Self Talk In Physical Education Psychology Essay

Published Date: 23 Mar 2015

Almost ninety-six percent of adults talk to themselves engaging in what is called internal dialogue, inner speech, self-statement, inner conversation, subvocal speech, self-verbalizations, or self-talk (Winsler, 2009). These concepts of internalization vary from theory to theory (Guerrero, 2005; Larrain & Haye, 2012) playing a key role in human processes and self-regulation (Meichenbaum, 1977; Berk, 1992). Meichenbaum viewed self-statements as indices of individuals' beliefs that may play a mediational role in behavioral performance and very recently Hatzigeorgiadis, Zourbanos, Galanis, & Theodorakis (2011) with the use of a meta-analytic approach revealed a positive moderate effect size (ES = .48) supporting the facilitative effects of self-talk on sport task performance. In sport and physical activity literature the term that has prevailed for the description of self-statements is self-talk. Despite the recent growth of self-talk research in sport, the lack of theoretical background in the self-talk literature is evident (for review see, Theodorakis, Hatzigeorgiadis, & Zourbanos, 2012). Hardy, Oliver, and Tod (2009) proposed a conceptual model for the advancement of the field. Their model postulates that personal and situational factors influence athletes' self-talk, which in turn has an impact on cognitive, motivational, behavioral and affective mechanisms, and subsequently on their sport performance. According to the model, among others (e.g., cognitive processing preferences, personality, anxiety) one of the personal antecedents that influences individual's self-talk are achievement goal orientations (e.g., Harwood, Cumming, & Fletcher, 2004; Hatzigeorgiadis & Biddle, 2002). Hardy et al. (2009) noticed that "While both self-concept and forms of anxiety may be antecedents of self-talk, preliminary evidence suggests that a motivation-based personality disposition, achievement goal orientation, might be another" (p. 41), stressing the importance of examining goal orientations as personal antecedents of self-talk. The main objective of the present research was to bridge the gap between positive and negative self-talk and achievement goals using two theoretical models of achievement goals which are presented below.

Achievement Goal Theory (AGT)

AGT is a central theoretical framework in the literature often used by researchers and sport psychologists to investigate and to understand why some individuals seem to be more motivated than others in sport and physical activity (e.g., Roberts, Treasure, & Conroy, 2007). Achievement goals were primarily examined using a dichotomous model that distinguished between two types of goals namely task and ego (Nicholls, 1984) or learning and performance goals (Dweck & Leggett, 1988). Task-oriented individuals adopt self-referenced criteria to define success, focus on mastery, use effective cognitive strategies to master a task, are intrinsically motivated, give high value to effort and seek for personal improvement (e.g., Roberts et al., 2007). On the other hand, ego-oriented individuals evaluate success through comparison with others' ability, focus on outperforming others, value high normative ability and pursue the exhibition of high normative ability. Biddle, Wang, Kavussanu, and Spray (2003) in their review of studies in physical activity concluded that task orientation was a significant predictor of enjoyment, satisfaction, intrinsic motivation, positive affect, and perceived competence and that ego orientation significantly predicted cognitive anxiety, stress and cognitive interference.

Later, Elliot and his colleagues modified the dichotomous model by proposing a trichotomous model that included mastery, performance approach, and performance avoidance goals (e.g., Elliot & Church, 1997) and then a 2x2 model that included mastery approach, mastery avoidance, performance approach and performance avoidance goals (e.g., Elliot & McGregor, 2001). Based on the 2x2 framework, Papaioannou, Zourbanos, Krommydas, and Ampatzoglou (2012) revealed in their review that mastery approach goals were connected with the most desirable motivational outcomes in sport and physical education while performance approach goals were associated with fewer but still positive motivational outcomes. Correspondingly, avoidance goals presented the less adaptive patterns of motivation and behavior.

Roberts et al. (2007) and Papaioannou et al. (2012) argued that differences between Nicholls' and Elliot's conceptualization have led to inconsistent findings due to measurement issues (Elliot & Murayama, 2008), with many questions still remaining unanswered in competitive sport and physical education settings (e.g., Harwood, Spray, & Keegan, 2008). For example, the concept of different definitions of success is inherent in the conceptualization and measurement of achievement goals in Nicholls' theory but is not considered at all in Elliot's. However, Papaioannou et al. (2012) noticed that high levels of motivation occur when task accomplishment conveys a meaning that is tied to an individual's long-term major outcome in life, which is defined as success. Another example concerns the notice of Papaioannou et al. (2002) and Roberts and Kristiansen (2012) that the standards to evaluate mastery are not specific in the wording of mastery approach items which are similar to the "do-your-best goals" that have been criticized by goal-setting researchers (Lock & Latham, 1990) (e.g., "My goal is to learn as much as possible" in Elliot & Murayama, 2008; "to answer a lot of questions correctly on the exams of this class" in Elliot, Murayama & Pekrun, 2011). These differences in conceptualization and item construction suggest that adaptive motivational outcomes might have stronger association with task goals than with mastery approach goals. A critical difference between the original dichotomous and the subsequent 2X2 models concern the role of perceived competence when individuals encounter performance difficulties (Elliot & Dweck, 1988). According to Nicholls (1984) individuals espousing high ego goals exhibit adaptive and maladaptive motivational patterns when they have high and low perceived competence respectively. It has to be noticed that the negative effects of ego orientation were originally predicted to occur when the individual encountered performance difficulties which led them to question their ability (Nicholls, 1984).  In the 2X2 model perceived competence has additive effects irrespective of individuals' goals (e.g., Elliot & Church, 1997).

Achievement goals and thoughts

In general two different research approaches are evident in the self-talk literature in sport. The first refers to self-talk as a cognitive strategy focusing on the beneficial effects of self-talk on performance enhancement (e.g., Mallett & Hanrahan, 1997). The second approach examines self-talk in the form of automatic thoughts exploring the factors that shape and influence athletes' self-talk content (e.g., Zourbanos, Hatzigeorgiadis, Tsiakaras, Chroni, & Theodorakis, 2010; Zourbanos et al., 2011). As stated above one of the personal antecedents that influences individual's self-talk are achievement goal orientations. Regarding research between achievement goals and thoughts in sport, Hatzigeorgiadis and Biddle (1999) revealed that task orientation was negatively related to disengagement thoughts irrespective of perceptions of competence. Furthermore, it was reported that for athletes with lower perceived competence ego orientation was positively related to experiencing disengagement thoughts, whereas for athletes with higher perceived competence no relationship between ego orientations and disengagement thoughts was shown. In another study, Hatzigeorgiadis and Biddle (2002) found that athletes with high ego and low task orientation goal were more vulnerable to disengagement thoughts than athletes with different goal profiles, whereas no consistent differences emerged concerning worrying thoughts. Finally regarding the relationships between perceived competence and cognitive interference (negative thoughts), Hatzigeorgiadis and Biddle (2000) revealed low but significant relationships. Overall, the results of the above studies seem to suggest that task orientation has more positive outcomes on the individual's thought patterns, whereas ego orientation doesn't and depends more on other personal factors such as perceived competence or situational factors, which can be failure (Nicholls, 1984). Nevertheless, all of the above studies have examined negative thoughts measuring them with the Thought Occurrence Questionnaire for Sport-TOQS (Hatzigeorgiadis & Biddle, 2000). Recently, Zourbanos, Hatzigeorgiadis, Chroni, Theodorakis and Papaioannou (2009) developed the Automatic Self-Talk Questionnaire for Sports- ASTQS for the evaluation of athletes' automatic thoughts. This questionnaire is more comprehensive and different to the TOQS in two ways: a) it measures four underlying factors of negative thoughts (adding somatic fatigue) instead of three (worry, disengagement, and irrelevant thoughts) and b) additionally to the athletes' negative thoughts, it also measures four positive ones (psych up, confidence, anxiety control and instruction). So far, only one study has examined the relationship between goal orientations and positive self-talk, but seeing it as a mental strategy and not as a content of thought (Harwood et al., 2004). They revealed that athletes with higher task and moderate ego orientations reported more positive thinking than athletes with lower task and moderate ego orientations and than athletes with moderate task and lower ego goal orientations. Furthermore, all the above mentioned studies examined athletes' self-talk and little research has been conducted in other fields such as physical education settings. Finally, the majority of the studies have used the dichotomous model investigating the relationships between task and ego goals and thoughts. Therefore, it would be interesting to examine the relationships of thoughts with mastery and performance approach and avoidance goals (Elliot & Murayama, 2008).

Expanding upon the studies of Hatzigeorgiadis and Biddle and also based on the theoretical postulations of self-talk (Hardy et al., 2009) and on Papaioannou et al.'s (2012) arguments about the differences between Elliot and Nicholls' conceptualizations on measurement issues, three studies were conducted aimed to examine more elaborately the relationships between achievement goals, perceived competence and student's thoughts. The first study tested the construct validity of ASTQS in Physical Education (PE) settings and examined the relationships between achievement goals, perceived competence and student's thoughts by using the dichotomous framework; the second study re-examined the relationships of the first study using this time the 2x2 framework; and finally, taking into consideration Harwood et al.'s (2008) suggestions that research should examine the moderating role of perceived competence on the relationship between achievement goals and psychological outcomes such as self-talk, the aim of the third study was to examine the interaction between perceived competence and achievement goals using both achievement goal frameworks on students' self-talk using multi-group path analysis.

Study 1

The ASTQS has demonstrated evidence of construct validity and reliability through experimental (Zourbanos et al., 2010) and cross-sectional research (Zourbanos et al., 2009; 2011) only in sport settings. The adaptation of the ASTQS in physical education would be a useful research tools to identify the nature and frequency of various thoughts students experience in this setting. Here we examined students' thoughts in PE during play with high and low sport achievers and we developed hypotheses based on theories of intrinsic motivation and anxiety (e.g., Csikszentmihaly, 1975) and previous findings in physical education related to this scenario (Papaioannou, 1995). Specifically we assumed that (a) students' negative self-talk (e.g., worry) would be more frequent but confidence-related self-statements would be less frequent when playing against the most competent student in a particular sport than during playing with the least competent student. We also expected that students would find more challenging to play against high achievers and therefore would score higher in psych up in comparison to play against low achievers. Based on assumptions of AGT and the preliminary findings of Hatzigeorgiadis and Biddle (1999, 2002) it was hypothesized that (b) task orientation would be positively related to students' positive self-talk dimensions and negatively to students' negative self-talk dimensions. Based on reviews of research in sport suggesting a weak positive association of ego goals with various cognitive processes (e.g., Duda & Hall, 2001), (c) a weak relationship between ego goals and positive self-talk was expected. Moreover based on theories of competence and motivation (e.g., Harter, 1978; White, 1959) it was expected that (d) perceived competence would positively correlate with positive self-talk dimensions, but would negatively correlate with negative self-talk dimensions.

Method

Participants and Procedure

Participants were 628 students (325 females and 303 males) with a mean age of 14.49 years (SD = .50) who attended physical education classes located in a city in central Greece. Each student assented to participate and provided written informed consent via a parent/guardian. Confidentiality and anonymity were assured throughout. Finally, instructions aimed at minimizing socially desirable responses were emphasized. The order of the questionnaires was counterbalanced. The questionnaires were completed under the supervision of one of the authors. Permission to conduct the study was obtained by the institution's research ethics committee.

Measures

Self-Talk in PE. An adapted version of the Automatic Self-Talk Questionnaire for Sports - ASTQS (Zourbanos et al., 2009) in PE was administered to assess students' self-talk. The instrument consists of 40 items assessing four positive (19 items) and four negative (21 items) ST dimensions. Positive self-talk consists of the dimensions of confidence (e.g., I believe in myself), anxiety control (e.g., Keep calm), psych up (e.g., Do your best), instruction (e.g., Concentrate on what you have to do right now). Negative self-talk consists of the dimensions of worry (e.g., I will lose), disengagement (e.g., I want to quit), somatic fatigue (e.g., I feel tired) and irrelevant thoughts (e.g., I am hungry). Participants were instructed to bring in their minds the most usual sport, game or activity that they play in the physical education lesson. Then they were asked to write down the game. After that they were told to imagine that they were playing against the most competent student in their age in this particular sport, game or activity. Furthermore, they were asked to recall their self-talk again but this time to imagine that they were playing against the least competent student in their age in this particular sport, game or activity. Finally, in both situations, they were told to indicate the frequency of thoughts that they usually experience or intentionally use while performing against the best/worst students in this sport, game or activity on a 5-point scale (0 = never, 4 = very often). Zourbanos et al. (2009; 2010; 2011) has supported the psychometric integrity of the ASTQS. In this study, Cronbach's alpha coefficients for both situations are displayed in Table 1.

Achievement goals in PE. Task and Ego Orientation in Physical Education

Questionnaire (TEOPEQ). This instrument (Duda & Nicholls, 1992), which has been used widely in Greece, has been adapted for physical education classes and has been shown to have very good psychometric properties (e.g., Biddle et al., 2003; Duda & Whitehead, 1998; Papaioannou & MacDonald, 1993; Marsh, Papaioannou, Martin & Theodorakis, 2006). Following the stem ''I feel most successful in physical education when. . .''), students respond to the seven task-oriented items (e.g. ''I learn something that is fun to do'') and six ego-oriented items (e.g. ''The others can't do as well as me'') of the instrument. Students respond to a 5-point Likert scale (1 = Strongly disagree, 5 = Strongly agree). Cronbach's alpha coefficients are displayed in Table 1.

Perceived competence. This subscale is part of the five-scale physical self-perception profile developed by Fox and Corbin (1989). It consists of six items (e.g. ''Some people feel that they are among the best when it comes to athletic ability'') and has been used several times in Greek physical activity settings and exhibits good psychometric properties (e.g. Papaioannou, Bebetsos, Theodorakis, Christodoulidis, & Kouli, 2006). Students responded to a 5-point scale (1 = Not at all like me, 5 = Very much like me to). Cronbach's alpha coefficient for the perceived competence subscale is displayed in Table 1.

Analysis

The 40 items of ASTQS were screened by the investigators and two physical educators, each of whom evaluated the wording and the applicability of the items in PE. Based on their recommendations the ASTQS was used in the current study with minor rewording (only one item was changed from the ASTQS: "What will others think of my poor performance" to "What will the teacher think of my poor performance") to be more suitable for the specific context of the study. The factor structure of the self-talk dimensions in PE were tested through confirmatory factor analysis (CFA) using the EQS 6.1 (Bentler & Wu, 2004). Three alternative models were tested. One where the original eight-factor structure was tested with the eight factors allowed to correlate, a 2-factor model where all positive self-talk items were set to load on a single positive factor and all negative self-talk items were set to load on a single negative factor (the two factors were allowed to correlate) and a 10-factor model where the four positive self-talk factors were set to form a second-order positive self-talk factor, and the four negative self-talk factors were set to form a second-order negative self-talk factor. To examine whether the chi square values differed significantly between the 3 models, the Satorra and Bentler's scaled difference qui-square test was conducted (Satorra & Bentler, 2001; Crawford & Henry, 2003). Four fit indices were used to assess the adequacy of the tested model, which have been shown to be more accurate at rejecting misspecified models (for review see, Martens, 2005): the Comparative Fit Index (CFI), the Incremental Fit Index (IFI), the Non-Normed Fit Index (NNFI), and the Root Mean Square Error of Approximation (RMSEA).

Results and Conclusions

Model parameters were estimated based on the covariance matrix and using the robust method, because examination of the descriptive statistics revealed small deviations from univariate normality for some of the items (kurtosis greater than 2.0). The results provided adequate support for the eight-factor model (Satorra and Bentler's χ2/ df = 1451.76 / 712, CFI = .90, IFI = .90, NNFI = .89, RMSEA = .04), but not for the two-factor model (Satorra and Bentler's χ2/ df = 2102.89 / 739, CFI = .82, IFI = .82, NNFI = .81, RMSEA = .05, Satorra-Bentler Scaled Difference between the eight and the two factor model = 545.01, df = 27, p < .001) and the ten-factor model (Satorra and Bentler's χ2/ df = 1649.32 / 731, CFI = .88, IFI = .88, NNFI = .87, RMSEA = .05, Satorra-Bentler Scaled Difference between the eight and the ten factor model = 200.20, df = 19, p < .001). Factor loadings for the eight-factor model ranged from .46 to .77 and for the two-factor model ranged from .28 to .73. The correlation between the two factors was -.54. The results provided supportive evidence for the construct validity of the ASTQS in PE.

To further assess the construct validity of ASTQS, eight dependent t-tests with Bonferroni adjustment were performed to examine significant differences between the self-talk dimensions in the two situations (playing with the best student versus playing with the worst student). The results were in accordance with hypothesis (a). Significant differences were observed for the negative self-talk dimensions revealing that students experience higher negative self-talk referring to worry, disengagement, somatic fatigue and irrelevant thoughts when they play against the best student than when they play against the worst student. Non-significant differences were observed for positive self-talk referring to instruction and anxiety control, but significant differences were observed for self-talk referring to psych up and confidence showing that students exhibit more psych up and less confidence statements when they play against the best student than when they play against the worst student.

In sum, the results of the eight-factor model showed good internal consistency except from the irrelevant thoughts factor. Similar result was also found in in Zourbanos et al.'s (2010) study. In general, ASTQS in PE showed factorial and structure validity, as well as discriminant validity, indicating that it can be an effective measure for examining students' thoughts during the physical education lesson.

Means, standard deviations, Cronbach's alpha coefficients and correlations between goal orientations, perceived competence and students' self-talk dimensions are reported in Table 1. In general, in accordance with hypothesis (b) correlation analyses revealed moderate relationships between task goal and students' positive self-talk dimensions. The results were in accordance with hypothesis (c) showing low positive relationships between ego goal and positive self-talk dimensions and non-significant relationships between ego and negative self-talk dimensions. Finally in accordance with hypothesis (d) the results revealed perceived competence had positive relationship with positive self-talk dimensions and negative relationship with negative self-talk dimensions.

Study 2

In the last decade Elliot's achievement goals model (e.g., Elliot & Harackiewitz, 1996; Elliot & Murayama, 2008 attracted the attention of several investigators in sport and exercise psychology (for reviews see Roberts et al., 2007; Papaioannou et al., 2012). The split of task and goals into approach and avoidance seems quite appealing and therefore we decided to re-examine the findings of the first study but this time by using the 2x2 framework and more specifically the Achievement Goal Questionnaire-Revised (AGQ-R; Elliot & Murayama, 2008). The purpose of the second study was twofold. Firstly to examine the reliability and construct validity of the AGQ-R in physical education and in the Greek language and secondly to investigate the relationship of achievement goals and self-talk as they are measured with the AGQ-R and the ASTQS . Following the same reasoning with that in Study 1 and based on meta-analysis of findings concerning the relationship of various cognitions with mastery and performance approach and avoidance goals in sport and physical education (Papaioannou et al., 2012) it was hypothesized that (a) mastery approach goal and (b) performance approach goal would be positively related to students' positive self-talk dimensions and negatively to students' negative self-talk dimensions, and (c) mastery avoidance goal and (d) performance avoidance goal would be negatively related to students' positive self-talk dimensions and positively to students' negative self-talk dimensions.

Method

Participants and Procedure

Participants were 313 students (151 females and 162 males) aged 12 (SD = .48) years old who attended physical education classes (elementary school) located in a city in central Greece. Similar procedures with study 1 were followed.

Measures

Self-Talk in PE. We distributed the ASTQ-PE that was used in Study 1. However, here we examined only self-talk during play with the most competent student. The ASTQ-PE demonstrated acceptable internal consistency (apart from irrelevant thoughts' factor, see Table 2).

Achievement goals in PE. Following the stem "In the Physical Education class…" students responded to the items of the Achievement Goal Questionnaire-Revised (AGQ-R; Elliot & Murayama, 2008) was used. Mastery approach items include, "My aim is to completely master the material presented in this class", Performance approach items include, "My aim is to perform well relative to other students", Mastery avoidance item include, "My aim is to avoid learning less than I possibly could", and Performance avoidance items include, "My aim is to avoid doing worse than other students". Ratings are made on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). The alpha coefficients in the present study are reported in Table 2). The questionnaire was translated from the English to the Greek language and back. More specifically, a bilingual translator familiar with the theoretical concepts of achievement goals translated the instrument. A back-translation was carried out by two translators. One of them completed a "blind" back-translation, which means that the blind translator was not familiar with the theoretical concepts of achievement goals, in contrast to the other translator who was. Discrepancies were finally decided by the three translators.

Perceived competence. To assess students' perceived competence we used the same measure as in study 1. Cronbach's alpha coefficient for the perceived competence is displayed in Table 2.

Analysis

The factor structure of the achievement goals were tested through CFA using the EQS 6.1 (Bentler & Wu, 2004). Three alternative models were tested. In Model 1 (M1) the original four-factor structure was tested, with the four factors allowed to correlate. Model 2 (M2) was a trichotomous model, in which the performance-approach and performance-avoidance items loaded on their respective latent factors, and the mastery-approach and mastery-avoidance items loaded together on a third latent factor. Model 3 (M3) was also a trichotomous model in which the mastery approach and performance approach items loaded on their respective latent factors, and the mastery-avoidance and performance-avoidance items loaded together on a third latent factor. Finally, the factor structure of ASTQS in PE was re-tested. Four fit indices as in Study 1 were used to assess the adequacy of the tested model.

Results and Conclusion

Confirmatory factor analyses

Model parameters were estimated based on the covariance matrix and using the robust method, because examination of the descriptive statistics revealed moderate deviations from univariate normality for some of the items (kurtosis greater than 2.49). To examine whether the chi square values differed significantly between M1 and M2 and between M1 and M3, the Satorra and Bentler's scaled difference qui-square test was conducted (Satorra & Bentler, 2001; Crawford & Henry, 2003). The results provided adequate support for the four-factor model (M1) (Satorra and Bentler's χ2/ df = 68.80 / 48, CFI = .97, IFI = .97, NNFI = .96, RMSEA = .04), which was superior to the trichotomous model (M2) (Satorra and Bentler's χ2/ df = 92.24 / 51, CFI = .93, NNFI = 91, IFI = .93, RMSEA = .05, Satorra-Bentler Scaled Difference = 20.16, df = 3, p < .001) and the trichotomous model (M3) (Satorra and Bentler's χ2/ df = 232.44 / 53, CFI = .70, NNFI = .62, IFI = .70, RMSEA = .10, Satorra-Bentler Scaled Difference = 112.46, df = 5, p < .001). Factor loadings for the four-factor (M1) ranged from .43 to .73. In conclusion the results provided adequate support for the factor structure of the AGQ-R in the PE settings.

Correlation analyses

Means, standard deviations, Cronbach's alpha coefficients and correlations between achievement goals, perceived competence and students' self-talk dimensions are reported in Table 2. In general, the results were in accordance with hypotheses (a) and (b) revealing low positive relationships of mastery and performance approach goals with students' positive self-talk dimensions and negative with students' negative self-talk. In order to compare the relationships between task goal and positive self-talk and mastery goal and positive self-talk, we used the Fisher r-to-z transformation for independent rs. If ra (relationships between task goal and positive self-talk dimensions) is greater than rb (relationships between mastery goal and positive self-talk dimensions), the resulting value of z will have a positive sign. The results showed that the relationships between mastery goal and self-talk were of a lower magnitude in comparison to the relationships between task goal and self-talk (z scores ranged between 3.25 and 3.73, p < .001, for all the positive self-talk dimensions). Furthermore, the results were not in accordance with hypothesis (c) and (d) revealing non-significant relationships between mastery and performance avoidance and negative self-talk dimensions. Finally, the results were in accordance with Study 1 findings revealing that perceived competence had moderate positive correlations with athletes' positive self-talk dimensions, whereas perceived competence had moderate negative correlations with athletes' negative self-talk dimensions. Furthermore, mastery approach and performance avoidance had low positive relationships with perceived competence, performance approach had moderate positive relationship with perceived competence and mastery avoidance had non-significant relationship.

In general the results of Study 2 re-confirmed the relationships between achievement goals, self-talk and perceived competence. We suspected that the weaker correlations of mastery approach goals with self-talk were due to different conceptualization and measurement of achievement goals between Elliot's and Nicholls' models (Papaioannou et al., 2012). In order to investigate it further we decided to use both dichotomous and 2x2 models and to investigate whether perceived competence moderates the relationship of ego goal with self-talk as it is outlined in the dichotomous model but not in the 2X2 model.

Study 3

Roberts et al. (2007) consider competence as the "energizing construct of the motivational processes of achievement goal theory" (p. 4). In the dichotomous model (Nicholls, 1984), task goals were assumed to lead to adaptive thoughts and behaviour in achievement situations irrespective of an individual's perceived competence. In contrast, the effects of ego goals were postulated to depend on perceived competence, that is, ego goals were suggested to lead to adaptive patterns of behavior when perceived competence is high, whereas in cases where perceived competence is low, ego goals were assumed to lead to maladaptive behavior, when performance difficulties are encountered. Nicholls' (1984, 1989) in his original work, described a process by which ego-involved individuals who encounter performance difficulties gradually lose confidence in their abilities and if they are unable to perform well eventually will abandon the goal of doing better than others and adopt the goal of avoiding a display of incompetence. On the contrary, individuals who do not have doubts about their ability should not experience this decline in perceived ability and consequent abandonment of the goal of performing better than others.

In an experimental study using a complex task, Elliot and Dweck (1988) found that mastery goals displayed adaptive learning patterns regardless of perceived competence. In contrast, performance goals exhibited adaptive patterns with high perceived competence, whereas with low perceived competence performance goals exhibited maladaptive patterns. More specifically noteworthy is that in the performance approach-low ability group their self-statements were negative such as "I'm not very good at this". In another study Kaplan and Midgley (1997) found that perceived competence functioned as a moderator in the relationship between mastery goals and learning outcomes, but not in the relationship between performance goals and learning outcomes. However, Elliot and Church (1997) with the revised social-cognitive model of achievement motivation viewed competence as an antecedent and not as a moderator. That is, higher levels of competence are associated with the performance approach and mastery approach goals. Furthermore, performance avoidance goals are predicted to develop for those who are low in perceived competence. In some cases competence has been viewed as a mediator as well (e.g., Cury, Elliot, Sarazzin, Fonseca, & Rufo, 2002). Based on the inconsistencies in the literature and suggestions about important differences in the theorization between the dichotomous and 2x2 models which are reflected in the measurement of achievement goals, the purpose of the third study was to explore the moderating effects of perceived competence on the relationship between achievement goals and positive and negative self-talk. Based on the original work by Dweck and Nicholls, participants were told to imagine that they were playing against the most competent student in their age in this particular sport, game or activity in order to put themselves into a situation with performance difficulties. Specifically we expected, for the dichotomous model, that low perceived competence and ego orientation would positively predict negative self-talk, whereas with high perceived competence no relationship between ego orientation and negative self-talk would occur. For the 2x2 model we expected that performance approach goal and perceived competence would have additive but not interactive effects. Furthermore, we hypothesized that low perceived competence and mastery approach goal would positively predict positive self-talk, whereas with high perceived competence no relationship between mastery approach goal and positive self-talk would occur. Finally, students with low perceived competence and performance/mastery avoidance goal would positively predict negative self-talk, whereas for students with high perceived competence performance/mastery avoidance goal would negatively predict negative self-talk.

Method

Participants and Procedure

Participants were 1169 students with a mean age of 13.34 years (SD = 1.30) who attended physical education classes located in a city in central Greece. Similar procedures with studies 1 and 2 were followed.

Measures

Self-Talk in PE. To assess students' self-talk we used the ASTQ-PE that was used in study 1 and 2. The ASTQ-PE demonstrated good internal consistency ranging from .70 to .81 for the negative self-talk dimensions except from athletes' irrelevant thoughts, which demonstrated low internal consistency .56. For the purposes of the present study and based on the low internal consistency that was also emerged in study 1 and 2 we didn't use the irrelevant thoughts factor. Finally, for the positive self-talk dimensions internal consistencies ranged from .82 to .83 except from athletes' anxiety control, which demonstrated lower (.62), but still acceptable internal consistency.

Achievement goals in PE. To assess students' achievement goals we used the TEOPEQ as in study 1 and the AGQ-R as in study 2. Both measures demonstrated acceptable internal consistencies ranging for AGQ-R from .67 to .72 and for the task and the ego orientations (TEOPEQ) were .78 and .84 respectively.

Perceived competence. To assess students' perceived competence we used the same measure as in study 1 and 2. Cronbach's alpha coefficient for the perceived competence subscale was .75

Results and Conclusions

Preliminary results

Means, standard deviations, and correlations between achievement goals, perceived competence and students' self-talk dimensions are reported in Table 3. The relationships of the third study were similar with the results of study 1 and 2.

Multi-group path analysis

One of the ways to test for moderator effects is through multiple-group path analysis (EQS; Bentler & Wu, 2004). To examine the hypothesized relationships between achievement goals and students' self-talk structural models with latent factors were tested (for dichotomous model, see Figure 1 and for 2x2 model, see Figure 2). In this study ASTQS was represented by a second order 10-factor model suggesting that the eight factors assess different self-talk dimensions, which however represents two broader positive and negative dimensions (for details see Zourbanos et al., 2009). The sample was dived into groups (high: n = 424, and low perceived competence: n = 375, top and bottom 33 percent of the sample). The multiple-group analysis allows investigation of moderation hypotheses through examination of path differences in identical models tested in different conditions. In the model specification the investigated paths are constrained to be equal and the analysis reveals whether the constraints should be released or not, that is whether the paths are statistically different or not. In addition, where indications of non-normality were identified, the Satorra and Bentler's χ2 and the respective robust fit indices, which have been shown to work well under non-normal distributions, were considered.

Dichotomous model

For the dichotomous model and the high perceived competence group (see Figure 1 for standardized coefficients, regular characters) task orientation was positively related to positive self-talk and negatively related to negative self-talk, ego orientation was neither related to positive self-talk nor to negative self-talk. For the low perceived competence group (see Figure 1 for standardized coefficients, bold characters) task orientation was once more positively related to positive self-talk and negatively related to negative self-talk, whereas ego orientation wasn't related to positive self-talk but was positively related to negative self-talk. Multi-sample analysis was calculated to test the moderation hypothesis. The paths between task orientation/ positive self-talk and negative self-talk and ego orientation/negative self-talk and positive self-talk were constrained to be equal for the high perceived competence and the low perceived competence groups, that is, it was hypothesized that for the two groups these paths coefficients were not statistically different. The LM test indicated that the constraint for the path between ego orientation and negative self-talk was significantly different (p < .05) for the two groups. More specifically students with low perceived competence and ego orientation were positively related to experiencing negative self-talk, whereas for students with high perceived competence no relationship between ego orientations and negative self-talk was shown. The fit indices for all models are displayed in Table 4.

2x2 model

Similar procedure was followed for the 2x2 model. For the high perceived competence group (see Figure 2 for standardized coefficients, regular characters) mastery approach orientation was neither related to positive self-talk nor to negative self-talk. Furthermore, performance approach orientation was neither related to positive self-talk nor to negative self-talk. Mastery avoidance orientation was neither related to positive nor to negative self-talk, and performance avoidance orientation was neither related to positive nor to negative self-talk. For the low perceived competence (see Figure 2 for standardized coefficients, bold characters) mastery approach orientation was positively related to positive self-talk and negatively related to negative self-talk, performance approach orientation was positively related to positive self-talk and was unrelated to negative self-talk. Mastery avoidance orientation was neither related to positive nor to negative self-talk and performance avoidance orientation was neither related to positive nor to negative self-talk. The paths between mastery approach orientation/positive and negative self-talk were constrained to be equal for the high perceived competence and the low perceived competence groups, that is, it was hypothesized that for the two groups these paths coefficients were not statistically different. The LM test indicated that the constraints for the paths between mastery approach and positive self-talk were significantly different (p < .05) for the two groups. More specifically, students with low perceived competence and mastery approach goal were positively related to experiencing positive self-talk, whereas students with high perceived competence no relationship between mastery and performance approach goal and positive self-talk was found. The fit indices for all models are displayed in Table 4.

General Discussion

The purpose of the study was threefold. First, to validate Zourbanos et al.'s (2009) questionnaire of self-talk in PE and Elliot and Murayama's (2008) 2x2 achievement goal questionnaire in the Greek language, second to explore using the dichotomous and the 2x2 achievement goal frameworks the relationships between achievement goals, perceived competence and students' self-talk dimensions, and third, using multi-group path analysis to investigate the moderating role of perceived competence on the relationship between achievement goals and students' self-talk. To the best of our knowledge, this is the first examination of the combined effects of achievement goals using both theoretical frameworks and perceived competence on positive and negative self-talk. Overall, the results provided support for the psychometric integrity of the ASTQS and AGQ-R in PE, and showed through path analysis the interaction between achievement goals and perceived competence on students' positive and negative self-talk.

Construct validity and reliability of ASTQS and AGQ-R

Zourbanos et al.'s (2009) scale of athletes' self-talk (ASTQS) has demonstrated evidence of construct validity and reliability (Zourbanos et al., 2010; Zourbanos et al., 2011) only in sport settings. In this study, following minor wording revisions, confirmatory factor analysis provided evidence of factorial validity in physical education and reliability analysis supported the internal consistency of the scale. However, one factor from ASTQS (irrelevant thoughts) demonstrated low internal consistency. Furthermore the results provided support of construct validity showing that students exhibit more negative self-talk when they play with students of a higher ability than when they play with students of a lower ability. Regarding the Greek version of AGQ-R (Elliot & Murayama, 2008) the results provided evidence for the factorial validity of the instrument in physical education. Furthermore, reliability analysis provided evidence for the internal consistency of the scale. However, one factor from AGQ-R (mastery avoidance) demonstrated low internal consistency. Finally correlation analysis provided evidence of concurrent validity showing that mastery approach was related to task goal and performance approach to ego goal. Nevertheless, the correlations were of a lower magnitude than we expected showing that AQG-R follows a different conceptualization of achievement goals (for review see Papaioannou et al. 2012). Overall, the results provided supportive evidence for the integrity of the instruments, however further research could examine the validity of the questionnaires using alternative research paradigms and new approaches in survey designs (Gehlbach & Brinkworth, 2011).

Achievement goals, perceived competence and students' self-talk

In Study 1, in which the dichotomous framework was adopted, task orientation was associated with positive thoughts related to psych up, confidence, instructions and anxiety control. Nicholls (1984) proposed that when individuals are task oriented, adaptive cognitive patterns are likely to emerge. Moreover task orientation was negatively related to negative self-talk indicating that increasing levels of task orientation were associated with lower worry, disengagement, somatic fatigue and irrelevant thoughts. Similar results have been also obtained for cognitive anxiety in the form of worry (e.g., White & Zellner, 1996). On the other hand ego orientation was unrelated to negative self-talk dimensions. Similar results, have been reported in the literature (e.g., Biddle et al., 2003) showing that ego orientation was not related to negative outcomes such as anxiety, feelings of pressure and worry.

Given the equivocal results between ego orientation and negative self-talk dimensions the aim of the second study was to re-examine the above relationships using the 2x2 framework. In general the results of the Study 2 were similar with Study 1 findings but as was expected the relationship of mastery approach goals with self-talk was of a lower magnitude than between task goal and self-talk. More specifically mastery and performance approach were positively related to students' positive self-talk dimensions and negatively related to negative self-talk. Similar results were reported in a recent meta-analysis (for more details see Papaioannou et al. 2012) which revealed that mastery and performance approach goals were related to several adaptive outcomes such as self-esteem, effort, positive affect and perceived competence. In study 3 performance avoidance goals were positively related to positive self-talk dimensions and were unrelated to negative self-talk and that mastery avoidance goals were unrelated to positive and negative self-talk dimensions. Urdan and Mestas (2006) found that participants had difficulty to distinguish between performance and mastery avoidance goals and that participants used approach explanations when they responded to performance avoidance items. Recently, Ciani and Sheldon (2010) found that athletes who endorsed mastery avoidance goals used mastery approach explanations. These findings suggest that children, have difficulty to understand the items of performance or mastery avoidance scales, which might explain the non-significant relationships between mastery avoidance goals and self-talk.

In general the moderating role of perceived competence on the relationships between ego goals and cognitive outcomes has not been always supported (for review see Roberts et al., 2007; Harwood et al., 2008). Based on the original work by Dweck and Nicholls, participants were told to imagine that they were playing against the most competent student in their age in this particular sport, game or activity in order to put themselves into a situation with performance difficulties. We sought to address this gap using both the dichotomous and the 2x2 framework. For students with low perceived competence ego orientation was positively related to experiencing negative self-talk, whereas for students with high perceived competence no relationship between ego orientations and negative self-talk was shown. Nicholls' (1984) moderating hypothesis was supported when the concept of success was used to capture an ego goal (Duda & Nicholls, 1992) but not when the definition of competence was used to capture a performance approach goal (Elliot & Murayama, 2008). Likewise, Papaioannou et al. (2012) reported findings supporting the moderating role of perceived competence between ego orientation and self-efficacy for normative performance when achievement goal measures were based on Nicholls' but not on Elliot's approach.

These findings are in line with our predictions. Elliot's measures do not capture different conceptions of success but different definitions of standards used to evaluate competence. Elliot and Murayama (2008) purposely separated "aim from reason" in their measurement of definition of competence. However, for Nicholls (1984; 1989) (a) the definition of competence was largely a by-product of the definition of success, (b) the association of aim with reason is central to individuals' decision to adopt different definitions of success and corresponding goals and motivational outcomes, (c) individual differences in achievement goal adoption reflect different world views that shape different definitions of success (Nicholls, Patashnick & Nolen, 1985) and trigger different reasons for striving to accomplish something important (Papaioannou et al., 2012). One should have an important reason to select consistently subjective or normative criteria of evaluation across situations. Indeed, personality psychologists suggest that reasons/values are connected with goals when individuals show consistent behavior across situations (Mischel & Shoda, 1998). Accordingly, Nicholls' measures seem more effective to assess dispositional differences in achievement goal adoption across situations, such as sport and school (e.g., Duda & Nicholls, 1992) than Elliot's which are more situation-specific (Papaioannou et al., 2012). In the current study ego goals were assumed to reflect dispositional differences, and therefore the support of the moderating hypothesis in in line with the original dichotomous model which conceptualized achievement goals as something relatively stable across situations, while the rejection of the moderating hypothesis is in line with Elliot's framework that considers achievement goals as situation-specific variables.

To exemplify this we can consider a low perceived athletic ability student who plays against the best student in the particular sport. If the student is high ego-oriented he would like to establish his superiority in the game because this is consistent with his notion of success in life which is to establish superiority. Because this is too difficult for him he is worried that he will not achieve this important goal for him. If the student has a high performance approach goal he wants to outperform others but his self-worth is not threatened because it is not necessarily tied to the expected outcome. In a competitive game almost all children want to win; those students who are quick to comply with the demands of the competitive situation (i.e., pursue performance approach goals) might consider win a temporary goal but do not necessarily an important goal for him as a person.

In accordance with previous research in sport (Hatzigeorgiadis and Biddle, 1999) and with Nicholls' (1984) predictions, it was also found that task orientation was negatively related to negative self-talk irrespective of perceptions of competence. It is important to note that task orientation was positively related to positive self-talk irrespective of perceptions of competence. When task oriented students play with high achievers they experience positive thoughts irrespective of their level of abilities because a challenging game offers them opportunities to improve and enjoy their involvement which is the definition of success for them in physical education. Regarding the 2x2 framework the results were different. It was revealed that for students with low perceived competence mastery approach goals were positively related to positive self-talk, whereas for students with high perceived competence no relationship between mastery approach goals and positive self-talk was shown. This finding suggests that individuals with low perceived competence think positively when they adopt mastery approach goals because they have positive expectancies to reach their goals since their achievement depends on themselves. Therefore, when they engage in PE activities tend to experience positive thoughts. In other words, mastery approach goal acts as a buffer of low perceived competence. On the other hand individuals with high perceived competence already have positive expectancies to reach their goals and therefore mastery approach goals that do not set specific criteria for self-improvement (i.e., do your best goals) do not increase further their challenge and do not add something to their positive thoughts that already exist due to high competence.

Main effects were found for performance approach goals which were positively related to positive self-talk irrespective of perceptions of competence and non-significant relationships were observed between mastery and performance avoidance goals and self-talk which were similar to the results of study 1 and 2.

The practical importance of this study is clear. Teachers and parents are encouraged to promote task-involving and mastery approach goals because this approach benefits both the person and society (Papaioannou et al., 2012). Our results showed that the adoption of task and mastery approach goals had the most adaptive patterns for students' positive self-talk. Although the adoption of performance approach goals was not maladaptive for students' positive self-talk, we don't encourage teachers' emphasis on performance approach goals because it might be perceived as controlling and sometimes stressful. As elite athletes said, every athlete knows that she has to win but when the coach tells them to win they feel stress (Kristiansen & Roberts, 2010). Moreover, a high evaluative, threatening and unsupportive climate might trigger avoidance goals and negative self-talk. On the other hand, teachers who help students to use positive self-talk to enjoy physical education and improve their performance might create a task-involving climate and boost students' self-esteem (Milosis & Papaioannou, 2007). The use of ASTQS items might be important to practitioners who want to investigate whether their teaching promotes positive thinking in their classes.

Limitations and future research

Given the exploratory nature of the present investigation, there are a number of limitations that further research should address. Firstly, it is important to notice that no causal link can be inferred from the present findings. That achievement goals may influence cognitive outcomes can be speculated based on theoretical grounds of motivation and models of self-talk antecedents, however, it is plausible that the identified links reflect bidirectional relationships. For example self-talk contains directions to self about doing, to do or not doing something which involve goal-directed actions. Furthermore, self-talk could play a mediating role between achievement goal theory and other cognitive, affective or behavioral responses. Also experimental research providing ego or task involving feedback could give us a deeper understanding on the relationship between achievement goals and self-talk. Nevertheless, considering that no previous research has examined this kind of relationships in PE, the present findings provide valuable evidence regarding the additive and interactive effects of achievement goals, perceived competence and students' self-talk. An issue that should be addressed is the retrospective verbal self-reports. Cognitive processes cannot be accurately assessed through external measures and the use of self-reports provide us with 'metacognitive knowledge' which can help us understand perceptions, motives, and generally what someone is thinking (Guerrero, 2005). Although ASTQS in PE was based on the established ASTQS, further validation in PE using discriminant validity is required to establish its psychometric properties. In summary, the present findings suggest that both students' achievement goals and perceived competence may impact students' self-talk. Taking into account the significant role of thoughts on performance, this line of research may contribute to subsequent explorations into the antecedents of students' self-talk.