Ellis and Barkhuizen (2005) stated that writing performance is composed of language complexity, language fluency, and language accuracy. The development of writing performance plays an important role in facilitating writing ability. Thus, the measurements of language complexity, language fluency, and language accuracy in EFL writing performance have long been acknowledged as a major goal of research in the development of writing ability. Numerous studies have been conducted to explore a variety of measurements of language complexity, language accuracy, and language fluency. For example, Wolfe-Quintero et al. (1998) examined more than one hundred measures of complexity, fluency, and accuracy in order to verify the characteristics of different indicators in measuring language accuracy, fluency, and complexity and the categories of these indicators. Cumming et al. (2005), based on Wolfe-Quintero’s research synthesis, by statistical analysis, found that the number of words per minute (W/M) can be used to measure the fluency of language, the number of errors per T-unit (E/T) and the ratio of errors to the total number of words (E/W) can test the accuracy of language, and the ratio of word type to word token (TTR) can be used to measure lexical complexity. Similarly, Lu (2010), by the research synthesis of Wolfe-Quintero et al. (1998) and Ortega (2003), described the automatic analysis of syntactic complexity in writing performance using 14 different measures proposed in studies of second language development, facilitating and extending in EFL writing performance research.

The measurement of language complexity, accuracy, and fluency has provided the basis for further investigating EFL writing performance, and the research has begun to concern about the development of language complexity, language accuracy, and language fluency from the perspective of cognitive factors, such as task complexity, task difficulty, task types, and task genres in EFL context (Robinson, 2001; Malicka, 2020; Awwad and Tavakoli, 2022; Tao and Wang, 2022). For example, the typical study conducted by Robinson (2001), based on the research findings conducted by Widdowson (1990), Long (1998) and Skehan (1998), described a framework for examining the effects of the cognitive complexity of tasks on language performance and learner perceptions of task difficulty, with the results indicating that the significant effects of task complexity and task difficulty on the dimensions of language complexity and language accuracy. Among the previous researches, the similarity was that the dimensions of writing performance were investigated from the perspective of cognitive factors or task factors, with the results showing the significant effects of both task complexity and different tasks on writing performance. The differences were that Robinson (2001), in his study, shed light on the effects of some affective factors such as confidence, motivation, and anxiety on writing performance, with the results that motivation, as well as confidence affected the writing performance, but concurrently emphasized the important role played by classroom anxiety in writing, and appealed to supplementing the following research about the relation between affective factors such as anxiety and the dimensions of language performance by adopting more extensive and detailed instrumentation.

To sum up, the development of EFL writing performance is not only affected by cognitive factors but rather affected by learning emotions. In order to fill the gap in the studies of the three indices of EFL writing performance, the present study attempts to incorporate task complexity and the learning emotions into EFL writing performance and gain insights into how task complexity and learning emotions such as boredom, enjoyment, and anxiety may influence language complexity, language fluency, and language accuracy.

Fehr and Russell (1984) have noted that emotion is not difficult to understand until we are asked to give its definition. Damasio (1994) defined the term emotions as changes in the body state in response to a positive or negative environment. Nowadays, emotions are considered to be the factors to direction the behavior. Oatley and Jenkins (1996) claimed that emotions are the central part of human mental life, which affects our physical well-being. Even research shows that afflictive emotions can make one ill, and positive emotions can tend to promote health (Goleman, 1997). Overall, the above definitions have presented the possible relationship between emotions, social environment, and human behaviors.

Not until positive psychology (PP) was introduced into second language learning, the new trends in the literature on emotions in SLA have emerged in the EFL educational context (Li et al., 2018; Dewaele, 2019; Li and Xu, 2019; Prior, 2019; Tsang and Lee, 2023). From the perspective of PP, the research on emotions in SLA is conducted under the theoretical framework of the broaden-and-build theory (Fredrickson, 2001) and control-value theory (Pekrun, 2006) to explore the influence of positive emotions and negative emotions on foreign language learning performance (Dewaele and Li, 2020). The broaden-and-build theory has played a part in building the “emotional wave” (MacIntyre and Gregersen, 2012; Dewaele and MacIntyre, 2016), which holds that positive emotion and negative emotion are two types of emotions with different functions; positive emotions have positive effects on cognition, motivation, physiology, psychology, and social resources, while negative emotions shrink short-term and long-term resources and affect individuals’ cognition of the environment (Fredrickson, 2001).

At present, control-value theory is introduced into SLA research and expands the perspective of SLA research (Li et al., 2018; Han and Hyland, 2019). Pekrun (2006) proposed the conceptual framework of learning emotion based on this theory, and then Pekrun and Linnenbrink-Garcia (2012) defined learning emotion from three dimensions: valence (positive emotion and negative emotion), activity (emotional arousal), and orientation (arousal of emotional objects). For example, enjoyment is classified as a positive high-arousal emotion and hopelessness as a low-arousal emotion (Dewaele and MacIntyre, 2016). There have been some studies on the relationship between emotions and learning achievement. For example, the relationship between anxiety and writing achievement (Zabihi et al., 2020) and the positive emotion (foreign language enjoyment) on foreign language achievement (see Dewaele and Alfawzan, 2018; Saito et al., 2018). The above-mentioned studies shed light on the positive effect of high-level positive emotions on language achievement and the negative effect of high levels of negative emotions on language learning. However, little attention has been concentrated on how the diverse-specific state emotions like writing anxiety, writing boredom and writing enjoyment appeared in the process of writing tasks administering, and how the different levels of emotions affect the dimensions of writing performance along with the cognitive factors like task complexity. EFL writing has a complicated cognitive process, and learners are prone to produce a variety of emotions (Cotterall, 2011). The diverse-specific state emotions may arise from the writing task complexity. Robinson (2001) examined the relationship between task complexity and task difficulty in task production, suggesting the extent to which task complexity can predict the learners’ stress, confidence, interest, and even motivation. Although the prior studies, in the domain of task-based instruction, paid attention to learners’ different emotions in perspective of task difficulty, the research instrumentation should have been extended and detailed. For example, adding more items to the different emotion variables based on preceding studies. Additionally, the effect of multiple high- or low-specific state emotions on the dimensions of EFL writing performance and the effect of high or low emotions at different writing task complexity on EFL writing performance also needed to be considered.

Task complexity is related to the attentional, memory, reasoning, and other information processing demands imposed by the structure of the task on the language learner (Robinson, 2001). Thus, it is a cognitive factor. The impact of task complexity on foreign language learning production has attracted scholars’ attention in the field of SLA (e.g., Kormos and Trebits, 2011; Kuiken and Vedder, 2011; Benson, 2015). Numerous studies are based on the cognitive hypothesis (Robinson, 2001, 2011, 2015) and competition hypothesis (Skehan, 2009) as the theoretical basis for exploring the influence of task complexity on EFL learning. According to the cognitive hypothesis, the two dimensions of resource-directing and resource-depleting are important factors affecting the complexity of the task; these two dimensions have different effects on the allocation of attention resources and the resource-directing dimension by increasing the difficulty of the task, and the cognitive requirements of conceptualization can direct attention resources to language performance, resulting in more accurate and complex language output and lower language fluency. As the resource-depleting dimension increases, learners will consume more attention resources and thus pay less attention to language performance, affecting language accuracy, complexity, and fluency. According to the competition hypothesis, learners are affected by the capacity of working memory, so the increase in task complexity can only make learners pay priority to either language accuracy or complexity. Nowadays, studies have explored the relationship between task complexity and L2 writing performance, but with inconsistent results. For example, Yan and Zhang (2015) pointed out that there was not a significant main effect of task complexity on language complexity and fluency in EFL writing performance, but a significant task complexity and self-efficacy interaction on EFL writing accuracy. Wang et al. (2020), based on 616 Chinese English learners, comprehensively investigated the impact of task complexity on language complexity, with the findings indicating that task complexity has an impact on different dimensions of language complexity. The above inconsistent findings may be the ignorance of the effects of individual factors, such as learning emotions, on the dimensions of writing performance. Furthermore, Jin and Wang (2021) not least stressed that future research could explore the interaction between task complexity and different learner factors as well as cultural environment, suggesting that cognitive factors like task complexity as well as learner factors like positive and negative emotional factors should be integrated into the development of EFL writing performance.

The present study seeks to address this gap. Drawing on Pekrun’s (2006) learning emotions such as anxiety, boredom, and enjoyment and Robinson’s (2001) concept about task complexity, it added the assessment of L2 proficiency based on a dedicated reading and writing proficiency test as the main dependent variable. It attempts to pursue two principal objectives: first, to explore the relation between task complexity and multiple learning emotions, and second, to investigate the possible interaction effects of task complexity and diverse high or low emotions on the dimensions of EFL writing performance. Specifically, we attempt to investigate the following research questions (RQS) among English major students in a Chinese EFL context.

To better elaborate on the preceding literature and the above research questions, we, drawing on the work by Rahimi and Zhang (2019), have synthesized the conceptual information in the current study, with the details indicated in Figure 1.

By convenient sampling, the participants were 84 Chinese mainland EFL learners, 8.9% males and 91.1% females, aged 19–20, who were university students majoring in English at a university in Hubei Province, China. They were all sophomores from three different classes, with 28 students from Class 1, 10.3% males and 89.7% females, with 28 students from Class 2, 7.4% males and 92.6% females, and 28 students from Class 3, which were recruited using convenience sampling.

The number of 84 students in the present study, based on the measurement results collected by questionnaires of writing anxiety, writing boredom, and writing enjoyment among 177 students of Class 1, Class 2, and Class 3, has been divided into six subgroups. To be specific, questionnaires of writing anxiety, writing boredom, and writing enjoyment are administered in three classes at the same time; we employed the way to sort the groups suggested by Argyrous (2000), with 25% of the high score and 25% of the low score as the boundary between high score group and low score group. Finally, we confirmed six subgroups, with the subgroup information demonstrated in Table 1. The six subgroups, with high or low levels of respective writing emotions, are selected to be between-subject factors due to high- or low-specific emotions exhibiting a big influence on EFL learning achievement (Dewaele and MacIntyre, 2016).

All participants were native Chinese speakers who shared a very similar social and cultural background. All learners, by the time of their admission into university, had learned English in primary school for 3 years and in high school for 6 years, and they were non-native English speakers, with a little concentration on spoken language use in high school education. Additionally, all the students have taken the National English Test authorized by the Chinese foreign test center, and additionally, it has been analyzed, by one-way between-subject ANOVA, that there is no significant difference in TEM-4 scores among six subgroups (p = 0.059 > 0.05). To further exclude the effect of other emotions on EFL writing performance between respective high or low emotions, paired-sample T-test was conducted among different emotions, indicating, between writing anxiety groups, that there is no significant difference in writing enjoyment (p = 0.090 > 0.05) and writing boredom (p = 0.802 > 0.05). Between writing enjoyment groups, there is no significant difference in writing anxiety (p = 0.342 > 0.05); however, there is a significant difference in writing boredom (p < 0.05), which needs to be further discussed. Between writing boredom groups, there is no significant difference in writing anxiety (p = 0.135 > 0.05) and writing enjoyment (p = 0.464 > 0.05).

The investigation and the different writing tasks were administered in three classes during class time. The specific process of research was presented in the following details.

To begin with, the investigation of writing anxiety, writing enjoyment, and writing boredom is administered to the students of three classes (N = 177) in the first week of the Autumn semester of 2023, with the aim of finding out six different subgroups as Between-subject factors and exploring the general state of writing anxiety, writing enjoyment, and writing boredom among the three class students. Then, the administration of the writing task 1 was given to the students of three classes in the second week of this semester to obtain the data of the indicators of language complexity, language fluency and language accuracy, followed by the administration of questionnaires of writing anxiety, writing enjoyment, and writing boredom with the same items as those at the beginning of this semester. Finally, after 1 week, writing task 2 was administered to the students of three classes again to extract the quantitative data of the indicators of language complexity, language fluency, and language accuracy, followed by the same questionnaires to obtain the data of six subgroups.

Prior to the investigation, we contacted the instructors of the three class students and also got their permission to collect data in the above classes. Participation was voluntary, and anonymity was guaranteed for all the participants. Data were matched with the help of a code system in SPSS 23.

The present study, based on fluency, accuracy, and complexity of writing performance (Ellis and Barkhuizen, 2005), collects the indexes of language performance. Specific language performance indicators are shown in Table 4. The accuracy measurement indicator is an error-free t-unit rate (EFT/T) (Ellis and Yuan, 2004); the complexity measures include sentence complexity and lexical complexity. Sentence complexity adopts the number of T-unit clauses (C/T), the number of subordinate conjunctions used (SC/T), and the average length of t-unit sentences (MLT). Lexical complexity adopts lexical diversity (TTR) (Johnson, 2017). Fluency is measured by the number of words produced per minute (W/M) (Cumming et al., 2005). The following indicators in Table 3 are assumed to be dependent variables in the present study, the six subgroups being between-subject factors and two tasks within-subject factors.

The language complexity and fluency indexes can be measured automatically by the online sentence analysis software Coh-Metrix 3.0 (see Ye, 2013). The language accuracy index in the present study is calculated by two raters who have been teaching English for 15 years. To begin with, two raters analyzed 20 essays to collect the grammatical errors in terms of syntactic errors, morphological errors, and lexical errors (see Plakans et al., 2019). It has been found that the inter-coder and inter-rater reliability for language accuracy in terms of grammatical errors is significantly correlated (syntactic errors = 0.832; morphological errors = 0.89; lexical errors = 0.901; error-free t-unit rate = 0.82). Then, the 354 essays were distributed so that each rater got 177 essays for errors. Error-free t-unit rate (EFT/T) was, by extension, computed.

All the data analysis was conducted using SPSS 23.0 and AMOS 7.0, and the specific data analysis procedure is presented as follows. First, all the data will be submitted to SPSS 23.0 to process the missing values of the data, and meanwhile, the reliability analysis of writing anxiety, writing boredom, and writing enjoyment is demonstrated using Cronbach’s alpha (N = 177). Then, the processed data by SPSS is imported into AMOS 7.0 to conduct the CFA of questionnaires (see Moustaki et al., 2004), with the common model fit indexes that AMOS provided in the current study presented in the Instruments section.

After the reliability and validity analysis of questionnaires, SPSS 23.0 is used to process the quantitative data of three-time questionnaires and the indicators of language complexity, language fluency, and language accuracy in different writing tasks. Following descriptive statistical analysis, repeated measures tests, and 2 × 6 two-factor analysis of variance are employed to shed light on the changes in writing anxiety, writing boredom, and writing enjoyment after administering the two different writing tasks, and the learning emotions and task complexity interaction on the indicators of language complexity, language fluency, and language accuracy involved in the present study.

In line with the previous studies on the effect of task complexity on learners’ perception of task difficulty, the first research question explores the effects of task complexity on learning emotions. The findings indicated, in Table 5, that the cognitively defined difficult writing task involved in the present study can lead to greater writing anxiety, while simple writing tasks can lessen writing anxiety, suggesting that English major students in China may demonstrate greater writing anxiety with more difficult writing tasks. The result of the present study corresponds with the study conducted by Robinson (2001), who concluded that the difficult task could result in higher stress significantly, causing, by extension, anxiety in L2 writing. Additionally, L2 writing itself is a very complicated cognitive process; learners often have a certain degree of writing anxiety, including fear and aversion, as well as anxiety behavior out of cognitive thinking obstruction (Jalili and Shahrokhi, 2017), thus, that is understandable, given that the simple writing task needs few resource depletion for learners to finish the writing task easily, difficult writing task would no doubt bring the learners some stress out of more cognitive load in achieving the task, and this stress, by extension, would result in writing anxiety. Meanwhile, in the present study, we unexpectedly found that English major students in China who were given difficult writing tasks significantly induced less writing boredom, and this result can be illustrated in the prior literature. According to the definition of foreign language boredom proposed by Li et al. (2023a), foreign language boredom, as a negative emotion with extremely low activation, arises from activities that are over-challenging or under-challenging. Without hesitation, writing boredom, as a specific negative emotion in writing, can be affected by the writing tasks. Writing tasks, whether they are over-simple or over-difficult, may induce writing boredom. Additionally, writing boredom can be affected by individuals’ L2 writing competence, which is evidenced by the empirical study conducted by Li et al. (2023b). Due to the difficult writing task in the present study not over-challenging for English major students in China, the students may have less writing boredom while engaging in the appropriately difficult writing task. However, we have not found that task complexity had an impact on writing enjoyment, suggesting that the learners, as writing tasks increase in cognitive complexity, may finish the writing task without the loss of foreign language enjoyment, which is more or less consistent with the previous study by Robinson (2001), who proposed that task complexity was not related to the interest or motivation. Overall, the findings, by comparison with the research conducted by Robinson (2001) and Li et al. (2023b), presented evidence that negative emotions like writing anxiety and writing boredom are likely to arise from writing difficult tasks in EFL writing education, the extent to which it can extend the scope of the research about the mechanism of how the different writing task complexity affects the learning emotions.

This present study examines the effects of task complexity and foreign language writing emotions’ interaction on the dimensions of EFL writing performance. In order to examine the task complexity and writing emotions interaction on the dimensions of EFL writing performance, we have divided the learning emotions of writing anxiety, writing enjoyment and writing boredom into six subgroups as between-subject factors and task complexity as within-subject variables. Then, A 2 × 6 between-in-subject ANOVA analysis, shown in Figures 2–4, found significant task complexity and learning emotions interactions on the dimensions of language fluency, language accuracy, and language complexity, respectively.

For task complexity and writing anxiety interaction on the dimensions of language fluency, language complexity, and language accuracy, we have found that English major students in China with low writing anxiety can lead more production of language fluency, language complexity, and language accuracy than any other subgroups in the current study, and these findings once further confirmed the essential effects of writing anxiety and task difficulty on EFL writing achievement, which has also been evidenced by the prior studies, for example, for the effect of task complexity on language achievement (see Kuiken and Vedder, 2008; Ong and Zhang, 2010), and for the effect of writing anxiety on language achievement (see Cheng, 2004; Rahimi and Zhang, 2019; Tahmouresi and Papi, 2021; Li et al. 2023b). However, compared with the previous studies, we looked at an investigation of writing anxiety and task complexity interaction and found there was significant writing anxiety and task complexity interaction for language fluency, language complexity, and language accuracy, respectively. To be specific, writing anxiety and task complexity interaction for language complexity dimensions is aligned with the work by Rahimi and Zhang (2019), who suggested that writing anxiety in highly complex writing tasks could negatively predict the effect on the production of language syntactic complexity. Additionally, the current study found high writing anxiety in difficult writing tasks could significantly lead to lower language accuracy, and this finding seemed to be inconsistent with the results obtained by Rahimi and Zhang (2019). The inconsistent results are due to the lack of consideration about the possible effect of different writing anxiety levels combined with different writing complexity tasks on language performance in their study. The components of writing anxiety and writing task complexity were used by Rahimi and Zhang (2019), while our study considered the effect of different writing anxiety levels (high writing anxiety and low writing anxiety) on language performance, which can also be evidenced negative effect of writing anxiety on language achievement in the previous study (Zabihi et al., 2020). Regarding the writing anxiety and task complexity interaction for language fluency, we also found writing task complexity could predict language fluency when the students have the same writing anxiety level. For example, students with low levels of writing anxiety can produce more fluency of language in difficult writing tasks than those in simple writing tasks; however, students with high levels of writing anxiety can perform less fluency in difficult writing tasks than those in simple writing tasks. The findings that the predictive effect of difficult writing tasks and simple writing tasks on language fluency was presented to be significantly different in terms of different writing anxiety levels support the work by Johnson (2017). Based on the previous studies (see Ellis and Yuan, 2004; Ishikawa, 2007; Johnson, 2017), increasing the resource-directing dimensions or decreasing the resource depletion dimensions in writing tasks has been confirmed to lead to more language fluency production, which can be explained the different predictive effect of writing tasks on language fluency in the present study. For instance, students with highly difficult writing tasks, due to the multiple elements, sufficient time available or low anxiety level, achieve higher language fluency. Thus, this finding does not only further complement the effect of task complexity or writing anxiety on language achievement but also verifies the interaction effect of writing anxiety and task complexity on writing performance.

Regarding the writing boredom and task complexity interaction for EFL writing performance, we only found significant writing boredom levels and task complexity interaction for some dimensions of language complexity, such as MLT, C/T, and SC/T, presented in Figure 3, indicating that those students majoring in English in China with low writing boredom are likely to produce more language complexity dimension scales in the difficult writing task, except for TTR of lexical complexity. This finding is in line with the results obtained by Li et al. (2023a), which gained insights into the significant negative relationship between writing boredom and writing achievement and showed the predictive power of writing boredom on writing achievement. Compared with the study by Li et al. (2023a), we investigated the different writing boredom levels and task complexity interaction for language fluency, language complexity, and language accuracy, presenting that different writing boredom levels combined with different writing complexity led to a significant difference in language complexity, despite no significant effect on language fluency and language accuracy. This finding further extends the EFL writing literature by integrating task complexity.

As for writing enjoyment and task complexity interaction for language fluency, language complexity, and language accuracy, there are significantly different writing enjoyment levels and task complexity interaction for language fluency, language complexity, and language accuracy, respectively, with the data shown in Figure 4, suggesting that those students majoring in English in China with high writing enjoyment are likely to produce more language fluency, language complexity, and language accuracy, not least in difficult writing task. What needs to be noted is that there is a significant difference in writing boredom between high writing enjoyment and low writing enjoyment groups; that is, the positive effect of high writing enjoyment on language fluency, language complexity, and language accuracy is likely to be affected by low writing boredom. However, according to the work by Li (2022), foreign language enjoyment is presented to be significantly and negatively related to foreign language boredom. Thus, high writing enjoyment may accompany low writing boredom, and meanwhile, low writing enjoyment may go with high writing boredom. Regarding the positive effect of high foreign language enjoyment and low foreign language boredom on language learning, the findings in the current study can still explain the effect of writing enjoyment on language fluency, accuracy, and complexity.

As for the predictive effect of high writing enjoyment together with high-difficulty writing tasks on language complexity and accuracy, the finding can be illustrated by preceding studies (Robinson, 2001, 2011; Johnson, 2017), who unanimously shed light on the positive effect of task complexity increased by resource-directing dimensions on syntactic or lexical complexity and accuracy. However, the writing enjoyment and writing task complexity interaction for language fluency counters with the work by Robinson (2001, 2011), who claimed that increasing task complexity would direct the L2 learners to engross in the linguistic form such as syntactic or lexical complexity, language accuracy. Thus, writing task complexity increased by recourse directing dimensions seems to play no role in language fluency. However, the effect of the writing task and writing enjoyment interaction on language fluency can highlight the salience of writing enjoyment on language fluency in L2 writing. There are a few studies examining the relationship between foreign language enjoyment and language fluency (Bielak, 2021), reporting that high enjoyment could even contribute to processing the writing materials, and then producing the language fluency.

We have also found the students in the high writing enjoyment and difficult writing task subgroup present even more language fluency, language complexity, and language accuracy than any other subgroups in the current study, indicating greater predictive power of writing enjoyment on EFL writing achievement than anxiety and boredom, and this finding has been confirmed by Dewaele and Alfawzan (2018), Tahmouresi and Papi (2021) and Li et al. (2023b). Compared to the prior studies, the present study examined the interaction effect of task complexity and different writing enjoyment levels on language fluency, complexity and accuracy of EFL writing, so we have found how different writing enjoyment levels with different task complexity presented significant differences in language fluency, complexity, and accuracy. Additionally, objective linguistic measures such as the dimensions of language fluency, complexity, and accuracy are used in the present study, so that we can further gain insights into such possible mediated variables as language fluency, complexity, and accuracy between writing enjoyment interacted with different writing task and writing achievement.

Above all, the research findings demonstrated the predictive power of positive and negative emotions like writing enjoyment, anxiety, and boredom on writing achievement in specific writing task complexity. Not least, the findings can still be theoretically illustrated by the control-value theory of educational psychology (Pekrun, 2006). Based on the control-value theory of educational psychology, writing enjoyment can be assumed as a positive high-arousal emotion, accompanied by boosted engagement and more access to cognitive resources. When it is combined with difficult writing tasks, learners with high writing enjoyment can access cognitive resources from difficult writing tasks to contribute to linguistic development, such as language fluency, complexity, and accuracy. While writing, boredom and anxiety can result in disengagement and less access to cognitive resources (Li et al., 2023a). When they are combined with difficult writing tasks, the learners with high writing anxiety or writing boredom, due to less access to cognitive resources from difficult writing tasks, present less language fluency, complexity, and accuracy.