REVISION

Affect and Achievement Goals in Physical Activity: A Meta-Analysis

Nikos Ntoumanis University of Exeter & Stuart J.H. Biddle Loughborough University. Manuscript published in Scandinavian Journal of Medicine and Science in Sport (Special Issue: European Perspectives in Sport Motivation Research), 9, 315-332. Manuscript submitted: 05-06-1998 Revision submitted: 15-04-1999 Running head: Meta-analysis on affect and goals Correspondence: Professor Stuart J.H. Biddle Department of Physical Education, Sports Science and Recreation Management Loughborough University Loughborough Leicestershire LE11 3TU United Kingdom Tel: +44 (0)1509 223287 Fax: +44 (0)1509 223971 E-mail: [email protected]

2 Abstract Achievement goal orientation theory has been the subject of extensive research in research years. In view of the importance of identifying the motivational antecedents of affect in physical activity, this study examined through meta-analysis the conflicting evidence regarding the links between different achievement goals and emotions. Using the formulas of Hunter and Schmidt (1), correlations were gathered from 41 independent samples and were corrected for both sampling and measurement errors. The results showed that task orientation and positive affect were moderately-to-highly correlated and in a positive fashion, whereas the relationship between task orientation and negative affect was negative and moderate-tosmall. Both correlations were found to be heterogeneous, and so moderators were sought. The relationships between ego orientation and positive and negative affect were positive but very small, with the former being heterogeneous. Moderators coded were the time frame of affect (independent of context vs. after an athletic event), the physical activity setting (school physical education vs. recreation vs. competitive sport), age (university vs. school students), nationality (British vs. American), nature of negative affect (high vs. low arousal), and the publication status of the studies (published vs. unpublished). Lastly, a subset of the corrected correlations were inserted into a structural equation modelling analysis in order to look concurrently at the relationships among all the variables. Keywords: affect, emotions, goal orientations, meta-analysis

3 In all areas of physical activity emotion plays a central role and one that psychologists can help understand and explain. For example, the role of emotion in high level competitive sport is clear to see. Emotional control is often believed to be one feature of the successful player (2). In addition, feelings of joy, satisfaction, enthusiasm, anxiety, fear, guilt, and shame, tears of happiness or sadness, hand-slapping and shouting, are all indicators that sport can offer a very rich variety of affective reactions and experiences. Furthermore, the emotion experienced by recreational players and health-related exercise participants may be crucial in determining whether involvement is continued or, as is often the case in structured exercise, ceased after initially high motivation (3). For all these reasons, it is obvious that what may cause certain affective reactions constitutes an important area of concern for psychologists. Since 1983, when Vallerand (4) underlined the importance and the need for research on emotion in physical activity, there have been many studies that have looked into the antecedents, the processes and the outcomes of various types of affect employing different theoretical perspectives. In trying to obtain a more comprehensive understanding of the mechanisms of positive and negative affect creation in sport, many researchers have underlined the importance and the predictive role of motivational variables. Various theories have been developed to describe how different motivational patterns can invoke different achievement reactions in achievement environments such as sport. The present study statistically synthesises through the use of meta-analysis part of this literature that has examined affect in physical activity under the framework of achievement goal theories of motivation (e.g. 5, 6). Achievement goal theory has become very popular in contemporary sport and exercise psychology. One of the outcomes of holding different achievement goals is that of affective reactions. For this reason it is important to quantify the effect of achievement goals on affect in physical activity settings.

4 Before some indicative studies are introduced, it is essential to present the theoretical framework this meta-analysis is based on. In order to understand affect adequately, one needs to explore its definitional properties and dimensions. In general psychology, the terms affect and emotion have often been used interchangeably. However, some authors, such as Lazarus (7), have differentiated between the two feeling states suggesting that emotion is a more generic term. He defined affect as the subjective experience of emotion, and he conceptualised emotions as “complex, patterned, organismic reactions to how we think we are doing in our lifelong efforts to survive and flourish and to achieve what we wish for ourselves” (7, p. 6). Despite this subtle distinction, in the physical activity literature emotion and affect have been referred to and measured as synonymous constructs. In the present meta-analysis both terms have been used to describe the feelings experienced by athletes. With regard to the dimensional properties of affect, there is an on-going controversy in psychology with important theoretical and methodological implications. This centres on whether emotions should be conceived as discrete categories (7,8) or if they can be reduced to a few higher order dimensions or factors (9). The advocates of the discrete analysis of emotions have urged researchers not to combine emotions into a small number of categories. For example, Lazarus (7), in his cognitive-motivational-relational theory of emotion, argued that the distinct qualities of different emotions can be lost or blurred when they are combined into a few dimensions. According to his theory, each emotion is unique because it is created by a different appraisal of the perceived significance of an event (the relational meaning of an event) which involves a distinct cost-benefit analysis. In the area we reviewed a variety of discrete emotions were measured, however, due to the relatively small number of studies that provided sufficient information for the meta-analysis, it was deemed appropriate to combine these emotions into general affect dimensions. The use of higher order affect factors was also justifiable by the theoretical motivational framework we adopted (see below). Although we

5 agree that each emotion is unique, we believe that the advocates of the discrete analysis should not overlook that there are shared qualities among emotions that can be represented through higher order factors or dimensions. In this regard, the dimensional analysis of affect is useful. In a series of studies during the 1980's, Watson and colleagues (10, 11, 9) showed that two major factors consistently appeared in factor analyses of a large pool of emotions. The first factor is Positive Affect and reflects the extent to which a person feels enthusiastic, active or alert. The second factor is Negative Affect and refers to whether one feels unpleasant affective states such as anger, contempt, and fear. Also using a dimensional analysis of affect, but from a different angle, Russell (12) presented the circumplex model of affect. In contrast to the previous affect structure, this model categorised affect into the dimensions of valence (pleasant vs. unpleasant) and arousal (high vs. low). Both dimensional models of affect were useful in our meta-analysis. Specifically, the discrete emotions reported in the studies of our review were classified into the positive and negative affect dimensions based on the work of Watson and colleagues. As it will be shown below, this classification is important because it facilitates the testing and comparison of results across studies that have used achievement goal theories of motivation to examine a variety of affective outcomes. As explained below, these theories refer to general in addition to specific emotional outcomes of different motivational orientations. Another dimensional model employed in this study was the circumplex model of affect; this was instrumental in deriving specific moderator variables. Achievement goals in physical activity As suggested earlier, much of the interest on affect in physical activity has been focused on the links between a number of different emotions and achievement motivation indices using an achievement goals theoretical framework (13, 5, 6). Reflecting the current

6 dominance of the cognitive paradigm in psychology, these theorists employ a socialcognitive perspective to examine how individuals cognitively process and develop their views about the nature of achievement. Specifically, they argue that individuals in achievement settings strive to demonstrate competence and avoid displaying incompetence. Perception of competence has two conceptions – one normatively-based and the other selfreferenced. These affect our understanding of what it takes to be ‘successful’ in an achievement context. These two conceptions lead to the formation of at least two goal states (task and ego goal involvement) in a specific achievement situation. In addition to the motivational climate of the context, these goal states are the product of achievement goal orientations. When performance evaluation is based on normative standards, that is when success and failure are defined in comparison to the performance of others, then an ego goal orientation is salient (Nicholls, 1989). Alternatively, when performance evaluation is selfreferenced, that is when it is based on personal improvement and learning criteria, then a task goal orientation is exhibited. As Ames (13) has contended, achievement goals affect less the amount of motivation of individuals and more the quality of their motivation. Specifically, the study of achievement goals concerns the understanding of the causes, the development, and the consequences of motivated behaviour, rather than the amount of motivation per se. As sport is an achievement-oriented context, researchers since the early 1980’s have been interested in applying the achievement goal theories to this area. The resulting volume of research has been very fruitful because it has aided our understanding of motivated processes and outcomes. In general, research (14) has shown that maladaptive motivational patterns (e.g. lack of effort and persistence, devaluation of activities, selection of inappropriate tasks and strategies) are more likely to be exhibited by ego-oriented individuals, especially those who doubt their competence. In contrast, task-oriented athletes

7 are more likely to exhibit adaptive motivational patterns (e.g. challenge seeking, use of effective strategies, high effort; 14). Emotions and achievement goals in physical activity With regard to emotions, the achievement goals theoretical framework postulates that ego- and task-oriented individuals have different affective responses in achievement situations, such as sport. Specifically, ego-oriented individuals are likely to experience high negative affect and low positive affect, especially those who hold perceptions of low competence (6). This is because, as Roberts (15) has argued, winning and losing in sport are highly unstable outcomes with relatively uncontrollable objective demands and, thus, they can create negative affective reactions in ego-oriented athletes. In contrast, task-oriented athletes are prone to experience high positive and low negative affect because they have internal standards of performance and the outcome they strive for is subjective and relatively controllable. Several studies have examined the hypothesised links between task and ego achievement goals and various types of affect, such as satisfaction, enjoyment, anxiety, tension, and boredom. The present meta-analysis is a quantitative review of all the above studies that have provided the necessary statistical information to calculate effect sizes of association. Several of these studies have confirmed the hypothesised links between achievement goals and affect (e.g., 16, 17, 18, 19, 20). However, several studies have not verified the predicted relationships between achievement goals and affect. For example, studies by Hom, Duda, and Miller (21), Newton and Duda (22), and Vlachopoulos and Biddle (23), showed task orientation and negative affect to be unrelated rather than negatively related. Furthermore, Duda and Nicholls (24), Rethorst and Duda (25), and Vlachopoulos, Biddle, and Fox (26) found ego orientation and positive affect to be unrelated or even positively related. In addition, no significant correlations between ego orientation and

8 negative affect were reported by Duda, Chi, Newton, Walling, and Catley (27), Kavussanu and Roberts (28), and Newton and Duda (22). In view of the previously presented arguments regarding the significance of emotional reactions in the different physical activity settings and the need to investigate their antecedents (4), this study reviewed quantitatively the motivational links of affect3. In brief, our purpose was to meta-analyse the previously presented conflicting evidence and provide some indices of strengths of association (effect sizes) between different goal orientations and affective outcomes. Potential explanations for inconsistent results across studies were sought through moderator analysis. It was hypothesised that (1) task orientation will correlate positively with positive affect, but (2) negatively with negative affect, and that (3) ego orientation will correlate positively with negative affect, but (4) negatively with positive affect. A further purpose of this study was to examine the concurrent associations between task and ego goals with positive and negative affect through the use of structural equation modelling. Unlike meta-analysis, which can examine the links between two variables at a time, structural equation modelling can estimate whether these links remain unchanged in the presence (and potential influence) of other variables. As it will be argued later on, these two techniques can compliment each other providing a more comprehensive understanding of an area. Method Search Procedures The relevant studies were identified by means of computer and manual searches as well as searches of personal files. Specifically, the computer databases searched were BIDS, First Search, Sport Discus, and PsychLit. The following keywords were used: goal orientations, achievement goals, task and ego goals, affect, emotions, enjoyment, satisfaction, happiness, anxiety, depression, boredom, anger, tension. Also, a manual search was

9 conducted on the 1986-1997 issues of the following journals: International Journal of Sport Psychology, Journal of Applied Sport Psychology, Journal of Sport and Exercise Psychology, Journal of Sport Behavior, Journal of Sports Sciences, Pediatric Exercise Science, Research Quarterly for Exercise and Sport, and The Sport Psychologist. Inclusion Criteria Studies were included in the meta-analysis if they had examined the relationships between task and ego achievement goals with various positive and negative emotions. Nearly 90% of the studies employed the task and ego subscales of the Task and Ego Orientation in Sport Questionnaire (TEOSQ; 24) and the Perception of Success Questionnaire (POSQ; 29). The remaining 10% used other subscales from the Sport Orientation Questionnaire (SOQ; 30), and the Achievement Orientation Questionnaire (AOI; 31). For the purposes of this meta-analysis we selected the Mastery Achievement and Sport Ability subscales of the AOI, and the Goal and Win subscales of the SOQ, as measures compatible to task and ego achievement goals respectively. The AOI is based on the early work of Nicholls (32) on achievement goals, and the two selected subscales reflect motivational orientations that Nicholls (6) subsequently named as task and ego orientation. As far as the SOQ is concerned, results reported by Duda (33) and Marsh (34) showed that the Goal and Win subscales correlate highly with task and ego achievement goals respectively. As far as measurement of affect is concerned, more than 50% of the studies did not use an established questionnaire. The remaining studies used predominantly (26%) the Enjoyment/Interest subscale of the Intrinsic Motivation Inventory (IMI; 35), and to a lesser extent the Sport Competition Anxiety Test (SCAT; 36), the Competitive State Anxiety Inventory-2 (CSAI-2; 37), the Exercise-Induced Feeling Inventory (EFI; 38) and a short version of the Profile of Mood States (POMS; 39).

10 With regard to publication status, both articles and conference presentations were included in the meta-analysis. Some critics of meta-analysis argue that unpublished papers should be excluded because they are often methodologically inferior to published articles. However, as Hunter and Schimdt (1) emphasised, such an assumption can be tested through meta-analysis itself. Methodological quality or publication status can be coded as a moderator variable and the results of the two sets of studies can be compared. From all studies located, 45 did not provide the necessary statistical information for the calculation of corrected mean correlations and variances. Specifically, the internal reliabilities and the zero-order correlation matrices of the variables of interest from each study were necessary for the meta-analytic computations. Therefore, a request letter was sent (with follow-up where necessary) to the authors of these studies. After six months, we had sufficient information from 37 studies with a total of 41 independent samples (N= 7950) which were subjected to the meta-analysis. These studies are indicated in the References section by an asterisk. The observed correlations among the variables of interest in the 41 samples are presented in a stem and leaf display in Table 1. The distribution of the correlations was homogeneous because the ratios of the skewness and kurtosis scores to their respective standard errors were not significant (40). In samples where more than one correlation was reported between an achievement goal and an affect (e.g. when task orientation correlated with several indicators of negative affect), a composite score was calculated for the affect variable based on the formulas of Hunter and Schmidt (1, pp. 454-457). The reason for combining the correlations within samples was to preserve the statistical independence of the data. Violations of the assumptions of data independence can affect more (i.e. increase) the observed variance of the mean correlations than the actual magnitude of the correlations (1). This artificial increase in the observed variance will erroneously indicate heterogeneity in the mean correlation and the

11 existence of moderator variables. Finally, for each study the following information was coded: number of participants, gender, age, nationality, publication status, questionnaire used to measure achievement goals, types of positive and negative affect, setting of physical activity, and whether affect was measured after an athletic event or independent of context. In line with Rosenthal's (41) suggestion, a summary of these study characteristics is presented in Table 2. Calculation Procedures The analysis was carried out according to Hunter and Schmidt’s (1) procedures for meta-analysis. These procedures emphasise that there are many kinds of error or “artifacts” in the results of primary studies which, if not corrected, can have an impact on the findings of a meta-analysis. Specifically, study artifacts can attenuate the magnitude of the population or “real” correlation between two variables, and at the same time artificially increase its variance, thus erroneously indicating the existence of moderators. Examples of study artifacts are sampling error, measurement error, dichotomisation of continuous variables, imperfect construct validity, and computational errors. The great advantage of meta-analysis over narrative reviews of literature is that it takes into account and corrects for several of these artifacts which can potentially give a distorted view of the area to the narrative reviewer. The present study, in line with recent meta-analyses in other areas of psychology (65b) has corrected for the two major sources of error in primary studies: sampling error and measurement error. Sampling error, which refers to the fact that samples are usually imperfect representatives of a targeted population, can decrease the magnitude of the correlation between two variables. It can also incorrectly show that the correlation is heterogeneous by inflating its observed variance across studies. Especially in small sample studies, it can create a large variability in study outcomes. Error of measurement, or imperfect reliability, in either of two variables can also deflate the magnitude of their

12 correlations and increase its variance. In the present study, measurement error was corrected using artifact distributions (1; pp. 173-177) due to the fact that some internal reliabilities were not reported in the primary studies. Artifact distribution analysis creates distributions for the reliabilities of each variable across studies. Then, based on these distributions, a composite reliability score is obtained for each variable which is used to correct for the impact of measurement error on the magnitude and the variance of the population correlations. Where appropriate, we calculated the number of studies needed with null results to reduce the estimated correlation to .10, a value characterised as small by Cohen (66). This practice tries to deal with the well-known problem of availability bias, that is, the fact that studies available to meta-analysis (or indeed to narrative reviews) are more likely to have reported statistical significance than studies remaining stored in file drawers (41). For each correlation obtained, its credibility and confidence intervals were calculated. Quite often credibility and confidence intervals are confused in the literature (67a). The calculation of the credibility interval is based on the corrected standard deviation and the corrected mean correlation. If the credibility interval is sufficiently large or includes zero, then it indicates that the mean correlation is heterogeneous and, therefore, moderators should be sought. If the interval is small or does not include zero, then it suggests that the mean correlation is homogeneous and there are no moderators. The confidence interval, on the other hand, is used to estimate the accuracy of the mean correlation and it is created using the standard error of that correlation. The homogeneity of each corrected mean correlation was assessed with the 75% rule and the Q test. The 75% rule, presented by Hunter and Schmidt (1), proposes that if in any data set the known and correctable artifacts account for 75% of the variance in the corrected mean correlation, then the remaining 25% is probably due to uncorrected artifacts. Therefore,

13 in such cases researchers should assume that the correlation is homogenous, that is, it is consistent across studies and no moderators should be sought. The Q test for homogeneity has a chi-square distribution and has been used extensively in previous meta-analyses to estimate the variability in corrected mean correlations. In a series of Monte Carlo computer simulations, Sackett, Harris, and Orr (68a) compared the statistical power to detect moderators of the 75% rule with that of the Q test and other tests of homogeneity. Under most conditions studied, the 75% rule had statistical power that was equal to or greater than that of the other methods. Results Due to the fact that the meta-analysis employed artifact distributions, there is no information (e.g. stem and leaf plots) for corrected effect sizes at the level of individual studies. Table 3 presents summary information of the four correlations (i.e. effect sizes) meta-analysed across studies. Given Cohen’s (66) suggestion that when analysing correlations, .10, .30, and .50 represent small, medium and large effect sizes respectively, the sampling error corrected correlation (r) between task orientation and positive affect can be considered as medium to large (r= .36). However, when the same correlation is corrected for both sampling and measurement errors (ρ), it can be characterised as large (ρ= .55). Furthermore, the strength of this correlation is indicated by the fact that 149 missing or yet unknown studies averaging null results (i.e. ρ= 0) would have to exist to bring the corrected correlation down to ρ= .10. Table 3 shows that the credibility interval for this correlation is quite large, that sampling error and measurement error account for only 33% of the observed variance, and that the homogeneity test Q is significant. Therefore, the variance of the correlation between task orientation and positive affect is not homogeneous or consistent across studies, suggesting the need to look for moderators.

14 Table 3 also shows that the correlation corrected for sampling error between task orientation and negative affect is small (r= -.11), whereas the correlation corrected for both sampling and measurement errors can be characterised as small to medium (ρ= -.18). Furthermore, 28 studies averaging null results are needed to bring down the population correlation to the value of .10. The credibility interval is large and includes the value of zero, suggesting that there is a significant variation in this correlation across studies. This finding is in accordance with the results provided by the 75% rule and the Q test and shows the need to search for moderators. Both the sampling error corrected and the population correlations between ego orientation and positive affect are very small (r= .07; ρ= .10). Furthermore, the homogeneity indices show that the variance of the population correlation is heterogeneous and, therefore, moderators should be sought. Both types of correlation between ego orientation and negative affect can be also characterised as very small (r= .02; ρ= .04). In addition, the confidence interval of the population correlation has its lower bound the value of zero, suggesting that the relationship between ego orientation and negative affect is not always different from zero. The amount of variance in the population correlation is very small and homogeneous, as the 75% rule and the Q test show. Therefore, there is no need to look for moderators for the ego orientation-negative affect relationship. Moderator analysis A moderator is a variable causing differences (variations) in the correlation between two other variables across studies. In meta-analysis, a hypothesised moderator variable is used to group the observed correlations into subsets. A variable can be characterised as a moderator when the corrected correlation will differ across the subsets, and the variance of this correlation will be lower in the subsets than in the data as a whole (1). These conditions imply that a moderator variable can, to a certain extent, account for the variation of a

15 correlation across studies. However, when comparing the mean correlations between two subsets in a moderator analysis, there is a possibility that differences between the correlations are due to second-order sampling error (1). That is, there is a possibility for sampling error in the meta-analytic estimates of corrected correlations and variances especially when the number of studies in moderator subsets is not very large. In such cases, second-order sampling error may occur because the results will depend on which studies randomly happen to be available. To check for second-order sampling error, we employed the z test, proposed by Hunter and Schmidt (1; p 438). The .05 critical value of the two-tailed z test is 1.96. Moderator analysis was conducted on the correlations between task orientation and positive affect, task orientation and negative affect, and ego orientation and positive affect, because their variances were found to be heterogeneous. The choice of potential moderators was limited by the few moderating characteristics that were consistently provided by the authors of the primary studies. The results of the moderator analysis are presented in Table 4. A variable was characterised as a moderator when the confidence intervals of the correlations in the different groups did not overlap substantially (1). A significant z test was also used to indicate whether a variable was a moderator of a particular goal orientation-affect relationship. In view of the arguments presented earlier, the publication status of the studies (published articles vs. conference presentations) was coded as a potential moderator. In general, no significant differences emerged between published and unpublished studies as indicated by the non significant z test. When such differences were found (i.e. in the ego orientation-positive affect relationship) the confidence intervals overlapped considerably. In brief, these results justified the use of both types of studies in the meta-analysis. A potential moderator of the task orientation-negative affect relationship is the degree of arousal when negative affect is experienced. As noted before, Russell (12) has classified

16 emotions not only on a pleasantness-unpleasantness dimension, but also on a high arousallow arousal dimension. The primary studies of this meta-analysis included negative affect of both high arousal (e.g. anger, anxiety, tension) and low arousal (e.g. depression, boredom). The results of the moderator analysis showed that the task orientation-negative affect relationship was greater when low arousal rather than high arousal negative affect was experienced. Nationality was also coded as a potential moderator with 33 of the 41 samples being American or British. The need to examine cross-cultural variations in achievement goals has been discussed by Duda and Allison (67b) and Whitehead (68b). As the results in Table 4 show, the only difference found was in the task orientation-positive affect relationship which was higher in the British than in the American samples. Also, most of the participants in the primary studies could be classified into school physical education, recreational, or competitive sport groups. Therefore, the setting of physical activity was coded as a potential moderator. The results revealed that the task orientation-positive affect relationship was higher in school physical education and competitive sport than in recreational contexts. Also, the task orientation-negative affect and ego orientation-positive affect relationships were higher in competitive sport than in recreational settings. Another potential moderator was the time frame of affect, that is, whether it was measured after an athletic event (e.g. after 800m running; 60) or independent of context (how one generally feels in one's physical activity context; e.g. 63). The results showed that when positive affect was measured independent of context (typical affect) and not as a post-event feeling (post-exercise affect), it correlated higher with both task and ego orientations. Similarly, when negative affect was measured independent of context rather than as a postexercise feeling, the task orientation-negative affect relationship was higher, but in a negative direction.

17 A recent multisample confirmatory factor analysis has shown that achievement goals may not be conceptualised similarly across different age groups (68c). To examine whether age is a potential moderator, the participants of the primary studies were classified into school (i.e. aged 9-18) vs. university students. Other age categorisations were not feasible because information about the age of the participants was provided at a group rather than individual level. The results in Table 4 show that the only significant difference found was in the task orientation-positive affect relationship that was higher in the younger sample. It is preferable that moderators are examined in combination rather than in isolation from each other. However, as Hunter and Schmidt (1) have noted, often meta-analyses do not have enough studies to analyse interactions among moderators and, unfortunately, the present meta-analysis is no exception. However, an interaction that was feasible to test with the present data was that between high arousal of negative affect and age, in the context of the relationship between task orientation and negative affect. Specifically, under conditions of high arousal of negative affect (e.g. anxiety, anger, tension), the relationship between task orientation and negative affect was significantly higher for school students than university students. The gender of the participants may have been a moderator variable because gender differences have been reported in the literature with regard to goal orientations (e.g., 27). However, although 80% of the samples included both males and females, the primary studies did not give separate results for the two genders and, therefore, moderator analysis was not possible. The kind of achievement goals questionnaire was not examined as a potential moderator because the conceptually very similar TEOSQ and POSQ were used in 88% of the samples and, therefore, moderator analysis with less compatible scales would have created greatly imbalanced subsets, thus increasing the probability of second-order sampling error (1). However, in view of the criticisms that have been addressed to the SOQ (e.g. 33, 34), we

18 also conducted the meta-analysis excluding the studies which have used this questionnaire. The results were very similar justifying, therefore, the use of the goal and win subscales of the SOQ in this study. The type of affect questionnaire was not tested as a likely moderator because most studies did not employ any established questionnaire. Structural Equation Modelling (SEM) Analysis The corrected mean correlations among task orientation, ego orientation, positive affect, and negative affect were analysed using SEM in order to look concurrently at the relationships among all the variables. The combination of SEM and meta-analysis is a very promising method because the two analyses can compliment each other (65b). As argued previously, SEM can capture interdependencies between variables that meta-analysis cannot examine, because the latter can look at the relationship of two variables at a time. On the other hand, meta-analysis removes the effects of artifacts from data before SEM analysis. To obtain the necessary full correlation matrix1 for the SEM, the correlations between task and ego goal orientations and between positive and negative affect also had to be metaanalysed. The population correlations, corrected for sampling error and measurement error, were ρ= .14 for the task-ego relationship, and ρ= -.46 for the relationship between positive and negative affect. A dilemma that arises when conducting SEM on meta-analysed correlations is whether to estimate the significance of the path coefficients or not. If it is assumed that the corrected correlations represent population values, there is no need to test the significance of the path coefficients because significance testing is appropriate only for sample estimates. However, the degree to which corrected correlations reflect population correlations depends on how large the number of studies is in a given meta-analysis. As Hunter and Schmidt (1) have suggested, the larger the number of studies in a meta-analysis, the more accurate the correction of study artifacts and the more likely that the results of the meta-analysis will

19 reflect population values. Because our meta-analysis was not based on a large number of studies, the reported population correlations are rather approximations of the real correlations and, therefore, we decided to estimate the standard errors of the path coefficients. However, most of the corrected correlations of the matrix are based on a different sample size. Following the suggestion of Hom et al. (65b), whose meta-analysis was based on a similar number of studies, we decided to use the sample size (N=2980) only of studies which examined all the correlations of Table 3. Therefore, only 15 of the 41 independent samples were retained for the SEM analysis. The EQS (Version 5.0; 69) statistical software was employed and the data were analysed using maximum likelihood analysis. The purpose of the SEM analysis was to examine whether the links between two variables that were independently estimated through meta-analysis, would remain unchanged at the presence (and potential influence) of other variables. The results of the SEM are presented in Figure 1 and show that the paths had quite similar values to those of the meta-analysed correlations, with the exception of the ego 2 orientation-positive affect relationship which was set to zero [Fit indices: x (1)= 64.44,

p<0.01; Bentler-Bonnet Normed Fit Index = .88, Comparative Fit Index = .96; Goodness of Fit Index= .98; Root Mean Square Residual= .05]. Discussion The purpose of the present study was to examine the conflicting evidence in the literature regarding the relationships between achievement goals and affect. The significance of identifying and measuring the impact of different motivational variables on affect stems from the fact that motivation and emotion play a central part in individuals' experiences in physical activity (14, 70, 4) and their relationship needs to be more fully understood. Four correlations which often appear in the achievement goals literature were meta-analysed: task orientation-positive affect, task orientation-negative affect, ego orientation-positive affect,

20 and ego orientation-negative affect. The correlations were gathered from 41 independent samples (Total N= 7950) and were corrected for both sampling and measurement errors using the formulas of Hunter and Schmidt (1). Task Orientation and Positive Affect The relationship between task orientation and positive affect was verified through the meta-analysis by demonstrating that the relationship was relatively high (ρ= .55) when both sources of error were removed. A theoretical explanation for the high correlation is that because individuals with a predominant task orientation focus on personal effort and have self-referenced standards of comparison, they are more likely to report perceived success and improved performance which will subsequently lead to high levels of positive affect (71). However, the meta-analysis showed that the variance of this correlation was not homogeneous suggesting that some moderators may be in operation. Four variables were found to moderate the task orientation-positive affect relationship: the participants’ nationality, the setting of physical activity, the time frame of affect and the age of the participants. With regard to the first moderator, the results showed that more pleasant emotional experiences in physical activity were related to high levels of task orientation, significantly more in British than in US samples. This finding cannot be explained easily because the two countries have many more cultural similarities than differences, although Whitehead (68b) has suggested that the American sporting culture appears more overtly competitive, in the ego-oriented sense, than that in Britain. Whether in the American culture task orientation is downplayed and whether this can account for the difference in the correlations remains to be seen. It would have been desirable, had enough studies existed, to conduct moderator comparisons among more diverse cultures. According to Duda and Allison (67b), cross-cultural research is needed to shed light on how cultural factors can shape the social context of physical activity.

21 It is quite possible that the difference between the British and the US samples was confounded with differences in physical activity settings because almost all the British participated in school physical education while most of the Americans took part in recreational activities. However, some of these ‘recreational’ activities were also required classes, so more needs to be known here. Nevertheless, in support of the above argument, when the setting of physical activity was coded as a potential moderator, the relationship between task orientation and positive affect was found to be higher in school physical education than in so-called recreational classes. Similarly, this relationship was higher in school physical education than in competitive sport. A possible explanation for this moderating effect has to do with the different nature of these contexts. Specifically, school physical education is a salient achievement context where individuals have to participate and where the emphasis can be on inter-personal comparison and/or on individual improvement and effort (72). These factors will produce wide variability in the task orientation scores of the individuals that will reflect in higher correlations between task orientation and positive affect. On the contrary, in recreational and competitive sport contexts there is a greater degree of voluntary participation that is likely to produce high levels of task orientation. Therefore, it is expected that the variability in the amount of task orientation in these contexts will be lower than in school physical education that will reflect on lower correlations between task orientation and some third variables such as affect. Some support for this argument is provided by the standard deviation of the task orientation scores which is larger in school physical education than in other contexts. The moderator analysis also showed that when positive affect was measured independent of context, it correlated more highly with task orientation than when it was measured after an intervention or a competition. This result shows that situational factors may lead task-oriented people to experience less positive affect than would normally be

22 expected. Such situational factors, for example, can be motivational climates that promote interpersonal comparison because they can lower the degree to which an individual will be focused on effort and individual improvement in a particular situation (73). This will reflect on lower correlations between task orientation and positive affect. Furthermore, the achievement goals literature (e.g. 6) emphasises that competition and especially sport competition, makes success and failure very salient thus enhancing the probability that one will experience less positive emotions. However, it seems that when participants are requested to recall their general emotional experiences in sport, high levels of positive affect are reported among those high in task orientation, thus providing further support to authors emphasising the adaptive role of task orientation (e.g. 14). Age was also a moderator of the task orientation-positive affect relationship which was higher in younger (school age) than older (university age) athletes. This difference can be attributed to the relatively higher pressures for winning at the senior levels of sport that can undermine task orientation. For example, Gould, Jackson, and Finch (74) and Scanlan, Stein, and Ravizza (75) found that pressures and expectations to win and the fear of failure were significant sources of stress among elite figure skaters. Task Orientation and Negative Affect The relationship between task orientation and negative affect was also meta-analysed. The population correlation between the two variables can be characterised as small to medium (ρ= -.18), bearing in mind Cohen’s (66) suggestions. This correlation may be smaller than expected or hypothesised in the literature. However, an important moderator that could account for the relatively low effect size is the nature of negative affect experienced. Specifically, as noted earlier, Russell (12) has suggested that affect can be classified along a high arousal-low arousal dimension. The present analysis showed that the arousal dimension was a moderator of the task orientation-negative affect relationship in that the lower the

23 arousal of negative affect, the higher (and more negatively) negative affect correlated with task orientation. It seems, therefore, that the theoretically hypothesised negative relationship between task orientation and negative affect is more likely to occur with low arousal negative emotions such as boredom and depression, rather than with high arousal negative emotions such as anxiety and tension. The reason may be that when examining intense negative feelings, such as anxiety and tension, other factors than goal orientations may have a higher predictive validity. For example, self-confidence and self-efficacy have been found to strongly relate to competitive anxiety (37). An alternative explanation for the moderating effect of the arousal of negative affect has to do with the fact that task orientation has been mainly measured in the literature as the degree of effort that one exerts in an achievement situation. Effort is very likely to be negatively related with boredom (the most common measure of low arousal negative affect employed in the primary studies). In contrast, individuals may feel anxious even when they try hard because, for example, they lack self-confidence. Furthermore, in the moderator analysis of the task orientation-negative affect relationship, it was possible to examine the interaction between the variables of high arousal of negative affect and age. A significant interaction emerged in that when high arousal negative affect was experienced, the above relationship was higher in school students rather than university students. In other words, younger task-oriented individuals were likely to experience more anxiety and tension than older task-oriented individuals. The reason for this difference may be that older individuals are more experienced and can control and cope more effectively with aversive feelings (76). Another moderator of the task orientation-negative affect relationship was the time frame of negative affect, that is, whether it was measured after an event or independent of context. The results showed that task-oriented individuals experienced less negative affect

24 when the latter was assessed independent of context than immediately after a specific athletic event. Similar to the correlation between task orientation and positive affect, it is possible that situational factors may cause more negative affective experiences for individuals with a predominant task orientation than would normally be expected. Competitive situations are also more likely to threaten valued goals thus producing high levels of negative affect (77). Lastly, the moderator analysis showed that low negative affect in competitive sport was related to higher task orientation, significantly more than in recreational settings. An explanation may relate to the fact that in sport there is a greater emphasis on learning and individual improvement of sport skills (which nurture task orientation) whereas in recreational contexts participation relates more to fitness and affiliation motives (77). Another factor could be that the assessment of achievement goals has taken place in nonachievement settings, such as recreational physical activity, thus raising the issue of the applicability of achievement goals in such contexts. Ego Orientation and Positive Affect The third relationship of interest in the present study was that between ego orientation and positive affect. The correlation between the two variables was very low (ρ= .10) even when both sources of error were removed. A better understanding of their relationship could have been provided if competence was measured as a moderator. According to social cognitive theories (6), different motivational patterns can be observed in high and low perceived competence ego-oriented individuals. However, the testing of the potential moderating role of perceived competence was not possible because a major limitation of most of the primary studies was that they did not provide separate correlations for different levels of perceived competence. A variable that was found to moderate the ego orientation-positive affect relationship was the temporal assessment of positive affect. When it was assessed independent of context,

25 it correlated higher with ego orientation than when it was measured after a competition. Again, a plausible explanation relates to the influence of situational factors in that less positive emotions can be induced in ego-oriented individuals in the evaluative environment of competition than in non-evaluative situations (e.g. training sessions). This is because in evaluative situations ego-oriented individuals are more likely to feel insecure about the adequacy of their ability (6). Ego Orientation and Negative Affect The fourth relationship that the present study meta-analysed was that between ego orientation and negative affect. The corrected mean correlation was very small (ρ= .04), while its confidence interval included the value of zero, indicating that the relationship is not always different from zero. The low links between ego orientation and positive and negative affect can partly be explained in relation to the work of Elliot and Harackiewicz (79). These authors have suggested that the conventional achievement goal dichotomy be expanded to include both approach (striving to attain success) and avoidance (striving to avoid failure) components within the ego goal orientation. In an unpublished meta-analysis by Elliot (cited in 79) it was shown that the maladaptive influence of ego goals is more salient in avoidance ego goals. When the two types of ego goals were confounded in the literature, results generally showed no significant links with motivational indices. It would be interesting to examine the usefulness of Elliot and Harackiewicz's (79) approach to achievement motivation in the area of physical activity. Lastly, the variance in the corrected correlation between ego orientation and negative affect was very small and homogeneous across studies suggesting that there were no moderators. Concurrent Relationships The corrected correlations derived from a subset of the primary studies were inserted into a structural equation modelling analysis in order to examine whether the relationship

26 between any two variables would change in the presence of the other variables in the model. Unlike meta-analysis, structural equation modelling can give a picture of the concurrent relationships between achievement goals and emotions. The findings showed that the impact of task orientation on both modes of affect was independent of the presence of ego orientation in the model. Similarly, the impact of ego orientation on negative affect was small and independent of the presence of task orientation in the model. However, when the link between ego orientation and positive affect was tested in the presence of the other variables, it became non-existent. This indicates that the impact of achievement goals on positive affect is predominantly exerted through task orientation. The model had a good fit to the data when the errors of positive and negative affect were correlated. This indicates that emotions share specific properties in addition to being part of a general dimension, that of positive or negative affect, a finding that is not surprising. Similar argument has been put forward by Diener, Smith, and Fujita (80) who have shown that global emotional dimensions cannot account for all the relations among discrete emotions. An equally plausible interpretation for the correlated errors is that the two goal orientations cannot explain on their own all the variance in emotional outcomes, and therefore, additional research directions have to be followed. Additional future research directions Studies in the area of achievement motivation have often been criticised for ignoring the theoretical and empirical independence of task and ego goal orientations (see 18, 2). The independence of the two goal orientations implies that varying degrees of task and ego orientations can be found within the same person and permits the categorisation of individuals in groups with different combinations of goal orientations (e.g. high task/ low ego, low task /low ego). Fox et al. (18) predicted that the motivational impact of the two goal orientations in combination may be different from their effects examined separately. It would

27 be interesting to examine this hypothesis, as far as the affective experiences of athletes are concerned, through comparative meta-analyses when more studies on goal groups are conducted. Furthermore, in order to have a more comprehensive view of the emotionachievement motivation relationship from a social cognitive perspective, it would be useful to meta-analyse research findings on the links between motivational climates and emotions. This is consonant with the suggestion by Treasure and Roberts (73) that research based on social cognitive theories of achievement should examine the influence of both dispositional goals and environmental factors (i.e. motivational climates) on the cognitive and affective responses of physical activity participants. Furthermore, it would be informative if longitudinal studies were carried out to investigate possible reciprocal relations between emotions and achievement motivation. Almost all the available studies to date are based on cross-sectional data which do not permit inferences regarding causal relations. In summary, the purpose of the present study was to examine the strength of links between ego and task achievement goals with positive and negative affect in physical activity. Drawing on the responses of nearly 8000 participants (a sample size larger than in many other meta-analyses in psychology), this study showed that these links are small (with the exception of the task orientation-positive affect relationship) and heterogeneous (with the exception of the ego orientation-negative affect relationship). Future meta-analyses in this area, which will include new studies, should combine their findings with those reported here so that a more comprehensive, or second-order meta-analysis is performed (1). Despite the limited information that the primary studies provided, the present meta-analysis offers some answers to contradictory findings in the literature, and along with its moderator analysis, highlights issues for more informative future research which can assist in theoretical modifications and expansions.

28

29 References 1. Hunter, J.E. & Schmidt, F.L. (1990). Methods of meta-analysis. Newbury Park, CA: Sage. 2. Hardy, L., Jones, G., & Gould, D. (1996). Understanding psychological preparation for sport: Theory and practice of elite performers. Chichester: John Wiley. 3. Biddle, S.J.H. & Mutrie, N. (1991). Psychology of physical activity and exercise: A healthrelated perspective. London: Springer Verlag. 4. Vallerand, R.J. (1983). On emotion in sport: Theoretical and social psychological perspectives. Journal of Sport Psychology, 5, 197-215. 5. Dweck, C.S., & Leggett, E.L. (1988). A social-cognitive approach to motivation and personality. Psychological Review, 95, 256-273. 6. Nicholls, J.G. (1989). The competitive ethos and democratic education. Cambridge, MA: Harvard University Press. 7. Lazarus, R.S. (1991). Emotion and adaptation. New York: Oxford University Press. 8. Clore, G.L., Ortony, A., & Foss, M.A. (1987). The psychological foundations of the affective lexicon. Journal of Personality and Social Psychology, 53, 751-766. 9. Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219-235. 10. Watson, D., Clark, L.A., & Carey, G. (1988). Positive and negative affectivity and their relation to anxiety and depressive disorders. Journal of Abnormal Psychology, 97, 346-353. 11. Watson, D., Clark, L.A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070. 12. Russell, J.A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161-1178. 13. Ames, C. (1992). Achievement goals and classroom motivational climate. In J. Meece & D. Schunk (Eds.), Students’ perceptions in the classroom (pp. 327-348). Hillsdale, NJ: Erlbaum. 14. Duda, J.L. (1996). Maximizing motivation in sport and physical education among children and adolescents: The case for greater task involvement. Quest, 48, 290-302. 15. Roberts, G.C. (1992). Motivation in sport and exercise: Conceptual constraints and convergence. In G.C. Roberts (Ed.), Motivation in sport and exercise (pp. 3-29). Champaign, IL: Human Kinetics.

30 16. *Boyd, M.P., & Yin, Z. (1996). Cognitive-affective sources of sport enjoyment in adolescent sport participants. Adolescence, 31, 383-395. 17. *Duda, J.L., Fox, K., Biddle, S.J.H., & Armstrong, N. (1992). Children’s achievement goals and beliefs about success in sport. British Journal of Educational Psychology, 62, 313323. 18. *Fox, K., Goudas, M., Biddle, S., Duda, J., & Armstrong, N. (1994). Children’s task and ego goal profiles in sport. British Journal of Educational Psychology, 64, 253-261. 19. *Goudas, M., Biddle, S., & Fox, K. (1994). Achievement goal orientations and intrinsic motivation in physical fitness testing with children. Pediatric Exercise Science, 6, 159-167. 20. *Goudas, M., Biddle, S., Fox, K., & Underwood, M. (1995). It ain’t what you do, it’s the way that you do it! Teaching style affects children’s motivation in track and field lessons. The Sport Psychologist, 9, 254-264. 21. *Hom, H.L., Duda, J.L., & Miller, A. (1993). Correlates of goal orientations among young athletes. Pediatric Exercise Science, 5, 168-176. 22. *Newton, M., & Duda, J.L. (1993). The relationship of task and ego orientation to performance-cognitive content, affect, and attributions in bowling. Journal of Sport Behavior, 10, 209-220. 23. *Vlachopoulos, S., & Biddle, S.J.H. (1996). Achievement goal orientations and intrinsic motivation in a track and field event in school physical education. European Physical Education Review, 2, 158-164. 24. *Duda, J.L., & Nicholls, J.G. (1992). Dimensions of achievement motivation in schoolwork and sport. Journal of Educational Psychology, 84, 290-299. 25. *Rethorst, S., & Duda, J.L. (1993). Goal orientations, cognitions, and emotions in gymnastics. In S. Serpa, J. Alves, V. Ferreira, & A. Paula-Brito (Eds.), Proceedings: VIII World Congress of Sport Psychology (pp. 379-382). Lisbon: International Society of Sport Psychology. 26. *Vlachopoulos, S., Biddle, S.J.H., & Fox, K. (1996). A social cognitive investigation into the mechanisms of affect generation in children’s physical activity. Journal of Sport and Exercise Psychology, 18, 174-193. 27. *Duda, J.L., Chi, L., Newton, M.L., Walling , M.D., & Catley, D. (1995). Task and ego orientation and intrinsic motivation in sport. International Journal of Sport Psychology, 26, 40-63. 28. *Kavussanu, M., & Roberts, G.C. (1996). Motivation in physical activity contexts: The relationship of perceived motivational climate to intrinsic motivation and self-efficacy. Journal of Sport and Exercise Psychology, 18, 264-281.

31 29. Roberts, G.C., Treasure, D.C., & Balague, G. (1998). Achievement goals in sport: The development and validation of the Perception of Success Questionnaire. Journal of Sports Sciences, 16, 337-347. 30. Gill, D.L., & Deeter, T.E. (1988). Development of the Sport Orientation Questionnaire. Research Quarterly for Exercise and Sport, 59, 191-202. 31. Ewing, M.E. (1981). Achievement motivation and sport behaviour of males and females. Unpublished doctoral dissertation, University of Illinois, Urbana. 32. Maehr, M.L., & Nicholls, J.G. (1980). Culture and achievement motivation: A second look. In N. Warren (Ed.), Studies in cross-cultural psychology (221-267). New York: Academic Press. 33. Duda, J.L. (1992). Motivation in sport settings: A goal perspective approach. In G.C. Roberts (Ed.), Motivation in sport and exercise (pp. 57-91). Champaign, IL: Human Kinetics. 34. Marsh, H.W. (1994). Sport motivation orientation: Beware of jingle-jangle fallacies. Journal of Sport and Exercise Psychology, 16, 365-380. 35. McAuley, E., Duncan, T.E., & Tammen, V.V. (1989). Psychometric properties of the intrinsic motivation inventory in a competitive sport setting: A confirmatory factor-analysis. Research Quarterly for Exercise and Sport, 60, 48-58. 36. Martens, R. (1977). Sport competition anxiety test. Champaign, IL: Human Kinetics. 37. Martens, R., Burton, D., Vealey, R.S., Bump, L.A., & Smith, D.E. (1990). Development and validation of the Competitive State Anxiety Inventory-2. In R. Martens, R.S. Vealey, & D. Burton (Eds.), Competitive Anxiety in Sport (pp. 117-190). Champaign, IL: Human Kinetics. 38. Gauvin, L., & Rejeski, W.J. (1993). The Exercise-Induced Feeling Inventory: Development and initial validation. Journal of Sport and Exercise Psychology, 15, 403-423. 39. McNair, D.M., Lorr, M., & Doppleman, L.F. (1971). Manual for the Profile of Mood States. San Diego: Educational and Industrial Testing Service. 40. Tabachnick, B.G., & Fidell, L.S. (1989). Using multivariate statistics (2nd ed.). New York: Harper Collins Publishers. 41. Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183192. 42. *Andree, K.V., & Whitehead, J. (1995). The interactive effect of perceived ability and dispositional or situational achievement goals on intrinsic motivation in young athletes. Paper presented at the meeting of the North American Society for the Psychology of Sport and Physical Activity, Monterey, CA.

32 43. *Biddle, S.J.H., Akande, A., Vlachopoulos, S., & Fox, K. (1996). Towards an understanding of children’s motivation for physical activity: Achievement goal orientations, beliefs about sport success, and sport emotion in Zimbabwean children. Psychology and Health, 12, 49-55. 44. *Boyd, M., & Callaghan, J. (1994). Task and ego goal perspectives in organized youth sport. International Journal of Sport Psychology, 22, 411-424. 45. *Dorobantu, M., & Biddle, S.J.H. (1997). The influence of situational and individual goals on the intrinsic motivation of Romanian adolescents towards physical education. European Yearbook of Sport Psychology, 1, 148-165. 46. *Gill, D.L., Dzewaltowski, D.A., & Deeter, T.E. (1988). The relationship of competitiveness and achievement orientation to participation in sport and nonsport activities. Journal of Sport and Exercise Psychology, 10, 139-150. 47. *Grieve, F.G., Whelan, J.P., Kottke, R., & Meyers, A.W. (1994). Manipulating adults’ achievement goals in a sport task: Effects on cognitive, affective and behavioural variables. Journal of Sport Behavior, 17, 227-245. 48. *Hall, H.K., & Earles, M. (1995). Motivational determinants of interest, and perceptions of success in school physical education. Paper presented at the meeting of the North American Society for the Psychology of Sport and Physical Activity, Monterey, CA. 49. *Kim, B.J., & Gill, D.L. (1997). A cross-cultural extension of goal perspective theory to Korean youth sport. Journal of Sport and Exercise Psychology, 19, 142-155. 50. *Martin, J.J., & Gill, D.L. (1991). The relationships among competitive orientation, sport-confidence, self-efficacy, anxiety, and performance. Journal of Sport and Exercise Psychology, 13, 149-159. 51. *Morgan, K., & Carpenter, P. (1996). Motivational climate, achievement goals, beliefs about success, task choice and affective responses in physical education classes. Paper presented at the annual meeting of the British Association of Sport and Exercise Sciences, Lilleshall, U.K. 52. *Newton, M., & Duda, J.L. (1995). Relations of goal orientations and expectations on multidimensional state anxiety. Perceptual and Motor Skills, 81, 1107-1112. 53. *Nyheim, M., Kavussanu, M., Roberts, G.C., & Treasure, D. (1996). The relationship between goal orientations, beliefs about the causes of success, and satisfaction in summer sports camp participants. Paper presented at the joint conference of the North American Society for the Psychology of Sport and Physical Activity and the Canadian Society for Psychomotor Learning and Sport Psychology, Ontario, Canada. 54. * Ommundsen, Y., Roberts, G.C., & Kavussanu, M. (1998). Perceived motivational climate and cognitive and affective correlates among Norwegian athletes. Journal of Sports Sciences, 16, 153-164.

33 55. *Pemberton, C.L., Petlichkoff, L.M., & Ewing, M.E. (1986). Psychometric properties of the achievement orientation questionnaire. Paper presented at the meeting of the North American Society for the Psychology of Sport and Physical Activity, Scottsdale, Arizona. 56. *Peng, W., Chi, L. (1996). The effects of goal orientations on attributions and emotions after a one-on-one basketball game. Paper presented at the joint conference of the North American Society for the Psychology of Sport and Physical Activity and the Canadian Society for Psychomotor Learning and Sport Psychology, Ontario, Canada. 57. *Roberts, G.C., Treasure, D.C., & Kavussanu, M. (1996). Orthogonality of achievement goals and its relationship to beliefs about success and satisfaction in sport. The Sport Psychologist, 10, 398-408. 58. *Roberts, G.C., & Ommundsen, Y. (1996). Effect of goal orientation on achievement beliefs, cognition and strategies in team sport. Scandinavian Journal of Medicine & Science in Sports, 6, 46-56. 59. *Swain, A.B.J., & Jones, G. (1992). Relationships between sport achievement orientation and competitive state anxiety. The Sport Psychologist, 6, 42-54. 60. *Vlachopoulos, S., & Biddle, S.J.H. (1997). Modelling the relation of goal orientations to achievement-related affect in physical education: Does perceived ability matter? Journal of Sport and Exercise Psychology, 19, 169-187. 61. *Vlachopoulos, S., Biddle, S.J.H., & Fox, K. (1997). Determinants of emotion in children’s physical activity: A test of goal perspectives and attribution theories. Pediatric Exercise Science, 9, 65-79. 62. *Williams, L., & Gill, D.L., (1995). The role of perceived competence in the motivation of physical activity. Journal of Sport and Exercise Psychology, 17, 363-378. 63. *Xiang, P., Lee, A.M., & Solmon, M.A. (1997). Achievement goals and their correlates among American and Chinese students in physical education. Journal of Cross-Cultural Psychology, 28, 645-660. 64. *Yin, Z., & Boyd, M.P. (1994). Achievement orientation and its psychological correlates in youth sport. Applied Research in Coaching and Athletics Annual, 50-65. 65a. *Yin, Z. (1993). Influence of socializing agents, achievement goal orientation, and perceived competence as sources of sport enjoyment. Research Quarterly for Exercise and Sport, 64, (Suppl. 1), 105. 65b. Hom, P.W., Caranikas-Walker, F., Prussia, G.E., & Griffeth, R.W. (1992). A metaanalytical structural equations analysis of a model of employee turnover. Journal of Applied Psychology, 77, 890-909. 66. Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159.

34 67b. Duda, J.L., & Allison, M.T. (1990). Cross-cultural analysis in exercise and sport psychology: A void in the field. Journal of Sport and Exercise Psychology, 12, 114-131. 67a. Whitener, E.M. (1990). Confusion of confidence intervals and credibility intervals in meta-analysis. Journal of Applied Psychology, 75, 315-321. 68a. Sackett, P.R., Harris, M.M., & Orr, J.M. (1986). On seeking moderator variables in the meta-analysis of correlational data: A Monte Carlo investigation of statistical power and resistance to Type I error. Journal of Applied Psychology, 71, 302-310. 68b. Whitehead, J. (1995). Multiple achievement orientations and participation in youth sport: A cultural and developmental perspective. International Journal of Sport Psychology, 26,431-452. 68c. Chi, L.K., & Duda, J.L. (1995). Multisample confirmatory factor analysis of the Task and Ego Orientation in Sport Questionnaire. Research Quarterly for Exercise and Sport, 66, 91-98. 69. Bentler, P.M. (1995). EQS Structural Equations Programme Manual. Encino, CA: Multivariate Software, Inc. 70. Hanin, Y. L. (1997). Emotions and athletic performance: Individual zones of optimal functioning model. European Yearbook of Sport Psychology, 1, 29-72. 71. Elliot, E.S., & Dweck, C.S. (1988). Goals: An approach to motivation and achievement. Journal of Personality and Social Psychology, 54, 5-12. 72. Biddle, S.J.H., Cury, F., Goudas, M., Sarrazin, P., Famose, J.P., & Durand, M. (1995). Development of scales to measure perceived physical education class climate: A crossnational project. British Journal of Educational Psychology, 65, 341-358. 73. Treasure, D.C., & Roberts, G.C. (1995). Applications of achievement goal theory to physical education: Implications for enhancing motivation. Quest, 47, 475-489. 74. Gould, D., Jackson, S., & Finch, L. (1993). Sources of stress in national champion figure skaters. Journal of Sport and Exercise Psychology, 15, 134-159. 75. Scanlan, T.K., Stein, G.L., & Ravizza, K. (1991). An in depth study of former élite figure skaters: II. Sources of stress. Journal of Sport and Exercise Psychology, 13, 103-120. 76. Compas, B.E., Banez, G.A., Malcarne, V., & Worsham, N. (1991). Perceived control and coping with stress: A developmental perspective. Journal of Social Issues, 47, 23-34. 77. Lewthwaite, R. (1990). Threat perception in competitive trait anxiety: The endangerment of important goals. Journal of Sport and Exercise Psychology, 12, 280-300. 78. Ryan, R.M., Frederick, C.M., Lepes, D., Rubio, N., & Sheldon, K.M. (1997). Intrinsic motivation and exercise adherence. International Journal of Sport Psychology, 28, 335-354.

35 79. Elliot, A.J. & Harackiewicz, J.M. (1996). Approach and avoidance achievement goals and intrinsic motivation: A mediational analysis. Journal of Personality and Social Psychology, 70, 461-475. 80. Diener, E., Smith, H., & Fujita, F. (1995). The personality structure of affect. Journal of Personality and Social Psychology, 69, 130-141. 81. Rothstein, H.R., Schmidt, F.L., Erwin, F.W., Owens, W.A., & Sparks, C.P. (1990). Biographical data in employment selection: Can validities be made generalizable? Journal of Applied Psychology, 75, 175-184. 82. Cudeck, R. (1989). Analysis of correlation matrices using covariance structure models. Psychological Bulletin, 105, 317-324. 83. Ntoumanis, N. & Biddle, S.J.H. (in press). A review of motivational climate in physical activity. Journal of Sports Sciences.

36 Table 1: Stem and Leaf display of the observed correlations in the primary studies. Observed Correlations Task OrientationPositive affect (Mdn=.36, min=.00, max=.61)

Task OrientationNegative affect (Mdn=-.11, min=-.35, max=.17)

Ego OrientationPositive affect (Mdn=.06, min=-.35, max=.44)

Ego OrientationNegative affect (Mdn=.02, min=.12, max=.23)

Stem

Leaf

.0 .1 .2 .3 .4 .5 .6 -.3 -.2 -.1 -.0 .0 .1 -.2 -.1 -.0 .0

56 36 1788889 001466999 2456688 2447 1 35555 004568 0123445 11778 001122369 7 4 01388 244

.1 .2 .3 -.1 -.0 -.0 .0 .0 .1 .1 .2

22355567789 22344457 26 9 12 578 1222234 00222334 5569 1334 7 1113

37 Table 2: Summary of the coded characteristics of the studies that were included in the meta-analysis. Studies

42. Andree & Whitehead (1995) 43. Biddle et al. (1996) 44. Boyd & Callaghan (1994) 16. Boyd & Yin (1996)

TEOSQ IMI

Enjoyment/ Interest

Pressure/ Tension

Sex Country Education Status (M= mean age) M+F England School, Competitive age=11-17

TEOSQ

Satisfaction

Boredom

M+F

TEOSQ

Performance satisfaction

TEOSQ

Enjoyment

45. Dorobantu & Biddle (1997) 24. Duda & Nicholls (1992) 17. Duda et al. (1992) 27. Duda et al. (1995) 27. Duda et al. (1995) 18. Fox et al. (1994) 46. Gill et al. (1988)

TEOSQ IMI

Enjoyment/ Interest

TEOSQ

Satisfaction

Boredom

TEOSQ

Satisfaction

Boredom

TEOSQ IMI

Enjoyment/ Interest Enjoyment/ Interest Satisfaction

46. Gill et al. (1988)

Goals Affect scale Scale

TEOSQ IMI TEOSQ SOQ

SCAT

SOQ

SCAT

19. Goudas et TEOSQ IMI al. (1994)

Positive Affect

Enjoyment/ Interest

Negative Affect

M

Learned helpless affect

M

School, M=15

USA

School, M=15

M+F England School, M=11 Pressure/ M+F USA University, Tension M=20 Pressure/ M+F USA University, Tension M=21 Boredom M+F England School, M=11 Competitive M+F USA School, Trait grades=9Anxiety 12 Competitive M+F USA University Trait Anxiety Pressure/ M+F England School, Tension age=12-15

Task- Task- Ego- Ego- Task Ego PA NA PA NA α α

Conference

139

0.45

0.03

0.13

Typical

Published

159

0.44 -0.35 0.07

0.02

Typical

Published

91

.00

Typical

Published

Typical

Wide variety

NA α

0.61

0.62

0.88

0.81

0.85

0.84

231

0.31 -0.33 0.05 -0.03 0.85

0.78

0.84

Published

145

0.39

0.77

0.78

0.73

Typical

Published

207

0.34 -0.28 0.05

.00

0.89

0.86

0.82

0.71

Physical Education Recreation

Typical

Published

142

0.54 -0.24 0.06

0.23

0.72

0.78

0.84

0.70

Typical

Published

107

0.21 -0.14 -0.24 0.21

0.72

0.82

0.91

0.83

Recreation

Typical

Published

121

0.36 -0.11 0.14 -0.02 0.83

0.78

0.77

0.82

Physical Education Physical Education

Typical

Published

231

0.57

Typical

Published

Recreation

Typical

Physical Education

0.80

0.84

0.63

Competitive

0.01

PA α

0.78

M+F Romania School, Physical age=15-16 Education M+F

N

Typical

Zimba- School, Competitive bwe age=12-14 Competitive USA School, M=11 USA

Type of Publication affect Status

0.09

-0.18

-0.2

0.26 -0.04

266

0.17

-0.02

Published

86

-0.07

0.03

PostPublished Exercise

255

0.3

0.06 -0.18 0.21

0.71

38 20. Goudas et TEOSQ IMI Enjoyment/ al. (1995) Interest 47. Grieve et TEOSQ POMS Short al. (1994) version 48. Hall & Earles (1995) 48. Hom et al. (1993) 28. Kavussanu & Roberts (1996) 28. Kavussanu & Roberts (1996) 49. Kim & Gill (1997) 50. Martin & Gill (1991) 51. Morgan & Carpenter (1996) 22. Newton & Duda (1993) 52. Newton & Duda (1995) 53. Nyheim et al. (1996) 54. Ommundsen et al. (1998) 55. Pemberton et al. (1986)

TEOSQ

Satisfaction

Pressure/ Tension Anxiety, Depression, Confusion Anger Boredom

TEOSQ

Satisfaction

Boredom

England School, M=13 M+F USA University, M=23

Physical Education Psychology students

PostPublished Exercise PostPublished Exercise

M+F England School, M=13 M+F USA School, M=11 M+F USA University

Physical Education Recreation

Typical

Conference

Typical

Recreation

University, Recreation M= 23

F

POSQ

IMI

Enjoyment/ Interest

Pressure/ Tension

POSQ

IMI

Enjoyment/ Interest

Pressure/ Tension

TEOSQ IMI

Enjoyment/ Interest

M+F S. Korea School, M= 14 Cognitive M USA School, state anxiety M=16 Boredom M+F England School, M=14

SOQ

CSAI2

TEOSQ

Satisfaction

TEOSQ

Enjoyment

TEOSQ CSAI2 POSQ

Enjoyment

POSQ

Satisfaction

AOI

SCAT

M+F

USA

24

0.61 -0.08 -0.35 0.17

0.8

0.86

-0.07 0.76

0.85

253

0.48 -0.35 0.13 -0.08 0.90

0.74

0.88

0.72

Published

57

0.39

0.44 -0.11 0.78

0.87

0.77

0.78

Typical

Published

285

0.46 -0.12 0.07 -0.02 0.88

0.90

0.86

0.73

Typical

Conference

131

0.28 -0.14 -0.04 0.13

0.85

0.87

0.88

0.84

Competitive

Typical

Published

344

0.39

0.75

0.67

0.76

Competitive

Typical

Published

73

-0.01

0.05

Physical Education

Typical

Conference

118

0.36 -0.35 0.05

0.03

0.81

0.89

0.88

0.3

0.14

0.75

0.68

0.21

0.73

0.82

0.81

0.86

0.71

0.72

113

0.02

0.03

0.14

Performance M+F worry

USA

University, Wide variety M=21

PostPublished Exercise

47

Competitive M+F State Anxiety M+F

USA

University, Recreation M=20

Typical

Published

107

Typical

Conference

250

0.13

-0.02

Typical

Published

148

0.06

0.12

Typical

Conference

460

School, Recreation age= 9-18 M+F Norway University, Competitive M=21

Competitive M+F Trait Anxiety

USA

USA

School, Recreation age=13-18

-0.25 -0.1

0

0.09

0.05

0.90

0.91 0.56

0.74

0.88

0.79

39 56. Peng & Chi (1996)

TEOSQ

25. Rethorst & Duda (1993)

TEOSQ

57. Roberts et al. (1996) 58. Roberts & Ommundsen (1996) 59. Swain & Jones (1992) 23. Vlachopoulos & Biddle (1996) 60. Vlachopoulos & Biddle (1997) 26. Vlachopoulos et al. (1996)

Satisfaction competence gratitude, pride, happy, confidence, surprise Joy, satisfaction, pride, relief

POSQ

Satisfaction

POSQ

Satisfaction

SOQ

CSAI2 TEOSQ IMI Enjoyment/ Interest

TEOSQ

TEOSQ

EFI

61. TEOSQ Vlachopoulos et al. (1997) 62. Williams TEOSQ IMI & Gill (1995) 63. Xiang et TEOSQ al. (1997)

Recreation

PostConference Exercise

80

0.16

0.02

0.06

0.77

0.75

Competitive

PostConference Exercise

11

0.54 -0.35 0.39

0.09

0.79

0.78

Competitive

Typical

Published

333

0.28

0.02

0.80

Wide variety

Typical

Published

148

0.05

Competitive M England University, Competitive state anxiety M=22 Physical Pressure/ M+F England School, Education Tension M=13

Typical

Published

60

PostPublished Exercise

304

Shame, guilt, M Taiwan University, pity, anger, M=22 depression, surprise, hopelessness DisappointF Germany School, ment, M=14 dissatisfaction, anger, shame, resignation Boredom M+F USA University M=21 M+F Norway University,

Satisfied, Pleased, Depressed, Happy, Proud, Ashamed, Confident Disappointed Embarrassed Positive Physical Engagement, Exhaustion Revitalisation Tranquillity Good, Satisfied, Depressed, Pleased, Happy, Ashamed, Proud, Confident Disappointed Embarrassed Enjoyment/ Interest Satisfaction Boredom

0.79

0.85

0.86

0.84

0.71

0.81

0.79

0.81

0.02 -0.04 -0.05 0.82

0.86

0.84

0.70

1070 0.46 -0.15 0.15 -0.02 0.83

0.84

0.87

0.73

0.77

-0.2

0.08 0.14

-0.1 0.28

0.04

M+F England School, M=14

Physical Education

Typical

M+F England School, M= 13

Physical Education

PostPublished Exercise

304

0.29 -0.13 0.02

.00

0.82

0.86

M+F England School, M=12

Physical Education

PostPublished Exercise

211

0.28 -0.01 0.17

0.02

0.82

0.80

0.9

Typical

Published

174

0.52

0.84

0.86

0.73

Typical

Published

121

0.42 -0.26 0.12 -0.01 0.86

0.91

0.70

M+F

USA

M+F

USA

Recreation School, M=13 School, Physical grades=4- Education 6

Published

.00

-0.13

0.50

40 63. Xiang et al. (1997)

TEOSQ

Satisfaction

64. Yin & TEOSQ SCAT Boyd (1994)

65a. Yin (1993)

TEOSQ

Boredom

Competitive trait anxiety, negative affect Enjoyment

M+F

China

M

USA

M+F

USA

School Physical grades=4- Education 6 Recreation School, M=15

Typical

Published

180

Typical

Published

270

Recreation

Typical

Conference

94

School, M=12

0.27

0.48

0.01

0.22 -0.12 0.73

0.75

-0.07

0.11

0.80

0.76

0.75

0.75

-0.11

0.70

0.50

0.83

0.62

41 Table 3: Results of the meta-analysis for the relationships between achievement goals and affect Correlation

N

K

r

ρ

Task-PA

6515

33

.36

.55

Task-NA

6556

33

-.11

Ego-PA

6515

33

Ego-NA

6556

33

Observed Variance .017

Corrected Variance .013

Credibility interval .88 to .23

Confidence interval .40 to .32

75%

Q

33

98.62**

Fail-Safe K for p=.10 149

-.18

.019

.014

.21 to -.58

-.07 to -.15

27

120.34**

28

.07

.10

.014

.009

.39 to -.19

.10 to .03

35

93.48**

.02

.04

.007

.002

.17 to -.10

.05 to .00

75

44.20

**p<.01 Note: PA= Positive affect, NA= Negative affect, N= number of participants, K= number of independent samples, r= correlation corrected for sampling error, ρ= correlation corrected for sampling and measurement errors, 75%= Hunter and Schmidt’s (1) rule for homogeneity, Q= chisquare test for homogeneity, Fail-Safe K= number of independent samples averaging null correlations required to bring down the ρ to the value of .10.

42 Table 4: Moderator analysis for the relationships between achievement goals and affect Correlation

N

K

r

ρ

Task-PA

Published

5348

24

.37

.57

Observed Variance .014

Task-PA

Unpublished

1076

8

.33

.51

.021

.015

.87 to .16

.42 to .25

35

23.01**

Task-NA

Published

5364

26 -.11

-.19

.015

.011

.15 to -.53

-.07 to -.15

33

78.36**

Task-NA

Unpublished

1192

7 -.09

-.15

.036

.030

.41 to -.71

.04 to -.22

17

42.42**

Ego-PA

Published

5439

25

.07

.11

.016

.011

.43 to -.20

.11 to .03

30

84.32**

Ego-PA

Unpublished

1076

8

.03

.04

.007

.000

-

.09 to -.03

113

Task-NA

2939

19 -.01

-.01

.011

.04

.20 to -.22

.02 to -.04

60

31.45*

2540

15 -.21

-.35

.016

.011

-.03 to -.67

-.17 to -.25

42

35.99**

Task-PA

High arousal negative affect Low arousal negative affect USA

2249

14

.32

.50

.017

.011

.80 to .20

.38 to .27

40

35.28**

Task-PA

Great Britain

3051

11

.41

.63

.010

.007

.84 to .42

.46 to .36

49

22.42*

Task-NA

USA

3015

17 -.09

-.15

.022

.017

.28 to -.57

-.02 to -.15

26

65.10**

Task-NA

Great Britain

3111

12 -.13

-.21

.014

.010

.11 to -.53

-.07 to -.19

29

41.08**

Ego-PA

USA

2249

14

.03

.04

.013

.007

.29 to -.21

.07 to -.02

47

29.75**

Ego-PA

Great Britain

3111

12

.08

.11

.010

.006

.35 to -.12

.12 to .03

39

30.84*

Task-NA

Moderator

Corrected Variance .011

Credibility Interval .85 to .28

Confidence interval .41 to .32

75%

Q

38

63.87**

7.05

z

.70

.31

2.22

7.03

2.59

.99

1.53

43 Task-PA

Physical Ed.

3532

14

.41

.63

.010

.007

.84 to .42

.45 to .36

52

26.95*

Task-PA

Recreation

1172

9

.31

.47

.018

.011

.78 to .17

.38 to .24

43

20.99**

2.57

Task-PA

Competitive

1456

8

.31

.47

.018

.014

.80 to .14

.39 to .22

32

24.99**

.02 with Rec. 2.30 with PE

Task-NA

Physical Ed.

3337

12 -.10

-.17

.020

.016

.24 to -.59

-.03 to -.18

17

61.68**

Task-NA

Recreation

1751

11 -.04

-.07

.009

.003

.09 to -.23

-.01 to -.07

71

15.41

1.53

Task-NA

Competitive

1350

8 -.15

-.25

.021

.015

.13 to -.64

-.07 to -.24

30

27.08**

2.47 with Rec. .91 with PE

Ego-PA

Physical Ed.

3532

14

.07

.11

.018

.014

.46 to -.25

.13 to .01

22

62.41**

Ego-PA

Recreation

1172

9

.01

.02

.019

.008

.34 to -.30

.08 to -.06

40

22.61**

Ego-PA

Competitive

1589

10

.08

.13

.002

.000

-

.13 to .03

255

Task-PA

Typical affect

5279

25

.37

.58

.019

.015

.93 to .22

.42 to .32

29

Task-PA

Post-exercise affect Typical affect

1236

8

.29

.44

.004

.000

-

.34 to .24

5177

24 -.13

-.22

.020

.016

.19 to -.63

-.08 to -.18

25

Post-exercise affect Typical affect

1349

9 -.03

-.04

.007

.001

.03 to -.12

.03 to -.08

93

.08

.13

.012

.007

.39 to -.13

.12 to .05

40

63.09**

Post-exercise affect

1236

8 -.02

-.03

.016

.009

.26 to -.32

.05 to -.09

41

19.51**

Task-NA Task-NA Ego-PA Ego-PA

5279

25

179

3.92

1.25 2 with Rec. .40 with PE

85.34** 4.48

3.60

97.49** 9.67

3.96

2.71

44 Task-PA

School

5115

24

.38

.59

.014

.010

.87 to .32

.42 to .34

40

59.32**

Task-PA

University

1400

9

.26

.41

.019

.013

.74 to .07

.34 to .19

35

25.85**

Task-NA

School

5086

22 -.11

-.18

.023

.019

.27 to -.63

-.05 to -.17

20

111.60**

Task-NA

University

1470

11 -.11

-.19

.005

.000

-

-.06 to -.17

Ego-PA

School

5115

24

.07

.11

.016

.011

.42 to -.20

.11 to .03

31

78.51**

Ego-PA

University

1400

9

.05

.07

.010

.003

.26 to -.11

.09 to .01

62

14.40

1802

9

.05

.08

.007

.005

.23 to -.08

.08 to .01

68

13.26

10 -.09

-.15

.004

.000

-

-.03 to -.15

Task-NA

School x High arousal negative affect Task-NA University x high arousal negative affect **p<.01 *p<.05

1137

137

207

8.03

4.84

2.83

.26

.71

8.51

Notes: 1) In some of the moderator subsets the variance explained by the artifacts is greater than the observed variance (i.e. the 75% test has values over 100%). Although this seems odd, it is due to the fact that the variance of observed correlations is a sample estimate. Unless the number of studies is infinite, there will be some error in that estimate (see 1, 81). 2) PA= Positive affect, NA= Negative affect, N= number of participants, K= number of independent samples, r= correlation corrected for sampling error, ρ= correlation corrected for sampling and measurement errors, 75%= Hunter and Schmidt’s (1) rule for homogeneity, Q= chisquare test for homogeneity, z= significance test for moderator differences.

45

Figure Caption Figure 1: Structural equation modelling analysis of the relationships between task orientation, ego orientation, positive affect, and negative affect based on the results of the meta-analysis.

46

.85 Task orientation

.52** (.52)

E1

Positive affect

-.34**

0 (.08)

-.26** (-.25) Ego orientation

.08** (.04)

Negative affect

.96

E2

**p<.01 Note: In parentheses are reported the correlations based on the subset of 15 independent samples who provided a full correlation matrix (N=2980).

47

Note: 1. Although many authors feel that it is more appropriate to analyse covariance over correlation matrices, the analysis of the present correlation matrix met the necessary requirements described by Cudeck (82). Specifically, the reproduced correlation matrix met the assumptions for a scale-invariant model. 2. Dr Ntoumanis is now at Warrington University College, UK. 3. see Ntoumanis & Biddle (83) for a review of motivational climate and a brief metaanalysis of the effects of climate in physical activity. Acknowledgements 1. Thanks are expressed to those researchers who kindly supplied additional data from their research thus making this meta-analysis possible. 2. The comments of Professor Glyn C. Roberts on an earlier version of this paper are gratefully acknowledged.

48