Journal of Gerontology: PSYCHOLOGICAL SCIENCES
1998, Vol. 53B, No. 2, P96-P104
Copyright 1998 by The Gerontological Society of America
The Role of Self-Perceived Usefulness and Competence
in the Self-Esteem of Elderly Adults:
Confirmatory Factor Analyses of the Bachman
Revision of Rosenberg's Self-Esteem Scale
Rob Ranzijn, John Keeves, Mary Luszcz, and N. T. Feather
School of Psychology, The Flinders University of South Australia, Adelaide.
This article reports on a confirmatory analytic study of the Bachman revision (1970) of Rosenberg's Self-Esteem
Scale (1965) that was used in the Australian Longitudinal Study of Ageing (ALSA). Participants comprised 1,087
elderly people aged between 70 and 103 years (mean 77 years). Five competing factor models were tested with
LISREL8. The best-fitting model was a nested one, with a General Self-esteem second-order factor and two firstorder factors, Positive Self-regard and Usefulness/Competence. This model was validated with data from a later
wave of ALSA. Usefulness and competence have received little attention in the gerontological literature to date. Preliminary results indicate that usefulness/competence may be an important predictor of well-being. Further work is
required on the relationships among usefulness, competence, self-esteem, and well-being in elderly people.
S
ELF-ESTEEM has a well-recognized relationship to
psychological well-being (Blascovich & Tomaka,
1991). Self-esteem can be conceptualized as self-regard, an
evaluation of one's worthiness (Rosenberg, Schooler,
Schoenbach, & Rosenberg, 1995). A number of scales have
been designed to measure self-esteem (Blascovich &
Tomaka, 1991), but there has been little research into the
psychometric properties of these scales, particularly using
data from elderly people. The aim of this research was to
assess the psychometric properties and.factor structure of a
self-esteem scale suitable for the study of elderly adults,
the Bachman revision (Bachman, 1970) of Rosenberg's
Self-Esteem Scale (RSE; Rosenberg, 1965).
Rosenberg's Self-Esteem Scale (Rosenberg, 1965) is the
most widely used measure of self-esteem (Blascovich &
Tomaka, 1991). It was designed for research into adolescent self-esteem (Rosenberg, 1965). However, it has been
used in well-being research on all age groups, and is considered to be an appropriate measure of self-esteem in the
elderly (Breytspraak & George, 1979). Recent studies that
have used it with older adults include those of Duffy and
MacDonald (1990) and Krause (1987).
Rosenberg (1965) designed the Self-Esteem Scale as a
unidimensional measure of global self-esteem. Other researchers, building on his work, subsequently designed
scales to measure different dimensions of self-esteem and
in some cases domain-specific self-esteem (Blascovich &
Tomaka, 1991). A number of principal components factor
analyses performed in the 1970s supported Rosenberg's
view that the RSE was unidimensional. However, some
investigators identified two highly correlated factors, positive and negative self-esteem (reviews in Blascovich &
Tomaka, 1991; Carmines & Zeller, 1979). In some studies,
such as research into the psychological impact of unem-
P96
ployment, positive self-esteem and negative self-esteem
have been used as separate variables (Feather & Bond,
1983; Warr & Jackson, 1983). However, Carmines and
Zeller (1979) claimed that the putative two-factor structure
was likely to be a statistical artifact due to response set,
since some items are positively worded while others are
negatively worded. They concluded that, instead of positive
and negative self-esteem, the two factors indicated by principal components analysis represented a global self-esteem
factor and systematic error variance due to the valence of
the item wording. The negatively worded items in the SelfEsteem Scale seem to describe the opposite poles, on the
same continuum, to the positively worded ones rather than
a conceptually different dimension. For instance, a positively worded item ("I take a positive attitude towards myself) is at the opposite pole to a negatively worded item
("At times I think I am no good at all"), both items referring to evaluation of worthiness. This could explain the
high correlation between the two putative factors.
In the last few years there have been major developments
in factor analysis. Confirmatory factor analysis (CFA), in
which a theoretically derived factor structure is tested for
its fit to the data, is now commonly used. Shahani, Dipboy e, and Phillips (1990) used confirmatory factor analysis
on the original RSE and reported that a structure with two
first-order factors, self-derogation and self-enhancement
(their terms for negative and positive self-esteem), was superior to a unidimensional one. Recently, Rosenberg
seemed to have departed from his earlier unidimensional
views concerning his scale and to have accepted the twofactor model ascribed to by Shahani et al. (Rosenberg et al.,
1995). A recent CFA study of the RSE was conducted by
Marsh (1996), who used a modified seven-item version. He
argued that his results supported the reasoning of Carmines
USEFULNESS AND SELF-ESTEEM
and Zeller (1979), that the second putative factor was a statistical artifact due to response set. Therefore, the evidence
is equivocal and the structure of the Self-Esteem Scale has
remained unclear.
The present study does not examine the original SelfEsteem Scale (Rosenberg, 1965) but the Bachman revision,
a scale that arose from a large set of personality items used
in the Youth in Transition project (Bachman, 1970). The
items of the revised scale are shown in Table 1. To date, the
Bachman revision has not been subjected to confirmatory
factor analysis. The revision is similar to Rosenberg's SelfEsteem Scale, both consisting of 10 items. However, the revision is an abbreviated composite of two measures, the
Self-Esteem Scale and a 10-item attitude scale devised by
Cobb, Brooks, Kasl, and Connelly (1966, in Bachman &
O'Malley, 1977) for a study of adult workers who had
changed jobs. In the process of reducing the composite
scale to 10 items, four items from the Self-Esteem Scale
and six from the scale of Cobb et al. were eliminated. The
Cobb et al. items that were retained (items 7, 8, 9, and 10 in
Table 1) were said to be "similar in content" to the SelfEsteem Scale items that were replaced (Bachman & O'Malley, 1977, p. 368). Bachman (1970) regarded the revised
scale as a unidimensional measure of global self-esteem.
This view has been accepted in subsequent research that
has used the revision (Bachman & O'Malley, 1986; Bynner,
O'Malley, & Bachman, 1981; Rosenberg & Rosenberg,
1978). However, the wording of the four replacement scale
items emphasizes usefulness and competence, while the
four that were replaced (scale of Rosenberg, 1965) describe
positive self-regard. Hence, there is reason to suppose the
factor structure of the Bachman revision is different from
that of the original RSE.
The Bachman revision, (called RSE-B subsequently for
the sake of brevity) has been used extensively in research
(e.g., Bachman & O'Malley, 1977, 1984, 1986; Bachman &
Schulenberg, 1993; Bynner et al., 1981; O'Malley & Bachman, 1979, 1983; Rosenberg & Rosenberg, 1978; Rosenberg et al., 1995; Wells & Sweeney, 1986). It has a reported
reliability of 0.75 (Cronbach's a) and good construct validity (Bachman & O'Malley, 1977).
It was noted above that the four replacement items contain references to usefulness and competence. Specifically,
items 8 and 9 in the RSE-B (see Table 1) contain words reTable 1. Bachman Revision (1970)
of Rosenberg's Self-Esteem Scale
Item
1.
2.
3.
4.
5.
6.
*7.
*8.
*9.
*10.
I feel that I'm a person of worth, at least on an equal plane with others.
I feel that I have a number of good qualities.
I am able to do things as well as most other people.
I feel that I do not have much to be proud of.
I take a positive attitude towards myself.
I think I am no good at all.
I am a useful person to have around.
I feel I can't do anything right.
When I do a job, I do it well.
I feel that my life is not very useful.
Note: Scale items marked * were added by Bachman (1970).
P97
lated to competence, whereas items 7 and 10 explicitly
refer to usefulness. Of the six items retained from the RSE,
item 3 seems to refer to competence and item 4 to achievement, a concept related to usefulness and competence. The
references to usefulness could make the revised scale especially applicable to the assessment of self-esteem in the elderly, since to feel useful is thought to be important to the
psychological well-being of this group (Butler, 1985; Ryff,
1989). The four items not yet discussed (items 1, 2, 5, and
6) refer to positive self-regard, the essence of self-esteem
(Rosenberg et al., 1995). We therefore argue that the RSEB may contain a usefulness/competence factor as well as a
self-esteem factor.
To summarize the argument so far, two possible factor analytic models for the RSE-B could be derived from the published literature to date, which has assumed that the RSE-B
represents the same construct as the RSE. The first possible
model comprises a unidimensional structure, measuring
only self-esteem (Bachman & O'Malley, 1986; Bynner et
al., 1981; Rosenberg & Rosenberg, 1978). The second
model comprises two first-order factors, self-derogation and
self-enhancement (following Shahani et al., 1990). A third
model, derived from the history of the scale, in which an attitude scale designed for adult workers was imported into
the existing self-esteem scale, and from an inspection of the
wording of the items, comprises two first-order factors, positive self-regard and usefulness/competence.
When performing confirmatory factor analysis, it is considered good practice to develop a number of competing
models and see which best fits the data (Hertzog, 1990;
Liang & Bollen, 1983; Marsh, 1996)," and this approach
was used in this research. The first two competing hypotheses were as follows.
Hypothesis I. The RSE-B is unidimensional and measures a
factor called global self-esteem. The factor structure to test
this hypothesis was labeled model 1. All 10 items were specified to load onto the global self-esteem factor.
Hypothesis 2. The RSE-B has a two-factor structure, the factors being self-enhancement and self-derogation. The factor
structure to test this hypothesis was labeled model 2. Items
1, 2, 3, 5, 1* and 9* were specified to load on the selfenhancement factor, and items 4, 6, 8,* and 10* on the selfderogation factor (* refers to items added by Bachman,
1970).
Model 3 was derived from the history of the scale and
from inspection of the wording of the scale items. The next
competing hypothesis was:
Hypothesis 3. The RSE-B has a two-factor structure, the factors being positive self-regard and usefulness/competence.
The first factor was labeled positive self-regard rather than
self-esteem to distinguish it from the general self-esteem
factor of models 4 and 5 (see below), but positive self-regard
is conceptualized in this article as equivalent to self-esteem.
Items 1, 2, 5, and 6 were specified to load on the positive
self-regard factor, and items 3, 4, 7,* 8,* 9,* and 10* on the
usefulness/competence factor.
We also tested two models, each of which included a second-order factor. Second-order factors have been identified
in addition to first-order factors in confirmatory factor anal-
P98
RANZIJNETAL.
yses of a number of personality scales in recent years. The
second-order factor is usually considered to be superordinate to the first-order factors—in other words, to represent
a construct common to the first-order factors. For this reason, models with second-order factors are often described
as hierarchical. For instance, Liang and Bollen (1983)
found that Lawton's (1975) Philadelphia Geriatric Center
Morale Scale had a second-order "subjective well-being"
factor, and McCallum, Mackinnon, Simons, and Simons
(1995) found a scale factor "depression" in Radloff's
(1977) Center for Epidemiological Studies-Depression
Scale. However, a second-order factor could also be nonsuperordinate but simply the factor common to all the items
of the scale. It is reasonable to state that any particular scale
is designed primarily to measure one major construct, and
so all the items of the scale should be related to this construct. If this were not the case, researchers would use two
or more different scales rather than one. Therefore, it may
be the case that scales with at least two first-order factors
contain a general factor shared by all the items in addition
to more specific factors each related only to some items.
The last two models tested were similar to models 2 and
3, with the addition of the second-order factor. The hypotheses therefore were:
Hypothesis 4. The RSE-B has a nested structure, with two
first-order factors (self-derogation and self-enhancement)
and one second-order factor (general self-esteem). The use
of the term nested rather than hierarchical will be explained
later. The factor structure to test Hypothesis 4 was labeled
model 4. It differed from model 2 in that all 10 items were
specified to load on the general self-esteem factor in addition
to their loading on the specificfirst-orderfactors.
Hypothesis 5. The RSE-B has a nested structure, with two
first-order factors (positive self-regard and usefulness/competence) and one second-order factor (general self-esteem).
The factor structure to test Hypothesis 5 was labeled model
5. It differed from model 3 in that all items were also specified to load onto the general self-esteem factor.
These models are structurally similar to one tested by
Gustaffson and Balke (1993), who suggested the term
nested because the statistical procedure to test a nested
model involves the extraction of the first-order factors from
the residuals produced by extracting the second-order factor
from the covariance matrix. However, whereas the firstorder factors of Gustaffson and Balke's study were specified to be orthogonal, in the present study they were modeled as correlated. This is because we theorized that the
first-order factors would be related, possibly in a causal
sense, to each other, with usefulness/competence influencing positive self-regard. A recent article by Mutran, Reitzes,
Bratton, and Fernandez (1997) concluded that competence
has a strong influence on self-esteem, which provides support for such a causal relationship among the factors in the
RSE-B. In spite of the difference between our model specification and that of Gustaffson and Balke (1993), we feel
that the term nested is appropriate for our models because
the analytical procedure used is basically the same.
In similar analyses previously reported, the term hierarchical has sometimes been used (e.g., Liang & Bollen,
1983; McCallum et al., 1995). In a true hierarchical structure the first-order factors are correlated with the secondorder factor and the individual items do not load directly
onto the second-order factor. A hierarchical structure was
rejected in this case for two reasons. First, the second-order
factor, general self-esteem, could not be superordinate to
positive self-regard because these two constructs are conceptually equivalent. Second, we thought that general selfesteem was common to all the scale items, and therefore
each item was specified to load onto the second-order factor
as well as one of the first-order factors.
In summary, five competing models were constructed for
the Bachman revision (1970) of the Self-Esteem Scale
(Rosenberg, 1965). Two of the models contained a factor,
usefulness/competence, that had not previously been associated with this scale. The first three models contained
single-order factors and the final two were nested models
that also contained a second-order factor. The data were derived from a longitudinal study of ageing, and the models
were validated with data collected 2 years after the first
wave of the survey. It was expected that model 5 would
provide the best fit to the data.
METHOD
Participants and Procedure
Data for this study were provided by 1,087 elderly people (mean age 77.43, range 70-103 years) who comprised a
subsample of the Australian Longitudinal Study of Ageing
(ALSA; Centre for Ageing Studies, 1992). The ALSA is a
multidimensional, multidisciplinary research project that
aims to increase understanding of how social, biomedical,
behavioral, economic, and environmental factors are related
to age-related changes in the health and well-being of elderly Australians. From late 1992 to early 1993, comprehensive baseline interviews, covering a broad range of demographic, health, social, and psychological issues, were
conducted with 2,087 elderly adults living in the Adelaide
metropolitan area. The ALSA also included a series of selfcomplete questionnaires, filled in by the participants after
the interviews, which included the self-esteem scale. Longitudinal follow-up occurred 2 years later. Details of the
ALSA have been published elsewhere (Centre for Ageing
Studies, 1992; Clark & Bond, 1995; Mawby, Clark, Kalucy,
Hobbin, & Andrews, 1996; Ranzijn & Luszcz, 1994).
The target sample for the ALSA (N = 3,623) was randomly generated from the South Australian Electoral Roll
and was stratified by sex and 5-year age cohorts from 70 up
to the age of 85 years and over. Older males were oversampled, as were the oldest groups, in an attempt to obtain reasonable numbers of each of these in subsequent waves of
the study. It could be argued that the stratified nature of the
sampling plan should have been taken into account so as to
minimize biases in subsequent analyses. However, it has
been found by Mawby et al. (1996), who used the same
sample of 1,799 people, that there were no significant differences in income, ADLS, IADLS, hospitalizations or selfrated health between analyses using raw data and those
weighted to adjust for stratification. To test for possible biases in self-esteem scores, both weighted and unweighted
USEFULNESS AND SELF-ESTEEM
P99
Table 2. Baseline Descriptive Statistics
for the Full Sample and Self-Esteem Subsamples
data were used to analyze the differences in the mean
scores for the four age groups and for men and women (see
Results). The values of mean scores for each cell were virtually identical. However, the very small differences between age groups became significant because of the very
large number of cases estimated through the weighting procedure. Therefore, unweighted data were used for the factor
analyses and subsequent analyses.
Of the 3,263 people approached for the first wave of the
ALSA, 558 were ineligible for reasons such as death, incorrect age or address, being outside the geographic region,
and language difficulties. Of the 2,705 in the eligible sample, 1,477 (54.6%) participated. Of the nonrespondents, 780
(28.8%) refused outright, the 2-hour length of the interview
and the long-term involvement sought by the ALSA designers being the major deterrents to participation. The other
major reason for nonparticipation was illness of self or
spouse (16.2%). Spouses older than 65 and other household
residents over the age of 70 were also invited to participate,
resulting in another 610 participants and hence a total
ALSA sample of 2,087. This number included people living in institutional care as well as in the community. However, we were interested only in community-dwelling people aged 70 years or more (N = 1,799). Full details of the
sampling frame and response statistics have been published
by the Centre for Ageing Studies (1992). Of the first-wave
pool of 2,087 people, 241 were deceased and 42 had moved
or could not be contacted at the 2-year follow-up, leaving
1,804 people. One hundred twenty-five refused to participate again, resulting in a retention rate of 93% (N= 1,679).
The self-esteem data used in this report were obtained
from the optional self-complete questionnaire. Consequently,
fewer people answered items from this questionnaire (N =
1,297 at the first wave and 957 at the 2-year follow-up) compared to those who took part in the interview (N = 1,799 and
1,331, respectively). There were 1,087 people (84% of those
attempting the self-complete questionnaires) who fully completed all the self-esteem items at the baseline survey of the
ALSA and 875 at the follow-up (91%), and they comprised
the subsamples for the analyses reported here.
Descriptive statistics from the baseline data collection on
demographic data and health and function for the full sample, for the sample that completed the full self-esteem
scale, and for those who completed only some optional
questionnaire items are shown in Table 2. In addition to the
variables shown, there were no differences in the proportions of male respondents, the percentages being 54.2%,
54.9%, and 50.5% respectively.
Table 2 shows there were few differences between the
groups. The full self-esteem group was younger, had higher
income, more education, fewer ADLs and IADLs, better
self-rated health, less depression, and better cognitive function than the other two groups. The differences between the
full sample and full self-esteem samples were less than
those between the full self-esteem and partial optional
questionnaire samples. However, the differences in every
case were small.
(Rosenberg, 1965) used in this research comprised the 10
items listed in Table 1. Participants were asked to report
how often each statement was true for them. The response
categories were "almost always true" (1), "often true" (2),
"sometimes true" (3), "not often true" (4), and "never true"
(5). Items 1, 2, 3, 5, 7, and 9 were reverse-scored in the process of data entry. Scores were summed and the total score
divided by 10, and subscale scores divided by their number
of items, to give a possible range from 1 to 5. Higher scores
indicate higher self-esteem. A reliability analysis of the full
scale showed that the RSE-B had good reliability, Cronbach's a = 80.
The following variables were assessed to confirm the independence of the first-order factors, using the method of
Carmines and Zeller (1979). Depression was measured by
the Center for Epidemiological Studies-Depression Scale
(Radloff, 1977), morale was measured by 15 items from the
Philadelphia Geriatric Center Morale Scale (Lawton, 1975),
and perceived control by items from the Expectancy of
Control subscale of the Desired Control Measure (Reid &
Zeigler, 1981). Further details of the adapted scales used in
the ALSA can be found in Luszcz (1996). It was found during data analysis that depression and morale had skewed
distributions, so transformed scores were used in subsequent analyses. For depression, the scores underwent a logarithmic transformation, and for morale reflected and
square rooted scores were used. Finally, two subscales of
the Adelaide Activities Profile (Clark & Bond, 1995),
namely, the Domestic Chores and Home Maintenance subscales, were used to measure functional ability to participate in everyday activities.
Materials
The Bachman revision (1970) of the Self-Esteem Scale
The participants completed the interview and other scales
in their own homes. The initial data preparation was performed with SPSS-X (SPSS/Norusis, 1993) and the confir-
Full
Self-Esteem
Full Sample
Variable
Age (years)
Income
Education
Hospital
ADLs
IADLs
S/R Health
Depression
Minimental
N
M(S.D.)
N
M(S.D.)
1,799 78.35 (5.89) ,087 77.43 (5.52)
,686 2.96(1.04) ,044 3.05(1.05)
,783 3.70(1.38) ,087 3.82(1.37)
1,799 0.38 (0.90) 1,087 0.34 (0.79)
,791 0.34 (0.93) ,087 0.28 (0.84)
,790 0.89(1.68) ,086 0.74(1.46)
,795 2.92(1.11) ,086 2.83(1.10)
,717 7.95 (7.09) ,059 7.48 (6.82)
,767 18.89(2.78) ,079 19.33(2.18)
Some
Self-Esteem
N
M (S.D.)
210 80.81 (6.26)
195 2.71 (0.87)
210 3.48(1.29)
210 0.48(1.21)
210 0.48(1.15)
210 1.14(1.83)
210 3.06(1.09)
200 8.43 (7.40)
206 18.36(2.70)
Note: For income, a score of 2 represents an annual income range of
$5,001-$12,000, whereas 3 is equivalent to $12,001-$20,000 Australian
dollars. For education, a score of 3 means the participant left school at 14
years, a score of 4 means 15 years. Hospital = number of times in hospital
in the previous 12 months. ADLs and IADLs are measures of everyday
function: ADLs = number of Activities of Daily Living problems, IADLs
= number of Instrumental Activities of Daily Living problems (Duke University, 1978). Self-rated health is the score on a single-item rating ranging from 1 (excellent) to 5 (poor). Depression = score on CES-D
(Radloff, 1977); a score of 8 indicates low depression. Minimental =
score on items from the Mini-Mental State Exam (Folstein et al., 1985); a
score of 18 indicates good cognitive function.
RANZIJNETAL.
P100
matory factor analysis using LISREL for Windows (Joreskog
& Sorbom, 1993).
Analysis
PRELIS2, the preprocessor for LISREL8 (Joreskog &
Sorbom, 1993), was used to generate the matrices for subsequent analysis by LISREL8. All RSE-B items were declared to be ordinal, because there were only a small number of response categories and the intervals between
adjacent categories could not be assumed to be equal. Accordingly, the correlation matrix generated consisted of
polychoric correlations (Joreskog & Sorbom, 1993). The
evaluation of goodness of fit was performed by LISREL8
with the generally weighted least-squares method, using the
asymptotic covariance matrix.
RESULTS
The mean total self-esteem score was 4.14 (SD .55). Individual item means ranged from 3.74 (item 4) to 4.50
(item 6), with standard deviations between 0.73 (item 2)
and 1.20 (item 4). There were statistically significant differences between different age groups, F(3,1083) = 4.73, p <
.01. Post-hoc Scheffe analyses showed that people aged 85
years or more had a lower mean self-esteem score (4.01,
SD .57) than the 70-74 (M 4.19, SD .52) and 75-79 year
age groups (M 4.16, SD .56), but the differences were very
small. There were no significant differences in the mean
scores of men compared to women. As mentioned earlier,
analyses using weighted data produced results virtually
identical to those using unweighted data (the equivalent
mean scores to those mentioned above were 4.01, 4.19, and
4.18) so unweighted data have been used for all subsequent
analyses.
The polychoric correlation matrix is shown in Table 3.
Confirmatory Factor Analyses
In CFA, the assessment of the fit of a model is based on
the relative importance attached to the various indices produced by the program used, a debate that is at present unresolved (for some points of view, see Bentler, 1990; Bollen,
1990; Joreskog & Sorbom, 1993; McDonald & Marsh,
1990; Marsh, 1996; Wade et al., 1996). Some authorities
consider that the best indicators are the root-mean-square
error of approximation (RMSEA), which should be below
.05 and as low as possible, and the probability that the
RMSEA is less than .05, this probability being as close as
Table 3. Polychoric Correlation Matrix of RSE-B
Item 2
Item 3
Item 4
Item 5
Item 6
Item 7
Item 8
Item 9
Item 10
Item
1
Item
2
Item Item Item Item Item
3
4
5
6
7
8
.74
.54
.25
.49
.33
.45
.26
.37
.29
.51
.23
.51
.30
.56
.24
.37
.27
.19
.45
.21
.52
.31
.48
.29
.25
.48
.23
.37
.21
.42
.38
.46
.32
.41
.40
.31
.56
.30
.55
.30
.56
.35
Item
9
.32
.54
Item
.29
possible to 1.00 (Gustaffson & Stahl, 1996). A lower
RMSEA and a higher p value indicate that one model is superior to another. One model is also considered better than
another if its x2 value is lower, the probability of this x2
value is higher, the x2 df ratio is lower, the root-meansquare residuals index (RMR) is lower, and the goodnessof-fit indices are higher, preferably close to 1.00 (Joreskog
& Sorbom, 1993), than the values of the other model. However, the final determination of best fit should be theoretically based as well as statistically driven (Bollen, 1990;
Breckler, 1990; Browne & Cudeck, 1989; Hertzog, 1990).
In the analyses reported here, no modifications were performed on the models in the interests of achieving a better
fit. This decision was taken because it was thought more
important to maintain theoretical consistency than to use
the results of model testing to adjust the models in order to
improve the fit, the latter being a statistically driven rather
than theoretically driven approach. The results of the confirmatory factor analyses are summarized in Table 4.
Table 4 shows that the only models that are well fitting
according to the RMSEA criteria (RMSEA < .05, and p of
RMSEA > .05) are models 4 and 5. Model 2 (the first twofactor model) is superior to model 1 (the unidimensional
model) in every respect except PGFI (parsimony goodness
of fit), where the results are the same. Model 2 is also superior to model 3 (the competing two-factor model). Model 4
(the first nested model) is superior to model 2 in every respect except PGFI, and the fit is excellent according to all
the criteria. The PGFI in model 4 is lower than in models 2
and 3 because of its increased complexity, having a secondorder factor, but is still acceptable. The reduction in x2 for
model 4 compared to model 2 is significant, x 2 (l) = 8.77, p
< .001 (two-tailed). Model 5 has a higher x2 value, df ratio,
and RMSEA than model 4, and is equal on the other criteria. The reduction in x2 from model 3 to model 5 is significant, X2(l) = 16.04, p = < .001 (two-tailed).
Using these indices alone, model 4 seems superior to
model 5. However, some of the factor loadings for model 4
were not meaningful, whereas all loadings for model 5
were meaningful (see Table 5). The determination of the
fit of a model should, in addition to the other criteria,
be based on the meaningfulness of the factor loadings
Table 4. Results of Model Testing of RSE-B
Model
Testing
1
2
3
4
5
x2
df
x2/df
RMSEA (p) RMR
GFI
AGFI PGFI
with Wave 1 ALSA data
274.63 35 7.85
.079 (~0)
143.41 34 4.22
.054 (.20)
230.82 34 6.79
.073 (~0)
55.72 24 2.32
.035 (.98)
70.38 24 2.93
.042 (.86)
.14
.07
.13
.037
.037
.96
.98
.97
.99
.99
.94
.97
.95
.98
.98
.61
.61
.60
.43
.43
Validation with Wave 3 ALSA data
4
66.22 24 2.76
.045 (.73)
5
63.01 24 2.62
.043 (.79)
.050
.051
.99
.99
.98
.98
.43
.43
Note: RMSEA = root-mean-square error of approximation; RMR =
root-mean-square of residuals; GFI = Goodness-of-Fit Index; AGFI = Adjusted Goodness of Fit (goodness of fit adjusted to take into account the
degrees of freedom); PGFI = Parsimony Goodness of Fit (goodness of fit
adjusted to take into account the complexity of the model).
P101
USEFULNESS AND SELF-ESTEEM
(Joreskog & Sorbom, 1993). On balance, model 5 was
thought to be the best-fitting model. In most cases the firstorder loadings were high, and in all cases they were significant. The sum of the first-order factor variances exceeded the variance of the second-order factor. In other
words, the contribution of the first-order factors to the factor structure was greater than that of the second-order factor. This means that the first-order factors are clearly identified by the analysis, given that the second-order factor
has already been extracted. The phi coefficient (the correlation between the first-order factors) for model 5 was .84
(p < .001).
Cross- Validation
The factor structures of models 4 and 5, the models that
provided the best fits to the baseline ALSA data, were validated with the follow-up data obtained 2 years later. The results of cross-validation are shown in Table 4. Model 5 was
superior on all the criteria except RMR, where model 4 was
marginally superior. A number of the factor loadings for
model 4 were again unstable, whereas all the loadings for
model 5 were meaningful (see Table 5). It was concluded
from the analyses of both data sets that model 5 provided
the best fit to the data.
Table 5. Factor Loadings for Models 4 and 5: \ Weights
Model 4
Item
First-Wave Subsample
Third-Wave Subsample
SE
SE
SD
GSE
-.03
.02
.84***
.62***
.57***
59***
.73***
.54***
.63***
.45***
10
.65***
Variance ( I \ 2 ) 2.71
1.68
.35*
.14
34*
.25*
.53***
.36***
.56***
.35***
1.17
SD
.87***
.86***
.73***
.63***
74***
.%!***
.72***
.72***
.69***
.75***
3.57
2.23
GSE
.30
.18
-.14
-.01
_08
.08
-.36
.02
-.39
-.11
.45
Analyses Using Subscales
The positive self-regard subscale was computed by using
scores on items 1, 2, 5, and 6, and the usefulness/competence subscale consisted of items 3, 4, 7, 8, 9, and 10.
Higher scores indicate more positive self-regard and a
stronger sense of usefulness/competence, respectively. Subscale scores were divided by their number of items, to give
a possible range from 1 to 5. This was done to enable comparisons to be made with the full scale scores. The reliability of the positive self-regard subscale was 0.67, and that of
the usefulness/competence subscale was 0.69 (Cronbach's
a), values that are quite adequate for scales with only a few
items (Nunnally, 1978). The subscale score for positive
self-regard was 4.34 (SD 0.57), and that for usefulness/
competence was 4.09 (0.59), mean scores being high in
both cases (range 1-5).
As a final test of the construct validity of the first-order
factors, the subscales were correlated with three variables
that have been associated with self-esteem, namely depression, morale, and personal control (Andrews & Robinson,
1991), and also with functional ability. The resulting correlations of the full self-esteem scale with these variables
were all significant, p < .001 (see Table 6).
If two factors are truly distinct, it would be expected that
they would be differentially related to relevant variables
(Carmines & Zeller, 1979). Following the procedure used
by Carmines and Zeller in their factor analysis of the SelfEsteem Scale (Rosenberg, 1965), correlations were computed between each subscale and the variables mentioned
above. These correlations are also shown in Table 6. Analyses of the differences between the correlations, using the
method for determining the significance of the difference
between dependent correlations described by Cohen and
Cohen (1983), showed there were no differences on depression and control for the first-wave subsample or on domestic chores for the third-wave subsample, but there were differences between all the other correlations. Because some
of the correlations were noticeably larger in the third-wave
Table 6. Matrix of Pearson Correlations
of Full Self-Esteem Scale and Subscales with Weil-Being,
Control, and Functional Ability Variables
Model 5
First-Wave Subsample
PSR
Item
U/C
.85***
.85***
2
3
4
5
6
7
8
9
10
.76***
.23*
.62***
.30*
.71
.29**
.63***
.33**
2
Variance (IX ) 1.91
1.62
First-Wave Subsample
Third-Wave Subsample
GSE
PSR
.12
.09
.08
.55***
.29**
74***
.20
.65***
.22
.66***
.85***
.84***
1.91
U/C
.74***
.19*
.66***
.26*
.76***
.21 *
.77***
.26**
1.93
1.86
GSE
.27*
.26*
.23*
59***
37***
Variable
Dep
Mor
Cont
Self-esteem
Positive self-regard
Usefulness/competence
-.25
-.20
-.25
.39
.30
.38
.53
.46
.48
Difference
Self-esteem
Positive self-regard
Usefulness/competence
.14
71***
Difference
Note: SE = self-enhancement, SD = self-derogation, PSR = positive
self-regard, U/C = usefulness/competence, GSE = general self-esteem.
*p< .05; **/>< .01; ***/>< .001.
ns
HM
.14
.09**
.15
.20
.11
.22
p < .001 p < .001
Third Wave Subsample
.86*
.23*
.71***
2.49
ns
DC
-.46
-.37
-.44
.46
.34
.46
.59
.50
.56
p<.05
.12
.08*
16
.23
.16
.25
p < .001 p<.00\
Note: Dep = depression, Mor = morale, Cont = personal control, DC =
domestic chores subscale, and HM = household maintenance subscale of
Adelaide Activities Profile; Difference = difference between correlations
with subscales.
*p < .05; **p < .01; all other correlations are significant, /? < .001.
P102
RANZIJNETAL.
subsample than in the first-wave subsample, the first-wave
correlations were repeated using only those people who had
full self-esteem data at both waves of the study (N = 875).
However, the results were virtually identical, the values of
the correlations differing only between 0 and .02.
DISCUSSION
This study has shown that the Bachman revision (1970)
of the Rosenberg Self-Esteem Scale (1965) (RSE-B) is a
multidimensional scale with two first-order factors, labeled
positive self-regard and usefulness/competence, and an orthogonal second-order factor, general self-esteem.
The original Self-Esteem Scale of Rosenberg (1965) was
conceptualized as unidimensional, and this has been verified
recently by Marsh (1996). However, the Bachman revision
(1970) is a different scale, with four items not in the original
scale, all construed as referring to self-perceived usefulness
or competence. In this study, the unidimensional model
(model 1) was the worst fitting of all five models tested.
The model based on the division of the scale into selfderogation and self-enhancement (model 2) provided a better fit to the data than the unidimensional model, but not as
good as later models. If one model demonstrates a better fit
than another, the superior model is the only one that can be
accepted as valid. We therefore agree with Carmines and
Zeller (1979) and Marsh (1996) that self-derogation and
self-enhancement are not separate factors.
The fit of the third model, with two first-order factors,
positive self-regard and usefulness/competence, was about
the same as model 1 but not as good as either model 2, 4, or
5, and was therefore rejected. The fourth model tested, a
nested model with a second-order general self-esteem factor
in addition to the first-order factors self-derogation and selfenhancement, provided a better fit than the previous three
models, but not as good as model 5, so it too was rejected.
Furthermore, model 4 was also unacceptable because the
factor loadings were not meaningful. Cross-validation verified that model 5 was superior to model 4.
Model 5 was a nested model, with positive self-regard
and usefulness/competence as first-order factors and a second-order factor, general self-esteem. The finding that a
nested model with a second-order factor provides the best
fit is consistent with previous studies of other scales in
which a general factor has been found (e.g., Liang &
Bollen, 1983; McCallum et al., 1995). We conclude that the
RSE-B contains a factor common to all items as well as
their specific first-order factors. It could generally be argued that all scales that are not unidimensional should contain such a common core, and therefore a nested model is
the most appropriate one to test with CFA.
The people in this study had high scores on both the positive self-regard and usefulness/competence subscales. It
seems that perceived usefulness is not necessarily a function of economic productivity, since the people in this study
had been retired for an average of 12 years. It may be misguided to link usefulness with economic productivity when
discussing elderly people. On the other hand, the respondents may have adjusted psychologically to any negative effects of retirement (see Lazarus & DeLongis, 1983, for a
discussion of adjustment mechanisms in the elderly). These
high scores, however, may not accurately reflect the wider
population of elderly Adelaide people, given that the response rate was not as high as would have been wished.
Nevertheless, outright refusals comprised just over a quarter of those people initially approached, and this was likely
to have been due to the long-term time commitment requested and the length of the interview and other aspects of
the questionnaire. Therefore, we concluded that the data are
a reasonable reflection of the population from which the
sample was drawn.
The usefulness/competence factor has not previously
been identified in studies of self-esteem. However, the concepts of usefulness, competence, achievement, and productive activity have lately received increasing attention in the
gerontological literature. For example, Stevens (1993)
found that involvement with others was significantly related
to feeling useful. Mclntosh and Danigelis (1995) found an
association between informal volunteering (such as performing chores for friends) and well-being in adults over
the age of 60 years. Furthermore, Mutran et al. (1997)
found a strong association between competence, measured
by a single item of self-assessment, and self-esteem. Research on younger adults has also shown a relationship of
self-esteem with competence (Feather, 1991a, 1991b) and
sense of structure and purpose in the use of time (Bond &
Feather, 1988; Feather & Bond, 1983), which may be related to usefulness. The relationships among usefulness,
competence, purpose, and self-esteem in older people deserves further exploration. The possible utility of the usefulness/competence subscale for future research is indicated
by its reliability, which is quite high for a scale with so few
items. The scale could be further developed by the addition
of other relevant items reflecting different aspects of usefulness, competence, and achievement.
As a further test of the robustness of the factor structure,
the differences in the correlations of the factors with other
variables relevant to self-esteem were tested for significance. These variables were depression, morale, personal
control, and functional ability. Functional ability was assessed by two subscales, measuring the extent to which
people carried out normal domestic chores and home maintenance. The findings indicated that the factors were conceptually distinct. Usefulness/competence had a stronger
relationship to the indicator variables than did positive selfregard. The possible causal pathways here also deserve further investigation. Usefulness/competence may be a good
predictor of both well-being and personal control. More
work could also be done on the association between usefulness/competence and functional ability. There could be a
causal pathway from functional ability to well-being via
usefulness/competence and/or positive self-regard.
One puzzling finding was that some of the correlations referred to in the previous paragraph were substantially higher
in the third-wave subsample than in the first wave. One possible explanation was that the subsamples were not the same,
since the first wave contained over 200 people who were not
in the third wave. However, even when the correlations were
recalculated for only those first wave respondents who were
also in the third wave, the results were virtually identical to
those using the full subsample. Two details should be noted.
USEFULNESS AND SELF-ESTEEM
The trends are the same for both waves, and the relationships
with functional activity are more stable than those with wellbeing and control. Perhaps the strengthened correlations had
something to do with the respondents being 2 years older and
more familiar with the inventories. It is not clear why the factors appear to become more differentiated. Overall, the validation analyses strengthen support for the conceptual distinctness of the two factors.
In summary, this study has demonstrated that the Bachman revision (1970) of the Self-Esteem Scale (Rosenberg,
1965) has a nested structure, with a second-order scale factor in addition to two first-order factors, usefulness/competence and positive self-regard. The newly identified usefulness/competence subscale, a six-item scale with adequate
reliability, could be further developed and gainfully employed in future research on elderly people. Little work has
been done so far into the importance of feeling useful and
competent to the well-being of elderly people. This study
supports the view that feeling useful and competent may be
related in important ways to the self-esteem of older adults
(Mcintosh & Danigelis, 1995; Mutran et al., 1997; Ryff,
1989). Preliminary results also indicate that usefulness/
competence may be an important predictor of well-being.
Further research is required to clarify the relationships
among usefulness, competence, and self-esteem in elderly
people and the ways in which they affect well-being.
ACKNOWLEDGMENTS
, The authors gratefully acknowledge the work of the Centre for Ageing
Studies, Flinders University of South Australia, which, under the direction
of Prof. Gary Andrews, carried out the Australian Longitudinal Study of
Ageing (ALSA) that was the source of the data for this study. This research was funded in part by a grant from the U.S. National Institutes of
Health (Grant AG 08523-02) and by grants from the Sandoz Foundation
for Gerontological Research, the South Australian Health Commission,
and the Australian Rotary Health Research Fund. We are also grateful to
Drs. J. G. Bachman and P. M. O'Malley and two anonymous reviewers for
their helpful and detailed comments.
Address correspondence to Dr. Rob Ranzijn, School of Psychology, The
Flinders University of South Australia, GPO Box 2100, Adelaide 5001,
Australia. E-mail: [email protected]
REFERENCES
Andrews, F. M., &. Robinson, J. P. (1991). Measures of subjective wellbeing. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.),
Measures of personality and social psychological attitudes (Vol. 1, pp.
61-114). San Diego, CA: Academic Press.
Bachman, J. G. (1970). Youth in Transition II: The impact offamily background and intelligence on tenth-grade boys. Ann Arbor, MI: The Institute for Social Research.
Bachman, J. G., & O'Malley, P. M. (1977). Self-esteem in young men: A
longitudinal analysis of the impact of educational and occupational attainment. Journal of Personality and Social Psychology, 35, 365-380.
Bachman, J. G., & O'Malley, P. M. (1984). Black-white differences in
self-esteem: Are they affected by response styles? American Journal of
Sociology, 90(3), 624-639.
Bachman, J. G., & O'Malley, P. M. (1986). Self-concepts, self-esteem, and
educational experiences: The frog pond revisited (again). Journal of
Personality and Social Psychology, 50(1), 35-46.
Bachman, J. G., & Schulenberg, J. (1993). How part-time work intensity
relates to drug use, problem behavior, time use, and satisfaction among
high school seniors: Are these consequences or merely correlates? Developmental Psychology, 29(2), 220-235.
P103
Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107, 238-246.
Blascovich, J., & Tomaka, J. (1991). Measures of self-esteem. In J. P.
Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (Vol. 1, pp. 115-160). San
Diego, CA: Academic Press.
Bollen, K. A. (1990). Overall fit in covariance structure models: Two types
of sample size effects. Psychological Bulletin, 107, 256-259.
Bond, M. J., & Feather, N. T. (1988). Some correlates of structure and purpose in the use of time. Journal of Personality and Social Psychology,
55(2), 321-329.
Breckler, S. J. (1990). Applications of covariance stucture modelling in
psychology: Cause for concern? Psychological Bulletin, 107, 260-273.
Breytspraak, L. M., & George, L. K. (1979). Measurement of self-concept
and self-esteem in older people: State of the art. Experimental Aging
Research, 5, 137-148.
Browne, M. W., & Cudeck, R. (1989). Single sample cross-validation indices for covariance structures. Multivariate Behavioral Research, 24,
445^55.
Butler, R. N. (1985). Health, productivity and aging: An overview. In R. N.
Butler and H. P. Gleason (Eds.), Productive aging: Enhancing vitality
in later life. New York: Springer-Verlag.
Bynner, J. M., O'Malley, P. M., & Bachman, J. G. (1981). Self-esteem and
delinquency revisited. Journal of Youth and Adolescence, 70(6),
407-441.
Carmines, E. G., & Zeller, R. A. (1979). Reliability and validity assessment. Beverly Hills, CA: Sage.
Centre for Ageing Studies. (1992). The Australian Longitudinal Study of
Ageing.and a collaborative study of ageing in the US and Australia.
Adelaide, S.A.: Flinders University.
Clark, M., & Bond, M. (1995). The Adelaide Activities Profile: A measure
of the lifestyle activities of elderly people. Aging: Clinical and Experimental Research, 7, 174-184.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Duffy, M. E., & MacDonald, E. (1990). Derminants of functional health of
older persons. The Gerontologist, 30, 503-509.
Duke University Center for the Study of Aging and Human Development
(1978). Multi-dimensional functional assessment: The OARS methodology. Durham, NC: Duke University.
Feather, N. T. (1991a). Human values, global self-esteem, and belief in a
just world. Journal of Personality, 59, 84-106.
Feather, N. T. (1991b). Attitudes towards the high achiever: Effects of perceiver's own level of competence. Australian Journal of Psychology,
43, 121-124.
Feather, N. T, & Bond, M. J. (1983). Time structure and purposeful activity among employed and unemployed university graduates. Journal of
Occupational Psychology, 56, 241-254.
Folstein, M., Anthony, J. C , Parhad, I., Duffy, J., & Gruenberg, E. M.
(1985). The meaning of cognitive impairment in the elderly. Journal of
the American Geriatrics Society, 33, 228—235.
Gustaffson, J.-E., & Balke, G. (1993). General and specific abilities as predictors of school achievement. Multivariate Behavioral Research, 28,
407^34.
Gustaffson, J.-E., & Stahl, P. A. (1996). STREAMS User's guide. Version
1.6for Windows. Molndal, Sweden: Multivariate Ware.
Hertzog, C. (1990). On the utility of structural equation models for developmental research. In P. B. Baltes, D. L. Featherman, & R. M. Lerner
(Eds.), Lifespan development and behavior (Vol. 10, pp. 257-290).
Hillsdale, NJ: Erlbaum.
Joreskog, K. G., & Sorbom, D. (1993). LISREL 8: Structural equation
modelling with the SIMPLJS command language. Hillsdale, NJ: Scientific Software International/Erlbaum.
Krause, N. (1987). Life stress, social support and self-esteem in an elderly
population. Psychology and Aging, 2, 349-356.
Lawton, M. P. (1975). The Philadelphia Geriatric Center Morale Scale: A
revision. Journal of Gerontology, 30, 85-89.
Lazarus, R. S., & DeLongis, A. (1983). Psychological stress and coping in
aging. American Psychologist, 36, 245-254.
Liang, J., & Bollen, K. A. (1983). The structure of the Philadelphia Geriatric Center Morale Scale: A reinterpretation. Journal of Gerontology,
38, 181-189.
Luszcz, M. (1996). Beliefs about control in later life: Implications of per-
P104
RANZIJNETAL.
ceptions of health, memory, and global control. Hong Kong Journal of
Gerontology. V0(Suppl.), 502-506.
Marsh, H. W. (1996). Positive and negative global self-esteem: A substantively meaningful distinction or artifactors? Journal of Personality and
Social Psychology, 70(4). 810-819.
Mawby, L., Clark, M. S., Kalucy, E., Hobbin, E. R., & Andrews, G.
(1996). Determinants of formal service use in an aged population. Australian Journal on Ageing, 75(4), 177-181.
McCallum, J., Mackinnon, A., Simons, L., & Simons, J. (1995). Measurement properties of the Center for Epidemiological Studies Depression
Scale: An Australian community study of aged persons. Journals of
Gerontology, 50B(3). S182-S189.
McDonald, R. E, & Marsh, H. W. (1990). Choosing a multivariate model:
Noncentrality and goodness of fit. Psychological Bulletin, 107,
247-255.
Mclntosh, B. R., & Danigelis, N. L. (1995). Race, gender, and the relevance of productive activity for elders' affect. Journals of Gerontology,
50B(4), S229-S239.
Mutran, E. J., Reitzes, D. J., Bratton, K. A., & Fernandez, M. A. (1997).
Self-esteem and subjective responses to work among mature workers:
Similarities and differences by gender. Journals of Gerontology,
526(2), S89-S96.
Nunnally, J. (1978). Psychometric methods (2nd ed.). New York: McGrawHill.
O'Malley, O. M., & Bachman, J. G. (1979). Self-esteem and education:
Sex and cohort differences among high school seniors. Journal of Personality and Social Psychology, 37, 1153-1159.
O'Malley, O. M., & Bachman, J. G. (1983). Self-esteem: Change and stability between ages 13 and 23. Developmental Psychology, 19(2).
257-268.
Radloff, L. S. (1977).The CES-D scale: A self-report depression scale for
research in the general population. Applied Psychological Measurement, 3,385-401.
Ranzijn, R., & Luszcz, M. (1994). Well-being of elderly Australians: The
role of parent-adult child contacts. Australian Journal of Ageing,
13(4), 186-189.
Reid, D. W., & Ziegler, M. (1981). The Desired Control measure and adjustment among the elderly. In H. Lefcourt (Ed.), Research with the
locus of control construct (Vol. 1). New York: Academic Press.
Rosenberg, F. R., & Rosenberg, M. (1978). Self-esteem and delinquency.
Journal of Youth and Adolescence, 7(3). 279-291.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton,
NJ: Princeton University Press.
Rosenberg, M., Schooler, C , Schoenbach, C , & Rosenberg, F. (1995).
Global self-esteem and specific self-esteem: Different concepts, different outcomes. American Sociological Review, 60, 141-156.
Ryff, C. D. (1989). Beyond Ponce de Leon and life satisfaction: New directions in quest of successful aging. International Journal of Behavioral Development, 12, 35-55.
Shahani, C , Dipboye, R. L., & Phillips, A. P. (1990). Global self-esteem
as a correlate of work-related attitudes: A question of dimensionality.
Journal of Personality Assessment, 54(1 &2), 276-288.
SPSS, Inc./Norusis, M. J. (1993). SPSS for Windows Release 6. Chicago:
SPSS, Inc.
Stevens, E. S. (1993). Making sense of usefulness: An avenue toward satisfaction in later life. International Journal of Aging and Human Development, 37(4). 313-325.
Wade, T., Tiggeman, M., Heath, A. C , Abraham, S., Treloar, S. A., & Martin, N. G. (1996). Structure of disordered eating in a twin community
sample. International Journal of Eating Disorders, 79(1), 63-71.
Warr, P. B., & Jackson, P. R. (1983). Self-esteem and unemployment
among young workers. Le Travail Humain, 46, 355-366.
Wells, L. E., & Sweeney, P. D. (1986). A test of three models of bias in
self-assessment. Social Psychology Quarterly. 49(\), 1-10.
Received July 19, 1996
Accepted September 19, 1997
Nominations for Prestigious Awards
DEADLINE...MAY 8,1998
Robert W. Kleemeier
Award
To a Fellow of The Gerontological Society ofAmerica in
recognition of outstanding research in the field of gerontology.
Donald P. Kent
Award
To a Fellow of The Gerontological Society of America who best
exemplifies the highest standards for professional leadership in
gerontology through teaching, service, and interpretation of
gerontology to the larger society.
Glenn Foundation
Award
Open to all scientists, regardless of field or nationality, for
significant research contributions to the biology of aging.
Nathan Shock New
Investigator Award
For outstanding contributions to new knowledge about aging
through basic biological research.
PGC Polisher Research
Institute Award
Open to all disciplines, for a significant contribution from
applied research that has benefited older people and their care.
For further information about these and additional awards, contact:
The Gerontological Society of America
1275 K Street NW, Suite 350
Washington, DC 20005-4006
Phone 202/842-1275 • FAX 202/842-1150
http ://w ww.geron.org