A Diachronic Investigation of Individual Differences and Writing Quality in Data-Driven Learning

2.2.1. Foreign Language Enjoyment and Anxiety

Foreign language enjoyment and anxiety have received growing attention in SLA research, with a plethora of empirical studies showing their strong predictability for language learning outcomes (e.g., N. Chen et al., 2025; J. M. Dewaele et al., 2025; Shao et al., 2020). Drawing upon the control-value theory (CVT; Pekrun, 2006), foreign language enjoyment and anxiety are directly related to language learning, which arise from learners’ perceptions of control and value on learning tasks. These foreign language affective attitudes can significantly influence learners’ cognitive, motivational, and behavioral processes, further affecting their engagement and learning outcomes (J. M. Dewaele et al., 2025; Trevors, 2021; Zare et al., 2022). Therefore, examining foreign language enjoyment and anxiety within pedagogical interventions is essential for understanding how instruction shapes learners’ affective and learning processes.

Foreign language enjoyment and anxiety are multidimensional, involving the object focus (i.e., an activity or its outcome), valence (i.e., positive or negative quality of performance), and activation (i.e., level of arousal). According to these dimensions, foreign language enjoyment is an activity-related positive affective attitude; that is, foreign language enjoyment emerges when learners are involved in specific learning tasks (Li et al., 2025). Foreign language anxiety, on the other hand, is an outcome-related negative affective attitude that responds to future results such as fear of failure in future learning outcomes. Foreign language enjoyment and anxiety, as the focuses of this study, can be recurrently activated through repeated learning experiences and may evolve over extended periods of time (J.-M. Dewaele & Li, 2020). Thus, the repeated exposure to a particular pattern of emotional experience becomes a trait-like affective attitude, which is “long-lasting and often lingered” (X. Wang et al., 2024, p. 8). Adopting this perspective, recent research has shown that foreign language enjoyment and anxiety may build progressively through ongoing classroom experiences and fluctuate over time in response to meaningful engagement and instructional environments (Horwitz, 2017; Saito et al., 2018).

Foreign language enjoyment refers to a positive emotional reaction resulting from students’ experiences in foreign language learning environments (Elahi Shirvan et al., 2024). Foreign language enjoyment enhances learners’ noticing, raises their awareness of linguistic features, and helps them consciously attend to, process, and learn language more effectively (J. M. Dewaele & Alfawzan, 2018; J. M. Dewaele & MacIntyre, 2016). As CVT posits, learners who experience foreign language enjoyment often have a higher sense of control over their learning process, which in turn contributes to greater persistence, focus, and effort, all of which have the potential to enhance writing quality (Fredrickson, 2001). Previous studies have provided evidence for the positive relationships between foreign language enjoyment and language achievement. For instance, Jin and Zhang (2018) explored this relationship in high school students’ language achievement via a data triangulation of surveys, classroom observations, and test scores, and found a positive relationship between foreign language enjoyment and their test scores. Similarly, Y. Guo (2021) conducted a longitudinal study to explore how foreign language enjoyment impacted language achievement and reported that students with greater foreign language enjoyment had higher performance. Similarly, Mierzwa (2018) found a positive relationship between university students’ foreign language enjoyment and their language achievement; that is, learners with higher levels of foreign language enjoyment produced higher-quality essays and demonstrated greater fluency in oral presentations.

Foreign language anxiety is an emotional response characterized by feelings of tension, apprehension, and worry experienced in the context of foreign language learning (Horwitz et al., 1986). From the perspective of CVT, learners experiencing high foreign language anxiety may feel a reduced sense of control over the learning process, especially when faced with tasks they perceive as overwhelming or beyond their capabilities. This perception of lack of control leads to a fear of failure, which can impair cognitive resources by consuming working memory and attention, leaving fewer cognitive resources available for task processing. This may further lead to avoidance behavior, where learners disengage from tasks due to anticipated failure, further impacting their writing performance (Heckel et al., 2021).

A number of studies have shown that language achievement is negatively affected by foreign language anxiety (e.g., Teimouri et al., 2019; X. Zhang, 2019). A meta-analysis by Teimouri et al. (2019) found a moderate negative correlation between foreign language anxiety and language achievement. This result was supported by another meta-analysis by X. Zhang (2019), which found that foreign language anxiety limited learners’ active learning and learning outcomes. Recent chronological studies in language learning also showed that the relationship between foreign language anxiety and achievement was curvilinear (Horwitz, 2017). The findings of these studies showed that foreign language anxiety could have dynamic effects on learners’ language acquisition, depending on its intensity and context (Bosmans & Hurd, 2016). Despite this, several studies have reported positive findings that foreign language anxiety may also facilitate language achievement to some extent (e.g., Sajedi, 2016; Salmani Nodoushan, 2015).

2.2.2. Behavioral, Cognitive, and Emotional Engagement

Engagement is defined as “the extent to which students contribute to activities” (Huang et al., 2022). It is widely recognized as a critical component of the learning process, encompassing active participation in educational activities, collaboration with peers, seeking guidance from instructors, and engaging in self-directed learning (Alsowat, 2016; L. Ding et al., 2017). Engagement is closely associated with self-determination theory (SDT), which posits that only when learners’ basic psychological needs are satisfied can they pay attention to and devote meaningful effort to learning tasks, thus contributing to greater performance (Ryan & Deci, 2017). Previous research has shown that learner engagement is a multidimensional construct, involving behavioral, cognitive, and emotional engagement (Ellis, 2010). Within the EFL classroom context, the three dimensions of engagement function as interrelated components that facilitate language development and achievement. Specifically, behavioral engagement refers to students’ observable behaviors in academic, social, and extracurricular activities, including attendance, effort, and persistence (Jelas et al., 2016; Kim & Kim, 2020). In the EFL context, students with high behavioral engagement are more likely to actively participate in classroom tasks such as peer feedback tasks and group discussions, thereby increasing their opportunities for language practice and feedback (Y. Ding & Zhu, 2025). Cognitive engagement refers to the level of investment in learning, including willingness and strategies for metacognitive and self-regulated learning, which improves learners’ understanding and promotes the internalization of language skills (Buil et al., 2020; Christenson et al., 2012). Emotional engagement involves the degree to which students react emotionally to the learning process, including interest, boredom, happiness, anxiety, and other feelings about their learning experience, and positive emotional engagement could promote sustained motivation and persistence (Jelas et al., 2016). Thus, behavioral, cognitive, and emotional engagement complement each other to increase active involvement, motivational effort, and the use of effective learning strategies, all of which facilitate learners’ language achievement (Fredricks et al., 2004; Y. Liu et al., 2024).

Previous findings suggest that engagement is positively associated with achievement. For instance, Bråten et al. (2022) examined how behavioral engagement enhanced comprehension performance. Using a sample of 116 Norwegian undergraduates, this study found that behavioral engagement significantly predicted comprehension performance. Liu et al. (2024) explored the relationship between engagement and achievement among Chinese high school EFL learners and university students and found a significant positive relationship between engagement and language achievement. This indicates that students who are behaviorally, emotionally, and cognitively engaged in language learning tasks are more likely to achieve greater learning outcomes.

2.2.3. Autonomy

Autonomy is defined as “the ability to take charge of one’s own learning” (Little, 2007, p. 15), which entails making informed choices and effectively managing resources to support learning. Based on SDT theory, autonomy influences L2 performance primarily by fostering learners’ autonomous learning behaviors (Hu & Zhang, 2017; Ryan & Deci, 2017). Specifically, when learners perceive a sense of psychological ownership over their learning, they are more likely to make independent choices about their learning activities, take initiative in seeking out language practice opportunities, and make decisions that align with their personal interests and values. These autonomous learning behaviors foster a greater willingness to participate, more active involvement in language learning tasks, and sustained efforts to learn, which ultimately support improved language performance. L. Chen (2023) explored the relationship between autonomy and English learning achievement among 291 Chinese undergraduates, and found that autonomy directly improved achievements and reduced foreign language anxiety. Hu and Zhang (2017) implemented a one-year action program in a Chinese university to examine how autonomy-driven teaching impacted English proficiency, and found positive effects of enhancing learners’ autonomy on their language skills.

Considering the crucial role of individual differences in foreign and L2 learning and teaching, it is important to investigate how different instructional contexts impact these factors. Currently, research on learners’ individual differences in DDL contexts is still limited, and few studies have simultaneously examined how these individual differences evolve within instructional contexts (as dependent variables) and influence learning outcomes (as independent variables). Examining individual differences from both perspectives is valuable, as individual differences are not only dynamically shaped through the instructional process but also can be stable predictors of learning outcomes (Parrott & Hertel, 1999). Therefore, this study aims to examine the dynamic trajectories of learners’ individual differences and writing quality under DDL instruction, as well as the relationship between individual differences and writing quality. Specifically, it investigates: (a) how learners’ individual differences evolve over time in a DDL context; (b) how learners’ writing quality develops over the same period; and (c) how individual differences are associated with variation in writing quality. This study addressed the following research questions:

3.4.1. Quantitative Data Collection

Data were collected in three waves (time points), that is, before (Week 1), mid (Week 8), and after (Week 17) instruction. Students’ writing samples were collected at three waves to assess their writing quality. Topics for the essays were “How to ride a motorcycle,” “The influence of childhood experiences on adult behaviors,” and “The application of technology in education.” All three writing prompts were adapted from the course textbook Writing Critically: Expository Writing 2, which aims to develop learners’ expository writing skills through different rhetorical modes. The first task aligned with process analysis, the second reflected cause and effect, and the third focused on classification and exemplification, all of which were development patterns for writing expository essays. A detailed description of the writing prompts is presented in Appendix C .

In order to capture learners’ individual differences (i.e., foreign language enjoyment and anxiety, engagement, and autonomy), three questionnaires were administered at each wave. Questionnaires were distributed via social media, allowing students to access them conveniently by scanning a QR code shared in a group chat. All questionnaires were completed in class. The first survey also included a demographic questionnaire.

Specifically, the first author translated the questionnaires into Chinese, and the accuracy was checked against the original materials by the second author. Several items were rephrased to ensure cultural relevance and clarity for Chinese EFL learners, and checked by an expert in translation. For example, we rephrased the item “I use online English resources such as YouTube or English forums” to “I use online English resources such as English forums or English learning platforms,” as YouTube is not generally accessible to students in the Chinese EFL context.

3.4.2. Qualitative Data Collection

One-on-one semi-structured interviews were conducted in Mandarin with three students. Participants were randomly selected to reduce potential selection bias, following previous L2 studies (e.g., De Saint Léger & Storch, 2009; Geluso & Yamaguchi, 2014). Each interview lasted approximately 30 minutes and was audio-recorded with the participant’s consent. The purpose of these interviews was to explore the reasons behind the observed changes in learners’ foreign language enjoyment and anxiety, engagement, and autonomy, and to understand their perceptions of how these changes impacted their writing quality during the semester. Each interview began with a brief introduction explaining the aims of the session, followed by a series of open-ended questions such as:

1. What do you think contributed to the changes in your foreign language enjoyment and anxiety, engagement, or autonomy throughout the course?

2. How did these changes influence your ability to write expository essays?

3. Were there specific features of the corpus-based activities that you felt supported or hindered your learning?

In addition, participants were encouraged to elaborate on their experiences, share challenges, and give suggestions for the instructional activities.

Classroom observations were conducted by the authors during three sessions (Weeks 1, 8, and 17) to explore students’ interactions with the SKELL while completing collaborative vocabulary-focused tasks. To reduce researcher bias, the following steps were adopted (Bell et al., 2019). First, prior to the observations, a structured observation protocol was developed by the two observers based on the research question, including: (a) search behaviors; (b) extraction of information from concordance lines; (c) difficulties encountered and coping strategies; (d) collaboration within groups; and (e) classroom reporting. Second, during each observation lesson, the observers took field notes independently, recording the salient behaviors or interactions listed in the protocol without influencing classroom proceedings. Third, following each observation session, the two observers met to review and compare their field notes in detail, in which they discussed the discrepancies in interpretation and jointly clarified ambiguous observations. Finally, both observers wrote observation logs, integrating their notes and reflections to capture both the observed events and their analytical insights.

3.5.1. Quantitative Data Analysis

The essays were evaluated using the analytical scoring rubric of expository writing developed by Kang (2020, 2022), which contains four categories: content, vocabulary, grammar, and organization (see Appendix D ). The justifications for this scoring method are as follows. First, this approach could capture students’ overall writing performance, and is therefore commonly used in educational research (Jonsson & Svingby, 2007). Second, DDL instruction may improve learners’ overall writing performance; thus, analytical scoring was used to assess learners’ progress across different dimensions of writing, following previous practice in DDL (e.g., Dong & Wang, 2025; Muftah, 2023). Third, to meet the class requirement of enhancing students’ overall writing performance, we used analytical scoring in the assessment. Two experienced EFL writing teachers (one rater was the course instructor, and the other was one of the authors) independently evaluated the essays based on this rubric. Prior to grading, a training session was conducted during which the raters thoroughly reviewed the rubric and discussed the rating procedures. Following this, all essays were anonymized and presented in randomized order, with the raters independently assigning scores to each essay. Essays were scored on a 100-point scale. Discrepancies in scores were resolved through a series of follow-up meetings. If the scores for any category differed by more than 2 points, the raters revisited their scores until full agreement was achieved or the difference was reduced to within 1 point. Inter-rater reliability was calculated using Pearson’s correlation coefficient, showing a high level of agreement (Pearson’s r: total score = .97; content = .94; vocabulary = .91; grammar = .90; organization = .92).

Statistical analyses were conducted in sequential steps. First, basic descriptive statistics for all constructs were examined across three waves. For the descriptive interpretation of the absolute levels of questionnaire scores, mean values were interpreted relative to the theoretical midpoint of each summed scale, following previous research (e.g., Kruk, 2018). Then, for determining the changes, a series of linear mixed-effects (LME) models were conducted in R (R Core Team, 2021) using the lmerTest package (Kuznetsova et al., 2017). LME has been attested to be “flexible and suitable for SND [small-N designs] and can provide standardized effect sizes and measures of statistical significance” (Wiley & Rapp, 2019, p. 1), and has been widely used in L2 research with small sample sizes (e.g., Brysbaert, 2025; Duan & Shi, 2024; Kyle et al., 2021). The summed score of all items for each scale was calculated to represent each individual difference factor. In the models addressing research question 1, each individual difference variable (e.g., foreign language enjoyment and anxiety) was entered as the dependent variable, time was entered as a fixed effect, and participants were included as random effects with random intercepts. The analyses provided estimates of variance explained by fixed and random effects, along with p-values to report the significance of the results. For the LME models, model-level effect sizes were reported by calculating R²c and R²m values using the MuMIn package, following Nakagawa and Schielzeth (2013). This approach was adopted in line with recent studies in language research (e.g., Kim & Lu, 2024; Monteiro & Kim, 2020). R²m reflects the variance explained by the fixed effects only, whereas R²c reflects the variance explained by both fixed and random effects combined.

In addressing research question 2, the same model equation was applied, while the dependent variable was set to learners’ scores and the independent variable was time. Additional post-hoc pairwise comparisons among scores were conducted using the emmeans package (Lenth et al., 2018), and the effect sizes were computed using Cohen’s d (Cohen, 1988), following Diez-Ortega and Kyle (2024). The thresholds follow Plonsky and Oswald (2014), which classify values of d = .6 as small, d = 1.0 as medium, and d = 1.4 as large.

For research question 3 to examine the impact of individual differences on writing quality, a series of LME models were constructed. Writing scores were entered as the dependent variable, individual difference factors were entered as fixed effects, and participants were included as random effects with random intercepts. Effect sizes were reported by calculating R²c and R²m values.

3.5.2. Qualitative Data Analysis

Thematic analysis was adopted to analyze students’ interviews and classroom observation data in sequential steps to reduce researcher bias (Braun & Clarke, 2006). The interview data and classroom observation notes were first transcribed and translated into English. Informed by the research questions, the first and second authors first familiarized themselves with the data by independently reading through each transcript and classroom observation notes. During initial coding, these excerpts were open-coded, capturing changes in learners’ foreign language enjoyment and anxiety, engagement and autonomy, the underlying reasons, and the impact of these factors on writing improvement, as well as learners’ behaviors in classes. This was followed by the two authors’ collaborative iterative phases of reading and re-reading the transcripts for categorization, axial coding, detecting recurrent keywords, and identifying and unifying themes relating to the phenomenon under study. The inter-coder reliability was high (Pearson’s r = .90), suggesting a high level of consistency. The coding discrepancies were resolved through discussions. In order to represent each theme, we reported its frequency of supporting quotes, following previous research (e.g., Lawless & Chen, 2019; Pizarro Milian, 2017).

4.1.1. Foreign Language Enjoyment and Anxiety

Table 2 lists the results of the LME models. No significant change over time was observed for foreign language enjoyment (estimate = −0.20, p = .39). In this model, time only explained a minor proportion of variance in enjoyment (R²m = .001), whereas the combination of fixed and random effects explained 89.00% of the variance (R²c = .890). Likewise, anxiety did not show significant changes over time (estimate = 0.07, p = .77). This model reported a R²m of .001 in anxiety, while the full model explained 71.2% of the variance (R²c = .712). The flat trajectory of the bold blue lines (Figures 1 and 2) indicated that learners who experienced DDL instruction maintained stable foreign language enjoyment and anxiety.

OPEN IN VIEWER

Table 2.

Results for Linear Mixed-Effects Models with Time as Fixed Factor.

Intercept					Time
Variable	Estimate	SE	t	p	Estimate	SE	t	p	R²m	R²c
Foreign language enjoyment and anxiety
Enjoyment	14.36	0.62	23.06	<.001***	−0.20	0.23	−0.89	.39	.001	.890
Anxiety	10.71	0.57	18.71	<.001***	0.07	0.22	0.30	.77	.001	.712
Engagement
Behavioral	12.28	0.45	27.25	<.001***	−0.17	0.19	−0.88	.39	.002	.811
Cognitive	11.86	0.26	45.26	<.001***	−0.17	0.17	−1.00	.33	.001	.311
Emotional	10.73	0.37	29.1	<.001***	−0.13	0.14	−0.95	.35	.006	.706
Autonomy	37.16	1.14	32.71	<.001***	0.80	0.32	2.51	.02*	.026	.839
Score	59.75	1.26	47.38	<.001***	8.15	0.86	9.44	<.001***	.491	.892

Note. *p < .05, ***p < .001.

The thematic analysis identified three themes influencing students’ emotional experiences: sense of control over the learning process (one quote), perceived value of acquired knowledge (one quote), and collaborative learning (one quote). First, participants’ stable level of foreign language anxiety could be attributed to their sense of control over the learning process. One student explained that she used to feel much more anxious when the teacher asked her questions in the classroom, because she was not sure if her answers were correct. After using the corpus, she felt more prepared and less anxious in answering teachers’ questions. She attributed this to the benefits of corpus use, particularly that the corpus could provide authentic language examples quickly. This suggested that using a corpus could reduce learners’ sense of uncertainty in classroom interactions, which allowed anxiety to remain at a manageable level:

Excerpt 1. Answering questions in class used to make me more anxious because I was never sure if what I said was correct. After learning how to use the corpus, I started checking how native speakers actually use certain words. I felt more confident, and my anxiety was relieved. (ID: 6, Linda)

Regarding foreign language enjoyment, a theme that explained learners’ fluctuations was the perceived value of the knowledge gained during the corpus activity. Although the overall level of foreign language enjoyment did not show a significant change, a few individual learners showed increased enjoyment over time in their descriptive scores. Interview data further suggested that this increase was associated with corpus use, particularly because it exposed learners to varied usage patterns. As shown in Excerpt 2, when asked to explain the reasons for this increase, the learner noted that corpus use enabled her to move beyond the limited, fixed expressions typically encountered in class and to identify new and varied usage patterns. She reported that this discovery process enhanced her interest in learning. This suggested that recognizing broader lexical patterning through corpus searches enhanced the perceived value of learning, thereby contributing to a sense of progress and fostering greater enjoyment:

Excerpt 2. Before the course, I only knew one or two common ways to use a word, like the fixed phrases we learned in class. But with the corpus, I started noticing new and different patterns, which made me more interested in learning. (ID: 7, Sara)

In addition, peer support in collaborative learning emerged as another main source of foreign language enjoyment. As illustrated in Excerpt 3, Sara emphasized how group discussions during the corpus searching activities helped her feel relaxed. This might be attributed to the sense of community and peer support that she received in collaborative learning, as working in groups allowed her to share ideas, clarify doubts, and learn from others: “In the corpus activity, discussing with my groupmates made me feel really happy” (ID: 7, Sara).

4.1.2. Engagement

Figures 3-5 illustrate the trajectories for behavioral, cognitive, and emotional engagement. As shown in Table 2, all three dimensions of engagement remained stable across the semester (p > .05; behavioral engagement: R²m = .002, R²c = .811; cognitive engagement: R²m = .001, R²c = .311; emotional engagement: R²m = .006, R²c = .706). This suggested that learners who experienced DDL instruction maintained their engagement across the semester.

The results suggested that the stability of learner engagement was related to the type of classroom interaction. In particular, learners’ behavioral engagement was sustained through hands-on interactions with corpora in DDL activities. As illustrated in Excerpt 4, by interacting with the corpus, Linda was motivated to actively participate in corpus-searching activities: “Using the corpus felt interactive because we could search and discuss examples in real time” (ID: 6, Linda).

4.1.3. Autonomy

The analysis revealed that learners receiving DDL instruction had a statistically significant increase over time (estimate = 0.80, p = .02). This model reported a R²m of .026 and a R²c of .839. As shown in Figure 6, the level of learner autonomy experienced a steady increase from Wave 1 to Wave 2, and a faster increase from Wave 2 to Wave 3, despite variation in individual trajectories. The results indicated that learners under DDL instruction consistently enhanced their autonomy throughout the course. A closer examination of the reported questionnaire results showed learners increased autonomous learning behaviors, including independently selecting learning materials, setting personal study goals, and engaging in self-directed learning activities.

The classroom observations further reinforced the quantitative results of autonomy by observing the progression in students’ independent use of corpus tools. In the first classroom observation, students relied heavily on peers and dictionaries to navigate corpus tools, often pausing to consult others and seeking advice from the teacher. In the second observation, students demonstrated improved corpus skills by quickly inputting keywords and selecting appropriate corpus modules. In the final session, most students exhibited purposeful and autonomous use of the corpus, effectively collaborated, and confidently shared findings.

The thematic analysis identified a primary theme influencing learners’ autonomy during the study; that is, opportunities for independent learning (three quotes). First, abundant learning resources in corpora increased opportunities for independent learning, which in turn supported the development of learner autonomy. One student mentioned that the use of corpora transformed her learning approach from passively relying on teachers’ instruction to more active exploration (see Excerpt 5). She believed that corpora provided abundant language input and allowed her to explore word collocations independently, thus reducing the reliance on teacher guidance. This result demonstrated that corpora could enable students to actively manage their learning process, thereby promoting greater learner autonomy:

Excerpt 5. I used to rely on passive learning, but now I’m much more proactive. The corpus website provided abundant content, allowing me to select collocations to improve my writing instead of waiting for the teacher’s feedback. This is a shift from the instruction I received previously, especially in my high school. (ID: 7, Sara)

Second, hands-on explorations contributed to learners’ independent learning, thus enhancing autonomy. As illustrated in Excerpts 6 and 7, students pointed out that they were required to engage in hands-on explorations and take ownership to solve problems. This active participation encouraged students to develop their independence of learning:

Excerpt 6. We needed to extract information applicable to expository writing from different corpus modules. Each of us had different tasks, and we needed to find ways to solve different problems independently. (ID: 6, Linda)

Excerpt 7. Using the corpus gave me the opportunity to find answers on my own. (ID: 13, Leo)

5.1.1. Foreign Language Enjoyment and Anxiety

The students in DDL instruction maintained a stable level of foreign language anxiety throughout the semester. Regarding practical meaningfulness, anxiety remained below the midpoint of the Likert scale across waves, indicating a low to moderate level of anxiety. The negligible effect size suggested that the change in anxiety over time was minimal, indicating that anxiety remained stable over time. This was consistent with Pawlak et al. (2025), who suggested that corpus-based activities may help learners maintain stable levels of affective experiences over time. As shown in the results (Excerpt 1), the immediate feedback provided by corpus tools allowed her to independently validate her understanding, which may have helped her manage the uncertainty associated with teacher evaluation. This sense of control over the learning process appeared to be a key mechanism for the stabilization of foreign language anxiety. In addition to the affordances of corpus tools, instructional support from the teacher may have contributed to the emotionally supportive learning environment observed in DDL contexts. In this study, the teacher guided learners throughout the learning process by first modeling the use of corpus tools using sample words, encouraging the group discussions, and providing feedback at the end, which might have created a pleasant and comfortable learning environment. During the DDL activities, students were also provided with downloadable worksheets, which were considered effective tools to maximize the teacher’s investment of time and effort and relieve learners’ cognitive load (Boulton, 2010; Johns, 1991; Lenko-Szymanska & Boulton, 2015).

In terms of foreign language enjoyment (the dependent variable for research question 1), learners showed no significant changes. Enjoyment was maintained at a moderate level across the semester. The minimal effect size indicated that time contributed only minimally to changes in enjoyment, indicating that enjoyment remained relatively stable across the semester. This supported the findings in J. M. Dewaele et al. (2018, 2025) that foreign language enjoyment was “trait-like” (J. M. Dewaele et al., 2025, p. 39). One possible explanation might be the complex nature of foreign language affective attitudes, which could be influenced by a diverse range of social, contextual, and psychological factors. Also, the characteristics of DDL activities may have helped sustain learners’ initial positive affective attitudes across the semester. This aligned with previous research suggesting that learners benefited from a “friendly atmosphere” and were “comfortable with the corpus technology and enjoyed discovery learning” (M. Chen & Flowerdew, 2018b, p. 108), as well as feeling “happy with the new skills they had acquired” (Cortes, 2011, p. 76).

In this respect, analysis of the students’ interviews indicated important contributors to the overall maintenance and individual fluctuation of these emotional experiences, including perceived control over the learning process and perceived value of acquired knowledge. Specifically, the DDL approach contributed to a sense of control over the learning process and perceived value of acquiring new knowledge by providing them with direct access to resources. The inclusion of collaborative learning may also have helped learners manage anxiety by providing peer support. This might be attributed to the fact that learners felt supported and connected during DDL activities, which fostered a sense of belonging and helped them manage the emotional burden of performing under pressure (Horwitz, 2017). This finding was consistent with CVT (Pekrun, 2006), which suggested that perceived control and value could foster positive emotional states.

5.1.2. Engagement

It was found that DDL helped learners maintain their engagement to some extent in classroom activities across the semester. In terms of practical meaningfulness, the three dimensions of engagement remained at a moderate level across the semester, with mean scores slightly above the midpoint of the scale. The small effect size values indicated that time explained only a small proportion of the variance, suggesting that engagement remained relatively stable across the semester. Previous research consistently showed that declines in engagement were a persistent issue in L2 language learning, and this pattern of decline in engagement was observed both in formal classroom settings and outside them (Elaish et al., 2019; He & Loewen, 2022); however, in DDL, learners’ engagement was sustained over time. This finding aligned with Zare et al. (2024), who reported that learners under DDL instruction maintained their engagement over the long term. The finding also provided support for the impact of DDL on sustaining learner engagement across time, consistent with previous studies showing that DDL could “engage them in active processing of the words” (Daskalovska, 2015, p. 138) and transform learner’s passivity into active engagement (Lin, 2016). From the perspective of SDT, this sustained engagement might be explained by the way DDL activities provide meaningful tasks, foster learner autonomy and support from peers, thereby meeting learners’ basic psychological needs and promoting continued classroom involvement (Fryer & Oga-Baldwin, 2019; Phung et al., 2021).

Analysis of the students’ interviews revealed the importance of the interaction with corpus data in maintaining engagement in DDL. This finding supported prior research emphasizing the benefits of DDL in fostering the involvement of learners in the active and independent explorations of linguistic patterns through corpus consultation (Chambers, 2019; Charles & Hadley, 2022; T. Liu & Gablasova, 2025). This aligned closely with a key characteristic of DDL, namely its inductive and discovery-oriented nature, which could help EFL learners engage in the learning process (Gilquin & Granger, 2010). First, DDL requires learners to directly engage with authentic corpus data, promoting active discovery and exploration and fostering deeper cognitive involvement (Boulton & Cobb, 2017). Second, DDL positions learners as active researchers and makes the learning process interactive and dynamic (Bernardini, 2004). Third, the inductive nature of DDL stimulates curiosity by challenging learners to uncover rules and structures themselves, making it an engaging and meaningful learning experience (O’Keeffe, 2023).

5.1.3. Autonomy

The analysis showed that learners’ autonomy significantly increased during a semester-long course under the DDL approach. The mean values of autonomy were at a medium-to-high level and demonstrated a clear upward trend across waves. The small effect size indicated that time explained a limited proportion of the variance in autonomy, suggesting limited practical change over time. This might be attributable to learners’ relatively high initial level of autonomy, which left limited room for further improvement, as well as the small sample size in this study. The development of autonomy has been repeatedly pointed out as an important long-term benefit of DDL (e.g., Chambers & O’Sullivan, 2004), where learners acted as researchers and the teacher acted as a facilitator (O’Keeffe et al., 2007). As previously mentioned, DDL provided learners with abundant resources. This was consistent with the constructivist view of learning, which viewed it as a dynamic process driven by the learner’s active explorations of language data (Flowerdew, 2015). DDL also enabled learners to engage in hands-on explorations, which required learners to adopt higher-order cognitive skills such as hypothesis formation, testing, and reasoning (Boulton, 2010; P. Lee & Lin, 2019; Zare et al., 2023). This empowered learners to take responsibility for their learning, thus remembering and internalizing language rules (Ellis, 2003).

The results of this study showed that after the corpus-based activities, learners demonstrated improvements in time management, self-regulation, and autonomous learning behaviors. This suggested that DDL not only maintained students’ engagement in the classroom but also contributed to the development of their long-term autonomous learning. This might be attributed to the task design of DDL, as it required learners to independently explore language resources and solve problems, thereby promoting higher-level cognitive skills (Gilquin & Granger, 2010). These acquired skills could thus be internalized and applied to other learning contexts outside the classroom. This finding was consistent with Benson’s (2013) description of learner autonomy, which refers to learners’ ability to actively regulate their learning processes both inside and outside the classroom. From a broader perspective, this also implied that the autonomy acquired in the DDL activities in class could be transferred to unstructured learning contexts, thereby supporting lifelong learning (Looi et al., 2025).