Journal of Gerontology: PSYCHOLOGICAL SCIENCES
2003, Vol. 58B, No. 3, P143–P152
Copyright 2003 by The Gerontological Society of America
Positive and Negative Affect in Very Old Age
Derek M. Isaacowitz1 and Jacqui Smith2
2
1
Department of Psychology, University of Pennsylvania, Philadelphia.
Center for Lifespan Psychology, Max Planck Institute for Human Development, Berlin, Germany.
The current study examined two issues involving the relationship between age and affect in very old age using data
from men and women (aged 70 to 1001 years, M 5 85 years) in the Berlin Aging Study (BASE). The first issue was
whether unique effects of age on positive and negative affect remained after we controlled for other variables that
would be expected to relate to affect in late life. We found no unique effects of age after we controlled for
demographic, personality, and health and cognitive functioning variables. Personality and general intelligence
emerged as the strongest predictors of positive and negative affect. Second, we evaluated patterns within
meaningful subgroups: young old versus oldest old and men versus women. Subgroup differences in predictor
patterns were minimal. Although we accounted for much of the age-related variance in positive and negative
affect, a significant amount of variance in the affect of older adults remained unexplained.
R
ECENTLY, psychologists studying the relationship
between age and affect have found evidence for
a ‘‘paradox of well-being’’ at least for the middle aged and
young old: Despite the presence of individual difference
variables (such as poor health) that disproportionately affect
older people and thus should lead to worse well-being in older
adults, mean-level age differences in well-being are null or
positive (Carstensen, Pasupathi, Mayr, & Nesselroade, 2000;
Mroczek & Kolarz, 1998). This apparent paradox highlights the
ways in which age and individual differences may be
intertwined in the study of well-being. Two empirical questions
emerge when one attempts to further disentangle the complicated relationship between age and affect.
First is the question of so-called unique age affects: Do
differences in mean levels of affect between individuals of
different ages exist when other possible sources of variance
are controlled for, including those individual difference variables known to predict affect more generally? Perhaps more
important, is there anything left to predict, by age or any other
variables, after age-related variables are entered into a model
predicting well-being? Age is a carrier for other variables and
is itself not a causal factor in psychological processes (P. B.
Baltes, Reese, & Nesselroade, 1977; Wohlwill, 1973). A typical
research strategy of life span developmentalists is to try to
explicate age-related variance. Mroczek and Kolarz (1998), for
example, found unique age effects on affect even when they controlled for a long list of age-related predictors. According to
life span developmental theory, this finding indicates that the
study may not have captured the complete set of age-related
predictors in their assessment.
Second, are there meaningful subgroups in which the profile
of affect predictors differs? For example, predictors of affect in
old age may differ by gender, or may vary in the young old and
oldest old. Together, analyses of these two questions can help
clarify how affect is and is not different in adults of different
ages and can help us understand both the apparent paradox as
well as its limitations.
A recent study on the relationship of age and affect in adults
aged 25–74 years considered each of the three above
approaches in its analyses (Mroczek & Kolarz, 1998). First,
the authors considered whether there was unexplained age-
related variance in positive and negative affect in the large,
representative sample from the MacArthur Study of Midlife
Development (MIDMAC) after they controlled for other
possible predictors of affect in middle-aged individuals (i.e.,
demographics, personality, and contextual factors particular
to midlife, such as occupational stress). Unique age effects did
emerge after controls such that older individuals, in general,
reported higher levels of positive affect and lower levels of
negative affect. However, the authors wondered whether these
effects might differ by gender, and investigated whether the
pattern of predictors and the unique age effects would hold true
for both men and women. Finally, the authors examined the
interaction of age and each positive predictor of affect in the
full sample. These analyses together suggested differential patterns of age–affect relationships based on gender, personality,
and demographics, as age interacted with marital status and
extraversion in predicting affect in males only (Mroczek &
Kolarz, 1998).
Although such in-depth analyses may only be possible in
large data sets such as that of MIDMAC, a consideration
of both age and age-relevant individual difference-type life contexts appears fundamental to understanding the unfolding of
affect in adulthood and old age. More important, these contexts may differ substantially by age, even between the range of
individuals studied in MIDMAC (those aged 25–74 years) and
their older peers. Advanced old age, for example, is a time of
life unlike middle age and even early old age (P. B. Baltes,
1997; P. B. Baltes & Smith, 2003; Smith, 2001a; Suzman,
Manton, & Willis, 1992). Unlike earlier in adulthood, when
most individuals experience good health, and levels of
disability and medical comorbidity are rather low, rates of
morbidity and functional impairment are impressively high
overall among those adults aged 80 and older (SteinhagenThiessen & Borchelt, 1999; Suzman et al., 1992). Similarly,
after age 75, few cognitive abilities appear spared from agerelated decline (Lindenberger & P. B. Baltes, 1997; Schaie,
1996). These age-related changes may in turn influence the
predictors of individual differences in affective experience. So,
although some traditional predictors (e.g., personality traits)
may predict affect in very old age as they do in earlier
adulthood, other affect predictors are likely unique to different
P143
P144
ISAACOWITZ AND SMITH
times of life. We expect additional factors such as physical
health, functional capacity, and cognitive functioning, for example, to be particularly important predictors of individual differences among the oldest old.
Below, we review evidence on predictors of affect other than
age in adults of different ages. Then, we consider previous
studies on age differences in affect, and we turn our attention
to what other than age should predict affect in late life, with
particular attention to differences between the young old and
the oldest old. Finally, we present our approach to studying
affect in the context of the Berlin Aging Study’s (BASE)
unique sample of older individuals.
Affect and Its Predictors in Adulthood
Personality may be the most consistent predictor of individual
differences in affect among adults regardless of age (Costa &
McCrae, 1980). Mroczek and Kolarz (1998), for example, found
that neuroticism and extraversion added more than any other
predictor, including age, to their model predicting positive and
negative affect in a large sample of middle-aged adults. A recent meta-analysis of the relationship between personality and
well-being foundstrong, consistent associations, particularly linking neuroticism with negative affect, and extraversion and
agreeableness with positive affect (DeNeve & Cooper, 1998).
Personality may influence affect directly through affective setpoints (Lykken, 1999) or indirectly by impacting an individual’s
environmental preferences and behavioral patterns. Interestingly, Costa, McCrae, and colleagues (McCrae et al., 1999;
Yang, McCrae, & Costa, 1998) have reported cross-cultural data
indicating small yet significant decreases in both extraversion
and neuroticism between younger and older adults. These age
differences, although small in magnitude, may have implications
for the contribution of personality factors to individual differences in affect late in life.
It is important to note, however, that personality does not
exclusively predict affect, even among middle-aged adults. In
the MIDMAC data, occupational and relative stress, health, and
age all predicted affect after personality was controlled for
(Mroczek & Kolarz, 1998). Other studies have found relationships with parents and children to be important determinants of
well-being among middle-aged individuals (Ryff, Lee, Essex,
& Schmutte, 1994; Welsh & Stewart, 1995). Work and family
are the most salient life domains to the middle-aged individual
(Lachman, 2001), so it is not surprising that variations in those
domains should affect individual differences in well-being
above and beyond personality in that age group.
Affect in Old Age
Although early cross-sectional studies on affect found few
differences between older adults and younger adults (Lawton,
Kleban, Rajagopal, & Dean, 1992; Malatesta & Kalnok, 1984),
more recent studies have presented a more positive view of the
affective life of older individuals. Charles, Reynolds, and Gatz
(2001) found that participants followed over 23 years from their
mid-60s to their mid-80s showed a small decrease in positive
affect (see also Stacey & Gatz, 1991). Negative affect declined
until age 60. Carstensen and colleagues (2000) found that older
adults were more likely to remain in positive emotional states
and to continue not experiencing negative emotional states over
time than were younger adults. However, no age differences
were found in this sample of 18–94 year olds in the rated
intensity of positive or negative affect or in the sheer frequency
of positive affect. Older adults reported less frequent overall
negative affect than did younger adults, but only up to age 60.
When age differences have been found in these studies, they
have been small in magnitude compared with the substantial
individual variation in positive and negative affect in any
age group. Furthermore, few studies have looked for age or
individual differences in affect among the oldest old. The
unique nature of very old age, starting around age 85, has even
led some theorists to call for the recognition of a fourth age
(e.g., P. B. Baltes, 1997; P. B. Baltes & Smith, 2003; Smith,
2001a). The population of the oldest old consists of people who
have outlived average life expectancy and the majority of their
birth cohort, and who have to deal with high levels of
medical comorbidity (Suzman et al., 1992).
Predictors of Affect in Old Age
Older adults are less likely to work and more likely to suffer
from health problems than are their middle-aged counterparts.
Thus, the set of life context variables that may predict variance
in affect in older adults should be different than those of
middle-aged individuals, and attempts to explicate age effects
in older individuals would need to first specify the impact of
these age-salient variables. For example, although occupational
stress would therefore not be an appropriate contextual variable
for older individuals, physical health status or functional abilities (e.g., strength, mobility, limitations in activities of daily
living [ADLs]) might be appropriate (see, for example, Birren,
Lubben, Rowe, & Deutchman, 1991; Kunzmann, Little & Smith,
2000; Lawton, 1991; Smith, 2001b).
Furthermore, the life contexts of the oldest old appear to
differ substantially from those of their young old counterparts.
Although the young old may be similar to their middle-aged
counterparts in relatively high levels of health, the lives of
the very old are likely to feature multiple comorbid health conditions (Manton & Soldo, 1992). A recent study of Swedish
older adults aged 90 and older, for example, found only 19%
of participants with no severe chronic diseases (von Strauss,
Fratiglioni, Viitanen, Forsell, & Winblad, 2000). Similarly, the
American oldest-old population has much higher levels of
disability than does the young-old population (Stump, Clark,
Johnson, & Wolinsky, 1997). Links between chronic health conditions, functional disability, and depression have been well established (see Williamson, Shaffer, & Parmelee, 2000).
Beyond health and functional status, cognitive abilities also
show age-related decline in the oldest old (e.g., Lindenberger
& P. B. Baltes, 1997; Sliwinski & Buschke, 1999). Although
surprisingly little research has looked at the relationship between
individual differences in cognitive functioning and affective
profiles in older adults (see, however, Arbuckle, Gold, & Andres,
1986; Wetherell, Reynolds, Gatz, & Pedersen, 2002), everyday
functioning has been linked with satisfaction in old age (Diehl,
1998; Lawton, 1991; Willis, 1996), and some evidence has found
a link between cognitive decline and depression in the early
stages of dementia (Schmand, Jonker, Geerlings, & Lindeboom,
1997; Zank & Leipold, 2001).
Social networks are not unimportant to the lives of the very
old. In fact, they may be even more important to those in worse
health, because of the impact social support has on physical
HAPPINESS IN VERY OLD AGE
health and health behaviors (Seeman, 2001). A recent metaanalysis of 286 studies on predictors of subjective well-being
in older individuals found quality of social relationships to
be a strong predictor of subjective well-being in older adults
(Pinquart & Sorensen, 2000). Interestingly, whereas quality was
in general a stronger predictor of well-being than was quantity
of social relationships, differences emerged for friendships and
adult family relationships. Frequency of contact was more important for friendships than for adult family relationships, but
quality of contact was more important for adult family relationships (Pinquart & Sorensen, 2000). Although some recent work
has suggested that personality may be more important than social
support in the prediction of well-being in the oldest old (Landau
& Litwin, 2001), it seems clear that social support—both quality
and quantity—does play some role in late-life well-being.
Chipperfield and Havens (2001) reported that stability of marital status in old age was differentially related to changes in life
satisfaction: Men were more negatively affected by widowhood,
whereas marital stability actually related to declines in satisfaction for women. These findings also suggest the need to evaluate affect predictors by gender, particularly given previous
findings showing differential patterns by gender (e.g., Mroczek
& Kolarz, 1998).
Hypotheses and Analytic Strategy
On the basis of Charles and colleagues’ (2001) and
Carstensen and colleagues’ (2000) findings in particular (both
of which included at least some individuals over the age of 80),
we hypothesized negative age differences for positive affect
and no age differences for negative affect in our sample
(average age 85 years). To test these hypotheses, we conducted
stepwise hierarchical regressions in which demographic variables such as education and marital status are entered in the first
step, personality is entered second, followed by the age-salient
context variables (health comorbidity, functional abilities, cognitive functioning, quality and quantity of social support), and
finally the age term in the fourth step.
However, we also expected different patterns of predictors
between the young old and oldest old when analyses were re-run
separately in those age groups. In particular, because of the
greater role of health-related comorbidity and disability, we
hypothesized that age-salient contextual variables such as these
would eliminate the unique age effect in the oldest-old sample,
but that the effect would remain in the young-old sample. In
other words, we expected the results in the young old to basically
replicate the findings of Mroczek and Kolarz (1998) in terms of
unique age effects on affect, whereas the oldest old would show
strong effects of health and cognition, but no unique age effects.
We tested this by conducting the full-model regression from the
previous analyses (Step 3 in the stepwise regression described
above, including all predictors except age) separately for positive and negative affect in the young old and the oldest old. We
also ran the analyses separately for men and women, although
we did not make differential hypotheses by gender.
METHODS
Participants
The present article uses cross-sectional data from the focal
BASE sample (N ¼ 516). Because details regarding the study
P145
sample, design, selectivity, and multidisciplinary assessment
battery are published elsewhere (P. B. Baltes & Mayer, 1999,
also Baltes & Smith, 1997; Smith et al., 2002), we provide
only a summary. Addresses for potential participants in an age
by sex stratified design were drawn from the city registry.
Participants included in the present analyses (N ¼ 516) had
completed a multidisciplinary assessment battery involving a
total of 14 individual sessions over a period of 3 to 5 months.
This cross-sectional sample included six age/cohort groups
(aged 70–74 years, born 1922–1915; aged 75–79 years, born
1917–1910; aged 80–84 years, born 1913–1905; aged 85–89
years, born 1908–1900; aged 90–94 years, born 1902–1896,
and aged 95–105 years, born 1897–1883). There were equal
numbers of men and women in each group (n ¼ 43) and
the mean age of both sexes did not differ (Men: M ¼ 84:7
years; Women: M ¼ 85:1 years). Data were collected from
1990 to 1993. Apart from some clinical medical assessments,
all testing was carried out in the participant’s place of
residence (e.g., private home, nursing home, hospital) by
trained research assistants and physicians. Seventy-eight percent of the participants resided in their own homes; 14%
resided in institutions (e.g., homes for seniors, nursing homes,
hospitals).
Measures
Sociodemographic characteristics. —In addition to age
(years) and gender (men 5 1; women 5 2), other sociodemographic variables in the analyses were current marital status
(not married 5 0; married 5 1) and years of education. At the
time of the cross-sectional study (1990–1993), 30% of the
BASE sample were married, 55% widowed, 7% divorced, and
8% had never married. Representative for these cohorts in
Germany, 65% of the sample had primary level education, 28%
secondary level, and 8% tertiary level (M ¼ 10:8 years).
Affect. —Raw total scores for Positive Affect and Negative
Affect were calculated using a translated version of the widelyused Positive and Negative Affect Schedule (PANAS, Watson,
Clark, & Tellegen, 1988). In BASE, participants rated on a 5point scale how often they had experienced 20 emotions over
the previous year: 10 of the emotion words were positive (e.g.,
enthusiastic, excited, proud), and 10 were negative (e.g.,
distressed, afraid, upset). Cronbach’s alphas were high for
the 10 positive items (a ¼ :78) and for the 10 negative items
(a ¼ :81).
Personality. —Personality dispositions were represented by
two factors: Neuroticism (derived from responses to six items
assessing the facets anxiety, depression, vulnerability, and
hostility), and Extraversion (derived from responses to six
items assessing the facets gregariousness, positive emotionality,
assertiveness, and activity). Items to assess these dimensions
were selected from the NEO Inventory (Costa & McCrae,
1985): Cronbach’s alpha for the six neuroticism items was .75,
and Cronbach’s alpha for the six extraversion items was .66.
Participants were asked to indicate how well items described
them using a 5-point scale. Each item was read aloud by an
interviewer, and the individual’s response was recorded.
P146
ISAACOWITZ AND SMITH
Health comorbidity. —On the basis of a full clinical medical
examination, BASE physicians determined medical diagnoses for each participant according to the ICD-9 (SteinhagenThiessen & Borchelt, 1999). Here we use as an index of
comorbidity a count of number of moderate to severe physical
illnesses (M ¼ 3:69).
Functional ability/disability. —We used two different measures of functional ability and disability. The first was a simple
count of those ADLs (Mahoney & Barthel, 1965) and instrumental ADLs (items from Lawton & Brody, 1969) on
which the individual did not show a limitation or need for
assistance; therefore, a higher score on this measure indicates
better functional ability and less disability. Scores on this scale
could range from 0 to 12.
We also used a measure of physical mobility, because ability
to move around may be a major constraint on functional health.
This mobility measure (described in more detail in SteinhagenTheissen & Borchelt, 1999) was a composite of four clinical
tests (balance, walking on a spot, bending over, and turning
around 360 degrees in as few steps as possible) and a subjective
report of the distance that could be walked without difficulty.
Social relationships and support. —We included two measures of social relationships and support: One was more objective, and one was more subjective. The more objective measure
was the size of the self-reported social network, as indicated
by each participant in Antonucci’s (1986) Hierachical Mapping
Task of the personal network of close partners. This measure
of network size is a sum of the number of individuals mentioned as being very close, close, and less close.
We assessed satisfaction with social networks by using a 5point self-report scale for (a) satisfaction with family life and
(b) satisfaction with friendships. Mean score in the entire
sample was 3.95 for family and 3.83 for friends. In the current
study, we used a composite of the 2 scales.
Intellectual functioning. —As reported by Lindenberger and
P. B. Baltes (1997), the various measures of intellectual functioning collected in the BASE battery could be reasonably summarized using a general intelligence factor (g). Therefore, we
used this measure in the current study, expressed as T-scores
(M ¼ 50, T ¼ 10).
RESULTS
Zero-order correlations between all study variables are shown
in Table 1 (see Appendix, Note). In contrast to the correlations
reported by Mroczek and Kolarz (1998) for a younger sample
but in line with Charles and colleagues’ results (2001), we
found a significant negative correlation between positive affect
and age in our sample, and no relationship between age and
negative affect (see also Smith, 2001b; Smith, Fleeson,
Geiselmann, Settersten, & Kunzmann, 1999; Smith & P. B.
Baltes, 1993).
ity, and contextual variables by creating hierarchical regression
models for each affect factor. As shown in Table 2, each step
except for the final step in which only age was entered added
significantly to the prediction of positive affect. In the final
model, however, only extraversion and general intelligence significantly predicted positive affect, with more extraversion and
higher intelligence predicting higher levels of positive affect.
However, approximately 70% of the variance in positive affect
remained unexplained.
Regression analyses for negative affect, shown in Table 3,
were quite similar to those for positive affect. Again, each step
significantly added to prediction of affect except for the final
step of adding age. Also similarly, the only unique predictors to
emerge in the full model were a personality variable, in this
case neuroticism, and general intelligence. Interestingly, after
controls for all other variables in the model, higher levels of
intelligence also predicted more negative affect, as did higher
levels of neuroticism. In this model, only 53% of the variance
remained unexplained.
Analyses in the Young Old and Oldest Old
Before conducting regressions separately for the young old
and oldest old BASE participants, we first ran descriptive
analyses comparing mean levels of affect and the predictor
variables in the two age groups. Means and age comparisons are
shown in Table 4. Although the oldest old reported lower levels
of positive affect than their young-old counterparts, there were no
differences between the two groups on negative affect. The
oldest-old participants were, however, significantly less likely
to be married, had fewer years of education, and were less
extraverted than the young-old participants. Not surprisingly, the
oldest old also had more functional disability, more medical
comorbidity, worse grip strength and mobility, as well as lower
levels of general intelligence, fewer total social network
members, and less satisfaction with those social partners.
It was in this context of significant mean-level differences
between the young old and oldest old that we then conducted
regression analyses separately for the two groups. These regression results are shown in Table 5 for positive affect and
Table 6 for negative affect. For positive affect, extraversion,
social network size, and general intelligence emerged as
predictors for the young old. However, only extraversion predicted positive affect in the oldest old. The models accounted
for similar amounts of variance (R2 ¼ :269 for young old
and .268 for oldest old). We then tested whether the betas for
social network size and general intelligence, which were significant predictors of positive affect in the young old but not
the old old, actually differed significantly using Cohen and
Cohen’s (1983, Eq. 3.6.11) formula for such comparisons.
These differences did not reach statistical significance.
Regression models predicted substantially more variance
for negative affect ðR2 ¼ :47 for young old and .47 for oldest
old). In both age groups, neuroticism and general intelligence
emerged as the only predictors of negative affect.
Analyses by Gender
Predictors of Affect
We tested for unique linear age effects on positive and
negative affect after we controlled for demographic, personal-
Mean levels of all relevant variables by gender are reported
in Smith and M. M. Baltes (1998). As shown in Table 1, the
correlation between sex and negative affect was significant,
HAPPINESS IN VERY OLD AGE
P147
Table 1. Intercorrelations Between Variables Included in the Study
Variable
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Age
Positive affect
Negative affect
Gender
Marital status
Education (years)
Extraversion
Neuroticism
No. diagnosed illnesses
ADL and IADL
Grip strength
General intelligence (g)
Mobility
Social network size
Satisfaction with network
1
2
3
4
5
6
7
8
9
10
11
12
13
14
.23**
.05
.02
.22**
.14**
.19**
.08
.24**
.56**
.49**
.56**
.53**
.33**
.09*
.03
.04
.13**
.17**
.46**
.13**
.12**
.25**
.20**
.34**
.22**
.29**
.19**
.17**
.05
.06
.11*
.64**
.14**
.11*
.10*
.04
.06
.01
.13**
.49**
.24**
.01
.23**
.11*
.19**
.62**
.09*
.22**
.06
.04
.22**
.06
.17**
.14**
.19**
.41**
.24**
.22**
.22**
.06
.07
.18**
.14**
.18**
.22**
.39**
.20**
.19**
.02
.12**
.09*
.15**
.10*
.19**
.17**
.23**
.26**
.19**
.24**
.23**
.22**
.22**
.15**
.16**
.34**
.23**
.24**
.37**
.12**
.01
.46**
.52**
.73**
.25**
.09*
.43**
.47**
.25**
.05
.47**
.46**
.13**
.31**
.08
.17**
Note: ADL 5 activity of daily living; IADL 5 instrumental activity of daily living; g 5 general intelligence factor.
*p , :05; **p , :01.
with women reporting higher levels of negative affect. When
we conducted regressions on positive affect separately for
men and women, only extraversion emerged as a significant
predictor for men (ß ¼ :33, p , :01, R2 ¼ :252). For women,
extraversion (ß ¼ :44, p , :01) and intelligence (ß ¼ :21,
p , :01), as well as functional ability/disability (ß ¼ :17,
p , :05) predicted positive affect (model R2 ¼ :365). Tests of
differences between betas did not reach two-tailed significance
levels.
Men who were not married (ß ¼ :11, p , :05), less extraverted (ß ¼ :11, p , :05), more neurotic (ß ¼ :66, p , :01)
and more intelligent (ß ¼ :23, p , :01) reported higher levels
of negative affect (R2 ¼ :494). For women, only neuroticism
(ß ¼ :63, p , :01) and general intelligence (ß ¼ :18, p , :05)
predicted more negative affect (R2 ¼ :430). Again, tests for
differences between regression betas were not significant.
After controls for age-related constructs, no significant age
variance was left unexplained for either men or women.
DISCUSSION
In this study, we attempted to contribute to the understanding
of affect in the very old by considering the relationship of age
to positive and negative affect in a sample of individuals aged
70 to 100 years once other age-related possible predictors had
been controlled. Although we specifically investigated unique
age effects, this is really a proxy for any variables that are left
to be explained after controlling for many other possible agerelated predictors. In this study, all of our predictors except
for neuroticism covary with age. We attempted to include
predictors that accurately reflect the life contexts of older
individuals, such as physical comorbidity as well as mobility,
Table 2. Hierarchical Regression on Positive Affect
Step 1
Predictor
B
Education (years)
Marital status
Gender
Extraversion
Neuroticism
Social network size
Satisfaction with network
No. diagnosed illnesses
Mobility
ADL and IADL
Grip strength
General intelligence (g)
Age
.66**
2.69*
1.22
R2
R2
F for R2
df
p,
.041
.041
7.36
3,512
.001
Step 2
SE
.19
1.0
1.0
B
.53*
1.9*
.96
.43**
.04
.237
.196
65.47
5,510
.001
Step 3
SE
B
.17
.97
.90
.03
.04
.22
.59
1.12
.37**
.03
.12*
.59
.03
.26
.28
.08
.15**
.29
.058
5.94
12,503
.001
Step 4
SE
.18
.97
1.08
.04
.04
.06
.54
.18
.43
.23
.06
.05
B
.21
.77
1.61
.38**
.01
.12*
.57
.02
.20
.34
.11
.16**
.07
SE
ß
.18
.98
1.16
.04
.04
.06
.54
.18
.44
.23
.07
.05
.06
.04
.03
.08
.38**
.01
.09*
.04
.01
.02
.09
.09
.16**
.07
.297
.002
1.47
13,502
ns
Note: The final column represents standardized betas for the full model. ADL 5 activity of daily living; IADL 5 instrumental activity of daily living; g 5 general intelligence factor.
*p , :05; **p , :01.
ISAACOWITZ AND SMITH
P148
Table 3. Hierarchical Regression on Negative Affect
Step 1
Predictor
Education
Marital status
Gender
Extraversion
Neuroticism
Social network size
Satisfaction with network
No. diagnosed illnesses
Mobility
ADL and IADL
Grip strength
General intelligence (g)
Age
R2
R2
F for R2
df
p,
B
.11
1.1
3.84**
.032
.032
5.7
3,512
.001
Step 2
SE
.19
1.1
1.0
B
.23
2.09*
1.75*
.04
.64**
.427
.395
176.1
5,510
.001
Step 3
Step 4
SE
B
SE
B
.15
.84
.78
.03
.03
.06
1.35
1.23
.06
.65**
.03
.59
.27
.75*
.37*
.02
.20**
.15
.85
.94
.03
.03
.05
.47
.15
.38
.20
.05
.04
.05
1.29
1.04
.06
.65**
.02
.58
.27
.72
.39
.03
.19**
.03
.464
.037
4.97
12,503
.001
SE
ß
.15
.86
1.01
.03
.03
.05
.47
.15
.38
.20
.06
.04
.05
.01
.06
.05
.06
.65**
.02
.04
.06
.09
.10
.03
.20**
.03
.465
.000
0.28
13,503
ns
Notes: The final column represents standardized betas for the full model. ADL 5 activity of daily living; IADL 5 instrumental activity of daily living; g 5
general intelligence factor.
*p , :05; **p , :01.
in addition to more standard and general predictors of affect,
such as personality and gender. We found that personality and
cognitive functioning predicted positive and negative affect,
with age contributing no unique variance after the entry of the
other predictors.
We then conducted analyses to determine whether predictors
of positive and negative affect in old age differ by meaningful
subgroup. We were particularly interested in differences between the young old and the oldest old, given recent findings
suggesting that the two groups are distinct (P. B. Baltes &
Smith, 2003; Smith, 2001a; Suzman et al., 1992). Although the
predictors from the full-sample analyses replicate in both age
groups for negative affect, an interesting difference emerges for
positive affect. Although extraversion, general intelligence, and
social network size (not significant in the full-sample analyses)
all predict positive affect in the young old, only extraversion
predicts positive affect in the oldest old. However, the regression betas for social network size and general intelligence
are not significantly different, suggesting that the differential
prediction pattern should be viewed with caution.
Finally, we conducted the analyses separately by gender.
Although additional predictors emerge in these analyses, such
as functional abilities for women and positive affect, and
marital status for men and negative affect, these analyses
primarily serve to reiterate the strong relationships between
personality and cognitive functioning, and the affect outcome
measures.
Personality: The Strongest Predictor of Affect
What is at first most striking about these findings is the strong,
consistent relationship between personality and affect: Extraversion predicts positive affect, and neuroticism predicts negative
Table 4. Comparison of the Young Old (N ¼ 258) and Oldest Old (N ¼ 258) in the Berlin Aging Study: Means and Standard Deviations
Construct
Young Old
Oldest Old
Age (years)
Positive affect
Negative affect
% women
% married
Education (years)
Extraversion
Neuroticism
No. diagnosed illnesses
ADL and IADL nonlimitations (max: 12)
Maximum grip strength (kg)
Mobility (max: 6)
General intelligence (g)
No. in social network
Network satisfaction
77.5 (4.3)
3.3 (.53)
2.3 (.54)
50
39
11.2 (2.4)
3.5 (.55)
2.3 (.75)
3.2 (2.1)
10.9 (1.7)
4.7 (1.01)
4.99 (7.6)
55 (8.8)
11.9 (7.9)
3.9 (.6)
92.4 (4.5)
3.0 (.62)
2.2 (.66)
50
20.5
10.4 (2.2)
3.2 (.6)
2.4 (.8)
4.2 (2.3)
8.4 (2.7)
3.7 (1.3)
45 (6.2)
44 (8.5)
7.6 (5.4)
3.7 (.8)
Significant Age and Cohort Differences
Fð1; 515Þ ¼ 1478:1, p , :001
Fð1; 515Þ ¼ 26:5, p , :001
ns
ns
v2 ð1; N ¼ 516Þ ¼ 21:2, p , :001
Fð1; 515Þ ¼ 115:8, p , :001
Fð1; 515Þ ¼ 25:2, p , :001
ns
Fð1; 515Þ ¼ 30:7, p , :001
Fð1; 515Þ ¼ 168:1, p , :001
Fð1; 515Þ ¼ 129:5, p , :001
Fð1; 515Þ ¼ 155:2, p , :001
Fð1; 515Þ ¼ 172:7, p , :001
Fð1; 515Þ ¼ 50:3, p , :001
Fð1; 515Þ ¼ 5:1, p , :05
Notes: Scores for positive affect, negative affect, extraversion and neuroticism are for a 5-point scale (5 5 high). When the original scale is noninformative
(cognition), means are reported as T-scores (M ¼ 50, SD ¼ 10). Mobility is coded on a 6-point scale (max 5 6: can walk 5 km without difficulty). ADL 5 activity
of daily living; IADL 5 instrumental activity of daily living; g 5 general intelligence factor.
HAPPINESS IN VERY OLD AGE
P149
Table 5. Regression Coefficients by Age Group for Positive Affect
Young Old (N ¼ 258)
Variable
Education
Marital status
Gender
Extraversion
Neuroticism
Social network size
Satisfaction with network
No. diagnosed illnesses
Mobility
ADL and IADL
Grip strength
General intelligence (g)
Oldest Old (N ¼ 258)
B
SE
ß
B
SE
ß
.12
.15
.72
.37
.02
.15
.52
.01
.00
.13
.11
.17
.23
1.26
1.73
.06
.06
.07
.87
.27
.62
.09
.09
.07
.03
.01
.04
.38**
.02
.13*
.03
.00
.00
.09
.11
.17*
.35
1.56
1.99
.39
.02
.12
1.14
.10
.21
.05
.12
.15
.29
1.60
1.66
.06
.06
.12
.75
.26
.65
.08
.13
.09
.07
.06
.10
.39**
.01
.06
.09
.02
.03
.05
.07
.13
R2
.269
.268
Note: ADL 5 activity of daily living; IADL 5 instrumental activity of daily living; g 5 general intelligence factor.
*p , 05; **p , :01.
affect, in both the young old and the oldest old samples. Not only
is this consistent with one of the earliest findings on predictors
of affect in adulthood (Costa & McCrae, 1980), it is also highly
consistent with more recent findings suggesting that personality
is more important than social relationships in predicting subjective well-being in the oldest old (Landau & Litwin, 2001).
Indeed, social network variables predict well-being in only one
case in the current analyses: total size of social network related
to positive affect in the young old.
More important, personality factors are also much stronger
predictors of positive and negative affect than is health. Although
previous studies have found health and subjective well-being
to be strongly linked in older adults (Kunzmann et al., 2000;
Smith, 2001b), we find that the inclusion of personality variables
eliminates any effect of health in the prediction of well-being.
The only exception is that functional abilities predict positive
affect in females. These findings strongly support Costa &
McCrae’s (1980) argument that personality is a strong predictor
of well-being at any age. Additionally, they are consistent with
Mroczek & Kolarz’s (1998) results in a sample of adults aged
25–74 years, in which personality was the single strongest predictor of affect as well. However, in the MIDMAC sample, other
sociodemographic variables also predicted affect, although not as
strongly as personality.
The personality findings provide a solid rationale for investigating positive and negative affect separately, rather than
investigating predictors of a single well-being measure such as
life satisfaction or depressive symptoms. Although controversies
still exist on the structure of affect, many in the field consider
positive and negative affect to be independent constructs
(Watson & Clark, 1997). Indeed, research with older individuals
has found that positive and negative affect are independent
(Kercher, 1992), and have both different age-related trajectories
(Carstensen et al., 2000; Charles et al., 2001; Mroczek & Kolarz,
1998) as well as differential relationships with other variables
(Isaacowitz & Seligman, 2002). Consistent with Costa and
McCrae’s (1980) early predictions that extraversion would
predict positive affect and neuroticism would predict negative
affect, we find exactly these relationships in the current study. In
only one case is there overlap: both neuroticism and extraversion
predict negative affect in men. Had we not evaluated positive and
negative affect separately, we may have missed these strong,
distinct relationships between personality and affect in our sample of very old adults.
One concern with these findings showing a strong link between personality measured with items from NEO, and affect
measured by the PANAS, is that there may be some conceptual
overlap between the items on the two measures. This is of particular concern because we asked participants to report positive and negative affect over the previous year, which is akin
to a trait-like rating (Watson et al., 1988). One recent study attempted to disentangle the two constructs by evaluating their
relationship to success on the job market among individuals
graduating from college (Burger & Caldwell, 2000). They
found that trait positive affect predicted behavior after
controlling for extraversion, but not vice versa. Interestingly,
another study found that premorbid personality predicted the
emotions of older individuals suffering from dementia, though
that study used different measures of personality and affect
(Magai, Cohen, Culver, Gomberg, & Malatesta, 1997). Clearly,
how personality and affect relate to each other more specifically
in older adults is a needed area of future research.
Intelligence and Affect
Perhaps a more surprising finding, and one that is less predictable from past research, is that, more so than any demographic or contextual variables, a general intelligence factor
predicts both positive and negative affect. Higher levels of cognitive functioning predict both more positive affect and more
negative affect, after demographic, health, and personality variables are controlled for. At the zero-order level, intelligence was
significantly correlated with positive affect, but not with negative affect.
Interestingly, intellectual functioning predicts mortality more
strongly in the young old than the oldest old in BASE (Maier &
Smith, 1999). It is difficult, however, to try to understand these
strong, distinct effects of cognitive fitness on well-being in both
the young-old and oldest-old participants because of the paucity
of other data on the links between cognition and affect. The
notable exception is data linking early stages of dementia with
declines in well-being (Schmand et al., 1997; Zank & Leipold,
2001).
ISAACOWITZ AND SMITH
P150
Table 6. Regression Coefficients by Age Group for Negative Affect
Young Old (N ¼ 258)
Variable
Education
Marital status
Gender
Extraversion
Neuroticism
Social network size
Satisfaction with network
No. diagnosed illnesses
Mobility
ADL and IADL
Grip strength
General intelligence (g)
Oldest Old (N ¼ 258)
B
SE
ß
B
SE
ß
.23
1.39
1.76
.05
.60
.01
.99
.08
.33
.07
.00
.17
.20
1.06
1.45
.05
.05
.06
.73
.22
.52
.08
.08
.06
.06
.08
.10
.05
.65**
.01
.07
.02
.04
.05
.04
.17**
.17
1.59
1.00
.09
.69
.09
.42
.42
.70
.08
.04
.20
.26
1.43
1.49
.05
.05
.11
.67
.24
.59
.07
.12
.08
.03
.06
.05
.08
.65**
.04
.03
.09
.08
.08
.02
.16**
R2
.471
.467
Note: ADL 5 activity of daily living; IADL 5 instrumental activity of daily living.
*p , :05; **p , :01.
Rowe & Kahn’s (1997; 1998) model of successful aging
provides one way of understanding the important role of
general intellectual functioning in the well-being of older
individuals. According to their model, maintaining high levels
of cognitive functioning is a critical part of the structure of
successful aging. But how is it possible that it relates to both
higher levels of positive and negative affect, as we find in
the current study? One possibility is that better cognitive
functioning allows individuals to stay more engaged with life;
such engagement may then bring with it both greater enjoyment from life, but also the risk for greater involvement
in disappointments (e.g., sadness about loss of friends, selfawareness of age-related changes taking place). It is important
to keep in mind, however, that mean levels of negative affect
are particularly low in this study.
Interestingly, one of the only differences between the young
old and oldest old is that intelligence predicts negative affect
in both age groups, but does not predict positive affect in
the oldest old. However, the regression betas do not differ
significantly between the two age groups, so it seems best to not
overinterpret this finding. Overall, predictors of affect in the
young old and oldest old are more similar than different.
Conclusions and Future Directions
It is perhaps surprising given recent findings on age and
well-being (e.g., Mroczek & Kolarz, 1998) that we do not find
any unique effects of age on positive or negative affect after we
controlled for a host of other predictors. Another way of looking
at this finding would be that we have successfully captured the
actual age-related predictors of affect in very old age, and thus
have left no variance to be explained by age. This argument
would be consistent with the assertion that age is primarily an
empty variable, carrying the effects of other more substantive
variables (P. B. Baltes, Reese, & Nesselroade, 1977; Wohlwill,
1973). Although health may be the most often considered
variable that age may carry, health status does not predict wellbeing in the current study. Instead, personality, consisting of
loosely age-related variables, and intellectual functioning,
a strongly age-related construct, emerge as the strongest predictors of affect in the very old.
Two lines of future research are suggested by these findings,
and by the obvious limitations of the study. First, even though
all age-related variance seems to have been accounted for in our
analyses, a large proportion of variance in affect is still left unexplained. Thus, there is clearly room for exploration of nonage-related factors that may predict affect in very old age.
Second, these cross-sectional results lead to obvious questions
for longitudinal work: In particular, what effect does change in
the predictor variables actually have on affect? Is it merely the
case that those individuals with lower mean levels of general
intelligence have more positive and negative affect, or do actual
declines in intellectual functioning over time lead to changes in
affect? The current study provides a glimpse into how age may
and may not relate to affect in very late life.
ACKNOWLEDGMENTS
This article was written while the first author was supported by the
National Science Foundation and the Max Planck Society to visit the Max
Planck Institute in Berlin. Presented are cross-sectional data from the Berlin
Aging Study (BASE), a multidisciplinary project which was financially
supported by two German Federal Departments: The Department of
Research and Technology (13 TA 011: 1988–1991) and the Department of
Family and Senior Citizens (1991–1998). The institutions involved in
BASE are the Free University of Berlin and the Max Planck Institute for
Human Development, Berlin, where the study is housed. The study is
directed by a Steering Committee co-chaired by the directors of the four
cooperating research units (P. B. Baltes and J. Smith, psychology; K. U.
Mayer and M. Wagner, sociology; E. Steinhagen-Thiessen and M.
Borchelt, internal medicine and geriatrics; H. Helmchen and M. Linden,
psychiatry). Field research was coordinated by R. Nuthmann. Valuable
discussions with the members of the Berlin Aging Study and colleagues at
the Center for Lifespan Psychology at the Max Planck Institute for Human
Development are gratefully acknowledged.
Derek M. Isaacowitz is now at the Department of Psychology, Brandeis
University, Waltham, MA.
Address correspondence to Jacqui Smith, Center for Lifespan Psychology, Institute for Human Development, Lentzeallee 94, 14195 Berlin,
Germany. E-mail: [email protected] or [email protected]
REFERENCES
Antonucci, T. C. (1986). Hierarchical mapping technique. Generations, 10,
10–12.
HAPPINESS IN VERY OLD AGE
Arbuckle, T. Y., Gold, D., & Andres, D. (1986). Cognitive functioning of
older people in relation to social and personality variables. Psychology
and Aging, 1, 55–62.
Baltes, P. B. (1997). On the incomplete architecture of human ontogeny:
Selection, optimization, and compensation as foundation of developmental theory. American Psychologist, 52, 366–380.
Baltes, P. B., & Mayer, K. U. (Eds.). (1999). The Berlin Aging Study: Aging
from 70 to 100. New York: Cambridge University Press.
Baltes, P. B., Reese, H. W., & Nesselroade, J. R. (1977). Life-span
developmental psychology: Introduction to research methods. Monterey, CA: Brooks/Cole.
Baltes, P. B., & Smith, J. (1997). A systemic-wholistic view of psychological
functioning in very old age: Introduciton to a collection of articles from
the Berlin Aging Study. Psychology and Aging, 12, 395–409.
Baltes, P. B., & Smith, J. (2003). New frontiers in the future of aging: From
successful aging of the young old to the dilemmas of the fourth age.
Gerontology, 49, 123–135.
Belsley, D. A., Kuh, E., & Welsh, R. E. (1980). Regression diagnostics:
Identifying influential data and sources of collinearity. New York:
Wiley.
Birren, J. E., Lubben, J. E., Rowe, J. C., & Deutchman, D. E. (Eds.).
(1991). The concept and measurement of quality of life in the frail
elderly. San Diego, CA: Academic Press.
Burger, J. M., & Caldwell, D. F. (2000). Personality, social activities, jobsearch behavior and interview success: Distinguishing between PANAS
trait positive affect and NEO extraversion. Motivation & Emotion, 24,
51–62.
Carstensen, L. L., Pasupathi, M., Mayr, U., & Nesselroade, J. R. (2000).
Emotion experience in the daily lives of older and younger adults.
Journal of Personality and Social Psychology, 79, 644–655.
Charles, S. T., Reynolds, C. A., & Gatz, M. (2001). Age-related differences
and change in positive and negative affect over 23 years. Journal of
Personality and Social Psychology, 80, 136–151.
Chipperfield, J. G. & Havens, B. (2001). Gender differences in the relationship between marital status transitions and life satisfaction in later life.
Journal of Gerontology: Psychologial Sciences, 56B, P176–P186.
Cohen, J., & Cohen, P. (1983). Applied multiple regression correlation
analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Costa, P. T. Jr., & McCrae, R. R. (1980). Influence of extraversion and
neuroticism on subjective well-being: Happy and unhappy people.
Journal of Personality and Social Psychology, 38, 668–678.
Costa, P. T., Jr., & McCrae, R. R. (1985). NEO: Five-Factor personality
inventory. Tallahassee, FL: Psychological Assessment Resources.
DeNeve, K. M., & Cooper, H. (1998). The happy personality: A metaanalysis of 137 personality traits and subjective well-being. Psychological Bulletin, 124, 197–229.
Diehl, M. (1998). Everyday competence in later life: Current status and
future directions. The Gerontologist, 38, 422–433.
Isaacowitz, D. M., & Seligman, M. E. P. (2002). Cognitive style predictors
of affect change in older adults. International Journal of Aging and
Human Development, 54, 233–253.
Kercher, K. (1992). Assessing subjective well-being in the old-old: The
PANAS as a measure of orthogonal dimensions of positive and negative
affect. Research on Aging, 14, 131–168.
Kunzmann, U., Little, T. D., & Smith, J. (2000). Is age-related stability of
subjective well-being a paradox? Cross-sectional and longitudinal evidence from the Berlin Aging Study. Psychology and Aging, 15, 511–
526.
Lachman, M. E. (Ed.) (2001). Handbook of midlife development. New
York: Wiley.
Landau, R., & Litwin, J. (2001). Subjective well-being among the old-old:
The role of health, personality, and social support. International
Journal of Aging and Human Development, 52, 265–280.
Lawton, M. P. (1991). A multidimensional view of quality of life in frail
elders. In J. E. Birren, J. E. Lubben, J. C. Rowe, & D. E. Deutchman
(Eds.), The concept and measurement of quality of life in the frail
elderly (pp. 3–27). San Diego, CA: Academic Press.
Lawton, M. P., & Brody, E. M. (1969). Assessment of older people: Selfmaintaining and instrumental activities of daily living. The Gerontologist, 9, 179–186.
Lawton, M. P., Kleban, M. H., Rajagopal, D., & Dean, J. (1992).
Dimensions of affective experience in three age groups. Psychology and
Aging, 7, 171–184.
P151
Lindenberger, U., & Baltes, P. B. (1997). Intellectual functioning in old and
very old age: Cross-sectional results from the Berlin Aging Study.
Psychology and Aging, 12, 410–432.
Lykken, D. (1999). Happiness. New York: Golden Books.
Magai, C., Cohen, C. I., Culver, C., Gomberg, D., & Malatesta, C. (1997).
Relation between premorbid personality and patterns of emotion
expression in mid- to late-stage dementia. International Journal of
Geriatric Psychiatry, 12, 1092–1099.
Maier, H., & Smith, J. (1999). Psychological predictors of mortality in old
age. Journal of Gerontology: Psychological Sciences, 54B, P44–P54.
Malatesta, C. Z., & Kalnok, M. (1984). Emotional experience in younger
and older adults. Journal of Gerontology, 39, 301–308.
Manton, K. G., & Soldo, B. J. (1992). Disability and mortality among the
oldest old: Implications for current and future health and long-term-care
service needs. In R. M. Suzman, D. P. Willis, & K. G. Manton (Eds.),
The oldest old (pp. 199–250). New York: Oxford University Press.
McCrae, R. R., Costa, P. T., Jr., Lima, M. P., Simoes, A., Ostendorf, F.,
Angleitner, A., et al. (1999). Age differences in personality across the
adult life span: Parallels in five cultures. Developmental Psychology, 35,
466–477.
Mroczek. D. K., & Kolarz, C. M. (1998). The effect of age on positive and
negative affect: A developmental perspective on happiness. Journal of
Personality and Social Psychology, 75, 1333–1349.
Pinquart, M. & Sorensen, S. (2000). Influences of socioeconomic status,
social network, and competence on subjective well-being in later life: A
meta-analysis. Psychology and Aging, 15, 187–224.
Rowe, J. W., & Kahn, R. L. (1997). Successful aging. The Gerontologist,
37, 433–440.
Rowe, J. W., & Kahn, R. L. (1998). Successful aging: The MacArthur
Foundation Study. New York: Pantheon Books.
Ryff, C. D., Lee, Y. H., Essex, M. J., & Schmutte, P. S. (1994). My
children and me: Midlife evaluations of grown children and of self.
Psychology and Aging, 9, 195–205.
Schaie, K. W. (1996). Intellectual development in adulthood: The Seattle
Longitudinal Study. New York: Cambridge University Press.
Schmand, B., Jonker, C., Geerlings, M. I., & Lindeboom, J. (1997). Subjective memory complaints in the elderly: Depressive symptoms and
future dementia. British Journal of Psychiatry, 171, 373–376.
Seeman, T. (2001). How do others get under our skin? Social relations and
health. In C. D. Ryff & B. H. Singer (Eds.), Emotion, social relationships,
and health (pp. 189–210). Oxford, England: Oxford University Press.
Sliwinski, M, & Buschke, H. (1999). Cross-sectional and longitudinal relationships among age, cognition, and processing speed. Psychology and
Aging, 14, 18–33.
Smith, J. (2001a). Old age and centenarians. In N. Smelser & P. B. Baltes
(Eds.), International encyclopedia of the social and behavioral sciences
(pp. 10843–10847). Oxford, England: Elsevier.
Smith, J. (2001b). Well-being and health from age 70 to 100: Findings from
the Berlin Aging Study. European Review, 9, 461–477.
Smith, J., & Baltes, M. M. (1998). The role of gender in very old age:
Profiles of functioning and everyday life patterns. Psychology and
Aging, 13, 676–695.
Smith, J., & Baltes, P. B. (1993). Differential psychological ageing: Profiles
of the old and very old. Ageing and Society, 13, 551–587.
Smith, J., Fleeson, W., Geiselmann, B., Settersten, R. A., Jr., & Kunzmann,
U. (1999). Sources of well-being in very old age. In P. B. Baltes &
K. U. Mayer (Eds.), The Berlin Aging Study: Aging from 70 to 100
(pp. 450–471). New York: Cambridge University Press.
Smith, J., Maas, I., Mayer, K. U., Helmchen, H., Steinhagen-Thiessen, E. &
Baltes, P. B. (2002). Two-wave longitudinal findings from the Berlin
Aging Study: Introduction to a collection of articles. Journal of
Gerontology: Psychological Sciences, 57B, P471–P473.
Stacey, C. A., & Gatz, M. (1991). Cross-sectional age differences and
longitudinal change on the Bradburn Affect Balance Scale. Journal of
Gerontology: Psychological Sciences, 46, P76–P78.
Steinhagen-Thiessen, E., & Borchelt, M. (1999). Morbidity, medication,
and functional limitations in very old age. In P. B. Baltes & K. U.
Mayer (Eds.), The Berlin Aging Study: Aging from 70 to 100 (pp. 131–
166). New York: Cambridge University Press.
Stump, T. E., Clark, D. O., Johnson, R. J., & Wolinksy, F. D. (1997). The
structure of health status among Hispanic, African American, and White
older adults. The Journals of Gerontology: Psychological Sciences and
Social Sciences, 52B (Special Issue), 49–60.
P152
ISAACOWITZ AND SMITH
Suzman, R. M., Manton, K. G., & Willis, D. P. (1992). Introducing
the oldest old. In R. M. Suzman, D. P. Willis, & K. G. Manton
(Eds.), The oldest old (pp. 3–14). New York: Oxford University
Press.
von Strauss, E., Fratiglioni, L., Viitanen, M., Forsell, Y., & Winblad, B.
(2000). Morbidity and comorbidity in relation to functional status: A
community-based study of the oldest-old (901 years). Journal of the
American Geriatrics Society, 48, 1462–1469.
Watson, D., & Clark, L. A. (1997). Measurement and mismeasurement
of mood: Recurrent and emergent issues. Journal of Personality
Assessment, 68, 267–296.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and
validation of brief measures of positive and negative affect: The
PANAS scales. Journal of Personality and Social Psychology, 54,
1063–1070.
Welsh, W. M., & Stewart, A. J. (1995). Relationships between women
and their parents: Implications for midlife well-being. Psychology and
Aging, 10, 181–190.
Wetherell, J. L., Reynolds, C. A., Gatz, M., & Pedersen, N. L. (2002).
Anxiety, cognitive performance, and cognitive decline in normal aging.
Journal of Gerontology: Psychological Sciences, 57B, P246–P256.
Williamson, G. M., Shaffer, D. R., & Parmelee, P. A. (Eds.) (2000). Physical
illness and depression in older adults: A handbook of theory, research,
and practice. Dordrecht, The Netherlands: Kluwer Academic Publishers.
Willis, S. L. (1996). Everyday competence in elderly persons: Conceptual
issues and empirical findings. The Gerontologist, 36, 595–601.
Wohlwill, J. F. (1973). The study of behavioral development. New York:
Academic Books.
Yang, J., McCrae, R. R., & Costa, P. T., Jr. (1998). Adult age differences in
personality traits in the United States and the People’s Republic of China.
Journal of Gerontology: Psychological Sciences, 53B, P375–P383.
Zank, S., & Leipold, B. (2001). The relationship between severity of dementia and subjective well-being. Aging & Mental Health, 5, 191–196.
Received July 13, 2000
Accepted December 13, 2002
Decision Editor: Carolyn Aldwin, PhD
Appendix
Note
Because of the high intercorrelations among several of our predictor
variables, we were concerned about potential multicollinearity in our
hierarchical regression models. To investigate whether multicollinearity might be substantially influencing our regression
findings, we evaluated variance inflation factors (VIFs) for the
predictors and found most to be below 2 and none above 3. These
were well below the traditional cut-off of 10 for substantial
multicollinearity (Belsley, Kuh, & Welsh, 1980). Nonetheless, we
did re-run the hierarchical regressions after standardizing all
predictor variables; this did not alter the results of the regression.