INTRODUCTION

Perfectionism is considered to be much more than a commitment to excellence and the sustained pursuit of exceedingly high standards. It is a complex personality characteristic that reflects a distorted and often irrational way of processing achievement information, and which underpins compulsive striving for flawlessness (^{Flett and Hewitt, 2002}). While various models of perfectionism (e.g., ^{Frost, Marten, Lahart and Rosenblate, 1990}; ^{Hewitt and Flett, 1991}; ^{Hill et al., 2004}; ^{Terry-Short, Owens, Slade and Dewey, 1995}; ^{Slaney, Rice, Mobley, Trippi and Ashby, 2001}; ^{Stoeber, 2011}) have emerged in the sporting literature, most researchers agree that perfectionism is multidimensional (^{Hill and Curran, 2016}; ^{Madigan, Stoeberg and Passfield, 2017}). Moreover, some authors have proposed a hierarchical structure, comprising two higher-order dimensions labelled perfectionistic strivings and perfectionistic concerns. The dimension of perfectionistic strivings represents striving for perfection and high personal standards for performance. In contrast, the dimension of perfectionistic concerns represents concerns over making mistakes, negative reactions to imperfection, and fears about failing to meet others' high expectations. The two dimensions broadly reflect either the energising or the critical, self-evaluative components of the construct, and because of this, each has a tendency to demonstrate a different pattern of association with a range of cognitive, affective and behavioural outcomes (^{González-Hernández, Capilla and Gómez-López, 2019}; ^{Haraldsen, Halvari, Solstad, Abrahamsen and Nordin-Bates, 2019}; ^{Stoeber and Otto, 2006}). Whereas perfectionistic strivings are often associated with adaptive processes and outcomes, the dimension of perfectionistic concerns is consistently associated with maladaptive processes and outcomes (^{Madigan et al., 2017}). In a recent review and reanalysis of existing evidence into perfectionism in sport, exercise, and dance, ^{Hill, Mallinson-Howard, Madigan and Jowett (2019)}concluded that perfectionistic strivings dimension is complex and ambiguous and that “this dimension (PS) would be unambiguously adaptive for athletes, dancers, and exercisers if it was not correlated with PC, but it typically is” (p.29).

Each model of multidimensional perfectionism, which has guided sport research, has adopted its own separate measurement methodology (e.g., ^{Frost et al., 1990}; ^{Hewitt and Flett, 1991}; ^{Hill et al., 2004}; ^{Terry-Short et al., 1995}; ^{Slaney et al., 2001}). The consequence of this has been the emergence of numerous subscales, which all purport to measure the various dimensions of the perfectionism construct.

Because there is growing evidence to suggest that perfectionism is a domain-specific construct (^{Dunn, Gotwals and Dunn, 2005}; ^{Mautz, Hill, Hueslman and Bazzini, 2017}) that tends to manifest differently across contexts, sport researchers have developed their own sport specific measures which are informed by different multidimensional models of perfectionism (Stoeber and ^{Madigan, 2016}). One sport specific instrument that has been developed is the Multidimensional Inventory of Perfectionism in Sports (MIPS; Stoeber, Otto and Stoll, 2004). This instrument that was developed in German and validated for use in both German and English languages, is based on a combination of different multidimensional models (^{Frost et al., 1990}; ^{Hewitt and Flett, 1991}) and the two-factor model described earlier, which suggests that perfectionism has two basic forms: a negative form and a positive form (^{Stoeber and Otto, 2006}). While the original version of the MIPS comprised nine subscales, previous research employing the MIPS with athletes (e.g., ^{Stoeber, Otto, Pesheck, Becker and Stoll, 2007}) administered a 10-item scale comprising two 5-item subscales that reflected striving for perfection (SP) and negative reactions to imperfection (NRI). ^{Stoeber and Madigan (2016)}report that structural equation modelling has shown these two dimensions to be reliable indicators of perfectionistic strivings and perfectionistic concerns. During the initial validation of the MIPS, the underlying factor structure was supported through the use of exploratory factor analysis (^{Stoeber et al., 2007}). More recently, ^{Madigan (2016)}has employed a confirmatory factor analysis and demonstrated that the factorial validity of the MIPS was acceptable in a sample of University athletes. As in other languages, the MIPS has been also validated in Spanish language, specifically, with elite athletes (M age = 23.3 years; SD = 5.62) representing several Latino American countries that participate in Central American and Caribbean games (^{Pineda-Espejel, Alarcón, López-Valle and Tomás, 2017}). Pineda-Espejel and his colleagues analysed the factorial structure, factorial invariance by gender and type of sport, and the validity based on the relationship with precompetitive anxiety. The results confirmed the two-factor structure of MIPS, the structural equivalence and partial invariance of factor loadings of the MIPS across sex and individual vs. collective sports, but not invariance of intercepts. Although the MIPS has been used previously with young athletes (e.g., ^{Madigan, Stoeber and Passfield, 2016}; ^{Stoeber, Stoll, Salmi and Tiikkaja, 2009}), the psychometrics of the scale in those sample, to our knowledge, have not yet being examined in any language. Hence, filling this gap was the initial interest of this work.

The sports psychologist has a long history in Spanish football and the interest in psychological work in football has been present in Spain since the late 1980s until today (^{García-Naveira, 2018}; ^{Peris, 2018}). Together, three were the objectives of the present study. First, to adapt the Multidimensional Inventory of Perfectionism in Sport (^{Stoeber et al., 2007}) items to be used with young Spanish footballers and to examine the factor structure and internal reliability of the instrument in a young adolescent sample and in a collective sport such as football. A second purpose of the present investigation was to ascertain the criterion related validity of the Spanish version of the MIPS in young footballers, by examining the associations between striving for perfection and negative reactions to imperfection with sport anxiety and contingent self-worth. Based on previous studies that have employed the MIPS in sport (e.g., ^{Gotwals, Stoeber, Dunn and Stoll, 2012}), it was hypothesized that negative reactions to imperfection and striving for perfection would show a positive relationship with both outcome variables, but that overlap between the two MIPS subscales would inflate the positive relationships between striving for perfection and maladaptive characteristics. Finally, in the conclusions to a recent paper, ^{Madigan (2016}) recommended that when undertaking further validation work on the MIPS researchers should examine the convergent validity of the instrument. Consequently, a third purpose of the present investigation was to determine the convergent validity by examining the association between the two subscales captured by the Spanish MIPS in young footballers and other perfectionism dimensions (i.e., self-oriented and socially prescribed perfectionism).

MATERIALS AND METHODS

RESULTS

Preliminary analysis

A missing value analysis was conducted on the data. Due to large amounts of missing data (> 5%), 10 participants were removed from the sample. An inspection of the pattern of missing data suggested a non-systematic mechanism for the missing data. Consequently, the missing data was considered as missing at random and estimated using the Expectation Maximization Algorithm for LISREL and SPSS analysis purposes. In the analysis for univariate outliers (standardised z-scores larger than 3.29, p < .001, two tailed), no outliers were found for MIPS, MPS, Anxiety and Contingent Self-worth items.

Analyses were based on 419 participants who had reasonably complete data for the MIPS, MPS, Anxiety, and Contingent Self-worth scales (i.e., they had missing values for no more than one item of the scales analysed).

The analysis of the normality distribution for the variables revealed moderate skewness and kurtosis in some items of the scales analysed. The univariate distributions of the items indicated that scores were non-normally distributed. Results showed negative skewness and negative kurtosis in some items. Kolmogorov-Smirnov test results also showed that scores were not normally distributed. Specifically, item 1 of MIPS (“I strive to be as perfect as possible”) had higher values of skewness and kurtosis, as well as the highest mean and lower standard deviation, indicating that most football players showed high levels of striving to be as perfect as possible (see Table 1).

Table 1. Descriptive statistics of the Spanish version of the Multidimensional Inventory of Perfectionism in Sport in young footballers 

^** = p < .01

Ske = Skewness; Kur = Kurtosis

Confirmatory factor analyses

A polychoric correlation matrix and asymptotic covariance matrix were derived from data, and each model was estimated with the weighted least squares method. Since the scores were non-normally distributed, the weighted least squares method for model estimation was appropriate in present study.

Table 2 shows the fit indices corresponding to the models analysed. First we tested a unidimensional model. Acceptance of this model implies that the subscales from the MIPS are not measuring different constructs (^{Kenny, 1979}). Secondly, we tested the two dimensional model proposed by the original authors of the MIPS (^{Stoeber et al., 2004}; ²⁰⁰⁷).

Table 2. Confirmatory factor analyses for the Spanish version of the Multidimensional Inventory of Perfectionism in Sport (MIPS) in young footballers 

df = degrees of freedom; χ2= Chi-Square; RMSEA = root mean square error of approximation; 90% CI = 90% confidence interval for the RMSEA; NNFI = non-normed fit index; CFI = comparative fit index.

^** = p < .01

The unidimensional model showed a ²= 3.29, suggesting a lack of fit, whereas both the NNFI and CFI suggested an acceptable fit with values of .91 and .93, respectively. The RMSEA suggested an acceptable fit with values of .07. The fit indices obtained for two-dimensional model showed a relative ²= 2.46, suggesting an adequate fit, and both the NNFI and CFI suggested a good and excellent fit with values of .95 and .96, respectively. The RMSEA suggested a good fit with value of .06. Figure 1 show the two-dimensional model. For this model, all parameter estimates were statistically significant. As demonstrated in Figure 1, representing the standardized solution of the two dimensional model, the lowest factor loading and higher uniqueness is for the item 1.

Note. All parameter estimates are standardized in relation to a common group metric, and all are statistically significant (p < .01); SP = Striving for perfection; NRI = Negative reactions to imperfection.

Figure 1. Factor loading, uniqueness and factor correlation for Model 2 

Internal consistency

The analysis of the internal consistency for the Spanish version of the MIPS in young footballers, the other subscales of perfectionism of the MPS (self-oriented perfectionism and socially prescribed perfectionism), anxiety and contingent self-worth showed Cronbach's alpha coefficients above .70 suggesting acceptable levels of internal reliability for each subscale/scale (see Table 3)

Table 3. Descriptive statistics, Cronbach's alphas, and correlations between perfectionism and maladaptive variables 

MIPS = Multidimensional Inventory of Perfectionism in Sport; MPS = Multidimensional Perfectionism Scale; Bivariate correlation = zero-order correlation; Partial correlation = partial correlation between striving for perfection (negative reactions to imperfection) and maladaptive variables, controlling for the influence of negative reactions to imperfection (striving for perfection). Internal reliability coefficients are presented on the diagonal.

^** = p < .01

DISCUSSION AND CONCLUSION

Three were the goals of the present study: First to examine the factor structure and internal reliability of striving for perfection and negative reactions to imperfection subscales from MIPS (^{Stoeber et al., 2007}) translated into Spanish in young football players. A further goal was to provide initial evidence of validity of the Spanish MIPS in young footballers by examining of the associations with measures of contingent self-worth and cognitive anxiety, and two additional measures of perfectionism. And the third purpose was to determine the convergent validity by examining the association between the two subscales captured by the Spanish MIPS in young footballers and other perfectionism dimensions (i.e., self-oriented and socially prescribed perfectionism).

Results from confirmatory factor analyses suggested better fit for a two-dimensional compared to a unidimensional model of the Spanish MIPS in young footballers. The findings from the confirmatory factor analyses are consistent with those recently reported by ^{Madigan (2016}) in a sample of University athletes, and also by the ones offered by ^{Pineda-Espejel et al. (2017}) in a sample of elite Latino American athletes' participating in a competition, providing additional evidence that the factor structure of the Spanish version of the MIPS replicates the original model proposed by ^{Stoeber and colleagues (Stoeber et al., 2007}). Furthermore, the current findings provide initial evidence regarding adequate internal consistency of the two subscales from the Spanish version of the MIPS in a sample of young footballers. Since in the study by Pineda et al (2017) the objective was to analyse perfectionism centered on competition (items preceded by the stem “during competition”) and in the present study we were interested in including the analysis of perfectionism both in training and competition (items preceded by the stem “when playing or practicing football”), it should be noted that both cases, the factor structure obtained supports the original model of striving for perfection and negative reactions to imperfection factors proposed by Stoeber and colleagues (2007).

Regarding the criterion validity of the Spanish MIPS in young footballers, correlation analyses revealed athletes' scores on the striving for perfection subscale were positively and significantly associated with contingent self-worth and anxiety at the bivariate level. A positive association between striving for perfection and contingent self-worth and anxiety has been reported elsewhere in the perfectionism literature (e.g., ^{Hill, Hall and Appleton, 2010}; ^{Gotwals et al., 2012}; ^{Pineda-Espejel et al., 2017}) and our findings lend further support to ^{Hill, Mallinson-Howard and Jowett (2018)}recent conclusion in their meta-analysis study that, although perfectionistic striving may have motivational qualities that enhance performance, it can also render athletes susceptible to motivational and psychological problems. Likewise, the finding that both associations became non-significant (albeit still remained positive) in the partial correlations provide support to the hypothesis that overlap between the two MIPS subscales would inflate the positive relationships between striving for perfection and maladaptive characteristics and complement the findings of Hill and colleagues' meta-analyses, suggesting that residual striving for perfection is more positive for indicators of young athletes' health and well-being than perfectionistic striving, in this case Spanish young footballers.

Regarding the negative reactions to imperfection subscale, correlations with athletes' contingent self-worth and anxiety scores were generally consistent at the bivariate and partial level; that is, the associations were positive and significant (albeit slightly weaker correlations were reported at the partial level). These findings are consistent with a main assumption of the theoretical model underpinning the scale (^{Gotwals et al., 2012}) and provide initial evidence that, as expected, negative reactions to imperfection and residual negative reactions to imperfection scores as measured by the Spanish MIPS in young footballers are related to a low sense of self-value and worries regarding performance in young footballers.

The analysis of the criterion related validity provides some interesting results regarding the Spanish version of MIPS in young footballers. A considerable overlap between the two perfectionism dimensions from the Spanish MIPS in young footballers was evident. This substantial construct overlap between striving for perfection and negative reactions to imperfection has been reported elsewhere in sport psychology research (see ^{Hill et al., 2018}), and further reinforces the importance of controlling for their overlap when researchers are interested in examining the unique relationship between each subscale and criterion variables (^{Stoeber, 2011}). Bivariate correlations also revealed positive and significant correlations between the striving for perfection and negative reactions to imperfection with self-oriented and socially prescribed perfectionism, providing initial support the convergent validity of the Spanish MIPS in young footballers.

Several strengths and limitations of this study are worth noting. Strength includes providing additional empirical evidence about the MIPS subscale structure and composition, validating a Spanish version of MIPS suitable for use with youth Spanish footballers. On a practical level, the availability of a validated instrument of perfectionism for use with Spanish youth athletes will allow to advance in knowledge of this complex personality characteristics as well as the relationship between perfectionism and youth athletes' well-and ill-being. A knowledge that is necessary for the development of strategies to improve the quality of the sport experience of young athletes. Limitations includes that the sample was predominantly male (87.8%). The gender of the sample is a biggest limitation. The results recently obtained by ^{Pineda-Espejel et al. (2017}) showing not invariance of intercepts, suggest that the responses of men and women in this instrument are not directly comparable. The sample of females obtained in the present study was too low to analyse the factorial invariance by gender of our Spanish version of the MIPS in young footballers. Thus, the findings should be taken with caution and further research is needed to evaluate the generalizability of the current. Another study limitation is the cross-sectional design that provides information on the relationships at a specific point of time and no indication of sequence of events in the relation between perfectionism and maladaptive characteristics. Future work may wish to use the Spanish version of the MIPS in young footballers to examine the influence of striving for perfection and negative reactions to imperfection overtime in other sport modalities.

Despite these limitations, the results of this study suggest that the Spanish version of the MIPS validated in young footballers is a well-adapted, reliable and valid sport specific instrument.