Content uploaded by Mitja D Back
Author content
All content in this area was uploaded by Mitja D Back
Content may be subject to copyright.
S.C. Schmukleetal.: Implicit Self-Concept of PersonalityEuropeanJournalof Psychological Assessment 20 08; Vol. 24(4):263–272© 2008Hogrefe& Huber Publishers
Validity of the Five-Factor Model
for the Implicit Self-Concept
of Personality
Stefan C. Schmukle, Mitja D. Back, and Boris Egloff
University of Leipzig, Germany
Abstract. The authors adapted the Implicit Association Test (IAT) in order to assess the implicit self-concept of personality. In two studies
(N= 106 and N= 92), confirmatory factor analyses validated the five-factor model for the implicit personality self-concept. Internal
consistencies of the IAT proved satisfactory for all Big Five personality dimensions. Correlations between the personality IAT and
different self-report measures of personality were generally small, and significant only for Extraversion and Conscientiousness. Patterns
of means and factor intercorrelations were, however, highly similar for implicit and explicit personality measures.
Keywords: IAT, implicit measures, Big Five, five-factor model
Introduction
At a superordinate level the personality structure can be
adequately described in terms of a few broad personality
dimensions (Saucier & Goldberg, 2001). The most com-
mon personality taxonomy is the five-factor model of Mc-
Crae and Costa (1987), which assumes that five broad and
robust factors, often referred to as the Big Five, account for
a considerable amount of covariation of personality traits.
These Big Five factors of personality are commonly la-
beled Extraversion, Neuroticism, Agreeableness, Consci-
entiousness, and Openness to Experience (Goldberg, 1990;
McCrae & Costa, 1987). The Big Five structure does not
imply that personality differences can be reduced to only
five dimensions. Rather, the Big Five represent a higher
level of a personality hierarchy, with each factor compris-
ing a large number of more specific, narrowly defined
traits. The Big Five taxonomy was originally developed by
analyzing peer rating scales (Tupes & Christal,
1961/1992), and was later confirmed in self-reports con-
cerning trait adjectives (Goldberg, 1990, 1992) and in per-
sonality questionnaire measures (McCrae & Costa, 1987).
Implicit Self-Concept of Personality
In all of the above cited studies participants either explicitly
judged themselves or were explicitly judged by others as
to the extent to which different personality traits applied to
them. Yet over the course of the past few years, an impres-
sive amount of evidence has been accumulated which
shows that individuals process information about them-
selves and their environment not only in an explicit (i.e.,
controlled or conscious) mode but also in an implicit (i.e.,
automatic or nonconscious) mode (Greenwald et al., 2002;
Strack & Deutsch, 2004). Accordingly, the explicit self-
concept of personality, which is based on propositional
structures, is to be distinguished from the implicit self-con-
cept of personality, which is based on associative structures
(Asendorpf, Banse, & Mücke, 2002). While the five-factor
model has already been validated for the explicit personal-
ity self-concept (McCrae & Costa, 1987), its validity for
the implicit personality self-concept remains untested.
Self-ratings of personality cannot be used when it comes
to assessing implicit self-concepts because information
processing in the explicit mode only has limited access to
automatic associations. Rather, implicit self-concepts are
to be assessed by procedures that measure personality more
indirectly. One of the most prominent indirect measures is
the Implicit Association Test (IAT; Greenwald, McGhee,
& Schwartz, 1998). The IAT measures the strength of as-
sociations between concepts by comparing response times
in two combined discrimination tasks. Participants are re-
quired to sort stimuli representing four concepts using only
two responses, each assigned to two of the four concepts.
The underlying assumption of the IAT is that if two con-
cepts are highly associated, the sorting task will be easier
(i.e., faster) when the two associated concepts share the
same response key than when they are assigned to different
response keys. The IAT can easily be adapted to assess the
self-concept of other personality traits by selecting appro-
priate categories and stimuli. For example, the IAT has
DOI 10.1027/1015-5759.24.4.263
© 2008 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2008; Vol. 24(4):263–272
been used to measure the implicit self-concepts of anxiety
(Egloff & Schmukle, 2002) and shyness (Asendorpf et al.,
2002).
Aims of the Present Research
The present two studies aimed to test the validity of the
five-factor model for the implicit self-concept of personal-
ity, as assessed by the IAT. The IAT has already been used
in assessing the Big Five personality dimensions in a recent
study (Steffens & Schulze-König, 2006). However, the fac-
torial structure was not tested but rather specified a priori.
Importantly, the structure of the implicit self-concept of
personality may not be identical to that of the explicit self-
concept. In particular, it may well be that the structure of
the implicit personality self-concept is less sophisticated
and less complex than the explicit personality self-concept.
For instance, the structure of the implicit personality self-
concept – as compared to the explicit concept – could po-
tentially be more strongly determined by the valence of the
different personality traits and might, thus, be better de-
scribed by a single evaluation factor.
We, therefore, analyzed the factorial structure and psy-
chometric properties of an IAT designed to assess the im-
plicit self-concept of personality in two studies. Our anal-
ysis comprised four components: (1) We tested the factorial
structure of the IAT using confirmatory factor analyses. (2)
We analyzed the reliability of the IAT, which is a prerequi-
site for assessing interindividual differences. (3) We tested
the correspondence between explicit and implicit person-
ality measures. (4) We investigated whether the patterns of
means and correlations between personality factors were
similar for implicit and explicit personality measures.
Study 1
Method
Participants
A total of 106 students (82 women, 24 men) participated
in exchange for research participation credit or monetary
compensation. Their average age was 23.4 years (SD =
5.5).
Personality IAT
The personality IAT was administered on a personal com-
puter and consisted of five subtests that were each de-
signed to measure one of the Big Five personality dimen-
sions. Each of the IAT subtests matched the seven-block
procedure proposed by Greenwald et al. (1998), with the
exception that the practice block for discriminating the cat-
egories me and others was only presented once at the be-
ginning of the whole test. Practice blocks consisted of 20
trials and critical blocks consisted of 60 trials. The IAT
subtests differed in the required attribute discrimination:
extraversion vs. introversion,anxiety vs. calmness,agree-
ableness vs. disagreeableness,conscientiousness vs. care-
lessness, and openness vs. narrow-mindedness. In selecting
these category labels, we refrained from using technical
terms of academic psychology, such as neuroticism or emo-
tional stability, in order to ensure comprehensibility for all
participants.
For the selection of stimuli, we chose prototypical items
for each category from a list of 823 adjectives assigned to
the Big Five personality dimensions (Ostendorf, 1994).
Table 1. Confirmatory factor analysis of item-specific Im-
plicit Association Test effects (Study 1)
Dimension Factors
Item-pair N E O A C
Neuroticism (N)
Anxious – calm .59
Nervous – relaxed .43
Fearful – restful .40
Uncertain – at ease .61
Afraid – balanced .38
Extraversion (E)
Sociable – shy .68
Talkative – reticent .72
Active – passive .63
Impulsive – deliberate .59
Outgoing – reserved .80
Openness (O)
Imaginative – unimaginative .57
Civilized – primitive .58
Well-educated – uneducated .48
Interested – indifferent .54
Gifted – limited .57
Agreeableness (A)
Trusting – obstinate .65
Well-meaning – quarrelsome .66
Friendly – hostile .61
Helpful – hard-hearted .57
Good-natured – resentful .53
Conscientiousness (C)
Meticulous – careless .53
Reliable – unreliable .43
Neat – chaotic .51
Fussy – frivolous .56
Thorough – erratic .73
Note. N = 106. All other factor loadings as well as covariances be-
tween error terms were set to zero.
264 S.C. Schmukle et al.: Implicit Self-Concept of Personality
European Journal of Psychological Assessment 2008; Vol. 24(4):263–272 © 2008 Hogrefe & Huber Publishers
Whenever feasible, we avoided the use of negated adjec-
tives common to many adjective lists (Goldberg, 1992;
Ostendorf, 1994), because such adjectives may allow for
alternative response strategies when completing the IAT
(e.g., sorting all stimuli that begin with “un-” into one cat-
egory). Stimuli from the categories me,others,extraver-
sion,introversion,anxiety, and calmness were identical to
those used in previous studies (Egloff & Schmukle, 2002;
Schmukle & Egloff, 2005). All adjective stimuli used in the
Big Five IAT are shown in Table 1. IAT data were treated
using an improved scoring algorithm (D1measure; Green-
wald, Nosek, & Banaji, 2003). We estimated the split-half
reliability of the IAT by applying the scoring algorithm to
two mutually exclusive subsets of the IAT combined-task
trials.
Item-Specific IAT Effects
In order to analyze the factorial structure of the personality
IAT, we computed separate IAT effects for each of the 25
bipolar adjective pairs shown in Table 1 (see Goldberg,
1992). These item-specific IAT effects were computed by
subtracting the mean reaction time for the two respective
adjective stimuli (e.g., sociable and shy) in the first critical
block from the mean reaction time for the two adjective
stimuli in the second critical block. Adjective pairs rather
than single adjectives were used to avoid confusion about
the reaction key. Furthermore, we did not use the D1mea-
sure for computing item-specific IAT effects, since this
measure requires an estimate of the intraindividual stand-
ard deviation which – because of the relatively small num-
ber of reaction times – is very fragile in the case of item-
specific effects. Instead, for the computation of item-spe-
cific effects, IAT reaction times were logarithmized and
treated in accordance with the procedure outlined by
Greenwald et al. (1998).
Self-Report Measures
Participants explicitly rated the 50 attribute stimuli of the
IAT on a scale ranging from 0 (not at all) to 5 (very much)
following the instruction: “Please indicate the extent to
which the following attributes apply to you.” Scores for the
five personality dimensions were computed as the mean of
the five stimuli describing the positive pole of the respec-
tive dimension and of the five reverse-scored stimuli de-
scribing the negative pole (e.g., the mean of the five extra-
version and the five reverse-scored introversion ratings).
Procedure
Explicit measures were presented before implicit measures.
The sequence of the five IAT subtests was counterbalanced
across subjects in a 5 × 5 Latin-square design with repeated
measures, in order to control for possible position effects.
Additionally, we counterbalanced the order of critical
blocks across subjects. Participants were randomly as-
signed to 1 of the 10 conditions.
Results
Confirmatory Factor Analysis
Before analyzing separate IAT scores for the five person-
ality dimensions, we first tested the validity of the five-fac-
tor model for the personality IAT by conducting a confir-
matory factor analysis of item-specific IAT scores. As de-
scribed in the Method section, these scores were based on
reaction times to a single specific pair of stimuli (e.g., so-
ciable – shy; see Table 1 for the stimulus pairs). Item-spe-
cific effects were allowed to load only on the factor to
which they belong according to the five-factor model, and
covariances between error terms were set to zero. This
model was estimated with a maximum likelihood (ML)
procedure by using Mplus (Muthén & Muthén, 2006). It fit
the data well, ML-χ2(265) = 346.42, p< .001, root-mean-
square error of approximation (RMSEA) = .05, stand-
ardized root-mean-squared residual (SRMR) = .08, partic-
ularly in view of the large number of degrees of freedom.
As Table 1 shows, factor loadings ranged from .38 to .80,
with a mean of .57. In sum, the confirmatory factor analysis
provides evidence for the validity of the five-factor model
for the implicit self-concept of personality.
Descriptive Statistics
Having presented evidence for the homogeneity of each
dimension of the personality IAT, we will now report re-
sults for IAT scores based on all critical trials of the respec-
tive Big Five dimension. The upper panel of Figure 1 dis-
plays descriptive statistics for the IAT and self-report mea-
sures relative to the midpoint of the scales (note that this
point does not necessarily represent the neutral value of the
underlying dimension since both scales have arbitrary met-
rics; see Blanton & Jaccard, 2006). For the IAT score, the
midpoint is zero indicating that both critical blocks are of
similar difficulty averaged across participants. For the trait
adjectives, this point was set to 2.5 because they were rated
with six response options ranging from 0 to 5. Overall, the
pattern of means relative to the midpoints were similar for
the IAT and self-report.
Multitrait-Multimethod (MTMM) Matrix
We next performed an MTMM analysis (Campbell & Fis-
ke, 1959) of the five personality dimensions measured by
either the IAT or self-report (see Table 2). All 10 measures
showed satisfactory internal consistencies. Importantly, re-
S.C. Schmukle et al.: Implicit Self-Concept of Personality 265
© 2008 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2008; Vol. 24(4):263–272
liabilities did not systematically differ between implicit and
explicit personality measures. Monotrait-heteromethod
correlations (in Table 2 in bold) were low and only signif-
icant for Extraversion and Conscientiousness, indicating
rather low convergence between implicit and explicit mea-
sures. Heterotrait-monomethod correlations were also rel-
atively low indicating satisfactory discriminant validity for
the five personality dimensions. Nevertheless, there were
low, yet significant associations between the five dimen-
sions both for the IAT and the self-report. Most interesting-
ly, the two heterotrait-monomethod matrices (in triangular
frames in Table 2) seemed to be very similar for implicit
and explicit measures. We estimated the size of the simi-
larity between the two correlation matrices obtained for im-
plicit and explicit personality measures using a coefficient
constructed analogously to the factor congruence coeffi-
cient (Burt, 1948):
where riis the ith element of q=p(p– 1)/2 distinct ele-
ments in each correlation matrix of pvariables. Similar to
the product-moment correlation and to the factor congru-
ence coefficient, possible values of cRrange from –1 to +
1, with zero indicating no similarity at all. The obtained
congruence coefficient of .79 indicated high similarity be-
tween the two heterotrait-monomethod matrices. An ad-
ditional test of similarity in the correlational pattern (Stei-
ger, 1980) using the computer program MULTICORR
Figure 1. Means and standard deviations of the Big Five personality dimensions measured with the Implicit Association
Test (IAT) compared to a self-rating of the IAT stimuli (Study 1; N= 106; upper panel) and to the NEO Five Factor
Inventory (NEO-FFI; Study 2; N= 92; lower panel).
266 S.C. Schmukle et al.: Implicit Self-Concept of Personality
European Journal of Psychological Assessment 2008; Vol. 24(4):263–272 © 2008 Hogrefe & Huber Publishers
(Steiger, 1987) indicated that the two matrices did not dif-
fer significantly from each other, χ2(10) = 12.12, p= .28.
Discussion
In summary, Study 1 yielded the following results: (1) A
confirmatory factor analysis provided evidence for the va-
lidity of the five-factor model for the implicit self-concept
of personality. (2) The reliability of the personality IAT was
satisfactory for all Big Five dimensions with internal con-
sistencies above .75. In fact, internal consistencies were
comparable to those obtained with self-report measures. (3)
Correlations between corresponding implicit and explicit
personality dimensions were rather low. The highest corre-
lations of approximately .30 were observed for Extraver-
sion and Conscientiousness. (4) In spite of the low implic-
it-explicit correlations, the patterns of means and factor in-
tercorrelations were highly similar between measures of
the implicit and the explicit personality self-concept.
Study 2
Study 2 aimed to replicate and extend the results of Study
1. In Study 1, the five-factor model was tested by using
item-specific IAT scores based on log transformed reaction
times (Greenwald et al., 1998). In order to rule out that the
validity of the five-factor model for the implicit self-con-
cept of personality is influenced by different scoring pro-
cedures, in Study 2 we tested the five-factor model with
IAT test halves that were computed by using the improved
scoring algorithm (Greenwald et al., 2003). We also com-
pared the five-factor model with a one-factor model as well
as with a hierarchical model consisting of five lower-order
factors loading on one higher-order factor. The first alter-
native model would hold if the implicit personality self-
concept was based only on the affective valence of person-
ality traits, and the latter model tests whether associations
between the five factors might be explained by an evalua-
tion factor.
Table 2. Multitrait-multimethod matrices of measures of implicit and explicit self-concepts of personality
Personality IAT Self-report
NEOACNE OAC
Study 1 (N= 106)
Personality IAT
Neuroticism (N) .77
Extraversion (E) –.18 .82
Openness (O) –.23 .32 .80
Agreeableness (A) –.17 .09 .22 .82
Conscientiousness (C) –.38 .13 .29 .20 .78
Self-report (personality traits)
Neuroticism (N) –.05 –.14 –.05 .01 .16 .87
Extraversion (E) .08 .31 –.05 .16 –.06 –.07 .82
Openness (O) .10 .05 .04 .09 –.06 –.14 .48 .77
Agreeableness (A) .07 –.04 .00 .08 –.10 –.17 .25 .38 .74
Conscientiousness (C) .05 .01 –.02 –.14 .26 –.09 .01 .18 .03 .78
Study 2 (N= 92)
Personality IAT
Neuroticism (N) .74
Extraversion (E) –.26 .88
Openness (O) –.26 .23 .77
Agreeableness (A) –.16 .11 .44 .78
Conscientiousness (C) –.44 .17 .07 .17 .80
Self-report (NEO-FFI)
Neuroticism (N) .11 –.38 –.04 .04 –.10 .90
Extraversion (E) .00 .32 .09 –.14 .00 –.39 .82
Openness (O) –.05 .07 .09 –.04 .03 –.31 .51 .70
Agreeableness (A) –.08 –.07 .01 –.11 –.13 –.11 .15 .18 .73
Conscientiousness (C) .04 .22 .06 .10 .22 –.30 .10 –.01 .10 .85
Note. IAT = Implicit Association Test. NEO-FFI = NEO Five Factor Inventory. Cronbach’s αis shown in the diagonal. Monotrait-heteromethod
correlations are in bold; heterotrait-monomethod correlations are shown in the triangular frames. Study 1: For all correlations ≥.20, p< .05,
and for all correlations ≥.26, p< .01. Study 2: For all correlations ≥.22, p< .05, and for all correlations ≥.30, p< .01.
S.C. Schmukle et al.: Implicit Self-Concept of Personality 267
© 2008 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2008; Vol. 24(4):263–272
Additionally, we investigated whether the similarity in
the pattern of means and the factor intercorrelations be-
tween implicit and explicit personality measures would
replicate. Given that identical trait adjectives were used in
the IAT and the self-report in Study 1, we wondered wheth-
er the structural similarities between the two modes of mea-
surement were dependent on the use of identical stimulus
materials, or whether this similarity is also generalizable to
self-report measures employing stimuli that differ from
those of the IAT. In addressing this question, we adminis-
tered the NEO-Five Factor Inventory (NEO-FFI; Costa &
McCrae, 1992) instead of adjective ratings as in Study 1.
Finally, Study 2 additionally explored whether implicit per-
sonality measures show incremental validity over explicit
measures in predicting self-reported behavior.
Method
Participants
A total of 93 students participated in the study in exchange
for research participation credit or monetary compensation.
One participant was excluded from the data analysis be-
cause of extraordinary high error rates (25% in average
across all subtests, with the highest error rate of 31% for
Agreeableness). The sample, thus, consisted of 92 students
(70 women, 22 men) with an average age of 23.1 years (SD
= 4.9).
Personality Measures
The personality IAT from Study 1 was used to assess the
implicit personality self-concept. Both the position of IAT
subtests and the order of critical blocks were again coun-
terbalanced using the same design as in Study 1. In assess-
ing explicit personality, we used the German version of the
NEO-FFI (Borkenau & Ostendorf, 1993). For presentation
purposes we report item means (ranging from 0 to 4) as
NEO-FFI scores instead of the more common item sums.
Additionally, we also assessed ratings of the 50 adjectives
of the IAT (as described in Study 1).
Behavior Report Form
Having completed the personality IAT, participants were
asked questions pertaining to everyday behaviors (see Table
3). These questions were inspired by the Behavior Report
Form introduced by Paunonen (1998). We included both be-
haviorsthat have been shown tobe related to explicit Big Five
measures (e.g., alcohol consumption, outgoingbehavior; see
Paunonen, 1998, 2003) and those behaviors that have proved
less predictable inprevious studies which, however, might be
related to implicit personality measures.
Results
Confirmatory Factor Analysis
We estimated and tested the five-factor model shown in Fig-
ure 2 using the program Mplus (Muthén & Muthén, 2006).
Scores of IAT test-halves were used as observed variables,
which were computed by separately applying the improved
scoring algorithm (D1measure; Greenwald et al., 2003) to
two mutually exclusive subsets of the IAT combined-task
trials. Factor loadings and error variances were set equal for
test-halves of each subtest. The five-factor model fitted the
data very well, ML-χ2(35) = 42.80, p= .17, RMSEA = .05,
SRMR = .08. Because the five factors were correlated, we
Table 3. Partial correlations of reported behavior with the personality IAT (sex partialed out), with the NEO-FFI (sex
partialed out), and with the personality IAT (sex and NEO-FFI partialed out) in Study 2
Personality IAT Self-Report (NEO-FFI) Personality IAT/NEO-FFI
Behavior NNEOACN E O A C NEOAC
Control variable: Sexa92 –.15 .02 –.18 –.08 .08 –.37** .09 .10 –.22* –.18 –––––
Alcohol consumption
Drinking?b92 –.10 .03 .06 –.01 .04 –.08 .27** .22 * .18 .01 –.09 –.01 .04 .05 .06
Quantity of drinking? 80 .00 –.12 –.01 –.24* –.16 –.04 .12 .07 –.02 –.19 .00 –.17 –.03 –.22 –.12
Drug consumption
Ever tested?b92 –.08 –.17 .08 –.04 –.08 .06 .13 .27** .13 –.32** –.06 –.13 .07 .01 .00
Using regularly?b92 .07 –.11 –.14 –.24* –.19 .04 .02 .09 .09 –.13 .09 –.07 –.14 –.22* –.15
Frequency of outgoingc92 .01 –.07 .15 –.16 –.17 –.12 .22* .23* .01 –.11 .02 –.15 .12 –.14 –.16
Maximum driving speed 88 –.23* .12 –.14 –.02 .09 –.01 .17 .15 –.10 –.11 –.26* .10 –.17 .00 .10
Giving charitable donationsb92 .00 –.17 –.07 –.02 –.08 .11 –.07 –.18 .00 –.04 .00 –.16 –.04 –.01 –.07
Freq. of cleaning windowsc87 –.07 .15 –.09 –.10 .22* –.08 .08 –.07 –.16 –.05 –.09 .08 –.11 –.10 .24*
Reading of classic literatureb92 .06 .10 .25* .03 –.12 –.20 .19 .34** –.05 .01 .08 .03 .21* .03 –.17
Note. aSex was coded female = 1, male = 2. bAnswers were coded yes = 1, no = 0. cFrequency data were square-root transformed to normalize
the distributions. IAT = Implicit Association Test. NEO-FFI = NEO Five Factor Inventory.
268 S.C. Schmukle et al.: Implicit Self-Concept of Personality
European Journal of Psychological Assessment 2008; Vol. 24(4):263–272 © 2008 Hogrefe & Huber Publishers
also tested a one-factor model, which would indicate that the
structure of the implicit personality self-concept is based only
on an affective evaluation of personality traits. Clearly, this
model failed to fit the data, ML-χ2(45) = 255.47, p<.001,
RMSEA = .23, SRMR = .18. Last, we tested a hierarchical
model with fivelower-order factors and one higher-orderfac-
tor that explains the associations among the five lower-order
factors. Although this model showed a better fit than the one-
factor model, ML-χ2(40) = 63.21, p= .01, RMSEA = .08,
SRMR = .11, the fit was, nevertheless, unsatisfactory and was
significantly worse than that of the five-factor model, χ2diff(5)
=20.41,p< .01. In sum, these analyses clearly confirmed the
validity of the five-factor model for the implicit personality
self-concept.
Descriptive Statistics
Means and standard deviations of IAT effects and NEO-FFI
scores are presented in the lower panel of Figure 1. Com-
parable to Study 1, the pattern of means relative to the mid-
points proved similar for the IAT and self-report.
MTMM Matrix
We next performed an MTMM analysis of the five IAT and
the five NEO-FFI variables (see Table 2). The correlations
between corresponding IAT and NEO-FFI scores were
rather low and only significant for Extraversion and Con-
scientiousness. Furthermore, correlational patterns of the
five personality dimensions did not differ between implicit
and explicit measures of personality, χ2(10) = 11.58, p=
.31 (computed by using MULTICORR; Steiger, 1987). The
congruence coefficient cRof .85 also indicated considerable
similarity of the two correlation matrices. For comparison,
we also estimated the similarity of correlational patterns for
adjective ratings and the NEO-FFI (cRof .92). This clarified
that correlational patterns of the IAT and self-report mea-
sures were only slightly less similar than those of the two
different self-report measures.
Behavior Report
Finally, we analyzed the associations of our measures of
implicit and explicit personality with self-reported behav-
ior. Because of observed significant sex differences for
some of our measures (in particular, self-reported Neurot-
icism and Agreeableness), we report all correlations with
sex partialled out. Table 3 shows that whereas both the IAT
and the NEO-FFI correlated with some of the reported be-
haviors, the explicit personality measures tended to show
stronger associations. Notably, after controlling for explicit
measures, implicit measures were still correlated with four
of the reported behaviors. That is to say, we obtained initial
Figure 2. Confirmatory factor analy-
sis of the Big Five Implicit Associa-
tion Test (IAT) in Study 2 (N= 92).
χ2(35) = 43.26, p=.16.Allfactor
loadings p< .001; all latent correla-
tions with absolute size ≥.57, p< .001,
and with absolute size ≥.29, p< .05.
S.C. Schmukle et al.: Implicit Self-Concept of Personality 269
© 2008 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2008; Vol. 24(4):263–272
evidence for the incremental validity of the Big Five IAT
compared to self-report scales. In particular, participants
with low implicit Neuroticism scores reported a higher
maximum driving speed; those with low implicit Agree-
ableness scores reported a higher level of alcohol and drug
consumption; those with higher implicit Conscientiousness
scores reported cleaning their windows more frequently,
and those with higher implicit Openness scores were more
likely to report reading classical literature.
Discussion
Study 2 successfully replicated the results of Study 1: (1)
Confirmatory factor analyses validated the five-factor
structure of the implicit self-concept of personality. The
implicit personality self-concept can neither be reduced to
differences in affective valence of personality traits, nor
can these valence differences exhaustively explain the as-
sociations among the personality dimensions. (2) Internal
consistencies of the personality IAT effects again proved
satisfactory. (3) We replicated the results of Study 1 with
respect to low implicit-explicit correlations in general, but
significant correlations for Extraversion and Conscien-
tiousness in particular. (4) As in Study 1, we observed sim-
ilar patterns of means and factor intercorrelations for im-
plicit and explicit measures of personality. Importantly, this
result was confirmed by the NEO-FFI, which uses different
stimuli compared to the IAT. We can, thus, rule out the
hypothesis that this similarity is dependent on the use of
identical stimulus material. In addition to Study 1, Study 2
provided initial evidence that a personality IAT has incre-
mental validity compared to personality questionnaires in
predicting self-reported behaviors.
General Discussion
We examined the psychometric properties of an IAT for
assessing the implicit self-concept of personality in two
studies. We analyzed (1) the factorial structure, (2) the re-
liability, (3) the correspondence between implicit and ex-
plicit self-concepts of personality, (4) the similarity of pat-
terns of means and intercorrelations between implicit and
explicit personality, and (5) incremental validity of the im-
plicit self-concept of personality with regard to behavior.
1. Results of confirmatory factor analyses yielded clear ev-
idence for five factors that were best described by the
Big Five personality dimensions. Our analyses, thus,
confirm that a five-factorial structure may not only be
an adequate description of the explicit (Goldberg, 1990;
McCrae & Costa, 1987), but also of the implicit self-
concept of personality. It is important to note that the
question pertaining to the validity of the five-factor
model for implicit personality is not a trivial one. For
instance, if the structure of implicit personality self-con-
cept had mainly been determined by emotional valence,
a model with one evaluation factor would also have fit
the data.
2. Internal consistencies of the personality IAT scores were
satisfactory for all Big Five dimensions in both studies.
This corresponds with previous studies that also showed
good internal consistencies for IATs assessing personal-
ity self-concepts such as anxiety (Back, Schmukle, &
Egloff, 2005; Egloff & Schmukle, 2002; Schmukle &
Egloff, 2006) or shyness (Asendorpf et al., 2002). How-
ever, stabilities of IAT effects are usually lower (Egloff,
Schwerdtfeger, & Schmukle, 2005, for an overview) and
should be explored for the personality IAT in future stud-
ies.
3. Both studies showed rather small correlations between
implicit and explicit personality dimensions. On aver-
age, these correlations amounted to approximately .13,
which was even lower than the mean correlation of .21
obtained in a meta-analysis of different self-concepts
(Hofmann, Gawronski, Gschwendner, Le, & Schmitt,
2005). Following the logic of the MTMM analysis
(Campbell & Fiske, 1959), the convergent validity of
measures of the implicit and explicit self-concept of per-
sonality is, thus, very low. However, in our view, mono-
trait-heteromethod correlations should, in this case, not
be interpreted as validity indicators since implicit and
explicit representations of the personality self-concept
represent different information processes (Greenwald et
al., 2002; Strack & Deutsch, 2004).
Interestingly, while explicit-implicit correlations were
generally low, the size of the correlation depended on
the construct. In both studies, the explicit-implicit cor-
relation was about .30 for Extraversion, was somewhat
lower yet still significant for Conscientiousness, and was
approximately equal to zero for the other three dimen-
sions. A variety of possible moderators of the associa-
tion between implicit and explicit measures has been
discussed (Hofmann, Gschwendner, & Schmitt, 2005;
Nosek, 2005). First, the implicit-explicit association
might be moderated by the degree of participants’
awareness of their implicit self-concept. For instance,
experimentally induced self-perception increases the as-
sociation of implicit and explicit anxiety (Egloff, Weck,
& Schmukle, in press). Since Extraversion and Consci-
entiousness are the two most easily observable of the
five dimensions (Borkenau & Liebler, 1992), partici-
pants might have a more accurate self-perception or re-
ceive more valid feedback from others regarding the for-
mer two dimensions, leading to enhanced explicit-im-
plicit associations. Second, the association might be
moderated by the degree to which participants adjust
their overtly expressed self-concept to social norms and
self-presentation concerns. Thus, the somewhat higher
implicit-explicit correlation for Extraversion compared
to Conscientiousness might be a result of the fact that
ratings of Extraversion are less influenced by social de-
sirability, and are, thus, less likely to be adjusted to re-
270 S.C. Schmukle et al.: Implicit Self-Concept of Personality
European Journal of Psychological Assessment 2008; Vol. 24(4):263–272 © 2008 Hogrefe & Huber Publishers
sponse factors such as self-presentation than are Consci-
entiousness ratings.
4. In spite of the rather low convergence between measures
of implicit and explicit personality, the pattern of means
was quite similar, with Neuroticism lying below the
midpoints of the scales, and Agreeableness, Conscien-
tiousness, and Openness consistently above the respec-
tive midpoints. Furthermore, intercorrelations between
Big Five dimensions were also highly similar for implic-
it and explicit personality measures.
Given the low convergence between implicit and explic-
it personality measures, the obtained evidence for a
structural equivalence of implicit and explicit personal-
ity self-concepts is a striking result. We replicated this
structural equivalence in a second study and, hence,
showed that it was observed neither purely by chance
nor exclusively in the restrictive case of identical stim-
ulus material in the IAT and the self-report scales. In an
unpublished study (N= 85), we found even more evi-
dence for the structural equivalence of implicit and ex-
plicit self-concepts. In this study, anxiety and self-es-
teem were both measured using the IAT and self-report.
IAT effects and self-report scores correlated neither for
anxiety nor for self-esteem, though the correlation be-
tween anxiety and self-esteem was very similar for IAT,
r= –.41, p< .001, and self-report measures, r= –.44, p
< .001. Thus, there appears to be accumulating evidence
that – while implicit and explicit self-concepts are rarely
associated – the associations within each of these self-
concepts are highly similar.
This structural equivalence might be influenced by the
valence of different personality traits. While apparently
neither implicit nor explicit personality self-concepts
can exhaustively be explained by a single evaluation fac-
tor (as, for example, indicated by the confirmatory factor
analyses in Study 2), the intercorrelations between per-
sonality dimensions might, nevertheless, be at least part-
ly explained by evaluation processes. For both implicit
and explicit personality measures, the more negatively
valenced dimension Neuroticism was consistently neg-
atively correlated with the other dimensions, whereas
the more positively evaluated dimensions Extraversion,
Agreeableness, Conscientiousness, and Openness were
all positively correlated with each other. This corre-
sponds with the fact that observed means were below the
midpoint exclusively for Neuroticism, and above the
midpoints for all other dimensions. Hence, basic evalu-
ation processes seem to be important for both implicit
and explicit personality self-concepts and might be one
reason for the structural similarities.
5. In Study 2, we broadened the focus of our research by
additionally assessing self-reported behaviors that are
conceptually related to the Big Five (Paunonen, 1998,
2003). Study 2 yielded evidence that measures of the
implicit self-concept are associated with some of these
behavioral criteria and, even more importantly, have in-
cremental validity compared to measures of explicit per-
sonality. This complements previous research showing
that a Big Five IAT has predictive validity for spontane-
ous behaviors (Steffens & Schulze-König, 2006).
To summarize, at a superordinate level the structure of the
explicit self-concept of personality can be adequately ex-
plained in terms of the five-factor model. Evidence for the
validity of the five-factor model for the implicit self-con-
cept of personality is provided for the very first time by the
current studies. One limitation of the present research per-
tains, however, to its reliance on a single method of implicit
assessment, namely the personality IAT. In contrast, the
five-factor model has been shown to be replicable across
different measures of the explicit self-concept of personal-
ity (Goldberg, 1990; McCrae & Costa, 1987). Future stud-
ies should, therefore, aim to replicate our results regarding
the structure of the implicit self-concept of personality with
reliable indirect measures other than the IAT.
Acknowledgments
This research was supported by Deutsche Forschungsge-
meinschaft (German Research Foundation) Grant EG
143/2–1 and 143/2–3.
References
Asendorpf, J.B., Banse, R., & Mücke, D. (2002). Double disso-
ciation between implicit and explicit personality self-concept:
The case of shy behavior. Journal of Personality and Social
Psychology, 83, 380–393.
Back, M.D., Schmukle, S.C., & Egloff, B. (2005). Measuring
task-switching ability in the Implicit Association Test. Exper-
imental Psychology, 52, 167–179.
Blanton, H., & Jaccard, J. (2006). Arbitrary metrics in psycholo-
gy. American Psychologist, 61, 27–41.
Borkenau, P. & Liebler, A. (1992). Trait inferences: Sources of
validity at zero acquaintance. Journal of Personality and So-
cial Psychology, 62, 645–657.
Borkenau, P., & Ostendorf, F. (1993). NEO-Fünf-Faktoren Inven-
tar (NEO-FFI) [NEO Five-Factor Inventory]. Göttingen: Ho-
grefe.
Burt, C. (1948). The factorial study of temperamental traits. Brit-
ish Journal of Psychology, 1, 178–203.
Campbell, D.T., & Fiske, D.W. (1959). Convergent and discrim-
inant validation by the multitrait-multimethod matrix. Psycho-
logical Bulletin, 56, 81–105.
Costa, P.T. & McCrae, R.R. (1992). Revised NEO Five Factor
Inventory. Professional manual. Odessa, FL: Psychological
Assessment Resources.
Egloff, B., & Schmukle, S.C. (2002). Predictive validity of an
Implicit Association Test for assessing anxiety. Journal of Per-
sonality and Social Psychology, 83, 1441–1455.
Egloff, B., Schwerdtfeger, A., & Schmukle, S.C. (2005). Tempo-
ral stability of the Implicit Association Test-Anxiety. Journal
of Personality Assessment, 84, 82–88.
S.C. Schmukle et al.: Implicit Self-Concept of Personality 271
© 2008 Hogrefe & Huber Publishers European Journal of Psychological Assessment 2008; Vol. 24(4):263–272
Egloff, B., Weck, F., & Schmukle, S.C. (2008). Thinking about
anxiety moderates the relationship between implicit and ex-
plicit anxiety measures. Journal of Research in Personality,
42, 771–778.
Goldberg, L.R. (1990). An alternative “description of personali-
ty”: The Big-Five factor structure. Journal of Personality and
Social Psychology, 59, 1216–1229.
Goldberg, L.R. (1992). The development of markers for the Big-
Five structure. Psychological Assessment, 4, 26–42.
Greenwald, A.G., Banaji, M.R., Rudman, L.A., Farnham, S.D.,
Nosek, B.A., & Mellott, D.S. (2002). A unified theory of im-
plicit attitudes, stereotypes, self-esteem, and self-concept. Psy-
chological Review, 109, 3–25.
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, 1464–1480.
Greenwald, A.G., Nosek, B.A., & Banaji, M.R. (2003). Under-
standing and using the Implicit Association Test: I. An im-
proved scoring algorithm. Journal of Personality and Social
Psychology, 85, 197–216.
Hofmann, W., Gawronski, B., Gschwendner, T., Le, H. & Schmitt,
M. (2005). A meta-analysis on the correlation between the Im-
plicit Association Test and explicit self-report measures. Per-
sonality and Social Psychology Bulletin, 31, 1369–1385.
Hofmann, W., Gschwendner, T., & Schmitt, M. (2005). On im-
plicit-explicit consistency: The moderating role of individual
differences in awareness and adjustment. European Journal of
Personality, 19, 25–49.
McCrae, R.R., & Costa, P.T., Jr. (1987). Validation of the five-
factor model of personality across instruments and observers.
Journal of Personality and Social Psychology, 52, 81–90.
Muthén, L.K., & Muthén, B.O. (2006). Mplus. user’s guide (4th
ed.). [Software manual]. Los Angeles, CA: Muthén & Muthén.
Nosek, B.A. (2005). Moderators of the relationship between im-
plicit and explicit evaluation. Journal of Experimental Psy-
chology: General, 134, 565–584.
Ostendorf, F. (1994). Zur Taxonomie deutscher Dispositionsbegrif-
fe [A taxonomy of German dispositional concepts]. In W. Hager
& M. Hasselhorn (Hrsg.), Handbuch deutschsprachiger Wort-
normen [Handbook of German word nouns] (S.382–441). Göt-
tingen: Hogrefe.
Paunonen, S.V. (1998). Hierarchical organization of personality
and prediction of behavior. Journal of Personality and Social
Psychology, 74, 538–556.
Paunonen, S.V. (2003). Big Five factors of personality and repli-
cated predictions of behavior. Journal of Personality and So-
cial Psychology, 84, 411–424.
Saucier, G., & Goldberg, L.R. (2001). Lexical studies of indige-
nous personality factors: Premises, products, and prospects.
Journal of Personality, 69, 847–879.
Schmukle, S.C., & Egloff, B. (2005). A latent state-trait analysis
of implicit and explicit personality measures. European Jour-
nal of Psychological Assessment, 21, 100–107.
Schmukle, S.C., & Egloff, B. (2006). Assessing anxiety with ex-
trinsic Simon tasks. Experimental Psychology, 53, 149–160.
Steffens, M.C., & Schulze-König, S. (2006). Predicting sponta-
neous Big-Five behavior with Implicit Association Tests. Eu-
ropean Journal of Psychological Assessment, 22, 13–20.
Steiger, J.H. (1980). Tests for comparing elements of a correlation
matrix. Psychological Bulletin, 87, 245–251.
Steiger, J.H. (1987). MULTICORR (Version 2.4) [Computer soft-
ware]. University of British Columbia.
Strack, F., & Deutsch, R. (2004). Reflective and impulsive deter-
minants of social behavior. Personality and Social Psychology
Review, 8, 220–247.
Tupes, E.C., & Christal, R.E. (1992). Recurrent personality fac-
tors based on trait ratings. Journal of Personality, 60, 225–251.
(Original work published 1961).
Stefan C. Schmukle
Department of Psychology
University of Leipzig
Seeburgstraße 14–20
D-04103 Leipzig
Germany
E-mail schmukle@uni-leipzig.de
272 S.C. Schmukle et al.: Implicit Self-Concept of Personality
European Journal of Psychological Assessment 2008; Vol. 24(4):263–272 © 2008 Hogrefe & Huber Publishers
A preview of this full-text is provided by American Psychological Association.
Content available from European Journal of Psychological Assessment
This content is subject to copyright. Terms and conditions apply.