; Bartlett p<0.05), three subscales (18 items) were used: writing-related abilities (e.g., organization, vocabulary, grammar, linking), reading-related abilities (select relevant ideas, justify opinions with texts, summarize, think about relationships), and source integration (paraphrase, quote, APA referencing, avoid plagiarism, integrate content). Reliability: overall alpha 0.885 (pre) and 0.994 (post); subscale reliabilities acceptable. Procedure: Ethics approval obtained. Participants provided consent, completed pre-writing test, then pre-task self-efficacy survey. Students received source materials and identical instructions and wrote the essay within 45 minutes. Post-task self-efficacy survey administered immediately after. Data analysis: Descriptive statistics, independent/paired t-tests to compare performance across groups and pre-post self-efficacy changes within groups; effect sizes reported where applicable; analysis of source integration frequencies by type.

- Overall writing performance: Multiple-source group mean=11.43 (SD=2.03) vs. single-source mean=10.69 (SD=1.54); difference not significant (t=-1.554, p=0.126). - Writing rubric subscales (means; t-tests multiple vs. single): • Content: 3.68 vs. 3.10; significant advantage for multiple sources (t=2.52, p=0.02, d=0.67). • Source use: 2.54 vs. 2.38; not significant (t=0.62, p=0.54). • Organization: 3.07 vs. 3.03; not significant (t=-0.22, p=0.83). • Language use: 2.14 vs. 2.17; not significant (t=0.12, p=0.91). - Source integration attempts: Single-source group totaled 19 attempts; multiple-source group totaled 87 attempts; all students in the multiple-source group made at least one attempt, whereas 16/29 in the single-source group made none. • By type (single vs. multiple): exact copying 9 vs. 24; borrowing words/strings 2 vs. 46; paraphrasing 8 vs. 17. - Self-efficacy (pre to post within groups, three subscales): • Reading-related abilities: Multiple-source increased by +0.51 and was significant (t(27)≈-3.563, p<0.05, d≈-0.94 sign reflects direction); Single-source decreased by −0.25, significant (t(28)=2.342, p=0.027, d=0.57). • Source use skills: Multiple-source increased by +0.41; Single-source increased by +0.08; changes not statistically significant for either group (p>0.05). • Writing-related abilities: Multiple-source +0.04; Single-source −0.10; changes not statistically significant (p>0.05).

Despite manipulating the number of sources, overall writing quality did not significantly differ across groups, possibly due to the brief time window (45 minutes) and the gradual nature of writing skill development. However, the multiple-source condition led to significantly higher content scores and more text-based, idea-rich essays, suggesting that exposure to multiple sources directs attention toward selecting and incorporating relevant ideas. Both groups struggled with source integration, often relying on copying or borrowing strings with limited paraphrasing, indicating the difficulty of discourse synthesis even when instruction had been provided. Organization and language use did not differ between groups; students generally adhered to a standard essay structure learned previously, and their language control was limited across conditions. Self-efficacy results indicate that multiple-source tasks can bolster reading-related self-efficacy, likely through mastery experiences and heightened engagement with source materials, whereas single-source tasks may yield less interaction and even reduce reading-related confidence. However, short-term exposure did not translate into broad self-efficacy gains in writing-related or source integration domains, underscoring the need for extended instruction and practice.

The number of sources alone did not improve overall source-based writing performance in a short, single-session task. Nonetheless, multiple sources promoted stronger content development and more frequent use of source material, and were associated with significant gains in reading-related self-efficacy. Educators designing source-based writing tasks may sequence activities to foster close reading, note-taking, organizing evidence, and explicit modeling of synthesis and citation practices to enhance both performance and confidence. Future research should examine longer-term interventions, include more diverse and balanced samples, and investigate writing processes (e.g., via think-alouds) to understand how learners engage with and synthesize multiple documents.

Generalizability is constrained by the small, predominantly male, single-institution sample; only one task per condition was administered. The study analyzed written products without process data; think-alouds or process tracing were not employed to avoid interfering with the main task. The short time frame (45 minutes) may have limited observable differences in performance and changes in self-efficacy.