Second language (L2) writing is a cognitively demanding process involving formulation, execution, and monitoring, often leading to a heavier cognitive load due to factors like topical knowledge, linguistic features, and code-switching. Simultaneously, L2 writers experience anxiety, which negatively impacts writing performance. While previous research has explored L2 anxiety, the role of cognitive load in L2 writing has received less attention, especially concerning its interaction with anxiety and writing performance. This study addresses this gap by hypothesizing direct and indirect interactions between cognitive load, anxiety, and L2 writing performance within the context of a story continuation writing task (SCWT). The SCWT, a novel reading-writing integrated task, requires learners to extend an incomplete story, potentially eliciting both linguistic alignment and anxiety. This study aims to investigate this interaction and determine how cognitive and affective variables contribute to L2 learners' SCWT performance, offering a new theoretical perspective on cognitive resource distribution in language learning and informing pedagogical strategies.

Cognitive load theory categorizes cognitive load into intrinsic (inherent task complexity) and extraneous (presentation-related) load. In L2 writing, factors like cultural background and register increase intrinsic load. Existing literature suggests interactions between cognitive load and anxiety in learning performance but lacks comprehensive understanding of the relationship between the three factors in L2 writing. L2 writing anxiety (L2WA) negatively impacts writing performance, including syntactic complexity. Previous studies showed positive correlations between cognitive load and learner anxiety, but the relationship between writing content-related cognitive load and different dimensions of L2WA is unclear. The study proposes four hypotheses: 1) Cognitive load directly affects L2WA; 2) Cognitive load directly affects SCWT performance; 3) L2WA directly affects SCWT performance; and 4) Cognitive load indirectly affects SCWT performance through L2WA. The SCWT, despite its potential for reducing cognitive load through language scaffolding, may also induce anxiety due to its integrated nature. Existing research on SCWT focuses primarily on linguistic aspects, with limited empirical examination of cognitive load and affective factors.

This study used convenience sampling to recruit 197 Grade 12 students from a Chinese public high school. After data cleaning, 182 participants (average age 17.21, 6.74 years of English learning) remained. Data were collected using three instruments: 1) An eight-item L2 writing cognitive load scale, adapted from existing scales, measuring overall difficulty, linguistic demand, time pressure, mental effort, frustration, and performance; 2) A nine-item short-version L2 writing anxiety scale from Cheng (2017), assessing cognitive, somatic, and avoidance anxieties; and 3) A SCWT from a mock college entrance exam, scored by experienced teachers based on connection to the source text, content, language, structure, and coherence (0-25 points). Data were analyzed using SPSS 27.0 for preliminary assumption tests (normality, correlation, multicollinearity) and AMOS 28.0 for structural equation modeling (SEM). Exploratory factor analysis (EFA) was performed to validate the questionnaire's factor structure. SEM was used to test the hypotheses, assessing model fit using χ², χ²/df, RMSEA, CFI, TLI, and SRMR. A bootstrapping procedure with 95% confidence intervals was applied to specify direct and indirect effects.

Exploratory factor analysis resulted in a five-item L2 writing cognitive load subscale and a nine-item L2 writing anxiety subscale (KMO = 0.724). Cronbach's α values for cognitive load and L2WA were 0.782 and 0.771, respectively. The mean SCWT score was 15.99 (SD = 2.00). Correlation analysis showed a positive correlation between cognitive load and anxiety (r = 0.216, p < 0.01) and a negative correlation between L2WA and SCWT scores (r = -0.310, p < 0.01). SEM showed a good model fit: χ²/df = 1.954, RMSEA = 0.073, CFI = 0.912, SRMR = 0.076 (TLI = 0.885 slightly below threshold). The model explained 15.6% variance in L2WA and 10.8% variance in SCWT performance. Bootstrapping confirmed a direct effect of cognitive load on L2WA (β = 0.396, p = 0.003) and a direct negative effect of L2WA on SCWT performance (β = -0.289, p = 0.048). No direct effect of cognitive load on SCWT performance was found (β = -0.080, p = 0.491). However, an indirect effect of cognitive load on SCWT performance through L2WA was significant (β = -0.114, p = 0.032). Cognitive load primarily accounted for cognitive anxiety (R² = 86.2%). A follow-up analysis showed that students with low writing anxiety performed significantly better in SCWT than those with high anxiety (t = 3.64, p < 0.01, d = 0.67).

The positive correlation between cognitive load and L2WA aligns with previous research. Cognitive load appears to mainly induce cognitive anxiety, potentially due to the interdependence of cognition and emotion. The lack of a direct effect between cognitive load and SCWT performance might be due to the task modality (writing allows for pausing and revision), the alignment strategy inherent in SCWT, or the low sensitivity of the global writing score. The negative effect of L2WA on SCWT performance is consistent with prior findings, potentially influenced by the competitive Chinese educational culture and the demands of the integrated task on attentional resources. The mediating role of L2WA in the relationship between cognitive load and SCWT performance suggests that reducing cognitive load can indirectly improve performance by lowering anxiety. The weaker predictive power of avoidance behavior compared to previous studies might be due to the higher English proficiency of participants and the high-stakes nature of the task.

This study integrates cognitive load theory into L2 writing research, demonstrating the mediating role of L2 writing anxiety in the relationship between cognitive load and writing performance. The findings highlight the importance of managing cognitive load and anxiety in L2 writing instruction. Future research should investigate these relationships in diverse contexts using objective measures of cognitive load and considering learners' motivational levels.

The study's limitations include the use of a convenience sample of Chinese students, the high-stakes testing environment potentially influencing results, the lack of reported inter-rater reliability for SCWT scoring, the potential low sensitivity of the SCWT and the global writing score to detect cognitive load effects, the cross-sectional design, and the use of subjective self-report measures for cognitive load. Future research should address these limitations by using more diverse samples, various task settings, objective cognitive load measures, and longitudinal designs.