Content uploaded by Reinout E. de Vries
Author content
All content in this area was uploaded by Reinout E. de Vries on Jan 04, 2023
Content may be subject to copyright.
How Genetic and Environmental Variance in Personality Traits Shift across the Lifespan:
Evidence from a Cross-National Twin Study
Christian Kandler1, Denis Bratko2, Ana Butković2, Tena Vukasović Hlupić2, Joshua M. Tybur3,
Laura Wesseldijk3, Reinout E. de Vries3, Patrick Jern4, & Gary J. Lewis5
OSF link for additional material: https://osf.io/jmz84/
This paper is an uncorrected preprint accepted for publication in
Journal of Personality and Social Psychology
______________________
1 University of Bremen, Department of Psychology, Germany
2 University of Zagreb, Department of Psychology, Croatia
3 Vrije Universiteit Amsterdam, Department of Experimental and Applied Psychology, Netherlands
4 Abo Akademi University, Department of Psychology, Turku, Finland
5 Royal Holloway University of London, United Kingdom
Correspondence concerning this article should be addressed to Christian Kandler (ORCID iD:
https://orcid.org/0000-0002-9175-235X), Department of Psychology, University of Bremen, PO:
330440, 28334 Bremen, Germany. E-mail: ckandler@uni-bremen.de
Funding. The present research used data from the longitudinal study “SPEADY – Study of
Personality Architecture and Dynamics”, which is supported by grants from the German Research
Foundation awarded to Christian Kandler (KA-4088/2-1 and KA-4088/2-2). TwinsUK is funded by the
Wellcome Trust, Medical Research Council, European Union, Chronic Disease Research Foundation
(CDRF), Zoe Global Ltd and the UK National Institute for Health Research (NIHR)-funded BioResource,
Clinical Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS
Foundation Trust in partnership with King’s College London. The present research used data from the
Croatian twin project, which is supported by grants from the University of Zagreb, Croatia, awarded
to Denis Bratko. Data collection for the Genetics of Sexuality and Aggression project was supported
by a European Research Council grant [(ERC) StG-2015 680002-HBIS] awarded to Joshua Tybur and
Grants No. 274521, 319493 and 284385 from the Academy of Finland awarded to Patrick Jern.
2
Sources of personality variance across the lifespan
Abstract
Decades of research have shown that about half of individual differences in
personality traits is heritable. Recent studies have reported that heritability is not fixed, but
instead decreases across the lifespan. However, findings are inconsistent and it is yet unclear
whether these trends are due to a waning importance of heritable tendencies, attributable
to cumulative experiential influences with age, or due to nonlinear patterns suggesting
Gene×Environment interplay. We combined four twin samples (N = 7,026) from Croatia,
Finland, Germany, and the UK, and we examined age trends in genetic and environmental
variance in the six HEXACO personality traits: Honesty-Humility, Emotionality, Extraversion,
Agreeableness, Conscientiousness, and Openness. The cross-national sample ranges in age
from 14 to 90 years, allowing analyses of linear and nonlinear age differences in genetic and
environmental components of trait variance, after controlling for gender and national
differences. The amount of genetic variance in Extraversion, Agreeableness, and Openness
followed a reversed U-shaped pattern across age, showed a declining trend for Honesty
Humility and Conscientiousness, and was stable for Emotionality. For most traits, findings
provided evidence for an increasing relative importance of life experiences contributing to
personality differences across the lifespan. The findings are discussed against the
background of Gene×Environment transactions and interactions.
Keywords: HEXACO personality traits; life experiences; cross-national twin study; lifespan;
heritability
3
Sources of personality variance across the lifespan
How Genetic and Environmental Variance in Personality Traits Shift across the Lifespan:
Evidence from a Cross-National Twin Study
Behavioral genetic research has robustly shown that almost all traits that vary
between humans are heritable. Indeed, a meta-analysis of twin studies on almost 18,000
human traits reported an average heritability of .49 (Polderman et al., 2015). Another meta-
analysis specifically examining the heritability of personality traits on the basis of identical
and non-identical twin data yielded a comparable estimate of .47 (Vukasović & Bratko,
2015), meaning that about 47% of population variance in personality involve genetic
contributions, with the remaining personality differences primarily attributable to individual
life experiences and error of measurement. When taking random and nonrandom error of
measurement into account, the heritability estimates typically exceed .50 (Kandler &
Papendick, 2017; Kandler, Richter, & Zapko-Willmes, 2017; Riemann, Angleitner, & Strelau,
1997).
Substantial heritability, however, does not imply immutability across the life span. A
sizable literature shows that genetic influences wax and wane across the lifespan depending
on the trait in question (Plomin, DeFries, Knopik, & Neiderhiser, 2016) with different
implications for underlying developmental processes (Briley et al., 2019). The existing
evidence for personality traits is equivocal, with some work reporting less support for
shifting genetic and environmental contributions across the lifespan (e.g., Loehlin & Martin,
2001), and more recent studies providing evidence for shifts, though the specific nature of
these shifts is inconsistent across studies (Briley & Tucker-Drob, 2014; Kandler & Papendick,
2017). The current study was designed to clarify this inconsistency by estimating linear and
nonlinear age trends in genetic and environmental contributions to the variance in
personality traits across the lifespan in a large-scale cross-national twin sample.
4
Sources of personality variance across the lifespan
Explanations for Genetic and Environmental Variance
Describing the potential explanations for estimates of genetic and environmental
variance in traits can help us understand why genetic and environmental contributions might
shift over the lifespan. Estimates of genetic variance could reflect individual differences in
people’s molecular genetic makeup. Despite substantial heritability estimates for personality
traits based on quantitative genetic (e.g., twin) designs, large-scale genome-wide
associations studies (GWAS) have struggled to identify any single gene that accounts for
more than 1% of the variance in complex personality traits (de Moor et al., 2012; Lo et al.,
2017). Several explanations for this so-called missing heritability problem have been
discussed (Maher, 2008; Plomin, 2013), such as a small number of rare within-population
genetic variants with large effects, a large number of multiple genetic variants with very
small main effects, or multiple interactions across different gene loci (see Supplement A for
a more in-depth discussion). Analogously, only tiny main effects of specific environmental
factors (e.g., work life, social experiences, life events, etc.) on personality have been
identified (Bleidorn, Hopwood, & Lucas, 2018; Bleidorn et al., 2020; Turkheimer & Waldon,
2000), pointing to a comparable missing “environmentality” problem.
Solving the puzzle of very small main effects of specific genes and certain events on
personality requires recognizing that genetic and environmental influences are intricately
interwoven. As a result, contributions of genetic factors and life experiences—and, thus,
estimates of genetic and environmental variance components—may vary across the lifespan
(Bleidorn, Kandler, & Caspi, 2014; Briley, Livengood, & Derringer, 2018; Kandler & Zapko-
Willmes, 2017). For example, different life stages and individual living conditions provide
different opportunities for heritable individual tendencies to arise. Increasing opportunities
allow more scope for unfolding, whereas narrow boundaries typically attenuate genetic
5
Sources of personality variance across the lifespan
contributions (Briley et al., 2019). Similarly, environmental influences are dependent on
genetic sensitivities (Belsky & Pluess, 2009). Thus, environmental influences vary as a
function of genetic differences and vice versa. Effects of these so-called Gene×Environment
interactions can change across the lifespan with changing sensitivity towards experiences
and changing environmental opportunities.
Age Trends in Genetic and Environmental Variance
Quantitative reviews have provided evidence for shifts in genetic and environmental
contributions to personality differences across the lifespan (Briley & Tucker-Drob, 2014;
Kandler, 2012): genetic contributions appear to decrease relative to environmental ones.
Such patterns could result from the individual opportunities and life experiences that
accumulate with personality maturation across the lifespan. Consequently, environmental
variance is expected to increase, whereas genetic variance remains constant (Hypothesis 1).
As a consequence, heritability estimates (i.e., the relative genetic contribution to the trait
variance) would decline with age.
Environmental factors can reduce the unfolding of genetic differences by constraining
an individual’s opportunities for expression or prompting scripted behaviors or normative
pressure to behave in a specific way within specific social roles (e.g., in family or at work).
Environmental factors can even influence gene expression by switching on and off the
genetic activity without altering the genome, a phenomenon known as environmentally
modified epigenetic regulation (Shah et al., 2014). Such influences can be viewed as
Gene×Environment interaction effects, because they regulate the genetic unfolding and
sensitivity to environmental influences. Cumulative intra-individual epigenetic changes that
arise during the human life course (epigenetic drift) have been found to be primarily driven
by environmental factors not shared by twins reared together (Tan et al., 2016). Hence,
6
Sources of personality variance across the lifespan
increasing epigenetic differences between genetically identical individuals (monozygotic
twins) must be environmental. As a consequence of those Gene×Environment interactions,
genetic contributions to the variance are expected to decrease across the lifespan, whereas
environmental variance is expected to increase (Hypothesis 2).
A meta-analysis of 20 longitudinal behavior genetic studies found evidence for
nonlinear declines in genetic variance relative to increases in environmental variance, with
more pronounced declines during childhood and adolescence (Briley & Tucker-Drob, 2014).
In contrast, a more recent review (Kandler & Papendick, 2017), which was partly based on a
different and newer set of longitudinal and age-group studies, did not find the sharp decline
in younger ages, but rather increases of heritability estimates between childhood and young
adulthood. There are several methodological explanations for the inconsistent results
between these quantitative reviews (see also Supplement B for a more in-depth discussion),
leading to the conclusion that the discrepancies can only be resolved by assessing
personality in different age groups using the same instrument and ensuring measurement
invariance across age (i.e., capturing the same personality constructs in different age
groups).
Two large-scale twin studies (Kandler, Waaktaar, Mõttus, Riemann, & Torgersen,
2019; Mõttus et al., 2019) using different rater perspectives and invariant personality
measures yielded evidence for increasing genetic differences relative to environmental
differences until adolescence. This finding is in line with the idea that young people are
challenged with identity formation and niche picking (McAdams, 2015; Scarr, 1992). They
select or avoid and create or manipulate environments to increase person-environment fit.
Such dynamic Gene×Environment transactions over time would amplify initial genetic
differences in personality traits with increasing age (Scarr & McCartney, 1983). A longitudinal
7
Sources of personality variance across the lifespan
twin family study, however, showed that increasing genetic variance was due to an
accumulation of novel genetic factors that come to play during adolescence rather than an
amplification of initial genetic differences (Kandler, Waaktaar et al., 2019). This could be due
to the activation and deactivation of genetic variants during development. The authors
argued that estimates of novel genetic factors during adolescence may also reflect
interactions between genetic and environmental influences shared by twins, which would be
captured in the estimate of genetic effects, if not directly estimated (see Purcell, 2002; for
mathematical derivations). Shared opportunities to unfold their partly heritable personality
in a very similar way are more probable by adolescent twins reared together.
After moving out of the parental home, twin siblings’ experiences diverge and more
opportunities are likely unshared within twin pairs. As a consequence, interactions between
genetic factors and nonshared environmental influences should accumulate across
adulthood. If not directly estimated, such interaction effects would be captured in the
estimate of nonshared environmental contributions (Purcell, 2002). Taken together, recent
studies (Kandler, Waaktaar et al., 2019; Mõttus et al., 2019) indicated a nonlinear – reversed
U-shaped – age trend of the genetic component with a peak in young adulthood beyond a
constant increase of environmental contributions across the entire lifespan (Hypothesis 3).
In sum, the literature provides three plausible explanations for declines in the
heritability of personality across age (versus null hypothesis: no age-related heritability
decline; see Figure 1). According to the idea that individual life experiences accumulate with
personality maturation across the lifespan (Hypothesis 1), environmental variance should
increase, resulting in increasing trait differences. Increasing environmental pressure against
innate differences and epigenetic drift (Hypothesis 2), however, should additionally come
along with a continuous decline in genetic variance with age, resulting in more balanced trait
8
Sources of personality variance across the lifespan
differences. A reversed U-shaped genetic variance pattern with a peak in young adulthood
could reflect interactions between genetic and environmental influences shared by twins
during childhood and adolescence and interactions between genetic and nonshared
environmental factors in adulthood (Hypothesis 3), resulting in increasing trait variance in
younger ages but balanced trait differences thereafter.
Figure 1. Illustration of four hypotheses on potential age trends of genetic and
environmental variance in personality traits across the lifespan. Hypothesis 0: No
age differences; Hypothesis 1: Increasing environmental component; Hypothesis
2: Declining genetic variance and increasing environmental variance; Hypothesis
3: Increasing environmental variance and reversed U-shaped age trend for the
genetic variance with a peak in young adulthood.
9
Sources of personality variance across the lifespan
Aims of the Present Study
Most previous behavior genetic studies relied on specific personality trait concepts
(see Johnson, Vernon, & Feiler, 2008, for an overview), such as those included in Eysenck’s
three-dimensional Psychoticism-Extraversion-Neuroticism model (Eysenck & Eysenck, 1985),
Tellegen’s hierarchical personality trait model (Tellegen & Waller, 2008), or the Big Five/Five-
Factor Model (B5/FFM; McCrae & John, 1992). Few behavior genetic studies on personality
trait dimensions, so far, were based on the HEXACO framework, which is built upon the very
same psycholexical approach that have yielded the B5/FFM, but which has yielded a
maximum set of six cross-culturally replicable personality trait dimensions rather than five
(Ashton & Lee, 2007; Ashton, Lee, & de Vries, 2014): Honesty-Humility, Emotionality,
Extraversion, Agreeableness, Conscientiousness, and Openness.
Although there is substantial construct-related and structural overlap between the
B5/FFM and the HEXACO framework, the inclusion of a sixth personality dimension Honesty-
Humility leads to different architectures of HEXACO Agreeableness and Emotionality when
compared to their B5/FFM counterparts (i.e., Agreeableness and Emotional Stability versus
Neuroticism; see Ashton & Lee, 2020). In short, HEXACO Agreeableness and Emotionality are
rotated versions of their B5/FFM counterparts, such that HEXACO Agreeableness contains
lack of irritability content associated with B5/FFM Emotional Stability, and HEXACO
Emotionality contains sentimentality content associated with B5/FFM Agreeableness.
Furthermore, some content associated with B5/FFM Agreeableness is shifted to Honesty-
Humility in the HEXACO framework. Currently, we know relatively little about genetic and
environmental contributions to individual differences in HEXACO personality traits, and the
few existing studies did not address age trends in genetic and environmental variance
(Kandler, Richter, & Zapko-Willmes, 2019; Lewis & Bates, 2014; Veselka et al., 2009).
10
Sources of personality variance across the lifespan
In addition, most previous studies were restricted in age range and used different
methods and measurement instruments to capture personality traits in different ages,
leading to inconsistenties of age effects on genetic and environmental components of
personality traits (Briley & Tucker-Drob, 2014; Kandler & Papendick, 2017). In sum, the state
of the literature calls for a systematic, genetically informative study with a large sample size
and age range based on the same well-established measure of personality.
In the current study, we combined data from four twin studies on the HEXACO
framework. We examined which of the potential patterns of age trends in genetic and
environmental variance in personality traits fit the data best (Figure 1). The data stem from
four European countries (Croatia, Finland, Germany, and United Kingdom), encompassing
more than 7,000 twins spanning an age range of 14 to 90 years. These data allowed an
examination of linear and nonlinear age trends in genetic and environmental sources in
personality differences from adolescence to old age.
Method
Participants and Procedure
For the current investigation, which was not preregistered, we combined data sets
from four twin samples collected in four different nations. Table 1 provides an overview of
the sample demographics. The combined sample includes 3,008 monozygotic (MZ) and
4,018 dizygotic (DZ) twins. The sample includes more women than men, with the ratio of
men to women varying across national subsamples and twin zygosity. Further, the samples
vary in their age means, variances, and ranges, with the Croatian sample only encompassing
late adolescents and young adults and the UK sample mainly containing women in middle
and higher age of life. We accounted for these unbalanced distributions (see Initial and
Preparatory Analyses for more details).
11
Sources of personality variance across the lifespan
Table 1. Sample Demographics
Study subsamples
Total
Croatian
Finnish
German
UK
sample
N total
830
2816
1142
2238
7026
n male
306
879
312
149
1646
n female
524
1937
830
2089
5380
n male MZ twin pairs
47
135
50
40
272
n female MZ twin pairs
100
414
171
547
1232
n male DZ twin pairs
45
120
57
31
253
n female DZ twin pairs
101
369
195
494
1159
n opposite-sex DZ pairs
122
370
98
7
597
Average age
22.15
28.44
38.99
64.73
40.97
Age range
19-28
18-45
14-88
26-90
14-90
Average age (male)
22.03
28.43
35.43
66.74
32.05
Age range (male)
19-25
18-45
14-84
35-90
14-90
Average age (female)
22.22
28.46
40.33
64.59
43.70
Age range (female)
19-28
18-45
14-88
26-90
14-90
Note. MZ: monozygotic; DZ: dizygotic.
Ethics statement. All specific study projects were approved by research ethics review
boards. Data collection involved no invasive procedures. Participants were informed that
participation was voluntary and that they were free to terminate their participation at any
time without giving a reason. All participants provided written, informed consent in
12
Sources of personality variance across the lifespan
accordance with the Declaration of Helsinki and EU data protection rules prior to responding
to the surveys.
Croatian sample. The Croatian data stem from a population-based twin sample
across six birth cohorts. The sample was formed in several steps. First, the National Centre
for External Evaluation of Education (NCEEE) was contacted. Following preliminary
agreement, a contract was signed between the principal investigator (Professor Denis
Bratko) and the NCEEE in order to secure the confidentiality of participants’ personal contact
details. Since NCEEE yearly administers exams based on which national standards of
students' academic achievement are evaluated (State Matura Exams), potential twin pairs
were identified based on the algorithm administered to six cohorts – school years 2009/2010
to 2014/2015. This resulted in a dataset of 3026 potential individual twins. At this time,
researchers were only given access to contact information. Potential twin pairs received a
postal letter inviting them to participate in the study, describing study goals, procedures,
and voluntary participation with no compensation. If potential participants did not want to
participate and/or did not allow their NCEEE identification data to be used in the study, they
could contact the researchers by postal mail, SMS, or e-mail and request deletion from the
database. These persons were not contacted again. A set of questionnaires was sent to 2649
individual twins together with an additional empty, stamped and return-addressed
envelope. A total of 836 individual twins provided written consent to participate in the study
and returned the questionnaires. The dataset used in this research is available as
anonymized scientific use file after sending a research proposal to the principal investigator.
The Ethics Review Board of the Department of Psychology, Faculty of Humanities and Social
Sciences, University of Zagreb reviewed and approved the research plan describing the data
collection procedure.
13
Sources of personality variance across the lifespan
Finnish sample. The Finnish data stem from Genetics of Sexuality and Aggression
(GSA) Project. Contact information for Finnish twins and their siblings were drawn from the
Central Population Registry in Finland with more than 33,000 addresses. Only those who live
in Finland were invited by letters in November 2018 to respond to an online survey and
provide agreements for data collection, which concluded in the first week of January 2019.
All participants were offered entry into a raffle that contained 40 gift vouchers to a Finnish
network of companies operating in the retail and service (e.g., restaurants, petrol stations,
hotels) sectors worth €100 each as prizes. In total, 9,564 individuals (6,965 twins and 2,592
siblings, 7 unknown) responded, resulting in a total response rate of 29%, with 9,319 (97%)
respondents giving consent for the use of their data for scientific purposes. The current
analysis includes the 2,816 participants who provided HEXACO data and were members of
complete twin pairs. Further details regarding the sample, recruitment, and zygosity
estimation are provided in Tybur et al. (in press). The Board for Research Ethics of Åbo
Akademi University in Turku, Finland, reviewed and approved the GSA project.
German sample. The German twin sample was collected between January 2016 and
January 2018 as genetically informative data of the Study of Personality Architecture and
Dynamics (SPeADy). Different strategies were used for the recruitment of twins, such as
media calls, attending twin club meetings, and getting contact details from registration
offices or former twin studies. The sample cannot be treated as population-based, but it can
be seen as heterogeneous with respect to age, gender, family status, and educational level
(see Kandler, Penner, Richter, & Zapko-Willmes, 2019, for more details on the recruitment
procedure and sample characteristics). All participating twins provided an informed consent
and contributed data via an online survey or mailed questionnaires. After both twins of a
pair had filled out the forms and questionnaires, they received a personality profile and a
14
Sources of personality variance across the lifespan
€10 voucher for compensation. Contact details and research data were entered into
different databases. Research data are available as anonymized scientific use file on request.
The Ethics Committee of the Bielefeld University and Medical School Hamburg, Germany,
reviewed and approved SPeADy. The twin data were analyzed previously as part of an
extended twin family study, but not with respect to age differences (Kandler, Richter et al.,
2019).
UK sample. The UK’s largest adult twin registry TwinsUK (http://twinsuk.ac.uk/about-
us/what-is-twinsuk/) contributed the fourth subsample of twins with personality data (see
Moayyeri, Hammond, Hart, & Spector, 2013, for more details). About 2,200 twins of almost
14,000 twins provided self-reports on their personality. The sample contains considerably
more women than men. This is due to the fact that women show greater prevalence rates in
the TwinsUK initial phenotypes of interest (e.g., osteoporosis and rheumatic diseases). Thus,
the UK twin sample cannot be seen as population-based with respect to basic demographics.
Access to the twin data is possible as scientific use file after sending a proposal form to the
data access manager (see https://twinsuk.ac.uk/resources-for-researchers/access-our-data/
for more details). TwinsUK has ethical approval from the Guys & St Thomas’ NHS foundation
Trust Ethics Committee. Lewis and Bates (2014) analyzed the twin data against the
background of the nature of the hierarchical structure of the HEXACO traits. Age differences,
however, were not the focus in this previous study.
Statistical Software
All descriptive statistics and preparatory analyses were done in part with the
statistical software packages R 4.0.1 (R Core Team, 2020) and SPSS 26.0 (IBM Corp., 2018).
Structural equation model analyses were run using the statistical software package Mx
(https://mx.vcu.edu/) and OpenMx (Boker et al., 2011) under R 4.0.1. All statistical analytical
15
Sources of personality variance across the lifespan
scripts are available at the open science framework (https://osf.io/jmz84/). Ethical review
does not allow unrestricted open access to the raw data, because the matched twin data
structure does not ensure full anonymity, but it does allow the sharing of variance-
covariance matrices. Thus, we added matrices for adolescent, young adult, middle adult, and
late adult MZ and DZ twins as eight data files to OSF. This allows an approximate replication
of the patterns described below.
Personality Measures
Participants completed Croatian, Finnish, German, or English versions of the 60-item
HEXACO Personality Inventory-Revised (Ashton & Lee, 2009).1 This questionnaire is a non-
commercial personality instrument available in 30 different language versions (see also
http://hexaco.org/hexaco-inventory). It captures the six broad HEXACO personality trait
dimensions with 10 items per dimension. Descriptive statistics are presented in Supplement
C. Given the proposed bandwidth of the HEXACO dimensions capturing a broad spectrum of
personality differences, internal reliabilities (McDonald’s ωt) for the six 10-item HEXACO
composites were acceptable (all ≥ .64) and comparable across national, gender, and four
age-group subsamples (see Table 2).
To estimate congruence of factor loadings between subsamples, we ran principal axis
factor analyses with promax rotation allowing for six factors for each national, gender, and
age group subsample. All factor loadings taken from the structure matrix are shown in Table
X1 (https://osf.io/jmz84/). Out of 78 coefficients (comparing six HEXACO factors across 13
group comparisons), only six were less than .85, and each of these surpassed .80 (see Table
3). Guided by recommendations that Tucker congruence coefficients surpass ϕ ≥ .85
1 For Croatian and Finnish twins, the 60-item version was extracted from the longer 100-item HEXACO version,
from which data are also available.
16
Sources of personality variance across the lifespan
(Lorenzo-Seva & ten Berge, 2006), we interpreted factor loadings as invariant across nations
and demographic categories.
Table 2. Internal Reliability (McDonald’s ωt)
HEXACO personality trait scores
Samples
HH
Em
eX
Ag
Co
Op
Croatian (n=830)
.75
.81
.81
.74
.76
.81
Finnish (n=1142)
.75
.79
.85
.75
.74
.79
German (n=1142)
.72
.77
.80
.73
.76
.71
UK (n=2238)
.64
.73
.76
.74
.70
.75
Female (n=5357)
.71
.73
.81
.74
.72
.77
Male (n=1638)
.76
.70
.83
.72
.76
.76
Age<20 (n=937)
.74
.78
.83
.72
.77
.73
Age:20-30 (n=2226)
.74
.80
.84
.75
.75
.79
Age:30-65 (n=2492)
.71
.74
.82
.74
.70
.77
Age>65 (n=1340)
.66
.72
.70
.72
.68
.75
Total (n=7026)
.73
.77
.82
.74
.73
.76
Note. McDonald’s ωt (the proportion of total common variance of items of a personality
dimension) was estimated using the open source functions available in the psych package for
R (Revelle, 2019); HH: Honesty-Humility; Em: Emotionality; eX: Extraversion; Ag:
Agreeableness, Co: Conscientiousness; Op: Openness.
17
Sources of personality variance across the lifespan
Table 3. Factor Congruence (Tucker’s ϕ)
HEXACO personality trait scores
Sample comparisons
HH
Em
eX
Ag
Co
Op
Female vs. Male
.90
.91
.97
.94
.94
.95
Croatian vs. Finnish
.95
.93
.96
.93
.93
.93
Croatian vs. German
.89
.91
.94
.92
.95
.93
Croatian vs. UK
.83
.89
.90
.89
.90
.91
Finnish vs. German
.91
.96
.95
.90
.91
.94
Finnish vs. UK
.87
.94
.92
.85
.82
.93
German vs. UK
.88
.94
.94
.94
.92
.96
Adolescents vs. Young adults
.95
.98
.98
.96
.97
.96
Adolescents vs. Mid adults
.93
.93
.95
.94
.94
.88
Adolescents vs. Late adults
.80
.90
.88
.92
.86
.88
Young adults vs. Mid adults
.96
.96
.97
.96
.97
.94
Young adults vs. Late adults
.83
.93
.90
.88
.84
.92
Mid adults vs. Late adults
.88
.96
.95
.92
.80
.95
Note. HH: Honesty-Humility; Em: Emotionality; eX: Extraversion; Ag: Agreeableness, Co:
Conscientiousness; Op: Openness; adolescents: age < 20; young adults: age 20-30; middle
adults: age 30-65; late adults: age > 65; Tucker’s ϕ (the cosine of the angle of two vectors of
factor loadings of all items on a personality dimension) was estimated on the basis of two
different sample vectors of factor loadings of all items on a personality dimension (see Table
X1; https://osf.io/jmz84/).
18
Sources of personality variance across the lifespan
Initial and Preparatory Analyses
Age distribution. As the current study aimed to investigate age differences in the
genetic and environmental components in personality differences across the lifespan, age
was the central predictor (or rather moderator) in our analyses. Although age was not
equally distributed (see Figure 2), each age year between 14 and 85 years of age had at least
10 twin pairs. Hence, the age distribution was sufficient to analyze differences in genetic and
environmental sources of personality.
Figure 2. Distribution of the frequency of twin pairs’ age in the entire sample.
Age, gender, and national differences in personality. Because of the unbalanced
distributions of demographic variables, we assessed and controlled for potential
confounding factors for age trends, such as the gender and nationality of participants. For
this purpose, we first computed three dummy variables capturing national differences. The
German group served as reference category, because it encompasses the broadest age
range (see Table 1). The three dummy variables indicated if a participant is Croatian or not (1
= Croatian; 0 = other), Finnish or not (1 = Finnish; 0 = other), and UK or not (1 = UK; 0 =
other). Then, we computed quadratic and cubic age variables as well as an age × gender
19
Sources of personality variance across the lifespan
interaction variable based on z-standardized age and a dichotomous gender variable (0 =
female; 1 = male).
To test for (confounding) effects of gender, nation, and age on personality, we ran
multiple regression analyses with a stepwise procedure for each HEXACO dimension. First,
the three dummy-coded nation variables and gender were included as predictors. Second,
we added age to test for linear age effects in the presence of potential confounding national
and gender differences. Third, nonlinear age and age × gender interactions were tested.
Results of all six stepwise multiple regression analyses are presented and discussed in detail
in Supplement D. In short, gender and age differences in personality were in line with
previous studies (e.g. Ashton & Lee, 2016). National differences in average levels of
personality traits were small and partially attributable to the age differences across the
national samples.
Because age and national differences could inflate variance and twin correlations,
and gender differences could bias the differences between same-sex and opposite-sex twins,
we used a regression procedure to correct each HEXACO score for national and gender
differences, linear and nonlinear age effects, and the age × sex interaction (McGue &
Bouchard, 1984). Unstandardized residual scores derived from these regressions were used
in the following analyses.
Age, gender, and national differences in trait variance. Variance in some of the
HEXACO dimensions varied across gender and nation (see Supplement E for Levene tests for
variance homogeneity). We thus standardized the above-mentioned residuals within each of
the eight gender-by-nation groups. This procedure resulted in an average total variance of 1
across nations and sexes, but the trait variance and the genetic and environmental
20
Sources of personality variance across the lifespan
components could vary across age. As a consequence, potential variance inequality across
age could not arise from unequal age distributions across the sexes and four nations.
Main Analyses
To examine how genetic and environmental variance components – and, potentially,
total trait variance – vary across ages, we first estimated twin correlations for different age
groups. If genetic variance increases in proportion to environmental variance (Hypothesis 3
suggests that this can be expected between adolescence and young adulthood), then either
twin correlations should increase with age, with a stronger increase for genetically identical
MZ twins compared to DZ twins, or the difference between MZ and DZ twin correlations
should increase (see Kandler, Waaktaar et al., 2019). If environmental variance due to
factors not shared by twins increases in relation to genetic variance across adulthood (in line
with Hypotheses 1, 2, and 3), then both MZ and DZ twin correlations should decline with
age.
Next, we ran variance component model analyses allowing for varying components
based on varying MZ and DZ twin covariances across age (Purcell, 2002). This twin model
(see Figure 3) allows a disentanglement of unstandardized variance components due to
additive genetic sources (A), environmental influences shared by twins (C), and
environmental influences not shared by twins plus error of measurement (E). These
unstandardized variance components are free to vary as linear and nonlinear functions of
age (M). The a, c, and e parameters represent main effects, the βA, βC, and βE parameters
reflect linear age effects, and the βA², βC², and βE² parameters constitute quadratic age
effects. Model parameters were estimated for each of the six HEXACO personality traits (see
https://osf.io/jmz84/ for Mx scripts).
21
Sources of personality variance across the lifespan
Figure 3. Illustration of the variance components model based on variance and
covariance in twin pairs’ trait scores. A: additive genetic factor; Var(A) = (a + βA ×
M + βA² × M²)²: additive genetic component as a function of age; C: shared
environmental factor; Var(C) = (c + βC × M + βC² × M²)²: shared environmental
component as a function of age; E: nonshared environmental factor; Var(E) = (e +
βE × M + βE² × M²)²: nonshared environmental component as a function of age; M:
age as continuous moderator; latent factors have unit variance and σ = 1 for MZ
twins and σ = .5 for DZ twins. The Local Structural Equation Modeling estimates
for each level of the moderator from –m to +m is shown in parantheses. These
estimates are based on weighted data rather than data precisely at the level of M.
The twin model relied on several assumptions. First, MZ twins share trait-relevant
environmental factors to the same degree as DZ twins. This equal-environment assumption
has been supported in several studies for a number of traits (e.g., Conley, Rauscher, Dawes,
Magnusson, & Siegal, 2013). Thus, differences between MZ and DZ twin similarities are
attributable to genetic influences, whereas a lack of differences in their similarities can be
22
Sources of personality variance across the lifespan
attributed to shared environmental influences, and within-pair differences in MZ twins are
due to nonshared environmental influences.
Second, nonadditive genetic influences and shared environmental influences cannot
be estimated simultaneously. However, to the extent that nonadditive genetic effects are
present, but unmodelled, they will inflate estimates of variance components due to additive
genetic sources derived from the differences of MZ and DZ twin covariances (Hahn et al.,
2012; Hill, Goddard, & Visscher, 2008). Therefore, estimates of additive genetic components
are good estimators of full genetic variance, including additive and nonadditive genetic
contributions.
Third, the model rests on the assumption that there is no assortative mating of twins’
parents regarding the traits investigated, which, if present, could lead to an underestimate
of genetic components and an overestimate of shared environmental components. Existing
extended twin family analyses suggest non-zero spouse similarity for Honesty-Humility (r =
.25) and Openness (r = .27), but not for the other HEXACO traits (Kandler, Richter, & Zapko-
Willmes, 2019). Potential contributions of assortative mating, however, do not bias age
trends in the genetic component.
Finally, the model assumes the absence of Gene×Environment interactions and
transactions. Thus, it allows estimates of the net contributions of genetic and environmental
sources to trait variance. Components that vary across age can be interpreted against the
background of Gene×Environment interplay across the lifespan (Briley et al., 2019; Kandler,
Zapko-Willmes, Richter, & Riemann, in press).
For the main analyses, age was both condensed to life phases (adolescence: 14-20;
young adulthood: 21-30; middle adulthood: 31-65; and late adulthood: >65) and included as
continuous moderator variable (M). The model fitting procedure works better when
23
Sources of personality variance across the lifespan
moderator values do not exceed 1,000, which would have been the case for age² (when age
> 32 years). Therefore, we recoded age by mean-centering (41 years) and by dividing it by
10. This coding resulted in a mean age value of 0 and a range from -2.7 (original age: 14
years) to 4.9 (original age: 90 years).
In addition to the parametric model fitting tests, we applied a non-parametric
approach – Local Structural Equation Modeling (LOSEM; see Briley, Harden, Bates, & Tucker-
Drob, 2015, and Mõttus et al., 2019, for more details). This approach produces locally
weighted estimates of genetic and environmental components (â², ĉ², and ê²) for each level
of the continuous moderator age (see Figure 3) – here: -2.7 (-m) to 4.9 (+m). LOSEM is an
explorative approach and as such inferior to the confirmatory model fitting tests. However,
it is more flexible and can thus give us a better impression of the exact age-related shift
underlying the linear and nonlinear age trends in genetic and environmental variance
components – if statistically significant. In particular, LOSEM can identify the exact data-
based trend and may thus allow us to avoid, for instance, interpreting a true nonlinear L-
shape as U-shape. Therefore, it is a valuable complement to the model fitting tests (see
https://osf.io/jmz84/ for the R code).
24
Sources of personality variance across the lifespan
Results
Twin Correlations
We first estimated MZ and DZ twin correlations for four age groups (see Table 4; see
also Supplement F for twin correlations separated by gender and national samples). MZ twin
correlations (on average: .46; range: .37 - .62) were consistently higher than DZ twin
correlations (on average: .19; range: .12 - .30) across traits and age groups, indicating
substantial genetic influences to trait variance across age. This is in line with findings from a
meta-analysis estimating average correlations between different family members, with MZ
twins reared either together or apart (r’s = .47 and .45, respectively) more than twice as
similar as DZ twins reared together or apart (r’s = .20 and .16, respectively) (Bratko, Butković,
& Vukasović Hlupić, 2017). These findings additionally show that all HEXACO dimensions
yielded results comparable with other personality models used in behavior genetic studies.
For all HEXACO traits except Openness, correlations were smaller for adolescent (age
< 20) compared to young adult twins (age: 20-30): .40 (.36 - .48) versus .48 (.41 - .57) for MZ
twins and .13 (.00 - .25) versus .18 (.12 - .27) for DZ twins (see Table 4). These patterns were
not consistent with Hypotheses 1 and 2, which implied the reverse pattern, as showed for
Openness. The general trend could be due to either larger genetic variance (in line with
Hypothesis 3), smaller variance due to nonshared environmental contributions in young
adulthood (not in line with any hypothesis), or both. However, the 95% confidence intervals
of twin correlations for adolescents overlapped with those for young adults and none of the
six comparisons between adolescent and young adult MZ twins on HEXACO traits were
statistically significant (Fisher’s z’s ≥ -1.78, p’s ≥ .07).
25
Sources of personality variance across the lifespan
Table 4. Twin Correlations
All ages
Adolescents
Young adults
Middle adults
Late adults
Trait
Stats
MZ
DZ
MZ
DZ
MZ
DZ
MZ
DZ
MZ
DZ
n
1504
2009
136
335
416
705
609
642
343
327
HH
r
.37
.20
.40
.20
.42
.25
.37
.16
.30
.15
p
<.001
<.001
<.001
<.001
<.001
<.001
<.001
<.001
<.001
.007
95% CI
.33-.42
.15-.24
.25-.54
.10-.30
.34-.50
.18-.32
.30-.44
.08-.23
.20-.39
.04-.25
Em
r
.44
.15
.38
.08
.41
.15
.45
.18
.50
.17
p
<.001
<.001
<.001
.165
<.001
<.001
<.001
<.001
<.001
.002
95% CI
.40-.48
.11-.29
.22-.51
-.03-.18
.33-.49
.08-.22
.39-.52
.10-.25
.42-.58
.07-.28
eX
r
.53
.22
.48
.25
.57
.27
.55
.20
.42
.12
p
<.001
<.001
<.001
<.001
<.001
<.001
<.001
<.001
<.001
.026
95% CI
.49-.56
.18-.26
.34-.60
.15-.35
.50-.63
.20-.34
.49-.60
.12-.27
.33-.50
.02-.23
Ag
r
.41
.12
.39
.00
.49
.12
.38
.21
.36
.09
p
<.001
<.001
<.001
.937
<.001
.002
<.001
<.001
<.001
.126
95% CI
.36-.45
.08-.16
.23-.52
-.11-.10
.41-.56
.04-.19
.31-.44
.14-.29
.27-.45
-.02-.19
Co
r
.40
.13
.36
.11
.50
.13
.36
.14
.35
.19
p
<.001
<.001
<.001
.047
<.001
.001
<.001
.001
<.001
.001
95% CI
.35-.44
.09-.18
.20-.50
.00-.21
.43-.57
.05-.20
.29-.42
.06-.21
.25-.44
.08-.29
Op
r
.62
.30
.66
.16
.57
.29
.62
.35
.76
.49
p
<.001
<.001
<.001
.003
<.001
<.001
<.001
<.001
<.001
<.001
95% CI
.58-.65
.26-.33
.55-.75
.06-.27
.50-.63
.22-.35
.56-.66
.28-.42
.71-.80
.41-.57
Note. MZ: Monozygotic; DZ: Dizygotic; HH: Honesty-Humility; Em: Emotionality; eX:
Extraversion; Ag: Agreeableness, Co: Conscientiousness; Op: Openness; significant
correlations (p < .01) are shown in bold; see text for more details.
26
Sources of personality variance across the lifespan
Compared to young adult twins, the correlations again tended to be smaller for
middle and late adult twins (age > 30) with smaller correlations for the late adults (age > 65),
except for Emotionality and Openness. This pattern indicated declining heritability
estimates. The differences in MZ twins’ correlations between young adult twins and twins
older than 65 were significant for Honesty-Humility (Fisher’s z = 1.97, p = .049), Extraversion
(Fisher’s z = 2.82, p = .005), Agreeableness (Fisher’s z = 2.09, p = .037), and
Conscientiousness (Fisher’s z = 2.58, p = .010). For Openness (Fisher’s z = -4.72, p < .001) and
Emotionality (Fisher’s z = -1.59, p = .112), however, the trend pointed to the opposite
direction and suggested increasing heritability estimates across adulthood (contrasting all
expectations). The twin correlations provided a first glimpse of age differences, but they did
not allow specific conclusions regarding the genetic and environmental variance underlying
the varying twin correlations – that is, whether the trend of heritability was due to varying
size of the genetic or the environmental component or both.
Parametric and Nonparametric ACE × Age Variance Component Analyses
To test the three hypotheses, we ran parametric variance component model analyses
allowing for varying ACE components across age. We first ran the full ACE × Age interaction
model (Figure 3): (a + βA × M + βA² × M²)² + (c + βC × M + βC² × M²)² + (e + βE × M + βE² × M²)²
(see Supplement G for full model results). Then, we tested whether five reduced and nested
models fit the data worse based on χ²-difference (p < .05): (1) a model without C effects, (c +
βC × M + βC² × M²)² = 0; (2) a nonlinear AE × Age (nAE) model removing the nonlinear E × Age
interaction following Hypothesis 3, (βE² × M²) = 0; (3) a linear AE × Age model dropping the
nonlinear A × Age interaction following Hypothesis 2, (βA² × M²) = 0; (4) a linear E × Age
model removing the linear A × Age interaction following Hypothesis 1, (βA × M) = 0; and (5)
an AE model (a² + e²) allowing for no age moderation according to the null hypothesis. We
27
Sources of personality variance across the lifespan
also used the sample-size adjusted Bayesian Information Criterion (ABIC) as descriptive
model fit index. The smallest ABIC indicates the model with a good compromise between
model fit and parsimony (Dziak, Coffman, Lanza, Runze, & Jermiin, 2019). A detailed
overview of model fit statistics and comparisons is provided in Table 5.
Table 5. ACE × Age Variance Component Model Fit Statistics and Model Comparisons
vs. full model:
vs. neighbor model:
Trait
Model
-2LogL
df
ABIC
Δχ²
Δdf
p
Δχ²
Δdf
p
HH
Full age moderation
19612.27
7017
-7689.85
No C effects
19614.84
7020
-7696.04
2.57
3
.463
nAE × age moderation
19615.01
7021
-7698.45
2.74
4
.603
0.17
1
.680
AE × age moderation*
19616.06
7022
-7700.42
3.79
5
.581
1.05
1
.306
E × age moderation
19628.78
7023
-7696.55
16.51
6
.011
12.72
1
<.001
No age moderation
19629.00
7024
-7698.94
16.73
7
.019
0.22
1
.639
Em
Full age moderation
19555.42
7017
-7718.27
No C effects
19555.42
7020
-7725.75
0.00
3
>.999
nAE × age moderation
19555.45
7021
-7728.23
0.02
4
>.999
0.02
1
.888
AE × age moderation
19555.47
7022
-7730.71
0.04
5
>.999
0.02
1
.888
E × age moderation*
19556.31
7023
-7732.79
0.89
6
.990
0.85
1
.357
No age moderation
19563.82
7024
-7731.53
8.40
7
.299
7.51
1
.006
Ex
Full age moderation
19308.72
7017
-7841.62
No C effects
19309.42
7020
-7848.75
0.70
3
.874
nAE × age moderation*
19310.73
7021
-7850.59
2.01
4
.733
1.31
1
.252
AE × age moderation
19325.33
7022
-7845.78
16.62
5
.005
14.61
1
<.001
E × age moderation
19350.40
7023
-7835.74
41.68
6
<.001
25.06
1
<.001
No age moderation
19352.71
7024
-7837.08
43.99
7
<.001
2.31
1
.129
28
Sources of personality variance across the lifespan
Continued Table 5
vs. full model:
vs. neighbor model:
Trait
Model
-2LogL
df
ABIC
Δχ²
Δdf
p
Δχ²
Δdf
p
Ag
Full age moderation
19633.29
7017
-7679.34
No C effects
19633.29
7020
-7686.94
0.00
3
>.999
nAE × age moderation
19634.38
7021
-7688.49
1.09
4
.896
1.09
1
.296
AE × age moderation
19638.24
7022
-7689.33
4.95
5
.422
3.86
1
.049
E × age moderation
19638.67
7023
-7691.66
5.38
6
.496
0.43
1
.512
No age moderation*
19640.93
7024
-7692.97
7.64
7
.365
2.26
1
.133
Co
Full age moderation
19631.94
7017
-7680.01
No C effects
19634.29
7020
-7686.31
2.36
3
.502
nAE × age moderation
19634.98
7021
-7688.46
3.05
4
.550
0.69
1
.406
AE × age moderation*
19636.78
7022
-7690.06
4.84
5
.435
1.79
1
.181
E × age moderation
19643.09
7023
-7689.40
11.15
6
.084
6.31
1
.012
No age moderation
19648.19
7024
-7689.34
16.25
7
.023
5.10
1
.024
Op
Full age moderation
19093.92
7017
-7949.03
No C effects
19094.01
7020
-7956.46
0.09
3
.994
nAE × age moderation
19098.89
7021
-7956.51
4.97
4
.291
4.88
1
.027
AE × age moderation
19104.85
7022
-7956.02
10.93
5
.053
5.96
1
.014
E × age moderation
19106.46
7023
-7957.71
12.53
6
.051
1.60
1
.206
No age moderation*
19107.51
7024
-7959.68
13.59
7
.059
1.06
1
.303
Note. HH: Honesty-Humility; Em: Emotionality; eX: Extraversion; Ag: Agreeableness; Co:
Conscientiousness; Op: Openness; best fitting models based on χ²-difference (Δχ²) tests (p < .05) are
shown in bold for each trait; see text for more details; *Model variants with the best balance
between parsimony and fit indicated by the smallest sample-size adjusted Bayesian Information
Criterion (ABIC).
29
Sources of personality variance across the lifespan
In line with previous research, fit for no-C-effects models was equivalent to full
models for all traits, indicating negligible shared environmental influences across the
lifespan. LOSEM also did not provide signal for systematic shared environmental influences
across age for any trait (ĉ² < .10).
For Honesty-Humility and Conscientiousness, linear AE × Age moderation models
(model 3) fit as well as more complex models and better than more parsimonious models
and showed the smallest ABIC. As can be seen in Figure 4, the respective age trends of the
genetic and environmental components are in line with Hypothesis 2. That is, the decline of
heritability for these traits across the lifespan (see h² in Figure 5) was attributable to both
declining genetic variance and increasing environmental variance across age, but due to a
steeper decline in the genetic variance component the total variance tended to decline
across the lifespan (see stacked variance components in Figure 5 and also the supplementary
Figures G1 and G2 for full ACE × Age variance component model results). LOSEM-based
trends also point to decreasing genetic and increasing environmental variance, in particular
between age 40 and 60 (see supplementary Figure H).
For Extraversion and Agreeableness, nonlinear AE × Age models provided a better fit
than models lacking nonlinear genetic parameters, indicating nonlinear age trends for the
genetic variance and linear trends for the environmental variance in both traits. Figure 4
illustrates that these trends were in line with Hypothesis 3: The genetic component tended
to increase until the 30s and sloped down thereafter, whereas the environmental
component linearly increased or remained stable across age. As a consequence, the total
variance in Extraversion and Agreeableness followed a reversed U-shaped pattern as a
function of the varying genetic component (Figure 5). These patterns were primarily
reflected by LOSEM, with the exception that the nonparametric approach indicated a
30
Sources of personality variance across the lifespan
plateau of the genetic variance until age 40 for both traits and a decline thereafter. The
LOSEM-indicated increase of the genetic variance in Agreeableness in old age (> 80) must be
treated with caution because of the low sample size beyond age 80 and thus lower precision
of estimates at the tails of the slopes (Briley et al., 2015).
Figure 4. Unstandardized variance components due to additive genetic (a²) and
nonshared environmental (e²) effects across age based on the best fitting
variance component models.
31
Sources of personality variance across the lifespan
Figure 5. Stacked unstandardized variance components due to additive genetic
(a²), nonshared environmental (e²) effects and resulting heritability estimates
(h²=a²/[a²+e²]) across age with highest and lowest values as well as with a third
value at the peak or valley for the nonlinear trends based on the best fitting
variance component models.
For Emotionality, a linear E × Age moderation model fit as well as more complex
models. The significant negative age trend of the environmental component (Figure 4) – a
pattern completely mirrored by LOSEM (see Figure H) – was not in line with any of the
32
Sources of personality variance across the lifespan
hypotheses, because its decline resulted in decreasing total variance and increasing
heritability estimates across age (Figure 5).
While model comparison tests also revealed a reversed U-shaped pattern of the
genetic variance in Openness, the Δχ²-tests suggested a No-C-effects model as most suitable
to fit the data, indicating nonlinear age effects on both genetic and environmental variance
components. The genetic variance increased until the 50s, reached a plateau, and declined
after age 60 (see Figure 4). This peak in middle adulthood is not in line with Hypothesis 3.
Not in line with any of the hypotheses, the environmental component followed a reversed
U-shaped pattern with increases until the 40s and steeper declines thereafter. This pattern
produced a left-skewed U-shaped trend of heritability estimates for Openness across age
(see h² in Figure 5).
Notably, for Openness and Agreeableness, descriptive model comparisons based on
the ABIC suggested that model variants favored the null hypothesis of no varying genetic and
environmental components. In addition, model comparison tests based on the Δχ² would not
have been significant with a more conservative p-value (p ≤ .01) and LOSEM-based trends
rather suggested small shifts in genetic and environmental variance components in
Openness and Agreeableness (see Figure H), indicating that these patterns should be
interpreted with caution.
To sum up, genetic differences declined at least in the second half of the lifespan for
five traits, whereas no significant age differences were found for Emotionality. That is, for
five out of six HEXACO dimensions, estimates pointed to smaller genetic variance in old age
compared to young and middle adulthood. Variance due to environmental effects shared by
twin siblings reared together (c²) could be treated as negligible, whereas individual
differences due to environmental influences not shared by twins (e²) followed a linear
33
Sources of personality variance across the lifespan
decrease for Emotionality, a decline for Openness in the second half of life, or a linear (or no)
increase for all other trait dimensions. In other words, variance components primarily due to
individualizing environmental influences tended to increase in relative terms across the
entire age range for personality dimensions, except for Openness and Emotionality. That is,
the older the identical and fraternal twins were, the larger the differences in Honesty-
Humility, Extraversion, Agreeableness, and Conscientiousness.
Discussion
This cross-national twin study provides a systematic investigation into age differences
in the genetic and environmental variance components in HEXACO personality traits from
mid-adolescence to old age. Results revealed evidence for declining heritability with age for
most but not all trait dimensions. We examined three competing hypotheses that might
account for declining heritability estimates across the lifespan (Figure 1). Strictly speaking,
only the age trends of genetic and environmental variance in four traits are directly in line
with proposed hypotheses, namely Hypothesis 2 for Honesty-Humility and
Conscientiousness and Hypothesis 3 for Extraversion and Agreeableness. For all four traits,
environmental differences linearly increased or remained constant across adult ages.
Genetic differences for Honesty-Humility and Conscientiousness linearly decrease with age,
whereas they were constant and even tended to increase until the 30s and declined
thereafter for Extraversion and Agreeableness. The age trends in genetic and environmental
differences in Emotionality and Openness are not in line with any of the hypothesized
patterns.
The constant and even increasing genetic differences during adolescence and young
adulthood for Extraversion and Agreeableness, and also for the first half of life for Openness,
are in line with Gene×Environment interplay accounts. With advancing age comes increasing
34
Sources of personality variance across the lifespan
autonomy – particularly during a period of identity formation between adolescence and
adulthood – and the opportunity to actively shape and regulate one’s own development.
People can be attracted to, create or invest in niches and social roles that are consistent with
their heritable traits and allow for self-expression, at least in the ecologies sampled from
here, which provide opportunities for personality to unfold (Kandler, Waaktaar et al., 2019;
Scarr, 1992; Scarr & McCartney, 1983). Moreover, the individual environment may feed back
to influence the development of individual traits. The resulting accentuation of the
observable variance in these traits is in line with the corresponsive principle of personality
development (Caspi, Roberts, & Shiner, 2005, p. 470): “traits that select people into specific
experiences are the traits that are most influenced in response to those experiences.”
Although a direct examination of the Gene×Environment interplay underlying the
increasing genetic and total variance in these traits is not possible, findings are in line with
indirect evidence from longitudinal studies. For example, Denissen and colleagues reported
transactional effects between job environment and Extraversion as well as Openness in a 5-
year longitudinal study of job beginners, job stayers, and job changers (Denissen, Ulferts,
Lüdtke, Muck, & Gerstorf, 2014). More extraverted individuals are more likely to select into
jobs that involve social interaction and extraverted behavior, and employment in such jobs
further increases Extraversion. In fact, most evidence for corresponsive Person×Environment
transactions stems from investigations of young and middle-aged adults (e.g., Denissen,
Luhmann, Chung, & Bleidorn, 2019; Lüdtke, Roberts, Trautwein, & Nagy, 2011; Roberts,
Caspi, & Moffitt, 2003; Roberts & Robins, 2004; Roberts, Walton, Bogg, & Caspi, 2006;
Zimmermann & Neyer, 2013). However, Gene×Environment transactions cannot account for
a decline in genetic differences during the adult years. Thus, other mechanisms must also
come into play.
35
Sources of personality variance across the lifespan
Declines in genetic differences for Honesty-Humility and Conscientiousness that are
already observable in young ages are in line with the idea that social environments already
and increasingly demand trustworthiness and diligence in emerging adults. Resulting
investments in social roles may reinforce corresponding or reduce conflicting trait levels
(Bleidorn et al., 2013). For example, initiating a career or job for the first time was found to
be linked to increases in Conscientiousness. People do not only seek to reach their own
inherent goals or express their preexisting tendencies but also follow social values and
standards that set the direction of personality development (Denissen, van Aken, Penke, &
Wood, 2013; Denissen et al., 2019). These social norms and values demand uniformity and
may thus limit the unfolding of genetic differences in relevant traits.
Beyond the mere “social limits and pressure” explanation of diminishing genetic
variance across the lifespan, increasing environmental variance in the presence of
decreasing genetic variance can reflect epigenetic drift with age. Environmental influences
can also act under the skin, affecting hormonal regulation, neurotransmitter release, or even
gene expression (e.g., via DNA methylation or histone modification). Epigenetic changes due
to environmental influences neither reflect genetic nor environmental influences per se.
They can be understood as specific kind of Gene×Environment interaction because they alter
the genetic sensitivity to environmental influences, such as the genetic sensitivity to stress
(Belsky & Pluess, 2009; Kandler & Zapko-Willmes, 2017). As increasing epigenetic differences
between genetically identical MZ twins have been found to be primarily driven by
individually unique (nonshared) environmental factors (Tan et al., 2016), the systematic shift
from genetic differences to environmental differences may reflect an epigenetic drift across
the lifespan.
36
Sources of personality variance across the lifespan
Irrespective of whether unqiue environmental factors reflect the individual social
opportunities of personality unfolding or represent epigenetic changes, our findings are
largely consistent with the idea that individual life experiences accumulate with personality
maturation from adolescence to mid-adult ages and enrich personality differences. The older
people get, the more life experiences mount in a highly idiosyncratic way (Bleidorn et al.,
2014). But this does not seem to be the case for all traits – at least not in our study.
Why did we find a declining trend in environmental differences in Emotionality and a
reversed U-shaped age trend for Openness? The age trend of the environmental differences
in Openness suggests that the mid-adult years involve the highest probability of experiencing
individualizing influences that act to increase Openness variance. In contrast to the other
personality traits, Openness is less about behavioral style and emotion regulation in the
context of social behavior and more about cognition and sensitivity to internal and external
sensory stimulation (McAdams, 2015). Young adults may be more attracted to and select
into environments that allow them to gain new experiences, such as cultural activity, as was
reported in a recent study (Schwaba, Luhmann, Denissen, Chung, & Bleidorn, 2018). The
observed increase in genetic variance during the first half of life may reflect active
Gene×Environment transactions underlying the openness-culture transactions, whereas the
parallel increase of individual environmental variance underlying the accentuation of the
total variance in Openness may be an accumulation of nonshared chances by twin siblings
with increasing age. The attenuation of individual differences in Openness in the second half
of life may come along with a general tendency to become less cognitively flexible with age
and thus more conservative, which could increase uniformity.
Why environmental differences in Emotionality decline across age is harder to
explain. De Vries and colleagues suggest that Emotionality might partially reflect vigilance
37
Sources of personality variance across the lifespan
against threats to kin (de Vries, Tybur, Pollet, & van Vugt, 2016). If the age and vulnerability
of offspring constitute part of the individualizing environment, then we might expect
decreased contributions after reproductive years and with offsprings age. However, this
account is speculative and should be treated with caution. This finding can also be
considered in the light of some important limitations that provides alternative accounts.
Limitations and Future Directions
Range restriction might have contributed to the decline in environmental
components in old age. Even though we corrected for national and gender differences, the
size-reduced, partly clinical, and primarily female sample of older adults might come along
with a range restriction for individual experiences contributing to trait differences. This
might be particularly true for Emotionality, which is associated with several clinical
symptoms, and could account for the decline of environmental variance in this trait with
age.
In a similar vein, the reduced age range of the adolescent sample with the smallest
age of 14 years along with a lower sample size could have led to difficulties to detect a
reversed U-shaped age trend for the size of the genetic variance (Hypothesis 3). Previous
studies on the ages between childhood and emerging adulthood indicated steeper
increasing genetic components at least until mid-adolescence (Kandler, Waaktaar et al.,
2019; Mõttus et al., 2019). The MZ twin correlations in our study that tended to be lower for
adolescent twins compared to young adult twins for all traits but Openness pointed to this
potential pattern, which could have appeared with more statistical power for this life stage.
Although the differences of relability estimates based on omegas across age groups
were small (see Table 2), they tended to be lower for late adulthood (on average .71 for age
> 65 versus for .76 for age ≤ 65). Despite twin correlations corrected for attenuation due to
38
Sources of personality variance across the lifespan
error of measurement (r/ω) yielded trends across age groups comparable to those of the
uncorrected twin correlations (see supplementary Table I), we cannot completely rule out
that the decline of the variance in old age might be triggered to some degree by lower
internal consistency of the personality trait measures for this age group. The lower internal
consistency in older age might be due to stronger differentiations of personality in old age.
In other words, personality facets could be less strongly related to each other in old age, and
thus the internal consistency of domain-level scales would be lower. Future projects with
more gender-balanced samples in old age could take this consideration into account.
Beyond the aforementioned limitations, the cross-sectional design of our cross-
cultural twin study raises the possibility that some of the observed age trends might reflect
birth-cohort and national effects. However, the observed age trends of declining heritability
for most traits are consistent with meta-analytic findings based on longitudinal designs
(Briley & Tucker-Drob, 2014; Kandler & Papendick, 2017) that argues against potential
cohort effects. Similarly, the age trends of average levels of personality traits in our study are
largely consistent with a Canadian study on about 100,000 individuals (Ashton & Lee, 2016),
arguing against potential national effects. Nevertheless, because of partly confounding
gender, national, and age differences, the correction procedure to adjust for national and
sex differences in variance may have resulted in an overcorrection and could have obscured
further age trends in the amount of variance and its genetic and environmental components.
Thus, future studies with population-based twin data with respect to age and other
important demographics are sorely needed.
Conclusions
Our results suggest that varying heritability estimates for personality traits across the
lifespan are due to different sources that partly differ for different traits and points to
39
Sources of personality variance across the lifespan
unique patterns for each personality trait. For most traits, however, genetic variance tended
to decline across age, whereas the reverse was found for the environmental differences. This
pattern provides evidence for an increasing relative importance of life experiences
contributing to personality differences across the lifespan. The specific trait patterns of
varying genetic and environmental contributions to trait variance suggest different kinds of
Gene×Environment interplay that are differently important for different traits and in
different ages. Or as McAdams put it (2015, p. 111): “The relationship between genes and
environments, therefore, is not much like a meeting of two independent forces (nature vs.
nurture) but instead resembles something more like conspiracy. […] genes and environments
conspire to make a person, and to shape the traits that structure how that person moves
through life…”
40
Sources of personality variance across the lifespan
References
Ashton, M. C., & Lee, K. (2007). Empirical, theoretical, and practical advantages of the
HEXACO model of personality structure. Personality and Social Psychology Review,
11, 150–166. https://doi.org/10.1177/1088868306294907
Ashton, M. C., & Lee, K. (2009). The HEXACO–60: A short measure of the major dimensions
of personality. Journal of Personality Assessment, 91, 340–345.
https://doi.org/10.1080/00223890902935878
Ashton, M. C., & Lee, K. (2016). Age trends in HEXACO-PI-R self-reports. Journal of Research
in Personality, 64, 102–111. https://doi.org/10.1016/j.jrp.2016.08.008
Ashton, M. C., & Lee, K. (2020). Objections to the HEXACO model of personality structure—
and why those objections fail. European Journal of Personality,
https://doi.org/10.1002/per.2242
Ashton, M. C., Lee, K., & de Vries, R. E. (2014). The HEXACO Honesty-Humility,
Agreeableness, and Emotionality Factors: A review of research and
theory. Personality and Social Psychology Review, 18, 139–152.
https://doi.org/10.1177/1088868314523838
Belsky, J., & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to
environmental influences. Psychological Bulletin, 135, 885–908.
https://doi.org/10.1037/a0017376
Bleidorn, W., Hopwood, C. J., Back, M. D., Denissen, J. J. A., Hennecke, M., Jokela, M.,
Kandler, C., Lucas, R. E., Luhmann, M., Orth, U., Roberts, B. W., Wagner, J., Wrzus, C.,
& Zimmermann, J. (2020). Longitudinal Experience-Wide Association Studies (LEWAS)
– A framework for studying personality change. European Journal of Personality, 34,
285–300. https://doi.org/10.1002/per.2247
41
Sources of personality variance across the lifespan
Bleidorn, W., Hopwood, C. J., Lucas, R. E. (2018). Life Events and Personality Change. Journal
of Personality, 86, 83–96. https://doi.org/10.1111/jopy.12286
Bleidorn, W., Kandler, C., & Caspi, A. (2014). The behavioral genetics of personality
development in adulthood: Classic, contemporary, and future trends. European
Journal of Personality, 28, 244–255. https://doi.org/10.1002/per.1957
Bleidorn, W., Klimstra, T. A., Denissen, J. J. A., Rentfrow, P. J., Potter, J., & Gosling, S. D.
(2013). Personality maturation around the world: A cross-cultural examination of
social-investment theory. Psychological Science, 24, 2530–2540.
https://doi.org/10.1177/0956797613498396
Boker, S., Neale, M., Maes, H., Wilde, M., Spiegel, M., Brick, T., … Fox, J. (2011). OpenMx: An
open source extended structural equation modeling framework. Psychometrika, 76,
306–317. https://doi.org/10.1007/S11336-010-9200-6
Bratko, D., & Butkovic, A. (2007). Stability and genetic and environmental effects from
adolescence to young adulthood: Results of Croatian longitudinal twin study of
personality. Twin Research and Human Genetics, 10, 151–157.
https://doi.org/10.1375/twin.10.1.151
Bratko, D., Butković, A., & Vukasović Hlupić, T. (2017). Heritability of personality.
Psihologijske teme, 26, 1–24. https://doi.org/10.31820/pt.26.1.1
Briley, D. A., Harden, K. P., Bates, T. C., & Tucker-Drob, E. M. (2015). non-parametric
estimates of Gene × Environment interaction using local structural equation
modeling. Behavior Genetics, 45, 581–596. https://doi.org/10.1007/s10519-015-
9732-8
Briley, D. A., Livengood, J., & Derringer, J. (2018). Behavior genetic frameworks of causal
reasoning for personality psychology. European Journal of Personality, 32, 202–220.
https://doi.org/10.1002/per.2153
42
Sources of personality variance across the lifespan
Briley, D. A., Livengood, J., Derringer, J., Tucker-Drob, E. M., Fraley, R. C., & Roberts, B. W.
(2019). Interpreting behavior genetic models: Seven developmental processes to
understand. Behavior Genetics, 49, 196–210. https://doi.org/10.1007/s10519-018-
9939-6
Briley, D.A., & Tucker-Drob, E.M. (2014). Genetic and environmental continuity in personality
development: A meta-analysis. Psychological Bulletin, 140, 1303–1331.
https://doi.org/10.1037/a0037091
Caspi, A., Roberts, B.W., & Shiner, R. L. (2005). Personality development: Stability and
change. Annual Review of Psychology, 56, 453–484.
https://doi.org/10.1146/annurev.psych.55.090902.141913
Conley, D., Rauscher, E., Dawes, C., Magnusson, P. K. E., & Siegal, M. L. (2013). Heritability
and the equal environments assumption: Evidence from multiple samples of
misclassified twins. Behavior Genetics, 43, 415–426. https://doi.org/10.1007/s10519-
013-9602-1
De Fruyt, F., Bartels, M., Van Leeuwen, K. G., De Clercq, B., Decuyper, M., & Mervielde, I.
(2006). Five types of personality continuity in childhood and adolescence. Journal of
Personality and Social Psychology, 91, 538–552. https://doi.org/10.1037/0022-
3514.91.3.538
de Moor, M. H. M., Costa, P. T., Terracciano, A., Krueger, R. F., de Geus, E. J. C., . . .
Boomsma, D. I. (2012). Meta-analysis of genome-wide association studies for
personality. Molecular Psychiatry, 17, 337–349.
https://doi.org/10.1038/mp.2010.128
43
Sources of personality variance across the lifespan
de Vries, R. E., Tybur, J. M., Pollet, T. V., & van Vugt, M. (2016). Evolution, situational
affordances, and the HEXACO model of personality. Evolution and Human Behavior,
37, 407–421. http://dx.doi.org/10.1016/j.evolhumbehav.2016.04.001
Denissen, J. J., Luhmann, M., Chung, J. M., & Bleidorn, W. (2019). Transactions between life
events and personality traits across the adult lifespan. Journal of Personality and
Social Psychology, 116, 612–633. http://dx.doi.org/10.1037/pspp0000196
Denissen, J. J. A., Ulferts, H., Lüdtke, O., Muck, P. M., & Gerstorf, D. (2014). Longitudinal
transactions between personality and occupational roles: A large and heterogeneous
study of job beginners, stayers, and changers. Developmental Psychology, 50, 1931–
1942. https://doi.org/10.1037/a0036994
Denissen, J. J. A., Van Aken, M. A. G., Penke, L., & Wood, D. (2013). Self-regulation underlies
temperament and personality: An integrative developmental framework. Child
Development Perspectives, 7, 255–260. https://doi.org/10.1111/cdep.12050
Dziak, J. J., Coffman, D. L., Lanza, S. T., Runze, L., & Jermiin, L. S. (2019). Sensitivity and
specificity of information criteria. Briefings in Bioinformatics, bbz016,
https://doi.org/10.1093/bib/bbz016
Eysenck, H. J., & Eysenck, M. W. (1985). Personality and individual differences: A natural
science approach. New York, NY: Plenum Press.
Ganiban, J. M., Saudino, K. J., Ulbricht, J., Neiderhiser, J. M., & Reiss, D. (2008). Stability and
change in temperament during adolescence. Journal of Personality and Social
Psychology, 95, 222–236. https://doi.org/10.1037/0022-3514.95.1.222
Hahn, E., Spinath, F. M., Siedler, T., Wagner, G. G., Schupp, J., & Kandler, C. (2012). The
complexity of personality: Advantages of a genetically sensitive multi-group design.
Behavior Genetics, 42, 221-233. http://dx.doi.org/10.1007/s10519-011-9493-y
44
Sources of personality variance across the lifespan
Hill, W. D., Arslan, R. C., Xia, C., Luciano, M., Amador, C., . . . Penke, L. (2018). Genomic
analysis of family data reveals additional genetic effects on intelligence and
personality. Molecular Psychiatry, 23, 2347–2362. https://doi.org/10.1038/s41380-
017-0005-1
Hill, W. G., Goddard, M. E., & Visscher, P. M. (2008). Data and theory point to mainly additive
genetic variance for complex traits. PLOS Genetics, 4, 1–10.
http://dx.doi.org/10.1371/journal.pgen.1000008
Hudziak, J. J., van Beijsterveldt, C. E. M., Bartels, M., Rietveld, M. J. H., Rettew, D. C., Derks,
E. M., & Boomsma, D. I. (2003). Individual differences in aggression: Genetic analyses
by age, gender, and informant in 3-, 7-, and 10-year-old Dutch twins. Behavior
Genetics, 33, 575–589. https://doi.org/10.1023/A:1025782918793
IBM Corp. (2018). IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY: IBM Corp.
Johnson, A. M., Vernon, P. A., & Feiler, A. R. (2008). Behavioral genetic studies of personality:
An introduction and review of the results of 50+ years of research. In G. J. Boyle, G.
Matthews, & D. H. Saklofske (Eds.), The sage handbook of personality theory and
assessment personality theories and models, Vol. 1, pp. 145–173. London: Sage,
https://doi.org/10.4135/9781849200462.n7
Kandler, C. (2012). Nature and nurture in personality development: The case of neuroticism
and extraversion. Current Directions in Psychological Science, 21, 290–296.
http://dx.doi.org/10.1177/0963721412452557
Kandler, C., Kornadt, A. E., Hagemeyer, B., & Neyer, F. J. (2015). Patterns and sources of
personality development in old age. Journal of Personality and Social Psychology,
109, 175–191. http://dx.doi.org/10.1037/pspp0000028
Kandler, C., & Papendick, M. (2017). Behavior genetics and personality development: A
methodological and meta-analytic review. In: J. Specht (ed.), Personality
45
Sources of personality variance across the lifespan
Development across the Lifespan (pp. 473–495). Elsevier Academic Press.
http://dx.doi.org/10.1016/B978-0-12-804674-6.00029-6
Kandler, C., Penner, A., Richter, J., & Zapko-Willmes, A. (2019). The Study of Personality
Architecture and Dynamics (SPeADy): A longitudinal and extended twin family study.
Twin Research and Human Genetics, 22, 548-553.
https://doi.org/10.1017/thg.2019.62
Kandler, C., Richter, J., & Zapko-Willmes, A. (2017). Genetic basis of traits. In: V. Zeigler-Hill &
T. K. Shackelford (eds.), Encyclopedia of Personality and Individual Differences.
Springer International Publishing. http://dx.doi.org/10.1007/978-3-319-28099-
8_1473-1
Kandler, C., Richter, J., & Zapko-Willmes, A. (2019). The nature and nurture of HEXACO
personality trait differences: An extended twin family study. Zeitschrift für
Psychologie, 227, 195–206. http://dx.doi.org/10.1027/2151-2604/a000378
Kandler, C., Waaktaar, T., Mõttus, R., Riemann, R., & Torgersen, S. (2019). Unravelling the
interplay between genetic and environmental contributions in the unfolding of
personality differences from early adolescence to young adulthood. European Journal
of Personality, 33, 221–244. https://doi.org/10.1002/per.2189
Kandler, C., & Zapko-Willmes, A. (2017). Theoretical perspectives on the interplay of nature
and nurture in personality development. In: J. Specht (ed.), Personality Development
across the Lifespan (pp. 101–115). Elsevier Academic Press.
https://doi.org/10.1016/B978-0-12-804674-6.00008-9
Kandler, C., Zapko-Willmes, A., Richter, J., & Riemann, R. (in press). Synergistic and dynamic
genotype-environment interplays in the development of personality differences. In: J.
46
Sources of personality variance across the lifespan
F. Rauthmann (ed.), The Handbook of Personality Dynamics and Processes. Elsevier
Academic Press.
Kornadt, A. E., Hagemeyer, B., Neyer, F. J., & Kandler, C. (2018). Sound body, sound mind?
The interrelation between health change and personality change in old age. European
Journal of Personality, 32, 30–45. https://doi.org/10.1002/per.2135
Lewis, G. J., & Bates, T. C. (2014). How genes influence personality: Evidence from multi-
facet twin analyses of the HEXACO dimensions. Journal of Research in Personality, 51,
9–17. https://doi.org/10.1016/j.jrp.2014.04.004
Lo, M.-T., Hinds, D. A., Tung, J. Y., Franz, C., Fan, C.-C., Wang, Y., . . . Chen, C.-H. (2017).
Genome-wide analyses for personality traits identify six genomic loci and show
correlations with psychiatric disorders. Nature Genetics, 49, 152–156.
https://doi.org/10.1038/ng.3736
Loehlin, J. C., & Martin, N. G. (2001). Age changes in personality traits and their heritabilities
during the adult years: Evidence from Australian Twin Registry samples. Personality
and Individual Differences, 30, 1147–1160. https://doi.org/10.1016/S0191-
8869(00)00099-4
Lorenzo-Seva, U., & ten Berge, J. M. F. (2006). Tucker’s congruence coefficient as a
meaningful index of factor similarity. Methodology, 2, 57-64.
https://doi.org/10.1027/1614-1881.2.2.57
Lüdtke, O., Roberts, B. W., Trautwein, U., & Nagy, G. (2011). A random walk down university
avenue: Life paths, life events, and personality trait change at the transition to
university life. Journal of Personality and Social Psychology, 101, 620–637.
https://doi.org/10.1037/a0023743
47
Sources of personality variance across the lifespan
Lykken, D. T. (2006). The mechanism of emergenesis. Genes, Brain, and Behavior, 5, 306–
310. https://doi.org/10.1111/j.1601-183X.2006.00233.x
Maher, B. (2008). Personal genomes: The case of the missing heritability. Nature, 456, 18–
21. https://doi.org/10.1038/456018a
McAdams, D. P. (2015). The art and science of personality development. New York: Guilford.
McGue, M., & Bouchard, T. J. Jr. (1984). Adjustment of twin data for the effects of age and
sex. Behavior Genetics, 14, 325–343. https://doi.org/10.1007/BF01080045
Moayyeri, A., Hammond, C. J., Hart, D. J., & Spector, T. D. (2013). The UK adult twin registry
(TwinsUK Resource). Twin Research and Human Genetics, 16, 144–149.
https://doi.org/10.1017/thg.2012.89
Mõttus, R., Briley, D., Zheng, A., Mann, F. D., Engelhardt, L. E., Tackett, J. L., … Tucker-Drob,
E. M. (2019). Kids becoming less alike: A behavioral genetic analysis of developmental
increases in personality variance from childhood to adolescence. Journal of
Personality and Social Psychology, 117, 635–658.
https://doi.org/10.1037/pspp0000194
Mõttus, R., Johnson, W., & Deary, I. J. (2012). Personality traits in old age: Measurement and
rank-order stability and some mean-level change. Psychology and Aging, 27, 243–
249. https://doi.org/10.1037/a0023690
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its
applications. Journal of Personality, 60, 175–215. https://doi.org/10.1111/j.1467-
6494.1992.tb00970.x
Plomin, R. (2013). Child development and molecular genetics: 14 years later. Child
Development, 84, 104–120. https://doi.org/10.1111/j.1467-8624.2012.01757.x
48
Sources of personality variance across the lifespan
Plomin, R., DeFries, J. C., Knopik, V. S., & Neiderhiser, J. M. (2016). Top 10 replicated findings
from behavioral genetics. Perspectives on Psychological Science, 11, 3–23.
https://doi.org/10.1177/1745691615617439
Polderman, T. J. C., Benyamin, B., de Leeuw, C. A., Sullivan, P. F., van Bochoven, A., . . .
Posthuma, D. (2015). Meta-analysis of the heritability of human traits based on fifty
years of twin studies. Nature Genetics, 47, 702–709. https://doi.org/10.1038/ng.3285
Purcell, S. (2002). Variance components models for gene–environment interaction in twin
analysis. Twin Research and Human Genetics, 5, 554–571.
https://doi.org/10.1375/136905202762342026
R Core Team (2020). R: A language and environment for statistical computing. R Foundation
for Statistical Computing. https://www.R-project.org/
Revelle, W. (2019). Package ‘psych’: Procedures for Psychological, Psychometric, and
Personality Research. (R package version 1.1.12)
Riemann, R., Angleitner, A., & Strelau, J. (1997). Genetic and environmental influences on
personality: A study of twins reared together using the self- and peer report NEO-FFI
scales. Journal of Personality, 65, 449–475. https://doi.org/10.1111/j.1467-
6494.1997.tb00324.x
Roberts, B. W., Caspi, A., & Moffitt, T. E. (2003). Work experiences and personality
development in young adulthood. Journal of Personality and Social Psychology, 84,
582–593. https://doi.org/10.1037/0022-3514.84.3.582
Roberts, B. W., & Robins, R. W. (2004). Person-environment fit and its implications for
personality development: A longitudinal study. Journal of Personality, 72, 89–110.
https://doi.org/10.1111/j.0022-3506.2004.00257.x
49
Sources of personality variance across the lifespan
Roberts, B. W., Walton, K. E., Bogg, T., & Caspi, A. (2006). De-investment in work and non-
normative personality trait change in young adulthood. European Journal of
Personality, 20, 461–474. https://doi.org/10.1002/per.607
Roberts, B. W., Walton, K. E., & Viechtbauer, W. (2006). Patterns of mean-level change in
personality traits across the life course: A meta-analysis of longitudinal studies.
Psychological Bulletin, 132, 1–25. https://doi.org/10.1037/0033-2909.132.1.1
Scarr, S. (1992). Developmental theories for the 1990s: Development and individual
differences. Child Development, 63, 1–19. https://doi.org/10.1111/j.1467-
8624.1992.tb03591.x
Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of
genotype → environment effects. Child Development, 54, 424–435.
https://doi.org/10.2307/1129703
Schaie, K. W. (1965). A general model for the study of developmental change. Psychological
Bulletin, 64, 92–107. https://doi.org/10.1037/h0022371
Schwaba, T., Luhmann, M., Denissen, J. J. A., Chung, J. M., & Bleidorn, W. (2018). Openness
to experience and culture-openness transactions across the lifespan. Journal of
Personality and Social Pschology, 115, 118–136.
https://doi.org/10.1037/pspp0000150
Shah, S., McRae, A. F., Marioni, R. E., Harris, S. E., Gibson, J., Henders, A. K., … Visscher, P. M.
(2014). Genetic and environmental exposures constrain epigenetic drift over the
human life course. Genome Research, 24, 1725–1733.
https://doi.org/10.1101/gr.176933.114
50
Sources of personality variance across the lifespan
Spengler, M., Gottschling, J., & Spinath, F. M. (2012). Personality in childhood: A longitudinal
behavior genetic approach. Personality and Individual Differences, 53, 411–416.
https://doi.org/10.1016/j.paid.2012.01.019
Tellegen, A., & Waller, N. G. (2008). Exploring personality through test construction:
Development of the Multidimensional Personality Questionnaire. In G. J. Boyle, G.
Matthews, & D. H. Saklofske (Eds.), The Sage handbook of personality theory and
assessment (Personality measuring and testing, Vol. 2, pp. 262–292). London: Sage.
Tan, Q., Heijmans, B. T., Hjelmborg, J. v. B., Soerensen, M., Christensen, K., & Christiansen, L.
(2016). Epigenetic drift in the aging genome: a ten-year follow-up in an elderly twin
cohort. International Journal of Epidemiology, 45, 1146–1158.
https://doi.org/10.1093/ije/dyw132
Turkheimer, E., & Waldron, M. (2000). Nonshared environment: a theoretical,
methodological, and quantitative review. Psychological Bulletin, 126, 78–108.
https://doi.org/10.1037/0033-2909.126.1.78
Tybur, J. M., Wesseldijk, L. W., & Jern, P. (in press). Genetic and environmental influences on
disgust proneness, contamination sensitivity, and their covariance. Clinical
Psychological Science
Veselka, L., Schermer, J., Petrides, K., Cherkas, L., Spector, T., & Vernon, P. (2009). A General
Factor of Personality: Evidence from the HEXACO Model and a Measure of Trait
Emotional Intelligence. Twin Research and Human Genetics, 12, 420–424.
https://doi.org/10.1375/twin.12.5.420
Viken, R. J., Rose, R. J., Kaprio, J., & Koskenvuo, M. (1994). A developmental genetic analysis
of adult personality: Extraversion and neuroticism from 18 to 59 years of age. Journal
51
Sources of personality variance across the lifespan
of Personality and Social Psychology, 66, 722–730. https://doi.org/10.1037/0022-
3514.66.4.722
Vukasović, T., & Bratko, D. (2015). Heritability of personality: A meta-analysis of behavior
genetic studies. Psychological Bulletin, 141, 769–785.
https://doi.org/10.1037/bul0000017
Ziegler, C., Richter, J., Mahr, M., Gajewska, A., Schiele, M. A., … Domschke, K. (2016). MAOA
gene hypomethylation in panic disorder—reversibility of an epigenetic risk pattern by
psychotherapy. Translational Psychiatry, 6, e773. https://doi.org/10.1038/tp.2016.41
Zimmermann, J., & Neyer, F. J. (2013). Do we become a different person when hitting the
road? Personality development of sojourners. Journal of Personality and Social
Psychology, 105, 515–530. https://doi.org/10.1037/a0033019
52
Sources of personality variance across the lifespan
Supplement
To the paper entitled:
How Genetic and Environmental Variance in Personality Traits Shift across the Lifespan:
Evidence from a Cross-National Twin Study
Christian Kandler1, Denis Bratko2, Ana Butković2, Tena Vukasović Hlupić2, Joshua M. Tybur3,
Laura Wesseldijk3, Reinout E. de Vries3, Patrick Jern4, & Gary J. Lewis5
______________________
1 University of Bremen, Department of Psychology, Germany
2 University of Zagreb, Department of Psychology, Croatia
3 Vrije Universiteit Amsterdam, Department of Experimental and Applied Psychology, Netherlands
4 Abo Akademi University, Department of Psychology, Turku, Finland
5 Royal Holloway University of London, United Kingdom
Correspondence concerning this article should be addressed to Christian Kandler (ORCID iD:
https://orcid.org/0000-0002-9175-235X), Department of Psychology, University of Bremen, PO:
330440, 28334 Bremen, Germany. E-mail: ckandler@uni-bremen.de
Funding. The present research used data from the longitudinal study “SPEADY – Study of Personality
Architecture and Dynamics”, which is supported by grants from the German Research Foundation
awarded to Christian Kandler (KA-4088/2-1 and KA-4088/2-2). TwinsUK is funded by the Wellcome
Trust, Medical Research Council, European Union, Chronic Disease Research Foundation (CDRF), Zoe
Global Ltd and the UK National Institute for Health Research (NIHR)-funded BioResource, Clinical
Research Facility and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust
in partnership with King’s College London. The present research used data from the Croatian twin
project, which is supported by grants from the University of Zagreb, Croatia, awarded to Denis Bratko.
Data collection for the Genetics of Sexuality and Aggression project was supported by a European
Research Council grant [(ERC) StG-2015 680002-HBIS] awarded to Joshua Tybur and Grants No.
274521, 319493 and 284385 from the Academy of Finland awarded to Patrick Jern.
53
Sources of personality variance across the lifespan
Supplement A: Molecular Genetic Explanations for the Missing Heritability Problem
The missing heritability problem (Maher, 2008) describes the discrepancy between
quantitative genetic and molecular genetic estimates of the heritability. Pedigree-based
quantitative designs, such as twin studies, have yielded substantial heritability (about .50) for
complex human traits (Polderman et al., 2015), whereas molecular genetic studies, such as
genome-wide association studies (GWAS), have had tremendous problems identifying genes
responsible for this heritability estimate (e.g., de Moor et al., 2012; Lo et al., 2017). One reason
for this apparent discrepancy could be that a small number of rare genetic variants in a
population could have large effects on personality traits. However, as far as we know, there is
little evidence for this explanation until now (see Hill et al., 2018, for an exception).
Another reason for the missing heritability problem is that a large number of multiple
genetic variants, each of them with a very small main effect, may account for a substantial
proportion of personality differences. In fact, molecular genetic analyses provided evidence
for personality traits as polygenic. That is, the common effect of multiple genes can account
for a substantial amount of variance in personality traits. However, a recent analysis involving
1,143,400 single nucleotide polymorphisms (SNPs; the smallest gene units that vary between
humans) accounts for only about 9% to 18% of variance in complex personality dimensions
(Lo et al., 2017). In other words, heritability estimates based on the combined effects of
multiple SNPs (<.20) are still not comparable to the estimates of substantial heritability
derived from twin studies (≈.50).
It is also possible that genes do not (only) unfold their effects additively, but rather
interact in multiple ways, giving rise to personality differences due to nonadditive genetic
effects. For example, two carriers of the same variant of a gene may differ in this gene’s
expression because of differences in their gene’s regulation. Similarly, two carriers with
different variants of SNPs may show similar gene expressions and thus comparable
phenotypes by virtue of gene regulations that promote similar outcomes. Those nonadditive
genetic effects are typically not considered in GWAS, which primarily focus on the additive
main effect of multiple genetic variants to estimate the SNP-based heritability, reflecting a
lower bound for the actual heritability of complex traits. The lower-bound (or narrow-sense)
heritability is defined as phenotypic (i.e., observable or measurable) variance component due
54
Sources of personality variance across the lifespan
to additive genetic sources, whereas the actual (broad-sense) heritability includes all genetic
variance components due to both additive and nonadditive genetic sources.
Additive genetic influences are shared between family members as a function of their
genetic relatedness. Nonadditive genetic influences encompass interaction effects between
two genetic variants within gene loci (i.e., allelic dominance, with the dominant allele
suppressing the effect of the recessive one) and between two or more genetic variants across
different gene loci (i.e., emergenesis; Lykken, 2006). Nonadditive genetic influences are
completely shared by monozygotic twins, but only marginally or not shared by other relatives.
Thus, those influences are implicated when monozygotic twin correlations are more than
twice as large as other sibling correlations or when twin studies show higher heritability
estimates compared to other study designs, such as adoption or other non-twin family studies.
Those indications for nonadditive genetic influences have often been reported for personality
traits. The meta-analysis of Vukasović and Bratko (2015) provided strong support for genetic
sources of personality variance, both from family and adoption studies (22%) and from twin
studies (47%), with marginal differences between men and women or across personality traits
and models. This finding suggests that nonadditive genetic effects probably play a role in
explaining individual differences in personality traits. Even though both the polygenic and the
gene × gene-interaction explanations and their empirical support allow a reduction of the
discrepancy between SNP-based and pedigree-based heritability estimates, a final gap
remains unexplained.
55
Sources of personality variance across the lifespan
Supplement B: Age Differences in Personality Measures Confound True Age Trends
Most behavior genetic studies investigating personality traits in childhood and young
adolescence used parent and teacher reports (e.g., De Fruyt et al., 2006; Ganiban, Saudino,
Ulbricht, Neiderhiser, & Reiss, 2008; Hudziak et al., 2003), whereas studies on late adolescent
and adult personality traits were mostly based on self-reports (Bratko & Butkovic, 2007;
Kandler et al., 2015; Viken et al., 1994). Parents typically knew all of their offspring, such as
both twins of a pair, and provided ratings for all of them. This dependence of raters increases
the probability for assimilation effects in case of identical twins and contrast effects between
non-identical twins leading to an overestimation of the genetic component. Even though
contrast effects can also be an explanation for differences in self-reports of twin siblings, this
would act to decrease the similarity of both identical and non-identical twins leading to an
underestimation of the genetic component. These rater biases alone can explain why the
meta-analysis by Briley and Tucker-Drob (2014) found a substantially declining genetic
component between childhood and young adulthood.
In addition, complex personality traits develop out of more basic temperament aspects
during childhood and adolescence (McAdams, 2015). Thus, most studies on children and
adolescents relied on temperament models and associated measures or are based on adapted
(mostly reduced) measures for capturing personality traits in childhood and young
adolescence (e.g., Ganiban et al.,2008; De Fruyt et al., 2006; Spengler, 2012). These measures
are different from those typically used to capture complex personality traits in adulthood and
are typically less reliable. As a consequence of this measurement invariance, lower heritability
estimates could artificially appear in childhood and adolescence compared to heritability
estimates based on more reliable measures of complex personality traits in adulthood. This
account provides an alternative explanation for why the quantitative review of Kandler and
Papendick (2017) found an increase of the genetic component between childhood and young
adulthood.
Of course, a cross-sectional study cannot disentangle true age from cohort effects
(Schaie, 1965). However, it follows from the aforementioned discussion that studies on age
(or cohort) differences in the sources of personality trait variance across the lifespan should
use the same personality measure for different age groups. In addition, measurement
invariance across age groups (or cohorts) should be established.
56
Sources of personality variance across the lifespan
Supplement C: Descriptive Statistics of HEXACO Personality Trait Measures
HEXACO Personality Trait Score Means and Standard Deviations
As can be seen in Table C1, women tended to show higher scores in Honesty-Humility
and Emotionality. These differences were consistent across subsamples and comparable to
previous studies (Ashton & Lee, 2009). Other gender differences did not appear as substantial
or consistent across subsamples. Standard deviations tended to be smaller in the older and
more clinical UK sample.
Table C1. Means and Standard Deviations for the HEXACO-60 Scale Scores
National subsamples
Croatian
Finnish
German
UK
Traits
Gender
M
SD
M
SD
M
SD
M
SD
Honesty-Humility
Male
3.45
0.65
3.50
0.64
3.44
0.61
3.71
0.59
Female
3.72
0.57
3.68
0.60
3.73
0.57
3.93
0.46
d
0.45
0.29
0.50
0.47
95%CI
[0.31;
0.59]
[0.21;
0.37]
[0.37;
0.63]
[0.30;
0.64]
Emotionality
Male
2.87
0.56
2.82
0.55
2.92
0.57
2.72
0.51
Female
3.58
0.57
3.57
0.56
3.35
0.60
3.28
0.50
d
1.25
1.35
0.73
1.12
95%CI
[1.10;
1.41]
[1.26;
1.43]
[0.59;
0.86]
[0.95;
1.29]
Extraversion
Male
3.44
0.59
3.24
0.71
3.55
0.58
3.53
0.47
Female
3.35
0.62
3.21
0.73
3.52
0.63
3.43
0.52
d
-0.15
-0.04
-0.05
-0.19
95%CI
[-0.29;
-0.01]
[-0.12;
0.04]
[-0.18;
0.08]
[-0.36;
-0.03]
Agreeableness
Male
3.23
0.53
3.32
0.54
3.16
0.53
3.13
0.56
Female
3.11
0.58
3.20
0.59
3.18
0.56
3.37
0.49
d
-0.21
-0.21
0.04
0.48
95%CI
[-0.36;
-0.07]
[-0.29;
-0.13]
[-0.09;
0.17]
[0.32;
0.65]
Conscientiousness
Male
3.43
0.58
3.51
0.53
3.53
0.60
3.68
0.48
Female
3.52
0.55
3.57
0.57
3.65
0.55
3.64
0.46
d
0.16
0.11
0.21
-0.09
95%CI
[0.02;
0.30]
[0.03;
0.19]
[0.08;
0.34]
[-0.25;
0.08]
Openness
Male
3.34
0.69
3.42
0.65
3.32
0.64
3.33
0.52
Female
3.42
0.69
3.39
0.69
3.34
0.60
3.31
0.58
d
0.12
-0.04
0.03
-0.04
95%CI [-0.03; 0.26] [-0.12; 0.04] [-0.10; 0.16] [-0.20; 0.13]
Note. See Table 1 for the numbers in the subsamples; significant Cohen’s ds based on 95%
confidence intervals are shown in boldface.
57
Sources of personality variance across the lifespan
HEXACO Personality Trait Score Correlations
The correlations between HEXACO personality trait scores based on the full sample are
shown in Figure C1. Statistically significant correlations (p < .001) are shown in boldface. No
correlation exceeded ׀ r = .25׀ and can thus be treated as moderate at maximum. Only five of
15 correlations were larger than ׀r = .10׀.
Figure C1. HEXACO personality trait correlations across all national subsamples.
As indicated by overlapping 99% confidence intervals, the correlations did not vary
significantly across the national subsamples, except for four associations (see Table C2). The
Emotionality-Extraversion link was significantly smaller for the Croatian sample (r = -.06)
compared to the UK sample (r = -.28). The correlation between Conscientiousness and
Emotionality was significant and higher in the Finnish (r = .09) than in the UK data (r = -.04),
but both correlations did not differ significantly from the correlations in the Croatian (r = .03)
58
Sources of personality variance across the lifespan
and the German sample (r = .03). The association between Extraversion and Openness was
higher in the UK sample (r = .22) than in the Croatian data (r = .07), but again both did not
differ significantly from the correlations based on Finnish (r = .11) and German data (r = .15).
Finally, the correlation between Conscientiousness and Openness was not significant and
smaller in the Finnish (r = -.01) than in the other samples. In sum, the correlations did not
differ systematically across subsamples and the differences can be treated as rather marginal.
All in all, the correlations are in the range comparable to those reported by Ashton and Lee
(2009).
Table C2. HEXACO-60 Trait Score Correlations for each Subsample
National subsamples
Croatian (n = 830)
Finnish (n = 2816)
German (n = 1142)
UK (n = 2238]
Variables
r
99% CI
r
99% CI
r
99% CI
r
99% CI
HH - Em
0.12
[.02; .21]
0.03
[-.03; .08]
0.03
[-.06; .11]
-0.01
[-.08; .06]
HH - eX
0.03
[-.06; .11]
0.02
[-.03; .08]
0.07
[-.01; .15]
0.01
[-.04; .06]
HH - Ag
0.22
[.12; .31]
0.23
[.17; .28]
0.17
[.09; .26]
0.28
[.22; .34]
HH - Co
0.20
[.11; .29]
0.18
[.13; .23]
0.25
[.16; .32]
0.20
[.13; .25]
HH - Op
0.10
[.00; .19]
0.08
[.03; .13]
0.02
[-.06; .10]
0.05
[-.00; .11]
Em - eX
-0.06
[-.14; .03]
-0.18
[-.23; -.14]
-0.19
[-.27; -.12]
-0.28
[-.33; -.22]
Em - Ag
-0.09
[-.19; .00]
-0.09
[-.14; -.04]
-0.08
[-.16; .00]
-0.11
[-.18; -.05]
Em - Co
0.03
[-.06; .13]
0.09
[.04; .15]
0.03
[-.05; .11]
-0.04
[-.10; .02]
Em - Op
-0.01
[-.10; .08]
-0.03
[-.09; .02]
-0.10
[-.18; -.01]
-0.09
[-.15; -.02]
eX - Ag
0.08
[-.02; .17]
0.05
[-.00; .11]
0.16
[.07; .23]
0.15
[.09; .21]
eX - Co
0.20
[.11; .28]
0.11
[.06; .16]
0.17
[.08; .24]
0.20
[.14; .26]
eX - Op
0.07
[-.02; .16]
0.11
[.06; .15]
0.15
[.07; .22]
0.22
[.17; .28]
Ag - Co
0.03
[-.07; .13]
0.06
[.00; .12]
-0.05
[-.13; .03]
0.01
[-.05; .07]
Ag - Op
0.03
[-.06; .12]
0.05
[-.00; .09]
0.00
[-.08; .09]
-0.01
[-.07; .05]
Co - Op
0.17
[.07; .26]
-0.01
[-.06; .04]
0.14
[.06; .22]
0.16
[.09; .22]
Note. HH: Honesty-Humility; Em: Emotionality; eX: Extraversion; Ag: Agreeableness, Co:
Conscientiousness; Op: Openness; 99% CI: 99% confidence intervals based on 1000
bootstrap samples.
59
Sources of personality variance across the lifespan
Supplement D: Age Trends in Personality Traits Controlled for Sex and National Differences
Each twin of a pair was randomly assigned to one of two Twin A and Twin B
subsamples. The analyses were run based on each subsample which allow a quasi-cross-
validation of results. We used the term “quasi” here, because the two samples are not
independent of each other. Because of the high statistical power provided by the large sample
size and separate tests for each personality trait, we only treated estimates at p < .01 across
both subsamples as statistically significant.
Honesty-Humility
The multiple regression analysis yielded linear age effects on Honesty-Humility beyond
significant sex and national differences (see Table D1 and Figure D1). The age trends are
comparable across Twin A and Twin B subsamples and those reported by Ashton and Lee
(2016). More specifically, women (vs. men) and UK (vs. other nations) reported higher levels
of Honesty-Humility. When age was added as predictor to the model, the differences between
the UK sample and other samples diminished, indicating that age differences underlie the
national differences. However, a new effect otherwise suppressed came up, indicating higher
levels of Honesty-Humility in Croatian compared to other national groups after controlling for
age differences. Men and women did not show different age trends as indicated by the non-
significant age × sex interaction.
Figure D1. Linear age trend of Honesty-Humility.
60
Sources of personality variance across the lifespan
Table D1. Results of the Multiple Regression Analysis for Honesty-Humility
Twin A (n = 3513)
Twin B (n = 3513)
Model 1:
Model 2:
Model 3:
Model 1:
Model 2:
Model 3:
Predictor
β
p
β
p
Β
p
β
p
β
p
β
p
Standardized regression weights
Sex
-.16
<.001
-.16
<.001
-.14
<.001
-.15
<.001
-.15
<.001
-.14
<.001
Nation 1
-.00
.853
.07
.002
.08
.001
-.01
.513
.07
.002
.07
.001
Nation 2
-.02
.299
.04
.085
.04
.192
-.01
.613
.06
.010
.05
.072
Nation 3
.16
<.001
.01
.843
.01
.671
.19
<.001
.01
.651
.02
.434
Age
.26
<.001
.31
<.001
.29
<.001
.31
<.001
Age²
-.08
.026
-.09
.008
Age³
-.02
.742
.02
.772
Age×Sex
.03
.192
.03
.119
Explained variance
R²
.072
<.001
.093
.099
.077
<.001
.104
.109
ΔR²
.021
<.001
.006
<.001
.027
<.001
.005
<.001
adj. R²
.070
.092
.097
.076
.103
.107
Note. Sex: 0 = female, 1 = male; Nation 1: dummy variable 1 (1 = Croatian, 0 = other); Nation
2: dummy variable 2 (1 = Finnish, 0 = other); Nation 3: dummy variable 3 (1 = UK, 0 = other);
ΔR²: increase in explained variance; adj. R²: adjusted R²; regression weights that were
statistically significant (p < .01) for both Twin A and B subsamples are shown in boldface.
Emotionality
The stepwise multiple regression analysis revealed significant larger scores of
Emotionality in women, particularly in younger women. There were no such age differences
in men, indicating larger gender differences for younger compared to older people (see Figure
D2). Initial national differences – Croatians and Finnish twins showed larger levels, whereas
the UK sample showed lower levels of Emotionality than Germans – diminished when age was
included as predictor (Table D2), except for Finnish people. Thus, sample differences were
mainly confounded with age differences. As Ashton and Lee (2016) also reported a negative
age trend for Emotionality in their study of about 100,000 English-speaking participants, the
national differences were obviously a result of underlying age trends.
61
Sources of personality variance across the lifespan
Figure D2. Different age trends of Emotionality in men and women.
62
Sources of personality variance across the lifespan
Table D2. Results of the Multiple Regression Analyses for Emotionality
Twin A (n = 2105)
Twin B (n = 2105)
Model 1:
Model 2:
Model 3:
Model 1:
Model 2:
Model 3:
Predictor
β
p
β
p
β
p
β
p
β
p
β
p
Standardized regression weights
Sex
-.46
<.001
-.46
<.001
-.43
<.001
-.45
<.001
-.45
<.001
-.43
<.001
Nation 1
.09
<.001
.06
.003
.07
.001
.06
.001
.04
.031
.04
.055
Nation 2
.10
<.001
.07
.003
.08
.001
.11
<.001
.09
<.001
.09
.001
Nation 3
-.09
<.001
-.03
.325
-.01
.695
-.11
<.001
-.08
.005
-.06
.035
Age
-.11
<.001
-.10
.061
-.07
.013
-.13
.012
Age²
.05
.170
-.01
.712
Age³
-.08
.154
.02
.711
Age×Sex
.09
<.001
.07
<.001
Explained variance
R²
.196
<.001
.200
.205
.196
<.001
.197
.200
ΔR²
.004
<.001
.005
<.001
.001
.013
.003
.003
adj. R²
.195
.199
.203
.195
.196
.198
Note. Sex: 0 = female, 1 = male; Nation 1: dummy variable 1 (1 = Croatian, 0 = other); Nation
2: dummy variable 2 (1 = Finnish, 0 = other); Nation 3: dummy variable 3 (1 = UK, 0 = other);
ΔR²: increase in explained variance; adj. R²: adjusted R²; regression weights that were
statistically significant (p < .01) for both Twin A and B subsamples are shown in boldface.
Extraversion
Controlling for small national (higher levels for German and Croatian compared to
Finns and UK) and non-significant gender differences, Extraversion showed a reversed U-
shaped age trend pattern with the highest level of Extraversion between 40 and 60 years of
age (Figure D3 and Table D3). Ashton and Lee (2016) reported an upward trend until the 70s.
However, this trend was negatively accelerated and their oldest age group was 74. Our sample
ranged in age until 90. Thus, in line with previous longitudinal studies on old age, people
tended to be more introverted in old age (Kandler et al., 2015; Mõttus, Johnson, & Deary,
2012).
63
Sources of personality variance across the lifespan
Table D3. Results of the Multiple Regression Analysis for Extraversion
Twin A (n = 2105)
Twin B (n = 2105)
Model 1:
Model 2:
Model 3:
Model 1:
Model 2:
Model 3:
Predictor
β
p
β
p
β
p
β
p
β
p
β
p
Standardized regression weights
Sex
.04
.019
.04
.014
.04
.052
.02
.183
.02
.159
.03
.136
Nation 1
-.07
.001
-.04
.078
-.05
.034
-.09
<.001
-.06
.012
-.05
.037
Nation 2
-.21
<.001
-.19
<.001
-.23
<.001
-.26
<.001
-.23
<.001
-.25
<.001
Nation 3
-.08
.002
-.14
<.001
-.13
<.001
-.04
.089
-.11
<.001
-.11
<.001
Age
.10
.001
.04
.521
.13
<.001
.16
.006
Age²
-.19
<.001
-.10
.004
Age³
.18
.004
.02
.713
Age×Sex
-.03
.110
.00
.964
Explained variance
R²
.026
<.001
.029
.038
.050
<.001
.055
.060
ΔR²
.003
.001
.009
<.001
.005
<.001
.006
<.001
adj. R²
.025
.028
.036
.049
.054
.058
Note. Sex: 0 = female, 1 = male; Nation 1: dummy variable 1 (1 = Croatian, 0 = other); Nation
2: dummy variable 2 (1 = Finnish, 0 = other); Nation 3: dummy variable 3 (1 = UK, 0 = other);
ΔR²: increase in explained variance; adj. R²: adjusted R²; regression weights that were
statistically significant (p < .01) for both Twin A and B subsamples are shown in boldface.
Figure D3. Reversed U-shaped age trend of Extraversion.
64
Sources of personality variance across the lifespan
Agreeableness
Regarding Agreeableness, the multiple regression analysis initially revealed higher
levels for participants of the UK sample. In addition, higher levels of Finns in Agreeableness
compared to Germans and Croatians also appeared after allowing for nonlinear age trends
and age × gender interaction effects (Table D4). The results indicated that older women
showed higher levels of Agreeableness than younger women and men in general (Figure D4).
In line with Ashton and Lee (2016), Agreeableness tended to show a slight downward age
trend until age 30 and an upward age trend thereafter, at least for women.
Table D4. Results of the Multiple Regression Analyses for Agreeableness
Twin A (n = 2105)
Twin B (n = 2105)
Model 1:
Model 2:
Model 3:
Model 1:
Model 2:
Model 3:
Predictor
β
p
β
p
β
p
β
p
β
p
β
p
Standardized regression weights
Sex
.02
.162
.02
.163
-.02
.281
.05
.004
.05
.003
.01
.540
Nation 1
-.04
.075
-.04
.095
-.03
.211
.00
.835
.02
.304
.03
.160
Nation 2
.03
.162
.03
.184
.09
.003
.07
.005
.09
.001
.14
<.001
Nation 3
.14
<.001
.14
<.001
.09
.002
.18
<.001
.14
<.001
.10
.001
Age
.00
.987
.15
.009
.07
.020
.23
<.001
Age²
.18
<.001
.16
<.001
Age³
-.18
.003
-.18
.003
Age×Sex
-.11
<.001
-.10
<.001
Explained variance
R²
.018
<.001
.018
.036
.021
<.001
.023
.038
ΔR²
.000
.987
.017
<.001
.002
.020
.016
<.001
adj. R²
.017
.017
.033
.020
.021
.036
Note. Sex: 0 = female, 1 = male; Nation 1: dummy variable 1 (1 = Croatian, 0 = other); Nation
2: dummy variable 2 (1 = Finnish, 0 = other); Nation 3: dummy variable 3 (1 = UK, 0 = other);
ΔR²: increase in explained variance; adj. R²: adjusted R²; regression weights that were
statistically significant (p < .01) for both Twin A and B subsamples are shown in boldface.
65
Sources of personality variance across the lifespan
Figure D4. Different nonlinear age trends of Agreeableness in men and women.
66
Sources of personality variance across the lifespan
Conscientiousness
Controlling for marginal national and gender differences, we found a nonlinear
(reversed U-shaped) age trend in Conscientiousness (Table D5 and Figure D5). This age pattern
again is consistent with previous studies who reported upward trends for Conscientiousness
until the 60s (Ashton & Lee, 2016) and downward trends in old age (Kandler et