Professional Documents
Culture Documents
Gow
Journal et al.: Social
of Individual SupHogrefe
port
Differences
© 2007 andVol.
2007; &Successful
Huber Aging
28(3):103–115
Publishers
Abstract. Social networks or support may contribute to successful aging. The Lothian Birth Cohort 1921 had their mental ability
assessed at age 11 and 79. Almost 500 participants also rated their life satisfaction, social networks, and support at age 80. After
controlling for age-11 IQ, sex, years of education, and social class, loneliness was the only social network/support characteristic adding
significantly to the prediction of age-79 IQ, explaining about 2% of the variance; in old age, increased loneliness was associated with
lower cognitive ability. Social network/support factors accounted for 23% of the variance in satisfaction with life ratings, with the
greatest contributions from reduced loneliness (~12%) and having someone to talk to (~6%). Social network/support characteristics
explained a greater proportion of the variance in life satisfaction ratings compared with later life cognition, although an individual’s
level of loneliness emerged as the largest single social support predictor of both outcomes. Possible causal pathways for these associ-
ations, which need to be studied in future research, are discussed.
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
DOI 10.1027/1614-0001.28.3.103
104 A.J. Gow et al.: Social Support and Successful Aging
terioration caused by the particular lifestyle factor under networks on cognitive ability or decline, which led the au-
investigation, or it might represent a low level of prior abil- thors to conclude there was evidence of a protective effect
ity in an individual. Only by controlling for a measure of from social lifestyle components (Fratiglioni et al., 2004).
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
early ability is it possible to determine whether factors from Although the mechanisms behind the reported associa-
a variety of domains can impact upon cognitive function. tions are not fully understood, a number of potential path-
The current study is unusual and valuable because of the ways have been suggested. If cognitive aging is delayed or
presence of childhood ability data in a cohort of elderly slowed by the level of social support received through con-
individuals. tact and interaction with others, then the physiological
pathways underlying this are likely to be shared with those
thought to link increased social support with better health
Social Engagement and Cognition outcomes. Social support may act as a buffer against stress-
ful life events and, thus, reduce exposure to the resultant
Many argue that pursuing a mentally engaged and active cumulative pathological effects of stress. Alternatively, so-
lifestyle may be a part of preserving one’s cognitive abili- cial support may be a constant, more generally available
ties into old age (Fratiglioni, Paillard-Borg, & Winblad, resource, accessed by an individual not just during periods
2004). In addition, social networks and support (the num- of increased stress, but across time and situations. Receiv-
ber of social ties, marital status, or the level of support from ing more support over time would lead to improved health
relatives and friends, for example) are predictive of later- as a direct result (Cohen & Wills, 1985), which might di-
life mental health outcomes including cognitive impair- rectly or indirectly impact upon cognitive outcomes.
ment and positive well-being (Fratiglioni, Wang, Ericsson, Pathways linking social networks and support to physi-
Maytan, & Winblad, 2000; Okabayashi, Liang, Krause, ological changes are plausible, with a major review of stud-
Akiyama, & Sugisawa, 2004). Fratiglioni and colleagues ies assessing physiological function and social support con-
(2000) reported that individuals who lived alone or had no cluding:
close social relationships were at an increased risk of de-
veloping dementia; those individuals defined as having a Social support has beneficial effects on physiological pro-
cesses across different age groups. The net effect of such
poor or limited social network had a 60% increased risk.
processes may be to biologically age the individual at a
Decreased social support may also increase cognitive de- slower rate (Uchino, Cacioppo, & Kiecolt-Glaser, 1996,
cline in the absence of dementia, as individuals aged 70–79 p. 525).
receiving a lower frequency of emotional support at base-
line were found to have poorer cognitive ability (a sum- This has obvious consequences for the development of dis-
mary score from six cognitive tests) after a 7.5-year follow- ease and, thus, premature mortality. Social relationships,
up (Seeman, Lusignolo, Albert, & Berkman, 2001). Al- networks, integration, and support could, therefore, also in-
though the percentage of variance in cognitive ability fluence cognitive aging, itself an integral part of the human
explained by the level of emotional support was only aging process, via physiological pathways shared with
around 1%, this was after other known confounders had those affecting health.
been controlled (including age, education, and health sta- However, this is not the only possible explanation. The
tus). It is possible that the lifetime impact of social support protection offered by being married or living with others,
could be much greater than that observed over a 7.5-year for example, might be the result of the mental stimulation
period. Social contact may have potentially far reaching required in dealing with other people (van Gelder et al.,
consequences for the health and mental functioning of the 2006). Most authors promote a similar idea for the ob-
elderly (Melchior, Berkman, Niedhammer, Chea, & Gold- served associations, often highlighting the notion of cogni-
berg, 2003). tive reserve (Barnes, Mendes de Leon, Wilson, Bienias, &
A recent review suggested that positive aspects of social Evans, 2004; Bassuk et al., 1999; Fratiglioni et al., 2004;
networks appeared to be generally positively related to bet- van Gelder et al., 2006; Zunzunegui et al., 2003). Measures
ter cognitive outcomes in later life (Fratiglioni et al., 2004). of social networks may give an indication of how active an
However, across studies, differences were apparent in the individual is in their social environment, such that good
assessment of social networks (and support, when as- social relations with friends and relatives, throughout the
sessed), from simple counts of close contacts to more de- life course and especially in the late life, might produce
tailed assessments of social integration. The review identi- continued mental stimulation and better cognitive strate-
fied seven longitudinal observational studies investigating gies or increase neural growth and synaptic density – the
the association between social networks and cognition, yet “use-it-or-lose-it” hypothesis – delaying cognitive impair-
only three of these (Bassuk, Glass, & Berkman, 1999; See- ment or protecting against pathological processes (Zunzu-
man et al., 2001; Zunzunegui et al., 2003) dealt with the negui et al., 2003, p. 98).
actual measurement of social ties or networks – the descrip- It is, of course, plausible that the reverse is true: Those
tions of the remaining four highlight their reliance on mea- individuals experiencing greater cognitive change in old
sures of social activity participation. Bearing this caveat in age may be less able to sustain their social networks (Zun-
mind, five of the seven studies reported effects of social zunegui et al., 2003). Studies have attempted to limit the
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 105
likelihood of this possibility by controlling for baseline satisfaction is perceived. Diener and Seligman (2004) re-
ability or by excluding participants with low scores at the viewed the importance of social relationships to well-being
initial assessment, arguing that the effects remain after such and suggested that the
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
106 A.J. Gow et al.: Social Support and Successful Aging
search in Education, 1933). Survivors of the SMS1932 participants completing each measure used in the analysis
(that is, individuals born in 1921 and at school in Scotland is reported throughout.
in 1932) from Edinburgh and the surrounding areas were
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
Procedure
Social Networks and Support
The LBC1921 were mailed the Satisfaction With Life Scale
(SWLS: Diener et al., 1985), the Significant Others Scale The following measures were used to assess social net-
(SOS: Power, Champion, & Aris, 1988) and additional so- works and support in the LBC1921.
cial network questions within a larger questionnaire book- 1. Significant others – the SOS (Power et al., 1988) was
let, the distribution of which is described elsewhere (Gow, administered. It is a self-report measure that assesses the
Whiteman, Pattie, & Deary, 2005). Booklets were sent to availability of, and perceived satisfaction with, the indi-
568 participants. This is a larger number than previously vidual’s support network. In the current study, five po-
reported from the LBC1921 Study (e.g., Deary et al., 2004), tential supports were specified: spouse (husband/wife)
as it includes those who were listed in the cohort and mailed or partner, closest brother or sister, other brother or sis-
the booklet, but who did not subsequently attend the clinic ter, closest son or daughter, and best friend. Participants
for age-79 cognitive testing. Return of the questionnaire were required to state how often they received emotional
booklet was by prepaid envelope: 29 responses were refus- (two items) and practical (two items) support from each
als, 3 participants had died, and one address was unknown. of these potential supports. For each item they were also
When the booklets were returned each was checked for asked what their ideal level of support would be. Re-
omissions. If an item had been missed or had multiple an- sponses were on a 7-point scale (from never to always).
swers, the participant was contacted to provide corrections. If the participant did not have one of the potential sup-
For those who had not returned the questionnaire or cor- ports (e.g., they were single or their spouse had died)
rections within 4–5 weeks after mailing, a reminder with they were asked to cross out that category. When this
another copy of the questionnaire/correction letter was sent occurred, the particular significant other was scored as
out. Booklets were received from 497 participants (87.5%). being unavailable to provide support, and received the
Twenty-eight (5.6%) of these had some missing data (after lowest possible score for each item. For each participant,
corrections were requested as appropriate). The number of the SOS responses generated the following set of scores:
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 107
– whether they had each of the named significant others guide variable selection in the regression analyses predicting
(yes/no); these outcomes. Adjusted R2 and adjusted R2 change are re-
– how many significant others they had in total (0–5); ported to indicate the percentage of variance accounted for
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
– the perceived level of support received from each indi- by each dependent variable in the models.
vidual significant other listed;
– their ideal level of support from each individual signif-
icant other listed;
– a discrepancy measure for each individual significant Results
other listed (ideal minus actual support score);
– and average support, average ideal support, and aver- Descriptives and Correlations
age discrepancy scores. The total support, total ideal
support, or discrepancy is summed across all the sig- Significant Others Scale (SOS)
nificant others and divided by the number of significant
others listed. This controls for the number of signifi- The SOS was completed by 488 individuals; 205 (42.0%)
cant others present in an individual’s network; that is, men and 283 (58.0%) women. Of this number, 220 (45.1%)
an individual is not penalized if they do not have sib- were married at the time of the mailing, 265 (54.3%) had
lings, for example. a close sibling, and 137 (28.1%) had another sibling. The
2. Household composition – participants filled in details of majority of participants had a close child (394; 80.7%) with
their age-11 and current (age-80) household composi- a similar percentage recording that they had a best friend
tion. They were first asked how many people were in (399; 81.8%). Overall, the majority of participants listed
their house at age 11. Next, they were asked how many two or three significant others; 147 (30.1%) and 137
individuals they currently shared their home with, even (28.1%) participants, respectively, giving a mean of 2.9
if they were not related. This gave the total number of (SD = 1.2). The number of significant others was not sig-
people they lived with at ages 11 and 80 (an indicator of nificantly related to any of the cognitive measures either in
their physical contact with others). The latter informa- the full sample (Table 1), or separately in men and women
tion was used to create a binary variable for living alone (Table 2). There was a small positive association between
at age 80 versus living with others. the number of significant others and life satisfaction scores
3. Loneliness – participants were asked two specific ques- in the full sample (r = .14, p = .003) suggesting that those
tions regarding feeling lonely. First, they were asked to with a greater number of close contacts tended to be more
rate on a 5-point scale (most of the time to never) wheth- satisfied, although the association was only significant in
er or not they felt lonely at the present time. The other men when the cohort was split (Table 2).
item simply asked whether the participants felt they had When the presence of each of the significant others was
people to talk to when they had problems, and required considered (Table 3), simply having a spouse or close child
a yes/no response. was related to life satisfaction scores, although the associ-
ations were again small; those with a spouse had signifi-
cantly higher satisfaction-with-life scores, mean SWLS of
Statistical Analysis 26.1 (6.0), compared with 24.7 (6.2); t(477) = –2.14, p =
.016, Cohen’s d = .22, as did those with a close child, mean
Descriptive statistics for the social support and network vari- SWLS of 25.6 (6.1) compared with 24.2 (6.1); t(477) =
ables are presented, followed by an examination of the uni- –2.04, p = .042, Cohen’s d = .24. Having a spouse was also
variate associations of each of these with age-11 IQ, age-79 positively related to age-79 IQ (Table 3) such that those
IQ, and life satisfaction ratings. Significant univariate asso- with a spouse had a significantly higher age-79 IQ, mean
ciations with age-79 IQ and satisfaction with life are used to of 102.5 (13.9) compared with 99.6 (14.9); t(464) = –2.15,
Table 1. IQ and satisfaction with life associations with social network and support factors
1 2 3 4 5 6 7 Mean (SD)
1. Age-11 IQ – .66*** .02 –.05 .05 .10* .07 100.0 (15.0)
2. Age-79 IQ – .06 .00 .08 .11* .04 100.0 (15.0)
3. SWLS – .14** .25*** .15** –.17*** 25.3 (6.1)
4. Number of significant others – –.10* –.08 .04 2.9 (1.2)
5. Average support – .77*** –.41*** 22.6 (4.4)
6. Average ideal support – .27*** 24.3 (4.2)
7. Average support discrepancy – 1.7 (2.9)
Note. SWLS = Satisfaction With Life Scale; Average support, average ideal support and average support discrepancy = the total support, total
ideal support, and total discrepancy summed across all the significant others, divided by the number of significant others listed. *p < .05, **p
< .01, ***p < .001
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
108 A.J. Gow et al.: Social Support and Successful Aging
Table 2. IQ and satisfaction with life associations with social network and support factors by gender
1 2 3 4 5 6 7
1. Age-11 IQ – .63*** .04 –.05 –.01 .11 .15*
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
Table 3. IQ and satisfaction with life associations with significant other variables
Age-11 IQ Age-79 IQ SWLS
Present Present Present
T M F A I D T M F A I D T M F A I D
Spouse .01 .04 .07 –.12 –.07 .10 .10* .04 .11 –.03 .00 .04 .11* .15* .05 .36*** .25*** –.30***
Closest sibling –.02 –.05 .00 .08 .06 –.04 .01 .01 .01 .07 .03 –.06 .02 .02 .01 .13* .05 –.13*
Other sibling –.11* –.13 –.09 .13 .06 –.12 –.06 –.07 –.06 .20* .08 –.20* .07 .02 .11 .12 .03 –.14
Closest child –.01 .09 –.06 –.12* .02 .15*** –.01 .09 –.08 –.10* –.03 .13* .09* .11 .08 .18*** .17** –.05
Best friend .00 –.04 .02 .02 .10 .12* –.05 –.09 .00 .06 .03 .10* .08 .13 .04 .23*** .11* –.19***
Note. Present = correlations with the presence or absence of the named significant other and the outcome for the total sample (T), and male (M)
and female (F) subgroups separately; A = correlations between the actual level of social support received from the named significant other and
the outcome, excluding those participants without this significant other; similarly, I = ideal level of social support and D = discrepancy between
ideal and actual level of social support. *p < .05, **p < .01, ***p < .001.
p = .032, Cohen’s d = .20. Having another sibling was neg- association with age-79 IQ remained in women when the
atively associated with age-11 IQ, mean age-11 IQ was sample was split (Table 2).
97.9 (15.5) in those with another sibling compared with The associations between the level of support received
101.4 (14.3) in those without; this difference was signifi- from each significant other and age-11 IQ, age-79 IQ, and
cant: t(422) = 2.18, p = .030, Cohen’s d = –.234. When satisfaction with life were then investigated (Table 3).
examined separately by gender (Table 3), the association When analyzing the associations with the level of spousal
between the presence of a spouse and greater life satisfac- support, we excluded people without a spouse. Exclusions
tion was only found in men; however, the presence or ab- were made in the same manner for the associations with the
sence of each of the significant others was not related to level of support from the closest and other sibling, closest
cognitive ability when the sample was split. child, and best friend. The level of support received from
The average level of social support received (the total each significant other was significantly positively related
support received divided by the number of significant oth- to satisfaction with life (except other sibling), with associ-
ers providing it) was positively related to satisfaction with ations ranging from .13 for closest sibling to .36 for spouse
life in the full sample (r = .25, p = .000: Table 1) and sep- (both p < .05; Table 3). That is, the higher the support from
arately in men and women (Table 2). Similarly, the partic- each of the people in a person’s social network, the greater
ipants’ average ideal level of social support was related to their life satisfaction.
life satisfaction ratings (r = .15, p = .001), although this was Those who received increased support from another sib-
only significant for men when the sample was split (r = .22, ling (not their closest) had higher age-79 IQ (r = .20, p =
p = .002). The discrepancy between the average level of .023: Table 3). Conversely, higher levels of support from a
support received and the participant’s average ideal level closest child were negatively related to IQ at ages 11 and
(ideal minus actual support) was negatively related to 79, r = –.12 (p = .030) and –.10 (p = .045), respectively.
SWLS (r = –.17, p = .000), which suggests that the most For each significant other, it was also possible to calculate
satisfied individuals are those who receive support that a discrepancy measure (the ideal level of support minus
more closely matches their desired level (this was only sig- actual support received), again excluding those who did not
nificant in women: r = –.31, p = .000). The average ideal list the specified significant other (Table 3). The discrep-
level of support was positively associated with both age-11 ancies between the actual support received and the ideal
IQ and age-79 IQ in the full sample (Table 1), but only the level from spouse, closest sibling, and best friend were neg-
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 109
atively related to SWLS; r = –.30 (p = .000), –.13 (p = .037) discrepancy and satisfaction with life in those with a
and –.19 (p = .042), respectively; a closer match between spouse, these were not included in the regression analyses
ideal and actual support (a lower discrepancy score) is as- because of the reduced sample sizes that would have oc-
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
110 A.J. Gow et al.: Social Support and Successful Aging
Table 4. Summary of regression analysis for social network/support variables predicting satisfaction with life
Variable B SE B β Adjusted multiple R2 Adjusted R2 change
Model 1 .02 .02*
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
Table 5. Summary of regression analysis for social network/support variables predicting satisfaction with life in women
Variable B SE B β Adjusted multiple R2 Adjusted R2 change
Model 1 .00 .01
Years education .09 .19 .03
Social class –.42 .46 –.06
Model 2 .16 .16***
Years education .02 .17 .01
Social class –.21 .43 –.03
Loneliness –2.28 .33 –.40***
Model 3 .22 .06***
Years education .10 .17 .04
Social class –.21 .41 –.03
Loneliness –2.01 .32 –.36***
Talk problems 7.61 1.69 .25***
Model 4 .23 .02*
Years education .10 .17 .04
Social class –.25 .41 –.04
Loneliness –1.91 .32 –.34***
Talk problems 6.54 1.74 .22***
Average social support .18 .08 .14*
Note. Talk problems = having someone to talk to about problems (yes/no); Average social support = the total support summed across all the
significant others, divided by the number of significant others listed. *p < .05, ***p < .001.
all, the final model accounted for about 23% of the variance The regressions were run separately for men and women
in SWLS scores, with the largest contribution from loneli- (with years of education and social class entered in Block
ness (β = –.35, p = .000, about 12% of the variance), fol- 1). In women, average support, loneliness, and having
lowed by having someone to talk to about problems (β = someone to talk to were entered stepwise in Block 2. The
.18, p = .000, about 6% of the variance). The regression results (summarized in Table 5) suggest that these variables
was repeated, but average support discrepancy was entered accounted for about 23% of the variance in life satisfaction
rather than average support and average ideal support. The ratings. When average support discrepancy was entered in-
results were essentially unchanged, with average support stead of average support, loneliness accounted for ~13%,
discrepancy accounting for about 1% of the variance in life followed by average support discrepancy (~5%) and hav-
satisfaction ratings. ing someone to talk to (~4%). In men, the number of sig-
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 111
Table 6. Summary of regression analysis for social network/support variables predicting satisfaction with life in men
Variable B SE B β Adjusted multiple R2 Adjusted R2 change
Model 1 .03 .04*
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
nificant others listed, spouse (yes/no), average support and Sherman, & Antonucci, 1998). In agreement with our life-
average ideal support, living alone, having someone to talk span cognitive change findings, previous research by See-
to, and loneliness were entered stepwise in Block 2, with man and colleagues (2001) suggested that the number of
the results summarized in Table 6. Loneliness again ac- close ties did not predict cognitive change over 7.5 years,
counted for the greatest proportion of the variance in satis- indicating that it may be that the adequacy of the support
faction with life ratings, accounting for about 14%. is of more importance than how many individuals are in the
social network.
Interestingly, living with a greater number of people at
age 11 was negatively associated with childhood and later
Discussion adulthood mental ability. It would be suggested that a more
overcrowded home in childhood might be a marker of so-
This study reported associations between social network cioeconomic status or deprivation. Furthermore, when
and support characteristics to age-79 IQ and satisfaction asked at age 80 about the presence of a range of significant
with life in a cohort of relatively healthy elderly individ- others, individuals who reported having another sibling (in
uals. After adjustment for sex, years of education, social addition to their closest sibling, that is, they were part of a
class, and age-11 IQ, loneliness remained a significant pre- larger family) scored lower at age 11. This item appears to
dictor of age-79 IQ, accounting for about 2% of the vari- be a very rough proxy for family size. The intelligence-
ance. Individuals reporting higher loneliness had poorer family size association was described in detail in a later
cognitive function at age 79. Social network and support Scottish Mental Survey, carried out in 1947 (Scottish
factors explained 23% of the variance in satisfaction with Council for Research in Education, 1949). Family size and
life ratings, with the greatest contributions to higher satis- overcrowding are closely related, and both are influenced
faction with life coming from reduced loneliness (~12%) by parental social class. It is, therefore, likely that lower
and having someone to talk to (~6%). social class and increased deprivation, or other associated
detrimental environmental factors, explain the relationship
between family size, household composition, and child-
Social Network Size hood ability, rather than there being a direct link (Scottish
Council for Research in Education, 1949). As childhood
These data show that having a greater number of individ- household composition and family size (indexed by having
uals in one’s social support network is not associated with another sibling in this instance) appear to be markers of
enhanced cognitive outcomes, and the impact on life satis- childhood circumstances rather than the presence of social
faction is small. The total number of significant others was networks or support, they will not be discussed further.
limited in this instance by the nature of the measurement
used; it remains possible that given an open response for-
mat, greater individual variability in the number of close Significant Others, Social Support and
social ties reported would be seen. Carstensen’s socioemo- Successful Aging
tional selectivity theory suggests that as individuals age,
they reduce the number of close contacts in their social The results from the LBC1921 suggest that the presence of
network, becoming more selective in their relationships; key significant others is associated with cognitive ability
however, emotional closeness to these (fewer) individuals or satisfaction with life in old age. For example, those who
increases (Carstensen, 1995; Carstensen, 1992; Lansford, had a spouse or a closest child had significantly higher life
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
112 A.J. Gow et al.: Social Support and Successful Aging
satisfaction ratings. Additionally, the level of support re- measured across the lifespan. Social factors may, therefore,
ceived from the spouse had the largest association (of all predict future, shorter-term changes in the LBC1921. Yet
the named significant others) with satisfaction with life, even over 7.5 years of follow-up in the MacArthur study
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
although it did not account for additional, independent vari- (Seeman et al., 2001), the level of emotional support re-
ance in the regression analyses. In a range of studies sum- ceived by individuals accounted for only 1% of the vari-
marized by Diener (1984), being married generally led to ance in cognitive change (a smaller percentage than that
increased subjective well-being. Indeed, a lasting marriage accounted for by loneliness in the current study). It is un-
adds a similar “worth” to life satisfaction ratings as does clear whether this would remain a significant predictor
an extra $100,000 income per annum (Blanchflower & Os- were it possible to control for an earlier measure of ability,
wald, 2004). as is possible with the LBC1921.
Those who were married also had better cognitive per- The current study suggests that increased social support,
formance in later life, with significantly higher age-79 IQ or a lower discrepancy between this and an individual’s
compared with those who were unmarried. This is in agree- ideal level, is associated with increased satisfaction with
ment with the finding that never-married individuals are at life. As previously reported by other researchers (Diener &
an increased risk of developing dementia (Helmer et al., Seligman, 2002; Ho et al., 1995; Okabayashi et al., 2004)
1999), although contradictory findings have been reported. good social relationships or support may be important for
In another study, unmarried individuals were shown to life satisfaction.
have better cognitive function at baseline (Seeman et al.,
2001), however, this relationship was no longer significant
in the longitudinal follow-up of these individuals.
And yet, no particular individual in the support network Living Alone and Feeling Alone
is more necessary than others, and there is a substitutability
of particular contacts (Bassuk et al., 1999). Seeman and Living alone was associated with poorer age-79 IQ and
Berkman (1988) reported that neither a spouse nor children lower life satisfaction in the current study, as has been pre-
are considered the primary sources of support, based on a viously shown for cognitive outcomes (Fratiglioni et al.,
study of almost 3,000 community dwelling, elderly indi- 2000). This finding is in agreement with previous literature,
viduals (aged 65 and over). This, however, does not rule whereby living alone was a risk factor for cognitive decline
out how important a spouse, for example, may be for cog- or dementia (Fratiglioni et al., 2000; van Gelder et al.,
nitive outcomes: Although the spouse may not be the most 2006). Living alone might be a marker for a lack of cogni-
important source of support received, their presence may tive stimulation from other individuals or a reduction in
play a role in increasing engagement, for instance. An in- social support, leading to poorer mental health outcomes.
creased level of support from the closest child is associated It is also possible that a transition to being alone has led to
with poorer mental ability at both ages 11 and 79. However, adverse lifestyle changes or increased depression, which
this increased support is perhaps more likely to be a con- are driving the association (van Gelder et al., 2006). Re-
sequence rather than a cause of poorer cognitive function. gardless of any hypothesized pathways, however, it is im-
Those with a lower ability in early life are more likely to portant to note that living alone versus with others did not
have a lower ability in later life; accordingly, such individ- enter into the regression predicting age-79 IQ. Continued
uals may be likely to have an increased dependency on follow-up with the cohort will help to determine whether
others and require extra help in their everyday lives and, living situation is associated with further cognitive change
thus, report greater levels of support. in the LBC1921, or whether transitions from one situation
While many previous studies report a link between de- to another are predictive of decline.
creased social support and poorer cognitive outcomes (Fra- However, simply living alone is not what may be the
tiglioni et al., 2004; Seeman et al., 2001), this was not ap- most important risk factor with respect to cognition, but
parent in the current study (only the level of support from rather actually being alone (Berkman, 2000). The current
another sibling or closest child were related to age-79 IQ). findings would support the view that it is feeling alone that
However, not all studies report such a finding: Okabayashi has the strongest association with cognition; loneliness was
et al. (2004) did not find a link between social exchanges the only significant predictor of age-79 IQ after adjusting
and cognitive outcomes in individuals who were married for early ability. It is plausible that sex is a confounder of
with children, although there was an effect of increased the relationship between loneliness and later life cognition,
support from children in those individuals without a as women may be more likely to have lost a spouse (be-
spouse. Mismatched results in this domain may be a reflec- cause of the shorter lifespan of men), and so might be ex-
tion of the different measures used to assess perceived and pected to experience greater feelings of loneliness. In the
actual support (Jones, Rapport, Hanks, Lichtenberg, & Tel- regression analysis, loneliness accounted for ~2% of the
met, 2003). The discrepancy may also be the result of the variance in age-79 IQ, which was adjusted for sex, educa-
fact that, in the current study, later-life cognition and social tion, social class, and age-11 IQ (although in the gender-
contacts were measured almost simultaneously, but cogni- specific analyses loneliness was not related to cognitive
tive change (age-79 IQ controlled for childhood IQ) was ability in men). That is, individuals who showed more neg-
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 113
ative cognitive change across the lifespan were lonelier in at present. Throughout, it has been suggested that social
later life. support (or not being lonely) is protective of a cognitive
Loneliness is clearly an important factor in people’s ability, but this is only one possible explanation for the as-
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
lives and may indicate a lack of possibility for social en- sociations. Those who maintain their cognitive abilities in-
gagement. Although not specifically assessing the structure to old age may be better able to retain a close, supportive
of an individual’s social network, or the actual level of so- network, and to seek help when needed, and, therefore, may
cial support received, the level of loneliness experienced be less likely to experience increased loneliness with age.
by an individual may indicate both an inadequate network An, as yet, unmeasured confounder may also underlie the
and a lack of support. Therefore, it would not be feeling association which would predict both social support and
lonely per se which is cognitively disadvantageous, but mental ability. Similarly, the life satisfaction findings are
rather it is the inherent lack of social support, contact, or currently cross-sectional: Although increased loneliness or
integration within a wider social network, which is driving poorer support might lead to reduced satisfaction with life,
the effect. The impact of loneliness appears to span do- it is also possible that individuals who are less satisfied
mains of successful aging as it also emerges as an important might be more likely to rate the quality of their relation-
predictor of life satisfaction. Feeling lonely would suggest ships as poor, or report being more lonely.
an inadequate or absent social support network and, thus, Because of the timing of the assessments, it is not pos-
limited opportunities for social engagement. sible to ascertain the directionality of the relationships re-
It is possible that those who are experiencing cognitive ported. However, the LBC1921 follow-up is continuing
decline are less able to gauge appropriately how lonely they and will be able to address this issue. The cohort are cur-
are, or that they have chosen to remove themselves from rently undergoing repeat cognitive testing, now aged about
social situations or have been left socially isolated because 83. Therefore, temporal (possibly causal) relationships be-
of this decline (Berkman, 2000). In this explanation, lone- tween social support and change in cognitive function and
liness would be a consequence rather than cause of cogni- quality of life between the ages of 79 and 83 will then be
tive decline. This important point requires further investi- able to be assessed.
gation before it would be possible to determine which caus-
al pathway takes precedence. Moreover, it is necessary to
consider whether the detrimental effect of feeling lonely (if
that is the causal direction) occurs later in life, or whether
there is a cumulative effect across the lifespan. This latter
Conclusions
issue would have implications for any suggested interven-
tions (Berkman, 2000). After controlling for the effects of childhood ability, sex,
education, and social class in this relatively healthy elderly
cohort, loneliness was the only feature of social support
Limitations of the Current Study related to later life ability. This is important as many studies
are unable to control for early ability, which is necessary
One of the major strengths of the current study is the pres- in order to determine what additional impact other factors
ence of a measure of early ability, which can be controlled may have on later functioning. Life satisfaction in old age
for when predicting later functioning; however, the study was much more strongly related to aspects of social sup-
also has limitations. For example, in later life, the method port, with loneliness accounting for the highest proportion
of assessing social support restricted the participants to five of the variance. Remedies for loneliness in the elderly pro-
named significant others. It is possible that a wider range vide one of several routes to improved quality of life for
of responses would have been received were the partici- older individuals.
pants given a freer choice. There is also debate about the
utility of discrepancy scores that can be analyzed as the
simpler constituents (for example Edwards, 1994), which
for current purposes would be: how much support would References
be ideally wanted, and how much support was obtained.
Discrepancy scores were computed following the scoring
Antonucci, T.C., Lansford, J.E., & Akiyama, H. (2001). Impact
procedures of the SOS (Power et al., 1988), but the simpler
of positive and negative aspects of marital relationships and
elements were also analyzed separately. friendships on well-being of older adults. Applied Develop-
In addition, other than the early cognitive data, only ba- mental Science, 5, 68–75.
sic background information was available from the early Barnes, L.L., Mendes de Leon, C.F., Wilson, R.S., Bienias, J.L.,
part of the lifespan in the LBC1921 participants making the & Evans, D.A. (2004). Social resources and cognitive decline
analyses primarily cross-sectional. The social contact and in a population of older African-Americans and whites. Neu-
support measures were administered shortly after the age- rology, 63, 2322–2326.
79 cognitive testing, making any casual explanations of the Bassuk, S.S., Glass, T.A., & Berkman, L.F. (1999). Social dis-
age-79 cognition and social support associations tentative engagement and incident cognitive decline in community-
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115
114 A.J. Gow et al.: Social Support and Successful Aging
dwelling elderly persons. Annals of Internal Medicine, 131, General Register Office (1956). Census 1951: Classification of
165–173. occupations. London: Her Majesty’s Stationary Office.
Baumeister, R.F., & Leary, M.R. (1995). The need to belong: Gow, A.J., Whiteman, M.C., Pattie, A., & Deary, I.J. (2005). The
Desire for interpersonal attachments as a fundamental human personality-intelligence interface: Insights from an aging co-
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
motivation. Psychological Bulletin, 117, 497–529. hort. Personality and Individual Differences, 39, 751–761.
Berkman, L.F. (2000). Which influences cognitive function: Helmer, C., Damon, D., Letenneur, L., Fabrigoule, C., Barber-
Living alone or being alone? The Lancet, 355, 1291–1292. ger-Gateau, P., Lafont, S. et al. (1999). Marital status and risk
Blanchflower, D.G., & Oswald, A.J. (2004). Well-being over of Alzheimer’s disease. Neurology, 53, 1953–1958.
time in Britain and the USA. Journal of Public Economics, Hendrie, H.C., Albert, M.S., Butters, M.A., Gao, S., Knopman,
88, 1359–1386. D.S., Launer, L.J. et al. (2006). The NIH Cognitive and Emo-
Bowling, A.P., Edelmann, R.J., Leaver, J., & Hoekel, T. (1989). tional Health Project: Report of the Critical Evaluation Study
Loneliness, mobility, well-being, and social support in a sam- Committee. Alzheimer’s and Dementia, 2, 12–32.
ple of over-85-year-olds. Personality and Individual Differ- Ho, S.C., Woo, J., Lau, J., Chan, S.G., Yuen, Y.K., Chan, Y.K.
ences, 10, 1189–1192. et al. (1995). Life satisfaction and associated factors in older
Carstensen, L.L. (1992). Social and emotional patterns in adult- Hong Kong Chinese. Journal of the American Geriatrics So-
hood: Support for socioemotional selectivity theory. Psychol- ciety, 43, 252–255.
ogy and Aging, 7, 331–338. Jones, T.G., Rapport, L.J., Hanks, R.A., Lichtenberg, P.A., &
Carstensen, L.L. (1995). Evidence for a life-span theory of so- Telmet, K. (2003). Cognitive and psychological predictors of
cioemotional selectivity. Current Directions in Psychological subjective well-being in urban older adults. The Clinical Neu-
Science, 4, 151–156. ropsychologist, 17, 3–18.
Cohen, S., & Wills, T.A. (1985). Stress, social support, and the Jorm, A.F., Christensen, H., Henderson, A.S., Jacomb, P.A.,
buffering hypothesis. Psychological Bulletin, 98, 310–357. Korten, A.E., & Mackinnon, A. (1998). Factors associated
Deary, I.J., Whalley, L.J., Batty, G.D., & Starr, J.M. (2006). with successful aging. Australasian Journal on Aging, 17,
Physical fitness and lifetime cognitive change. Neurology, 67, 33–37.
1195–1200. Lansford, J.E., Sherman, A.M., & Antonucci, T.C. (1998). Sat-
Deary, I.J., Whalley, L.J., Lemmon, H., Crawford, J.R., & Starr, isfaction with social networks: An examination of socioemo-
J.M. (2000). The stability of individual differences in mental tional selectivity theory across cohorts. Psychology and Ag-
ability from childhood to old age: Follow-up of the 1932 ing, 13, 544–552.
Scottish Mental Survey. Intelligence, 28, 49–55. Melchior, M., Berkman, L.F., Niedhammer, I., Chea, M., &
Deary, I.J., Whiteman, M.C., Starr, J.M., Whalley, L.J., & Fox, Goldberg, M. (2003). Social relations and self-reported
H.C. (2004). The impact of childhood intelligence on later health: A prospective analysis of the French Gazel cohort.
life: Following up the Scottish Mental Surveys of 1932 and Social Science and Medicine, 56, 1817–1830.
1947. Journal of Personality and Social Psychology, 86, National Research Council (2000). The aging mind: Opportuni-
130–147. ties in cognitive research. Washington, DC: National Acade-
Diener, E. (1984). Subjective well-being. Psychological Bulle- my Press.
tin, 94, 542–575. Okabayashi, H., Liang, J., Krause, N., Akiyama, H., & Sugi-
Diener, E., Emmons, R.A., Larsen, R.J., & Griffin, S. (1985). sawa, H. (2004). Mental health among older adults in Japan:
The Satisfaction with Life Scale. Journal of Personality As- Do sources of social support and negative interaction make a
sessment, 49, 71–75. difference? Social Science and Medicine, 59, 2259–2270.
Diener, E., & Seligman, M.E.P. (2002). Very happy people. Psy- Power, M.J., Champion, L.A., & Aris, S.J. (1988). The develop-
chological Science, 13, 81–84. ment of a measure of social support: The Significant Others
Diener, E., & Seligman, M.E.P. (2004). Beyond money: Toward (SOS) Scale. British Journal of Clinical Psychology, 27,
an economy of well-being. Psychological Science in the Pub- 349–358.
lic Interest, 5, 1–31. Rowe, J.W., & Kahn, R.L. (1987). Human aging: Usual and suc-
Edwards, J.R. (1994). Regression analysis as an alternative to cessful. Science, 237, 143–149.
difference scores. Journal of Management, 20, 683–689. Scottish Council for Research in Education (1933). The intelli-
Fillit, H.M., Butler, R.N., O’Connell, A.W., Albert, M.S., Birren, gence of Scottish children: A national survey of an age-group.
J.E., Cotman, C.W. et al. (2002). Achieving and maintaining London: University of London Press, Ltd.
cognitive vitality with aging. Mayo Clinic Proceedings, 77, Scottish Council for Research in Education (1949). The trend of
681–696. Scottish intelligence. London: University of London Press.
Folstein, M.F., Folstein, S.E., & McHugh, P.R. (1975). “Mini- Seeman, T.E., & Berkman, L.F. (1988). Structural characteristics
mental state.” A practical method for grading the cognitive of social networks and their relationship with social support
state of patients for the clinician. Journal of Psychiatric Re- in the elderly: Who provides support? Social Science and
search, 12, 189–198. Medicine, 26, 737–749.
Fratiglioni, L., Paillard-Borg, S., & Winblad, B. (2004). An ac- Seeman, T.E., Lusignolo, T.M., Albert, M., & Berkman, L.
tive and socially integrated lifestyle in late life might protect (2001). Social relationships, social support, and patterns of
against dementia. The Lancet Neurology, 3, 343–353. cognitive aging in healthy, high-functioning older adults:
Fratiglioni, L., Wang, H.-X., Ericsson, K., Maytan, M., & Win- MacArthur studies of successful aging. Health Psychology,
blad, B. (2000). Influence of social network on occurrence of 20, 243–255.
dementia: A community-based longitudinal study. The Lan- Uchino, B.N., Cacioppo, J.T., & Kiecolt-Glaser, J.K. (1996).
cet, 355, 1315–1319. The relationship between social support and physiological
Journal of Individual Differences 2007; Vol. 28(3):103–115 © 2007 Hogrefe & Huber Publishers
A.J. Gow et al.: Social Support and Successful Aging 115
processes: A review with emphasis on underlying mecha- ment determine cognitive decline in community-dwelling
nisms and implications for health. Psychological Bulletin, Spanish older adults. Journal of Gerontology, 58B, S93–S100.
119, 488–531.
van Gelder, B.M., Tijhuis, M., Kalmijn, S., Giampaoli, S., Nis-
https://econtent.hogrefe.com/doi/pdf/10.1027/1614-0001.28.3.103 - Sunday, November 24, 2019 11:16:30 PM - Universitätsbibliothek LMU München IP Address:129.187.254.47
© 2007 Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):103–115