Facets of self-esteem at an implicit level? Investigation of implicit–explicit correlations and development of four IATs

Facets of self-esteem at an implicit level? Investigation of implicit–explicit correlations and development of four IATs

Personality and Individual Differences 53 (2012) 693–698 Contents lists available at SciVerse ScienceDirect Personality and Individual Differences j...

361KB Sizes 0 Downloads 14 Views

Personality and Individual Differences 53 (2012) 693–698

Contents lists available at SciVerse ScienceDirect

Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid

Facets of self-esteem at an implicit level? Investigation of implicit–explicit correlations and development of four IATs Elena Klavina a,1, Michela Schröder-Abé b,⇑,1, Astrid Schütz b a b

Department of Psychology, Chemnitz University of Technology, Germany Department of Psychology, University of Bamberg, Bamberg, Germany

a r t i c l e

i n f o

Article history: Received 10 February 2012 Received in revised form 18 May 2012 Accepted 21 May 2012 Available online 19 June 2012 Keywords: IAT Implicit self-esteem Explicit self-esteem Meta-analysis Convergent validity Implicit–explicit correlation

a b s t r a c t Correlations between implicit and explicit self-esteem have consistently been shown to be smaller than implicit–explicit correlations in other areas. As past research has not taken the facet structure of selfesteem into account, the present studies investigated area-specific implicit–explicit correlations. In Study 1, we conducted a meta-analysis to investigate the link between the Standard Self-Esteem Implicit Association Test and facets of explicit self-esteem. Across 11 samples, we found small but significant implicit– explicit correlations for the facets of social self-esteem and self-regard. In Study 2 (N = 223), we developed new IATs to measure facets of implicit self-esteem. Confirmatory Factor Analyzes (CFAs) showed that a facet structure can also be distinguished at the implicit level. Concerning correlations between implicit and explicit self-esteem, a model allowing for implicit–explicit correlations only within but not across facets of self-esteem yielded the best model fit. Significant latent correlations between implicit and explicit self-esteem were found for the social self-esteem facet. Implications of these findings concerning the measurement of implicit self-esteem, congruency versus discrepancy between explicit and implicit self-esteem, and the relevance of the different facets are discussed. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Self-esteem has attracted considerable interest from researchers for more than half a century. Recent developments in the field of implicit social cognition (Greenwald & Banaji, 1995) and the invention of the Implicit Association Test (IAT, Greenwald, McGhee, & Schwartz, 1998) have led to a renewed interest in the field, now shifting the focus to implicit self-esteem, people’s overlearned, automatic, and nonconscious self-evaluation (Bosson, Swann, & Pennebaker, 2000; Greenwald & Banaji, 1995). The Self-Esteem IAT (Greenwald & Farnham, 2000) and the Name Letter Task (NLT, Nuttin, 1985) turned out to be the most popular measures of implicit self-esteem. Whereas the predictive validity of these measures has been demonstrated (e.g., Rudolph, Schröder-Abé, Riketta, & Schütz, 2010), low implicit–explicit correlations have puzzled scientists since the early days of research on implicit self-esteem (e.g., Bosson et al., 2000). Krizan and Suls (2008) meta-analytically found a weak correlation of .115 between explicit self-esteem measures and the Name Letter Task. In a metaanalysis on the IAT (Greenwald et al., 1998), implicit self-esteem ⇑ Corresponding author. Address: University of Bamberg, Department of Psychology, Personality Psychology and Psychological Assessment, Markusplatz 3, 96047 Bamberg, Germany. Tel.: +49 951 863 1817. E-mail address: [email protected] (M. Schröder-Abé). 1 These authors contributed equally to this work and thus share first authorship. 0191-8869/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.paid.2012.05.028

yielded the lowest implicit–explicit correlation (.128) as compared to other research domains (Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005). In an attempt to explain low correlations in the field of implicit self-esteem, Bosson et al. (2000) argued that ‘‘implicit self-esteem is a complicated and multifaceted construct’’ (p. 640). Still, most studies have investigated implicit self-esteem as a unitary construct and have not attempted to differentiate aspects of it. Building on the hierarchical multifaceted model of self-concept (Shavelson, Hubner, & Stanton, 1976), the present research aimed at an in-depth investigation of implicit–explicit correlations that takes facets of explicit and implicit self-esteem into account. The four facets proposed by the Shavelson model (academic, social, emotional, and physical) are intercorrelated, but can be measured separately (Byrne & Shavelson, 1996; Shavelson et al., 1976), for example by means of the Self-Rating Scale (Fleming & Courtney, 1984) or its German adaptation, the Multidimensional Self-Esteem Scale (Schütz & Sellin, 2006). Previous studies investigating the relation between implicit and explicit self-esteem have used only global measures of explicit self-esteem: Whereas Rosenberg’s (1965) scale assesses the global evaluation of the self, the scale by Tafarodi and Swann (1995) measures self-liking and self-competence as two broad components of global self-esteem, but does not differentiate more specific facets as proposed by Shavelson et al.’s (1976) model. Similarly, the Standard Self-Esteem Implicit Association Test (SSEIAT) is not designed to assess domain-specific

694

E. Klavina et al. / Personality and Individual Differences 53 (2012) 693–698

aspects of self-esteem. Its most often used affective variant uses highly positive and negative attribute stimuli (e.g., diamond, health, sunrise; agony, filth, poison), and the evaluative variant uses a variety of positive and negative trait words (e.g., bright, noble, honest; ugly, vile, guilty) to assess implicit associations of the self with positive or negative affect (Greenwald & Farnham, 2000). Three recent studies suggest that the differentiation of facets of implicit and explicit self-esteem might be a fruitful undertaking when investigating implicit–explicit correlations. First, in an attempt to explain the relation between implicit self-esteem and narcissism, Campbell, Bosson, Goheen, Lakey, and Kernis (2007) devised two different IATs, one measuring agency-related, the other measuring communion-related self-esteem. The correlations with narcissism differed between the two IATs, highlighting meaningful differences with regard to the content of the IATs. Second, Oakes, Brown, and Cai (2008) found that the specificity and selfrelevance of the stimuli used in the IAT influences correlations with explicit self-esteem. Third, Gschwendner, Hofmann, and Schmitt (2008) demonstrated in a theoretically and statistically more sophisticated way that specificity and content similarity of IATs and questionnaires moderate implicit–explicit correlations in the domain of anxiety. Building on these recent studies, but referring to established models on the structure of self-esteem and self-concept (Fleming & Courtney, 1984; Shavelson et al., 1976), the present research investigated implicit–explicit correlations by taking facets of selfesteem into account. Study 1 investigated correlations between the affective variant of the SSEIAT and facets of explicit self-esteem measured with the German adaptation (Schütz & Sellin, 2006) of Fleming and Courtney’s (1984) scale. If implicit–explicit correlations differ when facets of explicit self-esteem are taken into account (e.g., higher correlations between the SSEIAT and performance self-esteem compared to physical self-esteem or the global score), this would indicate that the SSEIAT taps some specific aspects of implicit self-esteem (e.g., performance self-esteem) but not others (e.g., physical self-esteem) and that using the global self-esteem score obscures such differences. Study 2 went one step further by investigating whether facets of self-esteem can be differentiated at an implicit level, too. Towards this goal, we developed four new IATs that assess facets of self-esteem. We tested their factor structure using Confirmatory Factor Analyzes (CFAs) and hypothesized that a model specifying the four facets of self-esteem should show the best model fit. We then analyzed latent correlations with the respective subscales of the explicit self-esteem measure and assumed that significant implicit–explicit correlations should be found within facets of self-esteem (e.g., implicit and explicit social self-esteem) but not across facets (e.g., implicit social and explicit physical self-esteem). 2. Study 1 In order to explore the association between the SSEIAT and facets of explicit self-esteem, we conducted a meta-analysis on 11 independent samples from studies conducted in our lab. We followed this approach to reduce the problems associated with measurement error while reducing heterogeneity due to factors such as language or stimuli used in the SSEIAT. 2.1. Method 2.1.1. Included studies All 11 studies used the affective generic variant of the SSEIAT (Greenwald & Farnham, 2000; Rudolph, Schröder-Abé, Schütz, Gregg, & Sedikides, 2008) as the implicit self-esteem measure, and the Multidimensional Self-Esteem Scale (Schütz & Sellin,

2006) or subscales thereof as the explicit self-esteem measure. Participants were German university students, except for one study that investigated trainees of about the same age. Five samples were from published studies (Rudolph et al., 2010, Study 2; Rudolph et al., 2008, Studies 1 and 2; Schröder-Abé, Rudolph, & Schütz, 2007, Study 1; Schröder-Abé, Rudolph, Wiesner, & Schütz, 2007, Study 2), and six samples from unpublished datasets (Ns = 60, 55, 104, 78, 87, and 69). In all studies, implicit measures were completed before explicit measures. 2.1.2. SSEIAT The SSEIAT followed the seven-block structure proposed by Greenwald et al. (1998). Participants were instructed to respond as fast and as accurate as possible. In Block 1, the participants classified stimuli into pleasant (left key) and unpleasant (right key) categories (attribute-concept discrimination). In Block 2, they classified stimuli into me (left) and not-me (right) categories (targetconcept discrimination). In Blocks 3 and 4, the target and attribute stimuli had to be classified simultaneously (initial combined task, practice and test). In Block 5, key assignment for the target stimuli was reversed (not me left and me right). In Blocks 6 and 7, the target and attribute stimuli again had to be classified simultaneously, but using the reversed key assignment introduced in Block 4 (reversed combined task). Blocks 1, 2, and 5 consisted of 24 trials; Blocks 3, 4, 6, and 7 consisted of 48 trials. As we were interested in individual differences and correlations, we kept the order of the combined blocks constant to reduce method variance (Banse, Seise, & Zerbes, 2001). Data from the combined blocks (3, 4, 6, and 7) were used to compute IAT scores (D measure). Sample stimuli of the target dimension were me, my, myself, they, them, and their. Sample stimuli of the attribute dimension were holiday, health, peace, agony, sickness, and war.2 2.1.3. Multidimensional Self-Esteem Scale Explicit self-esteem was measured using the Multidimensional Self-Esteem Scale (MSES, Schütz & Sellin, 2006). Responses were made on 7-point scales with one of two types of end points ranging from 1 (not at all) to 7 (very much) or 1 (never) to 7 (always). The scale includes subscales measuring the following facets of self-esteem: (1) self-regard and (2) performance self-esteem (which correspond to the emotional and academic facets in Shavelson et al.’s model), (3) social self-esteem, and (4) physical self-esteem. Global self-esteem was computed as the mean of all of the items. 2.3. Results and discussion Pearson product moment correlations between the SSEIAT and the subscales of the MSES were included as effect-size measures. Meta-analytic computations followed the procedures suggested by Hedges and Olkin (1985). Correlations were transformed to Fisher’s-z scores, weighted by sample size and aggregated across studies (weighted mean of effect sizes, q). Assuming that the true effect can vary from study to study, we computed a random-effects model in which the common effect represents the mean of the population of true effects (Borenstein, Hedges, Higgins, & Rothstein, 2009). Table 1 shows the random-effects model results for the correlations between facets of explicit self-esteem and the SSEIAT. The estimated population correlations (q) between the SSEIAT and the explicit self-esteem facets of self-regard and social self-esteem were small but significant. The global score of the MSES, which is computed from the items of all subscales, was also significantly correlated with the SSEIAT. For the facets of performance and phys2

The original German stimuli can be obtained from the authors upon request.

695

E. Klavina et al. / Personality and Individual Differences 53 (2012) 693–698 Table 1 Meta-analytic results for the correlations between the standard self-esteem IAT and facets of explicit self-esteem.

Self-regard Social SE Performance SE Physical SE Global SE

q

z

p

.081 .109 .058 .058 .095

2.277 2.817 1.584 1.102 2.008

.023 .005 .113 .270 .045

95% CI .011–.149 .033–.183 .014–.129 .045–.160 .002–.187

SDp

Q

df

p

K

N

.106 .120 .107 .169 .141

9.317 11.897 8.508 14.467 11.883

10 10 9 8 8

.502 .282 .484 .070 .157

11 11 10 9 9

828 832 775 692 691

Note: q = mean population correlation; K = total number of study correlations; N = total sample size; z = value of test statistic assuming population correlation of zero; 95% IC = the lower and upper values of the 95% confidence interval for average effect size; SDp = standard deviation of population correlations; SE = self-esteem.

ical self-esteem, however, there were no significant correlations. The variability among study outcomes was nonsignificant as indicated by the Q-statistic. Excluding the study with the largest sample size or the nonstudent sample from the analyzes did not meaningfully change the results. Due to the small size of the correlations and the small differences between them, a statistical comparison of the different facets as suggested by Borenstein et al. (2009), chapter 24, however, did not yield significant results, so the differences between the facets must be interpreted with great caution. In sum, our meta-analysis on the correlations between the SSEIAT and explicit self-esteem yielded results that were similar to past research using the IAT (Hofmann et al., 2005) or the NLT (Krizan & Suls, 2008). When taking the facet structure of self-esteem into account, we found significant correlations only for self-regard and social self-esteem, but not for performance and physical selfesteem.

3. Study 2 The aim of Study 2 was twofold. First, we wanted to investigate whether facets of self-esteem similar to those proposed by Shavelson et al. (1976) or Fleming and Courtney (1984) could also be differentiated at the level of implicit self-esteem. This question is not trivial given the fact that implicit and explicit self-esteem are located in different systems of information processing (cf. Strack & Deutsch, 2004) that are working in different ways (e.g., associative links and motivational orientations in the impulsive system, and propositional organization of knowledge and syllogistic inferences in the reflective system). Given those different operating principles, it is possible that the complex structure of explicit self-concept and self-esteem is not mirrored by a similarly differentiated structure at the implicit level. Implicit self-esteem may as well be a one-dimensional and unspecific association of the self-concept with positive or negative valence or affect that does not allow for further differentiation. But as past research has found low correlations between different measures of implicit self-esteem (e.g., Bosson et al., 2000), it seems reasonable to assume a multifaceted structure of implicit self-esteem. In addition, recent research on implicit self-esteem (Campbell et al., 2007) and other constructs such as anxiety (Gschwendner et al., 2008) has shown that a more fine-grained differentiation might be possible at the implicit level too. To test for a possible differentiation of implicit self-esteem according to the Shavelson model, we developed four IATs measuring facets of self-esteem (see Section 3.1). Assuming that facets of implicit self-esteem can be distinguished, our second aim was to investigate latent correlations of those facets with explicit self-esteem. Gschwendner et al. (2008) demonstrated that content specificity and similarity moderates implicit–explicit correlations in the field of anxiety. In line with that, we hypothesized that a model specifying implicit–explicit correlations within facets of self-esteem but not allowing for correlations across facets (e.g., implicit and explicit social self-esteem, but not implicit social and explicit physical self-esteem) would

show the best fit to our data. In light of the findings from Study 1, we also wanted to explore which facets would yield significant implicit–explicit correlations. 3.1. Method 3.1.1. Participants and procedure A total of 223 (171 female) participants completed the study (age: M = 24.5 years, SD = 9.5). Participants studying psychology (major or minor, n = 163) received partial course credit for participation. Participants first completed four IATs and then self-report measures. To control for possible position effects, we counterbalanced the sequence of the IATs across subjects using a 4  4 Latin-square design with repeated measures. Participants were randomly assigned to one of the four conditions. 3.1.2. Facets of self-esteem IATs The IATs assessing facets of implicit self-esteem followed the same structure as the SSEIAT and included the same target stimuli (see Study 1). Given that past research has shown that the specific stimuli used in the IAT have a powerful impact on the IAT (Bluemke & Friese, 2006), we used different attribute stimuli but the same category labels (pleasant vs. unpleasant) for each of the four IATs. This was to make sure that the participants focus on the positivity vs. negativity of the stimuli and the IATs capture evaluative associations of the self with positivity vs. negativity (i.e., self-esteem) rather than more descriptive association (i.e., self-concept). Stimuli for the attribute categories were selected in several steps. (1) A large set of positively and negatively connotated selfdescriptive stimuli representing facets of self-esteem was compiled using scale descriptions and model definitions (e.g., Fleming & Courtney, 1984; Shavelson et al., 1976). (2) We asked a sample of academic staff and graduate students working in the field of self-esteem and self-concept (N = 16) to come up with self-

Table 2 English translations of the German stimuli (attribute dimension) used for the IATs in Study 2. Self-regard

Social SE

Performance SE

Physical SE

Pleasant Happy Satisfied Valuable Cheerful Confident Optimistic

Open-minded Interested Sociable Easygoing Unperturbed Popular

Intelligent Reliable Competent Clever Diligent Successful

Beautiful Agile Attractive Athletic Pretty Strong

Unpleasant Bad Unhappy Worthless Dissatisfied Sad Unstable

Withdrawn Resentful Hypersensitive Self-conscious Hurt Uncertain

Stupid Incompetent Lazy Sloppy Incapable Unfit

Repulsive Saggy Fat Gross Ugly Weak

Note: SE = self-esteem. German stimuli can be obtained from the authors upon request.

696

E. Klavina et al. / Personality and Individual Differences 53 (2012) 693–698

Table 3 Means, standard deviations, and intercorrelations among implicit and explicit self-esteem measures (Study 2).

Note: SE = self-esteem, Cronbach’s a is shown in the diagonal. For all correlations P.25, p < .01, for all correlations P.13, p < .05. Correlations between corresponding IATs and MSES subscales in bold, intercorrelations among the different IATs and among the different self-report measures in triangular borders.

descriptive adjectives for each of the self-esteem facets. (3) After excluding multiple mentions and rare and foreign words, we presented the remaining 109 words to a sample of students (N = 68) in order to pretest and select stimuli with respect to positivity and negativity. (4) For each facet, we chose six positive and six negative adjectives that represented the dimension to be measured in a broad way, but did not overlap with other facets. Table 2 shows the stimuli of the attribute dimension for the four new selfesteem IATs. 3.1.3. Multidimensional Self-Esteem Scale As in Study 1, explicit self-esteem was measured using the MSES (Schütz & Sellin, 2006). 3.2. Results and discussion 3.2.1. CFAs of IATs measuring facets of self-esteem We first conducted a series of CFAs in order to test the validity of the assumed four-factor structure of the newly developed IATs. Equivalent test-halves of equal length (cf. Rudolph et al., 2008) were used as observed variables, and the factor loadings of each subtest were set equal. We first tested a one-factor model, which would indicate that implicit self-esteem consists only of an affective (positive vs. negative) structure, and that specific facets cannot be differentiated. The one-factor model did not fit the data, p < .001, CFI = .592, RMSEA = .232, CAIC = 387.535. By contrast, the CFA yielded a good to acceptable model fit for the four-factor model, v2(18) = 36.856, p = .005, CFI = .973, RMSEA = .069, CAIC = 152.185. Latent correlations between the factors were between .235 and .560. Finally, we also tested a hierarchical model with four lower order factors (self-regard, social, performance, and physical self-esteem) and one higher order factor (global self-esteem) accounting for the lower order factors. This model showed an adequate to mediocre model fit that was clearly better than that of the one-factor model but slightly worse than that of

the four-factor model, v2(20) = 51.546, RMSEA = .084, CAIC = 154.061.

p < .001,

CFI = .955,

3.2.2. Internal consistencies, descriptive statistics, and correlations The reliabilities of all measures were good to satisfactory (see Table 3). All IATs and self-report measures deviated significantly from the theoretical midpoints of the scales in a positive direction (see Table 3, one-sample t and Cohen’s d). In sum, the internal consistencies and the magnitudes of the IAT effects of the newly developed IATs were comparable to those found for the SSEIAT in past research (e.g., Rudolph et al., 2008). Intercorrelations among the different IATs were very similar to those among the different self-report measures. The similarity between those two correlation matrices (presented within triangles in Table 3) was estimated as suggested by Schmukle, Back, and Egloff (2008). With possible values ranging from 1 to +1, the congruence coefficient of .97 revealed high similarity between the two correlation matrices, indicating that common method variance among the IATs was not higher than among the self-report scales. Correlations between corresponding IATs and MSES subscales (presented in bold in Table 3) were small and comparable in size to past research on implicit self-esteem and the results of Study 1. 3.2.3. CFAs assessing the relation between facets of implicit and explicit self-esteem To analyze the intercorrelations of the IATs and self-report measures in a more stringent way, we conducted a series of CFAs following the procedure suggested by Gschwendner et al. (2008). Equivalent test-halves were used as observed variables for all IATs and self-report measures. Model 1 assumed a single common factor explaining the correlations between the 16 manifest variables. Whereas Model 2 included two correlated method factors (explicit and implicit, Gschwendner et al., 2008) irrespective of the facet of self-esteem assessed, Model 3 included four correlated domain factors (self-regard, social, performance, and physical self-esteem) irrespective of whether they were assessed using explicit or impli-

697

E. Klavina et al. / Personality and Individual Differences 53 (2012) 693–698 Table 4 Fit indices and model comparisons for CFAs, IATs, and self-report measures assessing facets of self-esteem. Model

v2

df

p

CFI

RMSEA

CAIC

0 1 2 3

2285.41 1256.73 850.79 1472.75

120 112 111 106

.000 .000 .000 .000

.000 .446 .642 .338

.278 .215 .173 .241

513.36

98

.000

.799

109.95

96

.156

99.66

84

.117

4 5a

5b

Null (no relations among 16 indicators) One common factor Two correlated method factors (implicit, explicit) Four correlated domain factors (self-regard, social, performance, physical) MTMM-model: Four correlated domain factors and two orthogonal method factors; domain and method factors uncorrelated Four correlated implicit domain factors and four correlated explicit domain factors; correlations between method factors (implicit, explicit) only within domains Four correlated implicit domain factors and four correlated explicit domain factors; correlations between method factors (implicit, explicit) within and across domains

Model comparison

Dv 2

Ddf

p

2287.93 1410.50 1010.97 1664.97

0–1 1–2 1–3

1028.69 405.94 216.03

8 1 6

.000 .000 .000

.138

756.83

3–4

959.40

8

.000

.993

.026

366.24

3–5a

1362.80

10

.000

.992

.029

432.84

5a-5b

10.29

12

.590

Note: CAIC = consistent Akaike information criterion, CFI = comparative fit index, RMSEA = root-mean-square error of approximation. v2-difference tests (Dv2) are always relative to the described model comparison.

cit measures. Model 4 constituted an extension of Model 3, adding two orthogonal method factors in order to separate trait and method variance. Finally, models 5a and 5b included eight latent factors, as they specified the two method factors within each facet of selfesteem. Both models specified correlations among the implicit and among the explicit latent factors. The models differed, however, with respect to which correlations were allowed between the method factors: Model 5a allowed for implicit–explicit correlations only within facets of self-esteem, whereas Model 5b also allowed for correlations across facets. The results that are presented in Table 4 show that models 1–4 did not fit the data. Both Models 5a and 5b fit the data very well, but the CAIC was smaller for Model 5a, and allowing for additional correlations between explicit and implicit factors across domains did not significantly increase the model fit (see Table 4). Thus, according to conventions, the more parsimonious model was chosen. Fig. 1 shows the parameters of Model 5a. The only significant latent correlation was found between implicit and explicit social self-esteem, whereas latent correlations between implicit and explicit self-esteem were nonsignificant for the remaining facets of self-regard, performance, and physical self-esteem.

.71

Self-regard IAT 1

.66

.74

Social SE IAT 1 Social SE IAT 2

.24

.81

.42 .64

Performance SE IAT 2

Performance SE IAT

Physical SE .78 IAT 1 Physical SE IAT 2

.04

.49 Physical SE IAT

.79

.55 SE 1

Performance SE

.56 -.05

.90

.89

.92 Performance

Performance .88 SE IAT 1

.76

.74 .95 Social SE 1

Social SE

.84

.87

Social SE 2

.76 .58

.11

Social SE IAT

.56 .60

.62

.86

.78 .62

.92 Self-regard 1 Self-regard

Self-regard 2

.47 .78

.02

Self-regard IAT

.54

.65

The correlation between implicit and explicit self-esteem has been of enduring interest since the development of the first implicit self-esteem measures (Bosson et al., 2000). The present research extended previous approaches by taking the facet structure of selfesteem into account. Using a meta-analytic procedure, we investigated the correlations between the SSEIAT and facets of explicit self-esteem. Correlations were small and comparable to past research (Hofmann et al., 2005; Krizan & Suls, 2008). Significant implicit–explicit correlations were found for the facets of self-regard and social self-esteem, but not for performance self-esteem and physical self-esteem. After having investigated facets of explicit self-esteem in relation to a global implicit self-esteem measure, we developed four new IATs to measure facets of implicit self-esteem. CFAs showed that facets of self-esteem can be distinguished at an implicit level, too. Investigating latent correlations between facets of implicit and explicit self-esteem, we found that a model allowing for implicit–explicit correlations only within but not across facets of self-esteem yielded the best model fit. Thus, content similarity moderates implicit–explicit correlations in the field

.85 .82

Self-regard IAT 2

4. General discussion

Physical SE

.92

.85

Performance SE 2

.84

.97 Physical SE 1

.95

.55

.85 Physical SE 2

.73

Fig. 1. Confirmatory factor analysis (Model 5a from Table 4). All latent correlations with absolute size P.24, p < .01, and with absolute size P.11, p < .05.

698

E. Klavina et al. / Personality and Individual Differences 53 (2012) 693–698

of self-esteem, too (cf. Gschwendner et al., 2008), even though a significant latent correlation between implicit and explicit self-esteem was found for the social self-esteem facet only. Although one must interpret the differences between the different facets with great caution, the social facet seems to stand out as the most common link between implicit and explicit self-esteem. This dovetails with the findings of Oakes et al. (2008), who found significant implicit–explicit correlations for self-liking, which represents a sense of social worth (Tafarodi & Swann, 1995). The present study showed that even after we enhanced specificity and content similarity by differentiating facets of implicit and explicit self-esteem, implicit–explicit correlations did not rise substantially above the size of correlations that were found in previous research. How can these findings be explained? In past research, it has been argued that low correlations can be due to low reliabilities (Bosson et al., 2000). However, reliabilities of all measures were satisfactory in the present study, ruling out this explanation. Furthermore, past research showed that implicit and explicit selfesteem are more strongly correlated when explicit measures are presented first (Bosson et al., 2000; Krizan & Suls, 2008). As implicit measures preceded explicit measures in both Studies 1 and 2, the present studies might constitute a particularly strict test of the implicit–explicit relationship. Apart from these methodological and design considerations, one might wonder whether specific features of the construct of implicit self-esteem are responsible for the particularly low implicit–explicit correlations. Congruency versus discrepancy between implicit and explicit self-esteem seems to be a relevant personality characteristic in itself (e.g., SchröderAbé et al., 2007; Schröder-Abé, Rudolph & Wiesner et al., 2007). Concluding from our findings, implicit self-esteem is not only an unspecific association of the self-concept with positive or negative valence but seems to be as multifaceted as explicit self-esteem. Implicit and explicit self-esteem are at best related within but not between facets, and implicit–explicit correlations remain lower than in other research domains, even when taking the facets into account. Following up on substantive aspects of implicit selfesteem (such as the relevance of congruency vs. discrepancy or the reasons for implicit–explicit correlations varying between the facets) might therefore be a promising endeavor for future research. References Banse, R., Seise, J., & Zerbes, N. (2001). Implicit attitudes towards homosexuality: Reliability, validity, and controllability of the IAT. Zeitschrift für Experimentelle Psychologie, 48(2), 145–160. http://dx.doi.org/10.1026//0949-3946.48.2.145. Bluemke, M., & Friese, M. (2006). Do features of stimuli influence IAT effects? Journal of Experimental Social Psychology, 42(2), 163–176. http://dx.doi.org/ 10.1016/j.jesp. 2005.03.004. Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to meta-analysis. Chichester: Wiley. Bosson, J. K., Swann, W. B., Jr., & Pennebaker, J. W. (2000). Stalking the perfect measure of implicit self-esteem: The blind men and the elephant revisited? Journal of Personality and Social Psychology, 79(4), 631–643. http://dx.doi.org/ 10.1037/0022-3514.79.4.631. Byrne, B. M., & Shavelson, R. J. (1996). On the structure of social self-concept for pre, early-, and late adolescents: A test of the Shavelson, Hubner, and Stanton (1976) model. Journal of Personality and Social Psychology, 70(3), 599–613. http://dx.doi.org/10.1037/0022-3514.70.3.599.

Campbell, W. K., Bosson, J. K., Goheen, T. W., Lakey, C. E., & Kernis, M. H. (2007). Do narcissists dislike themselves ‘deep down inside?’. Psychological Science, 18(3), 227–229. http://dx.doi.org/10.1111/j.1467-9280.2007.01880.x. Fleming, J. S., & Courtney, B. E. (1984). The dimensionality of self-esteem: II. Hierarchical facet model for revised measurement scales. Journal of Personality and Social Psychology, 46(2), 404–421. http://dx.doi.org/10.1037/00223514.46.2.404. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, selfesteem, and stereotypes. Psychological Review, 102(1), 4–27. http://dx.doi.org/ 10.1037/0033-295x.102.1.4. Greenwald, A. G., & Farnham, S. D. (2000). Using the implicit association test to measure self-esteem and self-concept. Journal of Personality and Social Psychology, 79(6), 1022–1038. http://dx.doi.org/10.1037/0022-3514.79.6.1022. Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74(6), 1464–1480. http://dx.doi.org/10.1037/ 0022-3514.74.6.1464. Gschwendner, T., Hofmann, W., & Schmitt, M. (2008). Convergent and predictive validity of implicit and explicit anxiety measures as a function of specificity similarity and content similarity. European Journal of Psychological Assessment, 24(4), 254–262. http://dx.doi.org/10.1027/1015-5759.24.4.254. Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press. Hofmann, W., Gawronski, B., Gschwendner, T., Le, H., & Schmitt, M. (2005). A metaanalysis on the correlation between the implicit association test and explicit self-report measures. Personality and Social Psychology Bulletin, 31(10), 1369–1385. http://dx.doi.org/10.1177/0146167205275613. Krizan, Z., & Suls, J. (2008). Are implicit and explicit measures of self-esteem related? A meta-analysis for the Name–Letter test. Personality and Individual Differences, 44(2), 521–531. http://dx.doi.org/10.1016/j.paid.2007.09.017. Nuttin, J. M. (1985). Narcissism beyond Gestalt and awareness: The name letter effect. European Journal of Social Psychology, 15(3), 353–361. http://dx.doi.org/ 10.1002/ejsp. 2420150309. Oakes, M. A., Brown, J. D., & Cai, H. (2008). Implicit and explicit self-esteem: Measure for measure. Social Cognition, 26(6), 778–790. http://dx.doi.org/ 10.1521/soco.2008.26.6.778. Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Rudolph, A., Schröder-Abé, M., Riketta, M., & Schütz, A. (2010). Easier when done than said!: Implicit self-esteem predicts observed or spontaneous behavior, but not self-reported or controlled behavior. Zeitschrift für Psychologie/Journal of Psychology, 218(1), 12–19. http://dx.doi.org/10.1027/0044-3409/a000003. Rudolph, A., Schröder-Abé, M., Schütz, A., Gregg, A. P., & Sedikides, C. (2008). Through a glass, less darkly? Reassessing convergent and discriminant validity in measures of implicit self-esteem. European Journal of Psychological Assessment, 24(4), 273–281. http://dx.doi.org/10.1027/1015-5759.24.4.273. Schmukle, S. C., Back, M. D., & Egloff, B. (2008). Validity of the five-factor model for the implicit self-concept of personality. European Journal of Psychological Assessment, 24(4), 263–272. http://dx.doi.org/10.1027/1015-5759.24.4.263. Schröder-Abé, M., Rudolph, A., & Schütz, A. (2007). High implicit self-esteem is not necessarily advantagous: Discrepancies between explicit and implicit selfesteem and their relationship with anger expression and psychological health. European Journal of Personality, 21(3), 319–339. http://dx.doi.org/10.1002/ per.626. Schröder-Abé, M., Rudolph, A., Wiesner, A., & Schütz, A. (2007). Self-esteem discrepancies and defensive reactions to social feedback. International Journal of Psychology, 42(3), 174–183. http://dx.doi.org/10.1080/00207590601068134. Schütz, A., & Sellin, I. (2006). Multidimensionale Selbstwertskala (MSWS) [Multidimensional Self-Esteem Scale]. Göttingen: Hogrefe. Shavelson, R. J., Hubner, J. J., & Stanton, G. C. (1976). Self-concept: Validation of construct interpretations. Review of Educational Research, 46(3), 407–441. http://dx.doi.org/10.2307/1170010. Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8(3), 220–247. http:// dx.doi.org/10.1207/s15327957pspr0803_1. Tafarodi, R. W., & Swann, W. B. Jr., (1995). Self-liking and self-competence as dimensions of global self-esteem: Initial validation of a measure. Journal of Personality Assessment, 65(2), 322–342. http://dx.doi.org/10.1207/s15327752 jpa6502_8.