Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Upload Folder
/
usctheses
/
uscthesesreloadpub_Volume1
/
Sources of stability and change in the trajectory of openness to experience across the lifespan
(USC Thesis Other)
Sources of stability and change in the trajectory of openness to experience across the lifespan
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
SOURCES OF STABILTY AND CHANGE IN THE TRAJECTORY OF OPENNESS
TO EXPERIENCE ACROSS THE LIFESPAN
by
Emily Schoenhofen Sharp
________________________________________________________________________
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
May 2012
Copyright 2012 Emily Schoenhofen Sharp
ii
ACKNOWLEDGEMENTS
Data were collected under the aegis of the National Institute on Aging (NIA)
grants AG04563 and AG10175. Analyses and writing were supported by in part by a
grant from the National Institute of Aging: Multidisciplinary Research Training in
Gerontology (5T32AG00037) and from a K. Patricia Cross dissertation grant awarded by
Road Scholar (formerly Elderhostel). No conflict of interest exists. Chapter 2 (Study 1)
was presented at the annual meeting of the Gerontological Society of America in Boston,
Massachusetts, 2011.
I gratefully thank my advisor and mentor, Margaret Gatz, PhD, for her insight and
guidance on this project. I thank Chandra Reynolds, PhD who was an essential advisor
and offered countless hours of statistical consultation. Finally, I thank my family for their
love and support throughout this adventure in dissertating.
iii
TABLE OF CONTENTS
Acknowledgements ii
List of Tables v
List of Figures vii
Abstract ix
Chapter 1. General Introduction 1
Chapter 2. Longitudinal Stability and Change in Openness to
Experience over the Adult Lifespan
Chapter 2 Abstract 8
Chapter 2 Background 9
Chapter 2 Method 13
Chapter 2 Results 22
Chapter 2 Discussion 27
Chapter 2 References 48
Chapter 3. Relationship between the Longitudinal Trajectory of
Openness to Experience and Mortality
Chapter 3 Abstract 55
Chapter 3 Background 56
Chapter 3 Method 59
Chapter 3 Results 66
Chapter 3 Discussion 70
Chapter 3 References 87
Chapter 4. Longitudinal Genetic and Environmental Variation in
Openness to Experience over the Adult Lifespan
Chapter 4 Abstract 90
Chapter 4 Background 91
Chapter 4 Method 96
Chapter 4 Results 102
Chapter 4 Discussion 105
Chapter 4 References 126
Chapter 5. General Discussion 130
Comprehensive References 139
iv
Appendices
Appendix A: Openness to Experience 6-item Scale 149
Appendix B: Self Rated Health Scale 150
Appendix C: Activities of Daily Living Scale 151
Appendix D: Cardiovascular Disease Items 152
v
LIST OF TABLES
Table 1: Number of openness measurements completed for the total sample
and by males and females with openness means and standard
deviations at baseline
34
Table 2: Attrition between openness measurement occasions with means
and standard deviations for openness at last measurement occasion
35
Table 3: Demographic and variable characteristics of the total sample and
for males and females
36
Table 4: Correlations between openness measurement waves and study
covariates by males and females separately
37
Table 5: Participation and means and standard deviations across openness
measurement occasions by age group
38
Table 6: Initial growth model estimates (standard errors) of openness to
experience for the total sample
39
Table 7: Latent growth model estimates (standard errors) with covariates
for the total sample
40
Table 8: Model estimates (standard errors) by sex
41
Table 9: Final model estimates (standard errors) by age group
42
Table 10: Participation in openness measurement occasion for total sample
and by censoring status (as of December 31, 2010)
75
Table 11: Participant attrition with means and standard deviations for
openness at last measurement occasion (as of December 31, 2010)
76
Table 12: Number of years between last openness measurement occasion
and death (for confirmed descendants) in the total sample
(N=1947)
77
Table 13: Means and standard deviations for individuals who died within 1
to 5 years of their last openness measurement and for participants
who remained alive
78
vi
Table 14: Correlations between openness measurement waves and study
covariates for dead (lower half) and alive (upper half)
79
Table 15: Initial survival analysis of baseline openness to experience
adjusted for age, sex, and education. Comparison of total sample,
twin Sample A and twin Sample B.
80
Table 16: Survival results estimates (standard errors) with covariates from
two-stage growth and survival models.
81
Table 17: Effects of openness to experience on age of death in simultaneous
longitudinal and Cox proportional hazard model.
82
Table 18: Number of openness measurement waves completed by zygosity
and rearing status (numbers represent individual twins)
111
Table 19: Attrition across openness measurement occasions by twin status
(MZ vs. DZ) with means and standard deviations for openness at
last measurement occasion
112
Table 20: Means and standard deviations for openness and age by
measurement occasion by zygosity, rearing status, and sex
113
Table 21: Intraclass correlations at each openness measurement occasion by
gender and age group
114
Table 22: Maximum likelihood estimates (standard errors) from phenotypic
LGM models
115
Table 23: Biometric parameter estimates from ASCE and AE models:
Cholesky factor loadings for males
116
Table 24: Biometric parameter estimates from ASCE and AE models:
Cholesky factor loadings for females
117
Table 25: Expected ASCE (a) variances and unexplained variance across
ages by sex and expected AE (b) variances and unexplained
variance across ages by sex
118
Table 26: Heritability and environmental contributions to openness to
experience at age 65 based on ASCE model (a) and AE model (b)
119
vii
LIST OF FIGURES
Figure 1: Participant attrition with mean and standard deviations for
openness to experience
43
Figure 2: Quadratic growth model with covariate
44
Figure 3: Plot of longitudinal openness scores for sample of male (a) and
female (b) participants with up to 6 measurement occasions
45
Figure 4: Plot of longitudinal openness trajectories for the youngest (a),
middle (b), and oldest (c) age groups. Randomly selected 1 twin
from each pair (Twin Sample A)
46
Figure 5: Trajectory of openness to experience across the adult lifespan (a)
and best fitting trajectory of openness to experience across the
adult lifespan by age group (b)
47
Figure 6: Latent growth model with shared survival function
83
Figure 7: Openness scores from participants who died prior to the next
measurement occasion (a) and participants who dropped out prior
to the next measurement occasion (b) compared to participants
who remained in the study.
84
Figure 8: Kaplan-Meier survival plot by baseline openness to experience
(a) and age (b)
85
Figure 9: Kaplan-Meier survival plot by sex (a) and education (b)
86
Figure 10: Quadratic growth model
120
Figure 11: Biometric quadratic growth model
121
Figure 12: Longitudinal trajectories of openness for sample of (a) MZ twins
reared apart, N=131 and (b) MZ twins reared together N=180
122
Figure 13: Longitudinal trajectories of openness for sample of (a) DZ twins
reared apart N=326, and (b) DZ twins reared together N= 284
123
viii
Figure 14: Predicted biometric parameters estimated from the growth model
for (a) the full ASCE, and (b) AE models for males
124
Figure 15 Predicted biometric parameters estimated from the growth model
for (a) the full ASCE, and (b) AE models for females.
125
ix
ABSTRACT
This dissertation examined sources of stability and change in openness to
experience across the lifespan. The data came from the Swedish Adoption/Twin Study of
Aging (SATSA), a large longitudinal study of twins that included up to six measurement
occasions of openness. Study 1 examined the longitudinal trajectory of openness and
variables hypothesized to account for individual differences in level and change in
openness. Using twins as individuals (and adjusting for the correlation between twins in
the modeling), results from phenotypic latent growth curve modeling confirmed previous
literature that openness exhibits small but significant decline in older adulthood. Model
estimates indicated a total decline of about 1.5 points in openness over twenty years,
generally occurring after age 60. Thus, age accounted for the decline in openness. Males
and females were not significantly different in level or rate of decline. Education was
important in predicting individual differences in level of openness, but not decline. Self-
rated health, activities of daily living, and cardiovascular disease did not explain
individual differences in level or decline in openness. Study 2 examined the relationship
between openness to experience and death. Results from two methods of modeling
growth and survival, adjusted for entry age, sex, education and twinness, suggested that
the slope of openness was significantly associated with death such that individuals who
had a faster rate of decline were at a greater risk for death. Level of openness (mean
openness at age 65) was unrelated to risk for death. The overall results from study 1 and
study 2 suggested that decline in openness was normative across older participants and is
x
perhaps best conceptualized in relation to theories of terminal decline - the period of time
prior to death during which multiple domains have been found to decline. Study 3
examined the pattern of genetic and environmental influences across the longitudinal
trajectory of openness. Results from a biometric latent growth model (a latent growth
model combined with a Cholesky decomposition model) suggested that a reduced
additive genetic (A) and nonshared environment (E) model provided the best fit. The AE
model estimates suggested that the individual differences in level, linear slope, and
quadratic slope were primarily accounted for by genetic influences. For both men and
women additive genetic influences generally increased with age. For men, nonshared
environmental influences increased similar to genetic influences but then declined after
age 65. For women, the contribution of nonshared environmental influences declined
steadily from age 40. The findings from Study 3 give initial support to one theory of
personality development that has suggested that normative patterns of change in adult
personality would be explained by primarily genetic influences.
1
CHAPTER 1. GENERAL INTRODUCTION
Recent studies have demonstrated that personality has similar predictive power to
that of socioeconomic status and cognitive ability on life outcomes such as occupational
attainment, physical health, and mortality (for a review see Roberts, Kuncel, Shiner,
Caspi, & Goldberg, 2007; Caspi, Roberts, & Shiner, 2005). The literature suggests that
personality imparts a powerful influence over an individual’s life experience across
emotional, cognitive, and physical domains of functioning. Given the role of personality
in shaping outcomes across the lifespan, it is important to investigate the trajectory of
personality longitudinally and the potential sources of stability and change in personality.
A fundamental question of developmental personality research is whether
personality traits are stable or exhibit change over time. Much of the recent debate in the
research on stability and change in personality has focused on the Five Factor Model
(FFM) of personality, which includes neuroticism, extraversion, openness to experience,
conscientiousness, and agreeableness (McCrae & John, 1992). These traits have also been
termed the “Big Five”. A recent review of the personality literature examined stability
and change across 80 studies that included individuals aged 10-101 (Roberts, Watson, &
Viechtbauer, 2006). The authors concluded that the big five traits followed specific
patterns of development during childhood and also continued to show small but
significant mean-level change in adulthood. Further there was evidence for substantial
fluctuations at the individual level across all traits. The results of this review were
2
important because much of the previous research had concluded that the big five traits
were relatively stable after age 30 (Costa & McCrae, 1994; 1997).
Personality stability. The generally accepted theory of personality development
has held that personality undergoes developmental changes and continued maturation
throughout childhood and early adulthood and then traits remain stable in mean level
across adulthood (Costa & McCrae, 1997; McCrae & Costa, 1990). This hypothesis of
stability has been supported by numerous cross-sectional and longitudinal studies that
have demonstrated rank-order and mean-level stability after age 30 (Terracciano, Costa &
McCrae, 2006; Terracciano, McCrae, & Costa, 2010). Further, given the developmental
patterns of personality, scholars of this theory have hypothesized that any significant
patterns of change after the traditional markers of maturity would be attributable to
primarily biologically-based or genetic influences (McCrae et al., 2000; McCrae & Costa,
2008).
Personality flexibility. More recently, new studies have challenged the hypothesis
of stability in adult personality with evidence of both mean level and individual level
change occurring after markers of maturity have been reached (Fraley & Roberts, 2005;
Roberts & DelVecchio, 2000; Robins, Fraley, Roberts, & Trzesniewski, 2001). This
hypothesis of change suggests that personality continues to develop across adulthood and
that change in personality may be explainable by individual differences in life events
(Roberts, Caspi, & Moffit, 2001; Roberts, Robins, Trzesniewski, & Caspi, 2003;
Mroczek & Spiro, 2003). This conceptualization of personality development suggests that
sources of stability and change in personality are outcomes of varying interactions
3
between genetic and environmental influences across the lifespan (Roberts, Wood, &
Caspi, 2006). However, the patterns of potential change in adult personality are not well
understood. For example, change in adult personality could represent individual specific
fluctuations or traits might show change in the same direction across a sample or both
fluctuation and normative change could occur.
Individual differences in intraindividual change. The goal of the current project
was to examine the longitudinal trajectory of openness to experience across the lifespan
and to examine individual differences in intraindividual change in openness to experience.
It is important to differentiate intraindividual change from intraindividual variability
(Alwin, 1994; Baltes, 1987; Baltes & Nesselroade, 1973). Briefly, intraindividual change
and intraindividual variability both describe within-person variability over time.
Intraindividual change is more commonly defined as slow, enduring changes over time
such as those that are reflective of, for example, developmental changes, long-term
learning and skill attainment (Mroczek, Spiro, & Griffin, 2006). In contrast,
intraindividual variability refers to potentially reversible fluctuations that that are
influenced by individual-specific factors such as emotional, cognitive, and physical
health (Nesselroade, 1991). Based on the literature, personality change seems best
described as individual change, where changes are developmental and long term.
However, the way change is conceptualized may be somewhat dependent upon the
spacing between measurement points in a longitudinal study design – shorter spacing has
generally been used to address questions of intraindividual variability; whereas longer
4
intervals are more appropriate for intraindividual change (Fleeson & Jolley, 2004;
Hultsch, Strauss, Hunter, & MacDonald, 2008).
Sources of Stability and Change. If personality traits change across the adult
lifespan, what factors might be associated with that change? Baltes (1997) suggested that
personality is best described within the context of a life span developmental model. In
this theoretical framework personality is plastic in nature and can exhibit both stability
and change depending on what is most adaptive for that individual (see also Baltes,
Lindenberger, & Staudinger, 1998). To examine this question, a few studies have begun
to explore the potential sources of stability and change in personality and have examined
whether demographic factors (e.g. sex) or life events (e.g. loss of a spouse) might be
associated with different trajectories of change in personality at the individual level
(Mroczek & Sprio, 2003: Small, Hertzog, Hultsch, & Dixon, 2003). Yet, to my
knowledge, the recent literature has not examined whether health related factors might be
sources of individual differences in level or change in personality traits (although a
substantial body of literature exits on the cross-sectional relationship between personality
and health, for a recent review, see Williams, Smith, & Cribbet, 2008).
Beyond demographic and chronic health factors, death may also be an important
source of individual differences in intraindividual change. For example, change in later
adulthood may be related to global declines as an individual approaches death. The years
preceding death have been found to be associated with functional declines across multiple
domains (see Berg, 1996). This period of decline has been termed “terminal decline” in
the aging literature. Yet, it is also possibility that continued change, perhaps even
5
declines in personality traits prior to death, may be adaptive. For example, in older ages,
personality change could be part of an optimization and compensation process (Baltes &
Baltes, 1990; Baltes et al., 1998). In this case, changes in personality might be associated
with adaptive responses to age-related changes that maximize an individual’s functioning.
Further, recent models of aging offer a framework for adaptive personality change in
relation to death. For example, the socioemotional selectivity theory (SST, see Carstensen,
2006; Carstensen, Mickels & Mather, 2006) suggests that as time-to-death becomes more
salient, individuals change their goals. This fundamental shift in individuals’
relationships with their environment may require adaptive shifts in personality as well.
However, very few studies have examined the relationship between longitudinal
personality change and mortality. Finally, in terms of the current debate, one fundamental
question is whether factors influencing stability and change in adult personality are
environmental or genetic in origin. Primarily genetic influences would suggest that traits
follow genetically based stages of development and change; whereas primarily
environmental influences would suggest the role of individual environments (e.g. life
events) as a source of change. A longitudinal behavior genetic approach is necessary to
examine whether the underlying sources of stability and change are attributable to genetic
or environmental processes (Krueger, Johnson, & Kling, 2006).
Openness to Experience. Openness was selected because it has been identified as
the least understood of the Big Five (Caspi, Roberts & Shiner, 2005; Costa & McCrae,
1997). Openness to experience has been described as an intrinsic wish for knowledge,
curiosity, and the ability to assimilate novel ideas (McCrae, 1994; McCrae & Sutin,
6
2009). The openness scale used in the current project was comprised of both cognitive
engagement items (e.g. thinking about philosophical ideas) and behavioral engagement
items (e.g. taking up new hobbies, trying out new and foreign foods). A recent study
using SATSA data examined the relationship between openness to experience and
cognitive functioning. Results indicated that individuals reporting greater levels of
openness performed better on tasks assessing five different domains of cognitive
functioning, even after adjusting for level of education, activities of daily living, and the
presence of cardiovascular disease (Sharp, Reynolds, Pedersen, & Gatz, 2010). Given
that openness has been described in terms of cognitive and emotional engagement with
experiences and that openness may have a unique relationship with cognitive aging,
change in self-reported openness may be an important indicator of broader age-related
changes within the individual.
Previously, Pedersen and Reynolds (1998) examined stability and change in the
personality traits of extraversion, neuroticism, and openness to experience in the SATSA
sample. Results specific to openness suggested mean-level stability in adults less than 55
years old and mean-level decline after age 55. Results from the longitudinal genetic
analyses modeling change over time indicated overall genetic stability but with different
patterns of genetic and environmental contributions for males and females. For males, the
greatest proportion of the total variance was attributable to nonshared environments, but
this declined with time; additive genetic influences were small to moderate but stable;
and shared environmental influences were negligible. For females, nonshared
environmental influences accounted for most of the total variance, additive genetic
7
influences were moderate and increased with time; influences due to correlated
environment were small and declined with time, whereas influences due to shared
environments were generally small but quite variable over time. The authors identified
intraindividual variability, time-to-death factors, increased number of measurement
points, and latent growth models as important areas for future study in personality
research. It was the aim of the current project to address all of these areas.
The current project. The goal of this dissertation was to examine stability and
change in openness to experience over the adult lifespan and to identify individual
differences in the trajectory of openness based on demographic factors, chronic health
indicators, and risk for mortality, as well as to examine the genetic and nongenetic
sources of stability and change in openness. The specific aims of this dissertation were
addressed in three studies. Study 1 investigated stability and change in the phenotypic
(non twin) trajectory of openness to experience across the adult lifespan and evaluated
demographic and health-related predictors of individual differences in longitudinal
openness. Study 2 examined the relationship between openness to experience and death,
specifically, whether level and slope (i.e. change in openness) were associated with risk
for mortality. Study 3 examined the longitudinal pattern of genetic and environmental
influences on openness to experience across the adult lifespan. Combined, these three
studies advanced the personality and aging literature by exploring potential sources of
stability and change in openness to experience across the adult lifespan.
8
CHAPTER 2. THE LONGITUDINAL TRAJECTORY OF OPENNESS TO
EXPERIENCE ACROSS THE ADULT LIFESPAN
CHAPTER 2 ABSTRACT
The purpose of this study was to examine the sources of stability and change in
openness to experience across the adult lifespan. Participants were 1947 individuals who
completed at least one measurement of openness from the Swedish Adoption/Twin Study
of Aging (SATSA). Openness to experience scores were available for up to 6
measurement occasions spanning 23 years. Both individual level change and mean level
change across the sample were identified. Latent growth models were used to examine
the longitudinal trajectory of openness to experience. Age was centered at 65 and the
models adjusted for the correlation between twins. Covariates included sex, age group,
education, self-rated health, activities of daily living, and cardiovascular disease. Results
from the best fitting quadratic model revealed significant mean level decline in openness
with additional accelerated decline. No sex differences were identified on either level or
slope of openness. Education explained individual differences in level of openness but
not decline. Age group was used to describe the differences in the rate of decline per
decade, such that the youngest participants did not decline, the middle-aged had small
linear and nonlinear change, and the oldest participants evidenced significant linear
decline. Neither self-rated health, activities of daily living, nor the presence of
cardiovascular disease were related to slope in openness. Thus, differences in age group
9
and education described mean-level stability and change in openness to experience better
than sex differences, disability, or disease. However, individual-level fluctuation was
independent of age and remains unexplained.
CHAPTER 2 BACKGROUND
A fundamental question of developmental personality research is whether
personality traits are stable or continue to exhibit change over time. The generally
accepted theory of personality has been that traits undergo development during childhood
and young adulthood, but after adulthood is reached traits remain relatively stable (Costa
& McCrae, 1997). This hypothesis is based on the results of cross-sectional and
longitudinal studies that have demonstrated rank-order and mean-level stability of
personality traits during adulthood (Terracciano, Costa, & McCrae, 2006). In contrast,
however, recent studies have found evidence to suggest that personality continues to
undergo change across the adult lifespan. Specifically, research has identified significant
individual-level and mean-level change occurring after the traditional markers of maturity
(for a review see Roberts, Watson, & Viechtbauer, 2006). Further, developmental
influences such as individual life events have been suggested to be potential sources of
change in adult personality (Caspi, Roberts, & Shiner, 2005; Lewis, 2001; Helson, Kwan,
John, & Jones, 2002).
This recent debate on stability and change in personality has focused on the Big
Five traits of personality. Generally, the traits of extraversion, neuroticism, and
conscientiousness have been the traits most studied, whereas openness to experience has
10
been identified as the least understood of the five factors (Caspi, Roberts & Shiner, 2005;
McCrae & Costa, 1997). Openness to experience has been conceptualized as an
emotional connection with experiences and has been found to be highly correlated with
both education and cognitive ability. Open individuals are described as having an
intrinsic wish for knowledge, curiosity, and the ability to assimilate novel ideas. In
contrast, closed individuals are more rigid in their beliefs (McCrae, 1994; McCrae &
Sutin, 2009). The facets of Openness (imaginative, creative and enjoyment of novel
experiences) and Intellect (clever and insightful), have been suggested to be the core of
this personality trait (John & Srivastava, 1999).
In a recent review, Roberts and colleagues (2003) reported that traits associated
with openness to experience exhibited increases in adolescence and young adulthood,
stability across midlife, and then mean-level declines in older adulthood. The review
indicated no sex differences in patterns of mean level change in openness over time,
suggesting that any differences in level of openness based on sex were maintained over
time. Terracciano, McCrae, Brant, and Costa (2005) also found significant linear mean-
level decline in trait openness over time, but within the construct of openness, the facets
of openness to ideas, values, and aesthetics were found to have almost no decline
between age 30 and 90. Further, a previous study using SATSA data (Pedersen &
Reynolds, 1998) suggested mean-level declines after age 55, but found evidence for
significant sex differences in the trajectory of openness to experience over time. This
pattern was similar to findings from a more recent analysis of SATSA data that suggested
mean-level stability for males, but significant mean-level decline in openness for females
11
(Sharp, Reynolds, Pedersen, & Gatz, 2010). Thus, while the literature has suggested some
consistent patterns in the trajectory of openness, questions remain regarding stability and
change at the individual-level as well as the mean-level. Further, to our knowledge, no
studies have examined the potential sources of the decline in openness.
What factors might influence declines in openness in later life? Proponents of a
hypothesis of change in adult personality have suggested that different life events may
affect individual differences in intraindividual change over time (Roberts & Mroczek,
2008). For example, Mroczek and Spiro (2003) examined individual differences in
extraversion and neuroticism using longitudinal personality data from the Normative
Aging Study. Latent growth modeling suggested about 30% of the total variation for both
traits was within-person. Furthermore, results from the growth models indicated
individual differences in the level and rate of change for neuroticism was partially
accounted for by memory complaints, marriage or remarriage, and death of a spouse.
With respect to openness, demographic factors such as sex, age (or birth cohort),
and education may be related to initial mean level differences as well as the trajectory of
openness with age. For example, cognitive ability has been found to be associated with
openness (Booth, Schinka, Brown, Mortimer, & Borenstein, 2006; Sharp et al., 2010). It
is likely that baseline openness will be strongly correlated to education but it is unknown
whether greater education might also act as a buffer to declines in openness over the
lifespan. Similarly, individuals born during different time periods might report different
initial levels of a personality trait due to historical social and environmental differences
and might also evidence different trajectories of change with age (see Elder, 1978; 1998).
12
Given that previous research has documented declines in openness beginning
around age 60, it seems possible that the decline in openness may be associated with
health-related life events such as the onset of disease or disability. In the current study,
we selected cardiovascular disease (CVD), self-rated health (SRH), and activities of daily
living (ADL) as indices of individual differences in health. CVD was selected as a global
indicator of vascular health and a disease that might limit individuals from previously
enjoyed activities. The SRH scale was included because it has been identified as an
important predictor of aging-related changes in cognition and physical functioning
(Svedberg, Gatz, Lichtenstein, Sandin, & Pedersen, 2009). Finally, the ADL scale was
included as a measure of individual disability in daily life.
The purpose of this study was to examine stability and change in openness to
experience at both the mean and individual level as well as to examine potential
predictors of change by examining the effects of sex, age group, education, SRH, ADL,
and CVD. It was hypothesized that a) openness scores would exhibit significant mean-
level decline over age, b) males and females would have significantly different
trajectories, with greater decline among females, c) the oldest participants would have the
steepest decline, and d) that there would be individual differences in intraindividual
change longitudinally such that sex, age group, and education would predict differences
in level of openness, whereas age group, SRH, ADL, and CVD would predict differences
in the slope (i.e. rate of change) in openness. A latent growth modeling approach was
selected for the longitudinal data analysis. To our knowledge, no studies have examined
13
the impact of health factors as a source of individual differences in stability and change in
the trajectory of openness to experience over the lifespan.
CHAPTER 2 METHOD
Participants
Participants were drawn from the Swedish Adoption Twin Study of Aging
(SATSA). SATSA data collection began in 1984 and continues to collect longitudinal
follow-ups approximately every 3 years, although there was a 10.5 year gap between the
4
th
and 5
th
waves. At the time of this study, six measurement occasions of openness were
available for analysis. The first questionnaire (Q1) was sent out in 1984, Q2 was sent out
in 1987, Q3 was sent in 1990, Q4 in 1993, Q5 in 2004, and Q6 in 2007. The SATSA
sample was selected because it offers a large, unique source of longitudinal twin data and
it has been found to be representative of the larger Swedish population on a variety of
environmental and sociological variables (Cederlof, Friberg, & Lundman, 1977).
Across waves, 1947 individual twins provided 5683 data points of openness
measurement. Of the total sample, 58% were female. The mean entry age was 59 years
(range=26-93), median study follow-up time was 9 years (maximum = 23 years), and the
median number of waves completed was 4. A total of 1352 (69%) participants completed
3 or more waves and 399 (20%) participants completed all 6 waves. The distribution of
participants by the number of waves completed is presented in Table 1. In terms of
attrition, the most common pattern of missing was for participants to have dropped out or
died between waves 4 (1993) and 5 (2004). During this 10.5 year gap, 600 participants
14
left the study. Figure 1 presents the pattern of participant attrition across waves.
Generally, participants who died prior to the next measurement occasion had a lower
openness score than participants who remained in the study (see Table 2).
It is important to note that the SATSA sample is comprised of twins. However,
the current study treated twins as individuals and neither genetic nor family components
were incorporated into the analyses. As twins are not independent of each other, the
correlation between twins was accounted for in the phenotypic analyses (see Statistical
Method).
Measures
Openness to Experience. Openness was measured by responses on a 6-item scale.
The items were identified by factor analyzing the 26-item scale from the widely used and
validated NEO-PI (Costa & McCrae, 1985), and choosing the 6 highest loading items
(see Bergeman et al., 1993). This scale tapped the intellectual component as well as
engagement in new experiences. It was scored in the traditional fashion of the NEO-PI
based on a 5-point likert scale ranging from strongly disagree to strongly agree. Items
were summed to create a total score (range 6-30). Openness to experience data were
collected via questionnaires mailed to all eligible participants. The openness measure
included the same six items across all waves of data collection. See Appendix A for the
scale items.
Education. Educational attainment was treated as a continuous variable ranging
from 1 (elementary school) to 4 (university or higher). In the current sample, the majority
(60%) have 6 years of education – the required education during this time-period in
15
Sweden. Education was known for 90% of the sample with an openness measurement
and was collected at Q1.
Age Group. Three age groups of approximately equal size were created that
categorized individuals with respect to the Great Depression in Sweden. Given the study
design, age group and birth cohort are difficult to separate, because participants entered
the study at different ages and also could have completed their first (baseline) openness
measurement at an occasion other than Q1. However, to make interpretation more
meaningful, age groups were defined by birth year. The oldest group included individuals
born 1891-1919, inclusive (N=732, mean entry age = 73.8). This group was born prior to
the Great Depression and would have been born or been young children during the
Spanish flu epidemic. The middle age group was made of individuals born 1920-1933,
inclusive (N=643, mean entry age=59.2). This age group was born during the Great
Depression. The youngest age group was born 1934-1958, inclusive (N=572, mean entry
age=42.9). This age group was born after the Great Depression. Further, individuals born
after 1945 had access to penicillin, and those born after 1950 had access to state provided
healthcare as children.
Self-Rated Health (SRH). This measure was included as part of the SATSA
questionnaire and a baseline measure was available for 99% of participants with an
openness measurement. This measure contained four items including a question about
global health, a question inquiring as to health in comparison to 5 years ago, a question
comparing the individual’s health to their peers, and a final question asking whether their
health impeded their activities. These items were answered based on different three-point
16
scales. Thus, items were standardized to a mean of 0 and a standard deviation of 1 before
summing (for more detail see Harris, Pedersen, McClearn, Plomin, & Nesselroade, 1992).
Higher scores indicated a higher self-rated health. See Appendix B for the SRH scale
items.
Activities of Daily Living (ADL). ADL was measured as part of the SATSA
questionnaire at Q1 and was available for 89% of individuals with an openness
measurement. The scale was constructed of 14 yes-no items, where seven of the
questions pertained to instrumental activities and another seven questions pertained to
physical activities (see Pedersen & Harris, 1990). Items were reverse scored and a total
score was summed across questions. Higher scores indicated better functioning. See
Appendix C for the ADL scale items.
Cardiovascular Disease (CVD). The presence of CVD was assessed as part of a
self-reported health measure included in the SATSA questionnaire and was available for
99% of participants with an openness measurement. If an individual reported yes to any
of the following: angina pectoris, high blood pressure, heart insufficiency, heart attack,
claudication, phlebitis, circulation problems, thrombosis, stroke, tachycardia, a heart
operation, or heart valve problem, then they were coded to have self-reported
cardiovascular disease. A portion of this measure was comparable to the Rose
Questionnaire (Rose, McCartney, & Reid 1977). See Appendix D for the CVD items.
Statistical Method
The goals of the statistical analysis were to examine the longitudinal trajectory of
openness to experience and to identify potential sources of individual differences in
17
stability and change within that trajectory. Four indices of individual variability in
openness scores were created to measure the amount of individual variation
longitudinally including, range in openness (participants’ highest openness score minus
their lowest score), the absolute change over time (the sum of the absolute difference
between each measurement occasion), and the annual rate of change (ARC) in openness
(the sum of the non-absolute difference between measurements occasions divided by the
total number of years across measurements). Finally, the standard error of prediction
(SEP) as presented by Dudek (1979, Formula 3) was calculated. Here, SEP = σ*(1-r
2
)
.5
,
where for Open1, σ = 4.08 and r = 0.70 (Cronbach’s α). Rounding to the closest integer
resulted in an SEP of 3 points. Thus, a change in score by more than 3 points from one
measurement to the next suggested greater variability than would be the expected given
the sample distribution. Of note, this was essentially equal to one standard deviation in
openness (4.08). Measures of variability could only be calculated for individuals with 2
or more measurement points. These measures were used to describe the amount of
fluctuation in openness across measurement occasions and to examine potential sex and
age group differences.
Latent growth curve modeling (LGM) was used examine the phenotypic pattern
of stability and change in openness longitudinally. Analyses were conducted using PROC
MIXED in SAS 9.2 (SAS Institute, 2000). LGMs measure and allow for comparisons of
individual trajectories of decline as well as an average trajectory of decline across the
entire sample. This technique allows for the use of both missing and non-sequential data
points. Further, data from individuals with only one measurement occasion can be
18
included in the analysis to stabilize both mean and variance estimates (Bryk &
Raudenbush, 1987; Finkel, Reynolds, McArdle, Gatz, & Pedersen, 2003; McArdle &
Anderson, 1990; McArdle & Hamagami, 1992). Latent growth curve models allow for
missing data by giving more weight to individuals with the most measurement occasions
or time points. A frequent concern of longitudinal studies is missing data and whether the
patterns of missing data are ignorable or nonignorable. A full maximum-likelihood
estimate (MLE) technique was used in the latent growth models. This technique
aggregates all available data on participants included in the analyses to estimate the
model parameters. A basic statistical assumption of MLE is that the incomplete data
points are missing at random (MAR). The MAR assumption has been typically applied to
incomplete longitudinal data (Little, 1995; McArdle et al., 2004).
In this study, latent growth models were defined by the intercept, which gives an
estimate of the typical score at a specific age or point in time, a linear slope, the
systematic longitudinal variation around the intercept, and a nonlinear quadratic slope,
which further characterize a trajectory, specifically acceleration (or deceleration) in a
curve (See Figure 2). The intercept and slopes are considered fixed effects – parameters
that describe the overall trajectory of the sample. In this type of developmental aging
model, particularly where age varies greatly at each measurement occasion, change is
best evaluated in relation to chronological age (rather than by measurement wave or time
point - see McArdle, Ferrer-Caja. Hamagami, Woodcock, 2002).
In the current study, age was centered at the sample mean age at first
measurement occasion (65 years) and divided by 10 to evaluate change per decade. Thus,
19
the fixed effect intercept estimated the sample-level mean openness score at age 65, the
slope predicted the amount of per decade linear change in openness, and the quadratic
term represented any additional nonlinear change per decade. The addition of covariates
allowed for an examination of individual differences in change in openness based on, for
example, the presence of cardiovascular disease. Education, SRH, and ADL were mean
centered. Sex and CVD were effect-coded (males=-.5, females=+.5; CVD absent=-.5;
CVD present=+.5). Age group was entered into the model as a class variable, but also
effect coded as (oldest=1, middle=0, youngest=-1). As a class variable, it was modeled
such that the first level (young) and second level (middle) were compared to the oldest
(reference group). Intraindividual differences in intraindividual change were further
evaluated by examining the significance of the random effects from the latent growth
models. The random effects are the variances and covariances that describe individual
variability around the fixed effects (i.e. intercept, slope, quadratic). Individual differences
in intraindividual change would be identified if the random effects (i.e. the variance
around the intercept, slope, and quadratic) were significant (see Singer & Willet, 2003).
It is important to note that the analyses in this study were phenotypic, making it
necessary to adjust for the correlation between twins. In the PROC MIXED data step,
pair dependency was accounted for by specifying random effects of growth parameters
within and between twin pairs (i.e. adding two RANDOM lines in the data step and
specifying TYPE=UN for an unstructured covariance matrix parameterized by the within
and between pair variances and covariances). As a conservative check, LGMs were re-
analyzed using a random selection of 1 twin from each pair (Sample A) and then again
20
using the other twin from the pair (Sample B). As the pattern of results did not change,
results from the total sample (adjusted for twinness) are presented.
Growth models were fit in a stepwise fashion. First, a no-growth model (e.g.
intercept-only model) was fit; age is not included in this model. Thus, the fixed effect
intercept is the grand mean across all measurements and the random effects are the
within- and between-person variance around the intercept plus error. This model is
critical for use in comparing successive models via the chi-square difference test. Next,
age centered at 65 was added into the model. This model is considered the linear growth
model (Equation 1a) where,
Equation 1a: OPEN
ij
= γ
0
+ γ
1
((Age
ij
–65)/10) + δ
0i
+ δ
1i
((Age
ij
– 65)/10) + ε
ij
OPEN
ij
represents a openness score for the ith individual at measurement point j; γ
0
reflects the average intercept at age 65; γ
1
represents the linear rate of change by decade;
AGE
ij
is the ith individual’s age at measurement point j; δ
0i
and δ
1i
reflect the ith
individual’s deviations from the average intercept and slope respectively, and ε
ij
reflects
the deviation of the ith individual’s score at measurement point j from their expected
linear trajectory (i.e. measurement error).
Next, a quadratic function was added to further characterize the change over age
(Equation 1b). This model estimates both linear and nonlinear growth (e.g. additional
acceleration) over age where,
Equation 1b: OPEN
ij
= γ
0
+ γ
1
((Age
ij
– 65)/10) + γ
2
((Age
ij
– 65)/10)
2
+ δ
0i
+ δ
1i
((Age
ij
– 65)/10) + δ
21
((Age
ij
– 65)/10)
2
+ ε
ij
OPEN
ij
represents an openness score for the ith individual at measurement point j; γ
0
reflects the average intercept at age 65; γ
1
represents the linear rate of change at age 65
21
by decade; AGE
ij
is the ith individual’s age at measurement point j; γ
2
represents the
quadratic rate of change in openness per decade; δ
0i
, δ
1i
and δ
2i
reflect the ith
individual’s deviations from the average intercept, slope, and quadratic parameters
respectively, and ε
ij
reflects the deviation of the ith individual’s score at measurement
point j from their expected linear trajectory.
Finally, centered/effect-coded covariates of sex, education, SRH, ADL, and CVD
were added to the model containing age and age-squared. The addition of covariates
allowed for an examination of individual differences in the sources of stability and
change in openness to experience. Covariates were evaluated individually, as interactions
with age and age-squared, and as interactions with each other and age/age-squared. The
quadratic model equation was expanded to include an individual covariate (Equation 1c)
where,
Equation 1c: OPEN
ij
= γ
00
+ γ
10
((Age
ij
– 65)/10) + γ
20
((Age
ij
–65)/10)
2
+ γ
01
EDUC
i
+
γ
11
(EDUC
i
((AGE
ij
– 65)/10) + γ
21
(EDUC
i
((Age
ij
– 65)/10)
2
) + δ
0i
+
δ
1i
((Age
ij
– 65)/10) + δ
21
((Age
ij
– 65)/10)
2
+ ε
ij
OPEN
ij
represents an openness score for the ith individual at measurement point j;
EDUC
i
represents the ith individual’s education score (centered at its mean); AGE
ij
is the
ith individual’s age at measurement point j; γ
00
reflects the average intercept at age 65
and average education score; γ
10
represents the linear rate of change at age 65 at the
average education score by decade; γ
20
represents the quadratic rate of change at the
average education score by decade; δ
0i
, δ
1i
and δ
2i
reflect the ith individual’s deviations
from the average intercept, slope, and quadratic parameters respectively, and ε
ij
reflects
22
the deviation of the ith individual’s score at measurement point j from their expected
linear trajectory.
A stepwise procedure was used to evaluate study covariates and model fit.
Covariates were first examined separately and then the final model was built using the
significant covariates. The fit of each model was evaluated using the chi-square
difference test. Models were fit for the total sample, males and females separately as well
as by each age group separately.
CHAPTER 2 RESULTS
Overall descriptives for longitudinal openness to experience by measurement
occasion for the total sample and by sex separately are presented in Table 3. For both
males and females, openness scores were found to be strongly correlated across all time
points (see Table 4). The weakest correlation was between measurements at waves 1 and
6 (Males, r=0.62; Females, r=0.59). In terms of study covariates, education was strongly
correlated with openness across all measurement occasions such that higher education
was associated with higher openness scores. Age at study entry was correlated with
openness such that older participants had lower openness scores. SRH and ADL were
associated with some but not all openness measurements. CVD was not associated with
openness at any measurement occasion for males, and only weakly negative associations
with two measurement points for females.
Results from the four measures of variability, indexing intraindividual change
across measurement occasions, indicated that the average maximum range of openness
23
scores was 4.5 points (median=4), the average absolute sum of change across
measurement points was 7.3 points (median=6). Using the method specified by Dudek
(1979), of the 1615 individuals with 2 or more measurement points, 23% (22% of males;
24% of females) increased, and 27% (23% of males; 31% of female) decreased from their
initial openness score by four or more points - greater than the predicted level of
variability. The annual rate of change indicated that yearly change averaged to -0.05
points in openness (-0.04 for males; -0.06 for females), suggesting that individual
fluctuation averaged to close to zero. However, this amount of yearly change established
a projected decline of about a half a point in openness per decade.
Longitudinal openness data from a randomly selected sample of 1 individual from
each twin pair (Sample A) are presented in Figures 3 (males) and 4 (females). The data
suggested that openness is initially fairly stable (in mean-level) and then declined after
approximately age 60 for females, and after 70 for males. Further, the decline appeared
steeper for females than for males. To examine the hypothesized differences in decline in
openness according to age group, descriptive statistics for openness are presented in
Table 5. Age groups differed in the pattern of openness such that the oldest participants
declined in mean level openness across almost all measurement occasions, the middle age
group started to decline in openness beginning with the fourth measurement occasion,
and the youngest participants did not show any decline across the measurement occasions.
Further, graphs of longitudinal openness data by age group suggest that the youngest and
middle groups seemed to remain stable until age 60 after which decline is present but
24
variable, whereas the oldest group was observed to have a steeper linear decline across
the entire age range during which they participated (see Figures 5-7).
With respect to potential sex differences in baseline openness and variability,
results from analyses of variance indicated no significant sex differences in baseline level
of openness, suggesting that any sex differences in openness longitudinally would likely
be associated with differences in rate of change rather than level. In terms of the created
indices of individual fluctuation, males and females were significantly different on both
overall range and absolute change in openness such that females had significantly greater
fluctuation on both measures. Examining the three age groups, analysis of variance
indicated a significant difference between the three age groups in baseline level of
openness. Post hoc Tukey comparisons revealed that the oldest age group had
significantly lower baseline openness scores compared to the middle and youngest age
groups, but there was no significant difference between the middle and youngest age
group. In terms of fluctuation, there was a significant difference between the age groups
in the amount of absolute change across measurement occasions (but no significant
differences in range). A post hoc Tukey test revealed that the oldest participants had
significantly greater variability than either the middle or youngest age group. The middle
and youngest participants were not significantly different from each other. For both sex
and age group these differences in fluctuation remained significant after adjusting for the
number of openness measurement occasions completed.
Next, latent growth models were built in a step-wise fashion to analyze the
trajectory of openness and to examine whether individual differences in level or slope
25
could be explained by study covariates. Initial graphs of longitudinal openness suggested
that openness did not decline until after age 60 for females and after 70 for males.
Multiple centering ages were evaluated and age centered at 65 best captured the change
point in openness across the different samples. Intercept-only (no growth), linear, and
quadratic models were compared and the quadratic model provided the best fit (see Table
6). The no-growth model served as both the comparison model and identified the
proportion of variance attributable to within-person and between person variance a no-
growth (intercept-only) model for openness was fit to the data. Model estimates indicated
that 70% of the total variance in level of openness was attributable to between-person
differences and the remaining 30% was accounted for by within-person differences and
measurement error - indicating individual differences in interindividual change.
Study covariates were first evaluated individually in the quadratic LGM and fits
were compared. Education, age group, and SRH were each significantly associated with
level of openness and improved model fit; whereas sex and CVD were not associated
with level or slope and did not improve fit. Significant interactions were found between
age group *slope and ADL*quadratic. CVD was dropped from the analysis at this stage.
Sex was retained because there were remaining questions about the potential for sex to
interact with other study covariates in the subsequent analyses.
To build the final model, sex was added to the quadratic LGM and, as noted
above, had no effect on level or slope of openness. Education was then added and had a
large effect on level of openness such that greater education was associated with a greater
level of openness at age 65. This difference in level was maintained over time such that
26
there was no interaction between education and linear slope or quadratic components.
There were no significant interactions between sex, education and age group. Next, SRH
and ADL were added to the model one by one. Neither ADL nor SRH were significantly
associated with level or slope and neither interacted with the other study covariates. The
model fit was not significantly improved by adding these health-related covariates. Figure
8 presents the trajectory of openness to experience adjusting for sex and education.
Age group was then added as a class variable to better describe the openness
trajectory - the young and middle groups were not significantly different from the oldest
group on level of openness. However, the youngest group was significantly different
from the oldest group on slope, such that earlier born participants had a steeper decline in
openness than later born participants. There was no difference between the middle and
old age groups on level or slope. When age group was added to the model the slope and
quadratic estimates were reduced to nonsignficance (see Table 7). Thus, as expected, age
accounted for the decline in openness.
The LGM model was re-analyzed by sex and age group independently. The
centering age of 65 was retained for all models. Despite no significant mean differences
in level or slope based on sex, there had been initial evidence that males and females
might differ in variability in openness with age. However, results presented in Table 8
suggest no obvious differences in the proportion of within- vs. between-person variance.
To further understand the trajectory of openness based on age, age groups were modeled
separately, adjusting for sex and education (See Table 9). Results revealed that the
youngest participants exhibited no significant linear or quadratic change per decade; the
27
middle age participants showed significant linear and additional nonlinear decline per
decade, and the oldest participants showed substantial linear decline, but no acceleration
in decline. Sex was not associated with level or slope in any age group. Education was
significantly associated with the level of openness across all age groups, but not
associated with slope. Similar to previous models, almost all of the variance was
explained by individual differences around the intercept - there was no significant
variance for the slope or quadratic parameters. Figure 9 presents the different trajectories
for openness over age based on the best fitting model. This graph illustrates how the age
groups map onto the predicted trajectory of openness and further highlights the notable
decline in openness in older adulthood.
CHAPTER 2 DISCUSSION
The purpose of this study was to identify the longitudinal trajectory of openness
to experience and to examine potential sources of stability and change across adulthood.
Whether or not personality is stable or exhibits flexibility over the lifespan is of
continued debate. Recent research has suggested that intraindividual change in
personality traits is an important part of continued personality development (see Roberts,
Wood, & Caspi, 2008). Given that personality has been identified as an important
predictor of major life outcomes (Roberts et al., 2007), questions surrounding change in
personality are relevant to lifespan development and aging.
A goal of this study was to describe intraindividual differences in intraindividual
change in openness over age. A notable finding was the substantial amount of fluctuation
28
in openness during adulthood. The total amount of change in openness scores between
measurement occasions was greater than the standard error of prediction in 50% of the
sample. This change was also almost always reversible such that a yearly rate of decline
was quite small. The extent of fluctuation was also reflected in the significant within-
person variability identified in the no-growth LGM, suggesting a notable amount of
individual variability across the lifespan. Females and individuals in the oldest age group
had significantly more variability over measurement occasions compared to males and
younger individuals. Further, the oldest age group had significantly lower baseline
openness scores compared to both the middle and youngest age groups. Age group might
have been confounded by the opportunity for more measurement points for some
participants; however, these significant patterns remained after adjusting for the number
of measurements completed. Together, these findings suggested notable individual
differences in intraindividual change in openness across advancing age and his pattern
was consistent with findings from recent investigations of adult. However, an explanation
for these fluctuations remains unexplained.
In terms of the longitudinal models, it was hypothesized that openness would
decline with age, demographic variables would explain individual differences in intercept
and health-related factors would explain individual differences in change in openness.
The overall model results for the total sample suggested small but significant declines in
openness at a rate of approximately 0.75 per decade or 1.5 points per 20 years, with the
majority of decline occurring after age 60. As expected, education was a predictor of
individual differences in level of openness. However, neither gender nor health-related
29
covariates explained the rate of change in openness. That is, there was no evidence that
the decline in openness was influenced by individual differences in self-perceived health,
level of disability, or the presence of cardiovascular disease. Overall, there was no
significant variance on the slope or quadratic parameters of the growth model suggesting
limited individual differences in the rate of decline. Thus, the trajectory of longitudinal
openness was not moderated by selective decline due to disability or disease as measured
by this study.
Previous research indicated that other life events such as divorce and loss of a
spouse were important predictors of differences in change in extraversion and
neuroticism (Mroczeck & Spiro, 2003). It may be that social life events introduced
around age 65 (e.g. retirement) may be more relevant than health related factors in
predicting differences in personality change. To this point, Finkel, Reynolds, Andel,
Pedersen, and Gatz (2009) found that individuals who had an occupation requiring a high
level of complexity declined on tasks of spatial cognition following retirement. Such
findings may be relevant to openness because trait openness has often been characterized
as a cognitive and emotional engagement with experience and has also been found to be
associated with intellectual ability (Booth, Schinka, Brown, Mortimer, & Borenstein,
2006; Schaie, Willis, & Caskie, 2004). It is possible that the decline in openness might be
accounted for by other environmental factors such as cognitive decline or retirement.
We used age group to further describe the pattern of change in openness. Model
results indicated no significant decline in the youngest age group; the middle age group
declined approximately two-thirds of a point per decade; and the oldest participants
30
declined almost one and one-half points. This decline in openness began around age 60
and showed the greatest rate of decline after age 75 (see Figure 9). Overall, openness
seemed to follow a normative pattern of decline, such that older participants were
changing in the same direction. The finding that openness declines were small and
focused in older adulthood was consistent with the literature (McCrae & Sutin, 2009;
Pedersen & Reynolds, 1998; Sharp et al., 2010; Strivasta, John, Gosling, & Potter, 2003).
Previous SATSA research had suggested sex differences such that males
remained stable while females exhibited additional accelerated decline in openness
(Sharp et al., 2010). However, the current LGM analysis did not reveal significant sex
differences in differences in level or rate of change in openness. The differences in
findings between the current study and Sharp et al. (2010) are likely due to the previous
study having a smaller sample of participants, specifically only those with both cognitive
testing and an openness measurement (N=857), the sample being restricted to individuals
who were 65 and older, and having fewer measurement occasions of openness available
for analysis - particularly important because the greatest amount of decline occurred as
individuals moved into older adulthood.
Why does openness decline in old age? Given the definition of openness, changes
in engagement might be related to declines in openness. In the context of a
developmental lifespan model of aging, declines in openness may be related to a model
of selection, optimization and compensation (see Baltes & Baltes, 1990; Freund & Baltes;
2007). As such, declines in openness might be an adaptive response to aging processes
that serve to maximize functioning in other domains. In particular, decline in openness
31
could be an optimization and compensation process as described by the socioemotional
selectivity theory (SST, see Carstensen, 2006; Carstensen, Mickels & Mather, 2006). SST
suggests that as an individual enters older adulthood and time-to-death becomes more
relevant, individuals change their goals and outlook on life. When individuals are
younger, importance is placed on goals associated with gathering of information,
experiencing novelty, and expanding breadth of knowledge. In contrast, as the time left
prior to death becomes more salient, individuals are less interested in new experiences
and tend to reduce and focus their interests. Because openness is a trait describing the
tendency to seek out new experiences and information, the normative pattern of decline
in openness may be a reflection of a systemic change of goals in older adults.
However, another possibility is that the decline in openness is not adaptive but
simply a reflection of nearness to death. Research in the aging literature has pointed to
the importance of examining the pattern of declines in functioning as an individual
approaches death. The years preceding death that are accompanied by functional declines
across multiple domains have been identified within a theory of terminal decline (see
Berg, 1996). However, very few studies have examined the relationship between
personality and terminal decline. The current study provided initial evidence that declines
in openness may be related to death, specifically that openness showed the greatest
declines after age 75 (the oldest age group), and that the pattern of participant attrition
suggested that individuals who died had a lower openness score than individuals who
remained in the study. This difference based on death status calls for a survival analysis
to more closely examine whether declines in openness might be related to risk for death.
32
Finally, proponents of the stability hypothesis have suggested that fluctuations or
patterns of change in personality at the individual level would be accounted for by
genetic influences (Costa & McCrae, 1997; McCrae, Costa, Ostendorf, Angleitner, &
Hrebickova, 2000). This would suggest that the normative pattern of mean-level decline
in openness is being driven by the same genetic forces across individuals. An analysis of
the pattern of genetic and environmental contributions to longitudinal openness is
necessary to better understand the decline in openness with advancing age.
The current study had a number of strengths including the longitudinal design, a
large sample size, up to 6 measurement points of openness, and the use of contemporary
statistical techniques. Generally, the results were that openness declined approximately
1.5 points over twenty years. At best, the rate of change is a third of a standard deviation.
While this might be of arguable clinical significance, the effect qualifies as a small effect
size in psychology (Meyer et al., 2001).Further, if the total individual level is combined
with the linear decline, the total change is greater than one standard deviation – a large
effect in psychology. Results indicated that sex, ADL, SRH, and CVD were not soures of
individual differences in decline in openness. However, it is certainly possible that other
measures of health or important life events might be associated with decline in openness.
Finally, as with any longitudinal study, attrition was an important factor in this study and
the sample size was notably smaller in the sixth wave of data compared to the previous
five. Additionally, openness scores were observed to be lower in individuals who died or
dropped out. Thus, more research is needed to identify potential factors driving the
decline in openness with age.
33
This study advanced the literature by examining stability and change in openness
to experience across the lifespan. Openness to experience was found to have notable
individual differences in individual change that were independent of age. Overall,
openness exhibited significant linear and nonlinear decline after age 65. Males and
females were not significantly different in level or rate of decline. Education was
important in predicting individual differences in level of openness. Health related
variables (SRH, ADL, CVD) were not associated with level or change in openness with
age. Age group best described the individual differences in decline in openness. Overall,
openness declined between about one and a half points over twenty years, with most of
the decline occurring after age 65. This decline was small in magnitude but significant
and may be important in the context of aging related processes in older adults.
34
Table 1. Number of openness measurements completed for the total sample and by males
and females with openness means and standard deviations at baseline
Number of Total Openness Measurements
Sample 1 2 3 4 5 6
Total 332 263 298 434 221 399
Open Mean 16.59 16.97 17.67 17.63 18.25 18.41
Open std. 4.84 4.06 4.00 3.67 3.60 3.50
Males 125 112 123 203 91 161
Open Mean 16.90 17.58 17.98 17.61 18.34 18.24
Open std. 4.94 3.66 4.09 3.89 4.38 3.88
Females 207 151 175 231 130 238
Open Mean 16.41 16.79 18.01 18.00 18.45 18.29
Open std. 4.79 5.00 4.39 4.00 4.38 3.98
Note. Std=Standard Deviation.
35
Table 2. Attrition between openness measurement occasions with means and standard
deviations for openness at last measurement occasion
Attrition Status Q1-Q2 Q2-Q3 Q3-Q4 Q4-Q5 Q5-Q6
Died before next wave
N 69 80 79 400 77
Mean 16.54 16.41 16.11 17.18 17.16
SD 4.85 4.18 4.40 4.73 5.03
Dropped out
N 88 115 106 200 123
Mean 17.00 16.34 16.94 17.24 17.95
SD 4.63 4.54 5.04 4.39 4.35
a. Died later 67 87 80 86 8
Mean 16.63 15.33 16.60 16.84 16.50
SD 4.76 4.18 5.20 3.91 6.16
b. Still Alive 21 28 26 114 115
Mean 18.19 19.46 18.00 17.54 18.03
SD 4.04 4.24 4.47 4.72 4.22
Tot. Lost to follow-up 157 195 185 600 200
Note. Q1= First possible wave of openness measurement in 1984; Q2 in 1987; Q3 in
1990; Q4 in 1993; Q5 in 2004; Q6 in 2007. SD= Standard Deviation
36
Table 3. Demographic and variable characteristics of the total sample and for males and
females
Note. SD= standard deviation.
Openness Measurement Wave
Variable Statistic
Q1
(1984)
Q2
(1987)
Q3
(1990)
Q4
(1993)
Q5
(2004)
Q6
(2007)
Total Sample
N 1428 1486 1364 1351 749 610
Mean openness 17.84 17.79 17.90 17.89 18.11 18.15
SD 4.10 4.30 4.40 4.40 4.20 4.20
Mean age 59.02 61.37 63.03 64.58 69.44 71.45
SD 13.50 13.50 12.80 12.90 10.90 10.30
Age range 26-92 29-97 32-96 35-93 46-103 49-97
Males
N 630 637 570 572 303 239
Mean openness 17.92 17.69 17.95 17.93 18.05 18.00
SD 3.90 3.99 4.10 4.09 4.15 4.19
Mean age 58.20 60.67 62.37 63.56 69.31 71.14
SD 13.10 12.96 12.10 12.16 10.40 9.72
Age range 27-90 30-91 33-88 36-91 47-94 50-94
Females
N 798 849 794 779 446 371
Mean openness 17.77 17.86 17.86 17.81 18.15 18.00
SD 4.20 4.50 4.60 4.63 4.25 4.15
Mean age 59.66 61.89 63.50 65.33 69.53 71.66
SD 13.8 13.92 13.30 13.34 11.23 10.70
Age range 26 -92 29-96 32-93 35-93 46-103 50-97
37
Table 4. Correlations between openness measurement waves and study covariates by
males and females separately
Note. *p<.05. Correlations for males are presented within the bottom diagonal half.
Correlations for females are presented within the top diagonal half. Age= study entry age;
EDUC= education; SRH=self-rated health; ADL=activities of daily living;
CVD=cardiovascular disease.
Variable 1 2 3 4 5 6 7 8 9 10 11
1. Open1 .68* .64* .70* .63* .59* -.14* .26* .08* .10* -.03
2. Open2 .73* .72* .71* .65* .66* -.17* .26* .11* .11* -.08*
3. Open3 .71* .75* .74* .68* .60* -.13* .31* .07* .08* -.08*
4. Open4 .68 * .70* .71* .70* .65* -.16* .30* .08* -.01 -.01
5. Open5 .64* .70* .68* .70* .74* -.11* .35* .05 -.02 .02
6. Open6 .62* .72* .67* .68* .74* -.14* .35* .10 .04 -.01
7. Age -.16* -.17* -.12* -.21* -.09 -.22* -.40* -.18* -.25* .29*
8. EDUC .30* .32* .26* .32* .30* .36* -.25* .19* .12* -.12*
9. SRH .11* .18* .11* .11* .04 .14* -.16* -.20* .36* -.35*
10. ADL .07 .08* .03 .01 .10 .09 -.21* -.06* .20* -.20*
11. CVD -.07 -.05 .01 -.01 .10 .05 -.20* .14* -.37* -.08*
38
Table 5. Participation and means and standard deviations across openness measurement
occasions by age group
Age Group Open 1 Open 2 Open 3 Open 4 Open 5 Open 6
Oldest
N 540 530 422 359 61 28
Mean Open 17.23 16.99 17.22 16.90 16.85 16.89
SD 4.3 4.4 4.7 4.5 4.9 3.9
Mean Age 72.27 75.22 77.00 79.84 88.42 91.16
SD 5.4 5.3 4.3 4.2 2.9 2.1
Middle
N 487 517 510 503 287 221
Mean Open 18.04 18.06 18.02 17.85 17.87 17.57
SD 3.9 4.2 4.3 4.3 4.3 4.2
Mean Age 58.79 61.68 64.76 67.62 77.22 80.49
SD 4.0 4.1 4.1 4.0 4.0 3.9
Young
N 401 439 432 489 400 361
Mean Open 18.40 18.44 18.41 18.57 18.48 18.60
SD 3.8 4.1 4.2 4.3 3.9 4.1
Mean Age 41.45 44.27 47.33 50.24 60.97 64.39
SD 6.5 6.4 6.5 6.5 6.2 6.3
Note. Oldest cohort born 1891-1919; Middle cohort born 1920-1933; Young cohort born:
1934-1958. SD= standard deviation.
39
Table 6. Initial growth model estimates (standard errors) of openness to experience for
the total sample
Model Parameters Model 1 Model 2 Model 3
Means
Intercept@65 β
0
17.64 (0.10)* 17.72 (0.10)* 17.92 (0.11)*
Slope β
1
- -0.32 (0.05)* -0.41 (0.05)*
Quadratic β
2
- - -0.29 (0.05)*
Total Variances / Covariances
Intercept σ
0
2
13.33 12.44 13.28
Slope σ
1
2
- 0.62 0.82
Quadratic σ
2
2
- - 0.35
Int.-Slope σ
01
2
- 0.44 -0.12
Int.-Quad. σ
02
2
- - -0.88
Slope-Quad. σ
12
2
- - 0.36
Residual σ
u
2
5.69 (0.11)* 5.29 (0.12)* 5.05 (0.11)*
Fit Statistics
-2LL 36041.1 35935.0
†
35841.4
†
Parameters 4 9 16
Note. * p<0.05. Means are intercept at age 65 and change by decade (slope & quadratic).
Model 1 = No growth (Intercept-only); Model 2= Linear; Model 3 = Quadratic.
Total variances and covariances are the sum of within-twin pair and between-twin pair
variances and covariances. -2 LL= -2* Log Likelihood.
†
Difference in fit between Model 1 and 2 is significant, p<0.0001.
†
Difference in fit between Model 2 and 3 is significant, p<0.0001
40
Table 7. Latent growth model estimates (standard errors) with covariates for the total
sample
Model Parameters Model 1 Model 2 Model 3
Means
Intercept@65 β
0
17.93 (0.11)* 17.88 (0.10)* 17.99 (0.22)*
Slope β
1
-0.44 (0.05)* -0.37 (0.05)* -0.13 (0.15)
Quadratic β
2
-0.28 (0.05)* -0.28 (0.05)* -0.11 (0.12)
Sex 0.08 (0.22) 0.28 (0.21) 0.64 (0.41)
Sex*Slope -0.04 (0.10) 0.007 (0.10) 0.36 (0.28)
Sex*Quadratic -0.18 (0.10) -0.16 (0.11) 0.08 (0.23)
Education - 1.36 (0.11)* 1.57 (0.19)*
Education*Slope - 0.03 (0.06) 0.25 (0.14)
Education*Quadratic - -0.04 (0.05) 0.07 (0.10)
Sex*Education*Slope - 0.09 (0.11) 0.05 (0.11)
Sex*Education*Quadratic - 0.03 (0.10) 0.006 (0.11)
Young Age Group - - 0.19 (0.36)
Middle Age Group - - 0.04 (0.27)
Young *Slope - - -1.13 (0.45)*
Middle*Slope - - -0.12 (0.18)
Young*Quadratic - - 0.38 (0.36)
Middle*Quadratic - - -0.28 (0.21)
Sex*Young*Slope - - -0.11 (0.90)
Sex*Middle*Slope - - -0.31 (0.35)
Sex*Young*Quad - - -0.62 (0.72)
Sex*Middle*Quad - - -0.28 (0.41)
Education*Young*Slope - - -0.32 (0.57)
Education*Middle*Slope - - -0.31 (0.18)
Education *Young*Quad - - -0.32 (0.43)
Education*Middle*Quad - - 0.15 (0.22)
Total Variance / Covariance
Intercept σ
0
2
13.02 11.51 11.43
Slope σ
1
2
0.82 0.81 0.78
Quadratic σ
2
2
0.37 0.36 0.33
Int.-Slope σ
01
2
-0.09 -0.05 -0.06
Int.-Quad. σ
02
2
-0.90 -.79 -0.71
Slope-Quad. σ
12
2
0.36 0.36 0.34
Residual σ
u
2
4.98 (0.12)* 4.99 (0.12)* 4.98 (0.12)*
Fit Statistics: -2LL (parms) 33384.3 (19) 33224.7 (25)
†
33200.6 (43)
†
Note. * p<0.05. Intercept age = 65. Slope is by decade. Two-way interactions were not
significant and are not shown. Total variances and covariances are the sum of within-twin
pairs and between-twin pairs. -2 LL= -2* Log Likelihood.
†
Difference in fit between
Model 1 and 2 is p<0.0001;
†
Difference in fit between Model 2 and 3 is p=0.02.
41
Table 8. Model estimates (standard errors) by sex
Model Parameters
Males Females
Means
Intercept@65 β
0
17.72 (0.15)*
17.64 (0.29)*
18.03 (0.14)*
18.56 (0.29)*
Slope β
1
-0.37 (0.07)*
-0.04 (0.20)
-0.36 (0.07)*
0.05 (.20)
Quadratic β
2
-0.18 (0.07)*
-0.07 (0.16)
-0.36 (0.07)*
-0.08 (0.16)
Education
1.29 (0.15)*
1.55 (0.27)*
1.45 (0.16)*
1.58 (0.26)*
Education*Slope
-0.002 (0.07)
0.14 (0.20)
0.08 (0.08)
0.35 (0.19)
Education*Quadratic
-0.05 (0.07)
-0.005 (0.14)
-0.03 (0.07)
0.12 (0.14)
Young Age Group
-
0.45 (0.51)
-
-0.05 (0.51)
Middle Age Group
-
0.28 (0.38)
-
-0.21 (0.38)
Old Age Group (ref.)
-
-
-
-
Young *Slope
-
-1.18 (0.64)
-
-1.21 (0.63)
Middle*Slope
-
-0.06 (0.24)
-
-0.28 (0.26)
Old*Slope (ref.)
-
-
-
-
Young*Quadratic
-
0.65 (0.52)
-
0.12 (0.48)
Middle*Quadratic
-
-0.27 (0.29)
-
-0.41 (0.28)
Old*Quadratic (ref.)
-
-
-
-
Educ*Young*Slope
-
-0.18 (0.70)
-
-0.52 (0.91)
Educ*Middle*Slope
-
-.19 (0.25)
-
-0.43 (0.25)
Educ*Old*Slope (ref.)
-
-
-
-
Educ *Young*Quad
-
-0.33 (0.54)
-
-0.22 (0.68)
Educ*Middle*Quad
-
0.34 (0.30)
-
-0.10 (0.32)
Educ*Old*Quad (ref.)
-
-
-
-
Variance / Covariance
Intercept σ
0
2
10.93 10.93 11.97 11.86
Slope σ
1
2
0.66 0.63 0.91 0.86
Quadratic σ
2
2
0.22 0.19 0.41 0.38
Int.-Slope σ
01
2
0.04 0.07 -0.13 -0.17
Int.-Quad. σ
02
2
-0.82 -0.77 -0.25 -0.67
Slope-Quad. σ
12
2
0.22 0.19 0.41 0.40
Residual σ
u
2
4.36 (0.16)* 4.35 (0.15)* 5.46 (0.17)* 5.46 (0.17)*
Fit Statistics: -2LL
(Parms)
13597.7
(18)
13586.0
(31)
19579.2
(19)
19562.8
(31)
Note. * p<0.05. Two-way interactions between covariates were not significant and are
not shown. Variances and covariances are the sum of within-twin pair and between-twin
pair variances and covariances. -2 LL= -2* Log Likelihood; Parms = parameters.
42
Table 9. Final model estimates (standard errors) by age group
Model Parameters
Age Group
Young Middle Old
Means
Intercept@65 β
0
17.98 (0.19)* 18.04 (0.17)* 18.24 (0.21)*
Slope β
1
-0.14 (0.15) -0.29 (0.10)* -1.34 (0.47)*
Quadratic β
2
-0.12 (0.12) -0.35 (0.17)* -0.34 (0.38)
Sex 0.71 (0.39) 0.23 (0.35) 0.29 (0.62)
Sex*Slope 0.39 (0.30) 0.11 (0.20) 0.03 (0.95)
Sex*Quadratic 0.10 (0.24) -0.21 (0.33) -0.32 (0.75)
Education 1.56 (0.18)* 1.24 (0.19)* 1.38 (0.40)*
Educ*Slope 0.24 (0.14) -0.07 (0.11) -0.10 (0.61)
Educ*Quadratic 0.07 (0.10) -0.18 (0.20) -0.20 (0.47)
Sex*Educ*Slope 0.23 (0.28) -0.08 (0.22) -0.27 (1.23)
Sex*Educ*Quad 0.11 (0.20) -0.24 (0.39) 0.30 (0.94)
Variance/Covariance
Intercept σ
0
2
10.22 11.48 9.55
Slope σ
1
2
1.36 0.57 0.34
Quadratic σ
2
2
0.51 0.97 0.45
Int.-Slope σ
01
2
-0.49 0.55 2.92
Int.-Quad. σ
02
2
0.63 -1.74 -3.39
Slope-Quad. σ
12
2
0.63 -0.64 0.54
Residual σ
u
2
3.94 (0.15)* 5.13 (0.20)* 6.24 (0.30)*
Fit Statistics: -2LL
(parameters)
11329.9 (25) 11932.7 (25) 9843.3 (25)
Note. * p<0.05. Youngest were born 1933-1958; Middle were born 1920-1932; Oldest
were born 1891-1919. Means give intercept at age 65. Linear and quadratic change is by
decade. Total variances and covariances are the sum of within-twin pair and between-
twin pair variances and covariances. -2 LL= -2* Log Likelihood
43
Figure 1. Participant attrition with mean and standard deviations for openness to
experience
44
Figure 2. Quadratic growth model with covariate
Note. The squares represent measured variables, the circles represent latent variables, the
circles within squares represent data that are potentially available at each measurement
occasions; Covariate* = standardized score of the covariate (e.g. Education); I =
intercept; S = slope; Q=Quadratic; I* = standardized score for the intercept; S* =
standardized score for the slope; Q*=standardized score for the quadratic; M
i
= mean of
the intercept; M
s
= mean of the slope; M
q
= mean of the quadratic; M
c
= mean of the
covariate; R
is
= correlation between the intercept and slope; R
iq
= correlation between the
intercept and quadratic; R
sq
= correlation between slope and quadratic; R
ic
= correlation
between the covariate and the intercept; R
sc
= correlation between the slope and the
covariate; R
qc
= correlation between the quadratic and the covariate; D
i
= deviation from
the intercept; D
s
= deviation from the slope; D
q
= deviation from the quadratic; D
c
=deviation from the covariate mean; B1-B6 = linear age basis coefficients; B12-B62=
quadratic age basis coefficients; u
1
-u
6
= random components from the openness
measurements; D
u
= a constant deviation from the openness scores.
45
Figure 3. Plot of longitudinal openness scores for sample of male (a) and female (b)
participants with up to 6 measurement occasions
(a)
(b)
Note. Graphs show randomly selected twin from each pair (Twin Sample A).
46
Figure 4. Plot of longitudinal openness trajectories for the youngest (a), middle (b), and
oldest (c) age groups. Randomly selected 1 twin from each pair (Twin Sample A)
(a)
(b)
(c)
47
Figure 5. Trajectory of openness to experience across the adult lifespan (a) and best
fitting trajectory of openness to experience across the adult lifespan by age group (b)
(a)
6
8
10
12
14
16
18
20
22
24
26
28
30
35 45 55 65 75 85 95
Age
Openness Score
(b)
6
8
10
12
14
16
18
20
22
24
26
28
30
35 45 55 65 75 85 95
Age
Openness
Young
Middle
Old
Note. Age groups (Young, Middle, Old) were modeled separately. Trajectories in both
graphs are adjusted for sex and education. Age was centered 65.
48
CHAPTER 2 REFERNCES
Alwin, D. F. (1994). Aging, personality, and social change: The stability of individual
differences over the adult span. In D. L. Featherman, R. M. Lerner, & M. Perlmutter
(Eds.), Life-Span Development and Behavior (Vol. 12, pp. 135-185). Hillsdale, NJ;
Lawrence Erlbaum.
Baltes, P. B. (1987). Theoretical propositions of life-span development psychology: on
the dynamics between growth and decline. Developmental Psychology, 23, 611-626.
Baltes, P. B., & Nesselroade, J. R. (1973). The developmental analysis of individual
differences on multiple measures. In J. R. Nesselroade & H. W. Reese (Eds.), Life-
span developmental psychology: Methodological issues (pp. 219-251). New York:
Academic Press.
Baltes, P. B. (1997). On the incomplete architecture of human ontogeny. American
Psychologist, 52, 366–380.
Baltes, P. B., Lindenberger, U., & Staudinger, U. M. (1998). Life-span theory in
developmental psychology. In W. Damon & R. M. Lerner (Eds.), Handbook of Child
Psychology (5th ed., Vol. 1, pp. 1029–1143). New York: Wiley.
Baltes, P. B., & Baltes, M. M. (1990). Psychological perspectives on successful aging:
The model of selective optimization with compensation. In P. B. Baltes & M. M.
Baltes (Eds.), Successful Aging: Perspectives from the Behavioral Sciences (pp. 1–
34). Cambridge, U.K.: Cambridge University Press.
Berg, S. (1996). Aging, behavior, and terminal decline. In J.E. Birren & K. W. Schaie
(Eds.), The Handbook of the Psychology of Aging. (4th ed., pp. 323-337). San Diego:
Academic Press.
Bergeman, C. S., Chipuer, H. M., Plomin, R., Pedersen, N. L., McClearn, G. E.,
Nesselroade, J. R., Costa, P. T. Jr., & McCrae, R. R. (1993). Genetic and
environmental effects on openness to experience, agreeableness, and
conscientiousness: An adoption/twin study. Journal of Personality, 61, 159-179.
Booth, J. E., Schinka, J. A., Brown, L. M., Mortimer, J. A., & Borenstein, A. R. (2006).
Five-factor personality dimensions, mood states, and cognitive performance in older
adults. Journal of Clinical and Experimental Neuropsychology, 28, 676–683.
Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to
assessing change. Psychological Bulletin, 101, 147-158.
49
Carstensen, L. L. (2006). The influence of a sense of time on human development.
Science, 312, 1913–1915.
Carstensen, L. L., Mikels, J. A., & Mather, M. (2006). Aging and the intersection of
cognition, motivation and emotion. In J. Birren & K. W. Schaie (Eds.), Handbook
of the Psychology of Aging (6th ed., pp. 343–362). San Diego, CA: Academic
Press.
Caspi A. & Roberts, B. W. (2001). Personality development across the life: The argument
for change and continuity. Psychological Inquiry, 12, 49-66.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and
change. Annual Review of Psychology, 56, 453-484.
Cederlof, R., Friberg, L., & Lundman, T. (1977). The interactions of smoking,
environment and heredity and their implications for disease etiology. Acta Medica
Scandinavia, 612, 1-128.
Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory Manual. Odessa,
FL: Psychological Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992a). Revised NEO personality inventory (NEO PI-R)
and NEO five-factor inventory (NEO-FFI). Odessa, Florida: Psychological
Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992b). Multiple uses for longitudinal personality data.
European Journal of Longitudinal Research and Personality, 6, 85-102.
Costa, P. T. & McCrae, R. R. (1994). Set like plaster? Evidence for the stability of adult
personality. In T. F. Heatherton & J. L. Weinberger (Eds.), Can Personality
Change? (pp. 21-40). Washington, DC: American Psychological Association.
Costa, P. T., & McCrae, R. R. (1997). Longitudinal stability of adult personality. In R.
Hogan, J. A. Johnson, & S. R. Brigss (Eds.), Handbook of Personality Psychology
(pp. 269-290). San Diego: Academic Press.
Dudek, F. J. (1979). The continuing misinterpretation of the standard error of
measurement. Psychological Bulletin, 86, 335-337.
Elder, G.H. (1975). Age differentiation and life course. Annual Review of Sociology, 1,
65-190.
Elder, G. H. (1998). The life course as developmental theory. Child Development, 69, 1-
12.
50
Finkel, D., Reynolds, C. A., McArdle, J. J., Gatz, M., & Pedersen, N. L. (2003). Latent
growth curve analyses of accelerating decline in cognitive abilities in late adulthood.
Developmental Psychology, 39, 535-550.
Finkel, D., Andel, R., Gatz, M., & Pedersen, N. L. (2009). The role of occupational
complexity in trajectories of cognitive aging before and after retirement. Psychology
& Aging, 24, 563-574.
Fleeson, W., & Jolley, S. (2006). A proposed theory of the adult development of
intraindividual variability in trait-manifesting behavior. In D. Mroczek & T. D. Little
(Eds.), Handbook of Personality Development (pp. 41-59). Mahwah, NJ: Lawrence
Erlbaum Associates.
Freund, A. M., & Baltes, P. B. (2007). Toward a theory of successful aging: Selection,
optimization, and compensation. In R. Fernandez-Ballesteros (Ed.),
Geropsychology: European Perspectives for an Aging World (pp. 239–254).
Cambridge, MA: Hogrefe & Huber.
Fraley, C., & Roberts, B. W. (2005). Patterns of continuity: A dynamic model for
conceptualizing the stability of individual differences in psychological constructs
across the life course. Psychological Review, 112, 60-74;
Harris, J. R., Pedersen, N. L., McClearn, G. E., Plomin, R., & Nesselroade, J. R. (1992).
Age differences in genetic and environmental influences for health from the Swedish
Adoption/ Twin Study of Aging. Journal of Gerontology: Psychological Sciences,
47, 213-220
Helson, R., Kwan, V. S. Y., John, O. P., & Jones, C. (2002). The growing evidence for
personality change in adulthood: Findings form research with personality
inventories. Journal of Research in Personality, 36, 287-306.
Hultsch, D. F., Strauss, E., Hunter, M. A., & MacDonald, S. W. S. (2008). Intraindividual
variability, cognition, and aging. In F. I. M. Craik & T. A. Salthouse (Eds.),
Handbook of Cognition and Aging (3rd ed., pp. 491-556). New York, NY, US:
Psychology Press.
John, O. P., & Strivasta, S. (1999). The Big Five trait taxonomy: history, measurement,
and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of
Personality (pp. 102-138). New York: Guilford.
Krueger, R. F., Johnson, W., & Kling, K. C. (2006). Behavior genetics and personality
development. In D. K. Mroczek & T. D. Little (Eds.), Handbook of Personality
Development (pp. 81-108). Mahwah, NJ: Erlbaum.
51
Lewis, M. (2001). Issues in the study of personality development. Psychological Inquiry,
12, 67-83.
Little, R. T. A. (1995). Modeling the dropout mechanism in repeated-measures studies.
Journal of the American Statistical Association, 90, 1112-1121.
McArdle, J. J. & Anderson, E. (1990). Latent variable growth models for research on
aging. In J. E. Birren & K. W. Schaie (Eds.), The Handbook of the Psychology of
Aging (pp. 21-43). New York: Plenum Press.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual
abilities over the life span. Developmental Psychology, 38, 115-142.
McArdle, J. J., & Hamagami, F. (1992). Modeling incomplete longitudinal and cross-
sectional data using latent growth structural models. Experimental Aging Research,
18, 145-166.
McArdle, J. J., Hamagami, F., Jones, K., Jolesz, F., Kikinis, R., Spiro, A., & Albert, M.
S. (2004). Structural modeling of dynamic changes in memory and brain structure
using longitudinal data from the normative aging study. Journal of Gerontology:
Psychological Sciences, 59, 294-304.
McCrae, R. R. (1994). Openness to experience: Expanding the boundaries of factor V.
European Journal of Personality, 8, 251-272.
McCrae, R. R. & Costa, P. T., Jr. (1990). Personality in Adulthood. New York: Guilford
Press
McCrae, R. R. & Costa, P. T., Jr. (1997). Conceptions and correlates of Openness to
Experience. In R. Hogan, J. A. Johnson, & S. R. Briggs (Eds.), Handbook of
Personality Psychology (pp. 825-847). Orlando, FL: Academic Presss.
McCrae, R. R., & Costa, P. T. (2008). Empirical and theoretical status of the five-factor
model of personality traits. In G. J. Boyle, Mathews, G., & Sakofske, D. H. (Eds.),
The SAGE handbook of personality theory and assessment Vol 1: Personality
theories and models (pp. 273-294). Thousand Oaks, CA, US: Sage Publications, Inc.
McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebícková, M., Avia, M.
D., et al. (2000). Nature over nurture: Temperament, personality, and life span
development. Journal of Personality and Social Psychology, 78, 173-186.
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its
applications. Journal of Personality, 60, 175-175.
52
McCrae, R. R., & Sutin, A. R. (2009). Openness to experience. In, M. R. Leary, & R. H.
Hoyle (Eds.), Handbook of Individual Differences in Social Behavior (pp.257-273).
New York, NY, US: Guilford Press, 2009.
Meyer, G. J., Finn, S. E., Eyde, L. D., Kay, G. G., Moreland, K. L., Dies, R. R. et al.
(2001). Psychological testing and psychological assessment. American Psychologist,
56, 128-165.
Mroczek, D. K., Almeida, D. M., Spiro, A., & Pafford, C. (2006). Modeling
intraindividual stability and change in personality. Mahwah, NJ, US: Lawrence
Erlbaum Associates Publishers.
Mroczek, D. K., & Spiro, A. (2003). Modeling intraindividual change in personality
traits: Findings from the normative aging study. The Journals of Gerontology:
Psychological Sciences and Social Sciences, 58, 153-165.
Mroczek, D. K., & Spiro, A. (2007). Personality change influences mortality in older
men. Psychological Science, 18, 371-376.
Mroczek, D. K., Spiro, A., & Almeida, D. M. (2003). Between- and within- person
variation in affect and personality over days and years: How Basic and applied
approaches can inform one another. Ageing International, 28, 260-278.
Mroczek, D. A., Spiro, A., III, & Griffin, P. W. (2006). Personality and Aging. In J. E.
Birren and K. W. Shaie (Eds.), Handbook of the Psychology of Aging (6th ed., pp.
363-377).
Nesselroade, J. R. (1991). The warp and woof of the developmental fabric. In R. Downs,
L. Liben, & D. S. Palermo (Eds.), Visions of aesthetics, the environment, and
development: The legacy of Joachim F. Wohlwill (pp. 213-240). Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.
Pedersen, N. L., & Harris, J. R. (1990). Functional capacity and activities of daily living.
Behavior Genetics, 20, 740.
Pedersen, N. L., Mcclearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., & DeFaire, U.
(1991). The Swedish Adoption/Twin Study of Aging: An update. Acta Geneticae
Medicae et Gemellologiae: Twin Research, 40, 7-20.
Pedersen, N. L., & Reynolds, C. A. (1998). Stability and change in adult personality:
Genetic and environmental components. European Journal of Personality, 12, 365-
386.
53
Roberts, R. W., Caspi, A., & Moffitt, T. E. (2001). The Kids are alright: Growth and
stability in personality development from adolescence to adulthood. Journal of
Personality and Social Psychology, 81, 670-683.
Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality
traits from childhood to old age: A quantitative review of longitudinal studies.
Psychological Bulletin, 126, 3-25.
Roberts, B. W., Kuncel, N. R., Shiner, R., Caspi, A., & Goldberg, L. R. (2007). The
power of personality. The comparative validity of personality traits, socioeconomic
status, and cognitive ability for predicting important life outcomes. Perspectives on
Psychological Science, 2, 313-345.
Roberts, B. W., & Mroczek, D. (2008). Personality trait change in adulthood. Current
Directions in Psychological Science, 17, 31-35.
Roberts, B. W., Robins, R. W., Trzesniewski, K., & Caspi, A. (2003). Personality trait
development in adulthood. In J. Mortimer, & M. Shanahan (Eds.), Handbook of the
life course (pp. 579-598). New York: Kluwer Acad.
Roberts, B. W., Walton, K. E., & Vichtbauer, W. (2006). Patterns of mean-level change
in personality traits across the life course: A meta-analysis of longitudinal studies.
Psychological Bulletin, 132, 3-127.
Roberts, B. W., Wood, D., & Caspi, A. (2008). The development of personality traits in
adulthood. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of
Personality: Theory and Research (3rd ed., pp. 375-398). New York, NY: Guilford.
Robins, R. W., Fraley, R. C., Roberts, B. W., & Trzesniewski, K. (2001). A longitudinal
study of personality in young adulthood. Journal of Personality, 69, 617-640.
Rose, G., McCartney, P., & Reid, D. D. (1977). Self-administration of a questionnaire on
chest pain and intermittent claudication. British Journal of Preventive and Social
Medicine, 31, 42-48.
SAS Institute. (2000). SAS Release 9.2. Cary, NC: Author.
Schaie, K. W., Willis, S. L., & Caskie, G. I. L. (2004). The Seattle Longitudinal study:
Relationship between personality and cognition. Aging, Neuropsychology, and
Cognition, 11, 304–324.
Sharp, E. S., Reynolds, C. A., Pedersen, N. L., & Gatz, M. (2010). Cognitive engagement
and cognitive aging: Is openness protective? Psychology and Aging, 25, 60-73.
54
Singer, J. D., & Willet., J. B. (2003). Applied longitudinal data analysis: Modeling
change and event occurrence. Oxford, UK: Oxford University Press.
Small, B. J., Hertzog, C., Hultsch, D. F. & Dixon, R. A. (2003). Stability and change in
adult personality over 6 years: Findings from the Victoria Longitudinal Study.
Journals of Gerontology: Psychological Sciences and Social Sciences, 58, 166-176.
Srivastava, S., John, O. P., Gosling, S. D., Potter, J. (2003). Development of personality
in early and middle adulthood: Set like plaster or persistent change? Journal of
Personality and Social Psychology, 84, 1041-1053.
Svedberg, P., Gatz, M., Lichtenstein, P., Sandin, S., & Pedersen, N. L. (2009). Self-rated
health in a longitudinal perspective: A 9-year follow-up twin study. Journal of
Gerontology: Social Sciences, 60, 331-340.
Terracciano, A., Costa, P. T., & McCrae, R. R. (2006). Personality plasticity after age 30.
Personality and Social Psychology Bulletin, 32, 999-1009.
Terracciano, A., McCrae, R. R., Brant, L. J., & Costa, P. T., Jr. (2005). Hierarchical
linear modeling analyses of the NEO-PI-R scales in the Baltimore longitudinal study
of aging. Psychology and Aging, 20, 493-506.
Terracciano, A., McCrae, R. R., & Costa, P. T. (2006). Longitudinal trajectories in
Guilford-Zimmerman temperament survey data: Results from the Baltimore
longitudinal study of aging. The Journals of Gerontology: Psychological Sciences,
61, 108-116.
Terracciano, A., McCrae, R. R., & Costa, P. T. Jr. (2010). Intra-individual change in
personality stability and age. Journal of Research in Personality, 44, 31-37.
Williams, P. G., Smith, T. W., & Cribbet, M. R. (2008). Personality and health: Current
evidence, potential mechanisms, and future directions. In G. J. Boyle, G. Matthews,
& D. H. Saklofske (Eds.), The SAGE Handbook of Personality Theory and
Assessment, Vol 1: Personality Theories and Models (pp. 145-173). Thousand Oaks,
CA: Sage
55
CHAPTER 3. INDIVIDUAL DIFFERENCES IN OPENNESS TO EXPERIENCE AND
RISK FOR MORTALITY
CHAPTER 3 ABSTRACT
Previous research has found that openness to experience exhibits small but
significant declines in older adulthood. The purpose of this study was to examine whether
individual differences in level or change in openness were related to risk for mortality. It
was hypothesized that individuals with a lower level and greater rate of decline in
openness would have a increased risk for death. Participants were 1947 individual twins
from the Swedish Adoption/Twin Study of Aging (SATSA) with up to 6 measurements
of openness. The first year in which participation was possible was 1984; the last year
was 2007. Nearly 70% had more than 3 measurements of openness. By 2010,
approximately 55% of the sample had died. This study combined a latent growth model
with a survival model to examine the relationship between openness intercept (level) and
slope (change) and mortality risk. Results of both a two-stage as well as a simultaneous
growth and survival model indicated that a steeper decline in openness was associated
with an increased risk of death, adjusting for age, sex, and education. Level of openness
was unrelated to risk for death. Both from graphs of participant drop out and from the
statistical model results, death accounted for a large part of the observation that openness
declines in older adulthood. Thus, the relationship between decline and death may be best
conceptualized by theories of terminal decline
56
CHAPTER 3 BACKGROUND
The purpose of this study was to examine whether longitudinal openness to
experience was related to mortality. The trajectory of openness to experience has been
found to be generally stable in adulthood but to then show small but significant declines
in older adulthood (Pedersen & Reynolds, 1998; Roberts & Mroczek, 2008; Sharp,
Reynolds, Pedersen, & Gatz, 2010). However, an explanation for this decline remains
elusive. Results from previous research has suggested that decline in openness begins
around age 60 with a greater rate of decline occurring after age 75. This pattern of
stability and change has been found to be fairly consistent across studies (Roberts &
Mrozcek, 2008).
Given that the decline in openness increases with age, one possible explanation
for the decline could be an individual’s proximity to death. Across the aging literature,
studies have suggested that a specific period of time exists prior to death that is
associated with dysregulation across physical, cognitive, and emotional domains
(Johansson & Berg, 1989; Small, Fratiglioni, von Strauss, & Backman, 2003; MacDonald,
Hultsch, & Dixon, 2003). This period of decline prior to death has been referred to as
terminal decline and is conceptualized as the period of a few years prior to death during
which an individual exhibits declines in functioning across multiple domains (Berg,
1996). Specific to the question at hand, Berg (1996) noted the existence of declines in
self-reported personality in the year or two prior to death. Thus, the question of whether
57
change in adult personality is associated with death is related to a larger theoretical
question within the field of gerontology.
Previous research on the relationship between personality and mortality has
focused on cross-sectional studies of conscientiousness, neuroticism, and extraversion. In
a recent review of the literature, Roberts, Kuncel, Shiner, Caspi, and Goldberg (2007)
found that traits relating to hostility or negative emotion (e.g. neuroticism) were
associated with an increased risk for death whereas traits such as conscientiousness or
those relating to positive emotion (e.g. extraversion) were found to be associated with a
longer lifespan. Of the 34 studies identified in this review, only 2 examined the
relationship between openness to experience and death. Specifically, Wilson, Mendes de
Leon, Bienias, Evans, and Bennet (2004) found no significant association between
openness to experience and mortality in a sample of Catholic clergy members. Similarly,
results from a large study of Medicare patients found no association between openness to
experience and mortality (Weiss & Costa, 2005). Beyond this review, in the health and
personality literature, Christensen and colleagues (2002) found no significant relationship
between openness to experience and mortality in a sample of patients with chronic renal
insufficiency. Yet, in a sample of patients with cardiac disease, Jonassaint et al. (2007)
found that at the domain level, there was no significant association between openness to
experience and all-cause mortality; however, when examining the specific facets of
openness, a greater endorsement of feelings and actions was associated with significantly
increased survival.
58
Overall, the existing literature does not suggest a relationship between level of
openness and risk for mortality. However, cross-sectional studies offer a statistical
account of whether the level of a specific personality trait measured at one point in time
is predictive of greater or lesser survival in a particular population. Yet, it is unknown
whether individual differences in stability and change in openness is related to death. The
advantage to a longitudinal design is that estimates from a latent growth curve model can
be used to examine whether not only level, but also change in personality (i.e. slope), is
related to risk for death. However, the statistical methods that allow for an analysis of the
relationship between both level and slope in personality and subsequent risk for death
have only recently become available (see Guo & Carlin, 2004; McArdle, Small, Backman,
& Fratiglioni, 2005). To our knowledge, only one group has employed these methods in a
longitudinal study of personality and death (see Mroczek & Spiro, 2007). Using data
from the Normative Aging Study, Mroczek and Spiro found that a higher mean-level as
well as an increasing amount of neuroticism were associated with a greater risk for death.
No relationship was identified between level or change in extraversion and risk for
mortality. To our knowledge no recent studies have examined the longitudinal
relationship between openness to experience and mortality.
The purpose of this study was to examine the relationship between openness to
experience and mortality using a longitudinal design. To achieve this goal, two
contemporary analytical approaches were used that combine a latent growth curve model
with a proportional hazard model – a two-stage growth and survival model, and a
simultaneous growth and survival model (see Guo & Carlin, 2004; McArdle, Small,
59
Backman, & Fratiglioni, 2005; Ghisletta, McArdle & Lindenberger, 2006; Ghisletta,
2008). Such analyses were advantageous because of the ability to identify whether level
and slope in personality scores were associated with risk for death. It was hypothesized
that lower openness scores would be associated with an increased risk of mortality and
that greater decline in openness scores would also be associated with an increased risk for
mortality. All analyses were adjusted for entry age, sex, and education. Age group was
also included as an additional covariate to aid with interpretation.
CHAPTER 3 METHOD
Participants. Participants were drawn from the Swedish Adoption/Twin Study of
Aging (SATSA; Pedersen et al., 1991). The initial total sample included 1947 individuals
with one to six openness measurement occasions. The first questionnaire (Q1) was sent
out in 1984, Q2 was sent out in 1987, Q3 was sent in 1990, Q4 in 1993, Q5 in 2004, and
Q6 in 2007. Death data were obtained for all decedents from the Swedish national death
data registry. For nondecedents, age was calculated from the censoring date, 12/31/2010.
The survival period was from study entry to either death or study end (max survival = 26
years). Of the total sample, 1080 (55%) were confirmed to have died prior to the
censoring date. Of the 815 men, 60% were deceased and of the 1132 women, 52% were
deceased. Decedents’ mean age was 68.3 at baseline openness measurement, 75.0 at last
measurement, and the average age-at-death was 81.7 years. The average interval between
baseline openness measurement and death was 13.3 years. Of the total sample, 1352
(69%) had three or more measurement occasions (see Table 1). Overall, the data had an
60
advantageous distribution given that the statistical analyses required a substantial
proportion of the sample to have three or more measurement occasions (to inform the
longitudinal model) and to have died (to inform the survival model). The pattern of
participant attrition is presented in Figure 1. The most common pattern of attrition was
for participants to have died or dropped out during the 10.5 year gap between waves 4
(1993) and 5 (2004).
It is important to note that the current study treated twins as individuals and
neither genetic nor family components were included in the analyses. However, because
twins are not independent of each other, it was necessary to adjust for the correlation
between twins (see Statistical Method).
Measures
Openness to Experience. Openness was measured a 6-item scale. This scale was
identified by factor analyzing the 26-item scale, from the widely used and validated
NEO-PI (Costa & McCrae, 1985), and selecting the 6 highest loading items (see
Bergeman et al., 1993). This scale tapped the intellectual component as well as
engagement in new experiences. It was scored in the traditional fashion of the NEO based
on a 5-point likert scale ranging from strongly disagree to strongly agree. Items were
summed to create a total score (range 6-30). Openness to experience data were collected
via questionnaires mailed to all eligible participants. The openness measure included the
same six items across all waves of data collection. Individuals who had at least one
measurement point were included in the sample. See Appendix A for the scale items.
61
Education. Educational attainment was treated as a continuous variable ranging
from 1 (elementary school) to 4 (university or higher). In the current sample, the majority
(60%) have 6 years of education – the required education during the time-period in
Sweden. Education was collected at Q1and available for 90% of the sample with an
openness measurement.
Entry Age. This covariate indicated the age at which each participant entered
SATSA.
Age Group. Three age groups of approximately equal sizes were created that
categorized participants with respect to the Great Depression in Sweden. Age group and
birth cohort are difficult to separate in this study design because participants entered the
study and completed the Qs at different ages. Age group was included as an additional
way to adjust for the obvious confound between age and death. The oldest age group was
composed of individuals born 1891-1919, inclusive (N=732, mean entry age = 72.8).
These participants were born prior to the Great Depression and would have been born or
been young children during the Spanish flu epidemic. The middle age group was made of
individuals born 1920-1933, inclusive (N=643, mean entry age=59.2) and were born
during the Great Depression. The youngest age group was born 1934-1958, inclusive
(N=572, mean entry age=41.9). These participants were born after the Great Depression.
Further, individuals in this age group were born after 1945 had access to penicillin, and
those born after 1950 had access to state provided healthcare.
62
Statistical Method
In order to examine the relationship between openness to experience and
mortality, a number of initial analyses were carried out to describe the cross-sectional and
longitudinal relationship between openness and death. Covariates in the analyses
included entry age (or age group), sex, and education. Entry age and education were
mean centered, sex was effect coded (males = +.5 and females=-.5), and age group was
effect coded (oldest age group = +1; middle = 0; young = -1). It is important to note that
entry age and age group were used as ways to adjusted for age effects in the model; the
latter was included to aid in interpretation and discussion.
Longitudinal Models. Latent growth models are defined by the intercept, which
gives an estimate of the typical score at a specific age, a linear slope, the systematic
longitudinal variation around the intercept. Models can also include a nonlinear quadratic
slope, which further characterize acceleration (or deceleration) in the trajectory. Age was
centered at the sample mean at first measurement occasion (65 years) and divided by 10
to evaluate change per decade. Thus, the fixed effect intercept estimates the sample-level
mean openness score at age 65 and the slope is the predicted amount of per decade linear
change in openness plus any additional nonlinear change per decade, as estimated by the
quadratic term.
The first step to building the model was to identify the appropriate (best-fitting)
model. A no-growth (intercept only), linear, and quadratic models were compared. Based
on initial analyses it was expected that a quadratic model would provide the best fit.
Models were analyzed using the total sample (N=1947) and adjusted for sex, education,
63
and the correlation between twins. To adjust for the correlation between twins, the
confidence intervals were adjusted in the PROC MIXED datastep in SAS 9.2 (SAS
Institute, 2000) by adding a second RANDOM line to identify the covariance matrix
within and between twin pairs. As a conservative check, all growth models were re-
analyzed using 1 twin from each pair (Sample A, N=879) and then again using the other
twin from each pair (Sample B, N=877). Next, using the best fitting model, individual
who were within 1 to 5 years of death at the time of their last openness score were
removed sequentially (i.e. within 1 year first, then within 2 year, etc.). If the decline in
openness disappeared when a group of individuals were removed, it would suggest that
the change in openness was attributable to the decline in individual scores within a
specific proximity to death.
Survival Models. In this study, death is the event of analysis. Years between study
entry and death or study end described the survival period during which death could
occur. Descriptive analyses and plots evaluated baseline individual differences in risk for
death based on openness and study covariates. First, a parametric survival analysis was
conducted to examine whether and how baseline openness, sex, age group, and education
were related to mortality. Baseline openness was dichotomized into high and low
openness (baseline mean = 17.70 and standard deviation = 4.38; thus high openness ≥
22.11 and low openness ≤ 13.38.). Sex and age group were effect coded, and education
was mean centered. Kaplan-Meier plots were used to examine the survival trajectories.
Models were analyzed using PROC LIFETEST in SAS. Next, a nonparametric survival
analysis was carried out using a Cox proportional hazard model (PROC PHREG). All
64
centered/effect-coded covariates and baseline openness to experience were analyzed in
one model. These analyses were carried out and compared across the total sample, twin
Sample A, and twin Sample B.
Two-Stage Growth and Survival Model. The two-stage growth and survival model
evaluated growth and survival models in two separate steps. Here, the quadratic
phenotypic latent growth model was estimated using PROC MIXED. From this model,
random effects parameter estimates were extracted to obtain Empirical Bayes (EB)
estimates. EB estimates are random effects estimates for the intercept and slope(s) (linear
and quadratic) for each individual across their respective measurement occasions. These
estimates take into account that the time points are correlated (unlike standard regression)
as well as weighing the reliability of data for each individual (i.e. more time points =
more reliable). It is important to note that the EB estimates for slope and quadratic were
also generated for participants with one measurement of openness, thus the overall
variance of the slope and quadratic estimates would be smaller than what would generally
be expected. To this point, data from individuals with one measurement point were
retained because the data help to stabilize mean estimates. The EB estimates for intercept,
slope, and quadratic were then added in a stepwise fashion as predictors in the Cox
proportional hazard model. Next covariates of education, entry age, and sex were added
to the model. Age group was added as a categorical variable, similar to the growth
models such that the youngest and middle groups were compared to the oldest age group
(the reference group). In the two-stage growth and survival model, the correlation
between twins was adjusted for in the survival analysis carried out in PROC PHREG.
65
Simultaneous growth and survival model. The simultaneous (also known as joint
or shared) growth and survival model is a relatively new method (see McArdle et al.,
2005; Ghisletta, McArdle & Lindenberger, 2006; Ghisletta, 2008). The advantage of this
model is that it allows for individual differences in growth trajectories to predict
mortality while accounting for censoring - all in one analysis resulting in gains in
statistical efficiency and power (Guo & Carlin, 2004). The simultaneous model produces
more accurate variance estimates in that the associated standard errors are not biased by
the smaller variance estimates produced by the EB estimates in the two-stage model.
Further, it was straightforward to adjust both the growth and survival pieces for
interactions. See Figure 1 for a path model example of the simultaneous growth and
survival model within a quadratic latent growth framework (adapted from McArdle et al.,
2005). This analysis allowed for the characterization of different trajectories of death
based on level and change in openness and whether particular trajectories, as informed by
covariates, predicted death. The simultaneous growth and survival model was analyzed in
Mplus (Muthén and Muthén, 2000). It was not feasible to adjust for the correlation
between twins within the model statement in Mplus, so twin Sample A (N=879) was used
in this analysis. The results were then re-analyzed using twin Sample B (N=877). The
results did not differ, thus the estimates from Sample A are presented.
It is important to note that the sample size available for analysis varied with
model selection. In the initial growth models, the total sample (N=1947) was used to
select the best fitting model and evaluate whether removing individuals close to death
affected model estimates. In the initial survival models where twinness could not be
66
adjusted for (e.g. Kaplain-Meier plots), 1 twin was selected from each pair (Sample A,
N=879). In the initial Cox proportional hazard models, as well as the two-stage growth
and survival model, twinness was adjusted for in the analysis and the sample was retained
at 1678 (participants with complete data across covariates, excluding twins with unknown
zygosity). Finally, as noted above, it was not possible to adjust for twinness in the
simultaneous growth and survival model in Mplus, thus Sample A was presented. For all
analysis using Sample A, analyses were re-analyzed using Sample B. No differences in
results were observed.
CHAPTER 3 RESULTS
Participant attrition is presented for individuals who died prior to the next
measurement occasion, dropped out and later died, or dropped out and are still alive as of
the censoring date (see Table 2). Data were graphed by those who died (Figure 2) or
dropped out (Figure 3) prior to the next openness measurement and compared to the
means of individuals who remained in the study. The plots suggest that individuals who
died or dropped out declined in openness scores compared to those remaining in the study.
An analysis of variance revealed that participants who died prior to the next measurement
occasion had a significantly lower baseline openness score compared to participants who
remained in the study or compared to participants who dropped out and were still alive at
the censoring date. There was no significant difference in baseline openness between
participants who died and those who dropped out and later died or between participants
who remained in the study and those who dropped out and remained alive.
67
To further examine the pattern of death in relation to openness measurement
occasions, I tabulated the number of years between last openness measurement occasion
and death as well as the total number of participants who died within 1, 2, 3… n years of
last measurement (see Table 3). Of the total deceased sample (N=1080), 51% died within
5 years of their last measurement occasion. Descriptives for each openness measurement
within the last five years prior to death are presented in Table 4 for the total sample and
by sex. A correlation matrix divided by dead and alive participants for openness
measurements and study covariates is presented in Table 5.
Longitudinal Growth Models. The quadratic model was found to provide the best
fit to the total sample. As expected, openness was found to exhibit small but significant
declines, primarily after age 65 (see the top portion of Table 7 for the quadratic model
estimates). Adjusting for sex and education did not change the results. Next, the
unadjusted model was used to examine whether excluding participants within a certain
number of years to death accounted for the rate of decline in openness. Models were
reanalyzed in a step-wise fashion removing the portion of individuals within 1 year of
death, and then 2 years, and then 3 years… and so on, up to 5 years from death. The
decline in openness continued to remain significant even after removing all individuals
within 5 years of death at their last measurement occasion.
Cross-sectional Survival Analysis. Initial parametric survival analyses were fit
using PROC LIFETEST. Results indicated that males, less educated individuals, and
older individuals were at a significantly greater risk for death. Further, individuals with a
high baseline openness had a greater survival compared to individuals with low baseline
68
openness (High =1 standard deviation above the mean). These unadjusted analyses are
presented in the Kaplan-Meier plots in Figures 4-7. However, results from the
nonparametric cross-sectional analyses of baseline openness, entry age, sex, and
education suggested that baseline openness was unrelated to risk for death whereas age,
sex and education were significant predictors. Analyses were carried out for the total
sample, adjusting for the correlation between twins in the survival model (PROC
PHREG), as well as modeling twin Sample A and twin Sample B separately. This
analysis confirmed that the twin samples resulted in nearly identical results (See Table 6).
Two-Stage Growth and Survival Model. As noted, the quadratic model provided
the best fit to the openness data (see the top of Table 7 for model estimates). From this
model, EB estimates were generated for each individual and then used as the growth
parameters in the Cox proportional hazard model (PROC PHREG). The correlation
between twins was adjusted for in the survival model and the “analysis of full maximum
likelihood estimates with sandwich variance estimate” was selected (as specified in the
SAS output). Results indicated that there was a significant association between openness
level and slope on risk for death such that individuals with higher openness at age 65 and
a lesser rate of decline per decade had a flatter survival curve, indicating a protective
effect. Because the quadratic estimate did not improve model fit and was unrelated to
mortality, it was dropped at this point.
Next covariates were added to the linear model. Sex was a significant predictor of
risk for death, but did not explain the relationship between intercept or slope in openness
and risk for mortality. When education was added, the effect of level of openness on risk
69
for mortality became nonsignificant, but the slope of openness continued to be a
significant predictor of mortality. Finally, entry age was added and as expected, it was
significantly associated with survival such that older participants had a significantly
greater risk for death than the youngest age group. However, entry age did not account
for the relationship between openness slope and risk for mortality. A similar analyses
adding age group as either a continuous variable (effect coded) or as a class variable to
the model did not change the results. Estimates from the two-stage growth and survival
model adjusted for entry age, sex, and education are presented in Table 7. Overall, the
results indicate that individuals with a steeper decline in openness (i.e. larger negative
slope), were at an increased risk for death.
Simultaneous-Growth and Survival Model. The advantage to this model was that
both fixed and random effects of the longitudinal model were estimated as well as the
survival estimates in one analysis. Due to the complicated nature of this analysis, twin
Sample A was used to avoid any bias from the correlation between twins. The model was
re-analyzed using twin Sample B and no differences were found. Model analyses were
performed in a step-wise fashion and covariates were added in the same order as the two-
stage model. The covariates of age, age group, sex, and education were all significantly
related to mortality. Final model results indicated that openness slope remained
significantly associated with mortality after age (or age group), sex, and education were
adjusted for. Specifically, a greater decline in openness was predictive of an increased
risk for death (see Table 8). Overall, the simultaneous model provided a more complete
70
picture by providing estimates for both growth and survival parameters and also
confirmed the pattern of results obtained from the 2-stage growth and survival model.
CHAPTER 3 DISCUSSION
The recent personality literature has identified individual differences in
intraindividual change as an important area for study because personality has been linked
to important life outcomes such as occupation, health, and death (Roberts, Kuncel, Shiner,
Caspi, & Goldberg, 2007; Caspi, Roberts, & Shiner, 2005). In terms of openness to
experience, research has suggested normative mean-level declines in older adulthood. To
our knowledge no studies had examined the relationship between openness and death
using a longitudinal-survival approach, although the level of openness had been found to
be unrelated to death in cross-sectional studies (Wilson, et al., 2004; Weiss & Costa,
2005; Christensen et al., 2002; Jonassaint et al., 2007). Thus, the purpose of this study
was to examine whether the level and slope in openness were related to death.
The results from two contemporary methods of modeling growth and survival
relationships revealed that after adjusting for sex, education, and entry age, the slope (or
rate of decline) in openness was significantly associated with risk of death. Specifically,
individuals with a steeper decline in openness were at greater risk for death. Thus,
individual differences in change in openness explained risk for death above the risk
predicted by age, sex, and educational differences. After the covariates were added, level
of openness was unrelated to risk for death. The lack of relationship between level of
openness and death in the growth-survival models was consistent with the initial cross-
71
sectional analyses in the current study as well as with the lack of finding documented by
previous cross-sectional studies in the literature (Wilson et al., 2004; Weiss & Costa,
2005; Christensen et al., 2002).
Why is the decline in openness associated with risk for death? Openness to
experience is the personality trait that describes an individual’s intrinsic motivation for
knowledge, novel experiences, and an emotional connection with experiences. The
decline in openness may be part of a larger global pattern of decline in an individual’s
functioning at the end of life. Gerontologists have suggested that the period of time prior
to death, known as terminal decline, is accompanied by functional declines across
multiple domains, including markers of physical and cognitive health (see Berg, 1996). In
the current study, participants who died prior to the next measurement occasion were
observed to have a lower openness score compared to individuals who remained in the
study. This pattern was consistent with the terminal decline theory. Yet removing
individuals within five years of death from the latent growth models did not account for
the decline in openness. This lack of finding might be related to the spacing between
measurement occasions such that the measurement occasions were too far apart to
consistently identify the change in openness prior to death in this analysis.
However, the decline in openness may not necessarily represent a detrimental
process. In contrast, the decline in openness may fit within a model of aging based on
individual optimization and compensation in older adulthood (see Baltes & Baltes, 1990;
Freund & Baltes; 2007). As such, declines in openness might be an adaptive response to
aging processes that serve to maximize functioning in other domains. Given the trait
72
description of openness, a framework for the decline in openness may be offered by the
socioemotional selectivity theory (SST, see Carstensen, 2006; Carstensen, Mickels &
Mather, 2006). SST suggests that as time-to-death becomes more salient, individuals
change their goals and preferences. When individuals are younger, goals associated with
the gathering of information, experiencing novelty, and expanding breadth of knowledge
are of greater importance. In contrast, as individual gets closer to death, they are less
interested in new experiences and tend to reduce and focus their interests. Because
openness is a trait conceptualized as the tendency to seek knowledge and novel
experiences, the fairly normative pattern of decline in openness may be a reflection of a
systemic change of goals during older adulthood.
Are individuals who decline less in openness protected from death? The results
indicated that a lesser rate of decline in openness was associated with a lower risk for
death. However, it is unlikely that openness itself extends life. More plausible is that
openness is a correlate associated with multiple other protective factors. For example,
openness has generally been regarded as “positive” trait associated with an emotional and
cognitive engagement with experiences and has also been referred to as trait “intellect”
(see McCrae & Sutin, 2009). In a previous study that used a portion of the same SATSA
sample, higher openness to experience was found to predict better cognitive functioning
after age 65 across 5 cognitive domains even after adjusting for education, disability, and
cardiovascular disease (Sharp, Reynolds, Pedersen, & Gatz, 2010). As such, openness to
experience may represent a trait that aids older adults in their experience of the aging
process via cognitive flexibility and engagement. Given this potential relationship,
73
declines in openness might be reflective of declines in an individual’s ability to engage
effectively with their environments. In this case, the results would fit better within the
theory of terminal decline, such that an individual’s overall decline in functioning is
behind the association with death (Berg, 1996).
The advantages of this study were a large sample size that was concentrated in
individuals age 65 and older, a large proportion of the study being deceased, and a large
proportion of individuals with more than 3 longitudinal measurement occasions. The
longitudinal design and use of two different contemporary statistical techniques made it
possible to examine individual differences in both level as well as the rate of decline as a
predictor of risk for death. To our knowledge, this latter question had not been previously
examined.
It may have been possible to identify a stronger relationship between openness to
experience and death if the openness measurement occasions had been closer together.
Similar to other studies, the period of terminal decline was identified as within 5 years of
death. However, the distance between last measurement and death was often greater than
5 years and this may explain why removing individuals did not account for the decline in
openness. Finally, this study examined all-cause mortality and so it remains unknown
whether openness might have a different relationship with specific causes of death,
although there is no evidence in the literature to suggest this.
Overall, the current study advanced the literature by examining the relationship
between the longitudinal trajectory of openness to experience and death. Results from
two contemporary statistical techniques for modeling growth and survival indicated that
74
individuals with steeper declines in openness were at a greater risk for death. To our
knowledge, this is the first study to identify a relationship between change in openness
and risk for death. This relationship remained significant after adjusting for age, sex, and
education. Overall, the decline in openness after age 65 is best conceptualized within a
terminal decline framework; however, this does not preclude the decline from being a
potentially adaptive progression in older adulthood.
75
Table 10. Participation in openness measurement occasion for total sample and by
censoring status (as of December 31, 2010)
Number of Openness Waves Completed
Sample 1 2 3 4 5 6
Total 332 263 298 434 221 399
Deceased 264 199 203 314 66 34
Alive 68 64 95 120 155 365
76
Table 11. Participant attrition with means and standard deviations for openness at last
measurement occasion (as of December 31, 2010)
Attrition Status Q1-Q2 Q2-Q3 Q3-Q4 Q4-Q5 Q5-Q6
Died before next wave
N 69 80 79 400 77
Mean 16.54 16.41 16.11 17.18 17.16
SD 4.85 4.18 4.40 4.73 5.03
Dropped out
N 88 115 106 200 123
Mean 17.00 16.34 16.94 17.24 17.95
SD 4.63 4.54 5.04 4.39 4.35
a. Died later 67 87 80 86 8
Mean 16.63 15.33 16.60 16.84 16.50
SD 4.76 4.18 5.20 3.91 6.16
b. Still Alive 21 28 26 114 115
Mean 18.19 19.46 18.00 17.54 18.03
SD 4.04 4.24 4.47 4.72 4.22
Tot. Lost to follow-up 157 195 185 600 200
Note. Std=Standard deviation. The first openness measurement (Q1) was collected in
1984; Q2 in 1987; Q3 in 1990; Q4 in 1993; Q5 in 2004; Q6 in 2007.
77
Table 12. Number of years between last openness measurement occasion and death (for
confirmed descendants) in the total sample (N=1947)
Years to
death
Last Measurement Occasion
Open 1 Open 2 Open 3 Open 4 Open 5 Open 6 Total
1 25 34 24 43 17 26 169
2 23 26 25 31 26 15 146
3 18 15 23 23 16 6 101
4 6 10 12 33 16 0 77
5 3 8 10 31 5 0 57
6 5 14 6 34 5 0 64
5 5 6 8 33 0 0 46
8 6 3 5 44 0 0 61
9 2 10 4 47 0 0 60
10 3 4 6 39 0 0 47
11 3 6 7 26 0 0 35
12 5 3 4 27 0 0 33
13 3 3 8 28 0 0 37
14 2 5 4 16 0 0 23
15 5 3 2 13 0 0 19
16 1 3 4 10 0 0 13
17 2 1 4 8 0 0 7
18 2 4 1 0 0 0 6
19 5 2 1 0 0 0 6
20 3 2 1 0 0 0 5
21 1 2 0 0 0 0 3
22 2 2 0 0 0 0 3
23 0 1 0 0 0 0 1
24 2 0 0 0 0 0 1
25 2 0 0 0 0 0 2
26 2 0 0 0 0 0 2
Total
Dead
136 167 159 486 85 47 1080
78
Table 13. Means and standard deviations for individuals who died within 1 to 5 years of
their last openness measurement and for participants who remained alive
Number of Years Prior to Death
Alive
1 2 3 4 5
Total 169 146 101 77 57 867
Mean 16.86 16.44 16.48 16.53 17.25 18.15
SD 4.79 4.28 4.68 4.27 4.56 4.26
Males 99 80 55 36 24 328
Mean 17.35 16.61 17.24 17.19 16.41 18.22
SD 4.73 4.02 4.56 16.41 3.79 4.26
Females 70 66 46 41 33 539
Mean 16.16 16.23 15.57 15.95 17.85 18.11
SD 4.80 4.60 4.70 4.41 5.03 4.26
Note. SD=standard deviation
79
Table 14. Correlations between openness measurement waves and study covariates for
dead (lower half) and alive (upper half)
Note. *p<.05. Sex (males=1; females=2). Age = study entry age
Variable 1 2 3 4 5 6 7 8 9
1. Open1 .74* .72* .74* .65* .61* -.03 .30* -.06
2. Open2 .66* .77* .73* .69* .69* .02 .30* -.02
3. Open3 .62* .69* .74 * .69* .63* -.004 .32* -.06
4. Open4 .63* .66* .72* .71* .66* -.02 .34* -.09*
5. Open5 .59* .61* .67* .64* .75* .01 .35* -.15*
6. Open6 .49* .64* .63* .66* .70* .02 .37* -.08*
7. Sex -.02 -.005 -.03 -.02 -.04 -.04 -.07 .12*
8. Educ .24* .22* .22* .22* .20* .13* -.17* -.27*
9. Age -.14* -.16* -.10* -.20* -.16 -.13 .20* -.19*
80
Table 15. Initial survival analysis of baseline openness to experience adjusted for age, sex,
and education. Comparison of total sample, twin Sample A and twin Sample B.
Parameter Estimate (s.e.) χ
2
Pr> χ
2
Hazard Ratio
[95% C.I.]
Total Sample
(N=1678)
Entry Age 0.12 (0.004) 808.79 <0.0001 1.13 [1.12-1.14]
Sex -0.55 (0.07) 64.41 <0.0001 0.58 [0.50-0.66]
Education -0.10 (0.04) 5.40 0.02 0.90 [0.83-0.98]
Openness -0.001 (0.01) 0.02 0.90 0.99 [0.98-1.01]
Twin Sample A
(N=879)
Entry Age 0.12 (0.005) 472.87 <0.0001 1.13 [1.11-1.14]
Sex -0.52 (0.09) 31.51 <0.0001 0.59 [0.49-0.71]
Education -0.12 (0.06) 3.75 0.05 0.89 [0.79-1.00]
Openness -0.006 (0.01) 0.30 0.58 1.01 [0.98-1.03]
Twin Sample B
(N=877)
Entry Age 0.12 (0.006) 453.64 <0.0001 1.13 [1.12-1.14]
Sex -0.54 (0.09) 34.42 <0.0001 0.58 [0.48-0.70]
Education -0.07 (0.07) 1.19 0.28 0.93 [0.81-1.06]
Openness -0.005 (0.01) 0.25 0.62 1.00 [0.97-1.02]
Note. Baseline openness, education and age are mean-centered; Sex is effect coded
(males=-.5 and women=+.5). The total sample results adjust for the correlation between
twins in the model.
81
Table 16. Survival results estimates (standard errors) with covariates from two-stage
growth and survival models.
Parameter
Estimate
(s.e.)
χ
2
Pr> χ
2
Hazard Ratio
[95% C.I.]
-2LL
Growth Estimates
Open Level 17.93 (0.10)*
Open Slope -0.45 (0.05)*
Open Quadratic -0.29 (0.05)*
Survival Model 1
Open Level -0.02 (0.02) 2.18 0.14 0.98 [0.96-1.01]
Open Slope -0.25 (0.07) 12.76 0.0004 0.78 [0.68-0.89] 10826.737
Sex -0.51 (0.08) 43.89 <0.0001 0.60 [0.52-0.70]
Education -0.13 (0.05) 6.44 0.01 0.88 [0.79-0.97]
Age Group -1.54 (0.06) 633.3 <0.0001 0.21 [0.19-0.24]
Survival Model 2
Open Level -0.01 (0.01) 0.56 0.46 0.99 [0.97-1.01]
Open Slope -0.27 (0.06) 17.28 <0.0001 0.77 [0.68-0.87]
Sex -0.56 (0.06) 66.81 <0.0001 0.57 [0.50-0.65] 10589.094
Education -0.09 (0.05) 3.82 0.05 0.92 [0.84-1.00]
Entry Age -0.12 (0.004) 808.5 <0.0001 1.13 [1.12-1.14]
Note. *p<0.001. Latent growth model estimates (PROC MIXED, N=1947) and Cox
proportional hazard model estimates (PROC PHREG, N=1678). Open level centered at
age 65; Open slope and quadratic is change per decade; Sex is effect-coded (males=-.5
and women=+.5); Age group is effect-coded: (Young=-1, Middle=0, Old=+1); Education
and entry age are mean centered; -2LL = - 2*Log Likelihood.
82
Table 17. Effects of openness to experience on age of death in simultaneous longitudinal
and Cox proportional hazard model.
Model Parameters
Model 1 Model 2
Estimate (s.e.) Estimate (s.e.)
Survival Estimates
Open Level β
h,L
-0.01 (0.02) -0.02 (0.02)
Open Slope β
h,S
-0.51 (0.18)* -0.41 (0.12)*
Sex β
h,sex
-0.46 (0.12)* -0.52 (0.11)*
Education β
h,educ
-0.10 (0.07)* -0.09 (0.07)
Entry Age β
h,age
--- 0.13 (0.01)*
Age Group β
h,BC
-1.50 (0.10)* ---
Growth Model Fixed Effects
Intercept @ 65 18.02 (0.14)* 17.94 (0.14)*
Slope/decade 0.41 (0.08)* -0.29 (0.08)*
Quadratic/decade -0.30 (0.11)* -0.66 (0.14)*
Sex 0.43 (0.27) 0.41 (0.27)
Sex*Slope 0.08 (0.14) 0.10 (0.15)
Sex*Quadratic -0.18 (0.13) -0.18 (0.13)
Education 1.24 (0.15)* 1.24 (0.15)*
Education*Slope 0.03 (0.08) 0.03 (0.08)
Education*Quad 0.05 (0.07) -0.03 (0.07)
Entry Age --- 0.002 (0.12)
Entry Age*Slope --- 0.006 (0.01)
Entry Age*Quad --- -0.02 (0.01)*
Age group -0.09 (0.18) ---
Age group*Slope 0.27 (0.12)* ---
Age group*Quad 0.17 (0.11) ---
Growth Model Random Effects
Var. Intercept 11.10 (0.71)* 11.08 (0.16)*
Var. Slope 0.76 (0.21)* 0.91 (0.23)*
Var. Quadratic 0.37 (0.13)* 0.40 (0.14)*
Cov. Int-Slope -0.19 (0.26) -0.26 (0.26)
Cov. Int-Quad -0.38 (0.27) -0.40 (0.27)
Cov.Slope-Quad 0.42 (0.13)* 0.51 (0.13)*
Residual Variance 4.89 (0.16)* 4.84 (0.16)*
Fit Statistics
-2LL 19884.290 (28) 19757.622 (28)
AIC 19932.291 19805.622
BIC 20046.982 19920.313
Note. * p <.05; Simultaneous model estimated using Mplus. Participants are 879
individuals from twin Sample A. Model 1 is adjusted for sex, education and age group.
Model 2 is adjusted for sex, education, and entry age. - 2LL= -2*Log Likelihood (number
of parameters).
83
Figure 6. Latent growth model with shared survival function
Note. Simultaneous growth and survival model is an expansion of the quadratic growth
model for openness to experience. The level of the intercept (I), and the growth estimates
from the linear slope (S) and Quadratic slope (Q) predict hazard rates (Log{h[t]}). The
hazard is indicated by a binary outcome of death at a particular age (adapted from
McArdle, et al., 2005).
84
Figure 7. Openness scores from participants who died prior to the next measurement
occasion (a) and participants who dropped out prior to the next measurement occasion (b)
compared to participants who remained in the study.
(a)
12
13
14
15
16
17
18
19
20
21
22
Open1 Open2 Open3 Open4 Open5 Open6
Q Measurement Occassion
Openness Score
Died After Open1 Died After Open4
Died After Open2 Died After Open5
Died After Open3 Remained in Study
(b)
12
13
14
15
16
17
18
19
20
21
22
Open1 Open2 Open3 Open4 Open5 Open6
Q Measurement Occassion
Openenss Score
Dropped out after Open1 Dropped out after Open4
Dropped out after Open2 Dropped out after Open5
Dropped out after Open3 Remained in Study
85
Figure 8. Kaplan-Meier survival plot by baseline openness to experience (a) and age (b)
(a)
(b)
Note. High openness =1 standard deviation above; Low openness = 1 standard deviation
below the mean). Oldest Age Group (born 1891-1919); Middle Age Group (born 1920-
1933); Youngest Age Group (born 1934-1958).
86
Figure 9. Kaplan-Meier survival plot by sex (a) and education (b)
(a)
(b)
87
CHAPTER 3 REFERENCES
Almada, S. J., Zonderman, A. B., Shekelle, R. B., & Dyer, A. R. (1991). Neuroticism and
cynicism and risk of death in middle-aged men: The western electric study.
Psychosomatic Medicine, 5, 165-175.
Berg, S. (1996). Aging, behavior, and terminal decline. In J.E. Birren & K. W. Schaie
(Eds.), Handbook of the Psychology of Aging. (4th ed., pp. 323-337). San Diego:
Academic Press.
Bergeman, C. S., Chipuer, H. M., Plomin, R., Pedersen, N. L., McClearn, G. E.,
Nesselroade, J. R., Costa, P. T. Jr., & McCrae, R. R. (1993). Genetic and
environmental effects on openness to experience, agreeableness, and
conscientiousness: An adoption/twin study. Journal of Personality, 61, 159-179.
Carstensen, L. L. (2006). The influence of a sense of time on human development.
Science, 312, 1913–1915.
Carstensen, L. L., Mikels, J. A., & Mather, M. (2006). Aging and the intersection of
cognition, motivation and emotion. In J. Birren & K. W. Schaie (Eds.), Handbook
of the Psychology of Aging (6th ed., pp. 343–362). San Diego, CA: Academic
Press.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and
change. Annual Review of Psychology, 56, 453-484.
Christensen, A. J., Ehlers, S. L., Wiebe, J. S., Raichle, K., Ferreyhough, K., Lawton, W.
J. (2002). Patient personality and mortality: A 4-year prospective examination of
chronic renal insufficiency. Health Psychology, 21, 315-320.
Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory Manual. Odessa,
FL: Psychological Assessment Resources.
Guo, X., & Carlin, B (2004). Separate and joint modeling of longitudinal and event time
data using standard computer packages (statistical practice). The American
Statistician, 58, 1-9.
Ghisletta, P. (2008). Application of a joint multivariate longitudinal-survival analysis to
examine the terminal decline hypothesis in the Swiss interdisciplinary longitudinal
study of the oldest old. Journal of Gerontology: Psychological Sciences, 63, 185-
192.
88
Ghisletta, P., McArdle, J. J., & Lindenberger, U. (2006). Longitudinal cognition-survival
relations in old and very old age. European Psychologist, 11, 204-223.
Jonassaint, C. R., Boyle, S. H., Williams, R. B., Mark, D. B., Siegler, I. C., & Barefoot, J.
C. (2007). Facets of openness predict mortality in patients with cardiac disease.
Psychosomatic Medicine, 69, 319-322.
Johansson, B., & Berg, S. (1989). The robustness of the terminal decline phenomenon:
Longitudinal data from the digit-span memory test. Journal of Gerontological and
Psychological Science, 44, 184-186.
McArdle, J. J., Small, B. J., Backman, L., & Fratiglioni, L. (2005). Longitudinal models
of growth and survival applied to the early detection of Alzheimer’s disease. Journal
of Geriatric Psychiatry and Neurology, 18, 234-241.
Mikels, J. A., Löckenhoff, C. E., Maglio, S. J., Carstensen, L. L., Goldstein, M. K., &
Garber, A. (2010). Following your heart or your head: Focusing on emotions
versus information differentially influences the decisions of younger and older
adults. Journal of Experimental Psychology: Applied, 16, 87-87-95.
Mroczek, D. K., & Spiro, A. (2007). Personality change influences mortality in older
men. Psychological Science, 18, 371-376.
Mroczek, D. K., Spiro, A., & Turiano, N. A. (2009). Do health behaviors explain the
effect of neuroticism on mortality? Longitudinal findings from the VA normative
aging study. Journal of Research in Personality, 43(4), 653-659.
Muthén, L. K., & Muthén, B. O. (1998-2007). Mplus User’s Guide. Sixth Edition. Los
Angles, CA: Muthén & Muthén
Pedersen, N. L., Mcclearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., & DeFaire, U.
(1991). The Swedish Adoption/Twin Study of Aging: An update. Acta Geneticae
Medicae et Gemellologiae: Twin Research, 40, 7-20.
Pedersen, N. L., & Reynolds, C. A. (1998). Stability and change in adult personality:
Genetic and environmental components. European Journal of Personality, 12, 365-
386.
Roberts, B. W., Kuncel, N. R., Shiner, R., Caspi, A., & Goldberg, L. R. (2007). The
power of personality. The comparative validity of personality traits, socioeconomic
status, and cognitive ability for predicting important life outcomes. Perspectives on
Psychological Science, 2, 313-345.
89
Roberts, B. W., & Mroczek, D. (2008). Personality trait change in adulthood. Current
Directions in Psychological Science, 17, 31-35.
SAS Institute. (2000). SAS Release 9.2. Cary, NC: Author.
Sharp, E. S., Reynolds, C. A., Pedersen, N. L., & Gatz, M. (2010). Cognitive engagement
and cognitive aging: Is openness protective? Psychology and Aging, 25, 60-73.
Small, B. J., Fratiglioni, L., von Strauss, E., & Backman, L. (2003). Terminal decline and
cognitive performance in very old age: Does cause of death matter? Psychology and
Aging, 18, 193-202.
Small, B. J., Hertzog, C., Hultsch, D. F. & Dixon, R. A. (2003). Stability and change in
adult personality over 6 years: Findings from the Victoria Longitudinal Study.
Journals of Gerontology: Psychological Sciences and Social Sciences, 58, 166-176.
Weiss, A., & Costa, P. T. (2005). Domain and facet personality predictors of all-cause
mortality among Medicare patients aged 65 to 100. Psychosomatic Medicine, 67,
715-723.
Wilson, R. S., Mendes de Leon, C. F., Bienias, J. L., Evans, D. A., & Bennett, D. A.
(2004). Personality and mortality in old age. The Journals of Gerontology:
Psychological Sciences and Social Sciences, 58, 110-116.
90
CHAPTER 4. GENETIC AND ENVIRONMENTAL SOURCES OF STABILITY AND
CHANGE IN OPENNESS TO EXPERIENCE ACROSS THE ADULT LIFESPAN
CHAPTER 4 ABSTRACT
The personality trait of openness to experience has been found to show mean-
level declines after age 65. The purpose of this study was to examine the pattern of
genetic and environmental contributions to level (intercept) and slope (linear and
quadratic) in longitudinal openness. The sample included twins from the Swedish
Adoption/Twin Study of Aging (SATSA). This longitudinal study included a twin reared-
apart, reared-together design and up to six measurement occasions, spanning 23 years.
For all models, age was centered at 65. Males and females were modeled separately.
Results from the biometric latent growth model revealed that for both men and women
genetic influences accounted for the majority of the total variance on the mean level,
linear, and quadratic slopes of openness. Nonshared environmental components were
important but of smaller influence. Shared and correlated environmental influences were
not statistically significant. Cholesky variance components suggested sex differences in
the longitudinal trajectory of openness. For both men and women, genetic effects were
larger than environmental effects and increased with age. For males, nonshared
environmental influences increased between ages 40-65 and then declined. For females,
nonshared environmental influences declined continually from age 40. While specific to
91
openness, these findings offer support to hypotheses that genetic influences might be a
primary source of the normative pattern of decline in older adulthood.
CHAPTER 4 BACKGROUND
Whether personality traits remain stable or change across adulthood is of current
debate. A substantial amount of both cross-sectional and longitudinal evidence supports
the hypothesis of rank-order and mean-level stability of personality traits after age 30
(Costa & McCrae, 1997; Terracciano, Costa, & McCrae, 2005). In contrast, other studies
have demonstrated notable individual differences in the stability of personality across the
lifespan (Caspi & Roberts, 2001; Lewis, 2001; Helson, Kwan, John, & Jones, 2002). A
recent meta-analysis suggested evidence for continued development (i.e. change) in
personality after adulthood as well as considerable individual differences in the trajectory
of change (Roberts, Walton, & Viechtbauer, 2006).
Personality development has been suggested to be accounted for by genetic
influences (McCrae, Costa, Ostendorf, Angleitner, Hrebickova et al., 2000). Specifically,
this hypothesis suggests that any personality change over the lifespan is part of an
inherent genetic process that surpasses individual environment. Thus, any systematic
change over the lifespan in personality would be attributable to increased genetic
influences. In contrast, an alternative hypothesis is that developmental life events and
individual experience (e.g. loss of a spouse, health factors, or proximity to death) affects
change in personality (Caspi, Roberts, & Shiner, 2005; Mroczek & Spiro, 2003). This
92
hypothesis suggests that individual environments are the source of continued change in
personality.
Recent applications of latent growth curve analysis to longitudinal personality
data have allowed for a more nuanced examination of individual differences in change.
The advantage of this approach is the ability to examine change in both level and slope in
personality across advancing age. Studies specific to openness have generally reported
small but significant mean decreases in trait openness beginning around the second half
of the lifespan. (Caspi et al., 2005; Terraciano et al, 2005; Pedersen & Reynolds, 1998).
Behavioral genetic analyses of longitudinal personality data extend the capability of the
latent growth model by decomposing the variance in the intercept and slope into genetic
and environmental components. Thus, this analysis describes the variance accounted for
by heritable factors versus environmental factors and how the proportion of each
contribution might change over the lifespan. Very few studies have examined the
longitudinal genetic and environmental influences on the trajectories of personality.
The purpose of the current study was to apply a biometric growth model approach
to examine the patterns of genetic and environmental sources of stability and change in
openness to experience across the lifespan. Openness to experience is one dimension of
the Big Five trait model of personality, which also includes extraversion, neuroticism,
conscientiousness, and agreeableness (McCrae & Costa, 1997). Generally, openness is
characterized by an emotional connection with experiences and is highly correlated with
both education and cognitive ability Open individuals are described as having an intrinsic
wish for knowledge, curiosity, and the ability to assimilate novel ideas (McCrae, 1994;
93
McCrae & Sutin, 2009). Previous studies have found that openness increases throughout
childhood and young adulthood, remains generally stable across adulthood, and then
evidences small but significant mean level declines after age 60 (Caspi, Roberts, & Wood,
2001). However, to our knowledge no studies have identified the source of decline in
openness in older adulthood. A fundamental question is whether the variation in the level
(intercept) and change (slope) is primarily influenced by genetic or environmental factors.
Generally, behavior genetics studies of cross-sectional self-reported personality
data have reported that additive genetic influences account for between 40-60% of the
total variance, nonshared environmental influences account for 40-60%, and shared
environmental influences account for little to none of the total variance (Krueger,
Johnson, & Kling, 2006, Viken, Rose, Kaprio, & Koskenvuo, 1994; Floderus-Myrhed,
Pedersen, & Rasmussen, 1980; Bouchard & Loehlin, 2001). Recently, Johnson, Vernon,
and Feiler (2008) aggregated results from 145 total studies in a meta-analysis that
reported weighted means. Specific to openness, 19 studies identified a correlation value
for monozygotic twins raised together, 17 studies provided a correlation value for
dizygotic twins raised-together, 4 studies had a correlation value for monozygotic twins
raised-apart, and 2 studies noted a correlation value for dizygotic twins reared apart.
Using these studies, weighted means were computed for genetic variance (.49, range
=.16-.81); nonshared environment (.48, range = .19-.81); and correlated environment (.14,
range= .00-.28). These estimates were based on cross-sectional analyses of personality
and thus provide a moment-in-time look at the biometrical variance structure of openness.
Missing in the literature are studies that have examined the potential for change in genetic
94
and environmental contributions to personality traits over the lifespan. Finally, potential
sex differences in the pattern of genetic and environmental influences over age remains
largely unexplored in biometric studies of personality.
In an earlier study of SATSA data, Pedersen and Reynolds (1998) examined a
genetic model of the NEO- Personality Inventory (NEO-PI) - a personality measurement
based on the five-factor model of personality (Costa & McCrae, 1985). Results for
openness indicated different patterns of genetic and environmental contributions for
males and females. For males, the greatest proportion of the total variance was
attributable to nonshared environments, but this declined with time; additive genetic
influences were small to moderate but stable; and shared environmental influences were
negligible. For females, nonshared environmental influences accounted for most of the
total variance, additive genetic influences were moderate and increased with time;
influences due to correlated environment were small and declined with time, whereas
influences due to shared environments were generally small but quite variable over time.
Despite the advantageous design of this initial SATSA study, it was limited by fewer
measurement points, statistical models that allowed only for linear change, and an
approach that modeled change over time (instead of over age). These limitations reflect
the fact that more sophisticated statistical analyses were not widely used at the time (e.g.
latent growth curve modeling).
Recently, Bleidorn, Kandler, Reimann, Angleitner, and Spinath (2009) examined
the 5 factor model of personality using a biometric latent growth curve approach. Results
specific to openness indicated that 44% of the variance in level was accounted for by
95
additive genetic variance and 23% was accounted for by nonshared environment.
However, the slope of openness was almost completely accounted for by nonshared
environmental factors. Thus, the stability (intercept) of openness was affected by both
genetic and environmental factors whereas the change (linear slope) was primarily
affected by environmental factors. This study was limited by a smaller sample size that
precluded an examination of sex differences, the availability of only 3 measurement
points, analyses that modeled change over time (i.e. by measurement occasion) rather
than by age, and statistical models that only considered linear change.
The purpose of the current study was to use the SATSA longitudinal twin reared-
together reared-apart design to examine the pattern of genetic and nongenetic variance on
the trajectory of openness to experience. An advantage of the current study was the use of
a biometric latent growth model and sufficient sample size to examine potential sex
differences and enough measurement occasions to model change over age as well as both
linear and nonlinear change. Based on previous SATSA research, sex differences were
expected. For level, it was expected that the genetic contributions to level of openness to
experience would remain stable over time while environmental and individual-specific
influences would increase with age. As for the sources of change in openness for males,
nonshared environment was expected to explain the majority of the variance in slope,
whereas for females, greater genetic influences on slope were expected. Shared and
correlated environmental components were not expected to be significant for either males
or females.
96
CHAPTER 4 METHOD
Participants. Participants were drawn from the Swedish Adoption/Twin Study of
Aging (SATSA). SATSA is made up of all Swedish twins who reported that they were
separated and reared apart from each other prior to age 11 (the majority of the twins in
this sample were separated before age 2). This sample was complemented by another
sample of twin pairs reared together who were matched to the sample of separated twins
on date of birth, gender, and county of birth (for a complete description of the SATSA
sample, see Pedersen et al. 1991). SATSA is an ongoing study that began data collection
in 1984 and continues to collect longitudinal follow-ups approximately every 3 years. At
the time of this study, data from up to six measurement waves were currently available
for analysis. The first questionnaire (Q1) was sent out in 1984, Q2 was sent out in 1987,
Q3 was sent in 1990, Q4 in 1993, Q5 in 2004, and Q6 in 2007.
A large sample size and a majority of participants with three or more
measurement occasions were the basic requirements to apply twin models to a
developmental personality question. The total sample consisted of 1850 individual twins
with known twin status (i.e. zygosity) and at least one measure of openness to experience.
Of the sample, 635 individuals were monozygotic (MZ) including 263 individual twins
who were reared apart (MZA) and 372 who were reared together (MZT). There were
1215 individual dizygotic (DZ) twins, including 646 individual twins who were reared
apart (DZA) and 569 who were reared together (DZT). Further, 70% of participants had 3
or more measurements of openness to experience (see Table 1 for participation by
97
zygosity). Attrition is an important concern for any longitudinal study. In the current
study, MZ and DZ twins were proportionately similar in attrition rates. Overall,
individuals who died prior to the next measurement occasion or dropped out and then
later died were likely to have a lower openness score compared to individuals who
remained in the study. There were no consistent differences in mean openness scores
between MZ and DZ twins who died or dropped out of the study (see Table 2).
Measures. Openness to experience was measured by a 6-item scale. This scale
was created by factor-analyzing 26-items from the widely used and validated NEO-PI
(Costa & McCrae, 1985), and selecting the 6 highest-loading items (see Bergeman et al.,
1993). The scale tapped the intellectual component as well as openness to new
experiences. It was scored in the traditional fashion of the NEO based on a 5-point likert
scale ranging from strongly disagree to strongly agree. Items were summed to create a
total score (range 6-30). See Appendix A for the scale items. Openness to experience data
were collected via questionnaires mailed to all eligible participants. The openness
measure included the same six items across all waves of data collection. Individuals who
had at least one measurement point were included in the sample.
Statistical Method
Phenotypic analyses. Initial latent growth curve analyses were performed to
determine the best growth model. Latent growth curve models measure and allow for
comparisons of individual trajectories of growth or decline as well as an average
trajectory of growth or decline across the entire sample. Two factors or more can be
defined based on longitudinal data: an intercept, the estimate of the typical score at a
98
specific age or point in time, and a slope, the systematic longitudinal variation around the
intercept. The slope can be modeled as linear and as nonlinear, using a quadratic
approach. In this study, both linear and quadratic (i.e. acceleration in change) were
included to best identify the trajectory of openness over age. The slope and intercept are
considered fixed effects – parameters (coefficients) that describe the overall trajectory of
the sample. Random effects are the parameters that describe the variability around the
fixed effects (Bryk & Raudenbush, 1987; Finkel, Reynolds, McArdle, Gatz, & Pedersen,
2003; McArdle & Hamagami, 1992). Although participants with three or more
measurements are needed to estimate the quadratic model, data from participants with
one or two measurements were included because they inform the intercept and stabilize
slope estimates.
A full maximum-likelihood estimate (MLE) technique was used in the latent
growth models. Longitudinal change was defined by chronological age (and age-squared)
rather than by time or measurement occasion (see McArdle, Ferrer-Caja, Hamagami, &
Woodcock, 2002). It is important to note that modeling change over age is more
appropriate when asking a developmental aging question compared to analyses centered
on an arbitrary measurement time point. Further, McArdle, et al (2002) found that
including a baseline age variable to the latent growth models based on time was not
equivalent to latent growth models based on age. Thus, a biometric latent growth model
centered on age was a more nuanced method for examining whether the same genetic and
environmental factors influence the level, linear slope and/or acceleration in change (i.e.
99
quadratic). This was important because it is unlikely that the sources of stability in
openness would be the same as the sources of change.
Age was centered at 65 for all models and slope was characterized as change per
decade. A no-growth (intercept only), linear model and quadratic model were fit and
compared. Model fit was evaluated by the chi-square difference test (or likelihood ratio
test). Based on pervious analyses of SATSA data (Sharp, Reynolds, Pedersen, & Gatz,
2010), it was expected that a quadratic model would provide the best model fit. The
equation for the quadratic model was:
OPEN
ij
= γ
0
+ γ
1
((Age
ij
– 65)/10) + γ
2
((Age
ij
– 65)/10)
2
+ δ
0i
+ δ
1i
((Age
ij
–
65)/10) + δ
21
((Age
ij
– 65)/10)
2
+ ε
ij
.
Here, OPEN
ij
represents an openness score for the ith individual at measurement point j;
γ
0
reflects the average intercept at age 65; γ
1
represents the linear rate of change at age 65
by decade; AGE
ij
is the ith individual’s age at measurement point j; γ
2
represents the
quadratic rate of change in openness per decade; δ
0i
, δ
1i
and δ
2i
reflect the ith
individual’s deviations from the average intercept, slope, and quadratic parameters
respectively, and ε
ij
reflects the deviation of the ith individual’s score at measurement
point j from their expected linear trajectory (see Figure 1 for a path diagram of the
quadratic growth model). Sex differences were not predicted in the phenotypic growth
model.
All phenotypic models were fit using PROC MIXED in SAS 9.2 (SAS institute,
2000). To include all twins in this phenotypic analysis, it was necessary to adjust for the
bias resulting from the inclusion of twin pairs (i.e. twins are not independent of each
100
other). Pair dependency was accounted for by specifying random effects of growth
parameters within and between twin pairs in the PROC MIXED data step. As a
conservative check, the phenotypic models were also modeled using 1 twin from each
pair (Sample A) and then again using the other twin (Sample B).
Biometric Latent Growth Model. The SATSA sample is made up of both
monozygotic (MZ) and dizygotic (DZ) twins who were reared-apart or reared-together.
Because the SATSA sample is a genetically informed longitudinal sample, it is possible
to extend the phenotypic growth model to a biometric growth model in order to examine
the genetic and environmental contributions to the total variance longitudinally. The
biometric growth model is similar to the latent growth model such that individual
differences in level (intercept) and slope can be decomposed into latent means (for level
and slope) and standard deviations (McArdle, 2006). These deviations are then
decomposed into latent sources that represent the biometric components. SATSA
included a twin reared-apart reared-together design. The biometric model is able to
estimate five potential components of phenotypic variance: additive genetic influences
(a
2
), nonadditive genetic influences (d
2
), shared family environmental influences due to
being reared together (s
2
), correlated environmental influences shared in twins reared
apart or together (c
2
), and nonshared environmental influences (e
2
). This method of
representation is helpful in interpreting the distribution of total phenotypic variance
across the five biometric components longitudinally.
Individual variation was decomposed using the guiding principles that (1) MZ
twins share all genes in common whereas DZ twins share half of their genes on average,
101
based on the products of previous expectations for twins (e.g. the genetic covariance for
MZ twins = 1.0; DZ twins = 0.50). (2) If rearing environment is important, twins reared
together will be more similar than twins reared apart. (3) In this twin design, it is not
possible to solve for both c
2
and d
2
, thus nonadditive genetic influence (d
2
) will be
reflected if the DZ correlation is less than half of the MZ correlation whereas correlated
environmental influences (c
2
) will be identified if the DZ correlation is greater than half
of the MZ correlation (regardless of rearing status). The decomposition model depends
on three main assumptions: (1) environmental similarity is equally important for MZ and
DZ twins within rearing status, (2) random mating for the traits being studied, and (3) no
gene-environment correlations or interactions (Reynolds, Finkel, McArdle, Berg, &
Pedersen, 2005). Based on previous findings (see Pedersen & Reynolds, 1998), males and
females were modeled separately.
The quadratic model results in the decomposition across three latent variables, the
intercept, linear slope, and quadratic slope. The decomposed model is then able to explain
how much variance is accounted for by genetic and nongenetic sources at each age, such
that the means over age indicate age changes at the phenotypic level and the deviation
around the means represents genetic and nongenetic sources of influence. The total
variance of openness was estimated as Vy = Va + Vs + Vc + Ve + Vu. Where V
indicated variance, y was observed openness value, a was additive genetic influences, s
was shared environment, c was correlated environment, and e was nonshared
environment. Vu was the unexplained variance in the growth model and was not further
decomposed (see Figure 2 for a path diagram of a latent growth biometric model adapted
102
from McArdle & Hamagami, 2003; McArdle, Prescott, Hamagami, & Horn, 1998;
Reynolds et al., 2005). The path models describe the trajectory of heritability across age
using the quadratic latent growth model variance estimates decomposed within the
Cholesky model structure.
The biometric model analyses were carried out in Mx (Neale, Boker, Xie, & Maes,
1999). Total phenotypic variance was decomposed into four sources of variance: additive
genetic (A), shared rearing environment (S), correlated environment (C), and nonshared
environment (E). Mx uses a traditional Cholesky model to decompose the variance across
A, C, S, and E. Of note, modeling of longitudinal data in a Cholesky framework can
create an artificial boundary if the variances are restricted to be nonnegative. This
problem was resolved by removing this constraint to the loadings and allowing all
variances to be negative. The full ASCE model was analyzed and nested models were
compared. Models were compared using the chi-square difference test. To improve
model identification, twins less than age 30 at study entry were removed (N=12 males;
N=15 females).
CHAPTER 4 RESULTS
A description of the total sample with at least one measure of openness to
experience by measurement occasion, twin status, and gender is presented in Table 3.
Sample raw trajectories of individual openness measurements across age are presented by
MZA, MZT, DZA, and DZT in Figures 3-6. Intraclass correlations by twin status, sex,
and age group are presented in Table 4. Initial evidence for the longitudinal pattern of
103
genetic and environmental effects can be found in this table. Overall, the DZ correlation
was greater than half of the MZ correlation (regardless of rearing status) at most
measurement occasions, indicating that the phenotypic models should be solved for c
2
(correlated environment) and not d
2
(nonadditive genetic influences).
For females, there was evidence for genetic influence (A) because MZT
correlations were two times greater than DZT correlations, and MZA correlations were
greater than DZA correlations in 5 of the 6 measurement occasions. Nonshared
environmental influences (E) were apparent because the correlations for MZA and MZT
twins were not 1.00. MZA correlations were lower than MZT in 4 of 6 measurement
occasions, yet DZA correlations were lower than DZT in only 3 of 6 measurement
occasions. Given this overall pattern of twin correlations, it was expected that genetic
effects would increase over time, whereas environmental effects would decrease.
For males, the pattern of genetic effects (A) was more varied such that in 4 of the
6 measurement occasions MZT correlations were two times greater than DZT
correlations, and in only 3 of the 6 measurement occasions MZA correlations were
greater than DZA correlations. Nonshared environmental influences (E) were evident
given that MZA and DZT correlations were not 1.00. However, in only 3 of 6
measurement occasions were MZA and DZA correlations lower than MZT and DZT
correlations, respectively. The overall pattern for males suggested that both genetic and
environmental influences would increase with age.
Phenotypic LGM. Results from initial LGM modeling of openness to experience
suggested that a quadratic model provided the best fit (see Table 5). The model estimates
104
indicated mean level declines in openness and additional acceleration in decline with age.
These declines were small but significant and similar for men and women. Estimates
were comparable when analysis were carried out using the full sample and a single-twin
sample (Samples A and B). For ease of interpretation, Sample A is presented.
Biometric LGM. Cholesky parameter estimates from the full ASCE model are
presented separately in Tables 6 (males) and 7 (females) and the raw variances are
graphed in Figure 7 for males and females separately. However, for both males and
females the AE model resulted in the best fit (see Tables 6 and 7, respectively) suggesting
that shared (S) influences and correlated (C) influences were not statistically important to
the pattern of genetic and environmental contributions to either level or slope in openness
centered at age 65. The raw variances per decade as implied by the AE model are
presented in Table 8b. Graphs of the raw additive genetic and nonshared environmental
variance from the AE model for males and females separately are presented in Figure 8.
It is important to note that due to sparseness of data and selection factors (e.g. death); the
values at the tails were generally less reliable. To present the most accurate pattern of
genetic and environmental influences over time, the estimates from age 40-80 are shown
in black, while the point estimates at age 30 and 90 are “grayed out” in the graphs.
The predicted variance components highlighted some sex differences in the
longitudinal trajectory of openness. For both men and women, genetic effects were larger
than environmental effects and increased with age. For males, the genetic and
environmental influences were nearly parallel until age 65 at which point they diverged.
Further, nonshared environmental influences were relatively stable between ages 40-65
105
but then steadily declined. For females, genetic variance and environmental influences
generally diverged from initial measurement such that genetic influences increased more
steeply after age 65, whereas nonshared environmental influences declined continually
from age 40.
Overall, the Cholesky estimates from the AE model suggested that at age 65, the
genetic estimates for the intercept were high (see Table 9b). Specifically that of the total
variance in mean openness, 60% of the variance for males and 74% of the variance for
females was attributable to genetic factors. Genetic influences on the linear slope were
also high, 60% of the total variance for males and 83% of the total variance for females
was attributable to genetic components. Genetic influences on nonlinear change (as
identified by the quadratic slope) were quite high. Specifically, 83% of the total variance
for males and 98% of the total variance for females was accounted for by genetic factors.
It is important to reiterate that these variance estimates were based on the centered age of
65 and change (slope) was characterized by decade.
CHAPTER 4 DISCUSSION
The goal of this study was to examine sources of stability and change in openness
to experience using a biometric latent growth curve analysis. Research on stability and
change in personality has suggested that change in personality beyond age 30 may be the
result of genetic influences (Costa & McCrae, 1997; McCrae et al., 2000). However,
recent studies have demonstrated continued mean-level and individual-level change in
personality across adulthood (Caspi & Roberts, 2001; Roberts, Walton, & Viechtbauer,
106
2006). This change has been attributed to primarily environmental influences such as loss
of spouse (e.g. Mroczek & Sprio, 2003). The literature has consistently identified small
but significant declines in openness to experience after midlife. To date, no studies have
identified the source of the normative pattern of decline in openness. One fundamental
question is how genetic (G) and environmental (E) influences shape the trajectory of
openness to experience across the lifespan.
In terms of the level of openness, it was hypothesized that genetic contributions
would remain stable across ages, whereas environmental influences would increase with
age. In terms of the slope of openness, nonshared environment was expected to explain
the majority of the variance in slope for males, whereas genetic influences were expected
to explain more of the variance on slope for females. Shared and correlated
environmental components were not expected to be significant. Results from the
Cholesky model estimates indicated that at the centered age of 65, the genetic effects
accounted for the greatest proportion of the total variance for level, linear slope, and
quadratic slope. It was not expected that genetic variances would account for such a large
proportion of the total variance in the slopes. Further, for females, the genetic influences
on level were notably higher than suggested in the cross-sectional literature (generally
40-60%), yet they were within the range of genetic variance identified by a recent meta-
analysis (Johnson et al., 2008). Although it seems more reasonable for nonshared
environmental effects to increase with age, other longitudinal behavior genetic studies
have found that genetic influences increase for memory (Reynolds et al., 2005) and
depression (Gatz, Pedersen, Plomin, Nesselroade, McClearn, 1992). It is also important
107
to note that measurement error is not included in this variance distribution and would
likely increase the environmental proportion.
The current findings supported sex differences in the patterns of genetic and
environmental influences over age; however, the differences were less than expected. For
both males and females, shared and correlated environmental influences were not
statistically important to model identification and could be dropped without a significant
loss of fit. Thus, a model including additive genetic (A) and nonshared environmental (E)
components provided the best fit; however, for comparison purposes the full model
estimates were also presented. Overall, results indicated that genetic influences accounted
for the majority of the influences on both level and slope of openness. For males, both
genetic and nonshared environmental influences increased until age 65. After age 65,
genetic influences continued to increase, whereas nonshared environmental influences
showed a steady decline. For females, genetic influences generally increased across the
lifespan whereas nonshared environmental influences declined, with a notably steeper
decline after age 60.
The initial hypotheses were based on previous research with SATSA data that
examined longitudinal openness to experience (Pedersen & Reynolds, 1998). Pedersen
and Reynolds found sex differences such that for both males and females, nonshared
environment accounted for the greater proportion of variance; nonshared environmental
influences declined with time for males but remained stable for females. For males,
genetic influences were small to moderate and stable; for females, genetic influences
were moderate and increased with time. Both correlated and shared environmental
108
influences were small. The difference in findings between the current study and Pedersen
and Reynolds (1998) are likely due to model selection in the statistical analyses and the
current study having 2 additional waves of data available - allowing for a greater
proportion of measurements in older adulthood.
The present study identified strong genetic effects on the change of openness (e.g.
linear and quadratic slopes). In contrast, Bleidorn and colleagues (2009) found evidence
for nonshared environmental influences as the primary source of linear change in
openness. It is likely that this dissimilarity speaks to differences in model identification,
particularly that Bleidorn et al., centered the growth models at age 30 (compared to 65 in
the current study) and used a linear-only model. Different centering ages are likely to
capture different patterns of genetic and environmental contribution. In the case of
openness, where a pattern of decline has been identified after age 65, it is less likely for a
model centered at age 30 to be able to identify accurate estimates of the genetic and
environmental contributions to slope in older adulthood.
McCrae and Costa have theorized that change in personality traits after age 30 is
likely attributable to genetic influences. More specifically, personality traits were
described as endogenous tendencies that supersede individual experience. In this context,
the finding from several studies that openness declines after middle age would not be
considered unaccounted for “change” in personality, but rather part of a normative
pattern of development and perhaps a biologically determined consequence of aging (see
McCrae et al., 2000). It seems plausible that after age 65, there may be more genetically
programmed processes taking place, for example, a genetically based selection effect for
109
those individuals choosing to stay in the study or perhaps declines across individuals
associated with a terminal decline process (Berg, 1996).
This study is one of the few to have examined the sources of stability and change
in personality using a longitudinal behavior genetic design. The SATSA data set is
particularly powerful for detecting heritability due to the sample of twins reared-apart and
twins reared-together. Other advantages included the availability of up to 6 waves of data
spanning almost a quarter of a century, the selection of contemporary statistical models,
inclusion of a quadratic term in the biometric growth model, and modeling change based
on age rather than on time. The growth model estimates provided the expected genetic
and environmental variance components for each age. However, all models were centered
at age 65 and thus model results are specific to what we would expect the genetic and
environmental contributions to be at age 65. It is also important to note that genes and
environments have been found to interact across the lifespan (see Johnson, Vernon, &
Feiler, 2008). It remains unclear how genes and environments that influence stability and
change in openness might interact, particularly in older ages.
The analyses included individuals aged 30-93, but the tails of this range were
decidedly less reliable due to the sparseness of data and likely selection effects. Thus, the
trajectories presented from age 40-80 were considered the most reliable representations
of the genetic and environmental sources of variation in openness to experience. Another
limitation to the current study and most behavior genetic studies of personality is the use
of self-report personality data. It is unknown to what extent the biometrical results were
reliant on the structure and inherent bias of self-report measures. The SATSA measure of
110
openness to experience was a factor analyzed 6-item questionnaire based on the NEO-PI.
This reduced measure precluded the ability to examine the facets within openness to
experience. Finally, the results present the estimated trajectory of longitudinal genetic
and environmental contributions to openness for this Swedish sample, it is unclear how
similar this pattern would be to other populations.
Overall, the findings suggested that genetic or heritable influences are of greater
and increasing importance compared to nonshared environmental influences in
explaining individual differences in openness to experience across advancing age.
Nonshared environmental influences were of lesser importance and declining,
particularly after age 60. Shared and correlated sources of variation were not statistically
important. Males and females differed in the predicted trajectory of genetic and
environmental sources of variation in openness over time. The findings are considered
reliable estimations of the sources of phenotypic variance in openness to experience at
age 65. Further, these findings generally support Costa and McCrae’s hypothesis that
stability and change in openness to experience after age 30 is primarily influenced by
genetic sources.
111
Table 18. Number of openness measurement waves completed by zygosity and rearing
status (numbers represent individual twins)
Number of Openness Waves Completed
1 2 3 4 5 6
MZ 111 87 92 151 68 126
DZ 189 162 191 264 147 262
MZA 58 35 45 62 30 37
MZT 58 51 47 89 38 89
DZA 93 92 94 144 75 148
DZT 96 70 97 120 72 114
Note: MZ, monozygotic; DZ, dizygotic, A, reared apart; T, reared together.
112
Table 19. Attrition across openness measurement occasions by twin status (MZ vs. DZ)
with means and standard deviations for openness at last measurement occasion
Attrition Status Q1-Q2 Q2-Q3 Q3-Q4 Q4-Q5 Q5-Q6
Died before next Q
MZ
22 29 24 134 29
Mean
16.00 16.24 16.50 17.81 17.38
SD
4.83 4.84 4.85 4.52 4.95
DZ
44 47 53 249 48
Mean
16.84 16.15 15.92 16.76 17.02
SD
4.96 3.84 4.26 4.67 5.13
Drop out before next Q
MZ
31 38 41 62 38
DZ
49 65 59 122 77
a. Drop out and later died
MZ
24 31 31 26 2
Mean
16.67 14.16 16.58 16.62 11.50
SD
5.81 3.46 5.66 3.38 3.54
DZ
37 48 44 54 4
Mean
16.70 15.73 16.36 16.94 17.75
SD
4.28 3.93 4.70 4.24 4.35
b. Drop out and still alive
MZ 7 7 10 36 36
Mean 18.29 20.57 17.10 17.86 17.39
SD 4.15 1.72 3.73 4.99 4.12
DZ 12 17 15 68 73
Mean 17.58 19.00 18.93 17.29 18.43
SD 4.10 4.76 4.85 4.68 4.26
Note. N=1850 (participants with known zygosity). MZ= monozygotic; DZ= dizygotic.
SD= Standard Deviation. Q1= First possible wave of openness measurement in 1984; Q2
in 1987; Q3 in 1990; Q4 in 1993; Q5 in 2004; Q6 in 2007.
113
Table 20. Means and standard deviations for openness and age by measurement occasion
by zygosity, rearing status, and sex
Open
Wave
MZ Twins DZ Twins
Reared Apart Reared Together Reared Apart Reared Together
Males Females Males Females Males Females Males Females
Open 1
N 100 89 129 293 177 162 197 224
Mean 17.95 17.03 17.90 17.96 17.73 17.52 17.96 18.00
SD 3.8 5.07 3.9 4.1 4.1 3.9 3.9 4.1
Open 2
N 96 98 130 323 185 164 202 227
Mean 17.33 17.14 18.06 18.01 17.57 18.16 17.70 17.62
SD 3.9 5.07 4.0 4.4 3.9 4.3 4.0 4.3
Open 3
N 86 85 122 296 159 152 185 225
Mean 17.72 17.05 18.10 18.16 17.96 17.92 17.95 17.85
SD 3.6 5.4 4.2 4.5 4.03 4.4 4.3 4.5
Open 4
N 90 85 112 293 175 140 177 221
Mean 17.51 16.96 18.27 18.0 18.17 18.09 17.70 17.65
SD 4.1 5.09 3.7 4.6 4.4 4.5 4.0 4.5
Open 5
N 32 49 67 174 98 86 97 119
Mean 17.97 18.24 17.46 18.69 18.24 17.31 18.36 18.00
SD 3.8 4.4 4.2 4.1 4.29 3.9 4.0 4.5
Open 6
N 28 42 51 148 77 66 76 101
Mean 17.68 17.57 17.49 18.8 18.36 18.17 18.13 18.02
SD 4.5 4.3 4.2 4.3 4.2 4.0 4.1 3.8
Note. N=1850 (twins with known zygosity); MZ= monozygotic; DZ= dizygotic.
SD= standard deviation.
114
Table 21. Intraclass correlations at each openness measurement occasion by gender and
age group
Openness Measurement Occasion
Twin Group Open1 Open2 Open3 Open4 Open5 Open6
Total Sample
MZA (N
pairs
= 22-61) .31 .23 .48 .50 .55 .44
MZT (N
pairs
= 36-104) .36 .43 .59 .47 .54 .60
DZA (N
pairs
= 60-133) .28 .23 .19 .21 .22 .11
†
DZT (N
pairs
= 51-131) .24 .09
†
.15 .17 .20 .26
Men
MZA (N
pairs
= 8-32) .23
†
-.27 .40 .39 .11
†
.34
MZT (N
pairs
= 15-47) .23 .34 .39 .32 .45 .54
DZA (N
pairs
= 17-44) .33 .008
†
.11
†
.24 .21
†
-.28
†
DZT (N
pairs
= 20-61) .31 .12
†
-.12
†
.04
†
.44 .20
†
Women
MZA (N
pairs
= 14-29) .36 .44 .50 .57 .61 .47
MZT (N
pairs
= 21-57) .46 .50 .73 .55 .59 .64
DZA (N
pairs
= 43-89) .24 .30 .20 .19 .23 .21
DZT (N
pairs
= 31-70) .17 .06
†
.33 .25 .03
†
.30
Age<65
MZA (N
pairs
= 22-44) .26 .23 .49 .53 .55 .44
MZT (N
pairs
= 36-85) .40 .43 .60 .50 .53 .60
DZA (N
pairs
= 60-124) .33 .22 .12
†
.18 .18 .11
†
DZT (N
pairs
= 50-96) .23 .03
†
.10
†
.17 .24 .26
Age≥65
MZA (N
pairs
= 14-23) .39 .44 .18
†
.39 -- --
MZT (N
pairs
= 13-21) .19
†
.52 .38 .20
†
-- --
DZA (N
pairs
= 2-35) .16
†
.33 .21 .21
†
.52
†
--
DZT (N
pairs
= 6-106)
.25 .22
†
.20 .14
†
-.93 --
Note. MZ= monozygotic; DZ=dizygotic, A = reared apart; T = reared together. Npairs =
range of twin pairs used to calculate intraclass correlations across measurement occasions.
†
Correlation nonsignificant.
115
Table 22. Maximum likelihood estimates (standard errors) from phenotypic LGM models
Males Females
Parameters
Intercept
Only
Model
Linear
Model
Quadratic
Model
Intercept
Only
Model
Linear
Model
Quadratic
Model
Fixed
Effects
Intercept 17.7(.19) 17.7(.19) 17.8 (.20) 17.7 (.17) 17.8 (.17) 18.0 (.19)
Slope - -0.18 (.08) -0.25 (.09) - -0.44 (.08) -0.53 (.09)
Quadratic
- - -0.19 (.09) - - -0.32 (.09)
Random
Effects
Intercept 13.2 (1.05) 12.8 (1.1) 13.4 (1.18) 14.2 (1.00) 12.9 (.96) 14.2 (1.14)
Slope - 0.29 (.17) 0.40 (.23) - 0.71 (.22) 0.75 (.38)
Quadratic - - 0.23 (.16) - - 0.62 (.27)
Int-Slope - 0.21 (.31) -0.19 (.35) - 0.46 (.31) -0.26 (.38)
Int.-Quad. - - -0.55 (.39) - - -1.34 (.46)
Slope-Quad. - - 0.25 (.16) - - 0.41 (.16)
U[t] - unique 4.64 (.20) 4.45 (.21) 4.33 (.21) 6.56 (.25) 6.08 (.26) 5.74 (.25)
Fit Statistics
-2 LL 7526.3 7517.5 7504.1 10501.0 10459.1 10423.6
Parameters (df) 3 (1) 6 (3) 10 (6) 3 (1) 6 (3) 10 (6)
Note. Data is from Twin A (randomly selected) for up to 6 waves (Males N=415;
Females N=567); -2LL, -2 Log Likelihood; df=degrees of freedom.
116
Table 23. Biometric parameter estimates from ASCE and AE models: Cholesky factor
loadings for males
Model
Parameter
Additive
Genetic
(A)
Shared
Environment
(S)
Correlated
Environment
(C)
Nonshared
Environment
(E)
1 2 3 1 2 3 1 2 3 1 2 3
ASCE
Model
Intercept 2.1 1.1 1.4 2.2
Slope .15 .006 .18 .004 .17 .65 .08 .51
Quad .11 .008 .0001 .07 .004 .0004 .01 .51 .0004 .21 .09 .0001
AE
Model
Intercept 2.69 - - 2.21
Slope .05 .63 - - - - -.04 .52
Quad -.08 .44 .14 - - - - - - -.20 .07 -.000
Note. ASCE (-2LL=13847.106); ACE (-2LL = 13839.406); AE (-2LL= 13837.924)
ASCE compared to ACE, p=0.96
ASCE compared to AE, p=0.69
ACE compared to AE, p=0.26
117
Table 24. Biometric parameter estimates from ASCE and AE models: Cholesky factor
loadings for females
Model
Parameter
Additive
Genetic
(A)
Shared
Environment
(S)
Correlated
Environment
(C)
Nonshared
Environment
(E)
1 2 3 1 2 3 1 2 3 1 2 3
ASCE
Model
Intercept 1.9 1.4 2.1 1.9
Slope .48 .43 -.04 .31 -.33 -.00 -.28 .33
Quad .20 .22 -.00 -.22 -.24 -.00 -.34 .03 .15 -.16 .14 -.00
AE Model
Intercept 3.1 - - 1.9
Slope .02 .80 - - - - -.21 .31
Quad -.26 .37 .43 - - - - - -.04 .07 .00
Note. ASCE (-2LL=19474.597); ACE (-2LL = 19481.17); AE (-2LL=19491.553)
ASCE compared to ACE, p=0.36
ASCE compared to AE, p=0.15
ACE compared to AE, p=0.11
118
Table 25. Expected ASCE (a) variances and unexplained variance across ages by sex and
expected AE (b) variances and unexplained variance across ages by sex
(a)
Males Females
Age Va Vs Vc Ve Va Vs Vc Ve
40 3.17 0.06 3.25 1.74 3.93 5.25 1.52
2.75
50 4.28 0.48 1.19 3.72 2.79 1.88 3.52
3.97
60 4.58 1.02 1.65 4.78 3.09 1.84 4.84 4.06
70 3.97 1.42 2.27 4.44 5.03 1.70 3.49
3.05
80 2.64 1.54 7.00 3.45 10.86 0.67 0.83
1.99
All Ages Males Vu =4.66 Females Vu=5.46
Note. Va= additive genetic variance, Vs=shared environmental variance, Vc=correlated
environmental variance, Ve= Nonshared environmental variance, Vu= unexplained
variance. Models included individuals aged 30-97, however, the most reliable estimates
from age 40-80 are presented.
(b)
Males Females
Age Va Ve Va Ve
40 6.32 1.82 9.41 4.46
50 6.01 3.68 7.49 4.33
60 7.05 4.82 9.53 3.79
70 7.44 4.65 9.81 3.05
80 10.48 3.72 11.83 2.48
All Ages Males Vu =4.66 Females Vu=5.46
Note. Va= additive genetic variance; Ve= Nonshared environmental variance; Vu=
unexplained variance. Models included individuals aged 30-97, however, the most
reliable estimates from age 40-80 are presented.
119
Table 26. Heritability and environmental contributions to openness to experience at age
65 based on ASCE model (a) and AE model (b)
(a)
Parameter a
2
s
2
c
2
e
2
Males Intercept .36 .10 .15 .39
Slope .03 .04 .58 .35
Quadratic .04 .01 .80 .15
Females Intercept .27 .14 .33 .26
Slope .51 .12 .14 .23
Quadratic .22 .29 .37 .12
Note. Rows indicated proportion of variance and add up to 1.00. Measurement error not
included.
(b)
Parameter a
2
e
2
Males Intercept .60 .40
Slope .60 .40
Quadratic .83 .17
Females Intercept .74 .26
Slope .83 .17
Quadratic .98 .02
Note. Rows indicated proportion of variance and add up to 1.00. Measurement error not
included.
120
Figure 10. Quadratic growth model
The squares represent measured variables, the circles represent latent variables, the
circles within squares represent data that are available at some but not necessarily all
measurement occasions; I = intercept; S = slope; Q=Quadratic; I* = standardized score
for the intercept; S* = standardized score for the slope; Q*=standardized score for the
quadratic; M
i
= mean of the intercept; M
s
= mean of the slope; M
q
= mean of the
quadratic; R
is
= correlation between the intercept and slope; R
iq
= correlation between the
intercept and quadratic; R
sq
= correlation between slope and quadratic; D
i
= deviation
from the intercept; D
s
= deviation from the slope; D
q
= deviation from the quadratic; B1-
B6 = linear age basis coefficients; B1
2
-B6
2
= quadratic age basis coefficients; u
1
-u
6
=
random components from the openness measurements; D
u
= a constant deviation from
the openness scores.
121
Figure 11. Biometric quadratic growth model
Note. Example of Cholesky decomposition of additive genetic (A) and nonshared
environmental (E) variation within the growth model parameters of intercept (I), slope
(S), and quadratic term (Q). This model depicts a reduced AE biometric model and not all
paths are presented. Twin A is presented on the right and Twin B on the left. D=
deviations; MZ = monozygotic; DZ = dizygotic. See Figure 1 for an explanation of the
growth model notation and abbreviations.
122
Figure 12. Longitudinal trajectories of openness for sample of (a) MZ twins reared apart,
N=131 and (b) MZ twins reared together N=180
(a)
(b)
Note. Individual twins in each graph randomly selected from Twin Sample A
123
Figure 13. Longitudinal trajectories of openness for sample of (a) DZ twins reared apart
N=326, and (b) DZ twins reared together N= 284
(a)
(b)
Note. Individual twins in each graph randomly selected from Twin Sample A
124
Figure 14. Predicted biometric parameters estimated from the growth model for (a) the
full ASCE, and (b) AE models for males
a.
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
30 40 50 60 70 80 90
Age
Variance
Va
Ve
Vs
Vc
Vu
b.
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
30 40 50 60 70 80 90
Age
Variance
Va
Ve
Vu
Note. Raw variances are presented. Va = additive genetic; Vs= shared environment; Vc=
correlated environment; Ve= nonshared environment; Vu= unexplained variation (not
decomposed into genetic and environmental components).
125
Figure 15. Predicted biometric parameters estimated from the growth model for (a) the
full ASCE, and (b) AE models for females.
a.
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
30 40 50 60 70 80 90
Age
Variance
Va
Ve
Vs
Vc
Vu
b.
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
30 40 50 60 70 80 90
Age
Variance
Va
Ve
Vu
Note. Raw variances are presented. Va = additive genetic; Vs= shared environment; Vc=
correlated environment; Ve= nonshared environment; Vu= unexplained variation (not
decomposed into genetic and environmental components).
126
CHAPTER 4 REFERENCES
Bergeman, C. S., Chipuer, H. M., Plomin, R., Pedersen, N. L., McClearn, G. E.,
Nesselroade, J. R., Costa, P. T. Jr., & McCrae, R. R. (1993). Genetic and
environmental effects on openness to experience, agreeableness, and
conscientiousness: An adoption/twin study. Journal of Personality, 61, 159-179.
Bleidorn, W., Kandler, C., Reimann, R., Angleitner, A., & Spinath, F. (2009). Patterns
and sources of adult personality development: Growth curve analyses of NEO-PI-R
scales in a longitudinal twin study. Journal of Personality and Social Psychology,
87, 142-155.
Bouchard, T. J., & Loehlin, J. C. (2001). Genes, evolution, and personality. Behavior
Genetics, 31, 243-273.
Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to
assessing change. Psychological Bulletin, 101, 147-158.
Caspi, A. & Roberts, B. W. (2001). Personality development across the life: The
argument for change and continuity. Psychological Inquiry, 12, 49-66.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and
change. Annual Review of Psychology, 56, 453-484.
Costa, P. T., & McCrae, R. R. (1997). Longitudinal stability of adult personality. In R.
Hogan, J. A. Johnson, & S. R. Brigss (Eds.), Handbook of Personality Psychology
(pp. 269-290). San Diego: Academic Press.
Costa, P. T., & McCrae, R. R. (1992a). Revised NEO personality inventory (NEO PI-R)
and NEO five-factor inventory (NEO-FFI). Odessa, Florida: Psychological
Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992b). Multiple uses for longitudinal personality data.
European Journal of Longitudinal Research and Personality, 6, 85-102.
Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory Manual. Odessa,
FL: Psychological Assessment Resources.
Finkel, D., Reynolds, C. A., McArdle, J. J., Gatz, M., & Pedersen, N. L. (2003). Latent
growth curve analyses of accelerating decline in cognitive abilities in late adulthood.
Developmental Psychology, 39, 535-550.
127
Flouderus-Myrhed, B., Pedersen, N., & Rasmuson, I. (1980). Assessment of heritability
for personality, based on a short-form of the Eysenck Personality Inventory: A study
of 12,898 twin pairs. Behavior Genetics, 10, 153-162.
Helson, R., Kwan, V. S. Y., John, O. P., & Jones, C. (2002). The growing evidence for
personality change in adulthood: Findings from research with personality
inventories. Journal of Research in Personality, 36, 287-306.
Krueger, R. F., Johnson, W., & Kling, K. C. (2006). Behavior genetics and personality
development. In D. K. Mroczek, & T. D. Little (Eds.), Handbook of Personality
Development (pp. 81-108). Mahwah, NJ: Lawrence Erlbaum, 2006.
Johnson, A. M., Vernon, P. A., Feiler, A. R.. (2008). Behavior genetics 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, Vol 1: Personality Theories and Models (pp.
145-173). Thousand Oaks, CA: Sage
Lewis, M. (2001). Issues in the study of personality development. Psychological Inquiry,
12, 67-83.
McArdle, J. J. (2006). Latent curve analyses of longitudinal twin data using a mixed-
effects biometric approach. Twin Research and Human Genetics, 9, 343-359.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual
abilities over the life span. Developmental Psychology, 38, 115-142.
McArdle, J. J. & Hamagami, F. (2003). Structural equation models for evaluating
dynamic concepts within longitudinal twin analyses. Behavior Genetics, 33, 137-
159.
McArdle, J. J., Prescott, C. A., Hamagami, F., & Horn, J. L. (1998). A contemporary
method for developmental-genetic analyses of age changes in intellectual abilities.
Developmental Neuropsychology, 14, 69-114.
McCrae, R. R. (1994). Openness to experience: Expanding the boundaries of factor V.
European Journal of Personality, 8, 251-272.
McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebícková, M., Avia, M.
D., et al. (2000). Nature over nurture: Temperament, personality, and life span
development. Journal of Personality and Social Psychology, 78, 173-186.
128
McCrae, R. R., & Sutin, A. R. (2009). Openness to experience. In, M. R. Leary, & R. H.
Hoyle (Eds.), Handbook of Individual Differences in Social Behavior (pp.257-273).
New York, NY, US: Guilford Press, 2009.
Mroczek, D. K., & Spiro, A. (2003). Modeling intraindividual change in personality
traits: Findings from the normative aging study. The Journals of Gerontology:
Psychological Sciences and Social Sciences, 58, 153-165.
Neale, M. C., Boker, S. M., Xie, G., & Maes, H. H. (2003). Mx: Statistical Modeling (6th
Ed.) Richmond, VA: Department of Psychiatry
Pedersen, N. L., & Reynolds, C. A. (1998). Stability and change in adult personality:
Genetic and environmental components. European Journal of Personality, 12, 365-
386.
Pedersen, N. L., Mcclearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., & DeFaire, U.
(1991). The Swedish Adoption/Twin Study of Aging: An update. Acta Geneticae
Medicae et Gemellologiae: Twin Research, 40, 7-20.
Reynolds, C. A., Finkel, D., McArdle, J. J., Gatz, M., Berg, S., & Pedersen, N. L. (2005).
Quantitative genetic analysis of latent growth curve models of cognitive abilities in
adulthood. Developmental Psychology, 41, 3-16.
Roberts, B. W., Walton, K. E., & Vichtbauer, W. (2006). Patterns of mean-level change
in personality traits across the life course: A meta-analysis of longitudinal studies.
Psychological Bulletin, 132, 3-127.
Roberts, B. W., Wood, D., & Caspi, A. (2008). The development of personality traits in
adulthood. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of
Personality: Theory and Research (3rd ed., pp. 375-398). New York, NY: Guilford.
Terracciano, A., McCrae, R. R., Brant, L. J., & Costa, P. T., Jr. (2005). Hierarchical
linear modeling analyses of the NEO-PI-R scales in the Baltimore longitudinal study
of aging. Psychology and Aging, 20, 493-506.
Terracciano, A., McCrae, R. R., & Costa, P. T. (2006). Longitudinal trajectories in
Guilford-Zimmerman temperament survey data: Results from the Baltimore
longitudinal study of aging. The Journals of Gerontology: Psychological Sciences,
61, 108-116.
Viken, R. J., Rose, R. J., Kapiro, J., & Koskenvuo, M. (1994). A developmental genetic
analysis of adult personality: Extraversion and neuroticism from 18 to 59 years of
age. Journal of Personality and Social Psychology, 66, 722-730.
129
SAS Institute. (2000). SAS Release 9.2. Cary, NC: Author.
Sharp, E. S., Reynolds, C. A., Pedersen, N. L., & Gatz, M. (2010). Cognitive engagement
and cognitive aging: Is openness protective? Psychology and Aging, 25, 60-73.
130
CHAPTER 5. GENERAL DISCUSSION
The purpose of this dissertation was to describe the longitudinal course of
openness to experience across the lifespan and to examine sources of stability and change.
Few studies have examined the trajectory of openness and none to my knowledge have
examined sources of change longitudinally. Openness to experience is a personality trait
associated with curiosity, cognitive flexibility and engagement. Further, openness has
been found to be a predictor of cognitive functioning in later life (Sharp et al., 2010).
Previous studies have provided evidence that openness increases between childhood and
adulthood, is relatively stable between age 30 and 60, and then declines after age 60
(Roberts, Watson, & Viechtbauer, 2006). This project focused on providing a better
understanding of why openness declines in older adulthood.
A review of the recent personality literature described the debate surrounding
longitudinal stability and change in personality traits (see Roberts, Wood, & Caspi, 2008).
Much of this debate has focused on the big five personality traits with concentrated focus
on neuroticism, extraversion, and conscientiousness. The most accepted theory of
personality development has been that personality undergoes developmental changes and
maturity throughout childhood and early adulthood, after which traits remain stable
(Costa & McCrae, 1997; McCrae et al., 2000). This model of personality development
has been supported by cross-sectional and longitudinal studies demonstrating rank-order
and mean-level stability after age 30 (Terracciano, Costa, & McCrae, 2006). Further,
scholars of this hypothesis have suggested that because traits follow normative patterns of
131
development, any significant pattern of change after the traditional markers of maturity
would be attributable to primarily biological or genetic influences ((McCrae et al., 2000).
However, the theory of personality stability has been challenged by new studies
presenting evidence of both mean-level and individual-level change in personality across
the lifespan (Roberts & DelVecchio, 2000; Robins, Fraley, Roberts, & Trzesniewski,
2001). A recent review of the literature concluded that personality traits followed
developmental patterns such that mean-level change continued after the traditional
markers of maturity and there was also evidence for significant individual differences, or
fluctuation, in change (Roberts, Watson, & Viechtbauer, 2006). These scholars have
suggested that continued change in personality may be explainable by individual
differences in life events (Roberts, Caspi, & Moffit, 2001; Roberts, Robins, Trzesniewski,
& Caspi, 2003; Mroczek & Spiro, 2003). This conceptualization of personality
development has theorized that stability and change in personality would be the tied to
interactions between genetic and environmental influences (Roberts, Wood, & Caspi,
2008).
The specific aims of this dissertation were divided into three studies. The major
advantages of the current project were the longitudinal twin design, a large sample size,
up to 6 measurement points of openness, and the use of multiple contemporary statistical
techniques to investigate individual differences in change, survival, and genetic and
environmental contributions. The goal of Study 1 was to identify potential individual
differences in intraindividual change in openness and to examine whether sex, education,
self-rated health, activities of daily living, or cardiovascular disease accounted for the
132
observed decline in scores. Results indicated that half of the sample had total fluctuation
of 4 or more points in openness (equal to one standard deviation). However, individual
positive and negative change tended to cancel each other out at the sample level such that
the average yearly rate of decline averaged to -0.05 points per year. This finding was
complimented by evidence of notable within-person variability from the intercept-only
latent growth models. This pattern of individual differences in intraindividual change in
level of openness had not previously been well documented in the literature. Interestingly
when age was added to the growth model, the variability was found to be restricted to the
intercept. Rather than being contradictory, this finding indicated that the individual
differences in intraindividual change were primarily influencing the average level of
openness, whereas the lack of variance around the slope indicated that individuals were
changing in the same direction at a similar rate – suggesting that the decline in openness
was fairly normative across individuals.
Results from the latent growth models centered at age 65 indicated significant
linear and nonlinear decline in openness per decade. Openness was found to decline by
1.5 points per twenty years, generally occurring after age 60 with greater decline
occurring after age 75. Education was important in predicting individual differences in
level of openness, but had no effect on slope suggesting that differences in level were
maintained over time. Males and females were not significantly different in level or rate
of decline. Health related variables (SRH, ADL, CVD) were not associated with level or
change in openness with age. However, the finding that indices of chronic health were
unrelated to the decline in openness does not preclude the potential for other health,
133
environmental or individual life events to account for some portion for the decline in
openness. For example, two potentially important factors that were not investigated were
whether baseline cognitive functioning or retirement status might be related to individual
differences in decline in openness.
The decline of 1.5 points over 20 years might seem of questionable importance
given that the standard deviation for this sample was approximately 4 points. Yet the
decline was more than one third of standard deviation, which has been suggested to be
equal to a small effect in the psychology literature (Meyer et al., 2001). However,
combining the substantial individual-level fluctuations with mean-level decline would
suggest total change greater than one standard deviation - a large effect in psychology.
Thus, focusing only on a mean-level pattern of decline tended to obscure the total amount
of intraindividual change observed throughout adulthood. This acknowledgement of both
mean level and individual level change is in line with Roberts et al. (2006) who argued
that the absolute change accumulated across the lifespan would be equal to or greater
than one standard deviation.
The purpose of Study 2 was to examine whether level or change in openness was
associated with death. Results from two different techniques for modeling growth and
survival indicated that openness was significantly associated with risk for death such that
individuals with steeper declines in openness were at a greater risk for death. This
relationship remained significant after adjusting for age at study entry, sex, and education.
Interestingly, greater openness has also been found to be associated with better cognitive
functioning in older adults. As such, trait openness may aid older individuals in
134
navigating the aging process (Sharp et al., 2010). Given this relationship, decreases in
self-reported openness might be an indicator of broader age-related changes within the
individual. Specifically, the pattern of decline in openness may be reflective of global
declines in individual functioning.
Findings from the survival analyses suggested that individuals who declined in
openness at a faster rate were at an increased risk for death. In examining the pattern of
attrition, individuals who died while in the study had a lower mean openness score at
their last measurement point compared to individuals who remained in the study (as well
as a lower openness score compared to individuals who dropped out of the study but were
still alive at the censoring date). This finding seems to be consistent with changes related
to proximity to death. Gerontologists have suggested that the period of time prior to death
is accompanied by functional declines across multiple domains, including markers of
physical and cognitive health. This period of time has been termed terminal decline (see
Berg, 1996). Further, Berg had hypothesized change in personality within two years of
death. The results of this study were supportive of the theory that personality may show
declines prior to death.
Openness has generally been conceptualized as a “positive” trait (McCrae & Sutin,
2009). Given this trait description, it is challenging to see the decline in openness as
anything other than a detrimental process in older adulthood. Yet, changes in personality
could be conceptualized as part of optimization and compensation processes (Baltes et al.,
1998). In this case, changes in personality might be associated with adaptive responses to
aging-related changes that maximize an individual’s functioning. This might fit with the
135
socioemotional selectivity theory (SST, see Carstensen, 2006; Carstensen, Mickels &
Mather, 2006). SST suggests that individuals change their goals as the time left before
death becomes more relevant. Generally, younger individuals have goals associated with
the gathering of information, experiencing novelty, and expanding breadth of knowledge.
In contrast, as individual gets closer to death, they more likely to reduce and focus their
interests with the goal of optimizing psychological well-being by regulating their
emotions. This model is quite tempting because of the “fit” with openness. In this case,
the fairly normative pattern of decline in openness may reflect a systematic change in
goals across older adults.
The aim of Study 3 was to examine the pattern of genetic and environmental
influence on the longitudinal trajectory of openness to experience. Findings from
biometric growth modeling indicated that genetic influences accounted for a greater
proportion of the individual differences in both level and slope of openness to experience
compared to environmental influences. For both males and females, genetic influences
increased in importance with age. In contrast, environmental influences were of lesser
importance and declined with age, particularly after age 60. Shared and correlated
sources of variation were not statistically important. Males and females differed in the
predicted trajectory of genetic and environmental sources of variation in openness over
time. For males, genetic and nonshared environmental influences increased until age 65,
after which nonshared environmental influences decreased whereas genetic influences
continued to increase. For females, genetic and nonshared environmental influences
diverged from early adulthood such that genetic influences increased, whereas nonshared
136
environmental influences decreased. The finding that genetic influences were greater and
increased with age has also been identified in other domains including memory
(Reynolds et al., 2005) and depression (Gatz etal., 1992).
Results from the biometric analyses suggested that individual differences in both
level and change (centered at age 65) were attributable to genetic factors. McCrae and
Costa have suggested that change in personality traits after age 30 is likely attributable to
genetic influences. Specifically, they have suggested that personality traits are
endogenous tendencies that supersede individual experience (McCrae et al., 2000). In this
context, the finding from several studies that openness declines after middle age would
not be considered “change” in personality, but rather part of a normative pattern of
continued development and perhaps a biologically determined consequence of aging
(McCrae et al., 2000). However, it is known that genes and environment also interact
with each other and that the pattern of interactions can change with age (see Johnson,
Vernon, & Feiler 2008). It is likely that there are multiple GxE interactions that influence
the decline in openness across adulthood and these influences remain unidentified.
The overall results highlight both similarities and differences with previous
investigations of SATSA data. In our recent study using SATSA data to investigate the
relationship between openness and cognition, there was evidence that females declined in
openness with age, whereas males did not decline (Sharp, Reynolds, Pedersen, & Gatz,
2010). Results of the present study failed to identify sex differences in the level or rate of
change in openness over age. It is likely that this difference is related to the respective
samples. The current study included all individuals with openness information (N=1947)
137
and up to six measurement occasions, whereas the previous study was focused on
individuals older than 65 with cognitive data (N=857) and up to five measurement
occasions, thus the current study had a greater number of older participants and also more
participants had died –allowing for more decline in openness.
Similar to the current findings, Pedersen and Reynolds (1998) identified mean-
level stability in openness in adults less than 55 years old and a mean-level decline over
time after age 55. In contrast to the current study, findings from the genetic analyses
suggested that nonshared environment accounted for more of the variance over time and
genetic contributions were stable or declined over time. Consistent with the current study,
the genetic and environmental components to openness over the lifespan differed by sex.
The differences between the present study and Pedersen and Reynolds are likely
attributable to differences in statistical approach, specifically that the current study used
latent growth models of change over age (rather than time), included a quadratic term to
examine acceleration in change, and had two additional measurement points available.
Pedersen and Reynolds concluded that intraindividual variability, time-to-death factors,
increased number of measurement points, and latent growth models were important areas
for future study. The current study was able to address all of these questions.
This dissertation advanced the current personality and aging literature by using a
multi-method approach to examine sources of stability and change in openness to
experience. Three studies were designed to incorporate contemporary statistical methods
to examine the effects of demographic, indices of chronic health, death, and genetic and
environmental influences on the trajectory of openness to experience. Results of the first
138
study indicated that openness exhibits small but significant mean-level declines after age
65 and also identified substantial individual-level fluctuations in openness across ages.
Individual differences in sex, education, SRH, ADL, CVD were unrelated to this decline.
Thus, age explained the mean-level decline in openness but not individual fluctuation.
Study 2 revealed that level of openness was unrelated to risk for death, but a faster rate of
decline in openness (i.e. change) was associated with an increased risk for death,
controlling for age, sex, and education. This finding fit within a terminal decline model of
aging. Given that openness is a trait linked to cognitive engagement, decline may herald
broader age-related changes within the individual. Yet, it is also possible that the decline
reflects an adaptive strategy in older adulthood. Results of Study 3, indicated that genetic
influences accounted for most of the variance in individual differences in both level and
slope in openness. This finding supports one hypothesis in the literature that change in
personality during adulthood is primarily influenced by genetic processes. To date,
openness has been an understudied trait in the literature. The contributions of the three
studies presented here enhance our understanding of stability and change in adult
personality across the lifespan.
139
COMPREHENSIVE REFERENCES
Almada, S. J., Zonderman, A. B., Shekelle, R. B., & Dyer, A. R. (1991). Neuroticism and
cynicism and risk of death in middle-aged men: The western electric study.
Psychosomatic Medicine, 5, 165-175.
Alwin, D. F. (1994). Aging, personality, and social change: The stability of individual
differences over the adult span. In D. L. Featherman, R. M. Lerner, & M. Perlmutter
(Eds.), Life-Span Development and Behavior (Vol. 12, pp. 135-185). Hillsdale, NJ;
Lawrence Erlbaum.
Baltes, P. B. (1987). Theoretical propositions of life-span development psychology: on
the dynamics between growth and decline. Developmental Psychology, 23, 611-626.
Baltes, P. B., & Nesselroade, J. R. (1973). The developmental analysis of individual
differences on multiple measures. In J. R. Nesselroade & H. W. Reese (Eds.), Life-
span developmental psychology: Methodological issues (pp. 219-251). New York:
Academic Press.
Baltes, P. B. (1997). On the incomplete architecture of human ontogeny. American
Psychologist, 52, 366–380.
Baltes, P. B., Lindenberger, U., & Staudinger, U. M. (1998). Life-span theory in
developmental psychology. In W. Damon & R. M. Lerner (Eds.), Handbook of Child
Psychology (5th ed., Vol. 1, pp. 1029–1143). New York: Wiley.
Baltes, P. B., & Baltes, M. M. (1990). Psychological perspectives on successful aging:
The model of selective optimization with compensation. In P. B. Baltes & M. M.
Baltes (Eds.), Successful Aging: Perspectives from the Behavioral Sciences (pp. 1–
34). Cambridge, U.K.: Cambridge University Press.
Berg, S. (1996). Aging, behavior, and terminal decline. In J.E. Birren & K. W. Schaie
(Eds.), Handbook of the Psychology of Aging. (4th ed., pp. 323-337). San Diego:
Academic Press.
Bergeman, C. S., Chipuer, H. M., Plomin, R., Pedersen, N. L., McClearn, G. E.,
Nesselroade, J. R., Costa, P. T. Jr., & McCrae, R. R. (1993). Genetic and
environmental effects on openness to experience, agreeableness, and
conscientiousness: An adoption/twin study. Journal of Personality, 61, 159-179.
Bleidorn, W., Kandler, C., Reimann, R., Angleitner, A., & Spinath, F. (2009). Patterns
and sources of adult personality development: Growth curve analyses of NEO-PI-R
140
scales in a longitudinal twin study. Journal of Personality and Social Psychology,
87, 142-155.
Booth, J. E., Schinka, J. A., Brown, L. M., Mortimer, J. A., & Borenstein, A. R. (2006).
Five-factor personality dimensions, mood states, and cognitive performance in older
adults. Journal of Clinical and Experimental Neuropsychology, 28, 676–683.
Bouchard, T. J., & Loehlin, J. C. (2001). Genes, evolution, and personality. Behavior
Genetics, 31, 243-273.
Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to
assessing change. Psychological Bulletin, 101, 147-158.
Carstensen, L. L. (2006). The influence of a sense of time on human development.
Science, 312, 1913–1915.
Carstensen, L. L., Mikels, J. A., & Mather, M. (2006). Aging and the intersection of
cognition, motivation and emotion. In J. Birren & K. W. Schaie (Eds.), Handbook
of the Psychology of Aging (6th ed., pp. 343–362). San Diego, CA: Academic
Press.
Caspi, A. & Roberts, B. W. (2001). Personality development across the life: The
argument for change and continuity. Psychological Inquiry, 12, 49-66.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and
change. Annual Review of Psychology, 56, 453-484.
Cederlof, R., Friberg, L., & Lundman, T. (1977). The interactions of smoking,
environment and heredity and their implications for disease etiology. Acta Medica
Scandinavia, 612, 1-128.
Christensen, A. J., Ehlers, S. L., Wiebe, J. S., Raichle, K., Ferreyhough, K., Lawton, W.
J. (2002). Patient personality and mortality: A 4-year prospective examination of
chronic renal insufficiency. Health Psychology, 21, 315-320.
Costa, P. T., & McCrae, R. R. (1985). The NEO Personality Inventory Manual. Odessa,
FL: Psychological Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992a). Revised NEO personality inventory (NEO PI-R)
and NEO five-factor inventory (NEO-FFI). Odessa, Florida: Psychological
Assessment Resources.
Costa, P. T., & McCrae, R. R. (1992b). Multiple uses for longitudinal personality data.
European Journal of Longitudinal Research and Personality, 6, 85-102.
141
Costa, P. T. & McCrae, R. R. (1994). Set like plaster? Evidence for the stability of adult
personality. In T. F. Heatherton & J. L. Weinberger (Eds.), Can Personality
Change? (pp. 21-40). Washington, DC: American Psychological Association.
Costa, P. T., & McCrae, R. R. (1997). Longitudinal stability of adult personality. In R.
Hogan, J. A. Johnson, & S. R. Brigss (Eds.), Handbook of Personality Psychology
(pp. 269-290). San Diego: Academic Press.
Dudek, F. J. (1979). The continuing misinterpretation of the standard error of
measurement. Psychological Bulletin, 86, 335-337.
Elder, G.H. (1975). Age differentiation and life course. Annual Review of Sociology, 1,
65-190.
Elder, G. H. (1998). The life course as developmental theory. Child Development, 69, 1-
12.
Finkel, D., Reynolds, C. A., McArdle, J. J., Gatz, M., & Pedersen, N. L. (2003). Latent
growth curve analyses of accelerating decline in cognitive abilities in late adulthood.
Developmental Psychology, 39, 535-550.
Finkel, D., Andel, R., Gatz, M., & Pedersen, N. L. (2009). The role of occupational
complexity in trajectories of cognitive aging before and after retirement. Psychology
& Aging, 24, 563-574.
Fleeson, W., & Jolley, S. (2006). A proposed theory of the adult development of
intraindividual variability in trait-manifesting behavior. In D. Mroczek & T. D. Little
(Eds.), Handbook of Personality Development (pp. 41-59). Mahwah, NJ: Lawrence
Erlbaum Associates.
Flouderus-Myrhed, B., Pedersen, N., & Rasmuson, I. (1980). Assessment of heritability
for personality, based on a short-form of the Eysenck Personality Inventory: A study
of 12,898 twin pairs. Behavior Genetics, 10, 153-162.
Fraley, C., & Roberts, B. W. (2005). Patterns of continuity: A dynamic model for
conceptualizing the stability of individual differences in psychological constructs
across the life course. Psychological Review, 112, 60-74;
Freund, A. M., & Baltes, P. B. (2007). Toward a theory of successful aging: Selection,
optimization, and compensation. In R. Fernandez-Ballesteros (Ed.),
Geropsychology: European Perspectives for an Aging World (pp. 239–254).
Cambridge, MA: Hogrefe & Huber.
142
Ghisletta, P. (2008). Application of a joint multivariate longitudinal-survival analysis to
examine the terminal decline hypothesis in the Swiss interdisciplinary longitudinal
study of the oldest old. Journal of Gerontology: Psychological Sciences, 63, 185-
192.
Ghisletta, P., McArdle, J. J., & Lindenberger, U. (2006). Longitudinal cognition-survival
relations in old and very old age. European Psychologist, 11, 204-223.
Guo, X., & Carlin, B (2004). Separate and joint modeling of longitudinal and event time
data using standard computer packages (statistical practice). The American
Statistician, 58, 1-9.
Harris, J. R., Pedersen, N. L., McClearn, G. E., Plomin, R., & Nesselroade, J. R. (1992).
Age differences in genetic and environmental influences for health from the Swedish
Adoption/ Twin Study of Aging. Journal of Gerontology: Psychological Sciences,
47, 213-220
Helson, R., Kwan, V. S. Y., John, O. P., & Jones, C. (2002). The growing evidence for
personality change in adulthood: Findings form research with personality
inventories. Journal of Research in Personality, 36, 287-306.
Hultsch, D. F., Strauss, E., Hunter, M. A., & MacDonald, S. W. S. (2008). Intraindividual
variability, cognition, and aging. In F. I. M. Craik & T. A. Salthouse (Eds.),
Handbook of Cognition and Aging (3rd ed., pp. 491-556). New York, NY, US:
Psychology Press.
Johansson, B., & Berg, S. (1989). The robustness of the terminal decline phenomenon:
Longitudinal data from the digit-span memory test. Journal of Gerontological and
Psychological Science, 44, 184-186.
John, O. P., & Strivasta, S. (1999). The Big Five trait taxonomy: history, measurement,
and theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of
Personality (pp. 102-138). New York: Guilford.
Johnson, A. M., Vernon, P. A., Feiler, A. R.. (2008). Behavior genetics 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, Vol 1: Personality Theories and Models (pp.
145-173). Thousand Oaks, CA: Sage
Jonassaint, C. R., Boyle, S. H., Williams, R. B., Mark, D. B., Siegler, I. C., & Barefoot, J.
C. (2007). Facets of openness predict mortality in patients with cardiac disease.
Psychosomatic Medicine, 69, 319-322.
143
Krueger, R. F., Johnson, W., & Kling, K. C. (2006). Behavior genetics and personality
development. In D. K. Mroczek & T. D. Little (Eds.), Handbook of Personality
Development (pp. 81-108). Mahwah, NJ: Erlbaum.
Lewis, M. (2001). Issues in the study of personality development. Psychological Inquiry,
12, 67-83.
Little, R. T. A. (1995). Modeling the dropout mechanism in repeated-measures studies.
Journal of the American Statistical Association, 90, 1112-1121.
McArdle, J. J. (2006). Latent curve analyses of longitudinal twin data using a mixed-
effects biometric approach. Twin Research and Human Genetics, 9, 343-359.
McArdle, J. J. & Anderson, E. (1990). Latent variable growth models for research on
aging. In J. E. Birren & K. W. Schaie (Eds.), The Handbook of the Psychology of
Aging (pp. 21-43). New York: Plenum Press.
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R. W. (2002). Comparative
longitudinal structural analyses of the growth and decline of multiple intellectual
abilities over the life span. Developmental Psychology, 38, 115-142.
McArdle, J. J., & Hamagami, F. (1992). Modeling incomplete longitudinal and cross-
sectional data using latent growth structural models. Experimental Aging Research,
18, 145-166.
McArdle, J. J. & Hamagami, F. (2003). Structural equation models for evaluating
dynamic concepts within longitudinal twin analyses. Behavior Genetics, 33, 137-
159.
McArdle, J. J., Hamagami, F., Jones, K., Jolesz, F., Kikinis, R., Spiro, A., & Albert, M.
S. (2004). Structural modeling of dynamic changes in memory and brain structure
using longitudinal data from the normative aging study. Journal of Gerontology:
Psychological Sciences, 59, 294-304.
McArdle, J. J., Prescott, C. A., Hamagami, F., & Horn, J. L. (1998). A contemporary
method for developmental-genetic analyses of age changes in intellectual abilities.
Developmental Neuropsychology, 14, 69-114.
McArdle, J. J., Small, B. J., Backman, L., & Fratiglioni, L. (2005). Longitudinal models
of growth and survival applied to the early detection of Alzheimer’s disease. Journal
of Geriatric Psychiatry and Neurology, 18, 234-241.
McCrae, R. R. (1994). Openness to experience: Expanding the boundaries of factor V.
European Journal of Personality, 8, 251-272.
144
McCrae, R. R. & Costa, P. T., Jr. (1990). Personality in Adulthood. New York: Guilford
Press
McCrae, R. R. & Costa, P. T., Jr. (1997). Conceptions and correlates of Openness to
Experience. In R. Hogan, J. A. Johnson, & S. R. Briggs (Eds.), Handbook of
Personality Psychology (pp. 825-847). Orlando, FL: Academic Presss.
McCrae, R. R., & Costa, P. T. (2008). Empirical and theoretical status of the five-factor
model of personality traits. In G. J. Boyle, Mathews, G., & Sakofske, D. H. (Eds.),
The SAGE handbook of personality theory and assessment Vol 1: Personality
theories and models (pp. 273-294). Thousand Oaks, CA, US: Sage Publications, Inc.
McCrae, R. R., Costa, P. T., Jr., Ostendorf, F., Angleitner, A., Hrebícková, M., Avia, M.
D., et al. (2000). Nature over nurture: Temperament, personality, and life span
development. Journal of Personality and Social Psychology, 78, 173-186.
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its
applications. Journal of Personality, 60, 175-175.
McCrae, R. R., & Sutin, A. R. (2009). Openness to experience. In, M. R. Leary, & R. H.
Hoyle (Eds.), Handbook of Individual Differences in Social Behavior (pp.257-273).
New York, NY, US: Guilford Press, 2009.
Meyer, G. J., Finn, S. E., Eyde, L. D., Kay, G. G., Moreland, K. L., Dies, R. R. et al.
(2001). Psychological testing and psychological assessment. American Psychologist,
56, 128-165.
Mikels, J. A., Löckenhoff, C. E., Maglio, S. J., Carstensen, L. L., Goldstein, M. K., &
Garber, A. (2010). Following your heart or your head: Focusing on emotions
versus information differentially influences the decisions of younger and older
adults. Journal of Experimental Psychology: Applied, 16, 87-87-95.
Mroczek, D. K., Almeida, D. M., Spiro, A., & Pafford, C. (2006). Modeling
intraindividual stability and change in personality. Mahwah, NJ, US: Lawrence
Erlbaum Associates Publishers.
Mroczek, D. K., & Spiro, A. (2003). Modeling intraindividual change in personality
traits: Findings from the normative aging study. The Journals of Gerontology:
Psychological Sciences and Social Sciences, 58, 153-165.
Mroczek, D. K., & Spiro, A. (2007). Personality change influences mortality in older
men. Psychological Science, 18, 371-376.
145
Mroczek, D. K., Spiro, A., & Almeida, D. M. (2003). Between- and within- person
variation in affect and personality over days and years: How Basic and applied
approaches can inform one another. Ageing International, 28, 260-278.
Mroczek, D. A., Spiro, A., III, & Griffin, P. W. (2006). Personality and Aging. In J. E.
Birren and K. W. Shaie (Eds.), Handbook of the Psychology of Aging (6th ed., pp.
363-377).
Mroczek, D. K., Spiro, A., & Turiano, N. A. (2009). Do health behaviors explain the
effect of neuroticism on mortality? Longitudinal findings from the VA normative
aging study. Journal of Research in Personality, 43(4), 653-659.
Muthén, L. K., & Muthén, B. O. (1998-2007). Mplus User’s Guide. Sixth Edition. Los
Angles, CA: Muthén & Muthén
Neale, M. C., Boker, S. M., Xie, G., & Maes, H. H. (2003). Mx: Statistical Modeling
(6th Ed.) Richmond, VA: Department of Psychiatry
Nesselroade, J. R. (1991). The warp and woof of the developmental fabric. In R. Downs,
L. Liben, & D. S. Palermo (Eds.), Visions of aesthetics, the environment, and
development: The legacy of Joachim F. Wohlwill (pp. 213-240). Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.
Pedersen, N. L., & Harris, J. R. (1990). Functional capacity and activities of daily living.
Behavior Genetics, 20, 740.
Pedersen, N. L., Mcclearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., & DeFaire, U.
(1991). The Swedish Adoption/Twin Study of Aging: An update. Acta Geneticae
Medicae et Gemellologiae: Twin Research, 40, 7-20.
Pedersen, N. L., & Reynolds, C. A. (1998). Stability and change in adult personality:
Genetic and environmental components. European Journal of Personality, 12, 365-
386.
Pedersen, N. L., Mcclearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., & DeFaire, U.
(1991). The Swedish Adoption/Twin Study of Aging: An update. Acta Geneticae
Medicae et Gemellologiae: Twin Research, 40, 7-20.
Reynolds, C. A., Finkel, D., McArdle, J. J., Gatz, M., Berg, S., & Pedersen, N. L. (2005).
Quantitative genetic analysis of latent growth curve models of cognitive abilities in
adulthood. Developmental Psychology, 41, 3-16.
146
Roberts, R. W., Caspi, A., & Moffitt, T. E. (2001). The Kids are alright: Growth and
stability in personality development from adolescence to adulthood. Journal of
Personality and Social Psychology, 81, 670-683.
Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality
traits from childhood to old age: A quantitative review of longitudinal studies.
Psychological Bulletin, 126, 3-25.
Roberts, B. W., Kuncel, N. R., Shiner, R., Caspi, A., & Goldberg, L. R. (2007). The
power of personality. The comparative validity of personality traits, socioeconomic
status, and cognitive ability for predicting important life outcomes. Perspectives on
Psychological Science, 2, 313-345.
Roberts, B. W., & Mroczek, D. (2008). Personality trait change in adulthood. Current
Directions in Psychological Science, 17, 31-35.
Roberts, B. W., Robins, R. W., Trzesniewski, K., & Caspi, A. (2003). Personality trait
development in adulthood. In J. Mortimer, & M. Shanahan (Eds.), Handbook of the
life course (pp. 579-598). New York: Kluwer Acad.
Roberts, B. W., Walton, K. E., & Vichtbauer, W. (2006). Patterns of mean-level change
in personality traits across the life course: A meta-analysis of longitudinal studies.
Psychological Bulletin, 132, 3-127.
Roberts, B. W., Wood, D., & Caspi, A. (2008). The development of personality traits in
adulthood. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of
Personality: Theory and Research (3rd ed., pp. 375-398). New York, NY: Guilford.
Robins, R. W., Fraley, R. C., Roberts, B. W., & Trzesniewski, K. (2001). A longitudinal
study of personality in young adulthood. Journal of Personality, 69, 617-640.
Rose, G., McCartney, P., & Reid, D. D. (1977). Self-administration of a questionnaire on
chest pain and intermittent claudication. British Journal of Preventive and Social
Medicine, 31, 42-48.
SAS Institute. (2000). SAS Release 9.2. Cary, NC: Author.
Schaie, K. W., Willis, S. L., & Caskie, G. I. L. (2004). The Seattle Longitudinal study:
Relationship between personality and cognition. Aging, Neuropsychology, and
Cognition, 11, 304–324.
Sharp, E. S., Reynolds, C. A., Pedersen, N. L., & Gatz, M. (2010). Cognitive engagement
and cognitive aging: Is openness protective? Psychology and Aging, 25, 60-73.
147
Singer, J. D., & Willet., J. B. (2003). Applied longitudinal data analysis: Modeling
change and event occurrence. Oxford, UK: Oxford University Press.
Small, B. J., Fratiglioni, L., von Strauss, E., & Backman, L. (2003). Terminal decline and
cognitive performance in very old age: Does cause of death matter? Psychology and
Aging, 18, 193-202.
Small, B. J., Hertzog, C., Hultsch, D. F. & Dixon, R. A. (2003). Stability and change in
adult personality over 6 years: Findings from the Victoria Longitudinal Study.
Journals of Gerontology: Psychological Sciences and Social Sciences, 58, 166-176.
Srivastava, S., John, O. P., Gosling, S. D., Potter, J. (2003). Development of personality
in early and middle adulthood: Set like plaster or persistent change? Journal of
Personality and Social Psychology, 84, 1041-1053.
Svedberg, P., Gatz, M., Lichtenstein, P., Sandin, S., & Pedersen, N. L. (2009). Self-rated
health in a longitudinal perspective: A 9-year follow-up twin study. Journal of
Gerontology: Social Sciences, 60, 331-340.
Terracciano, A., Costa, P. T., & McCrae, R. R. (2006). Personality plasticity after age 30.
Personality and Social Psychology Bulletin, 32, 999-1009.
Terracciano, A., McCrae, R. R., Brant, L. J., & Costa, P. T., Jr. (2005). Hierarchical
linear modeling analyses of the NEO-PI-R scales in the Baltimore longitudinal study
of aging. Psychology and Aging, 20, 493-506.
Terracciano, A., McCrae, R. R., & Costa, P. T. (2006). Longitudinal trajectories in
Guilford-Zimmerman temperament survey data: Results from the Baltimore
longitudinal study of aging. The Journals of Gerontology: Psychological Sciences,
61, 108-116.
Terracciano, A., McCrae, R. R., & Costa, P. T. Jr. (2010). Intra-individual change in
personality stability and age. Journal of Research in Personality, 44, 31-37.
Weiss, A., & Costa, P. T. (2005). Domain and facet personality predictors of all-cause
mortality among Medicare patients aged 65 to 100. Psychosomatic Medicine, 67,
715-723.
Williams, P. G., Smith, T. W., & Cribbet, M. R. (2008). Personality and health: Current
evidence, potential mechanisms, and future directions. In G. J. Boyle, G. Matthews,
& D. H. Saklofske (Eds.), The SAGE Handbook of Personality Theory and
Assessment, Vol 1: Personality Theories and Models (pp. 145-173). Thousand Oaks,
CA: Sage
148
Wilson, R. S., Mendes de Leon, C. F., Bienias, J. L., Evans, D. A., & Bennett, D. A.
(2004). Personality and mortality in old age. The Journals of Gerontology:
Psychological Sciences and Social Sciences, 58, 110-116.
Viken, R. J., Rose, R. J., Kapiro, J., & Koskenvuo, M. (1994). A developmental genetic
analysis of adult personality: Extraversion and neuroticism from 18 to 59 years of
age. Journal of Personality and Social Psychology, 66, 722-730.
149
APPENDIX A: Items from the Openness to Experience Scale
1. I like to solve problems or riddles.
2. I find it easy to empathize with others.
3. I have great intellectual curiosity.
4. I find it interesting to take up new hobbies.
5. I like to ponder on theories and/or philosophical ideas.
6. I often try out new and foreign foods.
150
APPENDIX B: Items from the Self Rated Health Scale (SRH)
1. How would you rate your general health status?
2. How would you rate your general health status compared to 5 years ago?
3. How would you rate your health status compared to others in your age group?
4. Do you think your health prevents you from doing things you would like to do?
151
APPENDIX C: Items from the Activities of Daily Living Scale (ADL)
1. Can able to use phone
2. Can able to go out of walking distance
3. Can shop for groceries/clothes
4. Can prepare meals
5. Can do own housework
6. Can take own medicine
7. Can handle own money
8. Can eat
9. Can dress and undress self
10. Can take care of appearance
11. Can walk
12. Can get in and out of bed
13. Can take a bath or shower
14. Can get to the bathroom on time
152
APPENDIX D: Items for Cardiovascular Disorder Scale (CVD)
1. Have had angina pectoris
2. Have had heart infarct
3. Have had claudicatio
4. Have had high blood pressure
5. Have had heart insufficiency
6. Have had heart attack
7. Have had phlebitis
8. Have had circulation problems in limbs
9. Have had thrombosis
10. Have had a stroke
11. Have had tachycardia
12. Have had heart operation
13. Have had heart valve problem
Abstract (if available)
Abstract
This dissertation examined sources of stability and change in openness to experience across the lifespan. The data came from the Swedish Adoption/Twin Study of Aging (SATSA), a large longitudinal study of twins that included up to six measurement occasions of openness. Study 1 examined the longitudinal trajectory of openness and variables hypothesized to account for individual differences in level and change in openness. Using twins as individuals (and adjusting for the correlation between twins in the modeling), results from phenotypic latent growth curve modeling confirmed previous literature that openness exhibits small but significant decline in older adulthood. Model estimates indicated a total decline of about 1.5 points in openness over twenty years, generally occurring after age 60. Thus, age accounted for the decline in openness. Males and females were not significantly different in level or rate of decline. Education was important in predicting individual differences in level of openness, but not decline. Self-rated health, activities of daily living, and cardiovascular disease did not explain individual differences in level or decline in openness. Study 2 examined the relationship between openness to experience and death. Results from two methods of modeling growth and survival, adjusted for entry age, sex, education and twinness, suggested that the slope of openness was significantly associated with death such that individuals who had a faster rate of decline were at a greater risk for death. Level of openness (mean openness at age 65) was unrelated to risk for death. The overall results from study 1 and study 2 suggested that decline in openness was normative across older participants and is perhaps best conceptualized in relation to theories of terminal decline - the period of time prior to death during which multiple domains have been found to decline. Study 3 examined the pattern of genetic and environmental influences across the longitudinal trajectory of openness. Results from a biometric latent growth model (a latent growth model combined with a Cholesky decomposition model) suggested that a reduced additive genetic (A) and nonshared environment (E) model provided the best fit. The AE model estimates suggested that the individual differences in level, linear slope, and quadratic slope were primarily accounted for by genetic influences. For both men and women additive genetic influences generally increased with age. For men, nonshared environmental influences increased similar to genetic influences but then declined after age 65. For women, the contribution of nonshared environmental influences declined steadily from age 40. The findings from Study 3 give initial support to one theory of personality development that has suggested that normative patterns of change in adult personality would be explained by primarily genetic influences.
Conceptually similar
PDF
Personality and cognitive aging: are positive traits protective?
PDF
Dementia, co-morbid depressive symptoms, and mortality in older adults: a twin study
PDF
Multivariate biometric modeling among multiple traits across different raters in child externalizing behavior studies
PDF
Patterns of EEG spectral power in 9-10 year old twins and their relationships with aggressive and nonaggressive antisocial behavior in childhood and adolescence
PDF
Attrition in longitudinal twin studies: a comparative study of SEM estimation methods
PDF
Heritability of schizotypal traits during adolescence
PDF
Dedifferentiation of emotion regulation strategies in the aging brain: an MVPA investigation
PDF
Influence of age and anxiety on recognition of facial expressions of emotion: exploring the role of attentional processes
PDF
Dynamic analyses of the interrelationship between mothers and daughters on a measure of depressive symptoms
PDF
Examining the effect of fertility on lifespan using data from the Swedish Twin Registry
PDF
An exploratory study of physical activity across different adult groups
PDF
Family conflict, negative mood, and adolescents' daily problems in school
PDF
Developmental etiology of reactive and proactive aggression, electrodermal reactivity and their relationships: a twin study
PDF
Statistical algorithms for examining gene and environmental influences on human aging
PDF
Age-related androgen depletion and the development of Alzheimer's disease
PDF
Evaluating social-cognitive measures of motivation in a longitudinal study of people completing New Year's resolutions to exercise
PDF
Aging and emotion regulation in the judgment of facial emotion
PDF
Effects of AT-1 receptor blockers on cognitive decline and Alzheimer's disease
PDF
The effects of time and space on health status of older adults in China
PDF
Nice by nature? A twin study of the development and physiology of prosocial personality
Asset Metadata
Creator
Sharp, Emily Schoenhofen
(author)
Core Title
Sources of stability and change in the trajectory of openness to experience across the lifespan
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Psychology
Publication Date
04/16/2014
Defense Date
11/09/2011
Publisher
Los Angeles, California
(original),
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
aging,genetics,mortality,OAI-PMH Harvest,openness to experience,personality development,Twins
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Gatz, Margaret (
committee chair
), Knight, Bob G. (
committee member
), McArdle, John J. (
committee member
), Pedersen, Nancy L. (
committee member
), Silverstein, Merril (
committee member
)
Creator Email
emschoen@gmail.com,schoenho@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-6495
Unique identifier
UC1112360
Identifier
usctheses-c3-6495 (legacy record id)
Legacy Identifier
etd-SharpEmily-611.pdf
Dmrecord
6495
Document Type
Dissertation
Rights
Sharp, Emily Schoenhofen
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
genetics
mortality
openness to experience
personality development