The multidimensional theory of anxiety proposes that cognitive anxiety shares a negative linear relationship with performance, and confidence shares a positive linear relationship with performance (Martens et al., 1990). Research on competitive state anxiety is based on the premise that cognitive anxiety has a negative linear relationship with performance and that confidence has a linear positive relationship with performance and is considered to be a protective factor against cognitive anxiety (Weinberg and Gould, 2014). When facing stress, athletes with higher confidence are better able to interpret anxiety as a positive emotion and promote sport performance (Martens et al., 1990; Woodman and Hardy, 2003; Hays et al., 2009). Similarly, confidence has a significant effect on sport performance in archery (Zhao et al., 2013; Bebetsos, 2015), even though it affects performance to the greatest extent on different factors among skills and mental and physical fitness (Yang et al., 2008; Kim et al., 2015). In addition, several studies have examined the effects of trait anxiety (Kaur and Shenoy, 2019) and state anxiety (Lim, 2016; Kaur and Shenoy, 2019) on archery performance, but it is not clear whether there is a difference in the effect of trait sport-confidence and state sport-confidence [concepts belonging to a sport-specific model of self-confidence (Vealey, 1986)] on archery performance. In summary, we propose the following hypothesis:

Attention is described as the sustained focus of cognitive resources on information while filtering or ignoring extraneous information. It is a vital prerequisite of successful performance in sports (Abernethy et al., 2007 Moran, 2012). Athletes with a higher attention ability are better able to achieve desirable performance in competition (Moran, 1996; Love et al., 2018). Likewise, researchers have concurred that attention, especially concentration, is critical for archery athletes (Lee, 2009; Zhao et al., 2013; Bebetsos, 2015; Taha et al., 2018). Gonzalez et al. (2017) pointed out that longer quiet eye durations (QED) are required for archers to aim accurately at the bullseye, and they noted that attention control plays an important role in determining QED performance. A study focusing on athlete selection indicated that attention allocation is an important indicator for selecting elite archers (Yang et al., 2008). Lu et al. (2021) also found that national team archers performed better than provincial team archers in terms of attentional network efficiency. During the aiming process, archers need to rely on their own proprioception and environmental information (e.g., wind direction) to adjust their posture or aiming skills (Lu et al., 2021), which requires strong attention allocation and a certain breadth of attention to cope with and process this information. At the same time, athletes seem to invoke conflict control to shift attention from distracting factors (e.g., focusing on the performance of other athletes or caring about ranking) to valid information (e.g., paying attention to coordinated body movements). Thus, following the facts above, the different aspects of attention, that is, shifting, breadth, and allocation, may all influence the archery performance. In summary, hypothesis 2 is proposed:

In sports (e.g., archery) that require movement of fine control (Eysenck and Calvo, 1992; Mellalieu et al., 2004, 2009) or a high level of concentration (Kubiak et al., 2019), anxiety could provoke a decline in sport performance. Emotion regulation is often used by individuals as a means of reducing anxiety (Cisler et al., 2010). Admittedly, emotional control or emotion regulation is an essential mental skill for archery athletes (Yang et al., 2008; Zhao et al., 2013; Kim et al., 2015). However, previous research has progressed mostly to emphasize the importance of emotion regulation, and there has been less exploration of specific means of emotion regulation applicable to archery athletes. Although there are a considerable number of ways to regulate emotions, one prominent and adaptive approach to control emotions is through cognitive reappraisal (Gross and John, 2003; Gross, 2010; Uphill et al., 2012). Studies have found that positive cognitive reappraisal is positively associated with motor performance in table tennis, whereas expression suppression is not (Kubiak et al., 2019), and it enhances archery athletes’ feelings of positive emotions and improves performance on experimental tasks (Kim and Oh, 2015). In addition, similar to cognitive reappraisal, arousal reappraisal prevents golfers from “choking” and to some extent ensures performance in putting tasks (Balk et al., 2013). In conclusion, we chose cognitive reappraisal as a specific strategy for emotion regulation and proposed the following hypothesis:

According to attentional control theory, anxiety impairs the goal-directed attention system and reduces individuals’ performance on attention control by inhibiting attentional transfer processes (Eysenck et al., 2007; Eysenck and Derakshan, 2011). Among archery athletes, it has also been shown that concentration increases when competitive anxiety decreases (Indahwati and Ristanto, 2016). Multidimensional anxiety theory states that sport-confidence is an important mental quality for individuals to avoid cognitive anxiety (Martens et al., 1990). Athletes with high levels of confidence improve their attentional control and motor performance by investing more mental effort, even when they are anxious (Eysenck and Calvo, 1992). As evidence, a recent questionnaire-based study showed that high levels of attentional control are related to lower levels of competitive state anxiety and higher levels of self-confidence, and competitive state anxiety and self-confidence are significant predictors of attentional control (Tomé-Lourido et al., 2019). As a result, sport-confidence may ensure attention control by resisting anxiety and maintaining confidence. Studies on shooting athletes found that confidence had a relatively greater effect on concentration among the mental skills of target setting, representational and anxiety control (Mun, 2011), suggesting to us that confidence contributes to shifting, breadth, and allocation of attention. In summary, it is proposed the following hypotheses:

Research on the relationship between confidence and cognitive reappraisal in sports is relatively scarce. We only know that cognitive reappraisal can enhance subjects’ confidence in memory tasks (Richards and Gross, 2000). Studies on arousal reappraisal have shown that the reappraisal group displayed higher self-confidence than the control group when completing a pressurized dart-throwing task (Sammy et al., 2017). Along these lines, cognitive reappraisal affects emotion-related outcomes by modulating the intensity of emotional responses (Gross, 2010) and stress responses (Jamieson et al., 2013), such as reducing anxiety (Hofmann et al., 2009; Uphill et al., 2012; Efinger et al., 2019) and depression (Uphill et al., 2012) and shifting negative stress states (Jamieson et al., 2013). A study of a sport-specific model of self-confidence also states that these negative emotions are related to sport-confidence (Vealey, 1986). In addition, archery athletes’ sport-confidence is significantly and negatively correlated with competition anxiety and negative emotions (Zhao et al., 2013; Lim, 2016). Furthermore, according to the theory of self-efficacy, emotions are one of the sources of self-efficacy, and successful regulation of emotional experiences can enhance individuals’ perceptions of personal competence (Bandura, 1977a,b). As a result, cognitive reappraisal may promote sport-confidence by improving emotional experience and consequently affect sports performance. Therefore, we propose the following hypotheses:

Both attention and cognitive reappraisal are related to cognitive control (Ochsner and Gross, 2005; Buschman and Miller, 2010). Although cognitive reappraisal may be associated with lower cognitive costs (Gross and Thompson, 2007), recent research suggests that cognitive reappraisal depletes cognitive resources (Sheppes and Meiran, 2008; Keng et al., 2013). The use of cognitive reappraisal requires the invocation of cognitive processes, including working memory, task switching, and antagonistic dominance responses (Ortner et al., 2016; Gan et al., 2017), which may conflict with attentional control and attentional orientation. Under the condition of stress, cognitive reappraisal not only takes away self-control resources but also reduces the control of negative emotions (Sheppes and Meiran, 2007; McRae et al., 2012; Ortner et al., 2016). When athletes deliberately seek to be “more in control” to gain control over emotions and movement skills, it often comes at the cost of depleting more cognitive resources, as evidenced by decreased performance on reaction time tasks and increased rates of missed and false manipulation tasks (Sun et al., 2013; Ortner et al., 2016). Thus, we propose the following hypotheses:

Combining the hypotheses above, a theoretical model of archery athletes’ sport-confidence, attention, and cognitive reappraisal influencing sport performance was obtained, as shown in Figure 1.

Sixty-one athletes from the Chinese National Archery Team participated in the study. Among them, 26 were males and 35 were females, 12 were international-level athletes, 30 were national-level athletes, and 19 were first-class athletes. These technical hierarchies were officially determined by the General Administration of Sport of China. International-level athletes are those who have won certain competition placement in international competitions, national-level athletes are those who have achieved corresponding competition results international, continental, or national competitions, and first-class athletes are those who have won certain competition results in national competitions (General Administration of Sport of China, 2019). Informed consent was obtained from all athletes and coaches before testing. The detailed demographic distribution is presented in Table 1.

All of the athletes took the psychological test as a group at the same time and participated in an archery competition 1 week after the test. The psychological tests were conducted 2 h after the end of training in a warm, well-ventilated, and quiet room. Before implementing the attention tests, the chief examiner explained each subtask of the EAAT one by one and confirmed that the athletes understood the content of the EAAT, and then counted down the time to start the attention tests on the athletes. All three subtasks of the EAAT were limited to 3 min (Yin et al., 2006). After completing the attention tests, the athletes filled out the ERQ-CRV, the SSCI-CV, and the ASCI. The archery competition was conducted in accordance with Olympic archery competition rules. During the competition, we collected the total ranking score in qualification phase and the total set score in the matchplay phase. In addition, the technical level data are collected from the “Athlete Technical Hierarchy Integrated Management System” before the competition.

In this cross-sectional study, structural equation modeling (SEM) was chosen to analyze the measurement and structural models. SEM was divided into two categories: covariance-based SEM (CB-SEM) and partial least squares-SEM (PLS-SEM; Chin, 1998). Compared to CB-SEM, PLS-SEM (1) is more suitable for addressing small sample data and does not require normally distributed data (Chin, 1998; Urbach and Ahlemann, 2010); (2) focused on predictions and theory building, rather than model verification, by maximizing the explained variance of endogenous latent variables because its power is relatively higher (Hair et al., 2011); and (3) The model constructed in this study is permeated with various theories or hypotheses and is a development of theory rather than a validation of the theory, which determines the applicability of PLS-SEM for this study. The data were analyzed in SmartPLS V.3.3.3 (Ringle et al., 2015).

We assessed the model using a two-stage analytical procedure for PLS-SEM as delineated by Hair et al. (2020): (1) We tested the measurement model to refine our measures and establish validity and reliability; and (2) We examined the structural model and tested the significance of path coefficients. The PLS algorithm was utilized to calculate the weights/loading, path coefficients, predictive accuracy (R²), effect size (f²), reliability, and validity of the model; bootstrapping (resamples = 5,000, confidence intervals method is bootstrap with corrected bias-corrected and accelerated bootstrap) was utilized to calculate the statistical significance of path coefficients; and blindfolding (omission distance = 7) was utilized to calculate Stone-Geisser’s Q², which indicates predictive relevance (Hair et al., 2017). The remaining calculation parameters were the software’s default values. To comply with the rule of the “minimum sample size should be equal to the larger 10 times the largest number of formative indicators used to measure one construct sample” in the PLS-SEM algorithm (Marcoulides and Saunders, 2006), we packaged the SC-state and SC-trait in the form of summation of each item’s score.

Items for which the indicator loading is less than 0.6 should not be included in the subsequent analysis (Sarstedt et al., 2017). The preliminary PLS-SEM algorithms showed that the outer loading of the first and second items of the latent variable cognitive reappraisal ranged from 0.475 and 0.586. Based on the results of both PLS algorithms and empirical analyses (the two items are correlated in connotation with other items, and removing them does not affect the content validity), we deleted the two items above. Meanwhile, the relationship between cognitive reappraisal and attention (H6) was removed from the original hypothesized model due to a non-significant path coefficient [β = 0.200, 95% CI ∈ (−0.069, 0.449)]. The following section will develop the analysis with the revised model (Figure 2).

To measure reliability, indicator loadings higher than 0.7, composite reliability (CR) ranging from 0.8 to 0.95, and Cronbach’s alpha (CA) higher than 0.7 have been used as common criteria to assess indicator reliability and internal consistency reliability (Sarstedt et al., 2017). However, some researchers criticized CA as an indicator to assess the reliability of the measurement model in PLS-SEM, in that the calculation of CA is based on the assumption that “all items have tau-equivalence reliability,” which may underestimate the true reliability (Hair et al., 2012; Peterson and Kim, 2013). As shown in Table 2, indicator loadings (0.708–0.912) of each item, CR values (0.826–0.897) of each latent variable, and CA values (0.771–0.807) of the latent variables above cognitive reappraisal, sport-confidence, and attention comply with the requirements of the test standard. Although the CA value for the latent variable archery performance was 0.696, we considered the CA for archery performance to be acceptable due to the inherent large uncertainty in archery performance and the general tendency of CA to underestimate the reliability. In all, the measurement model was considered to have acceptable reliability in this study.

The validity tests of measurement models focus on convergent validity and discriminant validity tests (Hair et al., 2020). Convergent validity, which reflects the extent to which different measures of the same construct converge or strongly correlate with one another (Taherdoost, 2016), was assessed through the average variance extracted (AVE). All AVEs (Table 2) in this study exceeded the recommended threshold of 0.5 (Hair et al., 2020), indicating that the measurement variables can effectively explain the latent variables and that the measurement models were both good in convergent validity. The discriminant validity assessment ensures that a reflective construct has the strongest relationships with its own indicators (Hair et al., 2020). Traditionally, the Fornell–Larcker criterion (Fornell and Larcker, 1981) discriminated whether A V E i > max | r i j | ( ∀ i ≠ j ) was used, but Henseler et al. (2015) pointed out that the PLS algorithm would overestimate indicator loadings and underestimate correlation coefficients among the constructs, and they suggested that a heterotrait–monotrait ratio of correlations (HTMT) lower than 0.85 should be used as a criterion of discriminant validity assessment. Tables 3 and 4 present the results of these criteria. Both the Fornell–Larcker criterion and HTMT test were approved. Therefore, we demonstrated satisfactory validity of our measurement model.

Assessing the structural model, we checked for multicollinearity issues by examining the variance inflation factor (VIF) values first. No multicollinearity issues were found, as all VIF values, ranging from 1.272 to 2.370, were below the threshold value of three (Hair et al., 2020). As shown in Figure 2, the endogenous variables’ R² value (range from 0.142 to 0.336) and Q² value (range from 0.085 to 0.173) also indicated good predictive accuracy and predictive relevance, where R² equals 0.67, 0.33, and 0.19 for large, medium, and small predictive power, respectively, and Q² values greater than zero are meaningful, whereas values below zero indicate a lack of predictive relevance.

To assess the direct effect of the path, path coefficients’ significance (p) and effect size (f  ²) were used, in which f  ² values above 0.02 and up to 0.15 are considered small effects, values of 0.15 and up to 0.35 are considered medium effects, and values 0.35 and above are considered large effects (Cohen, 1988). The relationships between sport-confidence and archery performance (H1; β = 0.444, p < 0.001), sport-confidence and attention (H4; β = 0.377, p < 0.001), and cognitive reappraisal and sport-confidence (H5; β = 0.455, p < 0.001) were found to have a medium effect size. Small effects were observed for the relationships between attention and archery performance (H2; β = 0.344, p < 0.001) and between cognitive reappraisal and archery performance (H3; β = −0.268, p < 0.001; see Table 5). Overall, these results support H1, H2, H4, and H5. Beyond our anticipation, the results do not support H3, but a significant negative relationship between cognitive reappraisal and archery performance was found. In addition, no significant relationships were found between cognitive reappraisal and attention (H6), which had been clarified in the model revision process.

The mediation analysis procedure recommended by Hair et al. (2017) was used, in which, if the coefficient of the path including the mediator variable is significant, then we need to determine the significance of the direct effect. Partial mediation is found when the direct effect is significant, whereas support for full mediation is demonstrated when the direct effect is not significant. Among the results of the mediating effects test (Table 5), we found that attention acted as a partial mediator of the relationships between sport-confidence and archery performance (H4a, β = 0.130, p < 0.05). Sport-confidence acted as a suppressing mediator between cognitive reappraisal and archery performance (H5a, β = 0.202, p < 0.05) due to the opposite sign of the coefficient of the indirect effect and the coefficient of the direct effect (β = −0.268, p < 0.05). Moreover, a mediating effect of sport-confidence was found in the path between cognitive reappraisal and attention (H6a, β = 0.172, p < 0.05). Certainly, sport-confidence and attention have no chain mediating effect between cognitive reappraisal and attention (H5b, β = 0.059, p > 0.05). Overall, the results of the indirect relationships provided support for H4a, H5a, and H6a, and did not support H5b.

To compare the magnitude of different independent constructs affecting the dependent constructs, we collated the total effects of sport-confidence, attention, and cognitive reappraisal to influence archery performance by summing the direct and indirect effects. As shown in Table 6, the ranking of the total effects of the different constructs on archery performance was, in descending order, sport-confidence [0.574, 95% CI ∈ (0.342, 0.796)], attention [0.344, 95% CI ∈ (0.068, 0.581)], and cognitive reappraisal [−0.007, 95% CI ∈ (−0.354, 0.271)].

The primary limitations found in this study concerned the cross-sectional nature of data collection. Cross-sectional designs limit the causality in the hypothesized relationships by examining only the relations among the variables; hence, we cannot assert that the antecedents are causally related to desired outcomes. In addition, given the difficulty of reaching 61 national archery athletes at the same time and place, we could not guarantee collecting data periodically or conducting a complete experimental study. As a result, future research could further test the model using a longitudinal design or quasi-experiment to make the findings more convincing.

The model constructed in this study involved three psychological predictive constructs and explained 33.6% of the variance in archery performance. However, cognitive reappraisal may not be a one-size-fits-all approach to regulate emotion in archery competition. We not only encourage additional investigations to further verify the psychometric properties of the measures used in this study, but also suggest the inclusion of other constructs of emotion regulation in the model, such as acceptance. Acceptance encourages individuals to become aware of and accept their feelings, rather than trying to actively change them, and it not only reduces negative perceptions of unpleasant emotional and physiological states and maintains or increases physiological arousal (Desbordes et al., 2015; Feldman et al., 2016), but more importantly, it occupies lower cognitive costs and is easier to operate than cognitive reappraisal (Troy et al., 2017).

Furthermore, psychometric measures alone cannot possibly capture a thorough and detailed explanation of the factors influencing performance in archery. Thus, the present results can be complemented by integrating skill constructs, fitness constructs and to shed light on the relative importance of archery performance factors and guide the training practice process. Besides, recent studies have also suggested that genetic polymorphisms are associated with athletes’ emotional control (Altamura et al., 2019) and sport performance (Ahmetov et al., 2016), and therefore, the inclusion of genetic polymorphism structure in the model could be considered and guide the athlete selection.

Finally, the differences in the experience levels and level of participation among the athletes could affect their psychological capability which may affect the result. This study only focused on an elite archery athlete population, and therefore, the applicability of the model is currently limited to this group. Future research could further explore the applicability of the model to archers with different levels of experience or participation, even to athletes in other sports that require movement of fine control and a high level of concentration, such as shooting and golf. Comparisons between athletes of different experience levels and between different sports will bring their own contribution to the field in practice and in the literature.