Understanding college students’ test anxiety in asynchronous online courses: the mediating role of emotional engagement

materials in AOCs can be studied multiple times or skipped without instructor supervision (Tseng et al., 2023), providing personalized choices for learners at different


Background
Asynchronous online courses (AOCs) have received increasing attention in higher education (Kim et al., 2018;Lu et al., 2023) and are courses that provide asynchronous online learning, a mode of learning that allows for time-independent communication and reflection (Hrastinski, 2008;Kim et al., 2018).AOCs have several advantages, including overcoming time and space limitations (Han et al., 2023;Lu et al., 2023).Moreover, learning materials in AOCs can be studied multiple times or skipped without instructor supervision (Tseng et al., 2023), providing personalized choices for learners at different learning paces.Thus, AOCs offer convenience and numerous benefits to contemporary college students through different learning platforms, such as massive open online courses and small private online courses (Chiu & Hew, 2018;Ruiz-Palmero et al., 2020).
Test anxiety is a common psychological condition that can have serious negative effects on college students' educational process and can present in the online learning context (Chapell et al., 2005;Hill & Wigfield, 1984;Alibak et al., 2019), leading to students' heavy cognitive load, low achievement and low continuous learning intention in the asynchronous online learning context (Hart, 2012;Jo et al., 2015;Kim et al., 2021;Moody, 2004).Previous studies have found that test anxiety negatively influences students' academic performance (Cassady & Johnson, 2002) and physical and psychological development (Fischer et al., 2016).Additionally, it can contribute to depression and lead to suicidal behaviors (Lee et al., 2006).Thus, exploring the factors and mechanisms that influence test anxiety among college students is necessary.However, few studies have systematically investigated the factors influencing test anxiety in AOCs (Alibak et al., 2019).Without understanding the potential influencing factors, clarifying students' rationale behind test anxiety in online courses is difficult, making it challenging to alleviate inefficiency, underachievement, and low performance in AOCs (Rana & Mahmood, 2010).Therefore, this study aimed to determine the factors influencing test anxiety in AOCs among college students.In the present study, we mainly examined the factors influencing the learning process, that is, the learning factors in asynchronous online learning (Lu et al., 2021b).
Due to their self-paced learning environments, AOCs pose higher requirements for students' learning autonomy (Kim et al., 2018).Learning autonomy is strongly related to self-efficacy (Csizér et al., 2021), as strong self-efficacy can promote increased learning confidence.Moreover, in asynchronous online courses where students learn independently, they may encounter feelings of loneliness or isolation (Reedy, 2019), while interaction is a proposed way to combat students' feeling of loneliness from their peers and instructors (Kaufmann & Vallade, 2022).In addition, interaction with peers and teachers could alleviate students' learning anxiety in an English language classroom (Fang & Tang, 2021).Thus, increased attention should be given to students' self-efficacy and interactions with instructors and classmates in AOCs.
In addition, the learning platform is an essential aspect of AOCs, and its ease of use and usefulness could affect students' ongoing learning intention and emotions while learning (Daneji et al., 2019).It has been validated by previous studies that the positive influence of expectation confirmation on perceived ease of use and perceived usefulness in various contexts, including asynchronous online courses and other technologyenhanced learning contexts (Al-Sharafi et al., 2023;Liesa-Orús et al., 2023;Lu et al., 2023).Still, while technology-enhanced contexts provide various conveniences for students to learn asynchronously, their complex operational nature may increase their pressure and anxiety to perform well on use of asynchronous online learning platforms (Hauser et al., 2012;Martin et al., 2008).If test anxiety is not well mediated, it will lead to adverse psychological reactions and even affect students' academic performance (Fischer et al., 2016).However, there is still limited research examining how the ease of technology use and its perceived usefulness influence test anxiety in AOCs.
As one of the important subdimension of learning engagement (Sun & Rueda, 2012), emotional engagement plays a critical role in learning online courses (Daniels et al., 2016;Hewson, 2018), as it correlates with the quality of students' learning motivation and learning performance in online learning (Özhan & Kocadere, 2020;Wang et al., 2022).Previous studies found that students who engaged more emotions in the learning process displayed a higher level of success and a lower prevalence of dropping out (Özhan & Kocadere, 2020).At present, it was found that anxiety levels among university students are extremely severe (Al Battashi et al., 2021;Husky et al., 2020), while student who is emotionally engaged is better able to focus on a task, and is better able to solve complex problems, which could significantly reduce anxiety (Schöbel et al., 2023).Thus, it is essential to pay attention to students' emotional engagement in AOCs.
Previous studies found that emotional engagement was influenced by students' learning factors, including self-efficacy, interaction, and perceived usefulness (Chen, 2017;Pellas, 2014).For example, Pellas (2014) found that computer self-efficacy was positively associated with students' emotional engagement factors in online courses.Moreover, it was found that adolescents' emotional engagement, such as empathetic joy was negatively related to their anxiety (Smith, 2015).From this point of view, emotional engagement could be influenced by learning factors, and at the same time, it acts on anxiety.Based on the above analyses, it is reasonable to infer that emotional engagement may mediate the association between learning factors and test anxiety.Previous relevant studies were mainly concentrated on synchronous online learning contexts (Li et al., 2024), rarely based on asynchronous online learning contexts.However, asynchronous online learning is different from synchronous online learning contexts, which needs stronger learning autonomy and self-regulation abilities (Kim et al., 2018;Lai et al., 2024), and the mediating role of emotional engagement between college students' learning factors and test anxiety in asynchronous online learning still needs to be investigated.Therefore, this study also aimed to investigate the mediating role of emotional engagement in the association between college students' learning factors and test anxiety in an asynchronous online learning context.

Research purpose
Test anxiety is a relatively common negative emotional response and has negative influences on learning, such as not conductive to flow experience and learning confidence (Watthanapas et al., 2021).However, a nuanced investigation of the reasons that explain test anxiety in AOCs is still needed.Without proper adjustment, potentially increasing students' cognitive loads and learning pressure, test anxiety could affect their learning (Mavilidi et al., 2014).
Therefore, the purpose of the present study was to propose and build an integrated model that examines (1) the relationship between college students' learning factors and their test anxiety in an asynchronous online learning context and (2) the mediating role of emotional engagement in the relationship between college students' learning factors and their test anxiety in an asynchronous online learning context.We also explored the reasons for the existence of these relationships.

Theoretical framework
In this study, social cognitive theory was used as the theoretical framework (Bandura, 1986), which posits that the regeneration of human function and behavior is dynamically influenced by interconnected and sometimes overlapping personal, behavioral, and environmental determinants (Bandura, 1986).Social cognitive theory is a comprehensive theoretical framework that has been widely used in many empirical studies (Cai & Shi, 2022;Lu et al., 2023).For example, Lu et al. (2023) proposed a research model based on social cognitive theory to investigate the mediating effect of deep approaches to using technologies on learning factors (e.g., intrinsic motivation, peer interaction, and connectedness) and higher-order thinking skills in a technology-enhanced open inquirybased learning context.
For the critical characteristics of asynchronous online courses, we proposed a personal belief model which consisted of five key independent variables: self-efficacy, instructorlearner interaction, learner-learner interaction, perceived ease of use and perceived usefulness.Since the progress of asynchronous online courses is mainly controlled by students themselves, their learning autonomy is important, and learning autonomy is affected by self-efficacy (Csizér et al., 2021;Kim et al., 2018), so self-efficacy was taken into account.Moreover, students who learn independently in asynchronous online courses are more likely to feel lonely (Reedy, 2019), while interaction is a suggested way to combat the lonely feeling from their peers and instructors (Kaufmann & Vallade, 2022), thus, interaction was considered.What is more, the learning platform is an essential aspect of AOCs, its ease of use and usefulness could affect students' ongoing learning intention and emotions (Daneji et al., 2019), thus, perceived ease of use and usefulness was incorporated into the investigation.
Among the social cognitive theory framework, the determining factors of the personal dimension are individuals' beliefs about their abilities and skills.The determining factors of the behavioral dimension are individuals' behavior while demonstrating their abilities and skills.The determining factors of the environment refer to the influence of the surrounding conditions on individuals' abilities, skills, and beliefs.Self-efficacy refers to "the perception and belief that an individual has of their skills and that they can mobilize effectively to succeed in a particular action" (Puozzo & Audrin, 2021, p. 1), so we identified it as the personal dimension.Interaction refers to the two-way reciprocal communication among learners and instructors (Moore & Kearsley, 1996), while communication is one of the learning behaviors, so we identified interaction in AOCs as the behavioral dimension.Perceived ease of use refers to "the degree to which a person believes that using a particular system is free of effort" (Davis, 1989, p. 320).Perceived usefulness is "the degree to which a person believes that using a particular system will enhance his or her job performance" (Davis, 1989, p. 320).Both of them two are students' perceptions of learning platform (the online learning environment), so we identified students' perceived ease of use and perceived usefulness in AOCs as the environmental dimension.
As seen in Fig. 1, we proposed a research framework in this study.It includes (1) independent variables (including self-efficacy, interaction, perceived ease of use and perceived usefulness), (2) mediating variable (emotional engagement), and (3) dependent variable (test anxiety).
Based on this research framework, several hypotheses emerged.The following sections discuss these hypotheses according to the relationship among these key variables.

The relationship between learning factors and test anxiety
Test anxiety is a multidimensional structure and has been defined as "a set of cognitive, physiological, and behavioral responses related to concerns about possible failure or poor performance on an exam or a similar evaluative situation" (Bodas et al., 2008, p. 387) that invoke "an unpleasant feeling or emotional state that has physiological and behavioral concomitants" (Dusek, 1980, p. 88).Test anxiety exists in numerous learning contexts, including asynchronous online learning (Alhazbi & Hasan, 2021;Sullivan, 2016).Research has indicated that test anxiety has a negative influence on students' academic performance and physical and psychological development (Cassady & Johnson, 2002;Fischer et al., 2016).Thus, identifying the mechanisms of action in test anxiety is necessary to help reduce college students' test anxiety in the asynchronous online learning context.

Self-efficacy and test anxiety
Self-efficacy refers to "the perception and belief that an individual has of their skills and that they can mobilize effectively to succeed in a particular action (in the sense of achieving a goal)" (Puozzo & Audrin, 2021, p. 2).This definition builds on Bandura's definition of self-efficacy, which has been widely adopted in other studies (Tasgin & Dilek, 2023).In educational contexts, self-efficacy is individuals' beliefs that they have the capacity to deal with school-related tasks (Lei et al., 2021;Putwain et al., 2013).Previous studies found that self-efficacy is significantly related to student learning performance (Bonaccio & Reeve, 2010;Hayat et al., 2021;Onyeizugbo, 2010).
When it comes to the relationship between self-efficacy and test anxiety, several studies demonstrated that self-efficacy was negatively related to student' test anxiety (Bonaccio & Reeve, 2010;Hayat et al., 2021).For example, Lei et al. (2021) found that academic self-efficacy was negatively correlated with students' test anxiety, which was supported by Adesola and Li's (2018) findings.Based on this body of research, we proposed the following hypothesis: Hypothesis 1 (H1): College students' self-efficacy would be negatively related to test anxiety in asynchronous online learning contexts.

Interaction and test anxiety
In an asynchronous online learning context, face-to-face communication between classmates and instructors is absent, making online interaction particularly important (Rahman et al., 2021).Instructor-learner and learner-learner interactions are vital in AOCs.Instructor-learner interaction refers to the two-way communication between the instructor and students in the asynchronous online learning process (Moore & Kearsley, 1996).Test anxiety of learners usually comes from the learning courses, the teachers or their classmates, so, positive and effective interaction with instructors and learners could significantly reduce test anxiety (Masomi, 2015).Learner-learner interaction is the two-way reciprocal communication among learners, with or without an instructor, in the asynchronous online learning process (Kuo et al., 2014).
Regarding the relationship between interaction and test anxiety, Peleg-Popko ( 2002) examined the relationship between family interaction, children's traits, and test anxiety and found significant negative correlations between family interaction, children's traits, and test anxiety.Cohen (1969) also investigated the effects of group interaction and progressive hierarchy presentation on test anxiety desensitization and reported that students in groups who were given an opportunity to interact reported a greater reduction in test anxiety than those without this opportunity.Cong-Lem and Hang (2018) investigated the relationship between high school students' willingness to communicate and their speaking test anxiety, and found that there was a negative relationship between them.Masomi (2015) found that collaborative learning could reduce test anxiety.In AOCs, students and instructors usually interact with each other through communication and collaboration.Given this, we proposed the following hypotheses: Hypothesis 2 (H2): College students' instructor-learner interactions would be negatively related to test anxiety in an asynchronous online learning context.
Hypothesis 3 (H3): College students' learner-learner interactions would be negatively related to test anxiety in an asynchronous online learning context.Davis (1989) defined perceived ease of use as "the degree to which a person believes that using a particular system is free of effort" (p.320).This definition comes from a paper that has been cited more than 80,000 times, with high authority.In AOCs, perceived ease of use is the degree to which students perceive using AOCs as effortless.Since AOCs are based on online platforms, they involve technical operations, including working with the internet and computers during examinations (Alibak et al., 2019); therefore, the ease of use of the platform merits further study.

Perceived ease of use and test anxiety
Regarding the relationship between perceived ease of use and test anxiety, Saadé and Kira (2007) found a significant negative relationship between anxiety and perceived ease of use.Alkis (2010) found that test anxiety was inversely related to students' behavioral intentions to use web-based assessments.In other words, if students perceived technology as easy to use, they had a high intention to use it (Mastuti & Seger Handoyo, 2019), and the more he/she use it, the more familiar he will become with it.At this time, students' test anxiety could be relieved by familiarity.Based on these studies, we proposed the following hypothesis: Hypothesis 4 (H4): College students' perceived ease of use would be negatively related to test anxiety in asynchronous online learning contexts.

Perceived usefulness and test anxiety
Perceived usefulness is "the degree to which a person believes that using a particular system will enhance his or her job performance" (Davis, 1989, p. 320).This definition comes from the same paper as the definition of perceived ease of use, which has been cited more than 80,000 times and is highly authoritative.While perceived usefulness in AOCs, refers to the degree to which students believe that AOCs can improve learning performance.Since AOCs represent only one form of online courses, their usefulness is a critical factor to explain learners' tendency to use this form of learning, therefore, paying attention to the usefulness of AOCs is also important.
Regarding the relationship between perceived usefulness and test anxiety, research has shown that computer anxiety is negatively related to the perceived usefulness of computers (Igbaria & Parasuraman, 1989;Igbaria et al., 1994).Liaw and Huang (2013) found that students' perceived anxiety was negatively related to the perceived usefulness of e-learning, but the relationship did not reach statistical significance.Computer anxiety is the fear of using computers and is an affective response (Chua et al., 1999).From an information-processing perspective, negative emotions related to high anxiety divert cognitive resources from task execution (Kanfer & Heggestad, 1999).Thus, we proposed the following hypothesis: Hypothesis 5 (H5): College students' perceived usefulness would be negatively related to test anxiety in asynchronous online learning contexts.

Mediating effects of emotional engagement between learning factors and test anxiety
Emotional engagement refers to learners' positive and negative emotions toward their instructors, peers, and learning activities (Christenson et al., 2012).Positive emotions include interest, enthusiasm, and enjoyment during learning (Renninger & Bachrach, 2015), whereas negative emotions include sadness, boredom, and frustration during learning (Skinner, 2016).Emotional engagement is vital for changing learners' behaviors (Bandura, 2012).In the present study, we mainly focus on students' positive emotions, including their high levels of interest and positive attitudes, which could enroll students' interest, identification, and positive attitudes or values about the learning process (Pellas, 2014).Previous studies have found that emotional engagement is correlated with test anxiety, especially among high school and secondary school students (Raufelder et al., 2015).However, these studies mainly looked at other self-cognition related learning factors, including computer self-efficacy, self-regulation and self-esteem in the computer cram schools learning environment and online learning programs (Chen, 2017;Pellas, 2014).
Regarding the relationship between self-efficacy and emotional engagement, Pellas (2014) found that computer self-efficacy was positively associated with students' emotional and cognitive engagement factors in online courses.In addition, Chen (2017) reported that learning engagement fully mediated the relationship between computer self-efficacy and learning performance and that computer self-efficacy was positively correlated with learning engagement among students who were in their late middle and older years.
Regarding the relationship between teacher-student, student-student interactions and emotional engagement, Miao et al. (2022) reported that teacher-student and student-student interactions could directly influence learning engagement in online environments.Wang et al. (2022) reported that learner-learner interactions could directly and positively predict learning engagement during online learning, whereas instructor-learner interactions could affect learning engagement through learnerlearner interactions.Qureshi et al. (2023) also found that peer interaction with peers and student-instructor interaction positively influenced students' engagement through the mediating role of active collaborative learning in the online collaborative learning context.
Regarding the relationship between perceived ease of use and emotional engagement, Kim et al. (2021) found that perceived ease of use could indirectly influence learning engagement through perceived usefulness as a mediator in massive open online courses.Iswanto et al. (2021) also reported similar findings for employee performance.Moreover, in the asynchronous online learning contexts, when students find an online learning platform easier to use, the higher interest they would use it, which could enhance their emotional engagement.
Regarding the relationship between perceived usefulness and emotional engagement, Kim et al. (2021) reported that perceived usefulness has a direct effect on student engagement in massive open online courses.Similarly, Jung and Lee (2018) found that college students' perceived usefulness had a significant direct effect on their learning engagement in massive open online courses.Gunness et al. (2023) also found this relationship in a semi-synchronous online learning context.In the asynchronous online learning contexts, when students feel AOCs more useful, they will have more interest to learn it, which could improve their emotional engagement.Based on these studies, we proposed the following hypotheses: Hypothesis 6 (H6): College students' self-efficacy would be positively related to emotional engagement in asynchronous online learning contexts.
Hypothesis 7 (H7): College students' instructor-learner interaction would be positively related to emotional engagement in an asynchronous online learning context.
Hypothesis 8 (H8): College students' learner-learner interactions would be positively related to emotional engagement in an asynchronous online learning context.
Hypothesis 9 (H9): College students' perceived ease of use would be positively related to their emotional engagement in an asynchronous online learning context.
Hypothesis 10 (H10): College students' perceived usefulness would be positively related to their emotional engagement in an asynchronous online learning context.Regarding the relationship between emotional engagement and test anxiety, it was found that adolescents' emotional engagement, such as empathetic joy was negatively related to their anxiety (Smith, 2015).In addition, students' agentic engagement was negatively related to test anxiety, with basic psychological needs acting as a mediator (Mehdipour Maralani et al., 2018).Based on this research, we proposed the following hypothesis: Hypothesis 11 (H11): College students' emotional engagement would be negatively related to test anxiety in an asynchronous online learning context.
Given the existing body of research on the relationship between our variables of interest (i.e., learning actors, emotional engagement, test anxiety), a research model was proposed for the present study to explain their relationship in an asynchronous online learning context (Fig. 2).

Research design
This study adopted a quantitative method research design to investigate college students' self-efficacy, instructor-learner interaction, learner-learner interaction, perceived ease of use, perceived usefulness, emotional engagement, and test anxiety and examine the relationship between these key variables in the asynchronous online learning context.Quantitative research methods are systematic investigative processes primarily involving the collection and analysis of numerical data to understand phenomena, establish patterns, test hypotheses, and make predictions.Survey research was used in the present study (Creswell & Creswell, 2017), this method has been used extensively across various disciplines including social sciences, natural sciences, economics, and health sciences (Lu et al., 2021b;Prentice et al., 2020).
We collected data through questionnaires and used statistical methods to analyze the relationships between these research variables.The questionnaire was issued during the middle of the spring semester.Both universities involved in this study offered online courses every semester, and all students were enrolled in the fall semester.Therefore, participants in this study had at least one semester of learning experience using the online learning platform.
Research design of this study is illustrated in Fig. 3, this figure outlines an eight-step process for conducting this study.Each phase ensures that the research is methodical, reliable, and valid, leading to insightful outcomes.First is identifying the research problems, this step involves recognizing and defining the research problems that need to be addressed.Second is constructing the research model, this involves creating a conceptual framework that outlines the key variables and their hypothesized relationships.Third is designing questionnaires, these questionnaires are tools for data collection that will help in measuring the variables specified in the research model.Fourth is collecting data, in this step, the designed questionnaires are distributed to collect data Fig. 2 The hypothesized research model from the relevant population.Once data is collected, the measurement model needs to be confirmed.This involves assessing the goodness-of-fit, construct reliability, validity, and discriminant validity to ensure that the measurement model accurately captures the constructs of interest.And then is analyzing the direct and indirect influence, this analysis helps in understanding the relationships and interactions between different factors.Next is confirming the research model, this involves validating the hypothesized relationships and making any necessary adjustments to the model based on the findings.The last step is discussion and conclusion, this involves interpreting the findings, discussing their implications, and suggesting potential areas for future research.This systematic approach ensures that the research is thorough, valid, and reliable, leading to meaningful and actionable conclusions.

Participants
A total of 316 higher-education students from eastern China, ranging from first-year undergraduate to second-year master's degrees, who had asynchronous online learning experiences for at least one semester from a range of disciplines, participated in this study.Participants primarily came from two universities, one is a normal university, and the other is a science and engineering university.The gender ratio at each school was approximately 3:7 and 7:3 respectively, with more boys at the science and engineering university and more girls at the normal university.The term normal university refers to a teacher-training institution of higher education.Among the 316 participants, 176 came from the normal university, 140 came from the other, 133 (42.1%) majored in liberal arts, 183 (57.9%) majored in science, 133 were men (42.1%), and 183 were women (57.9%).It is crucial to emphasize that the study employed random sampling, ensuring a degree of representativeness in the selected cohort.
During the asynchronous online learning process on the pre-agreed learning platform, the instructor first introduced an overview of the course, its objectives, its outline and its evaluation metrics (including the proportion of their regular grades and final exam grades) to all students.Second, the instructor explained how to use the learning resources based on an online learning platform.Subsequently, the course assessment requirements were introduced.After this introduction, students could learn about the materials at their own pace.Learning materials came in different forms, including video lectures, readings, assignments, quizzes and exams and so on.Students were required to Fig. 3 The research design submit homework and complete assignments and tests before the deadline.Generally, updates and important information posted by instructors on the course platform.When students encountered problems in the learning process, they could post them on the discussion board of the platform or send private messages and emails to their classmates or instructors.Their classmates and instructors responded when they logged onto the platform.Finally, when students finished all the learning assignments, the platform graded them according to their academic performance, including the completion of learning assignments, homework, and interactions with classmates and instructors.As seen in Fig. 4, this interface showcases a typical asynchronous online course platform, featuring a menu bar positioned on the left-hand side and a content bar on the right.Students can navigate through designated sections to study and complete their learning assignments as per their requirements.

Instruments
The questionnaire comprised two parts.The first part collected the participants' demographic data.The second part measured the students' perceptions of self-efficacy, instructor-learner interaction, learner-learner interaction, perceived ease of use, perceived usefulness, emotional engagement and test anxiety.

Self-efficacy
Student self-efficacy was measured using three items adapted from Cheng and Tsai (2011).Its overall reliability alpha in previous study was 0.84.Each item is rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).An example item assessing self-efficacy is "I believe that I can get excellent grades in asynchronous online learning courses."

Instructor-learner interaction and learner-learner interaction
Students' instructor-learner and learner-learner interactions were measured using three items adapted from Rahman et al. (2021).Their Cronbach's alpha in previous study were respectively 0.83 and 0.89.Each item is rated on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree).A representative item to measure instructor-learner interaction is "I have numerous interactions with the instructor during the asynchronous online learning courses." An example of the items measuring learner-learner interaction is "I receive lots of feedback from my classmates in asynchronous online learning courses."

Perceived ease of use and perceived usefulness
Students' perceived ease of use and perceived usefulness were measured using three items adapted from MacLeod et al. (2018) because this scale was closer to the characteristics of AOCs in the present study.In the preceding study, their Cronbach's alpha values stood at 0.87 and 0.86, respectively.Each item was rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).A sample item assessing perceived ease of use is "In asynchronous online learning courses, I can use technology, which is easy to navigate." An item used to measure perceived usefulness is "Asynchronous online learning courses can benefit my learning experience."

Emotional engagement
Students' emotional engagement was measured using five items adapted from Sun and Rueda (2012) that were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).Its Cronbach's alpha in previous study was 0.88.A representative item measuring emotional engagement is, "I like taking the asynchronous online learning courses."

Test anxiety
Students' test anxiety was measured using three items adapted from Pintrich et al. (1993) that were rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).Its Cronbach's alpha in previous study was 0.80.A representative item measuring test anxiety is, "I have an uneasy, upset feeling when I take an exam in asynchronous online learning courses."

Data collection and analysis
An online survey was administered during the spring semester at the two participating universities to collect data.The universities granted permission to conduct this research before conducting the surveys.Participants were told that their survey results would not influence their academic grades and that their information would only be used for educational research.Participation was both voluntary and anonymous.The questionnaire required approximately 8-10 min to complete.All responses were then imported into SPSS 21.0 and AMOS 21.0 for data analysis.
Structural equation modeling was performed to analyze the relationships between independent and dependent variables in the present study.First, the overall model fit was evaluated to confirm the measurement model.Second, a structural model was constructed to test hypotheses regarding college students' learning factors, emotional engagement, and anxiety.Path coefficients, standardized regression weights (β value), and t-test values were used to confirm the relationship between college students' learning factors and their test anxiety.

Results
To investigate the differences between independent variables that investigated in the present study, we used ANOVA to examine the differences in distributions between selfefficacy, instructor-learner interaction, learner-learner interaction, perceived ease of use and perceived usefulness in the individual level, and it was found there was a significant difference between these independent variables (F = 89.11,p < .001).Next, to further look into the relationships between independent variables, mediating variable and dependent variables, structural equation modeling was used to assess the study's proposed research model.The model's overall fit was evaluated to confirm the measurement model (Ullman & Bentler, 2012).After the overall model fit was found to be within the acceptable range, a follow-up analysis of each variable in the research model was conducted.

Confirming the measurement model
The measurement model was assessed for goodness-of-fit, construct reliability, and construct validity.Goodness-of-fit was determined based on the GFI, CFI, TLI, RMSEA, and SRMR.These coefficients were chosen based on their relevance to the specific type of model and the nature of the data.The goal is to provide a comprehensive assessment of how well the model fits the data, taking into account different aspects of model performance such as variance explained, prediction accuracy, and complexity.All model fit indices were within the acceptable ranges (χ 2 /df = 2.14 < 3, GFI = 0.905 > 0.90, CFI = 0.951 > 0.90, TLI = 0.939 > 0.90, RMSEA = 0.060 < 0.08, SRMR = 0.039 < 0.08) (Hair et al., 2010), indicating that the measurement model exhibited a satisfactory fit (See Table 1).
The construct reliability of the model was determined through composite reliability (CR) and Cronbach's alpha (Table 2).CR is a measure of the internal consistency and reliability of a set of indicators for a latent construct.The CR values in this study were all over 0.70, indicating satisfactory reliability (Nunnally & Bernstein, 1994).Cronbach's alpha values were all over 0.70 and within acceptable limits (Fornell & Larcker, 1981).The convergent validity was determined using the extracted average variance (AVE).AVE reflects the average amount of variance that a construct captures from its indicators relative to the variance due to measurement error.The AVE values here were all greater than 0.5, which were in the satisfactory range (Segars, 1997).
The means, standard deviations, skewness, kurtosis, and correlation analyses are presented in Table 3.The square roots of the AVE were compared to the correlations between latent variables to evaluate discriminant validity (Fornell & Larcker, 1981), all of  which were less than the corresponding AVE square roots, which was satisfactory.Overall, these results indicated that the proposed research model had a good fit.

The structural model and hypothesis test
The structural model was constructed to test the hypotheses.As can be seen in Fig. 5, the structural model was adjusted based on the research model, presented with the path coefficients marked by standardized regression weights (β value) and t-values to Note M = mean, SD = standard deviation, PU = perceived usefulness, EU = perceived ease of use, LLI = learner-learner interaction, ILI = instructor-learner interaction, SE = self-efficacy; EE = emotional engagement, TA = test anxiety   show the relationship between college students' learning factors and their test anxiety in the asynchronous online learning context.In Fig. 5, the solid line between every two variables represents a supported hypothesis, whereas the dashed line an unsupported hypothesis.

Analysis of indirect effects between key factors
The direct and indirect effects of each hypothesis were analyzed in the present study to investigate the mediation effects.Based on the above analyses, we performed percentile bootstrapping and bias-corrected bootstrapping at a 95% confidence interval with 2000 bootstrap sample (Arnold et al., 2015) to test full or partial mediation.We calculated the confidence intervals of the lower and upper bounds to test the significance of the indirect effects, as recommended by Preacher and Hayes (2008).The indirect effects of self-efficacy on test anxiety (β = − 0.05, p < .05,Z = -0.18),instructor-learner interaction on test anxiety (β = − 0.06, p < .05,Z = -1.87),learner-learner interaction on test anxiety (β = − 0.06, p < .05,Z = -2.07)and perceived usefulness on test anxiety (β = − 0.08, p < .01,Z = -2.38)were all significant (Table 5).While the indirect effect of perceived ease of use on test anxiety (β = 0.01, p = .69,Z = -0.35)was not significant.

Discussion and conclusion
This study aimed to provide in-depth insights into the relationship between students' learning factors, test anxiety, and emotional engagement in asynchronous online learning.Specifically, we investigated the relationship between college students' learning factors and their test anxiety and the mediating role of emotional engagement in the relationship between them in an asynchronous online learning context.

The relationship between learning factors and test anxiety
First, the relationship between college students' learning factors and test anxiety in an asynchronous online learning context was explored, and we found that students' selfefficacy, instructor-learner interaction, and perceived ease of use were significantly and inversely related to test anxiety, suggesting that students may experience less test anxiety when they experience greater self-efficacy, interaction, or perceived ease of use simultaneously.These findings were consistent with previous studies (Čeko & Reić, 2020;Hayat et al., 2021), including research in the online and high school learning contexts.These findings also indicate that instructors should help college students improve their selfefficacy, instructor-learner interactions, and perceived ease of use in an asynchronous online learning context, which could reduce their test anxiety.While students' learnerlearner interaction and perceived usefulness did not significantly influence students' test anxiety, these findings were not consistent with previous research (Igbaria et al., 1994;Lin & Hou, 2022;Peleg-Popko, 2002).This can be explained by the fact that in asynchronous online learning contexts, learners typically engage with their classmates via online platforms that do not offer real-time communication.When students seek to communicate with peers due to test difficulty, they may not receive timely responses, leading to test anxiety to some extent.Furthermore, perceived usefulness refers to the degree to which students believe that asynchronous online courses could improve their learning performance.During the test period, learners tended to evaluate whether the course is beneficial, thus, its usefulness may not significantly impact test anxiety at this stage.Instead, perceived ease of use may be crucial, if the asynchronous online platform is difficult to navigate, it can increase the difficulty of online tests, thereby significantly affecting test anxiety.

The mediating role of emotional engagement between learning factors and test anxiety
Second, we examined the mediating role of emotional engagement between college students' learning factors and test anxiety in an asynchronous online learning context.Students' self-efficacy, instructor-learner interaction, learner-learner interaction, and perceived usefulness were significantly and positively associated with emotional engagement, consistent with previous findings (Gunness et al., 2023;Miao et al., 2022;Pellas, 2014).Perceived ease of use was the only learning factor that not significantly related to emotional engagement, which was not consistent with previous findings (Iswanto et al., 2021).Moreover, emotional engagement was found to be significantly negatively related to test anxiety, which is consistent with previous research (Mehdipour Maralani et al., 2018).Therefore, emotional engagement emerged as a significant mediator between selfefficacy, instructor-learner interaction, learner-learner interaction, perceived usefulness, and test anxiety.However, it did not serve as a significant mediator between perceived ease of use and test anxiety.
The lack of significant relationship between students' perceived ease of use and emotional engagement may be attributed to the role of ease of use in determining the initial ease of starting to learn in asynchronous online courses.Once students have begun their learning journey, the perceived usefulness becomes the primary factor influencing their emotional engagement with the course.Consequently, students' perceived ease of use may not directly impact their emotional engagement.This finding indicates that instructors should be aware of students' self-efficacy, instructor-learner interaction, learnerlearner interaction, and perceived usefulness to improve their emotional engagement in asynchronous online courses.

Theoretical and practical value
This study highlights the importance of paying close attention to college students' test anxiety while learning in asynchronous online courses and has theoretical and practical implications.From a theoretical perspective, this study highlighted that social cognitive theory could be a comprehensive theoretical framework for understanding the relationship between learning factors and test anxiety in online asynchronous courses.The findings provide a more integrative perspective for reducing college students' test anxiety in asynchronous online learning courses than previously reported.
From a practical perspective, this study has several implications for reducing college students' test anxiety in asynchronous online learning courses.First, instructors should design moderately challenging tasks in asynchronous online learning courses to improve college students' self-efficacy.Instructors can help learners succeed by continually providing them with moderately challenging materials and tasks and increasing the task difficulty to reflect their progress.Consistent success with these materials and tasks creates a record of active mastery or performance that instructors can use to inform struggling learners to increase their opportunities for success.This highly motivating realization could further ensure academic engagement in similar tasks.Overall, success is crucial for enhancing confidence and the willingness to keep trying (Margolis & McCabe, 2006).Second, some interactive learning tasks should be arranged to improve interactions between instructors and students, such as collaborative inquiry-based learning activities and mutual evaluation activities.Third, improving the perceived usefulness and perceived ease of use of asynchronous online learning courses depends on both the course designers and platform developers, who must conduct many investigations about students' learning preferences before redesigning platforms so that the redesigned learning content and platforms can better meet students' learning needs and habits.Finally, instructors need to communicate more with students to improve students' emotional engagement since communication conveys support and security, helping students cope with stress, such as test anxiety (Peleg-Popko, 2002).Students should be encouraged to provide positive feedback to their peers during the learning process so that they can be more actively and enthusiastically to participate in learning activities.

Limitations and future research directions
Although this study provides new insights into students' test anxiety in AOCs, it had some limitations that should be addressed in further research.First, this study specifically focused on students' learning factors, some psychological factors, such as tension and excitement in learning asynchronous online courses, were included as variables in future studies.Second, the data used in this study were self-reported, which are subjective.Other objective measures, such as learning behavior observations and analyses, should be considered in future studies to support our current self-reported findings.Third, the number of participants in the present study was limited, and additional participants from other cultures or countries could be included to ensure diversity and increase the generalizability of the findings.

Fig. 5
Fig. 5 The Unconstrained Model Outputs for College Students Learning Asynchronous Online Courses

Table 1
Results of goodness-of-fit

Table 2
Results of construct reliability and validity

Table 3
Results of descriptive statistics and discriminant validity

Table 4
Results of the hypotheses