Validation of Sport Anxiety Scale-2 (SAS-2) among Polish athletes and the relationship between anxiety and goal orientation in sport

This study aims to assess the validity and reliability of the Polish version of the Sport Anxiety Scale-2, as well as to determine the relationship between anxiety and goal orientation among high-performance and recreational athletes. A total of 519 athletes aged M = 22.83 (SD = 4.92) participated in the study, including 266 males and 253 females. 242 athletes trained professionally and 277 recreationally. The Sport Anxiety Scale-2 (SAS-2) was used to assess anxiety levels, while the Task and Ego Orientation in Sport Questionnaire (TEOSQ) and the Perception of Success Questionnaire (POSQ) enabled to assess athletes’ goal orientation. Confirmatory factor analysis showed a good fit of the model to the data for the Polish version of the Sport Anxiety Scale-2 (CFI = 0.945, RMSEA = 0.072). The models obtained during analysis of high-performance and recreational athletes, women and men, also presented a satisfactory fit to the data (CFI 0.932–0.946). The configural, metric, scalar and strict measurement invariances were demonstrated for high-performance and recreational athletes as well as among women and men. High internal consistency coefficients (alpha 0.81–0.91) and a high test–retest reliability indexes were reported (ICC 0.74–0.87). Women presented higher level of competitive anxiety than men. A positive relationship between competitive anxiety and athletes’ ego orientation was also presented. This relationship concerned particularly women practicing sport recreationally.

www.nature.com/scientificreports/ athletes consider the fact that they perform the task better than other athletes as success [27][28][29][30] . In a validation study of the SAS-2 scale, Smith et al. 9 obtained negative association of anxiety with task orientation and positive association with ego orientation. Fairly similar results were obtained by Grossbard et al. 25 in the study of young athletes. comparable results were also obtained by Duica et al. 31 . The addressed problem concerning the relationship between goal orientation and anxiety may be important from the perspective of sports practice because it is possible that the coach's actions (strengthening or weakening a specific goal [ego/task]) orientation of athletes may be, to some extent, connected with experiencing specific emotions, e.g., anxiety before and during competition. It's possible that activating an ego-oriented motivational climate during a training process simultaneously promotes an increase in competitive anxiety, whereas a task-oriented climate promotes a decrease in anxiety 9 . For example, task-oriented individuals may experience less anxiety in competitive situations because their sense of competence, based on an internal criterion for success, is not threatened by their performance compared to other athletes 32 . However, taking into account the character of correlational research, another direction of the relationship cannot be ruled out 9 , e.g., that individuals with higher levels of anxiety may be characterized by higher emotional arousal of behavior, which reduces concentration on the task (task orientation), while by reducing control (increased arousal) it increases confrontational behavior (ego orientation).
In addition, proceedings of our study took into account gender and level of participation in sport (high-performance, recreational). It was also worth determining whether there was a difference between competitive and recreational athletes because these groups differ in a number of factors e.g., motives for taking up sport activities, expectations of sport performance, personality traits etc. [33][34][35] . In light of previous research findings, comparing these groups in terms of anxiety may be relevant. As indicated in the review on the importance of personality for sport functioning 35 , professional elite athletes are characterized by greater emotional stability (strongly associated with lower anxiety) compared to recreational athletes. These results are also supported by a recent study on volleyball players, where professional athletes were characterized by lower levels of anxiety as a trait than amateurs 36 . Therefore, it can be assumed that low levels of anxiety are conducive to effective performance in competitive and stressful professional sport. However, the prediction of differences in anxiety across our researched groups is not entirely clear, as the SAS-2 scale measures generalized levels of anxiety before and during competition 9 . In the research conducted by Marcel and Paquet 37 where an older version of the SAS scale was used the results showed that athletes with high athletic level were characterized by higher anxiety than less advanced athletes. Differences here may be due to different approaches to sports competition and expectations concerning sports performance in these groups, e.g., professional athletes may have a much higher level of expectations concerning sports performance, which may intensify the experience of anxiety. However, in our SAS-2 scale validation, even more important than the analysis of differences in the levels of anxiety may be the assessment of invariance of anxiety measurement in the groups of professional and recreational athletes 38 . To our knowledge, neither such analyses exist for anxiety measurement by the SAS-2 scale, nor has a study been conducted on the relationship between anxiety and goal orientation among competitive and recreational athletes separately.
The growing popularity of sport psychology in Poland in recent years has been visible in the educational offer in this field. Several Polish universities provide specializations or postgraduate studies in sport psychology. Coaches and athletes cooperate with psychologists more often. In the latter aspect, it is worth noting that the basis for effective psychological intervention in sport is an accurate diagnosis. Despite the validation of various internationally known sports psychology scales into the Polish cultural context, there is still a need to adapt new ones. Therefore, validation of the SAS-2 scale will significantly broaden the possibilities of anxiety measurement among Polish athletes. The value of Polish validation is also enhanced by the possibility to use the scale not only on professional but also on recreational athletes. There are not many adaptations in the Polish cultural context which take into account a separate division into these groups of athletes. Moreover, a new tool in sport psychology such as the SAS-2 scale increases possibilities of searching for relations between anxiety and other variables, important from the theoretical and empirical point of view, which have not been verified before.

Methods
Participants. The study included 519 athletes aged M = 22.83 (SD = 4.92), 266 men and 253 women, 242 athletes trained professionally and 277 recreationally. The disciplines that the respondents practiced most often were: football, running, combat sports, swimming, volleyball, basketball, dancing, handball. All of the surveyed high-performance athletes were affiliated with professional sports clubs and reported that they trained in sports primarily for maximum results, moreover they systematically participated in national level competitions. Recreational athletes were not affiliated with professional sports clubs however participated in recreational athletic activities with competition and competed sometimes in sporting events (e.g., individual regional running competitions or academic competitions). They specified that they exercised mainly for health, physical fitness, and appearance. Due to the anonymity of the study, this information was obtained through a self-administered metric completed by respondents. Measurement tools. The Sport Anxiety Scale-2 9 consists of 15 items which form 3 subscales (somatic anxiety, worry and concentration disruption), 5 items each, and was used to investigate anxiety. The questionnaire instruction highlights that many athletes feel tense or experience nervousness before or during competitions and games. The respondents are asked to mark the number (1-not at all to 4-very much) that most closely corresponds to how they usually feel during or before sport competition. At the beginning of the adaptation procedure, an English language expert translated the SAS-2 scale into Polish, and then another expert translated the scale back into English. Finally, the final version of the scale was agreed upon by a group of experts consisting of three sport psychologists (including two coaches) and an English language expert. At the beginning of the study, the subjects were asked about doubts in understanding the statements or instructions. Nobody made www.nature.com/scientificreports/ any comments in this respect. The research was completely anonymous and the survey was conducted via the internet (online questionnaire). After reviewing the study description, the Bioethics Committee of Poznan University of Medical Science (Poland) issued a statement that the study had no features of a medical experiment and according to the Polish law and Good Clinical Practice it was not subject to the opinion of the Bioethics Committee. All methods were carried out in accordance with relevant guidelines and regulations. The research was performed in accordance with the Declaration of Helsinki. The study instructions included information that taking part in the research and submitting the results would be treated as the respondent's informed consent to take part in the research. The Task and Ego Orientation in Sport Questionnaire-TEOSQ 29,39 and the Perception of Success Questionnaire-POSQ 40 in the Polish adaptation 41,42 were used to measure goal orientation. The TEOSQ consists of 13 statements; 7 items concern task orientation and six items concern ego orientation. The respondent determines to what extent a given statement applies to him/her on a scale of 1-5. The POSQ, on the other hand, consists of 12 statements, 6 for each goal orientation (ego, task), where the respondent's task is also to assess to what extent he/she agrees with a given statement on a scale of 1-5 (A-E).
Statistical analysis. First, to assess construct validity confirmatory factor analysis was conducted. According to theoretical assumptions, a hierarchical model was estimated with a general factor of anxiety (overriding factor) and three subscales: somatic anxiety, worry and concentration disruption. Next, the 3-factor model without the overriding factor of anxiety was estimated, taking into account the three subscales. Additionally, in order to check the validity of extracting the subscales of the tool, the 1-factor model (without extracting the subscales of the tool) was estimated and compared with two models, including the subscales. CFI and TLI above 0.90, RMSEA and SRMR below 0.08 were assumed to indicate a satisfactory fit of the postulated models to the data 43 . Due to deviations from multivariate normality of the distribution, the correction proposed by Satorra-Bentler 44 was applied.
Additionally, scale invariance by gender and type of sport participation (high performance, recreational) was examined. First, a model for configural invariance was estimated, then for metric invariance, where factor loadings within groups were fixed. Then a model for scalar invariance was estimated, where regression intercepts were additionally fixed, and finally a model for strict invariance, where residuals within comparison groups were fixed. It was assumed that a decrease in CFI below 0.01 and an increase in RMSEA above 0.015 indicate significant differences within groups. For SRMR, an increase of 0.01 for scalar and strict invariance, and 0.03 for metric invariance, indicated significant group differences 38 .
Convergent validity indices of the SAS-2 scale were also analyzed. Item reliability was assessed by analyzing the values of factor loadings and their values squared (the proportion of variance in an item due to a given construct). In general, criteria ranging from 0.40 (most liberal) to 0.70 (close to 50% of the variance in the item explained by the factor) are adopted for minimum standardized loadings values 45 . Considering these criteria, as well as the analysis of factor loading values presented in SAS-2 validation articles conducted in other countries, we considered that a minimum standardized loading value of 0.5 would be sufficient. Average variance extracted (AVE) and composite reliability (CR) values were then calculated for each subscale 46 . It is conventionally accepted that an AVE value above 0.5 and a CR value above 0.6 (or 0.7 under a more restrictive assumption) indicates sufficient convergent validity. However, it is also often assumed that a construct meets the requirement of convergent validity when AVE is below 0.5 but composite reliability (CR) requirement is fulfilled 46 . Discriminant validity, on the other hand, was assessed by comparing the AVE of a given latent variable with the maximum shared variance (MSV) (the greatest value of the square of the correlation of a given latent variable with another construct-latent variable). The discriminant validity of the construct is confirmed when the AVE value is higher than the MSV 46 .
The reliability of the scale was determined by calculating Cronbach's alpha and McDonald's omega coefficients, as well as the ICC between two measurements of the same athlete at a two-week interval (N = 189, all athletes completed the scale twice). The discriminatory power of a particular test item was calculated by its correlation with the total subscale score.
A two-factor analysis of variance was used to compare individuals by gender and level of sport participation in terms of mean anxiety. The r-Pearson correlation coefficient was used to assess the relationship between anxiety and goal orientation (N = 312).

Results
Construct validity: confirmatory factor analysis. As indicated by the data in Table 1, both the hierarchical model (Fig. 1) and the 3-factor model (Fig. 2) fit the empirical data well. They had much better fit than the 1-factor model, which did not obtain an adequate fit to the data. The 3-factor model also fit the data well within gender and in the group of high performance and recreational athletes (Table 1).
Factor loadings of the SAS-2 scale model achieved statistical significance (p < 0.001) and range from 0.53 to 0.91 (Fig. 2). Thus, squared multiple correlations for the worry scale were respectively: 3-0.69, 5-0. www.nature.com/scientificreports/ Invariance across gender and level of participation. The postulated models fit well to the data. The decrease in CFI greater than 0.01 and the increase in RMSEA greater than 0.015 were not observed in any case. Hence, measurement invariance of the SAS-2 scale can be assumed both across gender and level of sport participation. Analysis of the SAS-2 invariance by gender and level of participation is presented in Table 2.   Table 3.

Reliability and discriminatory powers of the Polish version of the Sport Anxiety
Correlations between competitive anxiety and goal orientation among high-performance and recreational athletes. Statistically significant, negative small associations were found between anxiety (general and subscales: somatic anxiety and concentration disruption) and ego orientation (TEOSQ subscale) in the whole group of the examined athletes. Correlations between anxiety and athletes' goal orientation are presented in Table 4. Moreover, significant negative associations were found between general anxiety and the subscales of somatic anxiety and ego orientation in the group of women and in the group of recreational athletes.
In the case of recreationally training women, the relationships were stronger and concerned both the TEOSQ and the POSQ. The positive correlation between concentration disruption and ego orientation assessed with TEOSQ was also noted in the group of high-performance male athletes. Correlations between anxiety and goal orientation in males and females, recreational and high-performance athletes are presented in Table 5.

Discussion
First of all, through confirmatory factor analysis, the theoretical validity of the Polish version of the SAS-2 was examined. The model of the Polish version of the scale fit the empirical data well both for the entire group of respondents and separately in the groups of women, men, high-performance and recreational athletes. The model fit was not perfect (e.g., CFI > 0.95, TLI > 0.95), but it was sufficient to accept the SAS-2 factor validity (CFI > 0.90, RMSEA < 0.08). The hierarchical model and the 3-factor model had a significant advantage of fitting  www.nature.com/scientificreports/ to empirical data over the single-factor model, which confirms the validity of separating the theoretically justified three subscales of the SAS-2 scale. Comparing the obtained results with the results of other authors, it can be concluded that both hierarchical and the 3-factor model of the Polish version of the scale estimated on the whole group of respondents had worse fit to the data than the models presented by the authors of the scale 9 , the models of the Spanish version of the scale 10 and the 3-factor model of the American version of the scale 13 . In turn, the models of the Polish version of the scale obtained a better fit to the data than the model (three factor model without correlating errors for items) of the Brazilian version of the scale 6 . Based on the CFI index, the models of the Polish version of the SAS-2 also obtained a better fit to the data and based on the RMSEA, a slightly worse fit than the 3-factor Malaysian model of the scale conducted on young athletes 12 and the Korean version of the SAS-2 (both hierarchical and the 3-factor model of the scale) 14 . High correlations were also found between the factors representing each of the anxiety subscales, which is consistent with previous research findings. It can be assumed that the values of factor loadings obtained in the Polish validation of SAS-2 were satisfactory. The AVE values for the worry and somatic anxiety scales were satisfactory, whereas the AVE value for the concentration disruption scale was slightly below the accepted conventional threshold value (0.5). However, convergent validity can be accepted in light of the high composite reliability values obtained (for each subscale above 0.8). Moreover, AVE for each subscale was higher than MSV, which allowed us to assume discriminant validity of the tested constructs. The lower AVE for the concentration disruption subscale of the SAS-2 scale was not surprising. Even the authors of the original version of the scale 9 obtained for this subscale lower factor loadings on average than the values obtained by us. From the data presented by Smith et al. 9 it is easy to calculate that those authors also obtained a lower AVE value than the one obtained in our study. However, due to the theoretical importance of the selected items, we assumed that it was not worth removing any of them. Lower AVE values are not that uncommon for the SAS-2 scale. This is partly due to conservativeness of the AVE coefficient 46 (it's necessary that the average factor loadings value is above 0.7). For example, slightly higher AVE values for the concentration disruption subscale than the values in our study were obtained by Correira and Rosado 47 , Putra et al. 48 and Cho et al. 14 . The two latter researchers also obtained values under 0.5 for the somatic anxiety subscale.
High Cronbach's alpha reliability coefficients obtained in our analysis were also observed for the individual subscales and the total score, which were close to the results obtained by the authors of the original scale 9 , higher than the coefficients obtained for the Spanish version 10 and medium higher than the coefficients obtained for the Brazilian version 6 . The high reliability of the Polish version of the SAS-2 scale was also confirmed by the high values of McDonald's omega coefficients and fair values of inter-class correlations for the study of the same athletes at a two-week interval. The measurement was also shown to be invariant considering gender (women, men) and the level of participation in sport (high-performance, recreational). Four basic types of measurement invariance were demonstrated in both cases, i.e. configural, metric, scalar and strict. Therefore, it can be assumed that women will use the scale similarly as men, and high-performance athletes will use the scale in a similar way   Table 4. Correlations between anxiety and athletes' goal orientation. W, worry; S, somatic anxiety; C, concentration disruption; A, anxiety (total score); E TQ, T TQ, ego and task assessed with the TEOSQ; E PQ, T PQ, ego and task assessed with the POSQ. *p < 0.05. www.nature.com/scientificreports/ to recreational athletes. Measurement invariance by gender was demonstrated previously [e.g., 5,13,25 ], whereas measurement invariance in high-performance and recreational athletes brings new information.

Variable
In Polish validation of the SAS-2 scale a comparison was also made of the average level of anxiety by gender and level of participation in sport. It was shown that women are characterized by higher levels of general anxiety and its individual components compared to men. The difference obtained can be explained by the greater overall emotionality and sensitivity of women, which is associated with higher levels of anxiety compared to men 49,50 . Our results are mainly consistent with the results obtained by Martinez-Moreno et al. 51 . It is partially different from the results obtained in the study conducted on children by Grossbard et al. 25 , who found higher scores on the worry subscale but lower scores on the concentration disruption subscale in girls than boys. There were no differences between boys and girls in overall anxiety levels 25 . In our study no differences were also found in the level of anxiety and its components in high-performance and recreational athletes. On the other hand, one might expect that high-performance athletes constitute a group selected in terms of various psychological characteristics, including correlates of emotional resilience 35 . Such a rationale argues for observing a lower level of anxiety in high-performance athletes. In turn, the questions of the questionnaire concern the level of anxiety before or during competition, and high-performance athletes are generally characterized by a greater focus on sporting results, are more subject to personal and social expectations and are subject to social evaluation 52 . These factors may further enhance the experience of anxiety. Hence, it is possible that these effects are somewhat opposite, which is ultimately associated with similar levels of anxiety in the compared groups.
During the proceedings of the validation of the Polish version of SAS-2 scale there were also demonstrated small positive associations of the anxiety (total score, somatic anxiety and concentration disruption) with ego orientation as determined by the TEOSQ questionnaire. Our results were partly similar to the results obtained by the authors of the original scale 9 , although their correlation values were higher, but also not very high. In addition, the relationships obtained by Smith et al. 9 concerned ego orientation measured by the POSQ questionnaire, which was not observed in our study. Similar relationships-when controlling for social desirabilitywere also shown by Grossbard et al. 53 in the study of young athletes. It is possible, as already mentioned in the introduction, that people who are more anxious, anxiously acting-with greater emotional arousal and at the same time with less behavioral control-concentrate more often on confronting the opponent and it is more difficult for them to concentrate, for example, on completing the task. On the other hand, it may be possible that individuals from the sports environment characterized by an ego-enhancing climate simultaneously become more emotionally aroused due to their confrontation orientation and treat the situation as more difficult and threatening 9 . In a more detailed analysis, we could show that the associations of anxiety with ego orientation Table 5. Correlations between anxiety and goal orientation in males and females, recreational and highperformance athletes. E TQ, T TQ; ego and task assessed with the TEOSQ; E PQ, T PQ, ego and task assessed with the POSQ; W, worry; S, somatic anxiety; C, concentration disruption; A, anxiety (total score). *p < 0.05, ** < 0.01, *** < 0.001. www.nature.com/scientificreports/ occurred mainly among recreational athletes, particularly among women (for ego orientation as determined by both the TEOSQ and the POSQ questionnaires). Thus, it is possible that the context of this relation is important here, i.e. the association concerns mainly an environment with lower social expectations of sporting performance (recreational athletes) and among women, who generally present less competitive orientation 54 . In a recreational sport area, there may be greater freedom in the interpretation of the sporting situation-it may be treated more or less confrontationally 55 . Moreover, one's own expectations of the results may also vary here. Therefore, it is possible that among women participating in recreational sports, there is a large variation in the perception and interpretation of the situation, where women with high expectations of the sporting outcome act more confrontationally (ego orientation) and at the same time experience greater anxiety. Furthermore, no relationship was found between anxiety and task orientation, as noted in previous studies with the SAS-2 scale 9,53 . The results obtained in these aspects are tough to interpret unambiguously and will require further research. The study is not without limitations, though. Primarily, in further studies on the Polish version of the SAS-2 scale, it is worth extending the research on relations between the SAS-2 and other scales, e.g., another scale measuring anxiety in sport and other variables important from the perspective of functioning in sport, which will additionally broaden the research on the validity of the scale. It is also worth testing the criterion validity of the scale by correlating its results with other variables e.g., various effectiveness indicators in sport. An analysis of differences in anxiety among individual and team athletes would also strengthen the validity of the scale. Although a distinction has been made between high performance and recreational athletes, it would be interesting to conduct validation on high performance athletes with different sport levels ranging from regional to international. Furthermore, the validation has been carried out on relatively young people, so it would be interesting to extend it to different age groups.
The Polish version of the SAS-2 scale has adequate psychometric properties indicating its high validity and reliability, i.e. satisfactory fit of the postulated models to empirical data was demonstrated in confirmatory factor analysis, invariance of the measurement with respect to gender and type of participation in sport was also noted, additionally high reliability indices as assessed by the test-retest method as well as high internal consistency indices were obtained. Women were characterized by lower levels of anxiety than men. Moreover, a negative correlation between anxiety and ego orientation was found, which was particularly evident among women involved in recreational sports activities. These latter issues still require further research. However, the validation results obtained by us support the possibility of using the SAS-2 scale as a valid and reliable tool in scientific research as well as in sports practice. Due to the proceedings of validation conducted by us, the Polish version of SAS-2 scale can be used not only for high-performance, but also for recreational athletes.

Data availability
The data generated and/or analyzed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/