Psychology of Sport and Exercise

Transcription

1 Psychology of Sport and Exercise 13 (01) 491e499 Contents lists available at SciVerse ScienceDirect Psychology of Sport and Exercise journal homepage: Goal orientations, perceived motivational climate, and motivational outcomes in football: A comparison between training and competition contexts Pepijn K.C. van de Pol *, Maria Kavussanu, Christopher Ring School of Sport and Exercise Sciences, The University of Birmingham, Birmingham B15 TT, UK article info abstract Article history: Received 30 September 010 Received in revised form 13 December 011 Accepted 14 December 011 Available online January 01 Keywords: Goal orientations Motivational climate Football Objectives: In this study, we examined (a) variability and (b) differences in football players goal orientations and perceptions of the motivational climate across training and competition contexts, and (c) whether the context moderates the relationships between goal orientations, motivational climate, and effort, enjoyment and tension. Design: Cross-sectional. Method: Football players (36 males and 48 females) completed questionnaires measuring goal orientations, perceived motivational climate, effort, enjoyment, and tension in training and competition. Results: Multilevel analysis revealed that the context explained significant variation in goals and climate perceptions beyond the athlete and team levels: Participants reported higher ego orientation and perceptions of performance climate in competition than in training, whereas task orientation and perceived mastery climate did not differ between the two contexts. Effort and enjoyment were both positively related to task orientation and unrelated to ego orientation in both contexts. Effort was positively associated with perceived mastery climate only in training, while enjoyment was positively associated with mastery climate in both contexts but more strongly in training than in competition. Effort was negatively associated with performance climate in both contexts. Tension was unrelated to task orientation. It was also inversely associated with ego orientation in training and unrelated to this goal in competition. Tension was unrelated to perceived mastery climate and positively related to performance climate but only in training. Conclusion: The findings highlight the importance of distinguishing between the training and competition contexts when examining achievement motivation in sport. Ó 011 Elsevier Ltd. All rights reserved. In the last two decades, achievement goal theory (e.g., Ames, 199; Nicholls, 1989) has been the main theory used to understand achievement motivation in sport. According to this theory, individuals engage in achievement contexts in order to develop or demonstrate competence (Nicholls, 1989). However, competence or ability can be evaluated using self or other-referenced criteria; thus, two conceptions of ability operate in achievement contexts (Nicholls, 1989). These two conceptions of ability are embedded within two distinct achievement goals, namely task and ego involvement (Nicholls, 1989). When individuals are task involved, they evaluate competence using self-referenced criteria and feel successful when they learn something new, master a task, or improve their skills. In contrast, when individuals are ego involved, they evaluate competence using other-referenced criteria and feel successful when they establish superiority over others (Nicholls, 1989). * Corresponding author. address: pkcvandepol@hotmail.com (P.K.C. van de Pol). The terms task and ego involvement refer to the goal states that people experience in a given situation. People also have a proneness to the two types of involvement known as task and ego orientation (Nicholls, 1989). Task and ego orientations have been found to be relatively orthogonal (e.g., Roberts, Treasure, & Kavussanu, 1996). Thus, individuals can be high or low on either or both goals, and research has shown that the two goals can interact with each other in predicting various outcomes (e.g., Sage, Kavussanu, & Duda, 006; van de Pol & Kavussanu, 011a, 011b). To date, the vast majority of sport research stemming from the achievement goal framework has been conducted in the broad domain of sport. However, this domain can be subdivided into training and competition contexts, which may affect athletes goal orientations in each context (Harwood, 00). Previous research has found moderate-to-large correlations between training and competition goals in tennis players (see van de Pol & Kavussanu, 011a), which indicate that there is some consistency but at the same time a sufficient degree of variability between training and competition goals that merit separate examination. The contexts of /$ e see front matter Ó 011 Elsevier Ltd. All rights reserved. doi: /j.psychsport

2 49 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e499 training and competition entail conditions that could differentially influence task and ego involvement: As an organised structure, training typically provides opportunities for athletes to practise and develop their skills, whereas organised competition is formally regulated so that athletes test these skills against those of other athletes. Normative comparison may take place in training; however, it is inherent in competition, because in this context outperforming other athletes is objectively rewarded (e.g., normative rankings based on win/loss records). Hence, the structural characteristics of training and competition may affect the extent to which athletes are task or ego involved in each context. As a result, athletes may develop the tendency to evaluate success that is specific to each context and to be more ego and less task oriented in competition than in training. The distinction between training and competition contexts has received some attention in the literature with studies examining differences between the two contexts in goal involvement and goal orientation. In a study of goal involvement, female softball players were more task involved before a training session than before a competitive game, but did not differ in ego involvement (Williams, 1998). With regard to goal orientation, tennis players reported higher task and lower ego orientation in training than in competition (van de Pol & Kavussanu, 011a), while athletes from a variety of individual and team sports reported higher ego orientation in competition than in training but did not differ in task orientation (van de Pol & Kavussanu, 011b). Thus, there is value in distinguishing between training and competition contexts when one examines achievement goals in sport. A second variable that could vary between training and competition is the motivational climate, which refers to the situational goal structure operating in an achievement context (Ames, 199). The motivational climate has been distinguished 1 in mastery climate where the emphasis is on effort, personal improvement, and skill development, and performance climate where the emphasis is on normative comparison and public evaluation (Ames, 199). The motivational climate is created by significant others such as teachers, parents and coaches (Ames, 199). In this study, we use the term motivational climate to refer to the coach-created climate and in line with past work (e.g., Newton, Duda, & Yin, 000) wehave examined motivational climate via athlete perceptions. In sport, perceptions of a mastery climate have been associated with adaptive motivational patterns, whereas perceptions of a performance climate have been linked to less adaptive or maladaptive motivational responses (see for reviews: Harwood, Spray, & Keegan, 008; Ntoumanis & Biddle, 1999). To date, no study has investigated whether perceived motivational climate varies between the training and competition contexts. The win/loss records and ranking systems that are salient in competition and the presence of spectators, who often reinforce normative success, may lead coaches to put more emphasis on normative criteria in that context, thereby creating a stronger performance climate in competition than in training. Mastery climate may not vary between the two contexts because coaches should reward effort and encourage personal improvement in both contexts. This is because effort is beneficial for both skill development that occurs in training and high level performance for which athletes strive in competition. Thus, differences between the two contexts may exist in performance but not mastery motivational climate. 1 Although the terms task and ego-involving climate are also used in the sport literature to refer to mastery and performance climate respectively, in this study we use the terms mastery and performance to be consistent with Ames (199) original terminology. The distinction between training and competition contexts may also have implications for the relationships between goal orientations and effort, enjoyment, and tension. These outcomes are important to examine in both contexts because, as indicators of intrinsic motivation (Ryan & Deci, 000), they have implications for the quality of the athletic experience as well as for athletes performance. Ego orientation has been typically unrelated to effort and enjoyment and positively linked to tension in sport in some studies (see Biddle, Wang, Kavussanu, & Spray, 003; Harwood et al., 008); however, its relationship with these outcomes may differ depending on the context. In training, ego orientation should be unrelated to effort, enjoyment and tension. Although egooriented athletes may also apply effort during training, for example during competitive training drills/games, in general, the training context does not provide the ideal environment that would mobilise effort in ego-oriented athletes because it focuses on skill development rather than normative success. Because in training normative success is not formally rewarded, high ego-oriented athletes may not typically enjoy training or feel pressure as they may perceive a lack of challenge to demonstrate normative competence in this context. In competition, ego-oriented athletes should apply effort because this is the ideal context for these athletes to demonstrate normative competence (Harwood & Hardy, 001). Ego-oriented athletes should derive positive affect from normative success (Nicholls, 1989); therefore, their enjoyment in competition should also depend on such success. The normative success criteria embedded in competition could make high egooriented athletes worried about receiving an approving evaluation leading them to experience tension in this context. The relationship between task orientation and effort, enjoyment and tension should not vary between the two contexts. In sport, task orientation has been associated with greater effort and enjoyment and lower tension (e.g., Biddle et al., 003). The training context is the ideal environment for task-oriented athletes e who have an intrinsic interest and desire to improve (Nicholls, 1989) e to feel successful, as in this context they have the opportunity to practise and develop their skills. is also a context in which task-oriented athletes should exert effort and experience enjoyment in their pursuit of their self-referenced goals (e.g., personal bests). These athletes evaluate their competence in relation to their own previous performance, and this should reduce potential tension in both contexts. Previous research which examined whether the training vs. competition context moderates the relationship between goal orientations and motivational outcomes found that in tennis players, task orientation predicted effort and enjoyment positively in both contexts, while ego orientation predicted effort positively only in competition and only when task orientation was low or average (van de Pol & Kavussanu, 011a). Interestingly, task orientation was a stronger positive predictor of enjoyment in competition than in training. In another study, which examined athletes from a variety of sports, ego orientation was positively related to enjoyment only in competition (van de Pol & Kavussanu, 011b). Overall, these findings suggest that the context may moderate the relationship between goal orientations and various outcomes in sport, and that it is important to examine whether task and ego orientations interact with each other in predicting outcomes across the two contexts. The context may also moderate the relationships between perceived motivational climate and effort, enjoyment, and tension. In both training and competition, mastery climate should predict effort and enjoyment positively and tension negatively, whereas performance climate should predict these outcomes in the opposite direction (cf. Harwood et al., 008; Ntoumanis & Biddle, 1999). However, due to the normative success being valued in

3 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e competition, performance climate may be more prominent in this context compared to training, which should strengthen the negative impact of this climate on motivational outcomes in competition. Thus, perceived performance climate should have negative effects on motivational outcomes in training and competition, but stronger effects in the latter context. Finally, goal orientations and perceived climates may interact with each other in predicting effort, enjoyment and tension. In a physical education context, high task-oriented students had higher intrinsic motivation when they perceived a strong mastery climate but lower intrinsic motivation when they perceived a weak mastery climate (Standage, Duda, & Ntoumanis, 003). However, no interaction effects were found between task and ego orientations and perceived climate on effort, enjoyment, and tension in a study of female volleyball players (Newton & Duda, 1999). Therefore, research is needed to examine interaction effects between goals and motivational climate on effort, enjoyment, and tension. Making the distinction between training and competition may reveal such interactions. The present study In sum, the distinction between training and competition contexts when examining goal orientations, motivational climate, and motivational outcomes in sport has received little attention. The present study was designed to address this gap in the literature in the sport of football. We focused on football because it is the most popular (team) sport in the UK (Sport England, 009), thus, the findings have implications for many individuals. Our study had three purposes: First, to examine variability in football players goal orientations and perceived motivational climate across training and competition contexts. We expected that both task and ego orientations would show significant withinsubjects variance across the two contexts (van de Pol & Kavussanu, 011a). Due to the lack of empirical evidence, we made no predictions regarding the variability in perceived climate as a function of the context. The second study purpose was to examine whether goals and perceived motivational climate differ across training and competition. Based on previous research (van de Pol & Kavussanu, 011a, 011b), we hypothesised that football players would report higher ego orientation in competition than in training. However, due to inconsistent previous findings (van de Pol & Kavussanu, 011a, 011b; Williams, 1998) we made no predictions for context differences in task orientation. We also expected higher performance climate in competition than in training but no difference in mastery climate. The third purpose was to examine whether the context moderates the relationships between goal orientations and motivational climate and effort, enjoyment and tension. We hypothesised that task orientation would predict effort and enjoyment positively and tension negatively in both contexts, and that ego orientation would be unrelated to effort and tension in training and predict effort and tension positively in competition (Biddle et al., 003; van de Pol & Kavussanu, 011a, 011b). We made no predictions regarding ego orientation and enjoyment in competition. We expected that mastery climate would predict effort and enjoyment positively and tension negatively in both contexts, while the reverse relationships were expected for performance climate (Ntoumanis & Biddle, 1999). No hypotheses were formed with regard to potential interactions between goals and climates and outcomes in each context as previous research has reported mixed findings on this issue (Newton & Duda, 1999; Standage et al., 003). To examine our study purposes we used Multilevel Linear Modelling (MLM). This technique recognises that data of athletes who are members of a team and respond with respect to training and competition have a nested structure. This is important because athletes within teams share experiences that may affect their responses. Thus, if the nested data structure is ignored, the assumption of independence of observations underlying regression analysis may be violated (e.g., Hox, 010). This approach could also provide information regarding whether interventions aimed to optimise athletes motivation in each context should focus on the individual athlete or the team. To date, no research has examined achievement motivation across training and competition using multilevel modelling. Method Participants Participants were 410 (36 males, 48 females) football players, recruited from 36 teams in the United Kingdom. The players mean age was 1.11 (SD ¼ 4.34) years and they had been playing football competitively for an average of (SD ¼ 4.55) years. Most players competed at club level (93%), with few competing at county (%) and regional (5%) levels. On average, they attended 1.5 (SD ¼ 0.57) training sessions per week. At the time of data collection, the number of competitive matches they had played that season varied from 1 to 5 (14%), 5e10 (41%), 10e15 (17%), 15e0 (10%), to 0 or more (18%). Finally, the average period that the players were coached by their current coach was 1.65 years (SD ¼ 1.1). Measures We used a questionnaire which was divided into two major sections, one referring to training and one referring to competition. For all measures, we used identical items for both contexts as this allowed us to make straightforward comparisons between training and competition responses. That is, if we identified a difference between the two contexts, we knew that this was due to variation in the context rather than due to variation in the items. Thus, the instruments were used twice to assess the variables first in training and then in competition. The players were oriented towards the two contexts via written instructions (e.g., Please think about your football experience in training, and respond honestly to the following statements. ). A similar procedure has been used in previous research that examined goal orientations in school and sport (e.g., Duda & Nicholls, 199). Goal orientations Athletes goal orientations in the training and competition contexts were assessed with the Perception of Success Questionnaire (POSQ; Roberts, Treasure, & Balague, 1998), which consists of two 6-item subscales measuring task and ego orientations. Participants were asked when they feel most successful in training/ competition. The stem for each item was In training/competition, I feel most successful when.. Example items were I work hard for task orientation and I outperform others for ego orientation. Participants responded on a Likert scale, ranging from 1 (strongly disagree) to 5(strongly agree). The POSQ has demonstrated very good reliability with alpha coefficients of.90 for the task and.84 for the ego orientation subscales (Roberts et al., 1998). The mean for each subscale was computed and used in all analyses. This procedure was followed for all variables. Perceived motivational climate Perceived motivational climate was assessed with an adapted version of the Perceived Motivational Climate in Sport Questionnaire- (PMCSQ-; Newton et al., 000), which measures

4 494 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e499 perceptions of mastery (17 items) and performance climate (16 items). In the present study, we used only eight items from the mastery climate subscale and eight items from the performance climate subscale because only these items referred to coach behaviours and were relevant to both contexts. Abbreviated versions of the PMCSQ- have been used in previous research (e.g., Kavussanu & Spray, 006). The stem was: During training/ competition, on this team the coach., and example items are:.rewards trying hard for mastery climate, and.gives most of his or her attention to the stars for performance climate. Responses were made on a Likert scale ranging from 1 (strongly disagree)to 5(strongly agree). The PMCSQ- has demonstrated very good internal consistency with alpha coefficients of.88 and.87 for the mastery and performance climate subscales respectively (Newton et al., 000). Effort, enjoyment/interest and tension Three subscales of the Intrinsic Motivation Inventory (IMI; Ryan, 198) were used to measure effort (5 items), enjoyment (7 items) and tension (5 items). Participants were asked to think about their experiences during training/competition when responding to the items. Example items used in the present study are: I put a lot of effort into training/competition, I enjoy training/competition very much, and I feel pressured during training/competition for effort, enjoyment, and tension, respectively. Each item was rated on a Likert scale ranging from 1 (not at all true) to7(very true). These subscales have demonstrated satisfactory to very good reliability in previous research (effort, a ¼.84; enjoyment, a ¼.78; tension, a ¼.68; McAuley, Duncan, & Tammen, 1989). Procedure Upon approval of the study by the University Ethics Committee, we identified 36 football teams, and contacted the coaches of these teams to request their help with the study. The general study purpose and procedure for data collection were explained to the coaches during a phone call. Data collection took place e3 months after the season had started. Questionnaires were administered by a research assistant during a training session; before completing the questionnaires, participants signed a consent form. Players were informed of the study purposes verbally by the research assistant and via the information sheet attached to each questionnaire. It was emphasised that participation in the study was voluntary and that all participants responses would remain confidential. Finally, the players were asked to think about how they usually experience training and competition when they completed the respective parts of the questionnaire. Data analyses To examine our study purposes we employed MLM using SPSS Linear Mixed Models (version 19). Our design comprised three hierarchical levels of analysis: The repeated measures within training and competition at level 1 (n ¼ 80) were nested within individual athletes at level (n ¼ 410), which in turn were nested within teams at level 3 (n ¼ 36) (see Snijders & Bosker, 1999). It is important to acknowledge this hierarchical data structure as when individuals are clustered into groups (e.g., athletes in teams) observations at a specified level may not be independent, which may inflate Type 1 error (Cohen, Cohen, West, & Aiken, 003). Multilevel modelling takes into account this potential nonindependence as it allows the decomposition of variance in outcomes at each hierarchical level and an explanation of this variance with variables specified at corresponding levels (Heck, Thomas, & Tabata, 010). Prior to the analyses, the data set was prepared for repeated measures MLM via restructuring the file from variables to cases (e.g., Peugh & Enders, 005). Before investigating our first study purpose, we examined for each variable (i.e., goals, perceived climates, and outcomes) a three-level model with context (dummy coded: 0 ¼ training, 1 ¼ competition) specified as repeated measures and the intercepts at level (athletes) and 3 (teams) allowed to be random. This random-intercept-only model (Model 1) estimates the amount of variance in the outcome variable explained at each hierarchical level (i.e., within-group variance at level 1 and between-group variance at levels and 3) adjusting for correlated residuals within the random context effect. These estimates of variance were used to calculate intraclass correlations to determine whether MLM was necessary. Intraclass correlations were computed by dividing the unique variance at each level by the sum of the variances in levels 1,, and 3 (Hox, 010). For all models examined, we used the Maximum Likelihood (ML) estimation algorithm to compute coefficients because this algorithm allows comparison of nested models which differ in their fixed components (Tabachnick & Fidell, 007). Our first study purpose was to examine variability in football players goal orientations and perceptions of the climate across training and competition contexts. To examine this purpose, we calculated separate variance components for training and competition. These components substitute the total context variance (as calculated by Model 1) except that one estimate was now obtained per context. Thus, these parameter estimates reflect the withinsubjects variance for each context separately after taking into account between-subjects variance at the athlete and team levels. Wald s Z-test was used to assess significance of variance components. Our second study purpose was to examine differences in goals and climates across the two contexts. To examine this purpose, we extended the three-level random-intercept-only model (Model 1) by including context as a fixed factor. Testing this model (i.e., Model ) is equivalent to a paired t-test and allows determining the difference in the outcome variable (i.e., goals and climates) across the two contexts taking into account the nested structure of our data (Verbeke & Molenberghs, 000). In addition, variance components were examined by comparing this model with the random-intercept-only model (Model 1) using the chi-square ( Log) likelihood ratio test (e.g., Snijders & Bosker, 1999). Our third study purpose was to examine whether the context moderates the relationships between goal orientations, motivational climate, and effort, enjoyment and tension. To examine this purpose, we adopted an exploratory build-up strategy for selecting MLM models as outlined by Hox (010). The random-intercept-only model (i.e., Model 1) served as a baseline deviance for examining improvement of subsequent models. We also estimated for each outcome a three-level random-intercept model with context as a fixed predictor (i.e., Model ). This allowed estimating the unique effect of context on each outcome averaged over all three levels. Next, we extended Model by adding goal orientations and perceived climates as fixed predictors. This resulted in a three-level random-intercept model (i.e., Model 3) that estimated the relationships between predictors and each outcome averaged over all three levels. Before entering the variables in this model, goals and perceived climates were centred. Centring predictors in multilevel models decreases multicollinearity among predictors and between random intercepts and slopes, thereby stabilising the analysis (Cohen et al., 003). Task and ego orientations were centred around the grand mean of all individual cases. As explained earlier, perceptions of the motivational climate referred to the climate unique to athletes own team. Mean differences between groups (i.e., teams) on an independent variable can be an important factor in predicting the dependent variable (Kreft & De Leeuw, 1998).

5 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e Hence, recognising group membership, mastery and performance climates were centred around the mean of each team (e.g., Snijders & Bosker, 1999). Then, we tested if this model would improve by allowing the slopes for context and task and ego orientation to be random. This three-level random-coefficients model tested if the relationships between the predictors and outcomes varied across the three levels. The slopes for mastery and performance climate were not allowed to be random because predictors at the highest level of the hierarchy are necessarily fixed as there is no higher level within which they can vary (e.g., Tabachnick & Fidell, 007). Next, interaction terms were added to the model. Interaction terms for all possible two- and three-way interactions between context, goals, and perceived climates were formed by multiplying the centred predictors (Cohen et al., 003). We tested all possible two-way interactions between goals, goals and context, climates and context, and goals and climates; and three-way cross-level interactions between context, goals and perceived climates. Interaction effects were tested for their fixed and random effects. To permit a more powerful test of the significant interaction effects in each model, a sequential step-down approach was used: Starting with the highest-order term in the regression equation, the nonsignificant interactions were removed one at a time, and each subsequent term was tested for significance (Aiken & West, 1991; Hox, 010). Identified interaction effects were explored further by: a) plotting the regression of the outcome variable on the predictor at low (1 SD below the mean) and high (1 SD above the mean) values of the moderator, and b) testing whether the slopes of the simple regression lines were significantly different from zero (Aiken & West, 1991). To control for the family-wise error rate that results from multiple comparisons, a BonferronieHolm (Holm, 1979) correction was applied to each full model; the p-value for significance (i.e., a ¼.05) was adjusted down by dividing the nominal level of significance by the number of total comparisons. Model improvement (i.e., a better fit indicated by lower deviance) was tested using the chi-square ( Log) maximum likelihood ratio test of differences between models (e.g., Snijders & Bosker, 1999). The above build-up strategy resulted in a final model (i.e., Model 4) which was used to examine our third study purpose. Results Preliminary analyses Preliminary analyses were conducted to examine missing values, outliers, and internal consistency of the scales. These analyses revealed that 0.7% of the data points were randomly missing across the data. When less than 5% of the data are randomly missing from a large data set, almost any procedure for replacing missing values yields similar results (Tabachnick & Fidell, 007). Accordingly, missing values were replaced with the series mean of the individual items. Outliers were examined using standardised z- scores; cases with scores in excess of 3.9 SD from the mean of the respective subscale were considered outliers (Tabachnick & Fidell, 007). In the entire data set, eight outliers were found and removed. Finally, all scales had good-to-very good internal consistency with alpha coefficients ranging from.78 to.89. Main analyses As outlined above, we adopted an exploratory build-up strategy to examine our study purposes (Hox, 010). Models in which the slopes for context, task and ego orientations, and/or their interactions were allowed to be random did not converge. Hence, the following three-level random-intercept models with context specified as repeated measures were further tested for improvement: Model 1, a random-intercept-only model; Model, a random-intercept model with context as a fixed factor; Model 3, a random-intercept model with context, goals and climates as fixed predictors (i.e., main effects); and Model 4, a random-intercept model with fixed main and interaction effects for context, goals and climates. As can be seen in Table, Model showed better fit than Model 1 for ego orientation, performance climate and all outcomes, but no better fit for task orientation and mastery climate. The models for all outcomes further improved when we included goals and climates as main predictors (Model 3). Finally, best fit was achieved when main and (significant) interaction terms were included (Model 4). Models 1, and 4 were selected to examine our first, second, and third study purpose, respectively. Intraclass correlations and variance components were evaluated by the random-intercept model without predictors (Model 1). Intraclass correlations for all three levels are presented in Table 1 and indicate that the data were clustered. The values for goals, climates, and effort and enjoyment suggest that the data should be analysed using a three-level model. However, for tension, the intraclass correlation (r ¼.0) and the random-intercept test (Wald Z ¼ 0.33, p >.05) showed a lack of variability among athletes (level ). Therefore, we used the chi-square likelihood ratio test to examine whether Models 1e4 for tension would improve when only the intercept at level 3 (team) was allowed to be random. This test revealed that the models in which only the intercept at level 3 was allowed to be random showed a poorer fit than the initial models (i.e., with random intercepts at levels and 3, see Table ) for Models, 3, and 4 [Model 1, c diff (3) ¼ 0.11, p >.05; Model, c diff (4) ¼ 39.80, p <.001; Model 3, c diff (8) ¼ 40.85, p <.001; Model 4, c diff (10) ¼ 48.3, p <.001]. Due to the better fit, we retained the models for tension with random intercepts at levels and 3 (i.e., athletes and teams as random second and third level units, respectively) for all further analyses (cf. Hox, 010). To examine variability in goals and climates across training and competition, we calculated from Model 1 the unique components Table 1 Intraclass correlations and context effects for all variables. Model 1 e Intraclass correlations Model e Context effects Level 1 Level Level 3 Variables r r Wald Z r Wald Z M SE M SE t (409) d Task orientation ***.1.65** Ego orientation *** ** *** 0.40 Mastery climate *** *** Performance climate *** *** *** 0.8 Effort *** *** *** 0.38 Enjoyment/interest *** *** *** 0.5 Tension ** *** 0.99 Notes. r ¼ intraclass correlation; d ¼ Cohen s d; values of 0.0, 0.50, and 0.80, constitute small, medium, and large effect sizes, respectively (Cohen, 199); **p <.01; ***p <.001.

6 496 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e499 Table Comparisons of multilevel models for all variables. Model 1 Model Model 3 Model 4 Variables LL df LL df c diff (M1 M) LL df c diff (M M3) LL df c diff (M3 M4) Task orientation N/A N/A Ego orientation *** a N/A N/A Mastery climate N/A N/A Performance climate *** a N/A N/A Effort *** *** * Enjoyment/interest *** *** *** Tension *** *** *** Notes. Model 1 ¼ random-intercept model; Model ¼ random-intercept model with context as fixed factor; Model 3 ¼ random-intercept model with main predictor terms; Model 4 ¼ random-intercept model with main and interaction predictor terms. M ¼ Model; LL¼ Log likelihood. N/A ¼ not applicable (i.e., not assessed). *p <.05; **p <.01; ***p <.001. a Effect sizes: ego orientation, h ¼.16, performance climate, h ¼.10; values of.01,.09, and.5, constitute small, medium, and large effect sizes, respectively (Cohen, 1988; Tabachnick & Fidell, 007). of variance of each context. We found the following parameter estimates: task orientation: training, estimate ¼.1, SE ¼.01, Wald Z ¼ 8.75, p <.001, competition, estimate ¼.13, SE ¼.01, Wald Z ¼ 8.90, p <.001; ego orientation: training, estimate ¼.3, SE ¼.03, Wald Z ¼ 9.53, p <.001, competition, estimate ¼.4, SE ¼.03, Wald Z ¼ 7.87, p <.001; mastery climate: training, estimate ¼.09, SE ¼.01, Wald Z ¼ 7.89, p <.001, competition, estimate ¼.11, SE ¼.01, Wald Z ¼ 8.87, p <.001; performance climate: training, estimate ¼.0, SE ¼.01, Wald Z ¼ 7., p <.001, competition, estimate ¼.31, SE ¼.03, Wald Z ¼ 9.60, p <.001. Thus, training and competition contexts explained a significant amount of variance beyond the athlete and team levels, in both goals and perceived motivational climates. To examine differences in goals and climates across the two contexts, we used a three-level random-intercept model with context as a fixed factor (Model ). The paired t-test obtained from this model showed that, averaged over all three levels, athletes ego orientation and performance climate were both significantly higher in competition than in training (see Table 1). The effect sizes were small-to-medium for both variables (see Cohen, 199). Task orientation and mastery climate did not differ significantly across the two contexts. Chi-square likelihood ratio tests comparing Models 1 and, confirmed the result of the paired t-test: The model improved significantly only for ego orientation and performance climate by including context as a fixed factor. Thus, by adding context as a fixed factor the explained variance in ego orientation and performance climate significantly increased. The effect sizes e calculated based on guidelines by Kreft and De Leeuw (1998) e were medium-to-large for ego orientation (h ¼.16) and medium for performance climate (h ¼.10) (Cohen, 1988; see Table ). To examine whether the context moderates the relationships between goals, climates and outcomes, we used a random-intercept model with fixed main and interaction effects for context, goals and climates (Model 4). Results for fixed effects of this model are presented in Table 3. On average, context, task orientation and mastery climate positively predicted effort and enjoyment. Ego orientation did not predict either outcome, while performance climate negatively predicted effort. Effort and enjoyment were also both negatively predicted by the context by mastery climate interaction. Effect sizes for these findings were calculated based on guidelines by Nakagawa and Cuthill (007) and ranged from small to large (Cohen, 199; see Table 3). Probing these interactions showed a similar pattern for effort (see Fig. 1a) and enjoyment (see Fig. 1b): A higher mastery climate was associated with higher effort in training (b ¼.41, SE ¼.09, t ¼ 4.38, p <.001) but not in competition (b ¼.14, SE ¼.09, t ¼ 1.55, p >.05) and with greater enjoyment in training (b ¼.66, SE ¼.09, t ¼ 7.74, p <.001) than in competition (b ¼.19, SE ¼.08, t ¼.8, p <.05). Tension was positively predicted by context, performance climate, and the context by ego orientation interaction, and negatively predicted by ego orientation and the context by performance climate interaction. Probing these interactions showed that higher ego orientation was associated with lower tension in training (b ¼.4, SE ¼.07, t ¼ 3.46, p ¼.001) but not in competition (b ¼.10, SE ¼.08, t ¼ 1.33, p >.05) (see Fig. ), and higher performance climate was associated with greater tension in training (b ¼.36, SE ¼.08, t ¼ 4.8, p <.001) but not in competition (b ¼.01, SE ¼.08, t ¼ 0.11, p >.05) (see Fig. 3). Discussion Although training and competition are integral contexts of sport, to date, the distinction between these two contexts has been largely overlooked in achievement goal research. It has been suggested that different achievement criteria may operate within these contexts (Harwood, 00), and empirical evidence in tennis and softball has provided support for this contention (e.g., van de Pol & Table 3 Fixed effects of context, goals and climate predicting effort, enjoyment and tension (Model 4). Predictors Effort Enjoyment/interest Tension Parameter Estimate SE t d Parameter Estimate SE t d Parameter Estimate SE t d Intercept *** *** *** Context *** *** *** 1.67 Task orientation *** *** Ego orientation (EO) *** 0.4 Mastery climate (MC) *** *** Perf. climate (PC) ** *** 0.31 Context EO *** 0.35 Context MC * *** 0.37 Context PC *** 0.9 Notes. Perf. ¼ performance; only significant interaction effects are displayed; d ¼ Cohen s d; values of 0.0, 0.50, and 0.80, constitute small, medium, and large effect sizes, respectively (Cohen, 199); *p <.05; **p <.01; ***p.001.

7 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e Effort Enjoyment a b (-1 SD) (+1 SD) Mastery Climate (-1 SD) (+1 SD) Mastery Climate Fig. 1. Regression of effort (a) and enjoyment (b) on mastery climate in training and competition. Kavussanu, 011a; Williams, 1998). In this study, we sought to replicate and extend this work to football, and examined the contextual influence of training and competition on motivational climate and achievement goals as well as on the relationship of these variables with motivational outcomes. In addition, this study was the first to examine these issues using multilevel modeling. The first study purpose was to examine variability in football players goal orientations and perceived motivational climate across training and competition contexts. The context explained significant variation in goals and perceived climate beyond the variation explained by the athlete and team levels. Previous research has shown that variability in perceived climate was higher among athletes than among teams (Smith, Smoll, & Cumming, 009). Our findings support this work and extend it to goal orientations. Thus, perceived motivational climate and goal orientations may be more individual-level rather than team-level constructs. However, recognising also athletes intra-individual experiences across training and competition could provide a more complete understanding of variability in goals and perceived climate in sport. The second study purpose was to examine whether goal orientations and perceived motivational climate differed across training and competition. Football players reported higher ego orientation in competition than in training, a finding that supports our hypothesis and previous research in tennis players (van de Pol & Kavussanu, 011a). This finding may be explained by the strength of the normative cues in the competition context (e.g., emphasis on social comparison). Ego-oriented players may be particularly sensitive to these cues which may strengthen their tendency to evaluate success using normative criteria in this context compared to training. Football players task orientation did not differ between the two contexts, a finding consistent with results in team-sport athletes (van de Pol & Kavussanu, 011b) but inconsistent with findings in tennis players, who reported higher task orientation in training than in competition (van de Pol & Kavussanu, 011a). This discrepancy may be due to differences in interest/enjoyment in the two samples. Specifically, in contrast to the tennis players in the van de Pol and Kavussanu (011a) study, our participants reported significantly higher interest/enjoyment in competition than in training (see Table 1). Previous research has shown that students high in interest are more likely to endorse a task goal than those low in interest (Harackiewicz, Durik, Barron, Linnenbrink-Garcia, & Tauer, 008). Hence, the high level of interest/enjoyment in competition may have strengthened football players task orientation, thereby maintaining this goal in competition at a high level (van de Pol & Kavussanu, 011b). With regard to the perceived motivational climate, perceptions of performance climate were higher in competition than in training, suggesting that coaches may place more emphasis on Tension Tension (-1 SD) (+1 SD) Ego orientation (-1 SD) (+1 SD) Performance Climate Fig.. Regression of tension on ego orientation in training and competition. Fig. 3. Regression of tension on performance climate in training and competition.

8 498 P.K.C. van de Pol et al. / Psychology of Sport and Exercise 13 (01) 491e499 normative success in competition than in training. Perceptions of mastery climate were similar across the two contexts, suggesting that coaches may reward behaviours such as trying hard and improving skills similarly in both contexts. Overall, our findings provide support for Ames (199) argument that achievement contexts can be structured as mastery oriented even under interpersonal competitive conditions. Our third study purpose was to examine whether the context moderates the relationships between goal orientations and motivational climate and effort, enjoyment and tension. Effort and enjoyment were both positively predicted by task orientation in both contexts, a finding that supports our hypotheses. Taskoriented athletes evaluate their competence in relation to their own personal standards of improvement and performance, apply effort, and experience enjoyment in both contexts. Effort was unrelated to ego orientation, a finding that is consistent with previous research in team-sport athletes (van de Pol & Kavussanu, 011b). However, in tennis players (van de Pol & Kavussanu, 011a) and in athletes from a variety of individual sports (van de Pol & Kavussanu, 011b) high ego orientation corresponded to more effort than low ego orientation when task orientation was low. Hence, these discrepancies in the findings may be due to the type of sport. It has been suggested that individual sports may provide more exact individual performance information than team sports (cf. Hanrahan & Cerin, 009). Therefore, ego-oriented athletes may find it easier to link their normative goal striving to a concrete personal success in individual sports compared to team sports, where personal success is intertwined with the overall team success. Thus, a more concrete prospect of a gain in normative competence may lead ego-oriented athletes in individual sports to put more effort in their achievement striving compared to teamsport athletes. Finally, enjoyment was unrelated to ego orientation, a finding which supports previous research (Biddle et al., 003; van de Pol & Kavussanu, 011a). Effort was associated positively with perceived mastery climate only in training, while enjoyment was associated positively with this climate in both contexts but more strongly in training than in competition. Thus, a mastery climate may be more beneficial for effort and enjoyment in training than in competition. This may occur because mastery climate better matches the purposes of training, such as skill development and improvement, than those of competition with its emphasis on an instant performance of these skills. In addition, both effort and enjoyment were higher in competition than in training. Thus, mastery climate may be most beneficial for these variables when they are not very high. Effort was negatively associated with performance climate in both contexts, which suggests that when coaches want to foster athletes effort in training and competition they should temper performance climate in these contexts. Tension was unrelated to task orientation in both contexts. This finding partly supports previous research, which found that task orientation was either unrelated or negatively related to tension (see for a review, Biddle et al., 003). Tension was inversely associated with ego orientation in training but unrelated to this goal in competition. The inverse relationship with ego orientation in training was unexpected, as this goal has been either unrelated or positively related to tension and/or anxiety in previous research (see Biddle et al., 003; van de Pol & Kavussanu, 011b). According to Nicholls (1989), high ego-oriented athletes feelings of tension and/or anxiety should depend on their expectations of success and perceived ability relative to that of others. Perhaps our ego-oriented football players had, on average, high normative success expectations and/or ability perceptions in training, which may explain why ego orientation was negatively related to tension in this context. In view of this, the fact that ego orientation was unrelated to tension in competition may indicate that ego-oriented athletes were less certain about their normative success expectations and/or ability perceptions in competition compared to training. However, these suggestions are speculative and need to be verified by future research. Tension was positively associated with perceived performance climate but only in training. Thus, the characteristics of a performance climate, such as unequal recognition, may lead athletes to experience tension in training. In competition, athletes may be better able to cope with tension that is derived from normative success striving and public evaluation, as these factors are inherent, and probably more frequently experienced, in this context (cf. Lazarus & Folkman, 1984). Accordingly, athletes may feel less tension evoked by a perceived performance climate in competition than in training. Also, tension was higher in competition than in training, which may have influenced its non-significant relationship with performance climate in this context. Perceived mastery climate was unrelated to tension in both contexts, supporting previous research (Newton & Duda, 1999; Seifriz, Duda, & Chi, 199). Thus, when coaches aim to reduce athletes tension they should focus on avoiding those behaviours that create a performance climate, such as encouraging social comparison and intrateam rivalry. Overall, our findings suggest that in order to achieve positive motivational consequences in training and competition, coaches need to promote athletes task orientation and create a mastery climate in both contexts. However, because performance climate may increase from training to competition, coaches need to reward athletes effort and personal progress in both contexts. Ego orientation may not be detrimental; however, the possibility of repeated failures in normative success striving makes this goal a vulnerable source of motivation. Nonetheless, in both contexts, coaches should avoid normative comparison and public evaluation, as these behaviours may lead athletes to perceive the motivational climate as performance oriented with subsequent negative effects on effort and tension. When implementing the above suggestions, practitioners may direct their interventions to the team as a whole as well as to the individual athlete; the latter strategy may have the most impact. Limitations of the study and directions for future research Although this study revealed some interesting results, our findings need to be interpreted in light of some limitations. First, the study was cross-sectional, thus assertions about the direction of causality cannot be made. Experimental studies that simulate training and competition contexts are needed to determine the direction of causality. Second, goal orientations and perceived climates may develop over time (e.g., Smith et al., 009; Williams, 1998). Hence, researchers should examine contextual and temporal stability of goals and climates via a longitudinal design measuring these constructs in training and competition at different time points during the season. Third, although we found variability in goals, climates, and outcomes among athletes and teams, models that tested if the relationships (i.e., random slopes) between these variables differed across athletes and teams did not converge. Future research could extend our findings by testing if these relationships remain constant among athletes and teams at different competition levels and sport types. Fourth, researchers could examine whether normative success expectations and/or perceived ability moderate the relationship between ego orientation and outcomes in the two contexts. Finally, future research could investigate approach-avoidance goals, which incorporate the valence (i.e., approach vs. avoidance) dimension of competence (e.g., Elliot & McGregor, 001). We focused on task and ego goals which reflect differences in the