A central tenet of humanistic learning theory, with psychologist Carl Rogers as the representative, is the importance of people, self-concept, and emotional factors in the learning process; thus, helping students acquire the ability to learn is more central than teachers’ knowledge delivery (Rogers, 1983). Constructivist theory holds that learning is student-centered, and it should emphasize learners’ active exploration of knowledge, their spontaneous discovery, and their active construction of the meaning of the knowledge they have acquired (Piaget, 1980). In the triarchic theory of intelligence, the crucial role of individual initiative in the construction of cognitive structure is similarly emphasized (Surhone et al., 2010). Therefore, student-centered as well as autonomous and spontaneous learning behavior is what education strives toward. In the education field, spontaneous learning is also called self-directed learning (Knowles, 1975; Lemmetty and Collin, 2020). Holec (1979) defined self-directed learning as an individual’s ability to be responsible for their own learning and specifically explained it as the ability of a learner to define their particular learning objectives, content, and progress as well as to select learning methods and techniques, track the learning process, and assess the learning results. Self-directed learning involves a learner’s desire and ability to independently make choices regarding their own learning (Littlewood, 1999). Self-determination theory suggests that individuals have a natural inclination to act on their own willpower. Therefore, people are more motivated by what they wish to do than by what they are required to do (Deci and Ryan, 1985). According to this theory, after learners have independently established their learning goals, the adoption of self-monitoring-based active learning behaviors can lead to greater learning results. Autonomous and independent learning are critical methods for college students given their adult status (Tough, 1989; Loeng, 2020). Khiat (2015) revealed that adult Students’ level of perceptive skills regarding indices of self-directed learning directly or indirectly affects their academic performance. A model constructed by Kim (2017) similarly indicated that, in path analysis, self-directed learning can directly affect academic achievement; thus, self-directed learning plays a positive role in college Students’ academic achievement (Cazan and Schiopca, 2014; Oducado, 2021). Therefore, on the basis of the theoretical perspectives and literature support provided in the preceding section, Hypothesis 1 was formulated as follows:

Scheier defined optimism as a positive expectation of future results (Scheier and Carver, 1985). When viewed collectively, self-determination theory and psychological capital theory suggest that self-determination behavior can generate positive and optimistic emotions; as a type of psychological capital, optimism is described as a trait that motivates an individual to achieve more favorable outcomes and performance (Deci and Ryan, 1985; Luthans et al., 2015). During self-directed learning, learners generate positive emotions and expectations of learning outcomes. Those who are skilled at self-learning more positively evaluate their learning skills, and thereby, through self-determination, set more optimistic learning goals (Deci and Ryan, 1985). Moreover, optimistic people are adept at adopting direct actions, plans, and coping strategies in learning, thus improving their academic outcomes (Luthans et al., 2015). Studies have similarly confirmed that students can more effectively observe their own performance and assess their own targeted progress when they participate in learning activities with a positive goal and a desire to achieve goals (Schunk, 1990). More positive emotions, accompanied by positive psychological implications, are induced in students with strong self-directed learning skills as their satisfaction with their targeted progress increases. In sum, students with greater self-directed learning skills are more likely to generate optimism as psychological capital. Furthermore, according to the theory of psychological capital, optimism pertains to positive emotions and can motivate an individual to achieve higher performance (Luthans and Youssef, 2007; Luthans et al., 2015). Research has revealed that, compared with individuals with high optimism, those with low optimism are less capable of adopting direct actions, planning, and coping strategies to manage and relieve the stress caused by life events when perceiving stress, thus lowering the efficiency of their work and learning (Brissette et al., 2002). Optimism can predict favorable academic results, and those who are optimistic possess sufficient social adaptability (Alkhedr, 1999; Dougall et al., 2001). El-Anzi and Owayed (2005) also highlighted that academic performance can be improved through optimism. Therefore, on the basis of the abovementioned theoretical perspectives and research findings, Hypothesis 2 was formulated as follows:

Mental health is defined as a state of well-being; in a state of mental health, people feel relaxed and can adapt to circumstances and thoroughly reach their psychophysical potential (Zhang et al., 2013). Social ecological systems theory (Bronfenbrenner, 1992) suggests that difficulties related to people’s intimacy and communication in school and family life as well as to people’s emotional response and behavioral control may cause mental health problems such as self-isolation and alienation. A healthy mentality provides learners with a favorable mental environment, and this positive mental state determines whether individuals can realize their academic potential (Maslow, 1954). In self-determination theory, environmental factors are also the basis of an individual’s self-directed learning behavior (Deci and Ryan, 1985). However, spontaneous learners can adopt active learning behaviors involving self-monitoring to adjust their psychological state and improve their academic outcomes more effectively; those with stronger self-directed learning skills exhibit a more positive and active learning status (Deci and Ryan, 1985; Luthans and Youssef, 2007). Such positive learning behavior is more consistent with the expectations of society, school, and parents, and the learner is more likely to obtain affirmation from teachers and parents, thus sustaining their relaxed psychological state (Tuffin et al., 2001) and achieving learning outcomes. Therefore, students skilled in self-directed learning exhibit behavior that is consistent with social expectations and receive social support to maintain their mental health (Tuffin et al., 2001; Hefner and Eisenberg, 2010). Favorable mental health provides learners with a positive psychological environment as support, which affects whether they can realize their academic potential (Maslow, 1954). According to studies on learning and well-being, a self-monitoring learning style influences Students’ perceptions of successes and improves their mental state (Field, 2009; Teal et al., 2015). Byrd and Mckinney (2012) reported that mentally healthy college students have greater advantages in academic performance in terms of interpersonal adaptation, communication, and acquiring cognitive ability relative to those with lower levels of mental health. Related studies have also revealed that mental health significantly and positively predicts Students’ academic performance (Scheier and Carver, 1985; Alkhedr, 1999; Albeg and Castro-Olivo, 2014; Cazan and Schiopca, 2014). Combining the aforementioned theories and research results, Hypothesis 3 was formulated as follows:

According to self-determination theory, a direct positive relationship exists between self-determination behavior and performance (Deci and Ryan, 1985). However, the relation of self-determination learning with learning outcomes may be jointly mediated by optimism and mental health. According to psychological capital theory, optimism is a critical positive resource in personality (Luthans et al., 2015). This positive personality tendency represents an effective internal ability for self-reparation and enhancement, acting as a buffer in the process of the psychological stress response and helping people resist mental illness (Chang, 2002; Gustavsson-Lilius et al., 2007). Zhang et al. (2011) revealed that optimism and psychological capital can help people withstand the impact on mental health resulting from the pressure of learning. Similarly, research in other fields has revealed that the mental health of optimistic patients with cancer seems to be superior to that of unoptimistic ones (Schou et al., 2004). Optimism negatively predicts psychological problems including suicidal tendencies and negative emotions (Yu and Zheng, 2011). Optimism is regarded as a predictor of mental health, supporting the notion that optimism and mental health may act as chain-mediators but not in mutually independent mediation roles. When viewed collectively, self-determination theory and psychological capital theory suggest that people with high-level self-learning skills more positively assess their learning skills, thereby inducing positive and optimistic emotional experiences. As positive psychological capital, optimism can help people resist mental health problems caused by stress (Deci and Ryan, 1985; Chang, 2002; Gustavsson-Lilius et al., 2007; Zhang et al., 2011; Luthans et al., 2015). Furthermore, needs theory suggests that a healthy mental state in learners determines whether they can successfully realize their academic potential (Maslow, 1954; Elliot and Dweck, 1989). Studies have also confirmed that people with strong self-learning skills exhibit more positive self-assessment, leading to optimistic emotional perception in such learners (Alkhedr, 1999; El-Anzi and Owayed, 2005), and those with optimistic positive expectations exhibit superior mental health (Chang, 2002; Schou et al., 2004; Gustavsson-Lilius et al., 2007; Zhang et al., 2011; Luthans et al., 2015). Favorable mental health also predicts higher academic performance (Kantomaa et al., 2010; Albeg and Castro-Olivo, 2014; Bostani et al., 2014; Agnafors et al., 2020). Therefore, this study posited that, for Wuhan college students, although optimistic learning attitudes and experiences result from spontaneous learning, such optimistic positive emotions may in turn affect mental health, thereby providing support for academic performance through a healthy psychological state. Thus, Hypothesis 4 was formulated as follows:

This study’s hypothetical model (Figure 1) was developed with reference to relevant theories and studies. In the model, self-directed learning is assumed to positively predict the academic performance of college students in Wuhan, and optimism and mental health are the two chain-mediating factors in this relationship.

Four constructs were investigated through independent scales; each item was scored using a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). The constructs and scales are described in the following section.

Self-directed learning is a learning approach that requires the learner to establish learning goals and a learning plan in addition to employing active self-management and learning self-evaluation (Holec, 1979). The self-directed learning scale contains five items related to self-directed learning ability (Hung et al., 2010). An example item is as follows: “I carry out my own study plan.”

Optimism plays a positive role in people’s adaptation to stress and coping under challenging conditions. Optimism refers to an expectation of a favorable future outcome (Scheier and Carver, 1985). Eight items are included in the optimism scale (Scheier and Carver, 1985). An example item is as follows: “In uncertain times, I usually expect the best.” After confirmatory factor analysis (CFA), three items for which the factor loading was considerably below 0.50 were excluded, and the remaining five items were used in the modeling (Kline, 2010).

Mental health refers to the positive state of an individual’s well-being. Those who are in this state have favorable adaptability and can employ their full mental potential (Zhang et al., 2013). The mental health scale contains six items (Zhang et al., 2013). An example item is as follows: “I enjoy my life.” In this study, after CFA, one item with factor loading < 0.50 was excluded, and the remaining five items were used in the modeling (Kline, 2010).

Academic performance in the present study refers to the comprehensive ability of Chinese college students in terms of learning cognitive, communicative, and interpersonal skills required for their future endeavors (Li et al., 2016). The academic performance scale contains 14 questions in three parts: learning cognitive, communicative, and interpersonal abilities (Yang and Swekwi, 2021). An example item is as follows: “I can use the knowledge that I learned flexibly.”

Because pandemic-related regulations in China would have severely hampered in-person surveying, an online questionnaire was employed for purposive sampling. Data were collected from 473 college students (age: 18–25 years) from three universities in Wuhan between April 1 and 15, 2020. All three universities, with the same educational goal, attempted to equip students for self-directed learning and implemented extensive and long-term training related to self-directed learning. To ensure students from all majors would be recruited, university counselors assisted the investigators in recruiting participants through the use of WeChat groups that included dormitory residents. After acknowledging their understanding of the study objectives, all respondents provided signed online informed consent. Respondents were informed that their data would remain confidential.

In total, 500 completed questionnaires were received, with 473 (94.6%) remaining after invalid questionnaires were excluded. Among the valid responses, 219 (46.3%) were from men and 254 (53.7%) were from women; 238 (50.3%) respondents were only children and 235 (49.7%) had siblings; 162 participants (34.2%) were freshmen, 79 (16.7%) were sophomores, 152 (32.1%) were juniors, and 80 (16.9%) were seniors (Table 1).

AMOS and SPSS were employed for statistical analyses, and the hypothetical model was assessed through structural equation modeling (SEM). The measurement and structural models were also verified (Bollen, 1989; Schumacker and Lomax, 2004; Byrne, 2010; Kline, 2010). CFA was employed to test a reasonable measurement model. The association of the Students’ data with the measurement mode was verified by using maximum-likelihood-estimated model parameters and fit indices obtained as statistical indicators. The statistical indicators were factor loadings per factor and measurement errors. Nine indicators of fit were calculated: normed chi-square (χ²/df), root mean square residual (RMR), root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), normed fit index (NFI), comparative fit index (CFI), Tucker–Lewis index (TLI), parsimonious normed fit index (PNFI), and HOELTER0.05 (Bollen, 1989; Schumacker and Lomax, 2004). Furthermore, to confirm the reliability and validity of the measurement model, the Cronbach’s α reliability coefficient, average variance extracted (AVE), and composite reliability (CR) were used. Harman’s one-factor test was employed for common method variance (CMV) verification of the study variables (Podsakoff et al., 2003). A bootstrap method was used to verify the mediator roles.

To evaluate the consistency of variables, Cronbach’s α was used. The relevant values were 0.79, 0.84, 0.82, and 0.91 in the self-directed learning scale, optimism scale, mental health scale, and academic performance scale, respectively, indicating high consistency in the measurement results. For model calibration, this study deleted four items with factor loadings much lower than 0.50. Thus, 29 valid items remained (Table 2). As a final step, the item–objective congruences of the four scales were assessed by five experts; all items in these scales were determined to have high content validity. The model fit indices are presented in Table 3, and the AVE and CR values pertaining to the measurement model and variable correlation matrix are listed in Table 4. The 95% confidence intervals of the correlation coefficients for 2,000 bootstrap replications were calculated for discriminant validity (Torkzadeh et al., 2003).

The normality assessment revealed a Mardia coefficient of 286.14, which was lower than N × (N + 2) = 899, with N being the number of questionnaire items; the absolute skewness and kurtosis values for the 29 items ranged from 0.04 to 0.61 and from 0.01 to 0.72, respectively. The results met the standard for the absolute value of skewness and kurtosis (both < 2; Curran et al., 1996), indicating normally distributed data.

CFA of the measurement model revealed an absence of negative error variances; moreover, all variances were significant. The factor loadings were all > 0.50 and < 0.95 (Figure 2), with no large standard errors (Table 2; Fornell and Larcker, 1981; Bagozzi and Yi, 2012). Thus, the measurement model was reasonable.

The model fit the data from the Wuhan students reasonably well, with the following fit index values: χ²/df=2.48, RMR = 0.04, RMSEA = 0.06, NFI = 0.90, CFI = 0.92, GFI = 0.88, TLI = 0.91, PNFI = 0.78, and HOELTER0.05 = 215 (Table 3; Bollen, 1989; Schumacker and Lomax, 2004).

The measurement model’s convergent validity and discriminant validity were subsequently assessed; its CR was > 0.60 and AVE was > 0.40 (Slater et al., 2007). Moreover, the CR and AVE of all variables were 0.80–0.87 and 0.44–0.63, respectively (Table 4). These results confirmed that the measurement model was acceptable and that it had convergent validity. The 95% confidence intervals of the correlation coefficients with 2,000 bootstrap replications were calculated; none of the confidence intervals in the lower and upper parameters included 1 (Table 4). Thus, the variables had discriminant validity (Torkzadeh et al., 2003).

Finally, because of epidemic prevention and control requirements, this study employed an online questionnaire to limit CMV (Peng et al., 2006). For CMV verification of the study variables, Harman’s one-factor test was employed. Exploratory factor analysis was conducted for all 29 items included in the scales of this study, followed by testing of the unrotated factor analysis results. According to the results, the Kaiser–Meyer–Olkin statistic was 0.95, greater than the threshold of 0.80. The Bartlett test of sphericity also yielded a significant result (p < 0.001). The explanatory power of the first factor was 37% (threshold value: 50%), indicating that this study had no severe CMV problem (Podsakoff et al., 2003).

SEM of the second-order model was used to obtain a structural model for examining the chain-mediating effect of two mediating variables, optimism and mental health, on the relation of self-directed learning with academic performance (Figure 3). The 29 items in the structural model had standardized regression coefficients of 0.55–0.84. The structural model exhibited a reasonable fit: χ²/df = 2.69, RMR = 0.05, RMSEA = 0.06, NFI = 0.90, CFI = 0.91, GFI = 0.86, TLI = 0.90, PNFI = 0.78, and HOELTER0.05 = 207 (Table 5; Bollen, 1989; Schumacker and Lomax, 2004).

As revealed in Figure 3, self-directed learning had positively directly influenced the college Students’ academic performance (γ = 0.19, p < 0.01), and 32% of the variation in academic performance could be explained by self-directed learning. The 95% confidence interval of the aforementioned direct effect, as tested using the bias-corrected percentile bootstrap method, was 0.03–0.33 (Table 6); that the confidence interval did not include 0 indicated a direct effect, highlighting self-directed learning as a partial predictor of academic performance. Thus, Hypothesis 1 was confirmed.

According to the mediation model diagram constructed on the basis of the hypotheses of this study, three indirect effect paths existed between self-directed learning and academic performance.

First, self-directed learning directly positively influenced optimism (γ = 0.56, p < 0.001), and optimism exhibited a direct positive effect on academic performance (γ = 0.37, p < 0.001), with both of these effects being significant. The indirect impact through optimism was of size 0.21 (γ = 0.56 × 0.37, p < 0.01), with γ referring to the outcome of the path values of Hypothesis 2 (Figure 3). The indirect influence of this path was determined using bias-corrected percentile bootstrapping with the 95% confidence interval; the resulting interval was 0.10–0.25, which excluded 0, indicating that the mediation effect of optimism in the first path was significant (Table 6); thus, Hypothesis 2 was confirmed and indicated a partial mediation because both direct and indirect effects were observed.

Second, self-directed learning exhibited a direct positive impact on mental health (γ = 0.39, p < 0.001), and mental health positively directly influenced academic performance (γ = 0.39, p < 0.001), with both effects being significant. The indirect effect through mental health was of size 0.15 (γ = 0.39 × 0.39, p < 0.01), with γ referring to the outcome of the path values of Hypothesis 3 (Figure 3). The bias-corrected percentile bootstrap method was used to assess the indirect influence of this path with the 95% confidence interval; the resulting confidence interval was 0.07–0.20, which excluded 0, indicating that mental health had a significant mediating effect in the second path (Table 6); thus, Hypothesis 3 was confirmed and indicated a partial mediation because both direct and indirect effects were present.

Third, self-directed learning had an effect on optimism (γ = 0.56, p < 0.001), optimism exhibited a direct positive effect on mental health (γ = 0.52, p < 0.001), and mental health positively influenced academic performance (γ = 0.39, p < 0.001), with all effects being significant. The indirect impact of self-directed learning on academic performance by means of two chain-mediating variables, optimism and mental health, was of size 0.11 (γ = 0.56 × 0.52 × 0.39, p < 0.01), with γ indicating the outcome of the path values of Hypothesis 4 (Figure 3). The bias-corrected percentile bootstrap method was used to assess the indirect influence of this path. The 95% confidence interval was 0.06–0.15, which excluded 0, indicating that the chain-mediated effects of mental health and optimism in the third path were significant (Table 6); thus, Hypothesis 4 was confirmed and reflected partial chain mediation because both direct and indirect effects were observed.

According to the preceding analysis (Table 6), the direct effect of self-directed learning on academic performance was of size 0.19 (p < 0.01). Self-directed learning’s total indirect impact on academic performance was obtained by summing the effects of three indirect paths: 0.21 + 0.15 + 0.11 = 0.47 (p < 0.01). The 95% confidence interval of the indirect effect was 0.37–0.59, which excluded 0, thus confirming the mediating effect. The model’s total effect was 0.19 + 0.47 = 0.66 (p < 0.01). The 95% confidence interval of the bootstrap method was used for assessing the total effect, with the resulting interval being 0.55–0.77, which does not include 0; the total effect thus existed. Because total, direct, and indirect effects were present, the structural model was partially mediated; moreover, the three variables—self-directed learning, optimism, and mental health—explained 71.5% of the variation in academic performance.

This research proposes that self-directed learning is an essential learning method that can help Wuhan college students adapt to online education and obtain favorable learning outcomes because of the positive direct and indirect effects prompted by self-directed learning that lead to academic success. On the basis of the importance of self-directed learning for Wuhan college students, this research offers the following two recommendations for administrators of Chinese higher education and student managers:

First, online learning is a safer teaching mode in the current global pandemic context, and self-directed learning can be regarded as a prerequisite for the smooth implementation of online teaching (Zhang W. et al., 2020; Zhang X. et al., 2020; Zhang Z. et al., 2020). This study found that self-directed learning is a predictor of academic achievement. Therefore, Chinese universities (such as those in Wuhan city) should prioritize Students’ self-directed learning skills, help these students develop into self-directed learners, and accommodate online education.

Second, the chain mediation results revealed that students with stronger self-directed learning skills may possess an optimistic mindset that also provides them with a healthy mental state for obtaining positive academic outcomes (Alkhedr, 1999; El-Anzi and Owayed, 2005; Zhang et al., 2011; Bostani et al., 2014; Hanssen et al., 2014; Agnafors et al., 2020). Therefore, the positive indirect effect of self-directed learning on academic performance may also serve as a reminder that higher education in China should follow the concept of constructivist teaching, in which students are encouraged to actively explore and discover knowledge as well as keenly construct the meaning of the knowledge they acquire (Piaget, 1980). Through this approach, students may become more optimistic and active by means of self-directed learning, thereby maintaining their mental health, resisting academic stress, and obtaining more favorable learning outcomes. Thus, this research recommends that counselors serve as guides for students in colleges and universities in China (Ministry of Education of the People’s Republic of China, 2017), training these students in self-directed learning and helping them become active learners with an optimistic attitude and relaxed mental state during online teaching.

Only the self-directed learning situation of Wuhan college students was investigated herein because Wuhan is the most representative area of where the pandemic first broke out in China. However, online teaching has also been applied in other countries and throughout China. Future research may compare the results obtained in diverse regions or countries.

Because of the restrictions related to pandemic prevention and control in China, this study collected data solely through an online survey questionnaire. In the present investigation of Wuhan college students, a group with long-term online learning experience because of the COVID-19 pandemic, only the survey method was employed and relevant behavioral details and manifestations of the interviewees could not be obtained. Therefore, if future policy permits, follow-up research may integrate face-to-face interviews to further supplement the research data.

This was a cross-sectional study, and only the potential relationships between variables could be analyzed using SEM. Future studies may use experimental techniques to examine the connection of self-directed learning with learning performance.