Introduction

During the COVID-19 pandemic the world witnessed an unprecedented rate of development of digital devices and online resources (including informal digital learning) in all areas of education including L2 teaching and learning. Thus, even after the pandemic era, the development of informal digital learning (such as using online dictionaries, listening to songs in English, watching short videos in English online, talking to people in English through online calls) has drawn the attention of language materials developers and language teachers in both online and in in-person English classrooms. In the online class context, almost all language learners have digital devices and take advantage of them as affordances to improve their language learning skills. In this case, the role of informal digital learning of English (IDLE) can be highlighted. Therefore, exploring the role of IDLE in the language classroom context in association with other individual differences (IDs) seems essential (Lee & Sylvén, 2021). In light of this, there is a need to investigate the role of IDLE as a new emerging ID construct in foreign language classrooms and explore its relationship with other language-domain specific IDs.

IDLE, an important and extremely interesting sub-domain of computer-assisted language learning, is defined as “self-directed English activities in informal digital settings, motivated by personal interests and undertaken independently without being assessed by a teacher” (Lee & Lee, 2021, p. 359). IDLE may offer psychological advantages to students with respect to grit, autonomy, community of practice, flow, and affect (Chik, 2011, 2014; Han & Reinhardt, 2022; Lee, 2022). Since learning a second or foreign language is primarily motivated by the need to communicate, a growing number of SLA researchers have begun to investigate whether there is a connection between IDLE and willingness to communicate in a second language (L2WTC) (Lee, 2022; Lee & Lee, 2020; Mulyono & Saskia, 2021; Soyoof, 2022). These studies have found that IDLE has a positive effect on L2 learners’ L2 WTC. Additionally, in light of the affective turn in the area of applied linguistics, some previous studies have already investigated the role of positive and negative emotions, such as foreign language enjoyment (FLE: Lee et al., 2021) and foreign language anxiety (FLA: Fang & Tang, 2021; Lee, 2022; Lee et al., 2021) as mediators in the relationship between the two constructs.

However, one of the potentially important mediators that has been neglected in previous studies is foreign language boredom (FLB). Considering this emotion while investigating the association between IDLE and L2WTC is crucial since when L2 learners are involved in IDLE activities, diminishing FLB can serve as a tool to facilitate their engagement and satisfaction (Derakhshan et al., 2021a, 2021b, 2022a, 2022b; Pawlak & Kruk, 2022; Solhi et al., 2023) resulting in enhancing L2WTC (Kruk, 2021a, 2021b). Another important issue ignored in prior research is the parallel investigation of the longitudinal development of IDLE, FLB, and L2WTC. Due to the dynamic and developmental nature of these variables, the current project aimed to illuminate the mechanisms of change in IDLE and L2WTC, while examining the way in which FLB can mediate the link between IDLE and L2WTC. To do so, the current study applied latent change score mediation (LCSM) modeling (McArdle, 2009). LCSM models are structural equation models which are suitable for the investigation of latent variables with short term and long-term developmental nature.

Literature Review

Informal Digital Learning of English (IDLE)

As an important and extremely interesting sub-domain of computer assisted language learning (CALL), IDLE is an under-researched area concerning attempts to learn an L2 that occur outside class (Benson, 2011). IDLE has emerged as self-initiated, informal use of English as a global language with the help of a wide array of digital tools (e.g., social networks, chatrooms, and online forums outside the formal context of language learning) (Lee, 2019b). Recent research has shown that IDLE is closely linked to L2 learning outcomes (Lai et al., 2015). Benson (2011) proposed an explanatory theory for the fundamental qualities of IDLE with four aspects of informal L2 learning which include: (1) formality level, (2) place (inside class, outside class, extramural and extracurricular, i.e., the physical setting of L2 learning), (3) pedagogy (the degree to which formal L2 learning procedures are involved, i.e., whether the additional language is taught, self-taught, or acquired naturalistically), and (4) locus of control (the degree to which L2 learners are in control of their L2 learning experience, i.e., whether it is self-driven or other-driven).

However, in previous studies, this type of ID has been treated as a static, monolithic construct (Lee, 2022; Lee & Lee, 2020; Mulyono & Saskia, 2021; Soyoof, 2022). With the growing influence of CDST in applied linguistics, it has been maintained that all IDs per se are complex, dynamic, and multifaceted (see Derakhshan et al., 2023; Freeborn et al., 2022; Hiver & Al-Hoorie, 2019). Thus, attention should move away from the static perspective toward IDLE to a dynamic one. More specifically, Dörnyei and Ryan (2015) argue that the investigation of IDs requires the adaptation of a more dynamic conceptualization with a focus on change in these constructs. Given the time-intensive nature of change in IDs, the dynamic approach to IDLE should incorporate longitudinal investigations. As Larsen-Freeman and Cameron (2008) highlight, longitudinal research is suitable for investigating IDs as complex systems because it enables researchers to explore changes in IDs over time.

Second Language Willingness to Communicate (L2WTC)

L2WTC was initially described as the overall inclination to start communication, referring to the freedom to choose whether or not to do so (McCroskey, 1992). L2WTC is regarded as the final psychological stage preceding the act of actually engaging in communication in L2 (Elahi Shirvan et al., 2019; MacIntyre et al., 1998). MacIntyre and Gregersen (2021) argue that the nature of the construct in question is dynamic and complex. Thus, it should be considered as a both permanent and contextually adaptive construct (MacIntyre & Gregersen, 2021). Hodis et al. (2010) highlighted four gaps in L2WTC studies: (1) no explanation of the developmental mechanism underpinning variations in L2WTC over time was provided, (2) prior studies investigating L2WTC mainly applied cross-sectional designs (e.g. MacIntyre, 1994), (3) previous correlational research explored variables measured once, and (4) research did not clarify “fundamental differences existing between static and dynamic influences of predictors” (Hodis et al., 2010, p. 251). To fill this gap, in the present study, we adopted a CDST approach to address these criticisms.

Foreign Language Boredom (FLB)

Foreign language boredom (FLB) is a relatively little explored and understood negative emotion in the field of SLA (Wang, 2023). The construct has been associated with attention deficit, dissatisfaction, disengagement, modified perception of time and limited vitality (Fahlman, 2009). FLB is considered as the strongest and most frequently felt emotion among students (Goetz & Hall, 2014; Pekrun et al., 2010). It may lead to a loss of interest, diminished engagement in the performance of instructional activities and lack of dedication to learning (Chen & Kent, 2020; Skinner et al., 2009; Veiga et al., 2014). This imperceptible emotional condition (Derakhshan et al., 2021a, 2021b, 2022a, 2022b; Merrifield & Danckert, 2014; Solhi et al., 2023) is still not sufficiently explored in the L2 learning domain, although it has been recognized as the most salient problem in L2 classes, especially in the case of online learning (Derakhshan et al., 2022a, 2022b; Kruk & Pawlak, 2022; Kruk et al., 2023a; Pawlak et al., 2021). The still limited empirical evidence portrays FLB as a distinctive affective condition which can adversely affect the L2 learning experience (Kruk et al., 2022a; Kruk, Pawlak, Elahi Shirvan, et al., 2022a, 2022b, 2022c) and can be experienced with varying degrees of intensity (Elahi Shirvan et al., 2021; Kruk et al., 2021).

Numerous studies have revealed that FLB has a multidimensional and dynamic nature (e.g., Elahi Shirvan et al., 2021; Kruk et al., 2022a; Kruk et al., 2023b). With regard to the dynamic nature of FLB, individual developmental processes (Kruk et al., 2022a; Yazdanmehr et al., 2021) and its sources (Kruk et al., 2022b) have been examined through a person-oriented approach, also known as the idiographic approach. In contrast, the variable-oriented methods, also called the nomothetic approaches, have been applied to investigate the longitudinal validity of its measurement scale (Elahi Shirvan et al., 2021), and its simultaneous development with other IDs (e.g., FLE: Kruk et al., 2022b and L2 playfulness: Kruk et al., 2023b).

Interplay of IDLE, L2WTC, and FLB

A number of research projects carried out in different countries have aimed to explore the degree to which the IDLE might play a part in shaping in L2 learners’ L2WTC in class (e.g., Indonesia: Lee & Drajati, 2019; Iran: Soyoof, 2022; Korea: Lee & Lee, 2020; Taiwan: Lee & Hsieh, 2019). At nearly the same time, SLA research was faced with an emotional turn (White, 2018) or an emotional shift (Prior, 2019), which greatly contributed to the growth of empirical investigations into the effect of emotions on L2WTC (Lee & Hsieh, 2019) and on IDLE (Lee, 2019a) in exam-based Asian settings. Despite the growing number of studies on the relationship between IDLE, emotions, and L2WTC, two gaps in the literature can be recognized.

Firstly, prior studies did not explore the potential influence of FLB, as the strongest and most commonly experienced emotion in the ecology of an L2 class, on the relationship between IDLE and L2WTC. The role of FLB in shaping L2WTC has been thoroughly investigated (Alrabai, 2022; Kruk, 2021a). For example, Kruk (2021b) attempted to capture the links among L2WTC, boredom, motivation, and language anxiety (along with motivation) in Second Life (i.e., a virtual world). The findings indicated that these factors intermingled in a dynamic and unpredictable manner. Kruk (2021b) argued that although a higher/lower level of L2WTC is linked to a lower/higher level of boredom, this relationship did not apply across the board and some discrepancies from this tendency were also noted.

According to this rationale, investigating the role of FLB as a mediator in the relationship between IDLE and L2WTC is necessary because an FLB-provoking learning environment includes physical and cognitive idleness, lack of purpose in learning, reluctance to engage in the learning environment, and disengagement or withdrawal from tasks or activities at hand (Li et al., 2022a, 2022b). All of these are important factors that might influence the link between IDLE and L2WTC. Secondly, previous research adopted a cross-sectional perspective on the relationship between IDLE and L2WTC. However, with regard to the developmental and dynamic character of IDLE, FLB, and L2WTC, researchers have been encouraged to move from cross-sectional to longitudinal studies (Hiver & Al-Hoorie, 2019) as the former fail to capture the details of the dynamics in these constructs. To fill this gap, the current study examined the dynamic developmental interrelationship of IDLE, FLB, and L2WTC based on longitudinal data.

Purpose of Study

Considering the above-mentioned research gaps, the present research employed an LCSM model within the structural equation modelling (SEM) framework to test the association between IDLE and L2WTC with a representative sample of Iranian EFL learners at a university level. The analysis also included testing the mediating role of FLB. The approach was longitudinal and the data were collected in four phases of an L2 course over 4 months. Both the initial state of IDLE, FLB and L2WTC, and their patterns of growth throughout the course were considered and the trajectory of changes was mapped. Three research questions were addressed:

RQ1:

What are the patterns (decreasing or increasing) and the degrees (decelerating or accelerating) of trajectories for IDLE, FLB and L2WTC within latent processes?

RQ2:

How do the trajectories of IDLE and FLB influence the trajectories of L2WTC via longitudinal mediation analysis?

RQ3:

Are there inter- and intra-individual variations in the development of IDLE-FLB-L2WTC system?

Methodology

Participants and Setting

This study included 354 university learners (218 females and 136 male) attending an online general English course at three Iranian universities in three major Iranian cities. The course was a three-credit unit with 16 sessions that started in September 2022 and finished in January 2023. We chose classes where professors employed more technology-based resources for online instruction.

Hair et al. (2010) proposed that the mean values are used to estimate the missing values for the participants on four measurements on circumstances that the response rate was at least 90%. The replies of 22 students were ignored since they had missed two of the four rounds of data collection. To find outliers and extreme values in our data, we used the boxplot method. An extreme outlier was defined as any value that differed considerably from the norm by more than three interquartile ranges (Hoaglin & Iglewicz, 1987). We identified two participants with outliers for IDLE, two participants with outliers for FLB, and three participants with outliers for L2WTC after reviewing the data on four measurement occasions, all of whom were excluded for final analysis. The final data were analyzed based on 325 respondents (197 female and 128 male).

The learners’ English proficiency ranged from lower-intermediate to upper-intermediate, as evaluated by the Oxford Placement Test, and their ages ranged from 18 to 33. All participants were studying English as a foreign language. The current study’s sample size was determined sufficient to achieve 0.80 power to discover a substantial effect size utilizing longitudinal mediation analysis (see Pan et al., 2018). Furthermore, given the nesting of learners within universities and classrooms, an investigation of intraclass correlations (ICC) was conducted and showed a low degree of data class-level dependency (0.03–0.05). Therefore, multilevel analysis was unnecessary.

Instrumentation

Questions were developed for an online survey according to the operational definitions of the variables. The questionnaire was organized into four parts. Reliability of the scales was established by means of Cronbach’s alpha (α) and McDonald’s omega coefficient (ω) in four waves of measurement (see Table 1).

Table 1 Reliability of the scales at the four times of data collection
Full size table

The first part of the questionnaire comprised the participants’ demographic data. It aimed to gain information about the students, including their age, sex, academic year, and international experience. The participants were only asked to respond to this part in the first round of data collection.

The second part of the instrument included the Informal Digital Learning of English (IDLE) questionnaire. The eight-item version of the scale was designed to measure the degree to which L2 learners engaged in receptive and productive IDLE tasks (see Appendix). The scale was extracted from Lee (2022). The IDLE questionnaire has two subscales: receptive IDLE activities (four items: e.g., “I watch English-language YouTube clips”), and productive ones (four items: e.g., “I chat with others in English on social media”). Responses to the question (“How often do you engage in the following IDLE activities?”) are provided on a 5-point Likert-type scale: 1—‘Never,’ 2—‘Rarely (Once a week),’ 3—‘Sometimes (2 or 3 times per week),’ 4—‘Fairly often (Once a day),’ and 5—‘Very often (Many times per day).’

The third part of the questionnaire encompassed the Boredom in Practical English Classes-Revised (BPELC-R) scale (Pawlak et al., 2020). This instrument assesses the degree of boredom that language learners manifest in English language classrooms. The tool consists of 23 items representing two sub-scales: disengagement, monotony and repetitiveness (14 items: e.g., “It would be very hard for me to find an exciting task in language classes”) and lack of satisfaction and challenge (nine items: e.g., “I often have to do repetitive or monotonous things in my language classes”). The items are responded to on a 7-point Likert scale ranging from 1 = ‘I totally disagree’ to 7 = ‘I totally agree’.

The fourth part consisted of the L2 willingness to communicate (L2WTC, see Appendix) questionnaire in the classroom (three items: e.g., “When you have a group discussion in an English class”), partially adapted from Lee (2022). The responses to the question (“How much are you willing to communicate in English in this situation?”) were rated on a five-point Likert scale, ranging from 1 which meant ‘definitely not willing’ to 5 which meant ‘definitely willing.

Data Collection

The questionnaire was distributed to the participants on four occasions, beginning with the first session of the EFL course over the time of 4 months. By collecting longitudinal data, it was possible to track changes in IDLE, FLB, and L2WTC over time. A step-by-step procedure was used to complete the questionnaires. At the start of the course in October 2022, the first round of data collection was conducted. The second, third, and fourth surveys were completed in November, December of 2022, and January of 2023, respectively. The participants were informed about the confidential nature of the information they supplied, and they gave their consent to participate in the study.

Data Analysis

All LCS analyses were performed using Mplus 7.4. (Muthen & Muthen, 2013). The univariate and multivariate normality tests were conducted. The dynamic parallel processes of three constructs (IDLE, FLB, and L2WTC) was investigated employing MLCSM models, which enabled modeling the dynamic progress of constructs over time in terms of both within-person variability and between-person variability.

Model Fit

Model fit was tested sequentially in line with Grimm et al. (2005). The following LCS models for the three variables were separately tested to determine the best model fit. According to Grimm et al. (2005), model fit was checked step by step. To obtain the best model fit, the following LCS models for three constructs were examined separately:

  • Model 1. No-change model: a lack of any change over time.

  • Model 2. Constant change model: linear variation through time

  • Model 3. Proportional change model: variation in the system between adjacent measurements

  • Model 4. Dual-change model: a combination of constant and proportional change parameters

After evaluating the univariate models, the mediation model was tested, with changes in FLB and FLE as mediators of the relations between variation in IDLE and variation in L2WTC. In order to evaluate the model fits, χ2 tests was used across each pair models (Harrington, 2009). Given that the χ2 assessment is extremely sensitive to the sample size (Browne & Cudeck, 1992), root mean square error of approximation (RMSEA: < 0.08), Tucker-Lewis Index (TLI: > 0.90), and comparative fit index (CFI: > 0.90). The researchers used Δχ2 and ΔCFI to evaluate overall model fit criteria. If the p-value of Δχ2 is significant and the ΔCFI value is more than 0.010, we may infer that the difference is statistically significant.

Measurement Invariance

It is critical to test the invariance of measurement models to guarantee that comparisons of latent variables are reliable throughout time at least at the configural and scalar levels (Wickrama et al., 2021). Thus, in the next stage, we ran longitudinal confirmatory factor analyses (longitudinal CFA; Wickrama et al., 2021) to assess measurement invariance of the predictor, the mediator, and the outcomes at T1, T2, T3, and T4.

Test of Mediation

To examine mediation, we computed confidence intervals for the result ab. Because the product of a and b has a nonnormal distribution, asymmetric confidence intervals must be calculated using Monte Carlo techniques such as bootstrapping (MacKinnon et al., 2007). The percentile bootstrap technique for establishing confidence intervals for mediation using structural equation models takes into account the possibility of the relationship between a and b and has a suitable balance of power and Type I error (Valente et al., 2018). In the current investigation, we employed these confidence intervals to evaluate the indirect impact ab alongside the joint significance evaluation.

Results

Preliminary Analyses

Bivariate Correlations among Variables

The correlations for the IDLE, FLB, and L2WTC among the four time-series are shown in Table 2. IDLE, FLB, and L2WTC demonstrated large within-system correlations. Besides, the analysis showed that there were overall moderate and positive cross-system associations between IDLE and L2WTC across the four waves and moderate and negative cross-system associations between FLB and L2WTC as well as between IDLE and L2WTC.

Table 2 Bivariate correlations among the three variables in the present study
Full size table

Model Fit

Univariate Latent Change Score Model

To assess within-system growth patterns across four times, competing models were evaluated independently for IDLE, FLB, and L2WTC. The evaluation of model fit was made via the consideration of several indices.

Table 3 displays fit metrics for univariate IDLE, FLB, and L2WTC models. Subsequent analysis of nested models showed that the univariate dual variation model of IDLE, FLB, and L2WTC fit relatively better than both the constant change model and proportional change model. Thus, due to the need for both constant change and proportional change to properly characterize IDLE, FLB, and L2WTC changes across four periods, these analyses together supported dual change as the optimal univariate model for IDLE, FLB, and L2WTC.

Table 3 Comparative Univariate IDLE, FLB and L2WTC Model Fit
Full size table

Latent Change Score Mediation Model (LCSM)

The LCSM model was tested in which changes in FLB as a mediating construct in the subsequent associations between changes in IDLE and that of L2WTC. Model indices showed an acceptable model fit for the LCSM model. Therefore, the results of the mediation models indicated that FLB is a mediating mechanism for the longitudinal association between IDLE and L2WTC.

Measurement Invariance

The researchers assessed the configural, weak, and strong invariance of latent variables in univariate and multivariate LCS models over time. The results of the configural (i.e., free factor loadings), weak (i.e., factor loadings invariant) and strong (i.e., factor loadings invariant) invariance measurements are presented in Table 4. The results of the analysis of measurement invariance supported that weak and strong invariance models did not significantly different in fit for univariate and multivariate LSC models. changes in CFI, TLI, RMSEA, and SRMR were less than suggested cutoff scores (ΔCFI ≤ 0.010, ΔRMSEA ≤ 0.015, and ΔSRMR ≤ 0.030, Cheung & Rensvold, 2002). These results supported measurement invariance for all models across time.

Table 4 Measurement invariance across time in different LCS models
Full size table

Pattern and Degree of Trajectories of IDLE, FLB and L2WTC

Regarding the pattern and degree of trajectories of the variables under investigation, estimations of different parameters from the LCSM model are shown in Table 5. The initial means of IDLE, FLB, and L2WTC during the first measurement were significant (μI−IDLE = 5.647, p < 0.001, μI−FLB = 7.142, p < 0.001, μI−L2WTC = 3.471, p < 0.001). Significant and positive linear growth in IDLE and L2WTC (μS−IDLE = 0.276, p < 0.001, μS−L2WTC = 0.314, p < 0.001) and negative linear growth in FLB (μS−FLB = −0.431, p < 0.001) was observed across the four measurements. The proportional change (self-feedback parameter) components were significant and negative for all three variables (βIDLE =  − 0.016, p < 0.001; βFLB =  − 0.039, p < 0.001; βL2WTC =  − 0.046, p < 0.001). Overall, the negative proportional change parameters taken with the positive slope mean for IDLE and L2WTC indicated that IDLE and L2WTC increased over time, and those increases decelerated with each successive wave. By contrast, FLB decreased over time, and this decrease decelerated with each successive wave.

Table 5 Estimates from the LCSM model
Full size table

Influence of IDLE and FLB Trajectories on the L2WTC Trajectory

Coupling parameters (i.e., a, b, and c') were used to define the mediation paths. The study revealed that earlier levels of IDLE predicted the later reduction in FLB adversely (i.e., coupling from IDLE to FLB; a = −0.071, p < 0.001), and they significantly predicted further increase in L2WTC over time (i.e., coupling from IDLE to L2WTC; c' = 0.045, p < 0.001). Earlier FLB levels had a significant influence on the prediction of a subsequent increase in L2WTC (coupling between FLB and L2WTC; b = −0.083, p < 0.001).

Thus, using joint significance analysis allowed the researchers to establish that earlier levels of IDLE substantially predicted later decrease in FLB, and earlier levels of FLB strongly and negatively predicted subsequent changes in L2WTC. The 95% percentile bootstrap confidence interval of the outcome of the coupling parameters ab likewise did not entail zero, 95% CI = [0.02, 0.07], supporting the existence of mediation and validating the results of the joint significance analysis.

Inter- and Intra-individual Variations in the Development of IDLE-FLB-L2WTC System

The interindividual variations at the latent level of the constructs were determined by variance (σ2), illustrated in Table 5. The results showed significant interindividual variations at the initial level (σ2I−IDLE = 0.476, p < 0.001; σ2I−FLB = 0.243, p < 0.001; σ2I−L2WTC = 0.211, p < 0.001) and at the subsequent development of three constructs (σ2S−IDLE = 0.017, p < 0.001; σ2S−FLB = 0.006, p < 0.001; σ2S−L2WTC = 0.007, p < 0.001). More specifically, the variances of three variables decreased over time, indicating that at the beginning of the study the levels of IDLE, FLB, and L2WTC differed substantially between participants, but eventually they reached closer levels over time.

Intraindividual variations were determined by self-feedback parameter (ß) and covariance between intercept and slope (σI−S), illustrated in Fig. 1 and Table 5. The self-feedback parameters were negative for all three variables (βIDLE =  − 0.016, p < 0.001; βFLB =  − 0.039, p < 0.001; βL2WTC =  − 0.046, p < 0.001). The negative self-feedback parameters of IDLE, FLB, and L2WTC mean that within the short-term changes of the three variables, the participants with higher IDLE, FLB, and L2WTC at t−1 experienced smaller changes at time t compared to the participants with lower IDLE, FLB, and L2WTC.

Fig. 1
figure 1

LDSM model of IDLE predicting L2WTC, mediated by FLB. Unlabeled paths are set at 1. IDLE informal digital learning of English, FLB foreign language boredom, L2WTC L2 willingness to communicate inside the class, I Intercept, S Slope. *p < .05; **p < .01; ***p < .001

Full size image

Also, within the long-term changes of the three constructs, intraindividual differences showed significant and negative covariances between intercept and slope in IDLE and L2WTC (σS−IDLE/I−IDLE = −0.021, p < 0.001; σS−L2WTC/I−L2WTC = −0.058, p < 0.001), which indicates that the participants with lower initial IDLE and L2WTC scores experienced more changes in these variables than those with higher initial IDLE and L2WTC scores. On the other hand, a significant and negative covariance between intercept and slope in FLB was found (σS−FLB/I−FLB = 0.061, p < 0.001) indicating that the participants with lower initial FLB scores manifested less change in their boredom than those with higher initial FLB scores.

Discussion

This research project was inspired by a lack of investigations on the impact of changes in IDLE, an implicit approach to learning, on changes in L2WTC. Additionally, the role of changes in FLB, the strongest and most commonly experienced emotion in language classes, as mediator between the development of IDE and L2WTC were investigated. The present study showed that it is essential to longitudinally explore L2 learners’ informal digital learning, boredom and willingness to communicate to provide a more realistic, dynamic view of these variables as they emerge out of the live experience of L2 learning.

Regarding the first research question, considering the constant change component of the three variables over time, the results indicated a systematic increase in IDLE and L2WTC but a systematic decrease in FLB during the language course. However, the proportional change component confirmed that the increasing development of IDLE and L2WTC and the decreasing trend of FLB were decelerated from Measurement 1 to Measurement 4. This means that the growth of IDLE and L2WTC was faster at the early stages compared to later phases of the language course. Additionally, learners’ levels of FLB decreased faster at the beginning of semester compared to the end of semester. These results confirm that the early sessions of the course can possibly represent a critical period to influence learners’ IDLE, FLB and L2WTC. Thus, L2 teachers should be very meticulous in the early sessions of their courses to develop their learners’ IDLE and L2WTC as well as decreasing their FLB. Lee and Drajati (2019, p. 176) provided an extensive explanation of how IDLE can be integrated using a three-stage continuum paradigm:

  • Stage 1 (in-class CALL): technology can be integrated into the classroom syllabus as a supplementary material.

  • Stage 2 (extracurricular CALL): teachers can engage learners in extracurricular tasks outside of class that enable them to learn and apply a variety of digital technology while accomplishing assignments.

  • Stage 3 (extramural CALL): according to evaluation of learners’ digital device equipment out of class as well as their learning requirements and interests, L2 instructors can assist learners in creating their own IDLE setting in which they can constantly learn and use English in different IDLE settings without the involvement of a teacher.

During Stage 1 and Stage 2, IDLE could be a valuable addendum to formal language classes. The final pedagogical aim for L2 instructors is to help learners develop their own IDLE activities in which they can continually learn and use English in different IDLE tasks independent of formal L2 learning.

With respect to the second research question, that is, the association between the trajectories of IDLE, FLB, and L2WTC, the findings revealed that changes in IDLE affected the L2WTC’s growth significantly through the trajectories of FLB. The significant role of IDLE on L2WTC is supported by the results of previous studies (e.g., Lee, 2022; Mulyono & Saskia, 2021; Soyoof, 2022). For example, Lee and Sylvén (2021) revealed a positive association between IDLE and L2WTC in English in class-based learning. The positive effect of changes in learners’ IDLE on changes in their L2WTC can be discussed in light of the psychological benefits that IDLE has for L2 learners by creating a sense of flow and lowering their affective filter (Chik, 2011, 2014; Han & Reinhardt, 2022; Lee, 2022; Li et al., 2022a, 2022b). Firstly, in online L2 settings, when individuals participate in implicit training while being engaged in IDLE, they tend to completely enjoy the process, resulting in a flow state or being “in the zone.” In fact, these IDLE learners frequently lose track of time and become fully engaged in L2 learning, as a result of which they can increase their L2WTC level. Thus, it could be concluded that engaging in IDLE appears to play a highly positive role in helping L2 learners increase their L2WTC, paving the way for easier and more genuine L2 communication (Lee et al., 2022). Secondly, learners in classroom contexts with IDLE-rich activities could learn and practice English in an environment conducive to lowering the affective filter (less monotony, boredom and anxiety) (Kiaer et al., 2021; Kruk, 2022; Li & Dewaele, 2021; Li & Wei, 2022). This, in turn, leads to an increase in their level of L2WTC.

As the results showed, the trajectories of FLB were a statistically significant mediator for the developmental relationship between IDLE and L2WTC. One plausible explanation for the negative link between IDLE and FLB can be the use of various modalities when engaging in IDLE activities. As previously stated, contemporary EFL learners have access to and use English through a variety of online resources (e.g., online dictionaries, news broadcasts, social media, online forums, digital games, email, e-books, movies, songs, comedies or dramas, YouTube, and Google). From a multimodal point of view, learning through and employing a range of digital texts, audio, pictures, and animated images can improve students’ comprehension and acquisition of a target language, resulting in a decrease in FLB. Overall, it seems that IDLE has the potential to trigger ideal psychological states characterized by low boredom, more engagement, an active and creative approach to learning as well as the use of English for its own sake when engaging in activities of one’s own choice. Consequently, a low self-perceived level of boredom can motivate EFL learners to become more willing to communicate in English in online classes (e.g., Kang, 2005; Lee, 2022).

Considering the third research question, the findings revealed two sources of variation in the development of the IDLE-FLB-L2WTC system: (1) interindividual variation and (2) intraindividual variation. The interindividual variations of IDLE, FLB, and L2WTC were reduced over time as they were constantly multiplied by the negative self-feedback component, indicating that the between-person differences converged with time. Therefore, it can be inferred that individuals’ levels of IDLE, FLB, and L2WTC initially diverged significantly, but subsequently converged.

The convergence of interindividual variation of the three constructs through time can be explained by the concept of peer contagion (Dishion & Tipsord, 2011). Peer contagion is a process of reciprocal impact that happens between an individual and a peer. Peer contagion mechanisms are prominent in natural peer interaction contexts such as the language classroom. In this context, learners can be involved in relational behaviors that lead to companionship, and unintentionally impact themselves or others. As reflected in the results of this study, it can be postulated that over time the levels of IDLE, FLB, and L2WTC in the participants tended to converge as a result of the interactions with their peers.

Intraindividual variations were another source of variation in the developmental system of IDLE, FLB, and L2WTC. As represented in the results, the participants with lower initial IDLE and L2WTC experienced more variation in these variables than those with higher initial IDLE and L2WTC. Those with lower initial FLB levels, on the other hand, showed less change than those with higher initial FLB levels. An important consideration at this juncture is the ergodicity issue concerning individual difference variables (Lowie & Verspoor, 2019). Consistent with the complex dynamic systems theory that takes into account individual differences, students are not regarded as ergodic ensembles and the mean scores of their trait or behavior are not always indicative of the reality of the target trait or behavior among learners. Thus, as shown in the current results, on average, learners were changing in a systematic way, yet there were interindividual variations around the mean score of change in the three variables. Case studies are thus needed to uncover the subtleties of the change trajectory in these variables over time. Besides, the findings of this study regarding intraindividual variations might provide supporting explanations for using LCSM modeling to analyze inter-and intraindividual variation in developmental trajectories.

Conclusion

The present study has extended our understanding of the impact of trajectories of FLB on the association between the growth of IDLE and L2WTC, highlighting the role of this negative emotion as an important mediator. The longitudinal mediation analysis confirmed the significant role of FLB as the mediator. The analysis also provided evidence for the different rates of growth of the three variables under investigation during different periods of the L2 course, providing detailed insights into the increasing and decreasing patterns of change. This said, in view of the fact that this study only included EFL students with almost no international experience, its results and implications might be more relevant to EFL learners in Asian contexts including students who practice the English language not only in the classroom but also in extracurricular digital contexts (Toffoli & Sockett, 2015).

The present investigation provides a basis for some pedagogical recommendations. First, teachers can be advised to help EFL students benefit from IDLE activities matched to their own needs and interests. Second, teachers could try to ask more active informal digital learners of English to describe their individual experiences in this respect by introducing the most effective digital media (e.g., English-language movies, podcasts, audiovisual tracks), learning strategies (e.g., frequently engaging in reviews or listing useful words and idiomatic phrases for subsequent use), and benefits of practicing English with the help of digital resources (e.g., increasing the knowledge of English lexis, honing receptive and productive target language skills; Lee, 2022). Third, teachers are also recommended to help students create groups with peers sharing their interests and engaging in similar IDLE practices. IDLE activities can help EFL learners become more conscious of IDLE via sharing content with peers or engaging in learning English on an individual basis by means of involvement in IDLE practices in and out of the class context.

Regarding the limitations of the current study, collecting longitudinal qualitative data was not feasible in the case of this empirical investigation. We are fully aware that the incorporation of a qualitative emic perspective could have provided us with potential explanations regarding the interplay of the changes in the three variables over time. Hence, we propose that future research projects consider a mixed-methods design in order to provide further evidence on mechanisms of change in the temporal co-development of the three variables.