Recent Advances in Research on the Xu-Argument and Future Directions

In 2021, we published our first special issue on the xu-argument in Chinese Journal of Applied Linguistics. Five years have passed since then. Over this period, second language acquisition (SLA) has undergone substantial changes. For example, the emergence of Large Language Models (LLMs) such as ChatGPT and DeepSeek has begun to reshape language teaching and learning. During the same period, the xu-argument has advanced considerably, both theoretically and methodologically. The present special issue brings together an interview and seven empirical studies, which collectively exemplify these developments.

First of all, the theoretical framework of the xu-argument is now more systematic and fully articulated. In the interview conducted by M. Wang, I offer a comprehensive account of the xu-argument, elucidating the learning mechanisms it posits, and its core principles of language teaching and learning. Building on this foundation, I discuss how the xu-argument sheds new light on the long-standing issues in the field of SLA. I take two crucial issues as examples. First, I explain how the xu-argument reconciles the tension between the complexity and predictability of the learning process (Atkinson et al., 2025; Lantolf et al., 2025). Second, I discuss how the xu-argument helps integrate theory into teaching practice (Michel et al., 2025) by xu-based pedagogy. Of particular interest, the recent advances of LLMs create opportunities for the application of the xu-argument even as they introduce challenges to language pedagogy. I discuss the reciprocal relationship between LLMs and the xu-argument in the interview. Two points are highlighted. On the one hand, the xu-argument’s account of language learning aligns in important aspects with LLMs’ operating mechanisms. For this reason, LLM-mediated learning environments provide a productive test bed for further examining the beneficial effects of the xu-based tasks on language learning (Wang, 2025). On the other hand, given xu’s capacity to promote interaction and alignment, the xu-argument may help address the concerns about using LLMs in language learning such as overreliance, reduced agency and weakened problem-solving skills. At the same time, the strong capabilities of LLMs help enhance the efficiency and effectiveness of xu by intensifying interactive alignment. Taken together, I argue that integrating LLM and xu is a promising avenue for future research.

Secondly, inquiry into the mechanism through which xu facilitates learning has continued to deepen. New research methods are adopted to explore the cognitive processes underlying learners’ continuation behavior. Utilizing eye-tracking technology, Gao, Li and Yuan investigated the effects of xu-argument-based continuation writing on learners’ processing of source texts. They compared learners’ behaviors of source-text reading in three conditions: continuation writing, summary writing, or reading-only. Results showed that both production groups exhibited significantly longer first-pass durations than the reading-only group. More importantly, the continuation group spent significantly longer go-past time and total fixation duration than the summary group which outperformed the reading-only group. This indicates that continuation tasks promoted deeper cognitive engagement during reading. Moreover, during writing, the continuation group spent more time rereading the source text with higher fixation counts than the summary group. These findings suggest that continuation writing triggers more intensive reader-text interaction during pre-writing and enhances comprehension-production coupling through sustained attention to input during writing. The eye-tracking data in this study provide direct and compelling evidence for the role of the Xu-argument-based continuation in connecting comprehension and production by mobilizing learners’ internal cognitive capacities.

X. P. Zhang and Chen examined the impact of the xu-argument-based continuation on the development of L2 verb-argument constructions (VACs), which is central to the usage-based approach to L2 learning (Ellis et al., 2014). In this way, this study relates the inquiry into the xu-argument with broader discussions of learning mechanisms in SLA. Specifically, they investigated how the xu-argument-based continuation facilitates Chinese high school students’ English syntactic complexity as captured by their use of VACs over an 8-week period. Two comparable groups of students were involved: one group worked with English input texts (i.e., E-E), while the other group engaged with Chinese input texts containing the same content (i.e., C-E). The results indicated that over time, the E-E group exhibited a greater tendency to use a wider range of VACs, such as caused-motion constructions, attributives, passives, and phrasal verbs. At the same time, they reduced their use of simpler VACs like intransitive-motion and simple transitive constructions, especially when compared to the C-E group. These findings suggest that tasks combining language input with output can significantly enhance learners’ ability to use more sophisticated verb-argument constructions.

Yang, Guo and Yan highlighted comparison as a key mechanism through which continuation promotes learning, reinforcing its pivotal role in xu-competence. This study combines the Xu-argument with research on the Competition Model (MacWhinney, 1997) and article learning. Several types of cue-focused enhancement were developed to investigate (1) how these enhancements shape Chinese-speaking EFL learners’ comparative engagement during Xu-based comparative continuation writing and (2) whether such engagement, in turn, supports the acquisition of English articles. Learners completed comparative continuation writing under one of three cue conditions—paired cues, randomized cues and implicit cues. Learners’ article knowledge was assessed through pretest, posttest and delayed test, allowing the authors to compare gains over time. The results showed that paired cues produced the strongest learning outcomes, yielding more accurate article use in continuation writing and a significant increase in article knowledge. The authors attribute this advantage to an enhanced contrast effect: Paired cues encouraged learners to (a) actively identify similarities and differences within each cue pair, (b) link explanations and examples to article use in the source text, and (c) monitor their own output by comparing it with the input.

X. Y. Zhang examines the facilitating effect of continuation writing from the perspective of corrective feedback timing. This study compares the relative efficacy of the continuation task and the model-as-feedback writing (MAFW) task for EFL writing development. Ninety intermediate-level Chinese EFL learners were randomly assigned to a continuation group, a MAFW group, and a control group. In the treatment session, following common continuation-task procedures, learners in the continuation group first read a model argumentative essay and then wrote an argumentative essay taking the opposite stance to that presented in the model. Following the typical MAFW procedure, learners in the MAFW group first wrote an argumentative essay and then received the model essay as feedback. A pretest and a posttest were administered to gauge L2 writing development. The results showed that the continuation task outperformed the MAFW task not only in improving overall writing quality, but also in enhancing three key components of L2 writing—content, organization, and language use. It is thus concluded that the efficacy of continuation writing derives, at least in part, from learners’ use of the input text as immediate and sustained feedback in the process of continuation writing.

Thirdly, xu-argument research has proactively sought ways to enhance the learning benefits of continuation writing by combining it with well-attested approaches to instructed SLA such as focus on form (Long, 1996). Zhai, Du and Xu explored whether adding explicit input enhancement, a commonly used technique to direct learners’ attention to form, to comparative continuation writing would improve Chinese EFL learners’ discourse competence and writing performance. They examined Chinese middle school students’ comparative continuation writing under three continuation conditions with different degrees/combinations of enhancement techniques. A control condition involving designated-topic writing was also included. Overall, students in all three comparative continuation groups outperformed the designated-topic writing group on measures of discourse competence. At the same time, comparisons among the three enhancement versions showed mixed and somewhat unstable patterns across specific indices, suggesting that not all enhancement configurations yield uniform benefits.

Last but not least, AI tools have been integrated into xu-based learning and assessment, creating a need to investigate their effects. In response, Zhan and Zhou investigated how automated written corrective feedback generated by Grammarly influences EFL college students’ grammar learning strategies, grammar grit, and grammar competence in an iterative continuation task (ICT). Using a mixed-methods sequential exploratory design, 56 students completed ICT either with Grammarly feedback (experimental group) or without Grammarly (control group), with pre-/post-quantitative measures supplemented by qualitative perceptions. Both groups improved in grammar-related outcomes over time, but the patterns differed. The Grammarly group showed significant pre-/post-gains across all measured variables (strategy use, grit, and competence), whereas the control group’s significant gains were limited to the affective strategy dimension, consistency of interest (a grit subcomponent), and grammar competence. Notably, despite Grammarly’s broad benefits across strategy and motivational measures, the control group improved more in grammar competence overall. Qualitative findings revealed mixed learner perceptions of Grammarly. Learners highlighted both affordances and drawbacks, prompting the authors to discuss how Grammarly can be integrated with ICT in ways that support grammar development while managing potential limitations.

In the assessment domain, J. Zhang and Ma evaluated the reliability and validity of AI-generated scores for continuation writing by comparing GPT-4 with eight experienced human raters on 21 student continuations, focusing on scoring consistency, severity/leniency, and fit with human scoring criteria. Analyses showed that GPT-4 demonstrated high self-consistency and could flexibly adjust its scoring behavior when adopting different rating roles (e.g., classroom teacher vs. high-stakes rater), indicating strong procedural reliability and contextual responsiveness. At the same time, GPT-4 tended to assign more lenient scores than human raters and differed in evaluative priorities: the AI foregrounded narrative coherence and emotional depth, whereas teachers placed more weight on linguistic accuracy and richness of detail. The authors argue that these divergences make GPT-4 promising as a supplementary, efficiency-enhancing assessment tool—especially for rapid, holistic feedback—but also underscore the need for calibration to better match educational standards. The study points toward hybrid human–AI scoring models that combine complementary strengths to support fairer and more pedagogically meaningful assessment of continuation writing.

Despite the above-mentioned advances, the xu-argument is still developing. Much remains to be done to integrate it with the new developments in the present-day SLA, particularly as the field is becoming more closely connected with cognitive science and a transdisciplinary framework has begun to take shape (see Douglas Fir Group, 2016; Ortega, 2013). Although the xu-argument foregrounds xu and uses it to account for language acquisition mechanisms, which might strike one as too simplistic in view of the great complexity associated with SLA processes, yet its underlying principles are broadly consistent with mainstream approaches. The key elements that this argument emphasizes—such as interaction, context, and experience—are central to contemporary distributed cognition and various usage-based theories. To strengthen its theoretical grounding and enhance its credibility, further validation from the perspective of the philosophy of language is needed. To advance this line of research and bring Chinese scholarship to a wider audience, it is also necessary for the xu-argument to engage with major ongoing debates in the international SLA forum. Doing so may yield new insights and provide a pathway for Chinese scholars to make original contributions to the field.