خودارائه کمالگرایی واسطه رابطه بین دغدغه های کمال گرایی و بهزیستی ذهنی: مطالعه طولی سه موج

Abstract Dimensions of perfectionism are often proposed, but seldom integrated. Perfectionistic concerns and perfectionistic strivings were conceptualized as traits (core, relatively unchanging aspects of personality) and perfectionistic self-presentation as a characteristic adaptation (a contextualized cognitive-behavioral strategy). Theory suggests traits predispose people to engage in corresponding characteristic adaptations, and that perfectionistic concerns confer vulnerability for subjective well-being (SWB). It was hypothesized that perfectionistic concerns – but not perfectionistic strivings – would have an indirect effect on SWB through perfectionistic self-presentation. Young adults (ages 18–24) transitioning into university for the first time (N = 127) participated in a three-wave, 130-day longitudinal study. As hypothesized, perfectionistic self-presentation mediated the relationship between perfectionistic concerns and SWB. In contrast, perfectionistic strivings did not predict longitudinal change in perfectionistic self-presentation or SWB. This research integrates prior theory, and provides a novel test of hypotheses using longitudinal data.

Introduction Most personality research focuses on stable personality traits which crystallize by midlife. This level of personality is referred to as primary stabilities (Wakefield, 1989), the “having” aspect of personality (Cantor, 1990), basic tendencies (McCrae & Costa, 1999), traits (Fleeson & Leicht, 2006) and dispositional signatures (McAdams & Pals, 2006). This first level of personality is thought of as core, relatively unchanging attributes and behavioral tendencies of people. This level represents internal features of people that do not rely on specific contexts or situations. For instance, people high in neuroticism are thought to experience negative affect more strongly than other people, regardless of the situation (Nettle, 2009). We refer to this level of personality as a “trait”. Human individuality is not composed of traits alone. Theorists propose a second level of personality, which is referred to as secondary stabilities (Wakefield, 1989), the “doing” aspect of personality (Cantor, 1990), states (Fleeson & Leicht, 2006), or characteristic adaptations (McAdams and Pals, 2006 and McCrae and Costa, 1999). This level refers to cognitive and behavioral strategies used by individuals to deal with everyday demands of life and includes contextualized features of personality which are contingent on particular situations or developmental milestones. For instance, people are unlikely to be intrinsically motivated in all situations, so intrinsic motivation is best conceptualized at this level. We refer to this level of personality as a “characteristic adaptation”. 1.1. Perfectionism: Trait or characteristic adaptation? There is growing consensus on two major dimensions of perfectionism: Perfectionistic concerns and perfectionistic strivings (Dunkley et al., 2003 and Stoeber and Otto, 2006). Perfectionistic concerns include doubts about personal abilities, extreme concern over mistakes and being evaluated, and strong negative reactions to perceived failure. Perfectionistic strivings include rigidly and ceaselessly demanding perfection of oneself. These dimensions combine constructs from two dominant perfectionism research traditions: Cognitive-behavioral theory (Frost, Marten, Lahart, & Rosenblate, 1990) and personality/interpersonal theory (Hewitt & Flett, 1991). Evidence suggests perfectionistic concerns and perfectionistic strivings are stable, trait-like aspects of perfectionism (Graham et al., 2010, Hewitt and Flett, 1991 and Rice and Aldea, 2006). Theory and research suggest these dimensions are not context-specific and widely impact virtually all aspects of a person’s life (Hewitt, Flett, Besser, Sherry, & McGee, 2003a). These dimensions are best considered a “trait”. However, perfectionistic self-presentation is better conceptualized as a characteristic adaptation. Hewitt et al. (2003b) identified three components of perfectionistic self-presentation: Perfectionistic self-promotion (showcasing one’s supposed perfection), nondisplay of imperfection (concealing one’s imperfect behaviors), and nondisclosure of imperfection (avoiding verbal admissions of imperfection). Perfectionistic self-presentation is a contextual, situationally-activated social strategy that becomes more salient in certain relational contexts (Hewitt et al., 2003b), which is within the purview of characteristic adaptations. Consistent with this conceptualization, daily diary research shows self-concealment – a close analogue of nondisplay of imperfection – changes from day-to-day (Uysal, Lin, & Knee, 2010). McCrae and Costa (1999) assert traits (perfectionistic concerns) will predict increases in characteristic adaptations (perfectionistic self-presentation), rather than the reverse. Supporting this idea, participants with high levels of perfectionistic concerns show greater desire to keep their mistakes and personal information secret (Kawamura & Frost, 2004), even when it would be clearly advantageous to discuss their problems or limitations (Hewitt, Habke, Lee-Baggley, Sherry, & Flett, 2008). 1.2. Vulnerability models of perfectionism and subjective well-being We test a theoretical model of perfectionism which includes traits, characteristic adaptations and subjective well-being (SWB). SWB includes presence of positive affect, absence of negative affect, and general life satisfaction (Busseri & Sadava, 2011). We use a composite model of SWB, which involves summing all three components into a single composite variable. We prefer a composite model to a separate components model (i.e., viewing all three components as separate, orthogonal constructs) because of intercorrelations among SWB components, and because factor analyses support a single underlying factor (Linley, Maltby, Wood, Osborne, & Hurling, 2009). Personality strongly predicts SWB and setting realistic aspirations congruent with one’s personal resources is important (Diener, Suh, Lucas, & Smith, 1999). One widely researched model asserts perfectionism confers vulnerability for decreased SWB, but not the reverse (Hewitt & Flett, 2002). Longitudinal research suggests perfectionistic concerns confer vulnerability for decreased SWB (Chang, 2000, Graham et al., 2010 and Rice and Aldea, 2006). Perfectionistic self-presentation also confers vulnerability for decreased SWB in longitudinal research (Uysal et al., 2010). Results for perfectionistic strivings are inconsistent, with most research suggesting null relationships with SWB (Graham et al., 2010 and Hill et al., 2010). Moreover, perfectionistic strivings are largely unrelated to depressive symptoms and perfectionistic self-presentation once perfectionistic concerns are taken into account (Graham et al., 2010). 1.3. Rationale and hypotheses Most perfectionism research focuses on negative affect, rather than absence of positive outcomes. We advance past work by using a more comprehensive measure of functioning which encompasses both positive and negative components of SWB. There is also a shortage of multi-wave longitudinal research in perfectionism literature. Research using more than two waves of data is necessary to make stronger causal inferences about directionality, and is particularly persuasive when examining developmentally important periods of time where change is expected (Cole & Maxwell, 2003). Our research uses a three-wave, 130-day design to study transition to university, following freshman students across their first two semesters at university – a developmental transition associated with changes in personality and SWB (Lodi-Smith, Geise, Roberts, & Robins, 2009). We also use a longitudinal panel test of mediation (Cole & Maxwell, 2003), which represents one of the strongest tests of mediation in the perfectionism literature to date. Two hypotheses are proposed: (a) Perfectionistic concerns will indirectly affect SWB through perfectionistic self-presentation when controlling for perfectionistic strivings (Fig. 1); (b) perfectionistic strivings will not predict longitudinal change in perfectionistic self-presentation or SWB when controlling for perfectionistic concerns. Cross-lagged panel model of mediation. Rectangles represent measured variables. ... Fig. 1. Cross-lagged panel model of mediation. Rectangles represent measured variables. Black arrows represent hypothesized significant effects; grey dotted arrows represent hypothesized nonsignificant effects. Paths sharing the same number were constrained to equality. The indirect effect of perfectionistic concerns on subjective well-being through perfectionistic self-presentation was calculated by multiplying paths 2 and 5. Residual error terms are not displayed. Though not shown, perfectionistic strivings was entered in the model as a control variable.

Results 3.1. Data analytic strategy Overall, 6.3% of data were missing and covariance coverage ranged from 0.86 to 0.97. Listwise deletion was used for preliminary analyses and full information maximum likelihood estimation was used for hypothesis testing. In path analyses, we used MLR estimation in Mplus 6.0, which is robust against violations of multivariate normality (Muthen & Muthen, 2010). Model fit was assessed using multiple fit indices. Well-fitting models are suggested by a comparative fit index (CFI) and a Tucker-Lewis index (TLI) around .95 and a root-mean-square error of approximation (RMSEA) less than .06 (Kline, 2005). Mediation occurs if independent variables (perfectionistic concerns) lead to the mediator (perfectionistic self-presentation), which in turn leads to the dependant variable (SWB). We used Cole and Maxwell’s (2003) procedure for testing mediation with longitudinal data (Fig. 1). This improves on cross-sectional mediation by controlling for prior levels of variables at Waves 1 and 2, allowing researchers to examine rank-order change in outcomes over time. Indirect effects were calculated by multiplying paths from the independent variable to the mediator (a-paths) by paths from the mediator to the outcome (b-paths). When indirect effects are statistically significant, mediation has occurred. Statistical significance of indirect effects were calculated using bias-corrected bootstrapping with 20,000 resamples. If 95% bootstrapped confidence intervals (95% CI) do not contain zero, mediation has occurred (Little, Preacher, Selig, & Card, 2007). Bootstrapping is a nonparametric alternative used because the indirect effect typically has a skewed distribution. When model comparison tests are conducted, we used ΔCFI. Models are significantly different from one another if ΔCFI > −0.01 (Cheung & Rensvold, 2002). When conducting tests of equivalence, we compared the unconstrained model (i.e. all paths allowed to freely vary) with a constrained model (i.e. paths constrained to equality across waves). If model fit significantly worsens when constraints are added, equivalence across waves cannot be assumed. The most parsimonious model is the constrained model, which assumes relationships do not vary across different time lags. Two types of correlated error were specified a priori: (a) Within-trait, cross-wave correlated error and (b) same-trait, within-wave correlated error (Cole & Maxwell, 2003). These correlated error terms are used to account for violations of the independence assumption which typically occurs in longitudinal research. Wave 1 perfectionistic strivings was entered as a control variable by allowing it to correlate with all Wave 1 variables, and by including paths to perfectionistic self-presentation at Waves 2 and 3. 3.2. Preliminary analyses Means and standard deviations appear in Table 1 and bivariate correlations appear in Table 2. Means fell within one standard deviation of means from past studies of undergraduates (Graham et al., 2010). Alpha reliabilities ranged from .89 to .96, and test–retest correlations ranged from .60 to .91, supporting reliability. Perfectionistic strivings, perfectionistic concerns, and perfectionistic self-presentation were strongly correlated across waves (rs from .45 to .79). SWB was strongly and negatively correlated with perfectionistic concerns and self-presentation (rs from −.42 to −.61), but more weakly correlated with perfectionistic strivings (rs from −.15 to −.38) across waves. Perfectionistic strivings were uncorrelated with perfectionistic self-presentation and SWB at Waves 2 and 3 once controlling for perfectionistic concerns (rs from .00 to .12; ps > .05). Table 1. Means and standard deviations. Variable Wave 1 Wave 2 Wave 3 M SD M SD M SD Perfectionistic strivings HFMPS self-oriented perfectionism 4.74 1.07 4.52 1.11 4.26 1.23 FMPS personal standards 3.39 0.82 3.38 0.90 3.30 0.92 EDI Self-oriented perfectionism 3.24 1.08 3.19 1.13 3.22 1.12 Perfectionistic concerns HFMPS socially prescribed perfectionism 3.88 1.39 3.80 1.35 3.60 1.42 FMPS doubts about actions 2.75 0.96 2.76 0.97 2.74 1.02 FMPS concern over mistakes 2.88 1.22 2.92 1.21 2.73 1.21 Perfectionistic self-presentation Perfectionistic self-promotion 3.65 1.15 3.60 1.23 3.57 1.18 Nondisplay of imperfection 4.18 1.31 3.93 1.44 3.76 1.42 Nondisclosure of imperfection 3.67 1.21 3.51 1.19 3.35 1.21 Subjective well-being Positive affect 3.37 0.79 3.25 0.84 3.33 0.87 Negative affect 2.15 0.69 2.17 0.81 1.98 0.74 Life satisfaction 4.52 1.42 4.65 1.39 4.81 1.51 Note. N = 100. HFMPS = Hewitt and Flett’s (1991) Multidimensional Perfectionism Scale; FMPS = Frost et al.’s (1990) Multidimensional Perfectionism Scale; EDI = Garner et al.’s (1983) Eating Disorder Inventory. Means and standard deviations are based on averages calculated by summing all subscale items together then dividing by the number of items. Table options Table 2. Bivariate correlations and alpha reliabilities. Variable Wave 1 Wave 2 Wave 3 1 2 3 4 5 6 7 8 9 10 11 12 α Wave 1 1. Perfectionistic strivings – .91 2. Perfectionistic concerns .59 – .89 3. Perfectionistic self-presentation .57 .77 – .95 4. Subjective well-being −.19 −.49 −.49 – .91 Wave 2 5. Perfectionistic strivings .77 .53 .45 −.16 – .92 6. Perfectionistic concerns .55 .91 .72 −.50 .61 – .91 7. Perfectionistic self-presentation .56 .72 .87 −.49 .55 .73 – .96 8. Subjective well-being −.15 −.43 −.43 .69 −.19 −.50 −.56 – .92 Wave 3 9. Perfectionistic strivings .68 .54 .45 −.20 .81 .56 .52 −.23 – .93 10. Perfectionistic concerns .51 .89 .72 −.53 .51 .90 .71 −.49 .65 – .92 11. Perfectionistic self-presentation .46 .73 .79 −.46 .45 .72 .87 −.50 .54 .79 – .96 12. Subjective well-being −.27 −.49 −.42 .60 −.34 −.52 −.55 .71 −.38 −.61 −.58 – .94 Note. N = 100. Test–retest correlations appear in bold. Correlations with absolute values greater than .20 are significant at p < .05; correlations greater than .25 are significant at p < .01; and correlations greater than .32 are significant at p < .001. Table options 3.3. Discriminant validity Though the correlation between perfectionistic concerns and perfectionistic self-presentation at Wave 1 was large (r = .79), the bias-corrected, bootstrapped 95% CI for this correlation ranged from .71 to .84. Because the 95% CI does not include 1.0, this provides evidence of discriminant validity ( Anderson & Gerbing, 1988). See Hewitt et al. (2003b) for further data supporting the discriminant validity of perfectionistic self-presentation. 3.4. Path analysis Cole and Maxwell’s (2003) procedure for testing mediation was used (see Fig. 1). We first conducted tests of equivalence to determine if paths could be constrained to equality across waves. The unconstrained model was not a significant improvement over the constrained model, ΔCFI = −.003. Thus, we constrained all paths to equality across waves. The direct effect from perfectionistic concerns to SWB (β = −.05, p > .05) and the direct effect from perfectionistic strivings to SWB (β = −.04, p > .05) were both nonsignificant, and did not improve model fit, ΔCFI = .001. These paths were not added to the final model. The model in Fig. 2 fit the data well: χ2(N = 127) = 33.66, p = .07; χ2/df = 1.46; CFI = .99; TLI = .98; RMSEA = .06 (90% CI: .00, .10). The bias-corrected, bootstrapped indirect effect from perfectionistic concerns to SWB through perfectionistic self-presentation was statistically significant, 95% CI [−.005, −.0003]. Standardized paths in Fig. 2 differ slightly despite equality constraints (see Kline, 2005). Cross-lagged panel test of mediation. Rectangles represent measured variables. ... Fig. 2. Cross-lagged panel test of mediation. Rectangles represent measured variables. Numbers beside paths represent standardized path coefficients or correlations. Italicized, bolded numbers represent the proportion of variance accounted for by exogenous variables. Residual error terms are not displayed. Though not shown, perfectionistic strivings was entered in the model as a control variable. Solid black lines are statistically significant (p < .05). Dotted grey lines are nonsignificant (p > .05). Figure options Perfectionistic strivings was positively correlated with Wave 1 perfectionistic concerns (β = .58, p < .001), Wave 1 perfectionistic self-presentation (β = .53, p < .001), and Wave 1 SWB (β = −.18, p < .05). However, paths from perfectionistic strivings to perfectionistic self-presentation at Waves 2 and 3 were nonsignificant (β = −.02, p > .05). The indirect effect of perfectionistic strivings on SWB through perfectionistic self-presentation was also nonsignificant, 95% CI [−.005, .016]. Alternative models of SWB suggest positive affect, negative affect and life satisfaction are orthogonal constructs (Busseri & Sadava, 2011). Data were re-analyzed examining each SWB component separately. Mediation occurred when predicting positive affect and negative affect, but not life satisfaction. Perfectionistic concerns had an indirect effect on positive affect through perfectionistic self-presentation, 95% CI [−.03, −.001]. A similar indirect effect was found for negative affect, 95% CI [.0005, .041], but not life satisfaction, 95% CI [−.03, .002].