Comparing a domain-specific and global measure of perfectionism in competitive female figure skaters.

Dunn, John G.H. ; Craft, Janelle M. ; Dunn, Janice Causgrove 等

Theorists and researchers generally conceptualize the personality trait of perfectionism as a multidimensional construct (Frost, Marten, Lahart, & Rosenblate, 1990; Hewitt & Flea, 1991), with the core facet representing an individual's tendency to set extremely high standards for personal performance (Stoeber & Otto, 2006). Although a variety of multidimensional self-report instruments have been developed to assess perfectionism (for reviews see Enns & Cox, 2002; Flett & Hewitt, 2002; Stoeber & Otto, 2006), test-users face a critical decision when choosing a perfectionism measure because the choice of instrument will have implications for the validity, interpretation, and generalizability of results (Flett & Hewitt, 2002). The choice of instrument seems especially important in studies examining perfectionism in competitive sport settings because many researchers have argued that perfectionism in sport should be conceptualized and measured as a domain-specific construct (e.g., Anshel & Eom, 2003; Dunn, Causgrove Dunn, & Syrotuik, 2002; Stoeber, Otto, Pescheck, Becker, & Stoll, 2007). Therefore, the primary purpose of the current study was to directly compare potential benefits of using a domain-specific versus global (i.e., domain-free or dispositional) measure of perfectionism in a competitive sport setting.

Measuring perfectionism as a domain-specific construct is not a novel idea. Indeed, long before the creation of global measures of perfectionism (i.e., instruments which provide no situational/contextual frame of reference for respondents when they evaluate their perfectionist tendencies) theorists argued that individuals' perfectionist tendencies are likely to vary across different life settings (see Hollender, 1965; Missildine, 1963). Empirical support for the view that perfectionist tendencies vary across situational contexts has been provided in both sport (e.g., Dunn, Gotwals, & Causgrove Dunn, 2005) and non-sport settings (e.g., Mitchelson & Burns, 1998). Dunn et al. (2005) found that male (n = 133) and female (n = 108) intercollegiate varsity athletes reported significantly higher mean levels of perfectionism across all three dimensions of Hewitt and Flett's (1991) Multidimensional Perfectionism Scale (Hewitt-MPS) when asked to respond in the context of sport versus school/academe or life in general. Similarly, Mitchelson and Bums found that 67 working mothers reported significantly greater mean levels of perfectionism (when measured with the Hewitt-MPS) at work than at home.

While the results of the aforementioned studies support the view that the situational context should be considered when measuring perfectionism, some sport psychologists have taken this idea further and have shown that there may be dimensions of perfectionism unique to the competitive sport environment. For example, Dunn et al. (2002) and Anshel and Eom (2003) independently identified a dimension of perfectionism that reflects athletes' perceptions of the role that coaches play in exerting pressures on athletes to reach certain performance standards. Dunn et al. (2002) posited that this Perceived Coach Pressure dimension of perfectionism is a domain-specific aspect of Socially Prescribed Perfectionism (SPP)--where SPP is defined as the degree to which an individual feels that personal performance standards and performance expectations are imposed and judged by significant others (Hewitt & Flett, 1991). If, as previous research suggests, perceived coach pressure is an important dimension of perfectionism to consider in sport-perfectionism research (see Dunn, Causgrove Dunn, Gotwals, Vallance, Craft, & Syrotuik, 2006; Dunn, Gotwals, Causgrove Dunn, & Syrotuik, 2006; Sagar & Stoeber, 2009; Vallance, Dunn, & Causgrove Dunn, 2006), the implications of choosing an instrument to measure perfectionism in sport are further reinforced because global perfectionism measures do not specifically identify the coach as a source of socially prescribed perfectionism.

According to Stoeber, Uphill, and Hotham (2009), the most widely used domain-specific measure of perfectionism in sport is Dunn et al.'s (2002) Sport Multidimensional Perfectionism Scale (Sport-MPS). The Sport-MPS was created around the conceptual framework provided by Frost et al.'s (1990) Multidimensional Perfectionism Scale (Frost-MPS) and measures four dimensions of perfectionism in sport: Personal Standards, Concern Over Mistakes, Perceived Coach Pressure, and Perceived Parental Pressure. The Sport-MPS has undergone extensive psychometric evaluation, has consistently demonstrated good factorial and criterion-related validity evidence, and has repeatedly demonstrated acceptable levels of internal consistency (see Dunn, Causgrove Dunn et al., 2006; Dunn et al., 2002). However, the extent to which the Sport-MPS provides a greater understanding of perfectionism in sport beyond that provided by existing global (or domain-free) measures of perfectionism is still undetermined. Consequently, the first purpose of the current study was to compare the predictive power of the Sport-MPS against an established measure of global perfectionism in a competitive sport setting.

When comparing the measurement benefits of different instruments (that measure similar constructs), researchers can also benefit from examining the degree to which the different instruments produce results that contribute to an enhanced understanding of the nomological network (Cronbach & Meehl, 1955) that encompasses the construct being measured (Messick, 1989). Stated differently, decisions to use domain-specific measures over global measures of personality traits should also be based on the extent to which the domain-specific instrument sheds light on theoretical issues surrounding the construct-in-question beyond that which can be achieved by a global instrument (see Gauvin & Russell, 1993). To this end, the second purpose of this study was to determine if a domain-specific measure of perfectionism was capable of adding to the theoretical knowledge-base surrounding perfectionism in sport beyond that which could be achieved by a global measure of perfectionism. The particular theoretical issue of interest in this study was the healthy versus unhealthy nature of perfectionism in sport.

Healthy vs. Unhealthy Perfectionism

Some theorists argue that perfectionist tendencies are primarily unhealthy in nature (e.g., Flett & Hewitt, 2002; Hall, 2006) whereas others have proposed that perfectionist tendencies can be either healthy or unhealthy (e.g., Hamachek, 1978; Rice & Ashby, 2007). On the basis of a review of 35 empirical studies, Stoeber and Otto (2006) recently presented evidence supporting the distinction between healthy and unhealthy perfectionism and concluded that healthy perfectionism was best conceptualized as the combination of high perfectionist strivings (i.e., high personal standards and high self-oriented perfectionism) with low perfectionist concerns (i.e., low concern over mistakes and low socially prescribed perfectionism). In contrast, Stoeber and Otto proposed that unhealthy perfectionism was best conceptualized as the combination of high perfectionist strivings with high perfectionist concerns. Unfortunately, none of the studies in Stoeber and Otto's (2006) review were conducted with athletes in competitive sport settings, so the debate over whether there is a distinction between healthy and unhealthy perfectionism in sport is still largely unresolved.

Although the majority of perfectionism research that has been conducted in sport tends to support the view that perfectionism is a dysfunctional personality trait (see Hail, 2006, for a review), it should be noted that the majority of sport-based perfectionism research has used global measures to assess the construct in athletes. However, a small but emerging body of research has employed domain (i.e., sport) specific measures to assess athletes' perfectionist orientations and has provided evidence of both unhealthy and healthy perfectionism in sport (e.g., Dunn et al., 2002; Stoeber & Becker, 2008; Stoeber et al., 2007).

The ongoing debate over the functional nature of perfectionism in sport (see Hall, 2006) may potentially be addressed by examining the relationship between athletes' perfectionist orientations and attitudinal body image. Attitudinal body image can be defined as the manner in which individuals feel satisfied or dissatisfied with some aspect (or aspects) of their physical appearance (Cash & Szymanski, 1995). Athletes who become highly dissatisfied with their bodies (especially as they pertain to being overweight) may be at risk of developing eating disorders or experiencing psychological distress (see Davis, 2002).

According to theory, unhealthy perfectionists are more self critical, less self-accepting of personal flaws, and have higher self-presentational concerns than healthy perfectionists (Hamachek, 1978; Stoeber & Otto, 2006). Consequently, unhealthy perfectionists would be expected to be more concerned or dissatisfied with their physical appearance than healthy perfectionists. Research that has investigated this theoretical proposition (e.g., Davis, 1997; Pearson & Gleaves, 2006) has relied on global measures of perfectionism. In line with the majority of perfectionism research in sport psychology, these studies have produced considerable evidence supporting the link between unhealthy perfectionist orientations and negative body image (see Davis, 2002) but little evidence supporting a link between healthy perfectionist orientations and positive body image. It is currently unknown if similar findings would have emerged had domain-specific measures of perfectionism been employed, and whether the domain-specific conceptualization and measurement of perfectionism would have provided evidence supporting the theorized link between healthy/adaptive perfectionism and positive/healthy attitudinal body image.

To summarize, given the absence of studies that directly compare the potential benefits of measuring perfectionism with domain-specific versus global measures of perfectionism in sport (see Flett & Hewitt, 2005) the present research had two major objectives. First, the study sought to compare the ability of a domain-specific measure of perfectionism (i.e., the Sport-MPS: Dunn et al., 2002) and a global measure of perfectionism (i.e., the Hewitt-MPS: Hewitt & Flett, 1991) to predict attitudinal body image in a sample of athletes. Second, the study sought to compare the ability of the two instruments to shed light on the ongoing debate over whether perfectionist tendencies are primarily unhealthy in nature or whether perfectionist tendencies may also play a healthy/adaptive role in sport. The Sport-MPS and Hewitt-MPS were selected as the measures of choice because both have been used to measure perfectionism in athletes and both have undergone extensive psychometric evaluation (see Dunn, Causgrove Dunn, et al. 2006; Enns & Cox, 2002).

Method

Participants

A total of I 19 competitive female figure skaters from clubs across Canada (M age = 14.56 years, SD = 3.42) voluntarily participated in the study. (1) The sample contained the full range of competition levels that are sanctioned by Skate Canada (the sport's national governing body in Canada) and included 40 juvenile (M age = 11.75 years, SD = 1.31), 20 pre novice (M age = 13.25 years, SD = 1.21), 32 novice (M age = 15.13 years, SD = 1.21), 12 junior (M age = 16.58 years, SD = 1.62), and 15 senior athletes (M age = 21.00 years, SD = 3.61). Each competitive level provides an opportunity for athletes to compete in national championship competitions. On average, the combined sample had been involved in figure skating for 6.21 years (SD = 3.30) and engaged in 2.69 hours of on-ice training (SD = .95) and 1.59 hours of off-ice training (SD = 1.11) each day. The sport of female figure skating was chosen because perfectionist tendencies and body image concerns are believed to be prevalent in female athletes who compete in aesthetic sports (such as figure skating) where pressures are often exerted on athletes to maintain an "ideal" lean body type (Taylor & Ste-Marie, 2001).

Instruments

Athletes completed five self-report instruments: (1) a demographic questionnaire, (2) a skating-modified version of the Sport-MPS, (3) the Hewitt-MPS, (4) a skating-modified version of the Multidimensional Body Self Relations Questionnaire-Appearance Scale (MBSRQ-AS: Cash, 2000a), and (5) a skating-modified version of the Body-Image Ideals Questionnaire (BIQ: Cash & Szymanski, 1995).

Sport-Multidimensional Perfectionism Scale (Sport-MPS). The Sport-MPS (Dunn, Causgrove Dunn, et al., 2006; Dunn et al., 2002) is a 30-item instrument that assesses four dimensions of perfectionism in sport: Personal Standards (PS), Concern Over Mistakes (COM) Perceived Parental Pressure (PPP) and Perceived Coach Pressure (PCP). Respondents rate the extent to which they agree with each item using a 5-point scale (1 = strongly disagree; 5 = strongly agree) with higher composite subscale scores reflecting higher levels of perfectionism on each respective dimension. The Sport-MPS has consistently demonstrated acceptable levels of internal consistency (as > .70) across all subscales in a variety of studies (see Dunn, Causgrove Dunn, et al, 2006; Dunn et al., 2002; Vallance et al., 2006), and factor analyses of Sport-MPS data have consistently identified the same four factors across independent samples of athletes (see Dunn, Causgrove Dunn, et al., 2006; Dunn et al., 2002).

The Sport-MPS was originally developed for use with Canadian Football players (Dunn et al., 2002). As such, the original items were written for team-sport contexts. Consequently, a number of items were slightly re-worded in this study to make them more relevant to the sport of figure skating. Specifically, the words "play", "player" and "players" in the original inventory were respectively replaced by the words "skate", "figure skater" and "figure skaters." Multidimensional Perfectionism Scale (Hewitt-MPS). The Hewitt-MPS (Hewitt & Flett, 1991) is one of the most widely used and well-established measures of perfectionism in the extant literature (Enns & Cox, 2002). It conceptualizes and measures perfectionism as a global personality trait, and contains 45 items that measure three dimensions of perfectionism: Self-Oriented Perfectionism (SOP), Socially Prescribed Perfectionism (SPP), and Other-Oriented Perfectionism (OOP). Respondents rate the degree to which they agree that items reflect their "personal characteristics and traits" on a 7-point scale (1 = strongly disagree; 7 = strongly agree). Higher composite subscale scores reflect higher levels of perfectionism. A large number of clinical-, social-, and personality-psychology studies have provided reliability and validity evidence supporting use of the Hewitt-MPS as a measure of multidimensional global perfectionism (see Enns & Cox, for a review). Internal consistency levels for the three subscales have been shown to be acceptable ([alpha]s > .70: see Hewitt & Flett) and criterion-related validity evidence has been established for the subscales by way of statistically significant theoretically meaningful correlations with other measures of global perfectionism (see Enns & Cox).

Multidimensional Body Self Relations Questionnaire-Appearance Scale (MBSRQ-AS). The MBSRQ-AS contains 34 items and is a shortened version of the original 69-item MBSRQ (see Cash, 2000a). The instrument measures both cognitive/motivational and affective aspects of attitudinal body-image and contains five subscales that are labeled Appearance Evaluation (AE), Appearance Orientation (AO), Overweight Preoccupation (OP), Self-Classified Weight (SCW), and the Body Areas Satisfaction Scale (BASS). Five-point scales are used by respondents to answer items in each subscale: AE, AO, and OP (1 = definitely disagree; 5 = definitely agree), SCW (1 = very underweight; 5 = very overweight), and BASS (1 = very dissatisfied; 5 = very satisfied). Unhealthy/negative attitudinal body image is reflected in high AO, OP and SCW scores and low AE and BASS scores. All subscales have demonstrated acceptable levels of internal consistency (i.e., [alpha]s > .70) and test-retest reliability over a one month period (rs > .74) with female participants (Cash, 2000a).

A number of minor changes were made to the wording of several MBSRQ-AS items in an attempt to increase the relevance of the instrument for the current sample of figure skaters. For example, the phrase "Before going out in public ..." was changed to "Before going out to the skating arena ..." and "... the way my clothes fit me" was changed to "... the way my skating outfits fit me." Given the potential for a few athletes within the current sample to be at a prepubescent stage of development, we deleted two appearance evaluation items that asked respondents to consider if they viewed their bodies as "sexually appealing" and whether they liked the way their bodies looked without clothing.

Body-Image Ideals Questionnaire (BIQ). The BIQ (Cash & Szymanski, 1995) is an 11-item measure of evaluative body image that asks respondents to rate themselves according to a variety of physical characteristics. For each item respondents rate (a) the perceived discrepancy (0 = exactly as I am; 3 = very unlike me) between their ideal self and actual self on the physical characteristic in question (e.g., skin complexion, weight, etc.) and (b) the degree to which it is important (0 = not important; 3 = very important) that they reach their ideal self on the physical characteristic in question. A weighted discrepancy score is calculated for each item whereby the cross product of the discrepancy and importance ratings is computed (see Cash & Szymanski, for detailed explanation of scoring). Higher positive "weighted discrepancy scores" reflect a heightened discrepancy between the actual self and ideal self, and for this discrepancy to assume importance in the individual's life (i.e., higher scores reflect a more unhealthy body image). The BIQ has demonstrated acceptable levels of internal consistency (i.e., [alpha] > .70), and convergent validity evidence has been demonstrated through statistically significant theoretically meaningful correlations with several measures of body image (see Cash & Szymanski) in female samples. An overview of the construct validity evidence surrounding the BIQ is reported by Cash (2000b).

A number of minor modifications were made to the BIQ to make the instrument more relevant to the sport of figure skating. First, the phrase "as a figure skater" was added to each item (e.g., "My ideal height is ..." was changed to read, "My ideal height as a figure skater is..."). Second, given the potential for a few athletes to be at a pre-pubescent stage of development, one item that asked respondents to rate their chest size was removed, leaving a total of 10 items in the instrument.

Procedures

Standard ethical and informed consent procedures were adopted throughout the study. Written informed consent was obtained from all athletes, and written parental consent was obtained for all athletes who were under the age of 18 years at the time of the study. Instruments were individually administered to athletes by one member of the research team at each skater's respective club. The demographic questionnaire was always administered first, with the remaining instruments being presented in eight different presentation orders (to minimize potential order effects). Athletes took approximately 45 minutes to complete the instruments.

Results

Preliminary data analyses. Results of three separate MANOVAs (one for Sport-MPS subscales [n = 4], one for Hewitt-MPS subscales [n = 3], and one for body image variables [n = 6] entered as dependent variables) with "competitive level" entered as the independent variable found only three (out of a possible 130) between-group differences across the perfectionism and body-image variables at the univariate level of analysis. All homogeneity tests of covariance (Box's M) that were conducted in conjunction with each MANOVA were not significant, indicating that the covariance matrices associated with the perfectionism and body image variables across each of the five competitive levels did not significantly differ. Given that there were very few between-group differences across the five competitive levels on the perfectionism and body image variables (and no differences in the structure of the covariance matrices), the data across the competition levels were combined into a single data set (N = 119) for all subsequent analyses.

Table 1 contains descriptive statistics and internal consistency values for all 13 subscales for the combined sample. With the exception of the other-oriented perfectionism (OOP) subscale, all subscales had acceptable levels of internal consistency ([alpha] > .70). On the basis of the marginal level of internal consistency associated with other-oriented perfectionism ([alpha] = .67) readers are advised to interpret results pertaining to this subscale with some degree of caution. Table 2 contains the bivariate correlations among all subscales for the combined sample. All correlations were < |.70 | indicating that there would be no multicollinearity problems among the variables that could potentially affect the results of the upcoming regression analyses (see Tabachnick & Fidell, 1996, p. 86).

Assessing the Predictive Power of the Sport-MPS vs. the Hewitt-MPS

To determine if the Sport-MPS and the Hewitt-MPS explained different amounts of unique variance in attitudinal body image, a series of hierarchical multiple regression analyses were conducted. For the prediction of each body-image variable, the three Hewitt-MPS subscales were entered simultaneously in the first step and the four Sport-MPS subscales were entered simultaneously in the second step. The change in [R.sup.2] values and associated F-statistics for each step are contained in Table 3. For four of the six body-image variables (AO, AE, BASS, and BIQ) the Sport-MPS explained a significant amount of variance over and above the variance explained by the Hewitt-MPS (all ps < .05) suggesting that the Sport-MPS was valuable in explaining variance in attitudinal body image that had not been explained by the Hewitt-MPS.

Upon completion of these initial regression analyses, the entry order of the perfectionism measures was then reversed such that the Sport-MPS subscales were entered in the first step and the Hewitt-MPS subscales were entered in the second step. This reversal of entry order was important because we wanted to determine if the Hewitt-MPS would be able to explain variance in attitudinal body image that was not explained by the Sport-MPS. In each of the six analyses, the Sport-MPS explained a significant amount of body-image variance when entered in the first step (all [R.sup.2] >. 13; all ps < .005). However, when the Hewitt-MPS subscales were entered in the second step in each analysis, no statistically significant changes in [R.sup.2] were obtained (all [DELTA][R.sup.2] < .036; all ps > .20). These findings indicate that the Hewitt-MPS was unable to explain significant amounts of body-image variance beyond that which had been accounted for by the Sport-MPS.

Assessing the Functional Nature of Perfectionism with the Sport-MPS and Hewitt-MPS

Given that perfectionism researchers and theorists have argued that the healthy vs. unhealthy nature of perfectionism is best determined by considering scores across all perfectionism subscales simultaneously (e.g., Dunn et al., 2002; Frost et al., 1990; Gotwals, Dunn, & Wayment, 2003; Rice & Ashby, 2007; Stoeber & Otto, 2006), separate canonical correlation analyses were employed to examine the multivariate relationships between (a) perfectionism in sport (as measured by the Sport-MPS) and attitudinal body-image, and (b) global perfectionism (as measured by the Hewitt-MPS) and attitudinal body-image.

In accordance with the recommendations of Tabachnick and Fidell (1996), prior to conducting the canonical correlation analyses, the data were screened for the presence of multi variate outliers. Separate screening procedures (using Mahalonobis distances and [chi square] tests of significance) were conducted on the Sport-MPS and body-image data (i.e., the data to be included in the first canonical correlation analysis), and on the Hewitt-MPS and body-image data (i.e., the data to be included in the second canonical correlation analysis). One multivariate outlier was identified in each analysis, and upon further inspection of the data it was revealed that the outlier was the same person, Consequently, this individual was removed from the two canonical correlation analyses. Subject-to-variable ratios exceeded the minimum criterion (i.e., 10:1) required for each canonical correlation analysis (see Tabachnick & Fidell).

The five MBSRQ-AS subscales and the BIQ weighted discrepancy score were entered as the criterion set in both canonical correlation analyses. The four subscales of the Sport-MPS were entered as the predictor set in the first analysis and the three Hewitt-MPS subscales were entered as the predictor set in the second analysis. Tabachnick and Fidell (1996) recommend that canonical functions should be retained only if they are statistically significant and practically meaningful. To this end, canonical functions were retained only if the canonical correlation coefficient ([R.sub.C) was statistically significant (p < .05) and exceeded a minimum criterion value of .30. Using these criteria, two significant canonical functions were extracted when the Sport-MPS subscales were entered as the predictor set ([R.sub.C1] = .63; [R.sub.C2] = .43), and one significant function was extracted when the Hewitt-MPS subscales were entered as the predictor set ([R.sub.C] = .58). Within each significant canonical function, only variables with canonical loadings [greater than or equal to] [absolute value of .30] on their respective canonical variates were considered for interpretation purposes.

Perfectionism and unhealthy attitudinal body image. As seen in Table 4, the canonical loadings associated with all of the body-image and perfectionism variables were [greater than or equal to] |.30| in the first canonical function produced by the initial canonical correlation analysis (i.e., when the Sport-MPS subscales were entered as the predictor set). Appearance orientation, overweight preoccupation, self-classified weight and the BIQ weighted discrepancy variables had moderate to strong positive loadings, while the appearance evaluation and body areas satisfaction subscales had moderate negative loadings on the body image variate. Collectively this pattern of loadings reflects an unhealthy/negative attitudinal body image. All four Sport-MPS subscales had moderate to strong positive loadings on the perfectionism variate in the first canonical function. This pattern of loadings is reflective of an unhealthy perfectionist orientation (cf. Dunn et al., 2002; Stoeber & Otto, 2006). The magnitude and direction of the canonical correlation between these two canonical variates ([R.sub.C1] = .63) indicates that the predisposition to adopt an unhealthy/negative attitudinal body image in figure skating is associated with a heightened tendency to possess an unhealthy perfectionist orientation in sport.

Similar results were obtained in the canonical correlation analysis conducted with the Hewitt-MPS subscales entered as the predictor set. In this analysis, all global perfectionism and body-image variables had interpretable canonical loadings (i.e., [greater than or equal to] |.30|) on their respective variates (see Table 4). The direction and magnitude of all six body-image loadings were very similar to those corresponding with the unhealthy body-image variate in the first canonical function of the initial canonical analysis (when Sport-MPS subscales had been entered as the predictor set). In addition, all three Hewitt-MPS subscales had positive loadings on the perfectionism variate; this pattern of perfectionism loadings reflects unhealthy/maladaptive perfectionism (see Rice & Ashby, 2007; Sherry, Hewitt, Besser, McGee, & Flett, 2004). The magnitude and direction of the canonical correlation ([R.sub.C]= .58) between these two variates indicates that the predisposition to adopt an unhealthy/negative attitudinal body image in figure skating is associated with a heightened tendency to possess an unhealthy/maladaptive profile of global perfectionism.

Perfectionism and healthy attitudinal body image. As reported earlier, the initial canonical correlation analysis with the Sport-MPS subscales entered as the predictor set produced a second significant canonical function ([R.sub.C2]= .43). The patterns of canonical loadings in this second canonical function were quite different from those in the first canonical function (see Table 4). Three of the six body-image subscales and three of the four sport perfectionism subscales had interpretable loadings. Appearance evaluation (AE) and the body areas satisfaction scale (BASS) had strong positive loadings and the BIQ weighted discrepancy had a moderate negative loading on the body image variate. Collectively, this pattern of loadings reflects a healthy or positive attitudinal body image. The personal standards subscale of the Sport-MPS had a moderate positive loading on the perfectionism variate of the second canonical function (see Table 4). In contrast, the concern over mistakes and perceived coach pressure subscales had low to moderate negative loadings on this perfectionism variate. This pattern of loadings is consistent with theorized descriptions of normal/healthy perfectionism (see Hamachek, 1978; Stoeber & Otto, 2006). The magnitude and direction of the canonical correlation between these two canonical variates ([R.sub.C2]= .43) indicates that the predisposition to adopt a healthy or positive attitudinal body image in figure skating is associated with a heightened tendency to possess a healthy or adaptive perfectionist orientation in sport.

Discussion

The first purpose of this study was to compare the predictive power of the Sport-MPS against the Hewitt-MPS in explaining attitudinal body-image variance in the current sample of female figure skaters. Regression analyses showed that for four of the six body-image variables the Sport-MPS subscales explained a significant amount of unique variance over and above that which was explained by the Hewitt-MPS subscales. In contrast, when the Hewitt-MPS subscales were entered in the second step of each regression analysis, the Hewitt-MPS was unable to explain a significant amount of unique variance in any of the six body-image variables beyond that which had already been explained by the Sport-MPS. In other words, the domain-specific measurement of perfectionism had greater predictive power than the global measurement of perfectionism with respect to accounting for variance in figure skaters' attitudinal body-image.

Similar results supporting the superior predictive power of domain-specific measures over global measures of psychological dispositions have been documented in the extant literature (e.g., Bing, Whanger, Kristl Davison, & Van Hook, 2004; Hunthausen, Truxillo, Bauer, & Hammer, 2003; Laux, Glanzmann, & Schaffer, 1985; Martens & Simon, 1976). However, to the best of our knowledge this is the first direct comparison of the predictive power of a domain-specific vs. global measure of perfectionism in sport. Although it would be difficult to generalize the current findings beyond the current sample of competitive female figure skaters, these results provide the first evidence that domain-specific assessments may be more useful than global assessments when measuring perfectionism in sport settings. Clearly more research is needed to determine if these results are sample-specific, and if they would generalize to other groups of athletes or to the prediction of other sport-related cognitive, affective, and behavioral variables such as competitive anxiety, competitive anger, or competitive performance. Also, it is important to recognize that other multidimensional sport-specific measures (e.g., the Multidimensional Perfectionism Inventory in Sport: see Stoeber et al., 2007) and global measures (e.g., the Frost-MPS: Frost et al., 1990) of perfectionism are available to researchers, therefore, the current results cannot be generalized to these inventories. Nevertheless, the superior predictive power of the Sport-MPS over the Hewitt-MPS in this study informs researchers about a potential benefit that the domain-specific measurement of perfectionism might have over the global measurement of perfectionism in a sport setting.

Gauvin and Russell (1993) argued that when decisions are being made about which instrument to choose for research purposes, instruments that can clarify or expand theoretical knowledge around a particular construct or phenomenon are generally preferred over measures that cannot achieve this research objective. To this end, the second purpose of the current study was to compare the Sport-MPS and the Hewitt-MPS in their abilities to shed light on the ongoing debate over whether perfectionism in sport has both healthy/adaptive and unhealthy/maladaptive qualities (see Dunn et al., 2002), or if perfectionism is primarily an unhealthy/debilitative construct in sport (see Flett & Hewitt, 2005). To achieve this goal, canonical correlation analyses were conducted that permitted a comparison between the Sport MPS and Hewitt-MPS regarding the information they provided about the functional nature of perfectionism in the current sample of figure skaters.

Regardless of whether the Sport-MPS or Hewitt-MPS was used to measure perfectionism, both canonical correlation analyses produced solutions which depicted profiles (i.e., canonical variates) resembling unhealthy/maladaptive perfectionism that were significantly related to profiles of negative/unhealthy attitudinal body image. The profiles of unhealthy/ maladaptive perfectionism from the Sport-MPS and Hewitt-MPS directly reflect Stoeber and Otto's (2006) description of unhealthy perfectionism that is characterized by the combination of high perfectionist strivings (i.e., high personal standards or high self-oriented perfectionism) with high perfectionist concerns (i.e., high concern over mistakes, high perceived coach pressure, and/or high socially prescribed perfectionism). These results are not only consistent with the body-image literature that has documented the association between unhealthy/maladaptive perfectionism and unhealthy/negative body image (see Davis, 2002), but are also consistent with the perfectionism literature in both clinical psychology (e.g., Flett & Hewitt, 2002; Rice, Bail Castro, Cohen, & Hood, 2003) and sport psychology (e.g., Flett & Hewitt, 2005; Hall, 2006) where perfectionism has been largely characterized as a debilitating, unhealthy, and self-defeating construct.

It seems reasonable to conclude that both the Sport-MPS and Hewitt-MPS were equally useful in identifying the anticipated relationship between unhealthy/maladaptive perfectionism and negative/unhealthy attitudinal body image. This position is further supported by the fact that the multivariate relationship between unhealthy perfectionism (as defined by the pattern of Sport-MPS subscale loadings) and unhealthy attitudinal body image ([R.sub.C]= .63) was similar in magnitude to the relationship between unhealthy perfectionism (as defined by the pattern of Hewitt-MPS subscale loadings) and unhealthy attitudinal body image ([R.sub.C]= .58).

In marked contrast to the ability of both the Sport-MPS and Hewitt-MPS to capture the positive relationship between unhealthy/maladaptive perfectionism and unhealthy/negative attitudinal body image, only the Sport-MPS provided results that supported the theorized link between healthy/adaptive perfectionism and healthy/positive attitudinal body image. Specifically, the canonical results showed a profile of healthy/adaptive perfectionism (as measured by the Sport-MPS) that directly reflected Stoeber and Otto's (2006) description of healthy perfectionism: namely, high perfectionist strivings (i.e., high personal standards) were combined with low perfectionist concerns (i.e., low concern over mistakes and low perceived coach pressure). This finding is important on a number of fronts. First, it provides evidence supporting the view that perfectionism can have healthy/adaptive functions in sport (see Dunn et al., 2002; Stoeber et al., 2007). Second, it highlights the usefulness of employing analytical techniques that consider scores across all perfectionism dimensions simultaneously when examining the functional (i.e., healthy vs. unhealthy) nature of perfectionism (Stoeber & Otto). Third, it provides empirical support for a link between healthy perfectionism and healthy/ positive attitudinal body image that has received little support in the perfectionism/body-image literature where global measures of perfectionism have been predominantly used. Lastly, and most importantly from the perspective of this study's objectives, this finding indicates that the Sport-MPS was capable of enhancing theoretical knowledge (through an understanding of healthy perfectionism in sport) that the Hewitt-MPS was apparently unable to achieve.

Overall, the regression results and canonical correlation results obtained in this study indicate that the Sport-MPS fared better than the Hewitt-MPS in (a) the prediction of unique variance in attitudinal body image, and (b) the production of knowledge that advances researchers' understanding of the functional nature of perfectionism in a sport setting. Taken together, such findings indicate that the Sport-MPS may be a better choice of instrument than the Hewitt-MPS when assessing perfectionism within sport contexts. Having said this, we wish to emphasize that we are not recommending the abandonment of the Hewitt-MPS as a measure of perfectionism in sport, nor are we suggesting that the Hewitt-MPS does not have an important role to play in the assessment of perfectionism in sport settings. For example, the Hewitt-MPS measures an aspect of perfectionism (i.e., other-oriented perfectionism) that the Sport-MPS does not measure, and it is conceivable that other-oriented perfectionism may play a role in certain team-sport contexts where athletes expect (or require) high performance standards from their team mates. Additionally, the contextual/domain specificity of the Hewitt-MPS can be modified quite easily to fit a variety of achievement domains (see Dunn et al., 2005; Mitchelson & Bums, 1998) whereas the item content of the Sport-MPS does not readily lend itself to this task. As such, the Hewitt-MPS is a much more useful measurement tool if the researcher is interested in comparing perfectionist tendencies across different achievement domains (e.g., Dunn et al.; Mitchelson & Bums). Indeed, the sport-specific nature of many items contained within the Sport-MPS (e.g., the perceived coach pressure items) would likely undermine the relevance of the instrument in any research setting other than competitive sport.

In closing, the present study represents an initial examination of the potential benefits that can be accrued by employing a domain-specific measure of perfectionism over a global measure of perfectionism in a competitive figure skating setting. Given that perfectionism levels differ across achievement domains (Mitchelson & Bums, 1998) and because individuals' behaviors vary across situational contexts (Mischel, Shoda, & Mendoza-Denton, 2002), we speculate that the superior performance of the Sport-MPS in this study may have been a function of athletes' uncertainties about the situational context they should choose when responding to Hewitt-MPS items (see Bing et al., 2004, for a related discussion). This uncertainty may have produced error variance in athletes' responses to the Hewitt-MPS that was not produced when the athletes responded to the Sport-MPS. Clearly answers to this type of speculation will not be achieved until more research is conducted that compares the ability of domain-specific and global measures of perfectionism to predict different cognitive, affective, and behavioral variables in a variety of competitive sport settings.

References

Anshel, M. H., & Eom, H. J. (2003). Exploring the dimensionality of perfectionism in sport. International Journal of Sport Psychology, 34, 255-271.

Bing, M. N., Whanger, J. C., Kristl Davison, H., & Van Hook, J. B. (2004). Incremental validity of the frame-of-reference effect in personality scale scores: A replication and extension. Journal of Applied Psychology, 89, 150-157.

Cash, T. F. (2000a). The Multidimensional Body-Self Relations Questionnaire: MBSRQ users 'manual (3rd revision, January, 2000). Norfolk, VA: Old Dominion University.

Cash, T. F. (2000b). Manual for the body-image ideals questionnaire. Norfolk, VA: Old Dominion University.

Cash, T. F., & Szymanski, M. L. (1995). The development and validation of the Body Image Ideals Questionnaire. Journal of Personality Assessment, 64, 466-477.

Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281-302.

Davis, C. (1997). Normal and neurotic perfectionism in eating disorders: An interactive model. International Journal of Eating Disorders, 22, 421-426.

Davis, C. (2002). Body image and athleticism. In T. F. Cash & T. Pruzinsky (Eds.), Body image: A handbook of theory, research, and clinical practice (pp. 219-225). New York: Guildford Press.

Dunn, J. G. H., Causgrove Dunn, J. L., Gotwals, J. K., Vallance, J. K. H., Craft, J. M., & Syrotuik, D. G. (2006). Establishing construct validity evidence for the Sport Multidimensional Perfectionism Scale. Psychology of Sport and Exercise, 7, 57-79.

Dunn, J. G. H., Causgrove Dunn, J. L., & Syrotuik, D. G. (2002). Relationship between multidimensional perfectionism and goal orientations in sport. Journal of Sport & Exercise Psychology, 24, 376-395.

Dunn, J. G. H., Gotwals, J. K., & Causgrove Dunn, J. (2005). An examination of the domain specificity of perfectionism among intercollegiate student-athletes. Personality and Individual Differences, 38, 1439-1448.

Dunn, J. (2 H., Gotwals, J. K., Causgrove Dunn, J., & Syrotuik, D. G. (2006). Examining the relationship between perfectionism and trait anger in competitive sport. International Journal of Sport and Exercise Psychology, 4, 7-24.

Enns, M. W., & Cox, B. J. (2002). The nature and assessment of perfectionism: A critical analysis. In G. L. Flett & P. L. Hewitt (Eds.), Perfectionism: Theory, research and practice (pp. 33-62). Washington, DC: American Psychological Association.

Flett, G. L., & Hewitt, P. L. (2002). Perfectionism and maladjustment: An overview of theoretical, definitional, and treatment issues. In G. L. Flett & P. L. Hewitt (Eds.), Perfectionism: Theory, research and practice (pp. 5-31). Washington, DC: American Psychological Association.

Flett, G. L., & Hewitt, P. L. (2005). The perils of perfectionism in sports and exercise. Current Directions in Psychological Science, 14, 14-18.

Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimensions of perfectionism. Cognitive Therapy and Research, 14,449-468.

Gauvin, L., & Russell, S. J. (1993). Sport-specific and culturally adapted measures in sport and exercise psychology research: Issues and strategies. In R. N. Singer, M. Murphey, & L. K. Tennant (Eds.), Handbook of research on sport psychology (pp. 891-900). New York: Macmillan.

Gotwals, J. K., Dunn, J. G. H., & Wayment, H. A. (2003). An examination of perfectionism and self-esteem in intercollegiate athletes. Journal of Sport Behavior, 26, 17-37.

Hall, H. K. (2006). Perfectionism: A hallmark quality of world class performers, or a psychological impediment to athletic development? In D. Hackfort & G. Tenenbaum (Eds.), Perspectives in sport and exercise psychology: Vol. 1. Essential processes for attaining peak performance (pp. 178-211). Oxford, U.K.: Meyer and Meyer.

Hamachek, D. E. (1978). Psychodynamics of normal and neurotic perfectionism. Psychology, 15, 27-33.

Hewitt, P. L., & Flett, G. L. (1991). Perfectionism and self and social contexts: Conceptualization, assessment, and association with psychopathology. Journal of Personality and Social Psychology, 60, 456-470.

Hollender, M. H. (1965). Perfectionism. Comprehensive Psychiatry, 6(2), 94-103.

Hunthausen, J. M., Truxillo, D. M., Bauer, T. N., & Hammer, L. B. (2003). A field study of frame-of-reference effects on personality test validity. Journal of Applied Psychology, 88, 545- 551.

Laux, L., Glanzmann, P., & Schaffer, P. (1985). General vs. situation specific traits as related to anxiety in ego-threatening situations. In C.D. Spielberger, I.G. Sarason, & P.B. Defares (Eds.), Stress and anxiety (Vol. 9, pp. 121-128). New York: Hemisphere.

Martens, R., & Simon, J.A. (1976). Comparison of three predictors of state anxiety in competitive situations. Research Quarterly, 47, 381-387.

Messick, S. (1989). Validity. In R. L. Linn (Ed.), Educational measurement (3rd ed., pp. 13-103). New York: American Council on Education.

Mischel, W., Shoda, Y., & Mendoza-Denton, R. (2002). Situation-behavior profiles as a locus of consistency in personality. Current Directions in Psychological Science, 11(2), 50-44.

Missildine, W. H. (1963). Your inner child of the past. New York: Simon and Schuster.

Mitchelson, J. K., & Bums, L. R. (1998). Career mothers and perfectionism: stress at work and at home. Personality and Individual Differences, 25, 477-485.

Pearson, C. A., & Gleaves, D. H. (2006). The multiple dimensions of perfectionism and their relation with eating disorder features. Personality and Individual Differences, 41,225235.

Rice, K. G., & Ashby, J. S. (2007). An efficient method for classifying perfectionists. Journal of Counseling Psychology, 54, 72-85.

Rice, K. G., Bair, C. J., Castro, J. R., Cohen, B. N., & Hood, C. A. (2003). Meanings of perfectionism: A quantitative and qualitative analysis. Journal of Cognitive Psychotherapy, 17,39-58.

Sagar, S. S., & Stoeber, J. (2009). Perfectionism, fear of failure and affective responses to success and failure. The central role of fear of experiencing shame and embarrassment. Journal of Sport & Exercise Psychology, 31,602-627.

Sherry, S. B., Hewitt, P. L., Besser, A., McGee, B. J., & Flett, G. L. (2004). Self-oriented and socially prescribed perfectionism in the Eating Disorder Inventory Perfectionism subscale. International Journal of Eating Disorders, 35, 69-79.

Stoeber, J., & Becker, C. (2008). Perfectionism, achievement motives, and attribution of success and failure in female soccer players. International Journal of Psychology, 43, 980-987.

Stoeber, J., & Otto, K. (2006). Positive conceptions of perfectionism: Approaches, evidence, challenges. Personality and Social Psychology Review, 10, 295-319.

Stoeber, J., Otto, K., Pescheck, E., Becker, C., & Stoll, O. (2007). Perfectionism and competitive anxiety in athletes: Differentiating striving for perfectionism and negative reactions to imperfection. Personality and Individual Differences, 42, 959-969.

Stoeber, J., Uphill, M. A., & Hotham, S. (2009). Predicting race performance in triathlon: The role of perfectionism, achievement goals and personal goal setting. Journal of Sport & Exercise Psychology, 31, 211-245.

Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York:

HarperCollins.

Taylor, G. M., & Ste-Marie, D. M. (2001). Eating disorders symptoms in Canadian female pair and dance figure skaters. International Journal of Sport Psychology, 32, 21-28.

Vallance, J. K.H., Dunn, J.G.H., & Causgrove Dunn, J. (2006). Perfectionism, anger, and situation criticality in competitive youth ice hockey. Journal of Sport & Exercise Psychology, 28, 383-406.

Footnote

With the exception of two athletes, the current sample formed a sub-sample in a study conducted by Dunn, Causgrove Dunn, et al. (2006). As such, the means, standard deviations, and correlations among the Sport-MPS and Hewitt-MPS subscales are based on the same data used in the Dunn, Causgrove Dunn, et al. publication. However, none of the current results associated with any of the body-image data have been previously published.

Author's Note

1.) John Dunn and Janice Causgrove Dunn are with the Faculty of Physical Education & Recreation, University of Alberta.

2.) Janelle Craft was a graduate student at the University of Alberta at the time of this study.

3.) John Gotwals is with the School of Kinesiology, Lakehead University.

4.) This paper was based upon the Masters thesis of Janelle Craft (nee Dunham) conducted under the supervision of John Dunn. The research was supported by a grant from the Sport Science Association of Alberta that was awarded to the first two authors.

Table 1. Subscale Means, Standard Deviations, and Internal
Consistency (a) Values for  Combined Sample (N=119)

Subscales                                   M     SD      a

Sport-MPS
  Personal standards                      3.35   0.85    .85
  Concern over mistakes                   2.39   0.96    .87
  Perceived parental pressure             2.41   0.94    .84
  Perceived coach pressure                2.16   0.87    .87

Hewitt-MPS
  Self-oriented perfectionism             5.05   1.00    .87
  Socially prescribed perfectionism       3.32   0.99    .85
  Other-oriented perfectionism            3.59   0.73    .67

MBSRQ-AS
  Appearance orientation                  3.51   0.72    .84
  Appearance evaluation                   3.47   0.83    .76
  Overweight preoccupation                2.52   0.98    .77
  Self-classified weight                  3.04   0.61    .92
  Body areas satisfaction                 3.64   0.72    .84

BIQ
  Body image ideal weighted discrepancy   1.06   1.49    .83

Table 2. Bivariate Correlatios (r) Among Subscales

                     Sport-MPS

Subscales    PS     COM    PCP    PPP

COM          .55    --
PCP          .26    .59    --
PPP          .12    .45    .47    --
SOP          .66    .64    .35    .18
SPP          .38    .69    .60    .61
OOP          .21    .16    .14    .03
AO           .50    .49    .27    .20
AE          -.02   -.34   -.33   -.22
OWP          .42    .49    .29    .25
SCW          .13    .33    .22    .26
BASS        -.04   -.36   -.28   -.21
BIQ          .14    .43    .30    .30

                  Hewitt-MPS                   Body Image

Subscales    SOP    SPP   OOP    AO     AE     OWP    SCW   BASS   BdQ

COM
PCP
PPP
SOP          --
SPP          .44    --
OOP          .22    .13    --
AO           .42    .33   .17    --
AE          -.12   -.30   .02   -.01    --
OWP          .39    .41   .22    .52   -.45    --
SCW          .31    .26   .03    .13   -.52    .59    --
BASS        -.21   -.35   .04   -.20    .66   -.47   -.45    --
BIQ          .32    .35   .12    .31   -.61    .56    .51   -.62    --

Note. Correlations > 1.181 are significant at p <.05.

Subscale abbreviations: PS= Personal standards COW =Concern over
mistakes. PCP = Perceived coach pressure. PPP =Perceived parental
pressure. SOP= Self-oriented perfectionism. SPP= Socially
prescribed perfectionism. OOP= Other-oriented perfectionism AO =
Appearance orientation AE = Appearance evaluation. OWP =
Overweight preoccupation. SCW = Self-classified vveight. BASS
=Body areas satisfaction scale. BIQ= Body image ideal
questionnaire (weighted discrepancy).

Table 3. Results of Hierarchical Regression Models Used for the
Prediction of Body Image Variables

                                         [DELTA]R
Block   Instrument                      [R.sup.2]    F Change     p

Appearance orientation (Final model: F [7, 111] = 6.849,
p < .001, [R.sup.2] = .30)

1       Hewitt-MPS                         .197       9.432    < .001
2       Sport-MPS                          .104       4.145    < .005

Appearance evaluation (Final model: F [7, 111 ] = 3.679,
p < .005, [R.sup.2] = .19)

1       Hewitt-MPS                         .091       3.815    < .05
2       Sport-MPS                          .098       3.344    < .05

Overweight preoccupation (Final model: F [7, 111] = 6.117,
p < .001, [R.sup.2] = .29)

1       Hewitt-MPS                         .227      11.268    < .001
2       Sport-MPS                          .051       1.969    = .104

Self-classified weight (Final model: F [7, 111 = 3.084,
p < .01, [R.sup.2] = .16)

1       Hewitt-MPS                         .116       5.040    < .005
2       Sport-MPS                          .047       1.545    = .194

Body areas satisfaction (Final model: F [7, 111] = 3.873,
p < .005, [R.sup.2] = .20)

1       Hewitt-MPS                         .116       5.040    < .005
2       Sport-MPS                          .080       2.765    < .05

BIQ weighted discrepancy (Final model: F [7, 1111 = 4.652, p < .001,
[R.sup.2] = .23)

1       Hewitt-MPS                      .157          7.116    < .001
2       Sport-MPS                       .070          2.522    < .05

Table 4. Canonical Loadings of Attitudinal Body Image and Perfectionism
Subscales for Significant Canonical Functions

                                     Canonical loadings

                          Sport perfectionism    Global perfectionism

Variables              Function 1   Function 2        Function 1

Attitudinal body
  image
  Appearance               .830         .213              .740
    orientation
  Appearance              -.450         .738             -.391
    evaluation
  Overweight               .861         .093              .852
    preoccupation
  Self-classified          .600        -.271              .674
    weight
  Body areas              -.448         .716             -.534
    satisfaction
  BIQ weighted             .691        -.490              .752
    discrepancy

Perfectionism
  (Sport-MPS)
  Personal standards       .770         .611              --
  Concern over             .946        -.305              --
    mistakes
  Perceived parental       .547        -.268              --
     pressure
  Perceived coach          .485        -.407              --
     pressure

Perfectionism
  (Hewitt-MPS)
  Self-oriented            --           --                .754
    perfectionism
  Socially                 --           --                .902
    prescribed
    perfectionism
  Other-oriented           --           --                .373
    perfectionism

Note. N = 118 (one multivariate outlier removed).