Effects of input enhancement and genre on L2 learners' performance in the continuation writing task

Complexity, accuracy, and fluency (CAF) are established indices for evaluating L2 writing performance, yet prior research has emphasized output while underexamining input. The xu-argument based continuation writing task links input to output by asking learners to continue an incomplete text, potentially leveraging alignment to promote learning. Prior studies have examined single factors such as genre, task complexity, and input enhancement, but little is known about the combined effects and interactions among factors. Input enhancement visually highlights target forms to draw attention, and genre shapes discourse structures and language features. This study uses a two-factor mixed experimental design to examine how input enhancement (between-subjects: enhanced vs. non-enhanced) and genre (within-subjects: story vs. argumentation) affect Chinese EFL learners’ CAF in continuation writing, addressing: (1) the effects of input enhancement and genre on continuations; and (2) whether their interaction affects performance.

Studies using CAF show it is a robust framework to assess L2 writing proficiency and development, but most work focuses on production without integrating input. The continuation writing task (Wang, 2012) aligns input and output and has been used to study input complexity, topic familiarity, iterative practice, input enhancement, and pre-task planning, though most studies isolate single factors. Input enhancement aims to increase attention to linguistic forms via visual cues and has been shown in continuation tasks to improve attention and alignment. Genre critically influences writing due to differing communicative purposes and linguistic features, and limited research has explored genre effects in continuation tasks; none has examined genre effects on CAF within continuation tasks or the interaction between genre and input enhancement. This study fills these gaps by treating CAF as dependent variables, with input enhancement (between-subjects) and genre (within-subjects) as independent variables.

Design: Two-factor mixed experimental design. Between-subjects factor: input enhancement (enhanced vs. non-enhanced). Within-subjects factor: genre (story vs. argumentation). Dependent variables: CAF indices. Participants: 80 Chinese senior high school students (ages 16–18) from two parallel classes taught by the same instructor. Randomly assigned to enhanced group (n=40; 30 males, 10 females) and non-enhanced group (n=40; 25 males, 15 females). Approx. 8 years of English study; no immersion experience. A proficiency test one week prior showed no significant difference between groups (t = -0.058, df = 63.18, p = 0.954). Materials: Two incomplete texts (one story, one argumentation), 350–400 words each (story 387; argumentation 367). Topics were familiar/relevant (family life for story; social media for argumentation). Story: a mother pawns a ring to buy pants for a child’s graduation, with opening sentences guiding continuation. Argumentation: discusses pros/cons of social media, with given first sentences guiding analysis of another perspective and conclusion. Enhanced versions visually marked 24 target items per text (vocabulary, lexical chunks, syntactic structures) with underlines; non-enhanced versions had no markings. Procedure: Both groups completed the story continuation first, then argumentation one week later. Each task: 50 minutes, target ~150 words. Participants had to read and understand the input and maintain coherence with it. No dictionaries/computers allowed. Data collection: Compositions compiled; participants who completed only once were excluded. Final dataset: 40 compositions per group (after balancing non-enhanced group by removing seven at random), yielding four subcorpora: enhanced story, enhanced argumentation, non-enhanced story, non-enhanced argumentation (each n=40). Measures and tools: Lexical diversity measured by Uber index (U = (logTokens)^2 / (logTokens − logTypes)) using Vocab Profile. Lexical sophistication measured by Lambda (0–4.5) via P_Lex. Syntactic complexity: C/S (clauses per sentence, general complexity) and CN/T (complex nominals per T-unit, phrasal complexity) via L2SCA (Lu, 2010) and TAASSC (Kyle, 2016). Lexical accuracy: 1 − LE/CW (lexical errors per content word; excluding names/numbers), with errors annotated in Word 2016, double-checked via pigai.org, content words counted using AntConc 3.5.0 with TreeTagger POS tags. Syntactic accuracy: EFT/T (error-free T-units per T-unit), with syntactic errors (word order, omission/redundancy of constituents, verb morphology, clause-combining) annotated in Word 2016 and double-checked via pigai.org; T-units labeled by researchers. Fluency: W/T (words per T-unit) from Vocab Profile and manual T-unit counts. Analysis: Repeated measures general linear model (SPSS 24.0) tested within-subjects (genre) and between-subjects (input enhancement) effects and their interaction on each CAF index. Simple effects tests were conducted where interactions were significant.

Main effects and interactions (from repeated measures GLM): Lexical diversity (Uber index): No main effect of input enhancement (F=0.407, p=0.525, η²p=0.005). Significant main effect of genre (F=9.664, p=0.003, η²p=0.110), with story > argumentation. No interaction (F=0.194, p=0.660). Lexical sophistication (Lambda): No main effect of enhancement (F=1.738, p=0.191, η²p=0.022). Significant main effect of genre (F=92.268, p<0.001, η²p=0.542), with argumentation > story. No interaction (F=0.386, p=0.536). General syntactic complexity (C/S): No main effect of enhancement (F=3.254, p=0.075, η²p=0.040). Significant main effect of genre (F=24.013, p<0.001, η²p=0.235), with story > argumentation. No interaction (F=1.308, p=0.256). Phrasal complexity (CN/T): No main effect of enhancement (F=0.655, p=0.421, η²p=0.008). Significant main effect of genre (F=406.788, p<0.001, η²p=0.839), with argumentation > story. No interaction (F=0.483, p=0.489). Lexical accuracy (1−LE/CW): Significant main effect of enhancement (F=13.903, p<0.001, η²p=0.151), with non-enhanced > enhanced. Significant main effect of genre (F=42.392, p<0.001, η²p=0.352), with story > argumentation. Significant interaction (F=6.275, p=0.014, η²p=0.074). Simple effects: within enhancement, story > argumentation (p<0.001) and within non-enhancement, story > argumentation (p=0.006). Within story, no difference by enhancement (p=0.170). Within argumentation, non-enhanced > enhanced (MD=0.030, p<0.001). Syntactic accuracy (EFT/T): Significant main effect of enhancement (F=6.068, p=0.016, η²p=0.072) and genre (F=4.697, p=0.033, η²p=0.057). Significant interaction (F=10.804, p=0.002, η²p=0.122). Simple effects: in enhanced condition, argumentation > story (MD=−0.102, p<0.001); in non-enhanced, no story–argumentation difference (p=0.431). Within argumentation, enhanced > non-enhanced (MD=0.112, p<0.001); within story, no enhancement effect (p=0.734). Fluency (W/T): No main effect of enhancement (F=1.227, p=0.271, η²p=0.015). Significant main effect of genre (F=23.534, p<0.001, η²p=0.232), with argumentation > story. No interaction (F=0.025, p=0.874). Descriptive means (W/T): enhanced story 11.74; enhanced argumentation 12.99; non-enhanced story 11.38; non-enhanced argumentation 12.55. Overall: Input enhancement significantly affected accuracy (lexical and syntactic) but not complexity or fluency; genre significantly affected all CAF dimensions; interactions occurred for accuracy but not for complexity or fluency.

The lack of enhancement effects on complexity and fluency may relate to restricted lexical/syntactic scope in materials (24 enhanced items per text), topic-constrained language use, and learners’ avoidance of less-mastered structures. Genre robustly shaped CAF: story continuations yielded higher lexical diversity and general syntactic complexity (C/S), while argumentation produced higher lexical sophistication and phrasal complexity (CN/T), consistent with genre-specific discourse demands and alignment with input. Accuracy showed genre-dependent enhancement effects: in story, enhancement did not change lexical or syntactic accuracy; in argumentation, enhancement reduced lexical accuracy (likely due to misuse of highlighted items without full mastery) but increased syntactic accuracy (enhanced attention to form). Fluency did not improve with enhancement, aligning with resource competition among CAF and the need for long-term practice to affect automatization; argumentation nevertheless showed higher W/T, reflecting longer, more complex sentences typical of argumentative discourse. Pedagogically, enhancement can raise attention to forms, but target structures should match learners’ proximal development, and genres can be leveraged differentially: stories to stimulate ideas and clause-level complexity; argumentation to foster sophisticated lexis and phrasal structures.

The study shows that genre significantly influences all CAF dimensions, while input enhancement significantly affects accuracy but not complexity or fluency. Interactions between enhancement and genre occur for accuracy only. Recommendations include: selecting engaging, level-appropriate continuation materials and enhancing target forms within students’ zones of proximal development; incorporating multiple genres (not just stories) to exploit their complementary strengths; and encouraging sustained accumulation of vocabulary, phrases, and collocations, as enhancement alone may not immediately improve complexity or fluency.

The study examined only two factors (input enhancement and genre) and only two genre levels (story and argumentation); future work could include additional factors and genres (e.g., practical writing, exposition). The design assessed short-term effects; longitudinal studies are needed to track developmental trajectories and dynamic CAF interactions.