Maturation of Personality in Adolescence
Theo A. Klimstra, William W. Hale III, Quinten A. W. Raaijmakers,
Susan J. T. Branje, and Wim H. J. Meeus
Utrecht University
The present research assesses adolescent personality maturation by examining 3 measures of change and
stability (i.e., mean-level change, rank-order stability, and profile similarity) of Big Five personality
traits, employing data from a 5-annual-wave study with overlapping early to middle (n ⫽ 923) and
middle to late (n ⫽ 390) adolescent cohorts. Results indicated that mean levels of Agreeableness and
Emotional Stability increased during adolescence. There was mixed evidence for increases in Extraver-
sion and Openness. Additionally, rank-order stability and profile similarity of adolescent personality
traits clearly increased from early to late adolescence. For all change facets, the authors found evidence
for gender differences in the timing of adolescent personality maturation, as girls were found to mature
earlier than boys.
Keywords: personality, adolescence, longitudinal, maturation, five-factor model
Adolescence is generally regarded as being the formative period
in a person’s life. In this period, an individual physically trans-
forms from a child into an adult (Petersen, Crockett, Richards, &
Boxer, 1988). These physical changes are paralleled by psychosocial
changes: A transition takes place from being a child who relies
strongly on his or her parents’ teachings to becoming an adult who
makes his or her own informed decisions (Erikson, 1950). Ado-
lescence has been shown to be a period where individuals, for
example, gain an increasingly more stable identity status (Meeus,
Iedema, Helsen, & Vollebergh, 1999), form stable cultural orien-
tations (Vollebergh, Iedema, & Raaijmakers, 2001), and establish
increasingly more salient and intimate relations with peers and
romantic partners (Furman & Buhrmester, 1992). It is likely that
these changes toward maturation in biological and psychosocial
domains are also reflected by changes in personality traits.
Adolescent personality maturation should be indicated by
changes, signified by (normative) growth of personality traits (e.g.,
Caspi, Roberts, & Shiner, 2005). Maturation should also be re-
flected by increases in stability, as interindividual differences
should become more settled (e.g., Costa & McCrae, 1994), and
personality profiles should become more stable and better orga-
nized (e.g., Roberts, Caspi, & Moffitt, 2001). Nevertheless, ado-
lescent personality maturation, as measured by indices of change
and stability, has received much less attention than adult person-
ality maturation (Caspi et al., 2005). In the last decade, this
situation has changed as an ever-increasing number of studies have
focused on maturation of personality during adolescence. Despite
this increased attention, studies that longitudinally track personal-
ity maturation across adolescence on an annual basis in large
population samples are still lacking. Therefore, the current study
sets out to assess various facets of change and stability in person-
ality maturation across the entire period of adolescence (i.e., ages
12 to 20), using five-annual-wave data on overlapping early to
middle adolescent and middle to late adolescent cohorts.
Measuring Personality Maturation in Adolescence
Throughout the last 2 decades there has been a growing con-
sensus on the higher order structure of personality, as a majority of
researchers now agree that personality can be subsumed into five
broad traits: the Big Five (Caspi et al., 2005). The Big Five
personality traits are Extraversion, Agreeableness, Conscientious-
ness, Emotional Stability, and Openness to Experience (McCrae &
Costa, 1987). Extraversion refers to dominance and activity in
interpersonal situations; Agreeableness refers to the willingness to
maintain positive and reciprocal relationships with others; Consci-
entiousness refers to organizational and motivational aspects of a
person’s behavior; Emotional Stability indicates the ability to deal
effectively with negative emotions; and Openness to Experience
refers to how a person deals with new information at a personal
and experiential level. It has been demonstrated that the Big Five
traits apply to adolescent personality in the same way that they do
to adult personality (Digman & Inouye, 1986; John, Caspi, Robins,
Moffitt, & Stouthamer-Loeber, 1994) and that self-report is a
reliable and valid method of measuring Big Five traits in adoles-
cents (e.g., De Fruyt, Mervielde, Hoekstra, & Rolland, 2000; Soto,
John, Gosling, & Potter, 2008).
Several recent empirical studies (Akse, Hale, Engels, Raaijmak-
ers, & Meeus, 2007; Branje, van Lieshout, & Gerris, 2007; De
Fruyt et al., 2006; McCrae et al., 2002; Pullmann, Raudsepp, &
Allik, 2006; Roberts et al., 2001) and meta-analyses (Roberts &
DelVecchio, 2000; Roberts, Walton, & Viechtbauer, 2006) have
examined aspects of adolescent personality maturation. For this
purpose, these studies have focused on the three earlier mentioned
indicators of maturation: (a) mean-level growth (i.e., increases) on
Theo A. Klimstra, William W. Hale III, Quinten A. W. Raaijmakers,
Susan J. T. Branje, and Wim H. J. Meeus, Research Centre Adolescent
Development, Utrecht University, Utrecht, The Netherlands.
Correspondence concerning this article should be addressed to Theo A.
Klimstra, Research Centre Adolescent Development, Utrecht University,
P.O. Box 80.140, 3508 TC Utrecht, The Netherlands. E-mail: t.a
.klimstra@uu.nl
Journal of Personality and Social Psychology © 2009 American Psychological Association
2009, Vol. 96, No. 4, 898 –912 0022-3514/09/$12.00 DOI: 10.1037/a0014746
898
the Big Five traits, (b) an increase in rank-order stability of the Big
Five traits, and (c) a more stable and consistent organization of the
personality profile (i.e., increases in profile similarity). These three
indicators of maturation are now explored.
Mean-Level Change
A large number of longitudinal studies on mean-level change
were summarized in a meta-analysis by Roberts et al. (2006). They
demonstrated that adolescents generally become more extraverted,
more emotionally stable, and more open to experience as they
grow older. The adolescent personality traits of Agreeableness and
Conscientiousness tend to stay relatively stable over time. There is,
however, a lack of agreement on the amount and the direction of
change, as the statistic indicating heterogeneity of results among
the studies included in the meta-analysis reached significance for
all Big Five dimensions, except for Emotional Stability. There are
several possible causes for this heterogeneity: (a) Several studies
included a limited number of participants (n ⬍ 100), limiting the
generalizability of results from specific studies; (b) a vast majority
of the included longitudinal studies had high attrition rates
(⬎20%); (c) studies were published across a long period of time
(1968 –2004), possibly causing birth cohort effects (e.g., Twenge,
2000, 2001); and, as the authors themselves already indicated, (d)
a wide variety of measures were used that were not all specifically
designed to measure the Big Five, thereby causing measurement
variance.
Recent longitudinal studies on adolescent personality develop-
ment (Branje et al., 2007; De Fruyt et al., 2006; McCrae et al.,
2002; Pullmann et al., 2006) that were not included in the meta-
analysis of Roberts et al. (2006) all used measures exclusively
designed to measure the Big Five. Nevertheless, these studies also
show considerable heterogeneity of findings. One study that spe-
cifically focused on early adolescents (De Fruyt et al., 2006) found
increases in Emotional Stability, decreases in Openness and Con-
scientiousness, and stable levels of Extraversion and Agreeable-
ness. Two other studies focused on the period from early to middle
adolescence (Branje et al., 2007; McCrae et al., 2002). Both of
these studies found increases in Openness and stable levels of
Emotional Stability but did not agree on the developmental course
of the other three dimensions. Whereas Branje et al. (2007) found
decreases in Extraversion but increases in Agreeableness and Con-
scientiousness, McCrae et al. (2002) found stable levels of Extra-
version and Agreeableness and decreases in Conscientiousness.
These inconsistencies might be due to specific sample character-
istics, as McCrae et al.’s sample included only gifted students and
Branje et al.’s sample included only adolescents from middle-class
two-parent families with at least two children. Finally, one study
(Pullmann et al., 2006) compared personality changes in early,
middle, and late adolescents. Overall, they found more mean-level
change in early and middle adolescence than in late adolescence.
Specifically, their early adolescents increased in Emotional Stabil-
ity and Extraversion, decreased in Agreeableness, and had stable
levels of Openness and Conscientiousness. The middle adolescents
displayed increases in Emotional Stability, Extraversion, and
Openness but stable levels of Agreeableness and Conscientious-
ness. The late adolescents increased in Openness but exhibited no
changes on the other four Big Five traits. Although the study by
Pullmann et al. (2006) provides valuable information on adoles-
cent personality maturation because it covers the entire adolescent
period, they were able to measure personality traits only twice in
the same individuals. Therefore, they were unable to test whether
personality maturation is described by a linear or a curvilinear
pattern. In addition, their attrition rates were high (22%–57%),
which could have influenced their results.
In sum, the meta-analysis (Roberts et al., 2006) and the recent
empirical studies (Branje et al., 2007; De Fruyt et al., 2006;
McCrae et al., 2002; Pullmann et al., 2006) agree that levels of
Openness increase after early adolescence, whereas there is also
strong evidence for increasing levels of Emotional Stability. Con-
cerning the other three dimensions, there is less agreement on
mean-level change. Nevertheless, when considering the results of
the aforementioned studies, the number of Big Five dimensions
where mean-level increases, and thus maturation, were found
exceeds the number of dimensions where decreases were found.
Overall, previous studies suggest that when personality traits
change, they tend to change in the direction of maturation, al-
though there are inconsistencies regarding the specific traits that
mature and regarding the period of adolescence in which matura-
tion takes place.
Rank-Order Stability
Indices of rank-order stability indicate whether the rank order of
individuals on a certain trait is maintained over time. Mean levels
on a certain trait might change, but the rank order can remain
stable at the same time. Hence, mean-level change and rank-order
stability can be interpreted as two independent constructs (Roberts
& DelVecchio, 2000; Roberts et al., 2006). If mean-level changes
are accompanied by high rank-order stability, the observed mean-
level changes reflect normative change. Normative changes are
changes that occur to a similar degree in most people in the
population, and they are considered to reflect universal maturation
processes (e.g., Helson, Kwan, John, & Jones, 2002; Kasen, Chen,
Sneed, Crawford, & Cohen, 2006; Roberts et al., 2006). Therefore,
to asses whether there are such universal maturation processes,
indices of rank-order stability need to be assessed alongside indi-
ces of mean-level change. In addition, an increase in rank-order
stability can, by itself, also be interpreted as maturation, because
interindividual differences should become more set with age
(Costa & McCrae, 1994).
A meta-analysis by Roberts and DelVecchio (2000) and empir-
ical studies by Akse et al. (2007) and Pullmann et al. (2006)
indicated that rank-order stability of personality indeed increases
as adolescents grow older. However, the meta-analysis by Roberts
and DelVecchio used not only Big Five measures but also consid-
ered a wide range of measures such as ego-identity questionnaires,
temperament scales, and Rorschach tests. Akse et al. and Pullmann
et al. had only a limited number of measurement occasions for
rank-order stability (i.e., two and three, respectively) and could
therefore not systematically examine increases. In sum, the studies
that have been available up to now indicate that rank-order stabil-
ity of personality traits increases as adolescents grow older, just as
Costa and McCrae (1994) predicted. However, the aforementioned
studies have only a limited number of measurement occasions and
do not provide information that helps us to understand exactly
when in adolescence personality traits start to become more set-
tled. For that purpose, longitudinal studies systematically assessing
899
MATURATION OF PERSONALITY IN ADOLESCENCE
year-to-year changes in rank-order stability in adolescent Big Five
traits are needed.
Profile Similarity
Profile similarity provides information on the stability of a
constellation of traits over time within a person and is therefore,
unlike rank-order stability and mean-level change, not informative
of changes in individual personality traits. It is also different from
other change indices, because profile similarity indicates to what
extent a person’s personality configuration changes over time; it is
not indicative of changes at the sample level. Profile similarity is
usually calculated with q-correlations. Like Pearson correlations,
q-correlations range from ⫺1 to 1. The higher the q-correlation,
the more stable a constellation of traits within a person (e.g., Ozer
& Gjerde, 1989; Roberts et al., 2001). A more stable constellation
of traits indicates that a person’s personality profile is well orga-
nized and is therefore an indicator of maturation (Roberts et al.,
2001).
Previous studies on profile similarity of personality by Block
(1971), Roberts et al. (2001), and De Fruyt et al. (2006; the only
Big Five study on adolescent profile similarity) found average
q-correlations exceeding .70 in adolescence. The only study re-
porting on longitudinal changes in profile similarity from early
adolescence to late adolescence (Ozer & Gjerde, 1989) found that
profile similarity was stable across time. However, because only a
small sample of adolescents (n ⫽ 84) was employed, the general-
izability of these results is debatable. Large-scale longitudinal
studies reporting on systematic increases of profile similarity in
adolescents are not yet available.
The State of the Art of Studies on Personality Maturation
in Adolescence
Our discussion of research on the three change measures reveals
that there have been quite a few studies examining aspects of
personality maturation in adolescence, specifically during the last
decade. However, studies on mean-level change do not agree on
exactly what Big Five dimensions maturation processes are most
evident, there are no studies available that can inform us on exactly
where in adolescence interindividual differences in personality
traits start to become more settled, and large-scale longitudinal
studies systematically tracking changes in profile similarity are not
yet available. In addition, most of the aforementioned studies did
not asses all three indicators of maturation (i.e., increases on the
Big Five traits, increasingly stable rank order with regard to
interindividual differences, and a better organized personality pro-
file) in one design, even though it is not possible to gain a
comprehensive perspective on maturation of personality unless all
three indicators are assessed (e.g., Caspi & Roberts, 1999; Roberts
et al., 2001). The one study (De Fruyt et al., 2006) that did examine
all three indicators only assessed maturation in childhood and early
adolescence. The three indicators of maturation have so far not
been assessed together in studies covering other periods in ado-
lescence.
Another issue is that a majority of the published studies, includ-
ing the one by De Fruyt et al. (2006), measured adolescent per-
sonality traits only twice for the same participants. Changes found
in studies with only two longitudinal measurement occasions may
be caused by temporary factors that influence scores on one of the
measurement occasions, whereas consistent change across several
measurement occasions is more likely to reflect reliable develop-
mental trends. Consequently, the reliability of change trajectories
has been shown to increase steadily as the number of measurement
occasions increases (Willett, Singer, & Martin, 1998). In addition,
two measurement occasions allow for the assessment of only linear
change, which implies that it is possible to find no changes if the
underlying pattern is curvilinear. Finally, studies with only two
measurement occasions are unable to investigate systematic in-
creases in rank-order stability and profile similarity.
In sum, to advance the knowledge of adolescent personality
maturation, studies that longitudinally and systematically measure
various indicators of personality change and stability in a large and
broad sample of adolescents, using multiple measurement waves,
are needed. The current study aims to provide such a comprehen-
sive perspective on personality maturation.
Gender Differences in Personality
Because previous studies found gender differences in adolescent
personality change, we also consider the role of gender in adoles-
cent personality maturation. We now discuss the gender differ-
ences found by these previous studies.
With regard to the first aspect of personality maturation (i.e.,
increases in mean levels of Big Five traits), three of the aforemen-
tioned longitudinal studies on mean-level change in adolescent
personality traits explicitly examined gender differences. Two of
those studies compared overall mean levels of personality traits of
adolescent boys and girls. Branje et al. (2007) found that boys
tended to be more extraverted and open to experience than girls,
whereas McCrae et al. (2002) concluded the exact opposite. In
addition, McCrae et al. found higher levels of Agreeableness and
lower levels of Emotional Stability in girls when compared to
boys, whereas Branje et al. found no gender differences on these
two dimensions. With regard to changes, there is also little agree-
ment among studies. Branje et al. found decreases for boys and
increases for girls in Extraversion. For Agreeableness, Conscien-
tiousness, and Openness, they found stability for boys, but in-
creases for girls. Both studies agreed that there were no gender
differences in Conscientiousness. The other two longitudinal stud-
ies reporting on gender differences (McCrae et al., 2002; Pullmann
et al., 2006) found a more advantageous developmental pattern for
Emotional Stability for boys than for girls, as Pullmann et al.
(2006) reported increases for boys and stability for girls, and
McCrae et al. reported stability for boys and decreases for girls.
Emotional Stability was the only dimension where Branje et al. did
not find gender differences in change rates. Thus, there is little
agreement on gender differences in both mean levels and mean-
level changes in Big Five traits.
Gender differences concerning the second aspect of maturation
(i.e., increases in rank-order stability) were examined in three
studies. In their meta-analysis, Roberts and DelVecchio (2000) did
not find gender differences in rank-order stability across the life
span. Due to the underrepresentation of studies assessing gender
differences in personality change in the meta-analysis, they were
unable to focus on a specific period, such as adolescence. Further-
more, studies that specifically focused on adolescents either found
that girls were somewhat more stable than boys (Pullmann et al.,
900
KLIMSTRA ET AL.
2006) or found no gender differences (McCrae et al., 2002).
However, because none of the aforementioned studies systemati-
cally measured year-to-year changes in rank-order stability in
adolescence, it is not possible to use these studies to explore
whether interindividual differences in personality traits start to
become more set at a similar age for adolescent boys and girls. To
examine such possible gender differences in the timing of longi-
tudinal changes in rank-order stability for boys and girls, studies
that assess year-to-year changes in rank-order stability are needed.
Finally, with respect to the third aspect of adolescent personality
maturation (i.e., an increasingly more organized personality pro-
file), only two studies have examined gender differences. Ozer and
Gjerde (1989) found no substantial differences, whereas Roberts et
al. (2001) reported higher profile similarity in girls when compared
to boys. However, as previously mentioned, the former study
employed a small sample and the latter focused on the transition
from adolescence into adulthood. In addition, neither of these two
studies focused on Big Five trait profiles. Thus, gender differences
in mean levels and increases in Big Five profile similarity still
need to be investigated.
In sum, previous studies have examined gender differences for
all three aspects of adolescent personality maturation. However,
there is little agreement (i.e., for mean-level change), or there is
mixed evidence (i.e., for rank-order stability and profile similar-
ity), with regard to gender differences in personality change. These
inconsistencies in findings of previous studies could be due to the
possibility that gender differences mainly exist in the timing of
personality development, as suggested by Branje et al. (2007).
They proposed that the gender differences they found at age 16
(i.e., more increases in girls than in boys, and hence higher mean
levels on several personality traits for girls) could be caused by a
pubertal timing effect in personality. Branje et al. therefore sug-
gested that boys could catch up with girls with regard to person-
ality in late adolescence. However, they were unable to test this
hypothesis. Such differences in the timing of personality matura-
tion might exist, as there are profound gender differences in
indices of pubertal timing and neurological maturation processes.
On one of the most used indicators of pubertal timing, peak
height velocity (i.e., the age at which an adolescent exhibits the
fastest growth rate), girls have been shown to be 2 years ahead on
average, when compared to boys (e.g., Beunen et al., 2000;
Petersen et al., 1988). In addition, brain development in several
areas is up to 1 year ahead in girls (Giedd et al., 1999; Lynn, 1994).
Due to these gender differences in pubertal timing and neurolog-
ical development, girls usually look more mature and have a
cognitive advantage (Colom & Lynn, 2004) in the earlier stages of
adolescence. In a review, Paikoff and Brooks-Gunn (1991) indi-
cated that adolescents who mature earlier are also ahead on several
factors related to Big Five traits, such as social maturity, peer
prestige, leadership, and self-esteem. Gender differences in puber-
tal timing and neurological maturation are, therefore, likely to be
reflected in gender differences in Big Five personality maturation.
Hypotheses
In sum, we will examine whether adolescent personality matu-
ration is reflected by (a) increasing mean levels of Big Five traits,
(b) increasingly more settled interindividual differences, and (c)
better organized personality profiles. We expect increasing mean
levels of Big Five traits through adolescence but also substantial
interindividual differences in these increases. Previous studies
consistently found evidence for increases in Openness in adoles-
cence, but it is unclear in exactly which other traits maturation is
reflected. Increasingly more settled interindividual differences are
expected to be signified by increases in rank-order stability of
personality traits, whereas a better organized personality profile is
expected to be indicated by increases in q-correlations. Finally, we
expect gender differences in the timing of all three aspects of
personality maturation. Specifically, girls should exhibit high
mean levels, rank-order stability, and profile similarity of person-
ality at an earlier stage in adolescence than boys.
Method
Participants
Data for this study were collected as part of a five-annual-wave
longitudinal research project on Conflict and Management of
Relationships (CONAMORE; Meeus et al., 2006), with a 1-year
interval between each wave. The longitudinal sample was com-
posed of 1,313 adolescents. They were divided into an early to
middle adolescent cohort (n ⫽ 923; 70.3%), who were 12.4 years
old on average (SD ⫽ 0.59), and a middle to late adolescent cohort
(n ⫽ 390; 29.7%), who were 16.7 years old on average (SD ⫽
0.80) during the first wave of measurement. Because both age
groups were assessed during five measurement waves, a total age
range from 12 to 20 years was available.
The early to middle adolescent cohort consisted of 468 boys
(50.7%) and 455 girls (49.3%), and the middle to late adolescent
cohort consisted of 169 boys (43.3%) and 221 girls (56.7%). In the
younger cohort, 85.1% indicated that they were living with both
their parents, and in the older cohort a similar percentage of
adolescents (84.3%) reported the same. Other adolescents lived
with their mothers (7.9% and 7.2% in the younger and older
cohorts, respectively) or elsewhere (e.g., with their fathers, with
one biological parent and one stepparent, or with other family
members). The composition of the two cohorts did not signifi-
cantly differ with regard to ethnicity. In the younger cohort, 83.4%
identified themselves as Dutch, and 16.6% indicated that they
belonged to an ethnic minority (e.g., Surinamese, Antillean,
Moroccan, Turkish). In the older cohort these figures were 87.4%
and 12.6%, respectively. In the year that the current study was
initiated (2001), 21% of all Dutch early to middle adolescents and
22% of all Dutch middle to late adolescents belonged to ethnic
minorities (Statistics Netherlands, 2008a). Thus, ethnic minorities
were slightly underrepresented in our sample. With regard to
education, all participants initially were in junior high and high
schools. Given the Dutch educational system, most participants
changed school at least once during the study. Specifically, par-
ticipants in the younger cohort switched from junior high school to
high school, whereas most of the participants in the older cohort
switched from high school to college. Because of the sample
recruitment procedure, 100% of our middle to late adolescents
were in high school or college, whereas national demographic
statistics (Statistics Netherlands, 2008a, 2008b, 2008c) reveal that
96% of all Dutch middle to late adolescents were in some form of
education at that age during the period covered in the current study
(i.e., 2001–2005).
901
MATURATION OF PERSONALITY IN ADOLESCENCE
Sample attrition was 1.2% across waves: In Waves 1, 2, 3, 4,
and 5 the number of participants was 1,313, 1,313 (923 early to
middle adolescents and 390 middle to late adolescents), 1,293 (923
early to middle adolescents and 370 middle to late adolescents),
1,292 (923 early to middle adolescents and 369 middle to late
adolescents), and 1,275 (913 early to middle adolescents and 362
middle to late adolescents), respectively. For each analysis we
used the optimal method to estimate missing values. For the latent
growth curve analyses (used to asses mean-level change), we
therefore used the full information maximum likelihood (FIML)
procedure in Mplus (Muthe´n & Muthe´n, 2007). As the analyses
concerning rank-order stability and profile similarity were run in
SPSS, where FIML is not available, we used the expectation-
maximization (EM) procedure to estimate missing values for these
analyses. Across waves, 4.9% of the data were missing. Little’s
(1988) missing completely at random test, a regular test to com-
pare imputed with nonimputed data, revealed a normed chi-square
(
2
/df) of 1.35. According to guidelines by Bollen (1989), this
indicates a good fit between sample scores with and without
imputation.
Procedure
The participating adolescents were recruited from several ran-
domly selected schools in the province of Utrecht, The Nether-
lands. Of the 20 schools that were approached, 12 schools (60%)
decided to participate. There were no general differences between
participating and nonparticipating schools. All offered comparable
educational programs. We were unable to collect data on person-
ality or other variables among the pupils of the nonparticipating
schools. Students at the participating schools and their parents
received an invitation letter describing the research project and
goals and explaining the possibility of declining participation.
More than 99% of the approached students decided to participate.
All participants signed the informed consent form. The question-
naires were completed at the participants’ own (junior) high school
or at home, during annual assessments. Confidentiality of re-
sponses was guaranteed. Verbal and written instructions were
offered. The adolescents received €10 (approximately U.S. $15) as
a reward for every wave they participated in.
Personality Measures
Personality was assessed with a shortened 30-item Dutch ver-
sion of Goldberg’s Big Five questionnaire (Gerris et al., 1998;
Goldberg, 1992). In this instrument, a 7-point Likert scale with a
response format ranging from 1 (completely untrue)to7(com-
pletely true) is used to assess five personality dimensions: Extra-
version, Agreeableness, Conscientiousness, Emotional Stability,
and Openness to Experience. All dimensions are measured with six
items each, such as “talkative” (Extraversion), “sympathetic”
(Agreeableness), “systematic” (Conscientiousness), “worried”
(Emotional Stability, reverse scored), and “creative” (Openness to
Experience). Previous studies (e.g., Branje, van Lieshout, & van
Aken, 2004; De Fruyt et al., 2006; Dubas, Gerris, Janssens, &
Vermulst, 2002; Scholte, van Aken, & van Lieshout, 1997) have
demonstrated that this measure provides a valid and reliable esti-
mate of adolescent Big Five personality traits. In the current study,
reliability across waves was high, as the ranges of internal consis-
tency coefficients (Cronbach’s alphas) for each Big Five trait in
early to middle adolescents were as follows: Extraversion (.76 –
.85), Agreeableness (.80 –.88), Conscientiousness (.81–.88), Emo-
tional Stability (.81–.84), and Openness to Experience (.76 –.78).
For middle to late adolescents, these figures were as follows:
Extraversion (.86 –.91), Agreeableness (.81–.88), Conscientious-
ness (.87–.92), Emotional Stability (.79 –.85), and Openness to
Experience (.74 –.79).
Strategy of Analyses
We examined three types of change and stability (i.e., mean-
level change, rank-order stability, and profile similarity) in early to
middle and middle to late adolescents, and we examined gender
differences in these change measures. We estimated mean-level
change with latent growth curve modeling (LGCM; e.g., Duncan,
Duncan, Stryker, Li, & Alpert, 1999), using maximum likelihood
(ML) estimation in Mplus (Muthe´n & Muthe´n, 2007). LGCM
provides mean initial levels (i.e., intercepts) and mean change rates
(i.e., slopes), which are based on individual growth trajectories of
all participants. To indicate the extent to which there are interin-
dividual differences in growth, intercept and slope variances are
also incorporated in the analysis and provided in the output.
Therefore, LGCM integrates individual-level change and mean-
level change in one analysis (e.g., Branje et al., 2007). Different
shapes of development can be tested, as linear and curvilinear
shapes of development can be specified with the slopes. Because
latent variables are used in LGCM to estimate mean-level change,
it controls for measurement error (e.g., Duncan et al., 1999), which
is not possible with classical techniques, such as repeated measures
analysis of variance, which has been used in a majority of previous
studies on adolescent personality maturation (except for Branje et
al., 2007, who also used LGCM).
To assess gender differences in mean levels and change, we
used a multigroup design with four groups: early to middle ado-
lescent boys and girls and middle to late adolescent boys and girls.
All participants in the younger cohort were in their 1st year of
secondary education at Time 1 (T1) and in their 5th year of
secondary education at T5, whereas the participants from the older
cohort were in their 5th year of secondary education at T1. Mea-
surement wave was fully confounded with the number of years
adolescents had experienced secondary education (i.e., grade level)
and was also largely confounded with age (see sample descrip-
tion). No data were available concerning the exact chronological
age of the participants. Therefore, we used measurement occasion,
which was fully confound with grade level, to specify our growth
models. For convenience of presentation, we refer to the stage of
adolescence (i.e., early, middle and late adolescence), instead of
measurement wave or grade level, when discussing our results.
One wave of overlap existed between the two age cohorts.
During this one wave of overlap (i.e., T5 for early to middle
adolescents and T1 for middle to late adolescents), adolescents
were 16 years of age on average, and all had experienced 5 years
of secondary education. Therefore, we placed the intercept at T5
for early to middle adolescents and at T1 for middle to late
adolescents (i.e., slope factor loadings were ⫺4, ⫺3, ⫺2, ⫺1, and
0 for the five consecutive measurement occasions for early to
middle adolescents, and they were 0, 1, 2, 3, and 4 for middle to
late adolescents). We used the same growth factor loadings for
902
KLIMSTRA ET AL.
