Full text of DOI:10.1002/mpr.143

International Journal of Methods in Psychiatric Research, Volume 12, Number 2
65
A new rating scale for adult resilience: what
are the central protective resources behind
healthy adjustment?
ODDGEIR FRIBORG, ODIN HJEMDAL, JAN H. ROSENVINGE, MONICA MARTINUSSEN, Department
of Psychology, University of Tromsø, Norway
ABSTRACT: Resources that protect against the development of psychiatric disturbances are reported to be a significant
force behind healthy adjustment to life stresses, rather than the absence of risk factors. In this paper a new scale for
measuring the presence of protective resources that promote adult resilience is validated. The preliminary version of the
scale consisted of 45 items covering five dimensions: personal competence, social competence, family coherence, social
support and personal structure.
The Resilience Scale for Adults (RSA), the Sense of Coherence scale (SOC) and the Hopkins Symptom Checklist
(HSCL) were given to 59 patients once, and to 276 normal controls twice, separated by four months.
The factor structure was replicated. The respective dimensions had Cronbach’s alphas of 0.90, 0.83, 0.87, 0.83 and
0.67, and four-month test-retest correlations of 0.79, 0.84, 0.77, 0.69 and 0.74. Construct validity was supported by
positive correlations with SOC and negative correlations with HSCL. The RSA differentiated between patients and healthy
control subjects. Discriminant validity was indicated by differential positive correlations between RSA subscales and SOC.
The RSA-scale might be used as a valid and reliable measurement in health and clinical psychology to assess the presence
of protective factors important to regain and maintain mental health.
Key words: resilience, scale development, validation, sense of coherence, psychiatric symptoms
• family support and cohesion; and
• external support systems.
Individuals who sustain normal development despite
long-term stress, adversity or maltreatment, are
frequently labelled ‘resilient’ (Garmezy and
Nuechterlein, 1972; Garmezy, 1981; Rutter, 1985;
Steele, 1987; Zimrin, 1987; Cowen and Work, 1988;
Egeland, Carlson, Sroufe, 1993; Block and Kremen,
1996). Longitudinal studies conducted over four
decades, such as the Kauai study (Werner, 1989, 1993,
2001) and the Lundby study (Cederblad, 1996), point
to several key features characterizing resilient people
who overcome difficult life conditions. Generally, they
are more flexible than vulnerable people and cope by
using several protective resources either within
themselves or in their environment. Several authors
(Werner, 1989; Rutter, 1990; Werner, 1993; Garmezy,
1993) now classify these protective resources into
These are the most significant determinants of a
healthy adjustment to long-term stresses. The overall
aim of this study was to develop a valid scale for
measuring the presence of such protective resources
and to examine whether these resources differentiated
between patients and non-patients.
Most prospective studies report that resilient people
draw heavily on favourable dispositional attitudes and
behaviours like internal locus of control, pro-social
behaviour and empathy, to face life stresses. They have
a positive self-image and display great optimism for the
future (Werner and Smith, 1992; Cederblad, Dahlin,
Hagnell and Hansson, 1993; Cederblad, 1996; Blum,
1998). Moreover, resilient people seem to have a strong
ability to organize their life (Clausen, 1993). Such
• psychological/dispositional attributes;

Friborg et al.
66
personal dispositions and attitudes promote in return,
supportive relationships with family members and
friends. Conversely, the lack of social support that
many patients experience is related to their problem
with reciprocating social support from others
(Kringlen, 1990). Resilient individuals also manage to
distance themselves psychologically from the trouble
that mentally ill parents inflict (Watt, 1995).
Compared to more troubled individuals, they generally
value siblings as a more important source of emotional
support (Werner and Smith, 1992). They work harder
to resolve marital conflicts (Werner, 2001). Obviously,
resilience does not protect the individual from negative
life events but resilient individuals seem to cope more
functionally and flexible with stress. These character-
istics are developed early in life by the formation of a
secure attachment to other people, which may reduce
the vulnerability to developing psychiatric disorders
significantly (Svanberg, 1998). Gender differences in
resilience have been investigated less often, but one
consistent finding is that resilient women tend to elicit
and provide more social support (Werner, 2001). In
sum, resilience is considered a multi-dimensional
construct (Luthar, Doernberger and Zigler, 1993). The
concept not only refers to important psychological
skills or abilities but also to the individual’s ability to
use family, social and external support systems to cope
better with stress. Measurement scales for assessing
overall improvements in mental health should thus
include these factors.
study (Hjemdal, Friborg, Martinussen and Rosenvinge,
2001). The resulting scale consisted of 45 items
covering five dimensions labelled ‘personal compe-
tence’ (16 items), ‘social competence’ (12 items),
‘social support’ (nine items), ‘family coherence’ (five
items) and ‘personal structure’ (four items). This factor
solution was in accordance with the overall classifi-
cation of resilience (Werner, 1989; Rutter, 1990;
Werner, 1993; Garmezy, 1993). The internal consis-
tency (Cronbach’s alpha) was high for all dimensions,
ranging from 0.92 to 0.74 and for all the items
combined (total α = 0.93) as well.
As it was desirable to level out the number of items
along the dimensions, 21 new items were generated
under the following sub-dimensions: personal compe-
tence and social competence (one item, respectively),
family coherence (five items), and personal structure
(four items). The initial factor solution excluded the
generated locus-of-control items. As the construct is
considered highly important for a resilient outcome
(Werner, Smith, 1992; Cederblad, Dahlin, Hagnell,
Hansson, 1993), 10 new items pertaining to
internal/external control were generated for the present
study (for example, ‘if I succeed in school, it is because I
am competent’ or ‘success comes from hard work’).
The purposes of the present study were
• to test the original factor structure of the RSA after
adding new items, and to estimate new internal
consistencies;
Two scales to measure resilience have appeared in
the literature. Jew, Green and Kroger (1999)
developed a scale for children and adolescents from
the cognitive appraisal theory of Mrazek and Mrazek
(1987), which emphasizes 12 essential skills that are
important for coping adequately with life stress. The
other measure (Wagnhild and Young, 1990, 1993) was
developed from interviews with 24 elderly American
women who had successfully dealt with various losses
typical of old age. A follow-up study of the scale on
elderly people failed to validate the scale (Aroian,
Schappler-Morris, Neary, Spitzer and Tran, 1997).
However, none of the scales included measurement of
social factors (such as family/external support or social
competence) known to be essential to withstand life
stress. Age-specific features may also make these scales
inappropriate to measure resilience in the adult group.
Hence, a preliminary version of the Resilience
Scale for Adults (RSA) was developed in an earlier
• to study the test-retest reliability in a non-patient
sample; and
• to test important aspects of the instrument’s
construct validity.
All the subscales of the RSA were expected to correlate
significantly and positively with a measure of psycho-
logical/personal adjustment – the Sense of Coherence
scale (SOC) (Antonovsky, 1993, 1998). However, as
SOC assesses personal adjustment skills, the social and
family subscales of the RSA were expected to correlate
less positively with the SOC scale. Next, the RSA was
expected to correlate negatively with an inventory of
psychiatric symptoms (Hopkin’s Symptom Checklist,
HSCL) (Nettelbladt, Hansson, Stefansson, Borgquist
and Nordström, 1993).
A widespread method in establishing construct
validity is to assess differences among variables known
to differentiate the groups (Streiner and Norman,

A new rating scale for adult resilience
67
1995). As such, the RSA was expected to correlate
positively with level of education and indices of
occupational adaptation. It has been a consistent
finding that people with psychiatric problems report
less protective factors in their environment alleviating
stress and preventing maladjustment (Werner and
Smith, 1992). Thus, it was expected that patients
seeking psychiatric treatment in an outpatient clinic
would report less degrees of available protective
resources in comparison to the control group.
Finally, and most importantly, the study investigated
whether the patient sample (N = 59) was systematically
different, in terms of psychiatric diagnoses (Table 1),
from the patients that were contacted (N = 183), and
all patients that started in treatment during the year
2000 (N = 398). Due to four cells with expected cell
frequencies less than five, Fisher exact tests were run
instead of chi-square statistics. However, no significant
differences in the proportions of diagnoses emerged
between the samples. The number of patients with two
psychiatric diagnoses was not significantly different
between the samples.
Method
Normal controls were selected by Statistics Norway
at random (N = 977, 51% women and 49% men)
among inhabitants from 25–50 years (M = 37.0, SD =
7.3) living in Tromsø. In all, 162 women (M = 35.6
years of age, SD = 7.5) and 128 men (M = 37.1 years of
age, SD = 7.3) responded. The response rate was 31%.
There were no gender differences between the original
sample and those who participated (p > 0.05).
Although those who participated (M = 36.9, SD =
7.3) were slightly older than the original sample (M =
36.1, SD = 7.4), it was of no practical importance.
Four months later, 130 women (M = 36.0 years of age,
SD = 7.6) and 97 men (M = 37.4 years of age, SD =
7.1) returned the second set of the questionnaires. The
response rate was 79%.
Subjects
Patients were recruited from an adult outpatient clinic
in Tromsø, northern Norway. A total of 183 patients,
117 women (63.9%) and 66 men (36.1%), offered
psychotherapy for the first time, were contacted.
Altogether, 45 women between 18 and 75 years (M =
33.7, SD = 13.0) and 14 men between 19 and 75 years
(M = 36.2, SD = 13.0) responded. The response rate
was 32.2%. Several investigations were done to
examine whether the low response rate produced a
biased sample. The age differences between patients
that participated (M = 34.6, SD = 12.8) and those who
refused (M = 34.8, SD = 13.5), was insignificant.
However, the number of men who refused to participate
(41.9%) were significantly higher than the number who
did participate (23.7%) χ 2 (1, n = 183) = 5.75,
p < 0.05). As a t-test indicated no significant differences
in psychiatric symptoms between women and men,
however, this was considered of minor importance.
Procedure
Information was collected once in the patient sample
and twice in the control sample, with a four-month
follow-up. The random control sample was invited to
Table 1. The proportion of psychiatric diagnoses among patients that participated (N = 59), all patients that were
contacted (N = 183), and all patients who started in treatment at the outpatient clinic during the year 2000 (N = 398)
ICD-10 diagnoses
Patients
(N = 59)
%
Total Contacted1
(N = 183)
%
All year 20002
(N = 398)
%
Schizophrenia (F20, F22)
Bipolar disorders (F31)
1
2
1.7
3.4
4
6
2.2
3.3
22
13
5.5
3.3
Depressive disorders (F32-F34)
Anxiety disorders (F40-F42, F45)
PTSD/adjustment disorders (F43)
Eating/personality disorders (F50/F60)
Other diagnoses
24
6
14
1+1
10
40.7
10.2
23.7
3.4
67
24
35
7+6
34
36.6
13.1
19.1
7.1
105
67
111
8+14
58
26.4
16.8
27.9
5.5
16.9
18.6
14.6
1 A Fisher exact test found no significant differences in the proportions of diagnoses between the patient sample
and the drop-outs (Fisher (6, n = 183 ) = 6.40, p = 0.37).
2 The same analysis found no differences between the patient sample and the total number of patients during the
year 2000 (Fisher (6, n = 398 ) = 10.31, p = 0.13).

Friborg et al.
68
participate through postal invitations. The invitation
used the word ‘personal skills’ instead of ‘resilience’.
For the clinical sample, appointments for starting
psychotherapy and the invitations to participate in the
study were mailed at the same time but in separate
envelopes. After returning the completed question-
naires, the participants received a lottery ticket (value
$ 2.5).
Acquiescence bias is an unconscious tendency to agree
or disagree independent of the content in the items.
The Norwegian short-version of the MCSDS (α =
0.65) contains 10 items, of which five are reversed
(Rudmin, 1999). The social desirability index was
computed by reversing half of the items before calcu-
lating the mean. The acquiescence bias index was
computed by calculating the mean of the 10 items
without reversing any scores.
Measurements
Demographic data were age, gender, level of
education, current employment status, years of work
experience.
Removal of participants
To ensure the data quality, participants who met the
following criteria were deleted from the data pool:
Sense of Coherence Scale (SOC)
• more than 10% missing data;
The 29-item SOC self-report scale uses a seven-point
Likert scale, with positive and negative semantic
phrases at each endpoint. Higher scores reflect
stronger SOC. Thirteen of the items are reversely
scored to avoid response set bias (acquiescence). The
scale is used worldwide and is highly reliable
(Cronbach’s alpha ranging from 0.82 to 0.95) and
valid as a measure of overall mental health adjustment
(Antonovsky, 1993, 1998). The SOC correlates
negatively with perceived stress, trait anxiety and
current depression (Frenz, Carey and Jorgensen, 1993;
Sammallahti, Holi, Komulainen and Aalberg, 1996).
• non-serious responses;
• obvious misunderstanding of the questionnaires;
• a z-score > 2 on the social desirability index, a z-
score > 2 on the acquiescence index or a z-score
< 2 on the nay-saying-bias index.
In sum, 35 participants were deleted (six patients and
29 controls). This yielded a patient sample of 59
participants and a control sample of 276 participants.
Four months later, the control sample consisted of 230
participants.
Results
The Hopkins Symptom Check List-25 (HSCL)
The HSCL is a 25-item short version of the Symptom
Check List (SCL-90). It is a self-report inventory that
rates the presence of depression, anxiety and somati-
zation. It uses a four-point scale ranging from one (‘not
at all’) to four (‘very much’). Higher scores indicate
more psychiatric/affective symptoms. A mean score of
≥1.55 indicates a probable psychiatric problem,
whereas a score of ≥1.75 indicates a probable need for
psychiatric treatment (Nettelbladt, Hansson,
Stefansson, Borgquist and Nordström, 1993). The
scale has also proven highly reliable in Norwegian
samples (α = 0.91) (Lavik, Laake, Hauff and Solberg,
1998; Moum, 1998).
Replicative findings
Before subjecting the resilience items to factor
analyses, we examined whether they correlated highly
with indices for social desirability, acquiescence or
nay-saying bias. Such correlations were small (range
0.11 to 0.27).
It was decided that each dimension of the factor
structure should contain at least five items. An
exploratory principal components analysis with a
varimax rotation, instead of a confirmatory analysis,
was preferred to further reduce the number of items.
The factor analysis was based on the control sample
only (N = 276), as it would be theoretically contra-
dictory to include subjects with current psychological
problems for the selection of items. Factors with
Eigenvalues less than 1 were excluded. This procedure
generated 13 dimensions explaining 63% of the
variance. However, the component solution was not
acceptable as only two or three items loaded on the
Marlowe-Crowne Social Desirability Scale (MCSDS)
The MCSDS was applied to identify biased answers,
looking for the effects of social desirability and acqui-
escence. Social desirability is a conscious choice to
present oneself in an overly positive manner.

A new rating scale for adult resilience
Relationship with other tests
69
sixth factor and beyond. The scree-plot was then
examined for a clear bend in the curve, which emerged
at the fifth component. Inspecting the items in each
dimension, the five-component solution was the same
as in the previous study (Hjemdal, Friborg,
Martinussen and Rosenvinge, 2001). This accounted
for 41 percent of the variance, and excluded 33 items.
Six items with high side loadings (<0.30) on other
dimensions were also removed. One item from the
fifth factor with a high side loading was, however,
retained to keep five items in the dimension. The
factor solution and the factor loadings consisted of 37
items (Table 2). The five dimensions were labelled
‘personal competence’, ‘social competence’, ‘family
coherence’, ‘social support’ and ‘personal structure’.
Intercorrelations between the factors were low to
moderate (r = 0.22 to 0.46, p < 0.01), except for a
non-significant correlation between ‘social support’
and ‘personal structure’.
A widely used and accepted method for assessing
construct validity (Streiner and Norman, 1995) is to
examine associations with a convergent/similar scale
(such as SOC) and a discriminant/dissimilar scale
(such as HSCL). The correlations between the
subscales of the RSA and the SOC were all positive
(Table 5), ranging from 0.29 to 0.75 (all p <0.01). As
expected, the ‘personal competence’ subscale corre-
lated highest (r = 0.75, p < 0.01) with SOC, while the
other four subscales correlated less positively with the
SOC, ranging from r = 0.29 to r = .45 (p < 0.01). A
similar pattern was found between the subscales of the
RSA and the HSCL, with the highest negative corre-
lation between ‘personal competence’ and HSCL (r =
–0.61, p < 0.01). The other subscales correlated less
negatively with HSCL, ranging from –0.19 to –0.37
(all p < 0.01). The correlation between the SOC and
the HSCL was thus negative (r = –0.75, p <0.01).
The RSA’s construct was further examined against
three demographic variables expected to be associated
with higher resilience:
Reliability estimates (internal consistency and test-retest)
The internal consistency (Chronbach’s α) of all the
contrast scales (SOC, HSCL, MCSDS) was satisfactory
high (Table 3), indicating adequate psychometric
properties. The internal consistency of the subscales of
the RSA was satisfactory, ranging from 0.67 to 0.90.
The test-retest correlations were all satisfactory for the
subscales of RSA, ranging from 0.69 to 0.84 (p < 0.01).
Item-total correlations were calculated for every
subscale. The item-total correlations belonging to
‘personal competence’ ranged from 0.51 to 0.75, ‘social
competence’ from 0.48 to 0.74, ‘family coherence’
from 0.56 to 0.74, ‘social support’ from 0.43 to 0.70,
and ‘personal structure’ from 0.37 to 0.48.
• number of years of education;
• employment; and
• number of years in work.
Using alpha-level <0.01 to reduce type I error, years of
education were not associated with any of the RSA
subscales. Concerning employment status (0 = not
employed, 1 = employed), three of the subscales,
‘personal competence’, ‘social competence’ and ‘family
coherence’, were positively and significantly correlated
with holding a job (r = 0.18, 0.22 and 0.17, respec-
tively, all p < 0.01).
Gender and age effects
Number of years in work was positively and signifi-
cantly correlated with ‘family coherence’ and ‘personal
structure’ (r = 0.17 and 0.17, respectively, p < 0.01).
All the subscales of the RSA were examined for gender
and age differences (Table 4). Analyses were only done
on the control sample to reduce the number of subjects
with psychiatric symptoms in the group. The signifi-
cance-level was set to <0.01 to reduce type I error. A
t-test indicated that women reported significantly
higher levels of ‘social support’ than men (t(273) =
4.27, p < 0.001), whereas men reported sub-signifi-
cantly higher levels of ‘prsonal competence than
women (t(273) = –2.21, p = 0.03). Correlational
analyses on age differences, found ‘personal structure’
to be positively correlated with age (r = 0.17, p < 0.01).
Differences between the patient and the control sample
The study was examined whether the patient sample
reported lower degrees of resiliency than the control
sample. Again, alpha level was set to p < 0.01. All the
subscales of the RSA differentiated between the
samples (Table 6), finding higher degrees of resilience
in the normal control sample, except from ‘social
support’ that reached a sub-significant level (p < 0.04).
To ensure that the patient sample was compared to a

Friborg et al.
70
Table 2. The factor solution for the non-clinical sample showing the distribution and loading of items in the five
dimensions; 1 = personal competence, 2 = social competence, 3 = family coherence, 4 = social support, and
5 = personal structure (N = 276)
Dimensions:
Translated items from Norwegian
1
2
3
4
5
I believe in my own abilities
Believing in myself helps me to overcome difficult times
I know that I succeed if I carry on
0.77
0.69
0.65
0.64
0.61
0.60
0.58
0.58
0.55
0.54
0.47
0.44
I know how to reach my goals
No matter what happens I always find a solution
a I am comfortable together with other persons
My future feels promising
I know that I can solve my personal problems
I am pleased with myself
0.52
I have realistic plans for the future
I completely trust my judgements and decisions
a At hard times I know that better times will come
I am good at getting in touch with new people
I easily establish new friendships
0.37
0.87
0.75
0.66
0.65
0.64
0.59
0.55
0.54
0.51
0.39
It is easy for me to think of good conversational topics
a I easily adjust to new social milieus
0.30
0.36
It is easy for me to make other people laugh
I enjoy being with other people
a I know how to start a conversation
I easily laugh
It is important for me to be flexible in social circumstances
a I experience good relations with both women and men
There are strong bonds in my family
0.34
0.82
0.78
0.72
0.66
0.65
0.64
0.61
I enjoy being with my family
In our family we are loyal towards each other
In my family we enjoy finding common activities
Even at difficult times my family keeps a positive outlook on the future
In my family we have a common understanding of what’s important in life
There are few conflicts in my family
I have some close friends/family members who really care about me
I have some friends/family members who back me up
I always have someone who can help me when needed
I have some close friends/family members who are good at encouraging me
I am quickly notified if some family members get into a crisis
I can discuss personal matters with friends/family members
I have some close friends/family members who value my abilities
a I regularly keep in touch with my family
There are strong bonds between my friends
Rules and regular routines make my daily life easier
I keep up my daily routines even at difficult times
I prefer to plan my actions
0.82
0.76
0.67
0.63
0.59
0.53
0.47
0.46
0.39
0.46
0.69
0.67
0.64
0.58
0.46
I work best when I reach for a goal
I am good at organizing my time
0.34
21
Variance explained (%)
7
5
4
4
Note: Total variance explained (41%). Factor-loadings < 0.30 are omitted. The table shows a crude English
translation of the Norwegian original version.
a Items were removed due to side loading.

A new rating scale for adult resilience
71
Table 3. The reliability estimates for all scales (the RSA, SOC and HSCL) based on the control sample (N = 217)
Internal consistency
Test-retest (four months)
Dimensions
Cronbach’s α
r
1
2
3
4
5
Personal competence
0.90**
0.83**
0.87**
0.83**
0.67**
0.94**
0.92**
0.67**
0.79**
0.84**
0.77**
0.69**
0.74**
0.86**
0.77**
0.47**
Social competence
Family coherence
Social support
Personal structure
Total SOC
Total HSCL
MCSDS
Note. ** p < 0.01 (two-tailed).
Table 4. The differences in mean and standard deviations between women and men on resilience subscale
scores, based on the control sample (N = 273); the effect sizes indicate the magnitude of the differences.
Women (N = 153)
Men (N = 120)
Dimensions
No. items
M
SD
M
SD
t
Hedge’s
d
1
2
3
4
5
Personal competence
10
7
7
8
5
5.28
5.19
5.13
5.91
5.11
0.96
1.02
1.26
0.88
0.97
5.51
5.14
5.13
5.45
5.10
0.77
1.01
0.98
0.86
0.91
–2.21*
0.47
0.05
4.27***
0.09
–0.29
0.05
0.00
0.53
0.01
Social competence
Family coherence
Social support
Personal structure
Note. * p < 0.05 (two-tailed), *** p < 0.001 (two-tailed).
Table 5. The correlations between the subscales of the RSA, the SOC and the HSCL Scales (N = 335)
Scales
SOC
HSCL
1
2
3
4
5
Personal competence
0.75**
0.44**
0.45**
0.29**
0.33**
–0.61**
–0.32**
–0.37**
–0.19**
–0.21**
Social competence
Family coherence
Social support
Personal structure
Note. ** p < 0.01 (two-tailed).
psychological healthy sample, all control subjects with
symptom-scores above 1.55 were removed. New
comparisons between the patient and the healthy
sample confirmed the results by indicating somewhat
greater differences. The effect size of the differences
between the samples was greatest for ‘personal compe-
tence’ (Hedge’s d = 1.59) and ‘family coherence’
(Hedge’s d = 1.06).
of psychiatric problems (HSCL). The mean symptom-
score in the patient sample (2.24) was far above the
cut-off value (>1.75), which indicates a probable
treatment need (Nettelbladt et al., 1993). In contrast,
the symptom score in the control sample (1.42) was
lower than the cut-off value (<1.55) indicating a
probable psychiatric problem (Nettelbladt et al., 1993).
Finally, the proportion of subjects in the control
sample who reported experiencing critical life events
at the second data collection, was calculated. Out of
Control subjects also reported significantly higher
degrees of sense of coherence (SOC) and less symptoms

Friborg et al.
72
Table 6. The differences in mean and standard deviations between the patient (n = 59) and the control sample
(N = 276), the patient and the psychological healthy sample (n = 199); the effect sizes indicate the magnitude of the
differences
Scales
Patient sample
(N = 59)
Control sample
(N = 276)
Healthy Sample1
(N = 199)
M
SD
M
SD
t, Hedge’s d
M
SD
t, Hedge’s d
Personal competence
Social competence
Family coherence
Social support
Personal structure
SOC
4.11
4.67
4.22
5.41
4.58
3.82
2.25
1.30
1.07
1.42
1.02
0.99
0.86
0.57
5.39
5.17
5.13
5.71
5.10
4.85
1.42
0.88
1.01
1.14
0.90
0.95
0.75
0.37
7.08***, 1.33
3.30**, 0.49
4.45***, 0.76
2.12*, 0.33
3.38***, 0.54
8.56***, 1.34
-10.55***, 2.01
5.56
5.25
5.37
5.80
5.14
5.10
1.24
0.76
1.00
0.97
0.83
0.93
0.63
0.14
8.07***, 1.59
4.02***, 0.57
5.64***, 1.06
2.79**, 0.44
3.84***, 0.59
10.50***, 1.86
-13.32***, 3.37
HSCL
Note. * p < 0.05, ** p < 0.01, *** p < 0.001 (two-tailed).
1 The psychologically healthy sample was derived by removing control subjects with a mean symptom-score (HSCL)
above 1.55.
230 subjects, 32 (13.9%) confirmed that they had
been experiencing stressors that considerably influ-
enced the ability to cope with everyday demands.
The second category, ‘family cohesion/warmth’, was
comprised by the dimension ‘family coherence’ that
measured amount of family conflict, cooperation,
support, loyalty and stability.
Discussion
The third and last category ‘external support
systems’ was comprised of the dimension ‘social
support’ that measured access to external support from
friends and relatives, intimacy, and the individual’s
ability to provide support.
In summary, the five-dimensional scale corresponds
well with the overall categorization of resilience,
recapitulated as characterized by (i) personal/disposi-
tional attributes, (ii) family support and (iii) external
support systems (Werner, 1989; Rutter, 1990; Werner,
1993; Garmezy, 1993).
The factor structure and the reliability from the
original RSA-study (Hjemdal, Friborg, Martinussen
and Rosenvinge, 2001) were successfully replicated.
Due to the low-to-moderate intercorrelations between
the subscales of the RSA, the dimensions should be
regarded as subscales measuring different, but all
various and positive aspects of the concept of
resilience. This supports the theoretical understanding
of resilience as a multidimensional phenomenon
(Cicchetti and Garmezy, 1993; Garmezy, 1993; Luthar,
Doernberger and Zigler, 1993). Along with the satis-
factory test-retest reliability, the RSA seems to be
satisfactorily operationalized. In contrast to the
existing resilience scales (Wagnhild and Young, 1990,
1993; Jew, Green and Kroger, 1999), the RSA covers
all three of the main categories of resilience.
The first category of ‘dispositional attributes’ was
comprised by the three dimensions ‘personal compe-
tence’, ‘social competence’ and ‘personal structure’.
‘Personal competence’ measured level of self-esteem,
self-efficacy, self-liking, hope, determination and a
realistic orientation to life. ‘Social competence’
measured extraversion, social adeptness, cheerful
mood, an ability to initiate activities, good communi-
cation skills and flexibility in social matters. ‘Personal
structure’ measured the ability to uphold daily
routines, to plan and organize.
Reliability and validity
The internal consistency of the RSA subscales was high,
although the last dimension was in the lower part of the
recommended range (Streiner and Norman, 1995).
Although the reliability was satisfactory, it was
somewhat lower than the original study (Hjemdal,
Friborg, Martinussen and Rosenvinge, 2001). This may
be explained by the lower number of items in the present
RSA-scale, as compared to the previous version, which
generally reduce the Cronbach’s alphas (Cronbach,
1990). The item-total correlations for all subscale were
moderate to high, further indicating adequate reliability,
except for the last factor, ‘personal structure’, in which
one item total correlation dropped below 0.40.
All three explorations of the construct validity were
strongly supported. First, all the RSA subscales were

A new rating scale for adult resilience
73
positively related to the SOC and negatively related to
the HSCL. Discriminant validity between the RSA
and the SOC was also established by finding differing
correlation coefficients between the RSA subscales
and the SOC. A high sense of coherence indicates
how confident an individual feels that the outcome of
a stressful situation will be favourable (Antonovsky,
1998). This is promoted by the ability to comprehend,
manage and find meaning in life challenges
(Antonovsky, 1993). Such an ability is more
indicative of a personal competence than of, for
example, social support. As expected, ‘personal
competence’ correlated highest with the SOC,
whereas the other subscales correlated poorly to
moderately with SOC. As a stronger sense of
coherence also foster stronger feelings of coherence in
family and social matters (for example, that other
people’s behaviour is understandable), the moderate
correlations with ‘social competence’ and ‘family
coherence’ indicate tentative convergent validity for
these subscales as well. Still, the study did not include
other established scales to confirm the convergent
validity of the social and family subscale, and conse-
quently, further validation studies are needed.
this have been found essential for successful
adaptation (Werner, 2001).
Gender and age differences
Women reported significantly more access to social
support than men, whereas men reported more
personal competence than women. Although the last
difference was much weaker, it was still not negligible.
In scale development such gender differences are
generally unwanted. However, these results match
Werner’s longitudinal results (Werner, 1989),
indicating that men feel personally more competent
than women and that women generally are more
skilled in using social support. Similar findings come
from a meta-analysis that found higher self-esteem and
assertiveness among men, than among women.
Women, on the other hand, reported more extra-
version, trust, gregariousness and nurturance
(Feingold, 1994). In another overview (Cross, Marcus,
1993), women were more socially sensitive and
showed greater signs of stress than men when intimate
persons experienced straining life situations. As these
differences are common findings in the literature,
these items were retained in the RSA.
Secondly, like previously developed scales
(Wagnhild and Young, 1990, 1993; Jew, Green and
Kroger, 1999), the RSA differentiated between a
patient and randomly chosen sample from the normal
population of Tromsø. Patients reported less access to
protective factors within all five resilience dimensions.
It has been a consisted finding in earlier studies
(Werner, 2001) that individuals with psychological
troubles have fewer personal, social and external
resources available to protect them from stress than
more resilient individuals have. ‘Social support’ was
the only subscale that reached a sub-significant level.
The lower effect size for this subscale may be explained
by the high number of women in the patient sample
(76.3%), and thereby elevating the ‘social support’
scores. Third, current employment status (holding a
job) was moderately and positively associated with
higher levels of self-reported ‘Personal competence’,
‘social competence’ and ‘family coherence’. ‘Years of
work experience’, on the other hand, were positively
associated with ‘family coherence’ and ‘personal
structure’. These findings are in line with previous
results from longitudinal findings (Werner, 1993).
Surprisingly, though, ‘years of education’ was not
associated with higher degrees of resilience, although
The only subscale to be positively associated with
age, was ‘personal structure’. The ability to organize,
plan and maintain important structures and routines
to succeed with career and educational goals, along
with a careful planning choice of spouse, is one central
characteristic of resilient individuals (Werner, 2001).
The present results support the importance of
organizing and planning for adult resilience, and that
these characteristics may take more time to develop.
Sample limitations
As it was difficult to reach a sample consisting of
individuals that had successfully adapted and dealt
with long-term stress and difficulties, a random sample
from the population of Tromsø was contacted instead.
This may raise doubts about the external validity of
the scale. The control sample may, however, resemble
a resilient group as the majority of the population
generally overcomes several stressors, like loss of family
members, disease, troubled relatives or financial
problems, without developing difficulties. Indeed 32 of
the 230 subjects in the control sample reported experi-
encing taxing stressors during the four-month period,
such as serious somatic diseases, loss of significant
persons (death, separation, conflicts) and children in

Friborg et al.
74
trouble with the law. People in the normal population
are resilient in the sense that most individuals hold
positive views of themselves (positivity bias), display
optimism for the future and an illusion of control
(Taylor and Brown, 1988). To increase the probability
of comparing the patient sample against a psychologi-
cally healthy and thus more resilient sample, subjects
in the control sample with elevated symptom scores
were removed. New comparison between these
samples confirmed the results by finding somewhat
greater effect size differences. The probability of
reaching individuals with resilient characteristics in
this sample was therefore considered acceptable. In
future studies the probability of reaching a truly
resilient sample (those who have experienced taxing
stressors but still managed to cope) may be increased
by including life-event scales together with resilience
and coping scales to identify the subgroup of people
that have experienced adversities but still managed to
cope effectively.
• dispositional optimism (Scheier and Carver, 1985),
which is a central aspect of resilience, to see how
these co-vary and interact in the prediction of
mental health;
• health locus of control (Levenson, 1973) to inves-
tigate how resilience is associated with causal
attributions/beliefs about what influence good
health;
• social intelligence (Silvera, Matinussen and Dahl,
2001) to test if self-reported social competence in
the RSA co-variates with self reported social skills;
• coping styles (Lazurus, 1993) to see if resilient
people cope more actively (for example, solve
problems and seek information), rather than
passively (for example, avoidance or wishful
thinking) with life problems;
• life events, to examine how strongly the RSA
predicts individual differences in adaptational and
coping capacity when adversities/life problems
occur.
For both samples, the mean age was representative
for an adult population (25 and 50 years). The response
rate was low, but still comparable with that of previous
survey studies (Green and Boser, 1998), particularly
when using repeated measures. Although the response
rate was within the expected range, a low response rate
threatens the external validity if the sample is systemat-
ically biased. This was not the case for the patient
sample in terms of age and psychiatric diagnoses, with
the last being the most important indicator. The
proportions of psychiatric diagnoses in the patient
sample were not significantly different from the patients
who were invited to participate, and all patients in
treatment during the year 2000. The patient sample
may thus be regarded as representative of the patients
that seek psychiatric services at this particular clinic.
In sum, then, the age distribution, response rate, the
small differences in the dropout characteristics, as well
as a successful replication of the factor structure,
indicate acceptable generalizability of the scale.
Clinical use and implications
General resilient characteristics are presumed to be
more stable over time, than, for example, psychiatric
symptoms. They might therefore be more prognostic of
psychological growth during psychotherapy, as well as
providing better predictions of relapse rate and
patients’ ability to cope with present and coming diffi-
culties. This contention comes from several studies on
developmental psychopathology, claiming that under-
lying developmental level on social/emotional and
cognitive indices (pre-morbid functioning) is a consid-
erable predictor of good prognostic outcome (for an
overview, see Glick, 1997). In these studies, pre-morbid
social functioning has been found to be the most
potent predictor of clinical outcome variables. As the
RSA scale not only measures social competence but
also other important protective resources, one would
expect similar, or perhaps better, predictive ability.
In clinical and health psychology, it may be used as
an assessment tool of protective factors important to
prevent maladjustment and psychological disorders. For
the practitioner it points out key areas of psychological
and psychosocial interventions to help patients building
strength, by fostering protective resources known to
strengthen patients’ adaptability and self-reliance.
Further validation
In upcoming studies on construct validity of the RSA,
the scale will be compared with other establishment
instruments, such as an inventory for:
• personality (Costa and McCrae, 1985), to examine
how resilience is related to, for example,
neuroticism, which is a significant parameter for
predicting long-term psychological functioning;
Acknowledgement
We are grateful for the financial aid that the Psychiatric
Research Center for the Counties of Finnmark and Tromsø
provided to support the project.

A new rating scale for adult resilience
75
Garmezy N, Nuechterlein K. Invulnerable children: the fact
and fiction of competence and disadvantage. Am J
Orthopsychiatry 1972; 42: 328–9.
References
Antonovsky A. The structure and properties of the Sense of
Coherence Scale. Soc Sci Med 1993: 36; 725–33.
Antonovsky A. The structure and propoerties of the Sense of
Coherence scale. In HI McCubbin, EA Thompson, AI
Tompson, JE Fromer (eds) Stress, Coping and Health in
Families: Sense of Coherence and Resiliency. Thousand
Oaks CA: Sage Publications, 1998, pp. 21–40.
Aroian KJ, Schappler-Morris N, Neary S, Spitzer A, Tran
TV. Psychometric evaluation of the Russian language
version of the Resilience Scale. J Nursing Meas 1997; 5:
151–64.
Glick M. The developmental approach to adult
psychopathology. In SS Luthar, JA Burack, D Cicchetti,
JR Weisz (eds) Developmental Psychopathology:
Perspectives on Adjustment, Risk, and Disorder. New
York: Cambridge University Press, 1997, pp. 227–47.
Green KE, Boser JA, Hutchinson SR. Response-rate differ-
ences and response-enhancement effects by population
type. Psychol Rep 1998; 83: 336–8.
Hjemdal O, Friborg O, Martinussen M, Rosenvinge JH.
Preliminary results from the development and validation
of a Norwegian scale for measuring adult resilience. J
Norw Psychol Ass 2001; 38: 310–17.
Block JH, Kremen AM. IQ and ego-resiliency: conceptual
and empirical connections and separateness. J Pers Soc
Psychol 1996; 70: 349–61.
Jew CJ, Green KE, Kroger J. Development and validation of
a measure of resilience. Measurem Eval Counsel
Developm 1999; 32: 75–89.
Blum RWM. Healthy youth development as a model for
youth health promotion. J Adolesc Health 1998; 22:
368–75.
Kringlen E. Psykiatri [Psychiatry]. Oslo: Universitetsforlaget,
1990.
Lavik NJ, Laake P, Hauff E, Solberg Ø. The use of self-reports
in psychiatric studies of traumatized refugees: validation
and analysis of HSCL-25. Nord J Psychiatr 1998; 53:
17–20.
Cederblad M, Dahlin L, Hagnell O, Hansson K. Salutogenic
childhood factors reported by middle-aged individuals:
Follow-up of the children from the Lundby study grown
up in families experiencing three or more childhood
psychiatric risk factors. Eur Arch Psychiatry Clin
Neurosci 1993; 244: 1–11.
Lazarus RS. Coping theory and research: past, present, and
future. Psychosom Medi 1993; 55: 234–47.
Cederblad M. Fifty years of epidemiologic studies in child
and adolescent psychiatry in Sweden. Nord J Psychiatry
1996; 50: 55–66.
Cicchetti D, Garmezy N. Prospects and promises in the study
of resilience. Dev Psychopathol 1993; 5: 497–502.
Clausen JA. American lives: looking back at the children of
the Great Depression. New York: The Free Press, 1993.
Costa PT, McCrae RR. The NEO Personality Inventory
Manual. Odessa FL: Psychological Assessment Resources,
1985.
Cowen E, Work W. Resilient children, psychological
wellness, and primary prevention. Am J Community
Psychol 1988; 16: 591–607.
Cronbach LJ. Essentials of Psychological Testing. New York:
HarperCollins, 1990, pp. 190–223.
Levenson H. Multidimensional locus of control in psychi-
atric patients. J Consult Clin Psychol 1973; 41, 397–404.
Luthar SS, Doernberger CH, Zigler E. Resilience is not a
unidimentional construct: Insights from a prospective
study of inner-city adolescents. Dev Psychopathol 1993;
5: 703–17.
Moum T. Mode of administration and interviewer effects in
self-reported symptoms of anxiety and depression. Soc
Indicat Research 1998; 45: 279–318.
Mrazek PJ, Mrazek D. Resilience in child maltreatment
victims: a conceptual exploration. Child Abuse Negl
1987; 11: 357–65.
Nettelbladt P, Hansson L, Stefansson CG, Borgquist L,
Nordström G. Test characteristics of the Hopkins
Symptom Check List-25 (HSCL-25) in Sweden, using
the Present State Examination (PSE-9) as a caseness
criterion. Soc Psychiatry Psychiatr Epidemiol 1993; 28:
130–3.
Cross SE, Marcus HR. Gender in thought, belief, and action:
A cognitiv approach. In AE Beall and RJ Sternberg (eds)
The Psychology of Gender. New York: Guilford, 1993, pp.
80–1.
Egeland B, Carlson E, Sroufe AL. Resilience as process. Dev
Psychopathol 1993; 5: 517–28.
Rudmin FW. Norwegian short-form of the Marlowe-Crowne
Social Desirability Scale. Scand J Psychol 1999; 40:
229–33.
Feingold A. Gender differences in personality: A meta-
analysis. Psychol Bull 1994; 116: 429–56.
Rutter M. Resilience in the face of adversity. Br J Psychiatry
1985; 147: 598–611.
Frenz AW, Carey MP, Jorgensen RS. Psychometric evalu-
ation of Antonovsky’s Sense of Coherence scale. Psychol
Assess 1993; 5: 145–53.
Garmezy N. Children in poverty: resilience despite risk.
Psychiatry 1993; 56: 127–36.
Garmezy N. Children under stress: perspectives on
antecedents and correlates of vulnerability and resistance
to pathology. In AL Rabin, J Arnoff, AM Barclay, RA
Zuckers (eds) Further Explorations in Personality. New
York: Wiley, 1981: 196–296.
Rutter M. Psychosocial resilience and protective mecha-
nisms. In JE Rolf, AS Masten (eds) Risk and Protective
Factors in the Development of Psychopathology. New
York: Cambridge University Press, 1990, pp. 181–214.
Sammallahti PR, Holi MJ, Komulainen EJ, Aalberg VA.
Comparing two self-report measures of coping - The
Sense of Coherence scale and the Defense Style
Questionnaire. J Clin Psychol 1996; 52: 517–24.

Friborg et al.
76
Scheier MF, Carver CS. Optimism, coping, and health:
assessment and implications of generalized outcome
expectancies. Health Pscyhol 1985; 4: 219–47.
Silvera DH, Martinussen M, Dahl TI. The Tromsø Social
Intelligence Scale, a self report measure of social intelli-
gence. Scand J Psychol 2001; 42: 313–19.
psychological resilience: a phenomenological perspective
on deflecting life’s slings and arrows. J Prim Preven 1995;
15: 209–46.
Werner EE. High-risk children in young adulthood: A
longitudinal study from birth to 32 years. Am J
Orthopsychiatry 1989; 59: 72–81.
Steele BF. Abuse and neglect in the earliest years:
Groundwork for vulnerability. Zero to Three 1987; 4:
14–15.
Werner EE. Risk, resilience and recovery: Perspectives from
the Kauai longitudinal study. Dev Psychopathol 1993; 5:
503–15.
Streiner DL, Norman GR. Health Measurement Scales: A
Practical Guide to Their Development and Use. New
York: Oxford University Press, 1995.
Werner EE. Journeys from childhood to midlife: risk,
resilience and recovery. Ithaca, New York: Cornell
University Press, 2001.
Svanberg POG. Attachment, resilience and prevention.
J Ment Health 1998; 7: 543–78.
Taylor SE, Brown J. Illusion and well-being: a social psycho-
logical perspective on mental health. Psychol Bull 1988;
103: 193–210.
Werner EE, Smith RS. Overcoming the odds. High risk
children from birth to adulthood. Ithaca and London:
Cornell University Press, 1992.
Zimrin H. A profile of survival. Abuse Neglect 1987; 10:
339–49.
Wagnhild G, Young HM. Resilience among older women. J
Nurs Scholars 1990: 22; 252–5.
Wagnhild G, Young HM. Development and psychometric
evaluation of the resilience scale. J Nurs Meas 1993; 1:
165–78.
Correspondence: Oddgeir Friborg, Department of
Psychology, University of Tromsø N-9037, Tromsø,
Norway.
Telephone (+47) 776 45 945.
Fax (+47) 776 45 291.
Walsh F. The concept of family resilience: crisis and challenge.
Family Process 1996; 35: 261–81.
Watt NF, David JP, Ladd KL, Shamos S. The life course of
E-mail: ofriborg@psyk.uit.no.