ArticlePDF Available

Development and Validation of an Improved Hardiness Measure: The Hardiness Resilience Gauge

Authors:
  • Independent Researcher

Abstract

Previous research shows that psychological hardiness is an important factor contributing to stress resilience in individuals. Of the various instruments available to measure hardiness, the most commonly used is the Dispositional Resilience Scale (DRS). Despite its demonstrated utility, the DRS-15 still has a number of serious limitations, including low subscale reliability and limited construct validity. The present work aims to create a new hardiness scale that addresses these limitations. A pool of new items plus the original DRS item set was administered to a census-matched stratified sample of N = 2,021 men and women across the United States. Items for the new scale were selected based on item distribution characteristics, item response theory plots, scale reliabilities, item-total correlations, and confirmatory factor analysis (CFA). CFA results showed the best fitting model reflected a hierarchical structure with three factors (commitment, control, and challenge) nested under a broad hardiness factor. This factor structure is replicated in two independent validation samples and also holds invariant across gender and age. The new scale shows much improved reliability coefficients (e.g., Cronbach’s α of .93, .85, .84, and .89 for total hardiness, challenge, control, and commitment, respectively), as well as structural equivalence across gender and age. Validity is demonstrated in multiple samples via predictive associations of hardiness scores with theoretically relevant outcome measures, including coping, life satisfaction, anxiety, and depression. The Hardiness Resilience Gauge (HRG) possesses excellent reliability and validity and appears to be a more effective tool for measuring hardiness in adult populations.
Multistudy Report
Development and Validation of
an Improved Hardiness Measure
The Hardiness Resilience Gauge
Paul T. Bartone
1
, Kelly McDonald
2
, Braden J. Hansma
2
, Jonathan Stermac-Stein
2
,
E. M. Romero Escobar
2
, Steven J. Stein
2
, and Rebecca Ryznar
3
1
Institute for National Strategic Studies, National Defense University, Washington, DC, USA
2
Multi-Health Systems, Inc., Toronto, Canada
3
Biomedical Sciences Department, Rocky Vista University, Parker, CO, USA
Abstract: Previous research shows that psychological hardiness is an important factor contributing to stress resilience in individuals. Of the
various instruments available to measure hardiness, the most commonly used is the Dispositional Resilience Scale (DRS). Despite its
demonstrated utility, the DRS-15 still has a number of serious limitations, including low subscale reliability and limited construct validity. The
present work aims to create a new hardiness scale that addresses these limitations. A pool of new items plus the original DRS item set was
administered to a census-matched stratified sample of N= 2,021 men and women across the United States. Items for the new scale were
selected based on item distribution characteristics, item response theory plots, scale reliabilities, item-total correlations, and confirmatory
factor analysis (CFA). CFA results showed the best fitting model reflected a hierarchical structure with three factors (commitment, control, and
challenge) nested under a broad hardiness factor. This factor structure is replicated in two independent validation samples and also holds
invariant across gender and age. The new scale shows much improved reliability coefficients (e.g., Cronbachsαof .93, .85, .84, and .89 for total
hardiness, challenge, control, and commitment, respectively), as well as structural equivalence across gender and age. Validity is
demonstrated in multiple samples via predictive associations of hardiness scores with theoretically relevant outcome measures, including
coping, life satisfaction, anxiety, and depression. The Hardiness Resilience Gauge (HRG) possesses excellent reliability and validity and
appears to be a more effective tool for measuring hardiness in adult populations.
Keywords: hardiness, measurement, stress, resilience, Hardiness Resilience Gauge
First identified by Kobasa (1979)inastudyofstressand
health in Chicago telephone executives, the concept of per-
sonality hardiness has proved to be an important factor
influencing human resilience. A sizable body of literature
now indicates that personality hardiness can enhance resili-
ence, protecting some individuals against the ill effects of
stress on health and performance (Bartone, 1989;Contrada,
1989; Eschleman et al., 2010; Kobasa & Puccetti, 1983;
Wiebe, 1991). For example, in a study of Gulf War veterans,
combat-exposed soldiers who were high in hardiness expe-
rienced fewer posttraumatic stress disorder (PTSD) symp-
toms than those low in hardiness (Bartone, 1999).
Hardiness is generally conceived as an attitudinal style or
world view that develops early in life and is reasonably
stable over time, though amenable to change under certain
conditions (Bartone, 2006; Maddi & Kobasa, 1984). Hardy
persons have a strong sense of life and work commitment, a
greater feeling of control, andaremoreopentochangeand
challenges in life. They tend to interpret stressful and painful
experiences as a normal aspect of existence, part of life that
is overall interesting and worthwhile. While hardiness com-
mitment, control, and challenge show some conceptual
similarity to other constructs such as organizational com-
mitment, locus-of-control, and Big Five openness, they
are theoretically and empirically quite distinct (Bartone
et al., 2009;Maddi,2002). For example, considerable work
has shown that high hardiness is associated with active,
problem-solving coping strategies and better mental health
and adjustment, especially under stressful conditions
(Eschleman et al., 2010), whereas low hardiness is linked
to avoidance coping approaches, poor mental health includ-
ing depression, anxiety and PTSD, and stress vulnerability
(Bartone & Homish, 2020; Bartone et al., 2015; Thomassen
et al., 2018). Hardiness has also been linked to greater well-
being(Lambertetal.,1989) and relationship satisfaction
(Nabizadeh & Mahdavi, 2016).
Of the various instruments available to measure hardi-
ness, the most commonly used is the Dispositional Resili-
ence Scale (DRS; Bartone, 1995,2013;Windle,Bennett&
Noyes, 2011). Although this scale has gone through a num-
ber of revisions, some notable limitations remain, including
low sub-scale reliability. The present study was undertaken
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
https://doi.org/10.1027/1015-5759/a000709 under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
to develop a new hardiness scale that addresses these pre-
vious limitations.
Measuring Hardiness
The measurement of hardiness was highly problematic in
the early years of work with the concept. It was originally
assessed by Kobasa (1979)withanamalgamof18 different
psychological scales that aimed at capturing the dimensions
of commitment, control, and challenge. This original collec-
tion of over 100 items was later reduced to several shorter
versions (Ouellette, 1993), but these still had a number of
serious problems. These scales were composed exclusively
of negative or non-hardyitems, opening the door to con-
tamination by neuroticism (e.g., No matter how hard I try,
my efforts will accomplish nothing; Funk, 1992). Item
response distributions were highly skewed, and many were
found to be confounded with political attitudes and beliefs
(e.g., Government should guarantee jobs for all; Bartone,
1984). Also, multiple studies were unable to replicate the
theoretical three-factor structure of hardiness, further call-
ing into question the validity of these scales (Funk, 1992).
A shorter and more coherent hardiness test with 50 items
was subsequently developed by Bartone (1984,1989)using
samples of bus drivers and telephone company managers.
Later, this scale was refined into a 45-item hardiness mea-
sure with a balance of positive and negative items and an
equal number of items to measure the facets of commit-
ment, control, and challenge (DRS; Bartone et al., 1989).
In a critical review of hardiness theory and research, Funk
(1992) recommended this scale as the best available hardi-
ness measure at the time. Also, using the DRS, Sinclair and
Tetrick (2000) confirmed a factor structure of three dimen-
sions commitment, control, and challenge nested under
a more general hardiness factor. The DRS was subsequently
shortened and improved in various ways, resulting in a 30-
item and 15-item version (Bartone, 1991,1995). The DRS-15
has been used extensively in military and non-military sam-
ples, with fairly good results (Andrew et al., 2013;Bartone
et al., 1989,2008; Britt et al., 2001). A final revision of
the short DRS-15 sought to improve scale reliabilities and
eliminate linguistic bias in the wording of items (Bartone,
2013). The revised DRS-15 shows solid psychometric prop-
erties (Hystad et al., 2010) and evidence of predictive valid-
ity (Bartone, Valdes, et al., 2016;Johnsen et al., 2013).
Despite its success as a measure of hardiness, the DRS-15
still has several serious limitations. At five items each, the
subscales of commitment, control, and challenge often
show lower reliability coefficients than desired. For exam-
ple, a recent study of 570 collegiate athletes found low
Cronbachsαcoefficients of .58,.67, and .67 for hardiness
commitment, control and challenge, respectively, and only
.69 for the total scale (Madrigal et al., 2016). This study also
found that the theoretical three-factor hardiness model did
not show a good fit for the data. Hystad et al. (2010)
reported an even lower αcoefficient of .62 for the hardiness
challenge dimension, with somewhat better coefficients for
commitment (.76) and control (.74). Similarly, low αcoeffi-
cients were found in a sample of Norwegian navy cadets,
with a low of .62 for challenge and .73 for commitment
(Hystad et al., 2015). In a study of Norwegian undergradu-
ate students, reliability coefficients again were low at .65 for
commitment, .72 for control, .71 for the challenge, and .71
for the total hardiness scale (Hystad et al., 2009). Thus,
scale reliability is a problem for the DRS-15, especially with
regard to the challenge scale.
A further limitation of the DRS-15 is that the short 5-item
scales may not fully capture the complexity of the hardiness
facets, which is a construct validity weakness. For example,
the challenge items for the DRS-15 all focus on the dimen-
sion of appreciation for variety and novelty in life and fail to
address other conceptually important aspects of challenge.
Missing is any attention to the tendency to view changes
and disruptions as challenges to overcome and appraising
failures as opportunities to learn and grow. The present
work seeks to address these shortcomings by creating a
new, more reliable, and construct valid hardiness measure
with robust psychometric and structural properties across
different populations.
Study 1: Scale Development
and Initial Evaluation
In the first study, we describe the item development and
selection process and the preliminary evaluation of the
new scale on a large US-based sample. This group also pro-
vides the initial normative data for the new scale.
Materials and Methods
Item Development
Apoolof36 potential items was created that included the
original 15 items of the DRS, with 21 new items. In develop-
ing new items, two authors (PB and KM) conducted an
extensive review of the hardiness empirical and theoretical
literature and then independently wrote 1215 new items.
New items aimed to measure the hardiness facets that were
not redundant with the original 15 items of the DRS. Profes-
sional, academic, and theoretical jargon was avoided to cre-
ate easily understood items that would be relevant for all
adult respondents. Items were also designed to be free of
linguistic bias and idiomatic expressions. Following exten-
sive review and discussion, the most theoretically relevant
and unique items were retained for further consideration.
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 223
This resulted in 21 new items, 8for challenge, 6for control,
and 7for commitment. In addition, 7of the original 15 DRS
items were slightly adjusted for clarity. For example, mod-
ifying adverbs (e.g., somewhat, really) were removed to
make a point of the statement clearer for respondents.
The end result of this process was 36 total items retained
for more detailed analysis: 13 for challenge, 11 for control,
and 12 for commitment.
Normative Sample
Using an online study panel obtained through Amazon
Mechanical Turk (Mortensen & Hughes, 2018), a sample
of N=2,016 respondents answered a web-based survey
that included the 36 hardiness items, as well as a series
of demographic questions and several additional instru-
ments (described below). The sample was representative
of the adult population (18 years and older) of the United
States and was matched to 2016 US census data for age,
gender, race/ethnicity, education, geographic region, and
employment status (United States Bureau of the Census,
2016). Participants were recruited from all 50 US states,
with specific demographic targets.
To ensure data integrity, responses were screened for
patterns of inconsistency or dubious validity (Goldammer
et al., 2020). Cases showing evidence of careless respond-
ing were dropped from further analysis. Examples included
unusually rapid (< 2min to answer all 36 items) or delayed
(> 35 min) completion of the survey. Cases showing identi-
cal responses to 13 or more consecutive items were also
excluded, as these subjects appeared to be checking boxes
without reading the items. Respondents who failed to
answer four or more items were also dropped from further
analysis. Applying these criteria, a total of 143 cases were
thus excluded for spurious or incomplete responses. Of
the remaining 1,873 cases, cells were created containing
an equal number of cases for different age and gender
groups, with proportional representation in geographic
region, race, education, and employment groups. Table 1
presents additional details on the normative sample.
Item Selection
From the pool of 36 items, our goal was to identify the most
psychometrically sound items, retaining at least eight items
for each hardiness facet to assure construct validity and
scale reliability. The following criteria guided the selection
of the final items:
(1)Response distribution: An ideal test item should show
good distribution characteristics with all possible
response options endorsed and a lack of extreme
skewness. Items were selected that displayed distribu-
tions in which all possible response options were
endorsed to acceptable degrees (extreme responses
were less frequent relative to other response options).
(2)Item-total correlations: The degree to which each item
correlates with other items on its purported subscale
was also examined. Selected items displayed moder-
ate to high correlations with the remaining items on
their respective subscales, with a minimum r=.30.
(3)Confirmatory factor analyses (CFAs): CFA was applied
to assess the degree to which each item contributed
to its presumed subscale, as well as to determine the
appropriateness of a three-factor hierarchical model.
Items were retained that loaded .32 or greater on their
intended factor while not cross-loading on other fac-
tors (Brown, 2015). An item was determined to
cross-load if it: (a) loaded at .32 or greater on more
Table 1.Demographic variables and their distribution in the normative
sample
Frequency
Variable N%
Gender
Male 750 50.0
Female 750 50.0
Age group (years)
1824 300 20.0
2534 300 20.0
3554 300 20.0
5564 300 20.0
65+ 300 20.0
Median age (SD) 43 (18.7)
Racial/ethnic group
Asian 82 5.5
Black 177 11.8
Hispanic 241 16.0
White 957 63.8
Other 43 2.9
Geographic region
Northeast US 264 17.6
Midwest US 372 24.8
Southern US 528 35.2
Western US 336 22.4
Education level
No high school diploma 147 9.8
High school graduate 350 23.3
Some college or associate degree 505 33.7
Bachelors degree 309 20.6
Graduate school 189 12.6
Employment status
Employed or self-employed: Full-time 589 39.3
Employed or self-employed: Part-time 190 12.7
Retired 382 25.4
Student: Full-time 167 11.1
Student: Part-time 43 2.9
Unemployed 129 8.6
Total 1,500
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
224 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
than one scale, and (b) the difference between load-
ings on two factors was < .10. Additional CFAs were
conducted to identify an overall model with the best
fit for the data. Three models were tested: (1) a single
general factor; (2) three independent factors; (3) three
independent factors nested under a general hardiness
factor (hierarchical model).
(4)Item-response theory (IRT): IRT was applied as an addi-
tional step in identifying well-performing items to
retain. Ideally, good items should (a) yield adequate
information about the underlying constructs (chal-
lenge, control, commitment) and (b) provide useful
information about respondents at all levels of the con-
struct being measured (Zanon et al., 2016). Item char-
acteristic curves and item information curves were
plotted and examined for desirable and undesirable
properties. Items showing the most positive IRT pat-
terns, such as equally dispersed trace distributions
for polytomous responses and high levels of item
information across all levels of the latent trait (theta),
were retained for the final measure.
Norming Procedure
In establishing normative reference groups, the first step
was to determine if any age or gender differences were pre-
sent in the data. If significant differences were obtained for
different respondent groups, separate norms would be
appropriate. To test this, measurement invariance of the
new Hardiness Resilience Gauge (HRG) scale was first
assessed using multigroup CFA procedures (Horn et al,
1983;Jöreskog,1971; Meredith, 1993). Following this, a ser-
ies of analysis of variance (ANOVA) tests were performed
to check for gender and age group mean differences. For
measurement invariance, a factor structure consistent with
hardiness theory (a general hardiness factor with three sub-
dimensions) and confirmed with our own CFAs, was exam-
ined across gender (male and female) and age groups.
Here, age was collapsed into five groups to improve inter-
pretability (i.e., 1824,2534,3554,5564,and65+ years).
Measurement invariance was tested through a series of
hierarchical tests where increasingly restrictive equality
constraints were made on parameters across groups. The
goal of this process was to determine whether the HRG
met the conditions of scalar invariance (i.e., the indicator
thresholds and unstandardized factor loadings are statisti-
cally equivalent across groups). For the invariance test,
Wu and Estabrooks(2016) approach to model identifica-
tion and measurement invariance testing was used. With
this approach, threshold invariance is tested before factor
loading invariance (Svetina et al., 2019). The sequence of
models tested was: (1) configural model, equating the pat-
tern of factor loadings across groups; (2)metricmodel,test-
ing for weak invariance, or equating item thresholds across
groups; and (3) scalar model, testing for strong invariance,
or equating both thresholds and unstandardized factor
loadings across groups. Changes in model fit were assessed
through Satorra-Bentler scaled chi square difference tests
(Satorra & Bentler, 2001) and decreases in robust goodness
of fit indices (ΔGFI), such as ΔCFI and ΔRMSEA. Because
our observed variables were ordinal and not continuous,
and our factor structure was multidimensional, we used
Svetina and Rutkowskis(2017) recommended cut-offs for
ΔGFI, namely, for the metric model, we used a ΔRMSEA
cut-off of .05, and for the scalar model a cut-off of
ΔRMSEA .01 and ΔCFI .002.
For analysis of variance (ANOVA) tests, effect sizes are
reported in addition to statistical significance. Effect sizes
were assessed with Cohensdstatistic for gender (Cohen,
1988), applying Cohens benchmark guidelines of .20 for
small effects, .50 for medium, and .80 for large effect sizes.
Since age was a categorical variable with multiple groups,
the more appropriate partial-eta squared (η
p
2
) statistic was
used to assess effect size (Lakens, 2013). Here, guidelines
suggested by Cohen (1988)indicate.01 as a small effect,
.06 as medium, and .14 as large effect size.
Reliability and Validity
Reliability of the new scale was assessed using Cronbachs
αcoefficient, McDonaldsω, and also via a 3-week test-ret-
est reliability analysis. McDonaldsωwasusedinaddition
to the more common Cronbachsα, since it is believed to be
a more accurate indicator of reliability when assumptions
regarding tau equivalence are not met (Hayes & Coutts,
2020). Construct and factorial validity were assessed with
CFA to determine if the factor structure was consistent with
hardiness theory (a general hardiness factor with three sub-
dimensions). Convergent and divergent validity were eval-
uated by assessing Pearson correlations between the new
hardiness test and scores on several theoretically relevant
measures, including coping styles and life satisfaction. Here
we also refer to Cohens(1988) conventions for effect size,
such that a correlation of .10 is considered small, .30 mod-
erate, and .50 or more as large.
Measures
In addition to the DRS-15 and the HRG items under devel-
opment, three additional measures were administered to
assess the validity of the new instrument in the normative
sample.
Coping was measured with two scales from the Coping
Inventory for Stressful Situations (CISS; Endler & Parker,
1994). Relationships between scores on the HRG and the
CISS subscales of Task-Oriented Coping and Emotion-
Oriented Coping were explored. Task-Oriented Coping
entails actively pursuing long-term solutions to address
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 225
problems that are causing stress. Emotion-Oriented Coping
involves emotional reactions to stressful events (e.g., getting
angry), self-preoccupation, and fantasizing (e.g., daydream-
ing). These types of reactions are generally maladaptive. It
was expected that respondents higher in hardiness would
also be higher in Task-Oriented Coping and lower in Emo-
tion-Oriented Coping. Each scale contains 16 items, with
Cronbachsαcoefficients ranging from .87 to .92 for
Task-Oriented Coping and .82 to .90 for Emotion-Oriented
Coping scale (Endler & Parker, 1994). In the present sam-
ple, Cronbachsαfor Task-Oriented Coping was .93,and
for Emotion-Oriented Coping .90.
The Satisfaction With Life Scale (SWLS) is a 5-item mea-
sure designed to capture global judgments of a persons
overall life satisfaction (Diener et al., 1985). Cronbachsα
in the present sample was .90. Hardiness scores should
be associated with increased life satisfaction.
Relationship satisfaction was assessed using a single-item
indicator (I am satisfied with my relationships)ona
response scale ranging from Strongly Disagree to Strongly
Agree. Previous research has found that assessing relation-
ship satisfaction with a single-item is an adequate approach
for measuring relationship satisfaction (Fülöp et al., 2020).
Hardiness scores should be associated with higher levels of
relationship satisfaction.
Data Analysis and Statistics
All analysis was performed in R Version 4.0.4(R Core
Team, 2020). As detailed above, steps in the evaluation
of test items for the HRG included examination of response
distributions (including IRT techniques performed with the
lordif and mirt package version 0.3-3and 1.34,respec-
tively; Choi et al., 2011;Chalmers,2012), assessment of
skewness and kurtosis, and item-total correlations. CFA
was applied using the Lavaan package (version 0.6-8;Ros-
seel, 2012) to test several different measurement models
for the HRG data, with robust model fit statistics used to
determine the best-fitting model, including the Satorra-
Bentler scaled chi square test, CFI, TLI, RMSEA, and
SRMR. Lavaan was also used for testing measurement
invariance. Reliability of the new HRG scale and subscales
was assessed with CronbachsαstatisticaswellasMcDon-
aldsω(Hayes & Coutts, 2020)both statistics were calcu-
lated using the psych package (version 2.1.9; Revelle, 2021).
Pearson correlations were used to evaluate the association
of hardiness scores with relevant outcome measures (using
the stats package version 3.6.2; R Core Team, 2021)
Results
Factor Structure
Application of all item selection criteria resulted in the
retention of 28 items, 10 challenges, 10 commitments,
and 8controls (sample items for each subscale can be
foundinTable2). The factor structure of the 28 items
was expected to parallel the established hierarchical factor
structure of the previous DRS scale, in which the three
dimensions of challenge, control, and commitment are
nested under the higher-order hardiness construct. CFA
was performed using data from the normative sample.
Three models were tested. The first was a single-factor
model in which all items were loaded onto a single hardi-
ness factor. The second model was a three-factor solution
in which the facets of challenge, control, and commitment
were uncorrelated. The third model was the hierarchical
one in which the three facets were nested under a general
hardiness factor. For all models, a polychoric variance-cov-
ariance matrix was estimated, and Weighted Least Squares
Mean and Variance estimation was used. This was done to
account for the ordinal nature of the data, which also did
not fully meet assumptions of multivariate normality and
homoscedasticity. Also, for this reason, we used the
Satorra-Bentler scaled chi-square test (Satorra & Bentler,
2001), and the more stringent robust goodness of fit indices
where applicable to determine the best fitting model,
including the Satorra-Bentler scaled chi square test (Satorra
& Bentler, 2001), the Comparative Fit Index (CFI; Bentler,
1990), the Tucker-Lewis Index (TLI; Tucker & Lewis,
1973), the Standardized Root Mean Square Residual
(SRMR; Hu & Bentler, 1999), and the Root Mean Square
Error of Approximation (RMSEA; Steiger & Lind, 1980).
RobustCFIandTLIvalues>.90 and SRMR and robust
RMSEA values < .10 were used as indicators of adequate
model fit (Hu & Bentler, 1999).
Results confirmed that the hierarchical model provided
the best fit to the data, superior to alternative models. For
the hierarchical model, robust fit indices were CFI = .91,
TLI = .90,RMSEA=.09,andSRMR=.06. The comparable
non-robust fit indices were considerably higher, at CFI =
.98,TLI=.98,andRMSEA=.08. Complete model statistics
are provided in Table 3.
Measurement Invariance
Measurement invariance analysis following Svetina and
Rutkowskis(2017) approach revealed that males and
Table 2.Sample items from HRG Subscales
Subscale Item
Challenge I find the positives in any life change.
Even if I fail at something, I look for ways to improve.
Control I am confident I can accomplish whatever.
I set out to do.
I am responsible for my own success in life.
Commitment I have a clear sense of purpose in my life.
I feel energized about life.
Note. For access to the full Hardiness Resilience Gauge, please contact the
publisher, Multi-Health Assessments (2018). https://www.storefront.
mhs.com/collections/hrg
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
226 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
females were invariant on the HRG (i.e., the conditions for
strong invariance were met; Table 7, top portion). ANOVA
results also showed that males (N=750)andfemales(N=
750) did not differ significantly on their total standardized
hardiness scores, nor on any of the three subscales. Further,
none of the observed gender differences reached the min-
imum criterion for even a small effect size. These results
suggest that the HRG is measuring males and females in
the same way and that hardiness as measured by the
HRG is similar for both males and females. Descriptive
statistics, significance tests, and effect sizes are displayed
in Table 4.
Age effects were also found to be negligible. Here, the
sample was stratified into five age groups of N=300 each,
as follows: 1824 years, 2534 years, 3554 years, 5564
years, and 65+ years. Results verified measurement invari-
ance across age groups (see the bottom portion of Table 7).
It was also found that respondents of different ages did not
differ significantly on their total standardized hardiness
scores, nor on any of the three subscales this finding
was corroborated through examination of effect sizes, none
of which reached the minimum criterion for even a small
effect size. Taken together, the HRG appears to measure
individuals of different age groups similarly, and hardiness
scores on the HRG do not appear to substantially vary with
age. Table 5provides additional descriptive statistics, signif-
icance tests, and effect sizes for age.
Finally, there was no significant age by gender interac-
tions, and none of the interactions reached even the small
effect size level. Given the negligible age and gender effects
(i.e., measurement invariance and no group mean differ-
ences), a single normative group representing the US general
population was deemed most appropriate. Standard scores
with a mean of 100 and a standard deviation of 15 were esti-
mated for the challenge, control, and commitment subscales.
For the challenge and commitment subscales, the distribu-
tions of standard scores showed appropriate dispersion and
shape. For the control subscale, the normative means and
standard deviations of raw scores were statistically smoothed
at the high end of the distribution to ensure a similarly appro-
priate dispersion and shape (Roid, 1992;Zachary&Gorsuch,
1985). To compute total hardiness scores, standard scores of
the three subscales were summed and standardized to a
mean of 100 and a standard deviation of 15.
Reliability
Internal Consistency
Internal consistency is typically measured using Cronbachs
α(Cronbach, 1951), which is a function of both the interre-
latedness of the test items and the length of the test (John &
Benet-Martinez, 2000). In the normative data, Cronbachs
αfor total hardiness was .93, indicating high reliability. The
αvalues for the subscales were also high, at .89 for commit-
ment, .84 for control, and .85 for challenge.
Hayes and Coutts (2020), among others, have argued for
the use of McDonaldsω(1999) rather than Cronbachsα
as a more robust measure of scale reliability (Peters,
2014). Unlike Cronbachsα,McDonaldsωdoes not
Table 3.Goodness of fit indices for three alternative HRG models
One-factor Three-factor orthogonal model Three-factor hierarchical model
Number of parameters 112 112 115
Scaled w
2
7,470.81 27,291.95 4,415.53
df 350 350 347
p< .001 < .001 < .001
CFI .84 .40 .91
TLI .83 .35 .90
RMSEA .12 .23 .09
SRMR .08 .31 .06
Note. CFI = Comparative Fit Index; df = Degree of Freedom; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square
Residual; TLI = Tucker-Lewis Index. Robust estimates for CFI, TLI, and RMSEA are reported due to some deviation from multivariate normality and
homoscedasticity. The non-robust estimates were considerably higher (hierarchical model, CFI = .98, TLI = .98, and RMSEA = .08).
Table 4.HRG distribution by gender in the normative sample
Males (N= 750) Females (N= 750)
M SD M SD F(1, 1,490) pCohensd
Total hardiness 99.4 15.5 100.6 14.4 0.06 .808 0.09
Challenge 99.7 14.8 100.4 15.2 0.77 .381 0.05
Control 100.3 17.1 102.1 15.7 1.26 .262 0.12
Commitment 99.6 15.7 100.5 14.3 0.13 .723 0.06
Note. Recommended guidelines for evaluating Cohens|d| are 0.20 = small, 0.50 = medium, 0.80 = large (Cohen, 1988).
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 227
assume tau equivalence (the variance between a scale item
and the latent variable measured by the scale is approxi-
mately equal across items in the scale). When tau equiva-
lence is not met, Cronbachsαis likely to be a negatively
biased estimator of reliability (Sijtsma, 2009). An examina-
tion of the standardized factor loadings, as well as fitting a
CFA model in which factor loadings were constrained to be
equal across items, revealed that assumptions of tau equiv-
alence were not met for all HRG scales in the normative
sample (i.e., model fit was worse than an unconstrained
model). Thus, McDonaldsωis the more appropriate relia-
bility coefficient. Table 8(top portion) displays the McDon-
aldsωand Cronbachsαcoefficients for the HRG and its
subscales.
Test-Retest Reliability
In assessing test-retest reliability, a 3-week interval was
deemed optimal. A shorter interval would increase the
opportunity for memory effects to influence responses
(Downie & Heath, 1970), while a longer interval provides
the chance for real developmental shifts to affect responses.
Asubsetof168 randomly chosen individuals was selected
from the normative sample to complete the HRG test twice
over a three-week interval (mean interval = 22.1days, SD =
1.6days). The test-retest correlation for HRG total hardi-
ness was high at r=.81. The hardiness subscales also
showed good test-retest reliability with challenge at r=
.80, control at r=.74,andcommitmentatr=.79. Effect
sizes for the differences over time were small, indicating
that the differences between Time 1and Time 2scores
did not reveal any meaningful change in scores. Descriptive
statistics, effect sizes, and test-retest correlations are shown
in Table 6.
Validity
Relationship to Other Constructs
In addition to factorial validity as established by the CFA,
the validity of the new scale was further evaluated by exam-
ining its association with other relevant psychological con-
structs. These analyses provide evidence for convergent
and discriminant validity, evaluating the extent to which
the new scale (HRG) is measuring the intended hardiness
construct.
Hardiness Resilience Gauge and Coping Styles
As expected, results showed that HRG scores correlated
positively with Task-Oriented Coping, r=.68,p<.001
and negatively with Emotion-Oriented Coping r=.30,p
<.001. An examination of the relationships between the
HRG subscales and the CISS coping styles revealed similar
results. Commitment, control, and challenge scale scores
were all strongly positively correlated with Task-Oriented
Coping and negatively correlated with Emotion-Oriented
Coping. With Task-Oriented Coping, correlations with com-
mitment, control and challenge were .61,.55,and.62
respectively (all p<.001). With Emotion-Oriented Coping,
correlations with commitment, control, and challenge were
.31,.22,and.26, respectively (all p<.001). Overall,
HRG scores were related to coping styles in expected ways.
Respondents with higher HRG scores were more likely to
use positive coping strategies and less likely to use nega-
tive coping strategies compared to those with low HRG
scores.
Hardiness Resilience Gauge and Life Satisfaction
People who are high in hardiness experience a range of pos-
itive life outcomes, including less life and work stress and
increased psychological well-being (Eschleman et al.,
2010). As expected, total hardiness was positively related
to SWLS scores, r=.48,p<.001. Thus, persons higher in
hardiness reported more life satisfaction. Similar findings
were observed when looking at the HRG subscales individ-
ually. For Satisfaction with Life, correlations were .56,.39,
and .33 with commitment, control, and challenge, respec-
tively, all p<.001.
HRG and Relationship Satisfaction
HRG total hardiness scores correlate as expected with Rela-
tionship Satisfaction, r=.46,p<.001. Those high in hardi-
ness are more satisfied with the quality of their social
relationships. For the hardiness facets, correlations are also
positive with Relationship Satisfaction, at .52,.38,and.32
with commitment, control, and challenge, respectively, all
p<.001. The strongest correlation is with commitment,
which makes theoretical sense considering that this hardi-
ness facet includes a commitment to ones social world.
Table 5.HRG distribution by age in the normative sample
1824
(N= 300)
2534
(N= 300)
3554
(N= 300)
5564
(N= 300)
65+
(N= 300)
MSDMSDMSDMSDMSDF(4, 1,490) pη
p2
Total hardiness 100.5 15.1 101 14.4 98.9 15.4 100.6 15.2 99 14.8 1.70 .148 .003
Challenge 100.7 14.7 101.5 14.5 99.2 15.2 100.0 15.4 98.7 15.1 2.98 .018 .004
Control 102.7 16.5 102.2 15.8 100.6 16.8 101.7 16.5 98.8 16.6 2.16 .077 .007
Commitment 99.1 15.4 100.1 14.7 98.6 15.4 101.3 15.0 101 14.4 0.62 .646 .005
Note. Guidelines for evaluating η
p2
are .01 = small, .06 = medium, .14 = large (Cohen, 1988).
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
228 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
Discussion
This first study documents the development and evaluation
of an improved hardiness scale, the HRG. Applying careful
item selection strategies, 28 items were retained for the final
scale, representing the three hardiness facets of commitment,
control, and challenge. The HRG shows the predicted hierar-
chical structure of three factors nested under a more general
hardiness factor and proved to be invariant across gender and
agegroups.InanormativesampleofN=1,500,whichwas
census-matched to the United States population, the new
scale shows superior reliability for total hardiness, as well as
for its subscales. The new scale also shows appropriate con-
struct validity, as assessed through factor analysis, and con-
vergent and divergent validity, as assessed through
correlations with other measures, including coping styles
and life satisfaction. This preliminary evidence is supported
by additional validation efforts using two independent, non-
US samples. This is described in detail in Study 2below.
Study 2: Cross-Validation of the
Hardiness Resilience Gauge on
Independent Samples
Two additional independent samples were obtained to fur-
ther evaluate the psychometric and structural properties of
the new hardiness scale and assess its validity.
Canadian Sample
Using an online research panel, a sample of N=394
respondents answered a web-based survey that included
the HRG and the DRS-15 (there were 5identical items
between the two measures that were administered only
once). Participants also completed a series of demographic
questions and additional measures (described below). All
respondents resided in Canada, and the sample was tar-
geted to contain an equal representation of males and
females. To ensure that participantsages contained suffi-
cient variability, the sample was evenly split between par-
ticipants who fell within the 1844 age range and
participants who were above 45 years of age. Respondents
were omitted if they violated at least two of the following
criteria: (1) Excessively short (< 5min) or long survey dura-
tions (> 40 min), (2) A large number of repeated consecu-
tive responses to the survey questions (> 61% of the scale
for the HRG and 40% for the DRS-15), (3) Large summed
absolute differences (defined as > 2standard deviations
above the mean) between forward and reverse scored
items for both the HRG and DRS-15,and(4) Being identi-
fied as a multivariate outlier as assessed through Maha-
lanobis Distance. Participants were also excluded if they
had incomplete responses to either the DRS-15 or HRG.
This resulted in the exclusion of 31 participants, leaving a
final sample size of 363 participants. Of these, 69.7% iden-
tified as White, 21.4% as Asian, 2.8%asBlack,and6.1%as
other. In terms of education, 20.1%werehighschoolgrad-
uates, 35.4% some college, 32% bachelors degree, and
9.9% graduate school. 63.4%w
ereemployedfullorpart-
time, 24%retired,and10.2%unemployed.
International Sample
An anonymized dataset containing N=4,994 was provided
by MHS Multi-Health Systems (the HRG publisher) from
their data archive. Respondents had completed the HRG
and several other measures as part of a professional devel-
opment activity or research project. Although a detailed
breakdown of nationality is not available for this sample,
we do know that respondents came from multiple coun-
tries, including Canada, Chile, China, Estonia, Finland,
Germany, Greece, Hong Kong, Ireland, Korea, Lithuania,
South Africa, Thailand, United Kingdom, and the United
States of America. All participants completed the assess-
ment in English. In addition to the HRG, participants were
asked a series of optional questions on demographics and
work-related outcomes. Excluded from the sample were
any respondents who did not provide demographic infor-
mation (e.g., gender, occupation) or answer the work-
related outcome questions. This accounted for the majority
of those excluded. Participants were also omitted if they
had incomplete responses for the HRG. Of the N=1,313
remaining participants, 14 cases were removed for violating
at least two of the screening metrics defined above. The
final sample consisted of N=1,299 participants with com-
plete data on the HRG, demographics, and work-related
questions. Of these participants, 73.2% identified as female
and 26.8% identified as male.
Table 6.Three-week test-retest reliability for the HRG (N= 168)
Time 1 Time 2
N M SD M SD CohensdCorrelation between HRG Time 1 and Time 2
Total hardiness 168 101.1 18.7 103.7 17.6 0.14 .81
Challenge 168 101.7 18.9 104.2 18.2 0.13 .80
Control 168 99.8 18.1 102.7 17.1 0.16 .74
Commitment 168 102.7 18.0 104.1 16.6 0.08 79
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 229
Measures
Canadian Sample
General anxiety was measured with the GAD-7(General
Anxiety Disorder-7; Spitzer et al., 2006). A 7-item scale,
the authors report a Cronbachsαof .92 and test-retest reli-
ability of .83. Earlier research on hardiness and anxiety has
shown that hardiness is associated with lower levels of anx-
iety (Kovács & Borcsa, 2017; Kowalski & Schermer, 2019).
Thus, we expected that people scoring higher on the HRG
would report lower levels of anxiety. In the present sample,
Cronbachsαfor the GAD-7was .94.
Depression was measured with the 9-item PHQ-9scale
(Patient Health Questionnaire-9; Kroenke et al., 2001).
The authors report Cronbachsαranging from .86 to .89.
Consistent with previous research on hardiness and depres-
sion, it was expected that people who score higher on the
HRG would report lower levels of depression (Bartone &
Homish, 2020; Maddi et al., 2006; Ng & Lee, 2020;Sinha
&Singh,2009). In the present sample, Cronbachsαfor the
PHQ-9was .92.
Coping in this sample was measured with 5scales drawn
from the Carver COPE Inventory (Carver, 1997). These are
short 2-item scales, with reported Cronbachsαcoefficients
of .68 for Active Coping, .64 for Positive Reframing, .54 for
Denial, .65 for Behavioral Disengagement, and .90 for Sub-
stance Abuse. In the present sample, Kendalls tau coeffi-
cients are reported as well as Cronbachsα, because the
tau coefficient often provides a more accurate estimate of
reliability for 2-item scales (Eisinga et al., 2013). Cronbachs
αand Kendalls tau coefficients (in parentheses) were .80
(.63) for Active Coping, .80 (.60) for Positive Reframing,
.46 (.31)forDenial,.79 (.68) for Behavioral Disengagement,
and .91 (.83) for Substance Abuse. Because hardy people
tend to use more effective coping strategies in dealing with
stress, it was expected that people who score higher on the
HRG would also score higher on the Active Coping and Pos-
itive Reframing subscales, and lower on Denial, Behavioral
Disengagement, and Alcohol/Substance Abuse subscales.
International Sample
Stress appraisal was measured with five items. Three of the
items related to positive stress appraisal, asking participants
to indicate how strongly they believed stress is something
that is useful and beneficial. An example item is: There
are benefits to experiencing stress.Two of the items
related to negative stress appraisal, asking participants to
indicate how strongly they believed stress is something to
be avoided and that stress prevents them from doing the
things that they want to do. Responses were on a 5-point
scale, ranging from Strongly Disagree to Strongly Agree.
The five questions were summed to create an aggregate
measure (both scales for the negative appraisal questions
were reversed). Cronbachsαwas acceptable at α=.74.
In line with previous research, we hypothesized that people
scoring higher on hardiness would be more likely to see
stress as potentially beneficial and not something that is
negative and to be avoided (Florian et a, 1995;Sko-
morovsky & Sudom, 2011; Stein & Bartone, 2020).
Work performance was measured with two items: Iam
engaged at work,and Iamsuccessfulinmyjob.
Responses were on a 5-point Likert scale, ranging from
Strongly Disagree to Strongly Agree. The two questions were
summed to create an aggregate measure. Cronbachsα
(α=.70) was acceptable given the 2-item nature of the scale,
while Kendallstau(τ=.36) indicated a moderate correlation
between the two measures. It was hypothesized that higher
hardinessscoreswouldbeassociated with greater work per-
formance (i.e., higher engagement and satisfaction).
Data Analysis and Statistics
The same statistical procedures described in Study 1above
for the normative sample were also applied with the cross-
validation samples in order to further assess the reliability
and validity of the new HRG hardiness scale.
Results
Measurement Invariance
The earlier result showing measurement invariance for gen-
der in the normative sample was confirmed in the Interna-
tional cross-validation sample. That is, males and females
were invariant in terms of how they responded on the
HRG (i.e., the conditions for strong invariance were met).
These results are presented in Table 7(middle portion).
As reported earlier, HRG responses were found to be invari-
ant across age groups in the Normative sample (Table 7,
bottom portion). Due to the relatively small sample size,
we were unable to test measurement invariance for gender
or age in the Canadian sample, and age information was
not available in the International sample.
Reliability
Cronbachsαvalues were obtained in both cross-validation
samples. For the Canadian sample, αvalues were .95 for
total hardiness, .91 for commitment, .86 for control, and
.86 for challenge. For the International sample, these values
were .91 for total hardiness, .87 for commitment, .77 for
control, and .82 for challenge. Examination of the standard-
ized factor loadings, as well as fitting a CFA model in which
factor loadings were constrained to be equal across items,
showed that the assumption of tau equivalence was not
met for HRG scales in both cross-validation samples. As
with the Study 1normative sample, McDonaldsωwas con-
sidered to be the more appropriate reliability coefficient for
these samples. For comparison purposes, Table 8displays
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
230 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
both the McDonaldsωand Cronbachsαcoefficients for
the HRG and its subscales in all three samples.
Validity
Factor Structure
Once again, the factor structure of the HRG was expected
to parallel the hierarchical model demonstrated in the
(Study 1) normative sample, in which the three dimensions
of challenge, control, and commitment are nested under
the higher-order construct of hardiness. The same CFA pro-
ceduresusedwiththenormativesamplewerealsoapplied
for both cross-validation samples in order to verify the best-
fitting model. Results confirmed that the 3-factor hierarchi-
cal model showed the best fit in both samples, although
goodness of fit indices were slightly below the acceptable
threshold for the International sample. For the Canadian
sample, the fit indices for the three-factor hierarchical
model were: CFI = .93,TLI=.92,RMSEA=.09 and SRMR
=.07. For the International sample, the fit indices were: CFI
=.89,TLI=.88,RMSEA=.08 and SRMR = .07.These
results are summarized in Table 9.
Relationships to Other constructs
Validation Samples
Canadian Sample
Hardiness Resilience Gauge and Generalized Anxiety.As
expected, total hardiness was negatively correlated with
Generalized Anxiety Disorder scores as assessed through
the GAD-7,r=.38,p<.001. Thus, people who are higher
in hardiness have lower self-reported scores of Generalized
Anxiety Disorder. All subscales of the HRG were signifi-
cantly negatively associated with GAD-7scores, with the
commitment scale showing the strongest negative correla-
tion at r=.39,p<.001.
Hardiness Resilience Gauge and Depression.Also,as
expected, total hardiness was negatively correlated with
depression scores on the PHQ-9,r=.38,p<.001.
Respondents who are higher in hardiness also report fewer
depression symptoms. All subscales of the HRG were signif-
icantly negatively correlated with PHQ-9Depression
scores, with the commitment scale again showing the stron-
gest negative correlation, r=.41,p<.001.
Hardiness Resilience Gauge and Coping.TheHRGandfive
coping styles were examined. As expected, total hardiness
was positively correlated with both Active Coping, r=.55,
p<.001, and Positive Reframing, r=.51,p<.001.Also
as expected, total hardiness was negatively correlated with
Behavioral Disengagement, r=.16,p<.01 and Substance
Use, r=.11,p<.05, although the latter relationships were
weaker than for the positive coping indicators. Total hardi-
ness score also showed a weak but positive correlation with
Denial, r=.11,p>.05.
International Sample
Hardiness Resilience Gauge and Stress Appraisal.Asexpected,
total hardiness scores were positively correlated with Stress
Appraisal, r=.43,p<.001. Thus, people who are more
hardy are more likely to interpret stress as something that
is potentially positive rather than uniformly negative. The
Table 7.Measurement invariance tests for gender and age groups
Model Scaled w
2
df CFI RMSEA Δw
2
Normative: Male vs. Female
Configural 4,551.10 694 .914 .086
Metric 4,590.10 722 .914 .085 Δw
2
(28) = 30.08, p> .05
Scalar 4,409.91 747 .918 .081 Δw
2
(25) = 32.11, p> .05
International: Male vs. Female
Configural 3,302.32 694 .901 .076
Metric 3,305.05 708 .902 .075 Δw
2
(14) = 15.94, p> .05
Scalar 3,194.66 733 .907 .072 Δw
2
(25) = 26.68, p> .05
Normative: All 5 age groups
Configural 5,383.50 1,735 .921 .084
Metric 5,469.28 1,843 .921 .081 Δw
2
(108) = 108.68, p> .05
Scalar 5,233.49 1,943 .929 .075 Δw
2
(100) = 106.65, p> .05
Note. The metric model includes equality constraints across thresholds, whereas the scalar model includes equality constraints across thresholds and
factor loadings (Svetina & Rutkowski, 2017). Chi square is Satorra-Bentler scaled chi square. Delta parameterization was used for all models.
Table 8.McDonaldsωand Cronbachsαreliability coefficients for the
HRG and its subscales
Commitment Control Challenge
Hardiness
total
Normative sample
McDonaldsω.92 .90 .89 .87
Cronbachsα.85 .84 .89 .93
Canadian sample
McDonaldsω.94 .91 .90 .91
Cronbachsα.91 .86 .86 .95
International sample
McDonaldsω.92 .86 .88 .85
Cronbachsα.87 .77 .82 .91
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 231
challenge subscale showed the strongest correlation with
Stress Appraisal, r=.45,p<.001.
Hardiness Resilience Gauge and Work Performance.Alsoas
expected, total hardiness scores were positively correlated
with work performance, r=.42,p<.001. Thus, people
who are more hardy are also more likely to report being
successful and more engaged in their jobs. The commit-
ment sub-scale showed the strongest correlation with Work
Performance, r=.45,p<.001.
For ease of reference, correlations between HRG scales
and all criterion indicators for all three samples are summa-
rizedinTable10.
Correlations With Earlier Hardiness Test
As an additional check on validity, we assessed the Pearson
correlation between the scales of the new HRG and the pre-
vious hardiness measure, the DRS-15, in the Canadian sam-
ple (N=363). HRG scores were first standardized to correct
the unequal number of items in the subscales. Results
showed the HRG correlated strongly with the DRS, for over-
all hardiness (r=.87), for the commitment scale (r=.84), for
the control scale (r=.91), and for the challenge scale (r=
.61).
Hardiness Resilience Gauge vs. Dispositional Resili-
ence Scale in Predicting Outcomes
The relationship between the HRG and various outcome
variables in the Canadian sample was explored while con-
trolling for DRS-15 scores. The HRG should predict the out-
comesaboveandbeyondtheDRS-15 if it is to offer
enhanced predictive utility over the original assessment.
In other words, does the HRG show incremental predictive
validity beyond what the previous measure accounts for?
For these analyses, five highly similar items between the
HRG and DRS-15 were removed from the HRG raw score
in order to reduce potential multicollinearity between the
scales. Hierarchical linear regressions were performed
using the stats package (version 3.6.2) in R (R Core Team,
2021), 7regressions with the total raw scores for both the
HRG and DRS-15,and7regressions with the raw scores
for just the challenge subscale of both instruments. The
challenge scale was of special interest because that is where
the greatest differences exist between the DRS-15 and the
HRG. For all regressions, DRS-15 raw scores were entered
in the first block, and HRG raw scores were entered in
the second block. The change in R
2
was examined.
For total hardiness, results showed that the HRG
accounted for significant unique variance in Positive
Reframing Coping, Active Coping, and Denial Coping.
For the challenge subscale, the HRG accounted for signifi-
cant unique variance in PHQ-9Depression, GAD-7Anxiety,
Positive Reframing Coping, Active Coping, and Denial Cop-
ing. HRG total scores did not account for significant unique
variance in PHQ-9,GAD-7, Behavioral Disengagement
Coping, or Substance Use. HRG challenge likewise did
not account for significant unique variance in either the
Behavioral Disengagement Coping or Substance Use.
Table 11 summarizes these results.
Table 9.Goodness of fit indices for three HRG factor models for the Canadian and International samples
One-factor Three-factor orthogonal model Three-factor hierarchical model
Canadian sample
Number of parameters 111 111 114
Scaled w
2
1,805.31 9,810.07 1,324.77
df 350 350 347
p< .001 < .001 < .001
CFI .90 .33 .93
TLI .89 .28 .92
RMSEA .11 .27 .09
SRMR .08 .37 .07
International sample
Number of parameters 112 112 115
Scaled w
2
5,423.98 15,540.15 3,385.76
df 350 350 347
p< .001 < .001 < .001
CFI .81 .44 .89
TLI .80 .39 .88
RMSEA .11 .18 .08
SRMR .09 .28 .07
Note. CFI = Comparative Fit Index; df = Degree of Freedom; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Square
Residual; TLI = Tucker-Lewis Index. CFI, TLI and RMSEA values are robust estimates. Also, the Canadian sample had 1 less parameter estimate than the
International sample due to one less threshold parameter being estimated for an item (i.e., not all response options were endorsed for that item, so it was
collapsed with an adjacent category).
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
232 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
Discussion
In this Study 2, we explored the reliability and validity of the
HRG in two independent samples, one consisting of Cana-
dian nationals and the other made up of respondents from
multiple countries who completed the HRG as part of a pro-
fessional development activity or for a research study. As
with Study 1, results showed high-reliability coefficients in
both samples, as well as good evidence for validity. For
HRG total hardiness and the three subscales, Cronbachs
αranged from .77 to .95, whereas McDonaldsωranged
from .85 to .94. The HRG also showed measurement invari-
ance across gender, lending further support to its general
applicability. Factorial validity was also confirmed, with a
three-factor hierarchical model showing the best fit in both
independent samples. The HRG also showed predicted
associations with relevant outcome measures, including
anxiety, depression, positive and negative coping strategies,
stress appraisals, and work performance. Finally, the HRG
was found to account for significant unique variance in a
number of the outcome variables while controlling for the
DRS-15; this was especially true for the challenge subscale,
which underwent the most extensive revision in the devel-
opment of the HRG. Overall results from Study 2provide
strong support for the reliability and validity of the HRG
as a measure of psychological hardiness.
General Discussion
Since it was first reported by Kobasa (1979), multiple stud-
ies have shown that psychological hardiness is an important
factor influencing individual resilience, health, and human
performance under stress. Nevertheless, research and
applications involving hardiness have been somewhat ham-
pered by the lack of a fully reliable and comprehensive,
standardized tool for measuring it. This is due in part to
the complex nature of the hardiness construct, which con-
tains multiple facets that are themselves multidimensional.
The present work addresses this gap, providing a hardiness
measure with increased reliability, shown in terms of inter-
nal consistency (Cronbachsα=.93;McDonaldsω=.87)
and 3-week test-retest reliability (r=.81). Reliability coeffi-
cients for the three hardiness facets of commitment, con-
trol, and challenge, are also uniformly high. Further, the
HRG shows enhanced validity, with better content area
coverage for the hardiness facets of challenge, control,
and commitment. In particular, challenge has been poorly
measured in earlier hardiness tests, where items have
focused almost exclusively on attitudes toward variety
and novelty, and neglecting other theoretically important
aspects such as the tendency to regard change and disrup-
tion as interesting life challenges and opportunities to learn.
The new HRG challenge scale incorporates these additional
components of challenge, thus providing a more construct-
valid indicator of this important hardiness dimension.
The HRG hardiness measure also shows the expected
hierarchical structure of three distinct factors nested under
a general hardiness dimension. This supports the theoreti-
cal notion that the three factors work together, creating a
whole that is more than the sum of its parts. For example,
having a strong sense of control the belief that one can
influence outcomes can reinforce and encourage the
sense of challenge or willingness to take risks and try new
things. Likewise, having a strong sense of purpose or mean-
ing in life can encourage and facilitate a sense of control
Table 10.Correlations between the HRG hardiness scales and various criterion indicators for all three samples
Commitment Control Challenge Hardiness total
Normative sample
Task-oriented coping (N= 1,471) .61*** .55*** .62*** .68***
Emotion-oriented coping (N= 1,473) .31*** .22*** .26*** .30***
Life satisfaction (N= 1,500) .56*** .39*** .33*** .48***
Relationship satisfaction (N= 1,463) .52*** .38*** .32*** .46***
Canadian sample
Depression PHQ-9 (N= 346) .41*** .35** .26*** .38***
Anxiety GAD-7 (N= 351) .39*** .33*** .32*** .38***
Active coping (N= 353) .51*** .46*** .51*** .55***
Positive reframing (N= 353) .48*** .38*** .52*** .51***
Denial (N= 353) .12* .03 .22*** .11*
Behavioral disengagement (N= 353) .14** .19*** .08 .16**
Substance use (N= 352) .13* .12* .06 .11*
International sample
Stress appraisal (N= 1,270) .31*** .35*** .45*** .43***
Work performance (N= 1,260) .45*** .30*** .35*** .42***
Note.*p< .05; **p< .01; ***p< .001.
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 233
over ones own destiny. Past research applications have
fruitfully examined the influence of overall hardiness as
well as the three facets separately, although results with
the facets are sometimes ambiguous (Eid et al., 2008; John-
sen et al., 2009). And while the three facets generally tend
to intercorrelate and go up or down together, this may not
always be true. In some cases, people can be high in certain
hardiness facets and low in others. For example, in their
study of Norwegian Army soldiers deployed to Kosovo,
Johnsen et al. (2014) found that while for most subjects
their hardiness facet scores were internally consistent (all
high, medium, or low), some subjects showed deviant or
unbalanced profiles. The largest deviant group was the
Rigid Controls,who were low in hardiness challenge
despite being high in commitment and control. Also identi-
fied were Sensation-seekerswhowerehighinchallenge
but low in commitment and control. In this study, soldiers
with unbalanced Rigid-Controlhardiness profiles also
reported more health problems, similar to the low hardy
group. In another study that looked at hardiness pro-
files, Norwegian navy cadets with unbalanced hardiness
profiles commitment, control, and challenge scores not
in agreement also showed more extreme and unhealthy
physiological stress reactions (Sandvik et al., 2013). With
the improved hardiness facet scales in the HRG, similar
investigations into the influence of the separate hardiness
facets are more feasible and should yield more clear and
valid results.
One criticism of earlier hardiness measures is that the
tests contained a male bias and did not apply equally well
to women (Riska, 2002;Klag&Bradley,2004). This is
an issue of measurement equivalence or invariance. Mea-
surement invariance across gender has been previously
shown for the DRS hardiness scale but not for all items
Table 11.Regressions with both HRG and DRS hardiness measures predicting various outcome indicators
Test statistic pR
2
ΔF
Total HRG and DRS
PHQ9 DRS F(1, 344) = 115.30 < .001 .25 ΔF(1) = 1.04, ns
PHQ9 HRG + DRS F(2, 343) = 58.19 < .001 .25
GAD7 DRS F(1, 349) = 106.40 < .001 .23 ΔF(1) = 0.17, ns
GAD7 HRG + DRS F(2, 348) = 53.15 < .001 .23
AC DRS F(1, 351) = 103.40 < .001 .23 ΔF(1) = 36.08, p< .001
AC HRG + DRS F(2, 350) = 74.90 < .001 .30
PR DRS F(1, 351) = 80.13 < .001 .19 ΔF(1) = 35.85, p< .001
PR HRG + DRS F(2, 350) = 61.97 < .001 .26
DDRS F(1, 351) = 0.0006 ns .00 ΔF(1) = 16.88, p< .001
DHRG + DRS F(2, 350) = 8.44 < .001 .05
BD DRS F(1, 351) = 18.87 < .001 .05 ΔF(1) = 0.95, ns
BD HRG + DRS F(2, 350) = 9.91 < .001 .05
SU DRS F(1, 350) = 12.61 < .001 .03 ΔF(1) = 1.45, ns
SU HRG + DRS F(2, 349) = 7.04 < .001 .04
Challenge subscale for HRG and DRS
PHQ9 DRS F(1, 344) = 13.75 < .001 .04 ΔF(1) = 11.53, p< .001
PHQ9 HRG + DRS F(2, 343) = 12.85 < .001 .07
GAD7 DRS F(1, 349) = 25.07 < .001 .07 ΔF(1) = 16.56, p< .001
GAD7 HRG + DRS F(2, 348) = 21.37 < .001 .11
AC DRS F(1, 351) = 19.05 < .001 .05 ΔF(1) = 106.09, p< .001
AC HRG + DRS F(2, 350) = 65.42 < .001 .27
PR DRS F(1, 351) = 16.11 < .001 .04 ΔF(1) = 106.09, p< .001
PR HRG + DRS F(2, 350) = 65.42 < .001 .27
DDRS F(1, 351) = 3.24 ns .01 ΔF(1) = 2.02, ns
DHRG + DRS F(2, 350) = 9.37 < .001 .05
BD DRS F(1, 351) = 0.53 ns .001 ΔF(1) = 15.37, p< .001
BD HRG + DRS F(2, 350) = 1.28 ns .01
SU DRS F(1, 350) = 1.23 ns .004 ΔF(1) = 0.24, ns
SU HRG + DRS F(2, 349) = 0.74 ns .004
Note. AC = Active Coping; BD = Behavioural Disengagement; D = Denial; DRS = Dispositional Resilience Scale; GAD7 = General Anxiety Disorder-7 Scale;
HRG = Hardiness Resilience Gauge; PHQ9 = Patient Health Questionnaire-9; PR = Positive Reframing; SU = Substance Use. The tilde () symbol indicates
that the variable on the left was regressed on the right-hand side variable(s).
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
234 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
(Kardum et al., 2012). Hystad (2012) demonstrated gender
equivalence in the DRS-15 but noted some non-equivalence
in the control scale. The new HRG hardiness scale is an
advance in this regard, with demonstrated measurement
invariance across two large and independent samples.
Another important consideration is if measures such as
the HRG are invariant or hold equally well across age
groups. Our results show the HRG displays measurement
invariance across the broad age range of 1865 years.
Future studies will be needed to determine the validity of
the HRG in younger age groups.
In both studies and all three of the present samples, the
HRG showed expected associations with relevant criterion
indicators. Hardiness was negatively associated with anxi-
ety (PHQ-9) and depression (GAD-7) scores. Previous stud-
ies have also linked hardiness to lower anxiety. For
example, in a study of university undergraduates, Kowalski
and Schermer (2019) found that hardiness was negatively
correlated with anxiety, even after controlling for neuroti-
cism. It is worth noting that PTSD is also an anxiety disor-
der, and there are multiple studies showing that hardiness
isaprotectivefactoragainstPTSD(e.g.,Bartone,1999;
Escolas et al., 2013; Thomassen et al., 2018).
Consistent with previous studies, the present research
also finds fairly strong associations of HRG hardiness
scores with depression. For example, Maddi et al. (2006)
found that hardiness was negatively related to depression
in a sample of US Army War College students. Other stud-
ies have found a link between hardiness and depression in a
variety of samples (Bartone & Homish, 2020;Ganellen&
Blaney, 1984; Ng & Lee, 2020). HRG hardiness scores also
showed the expected positive association with measures of
well-being, life satisfaction, and satisfaction with relation-
ships. The positive relation of hardiness to well-being has
also been frequently reported (e.g., Kowalski & Schermer,
2019;Bartone&Bowles,2021). In the present study, the
HRG subscale of commitment showed the strongest corre-
lations with relationship satisfaction and life satisfaction.
This makes sense considering that hardiness commitment
involves a strong engagement with the social world and a
sense of passion and deep involvement in life (Kobasa &
Maddi, 1977; Stein & Bartone, 2020).
People high in hardiness tend to rely on more adaptive
coping styles that involve taking action to address the
source of their stress (Eschleman et al., 2010). Likewise,
they tend not to use avoidance or emotional coping strate-
gies that are not aimed at solving problems. Therefore, we
expected that higher scores on the HRG would be related to
more adaptive coping styles. This was largely the case with
HRG hardiness scores. In Study 1, hardiness correlated
quite strongly with task-oriented coping and negatively with
emotion-oriented coping, and these results were consistent
across the HRG subscales (Table 10). This trend also holds
in Study 2with the Canadian sample, where HRG hardiness
was positively associated with active coping and positive
reframing, while negatively correlated with behavioral dis-
engagement and alcohol/substance abuse.
A somewhat surprising finding is that the HRG challenge
scale shows a small but positive correlation with denial cop-
ing. This may reflect the tendency of high challenge people
to welcome change and disregard risk to some extent when
facing uncertain situations. Just such a situation is pre-
sented by the COVID pandemic, which was a dominant
concern when the present data were being collected in early
2021. It is also the case that when facing stressful situations,
especially those outside of our control, some amount of
denial can be adaptive. For example, Horowitz (2001)has
described the normal, healthy grief process following the
loss of a loved one as involving periods of denial, when
one puts the loss out of mind in order to go on with life
and accomplish necessary tasks, and periods of intrusion
when one is fully aware and experiencing the grief pain
of loss. Similarly, when facing stressors like an infectious
disease pandemic that is largely outside of individual con-
trol, a tendency to minimize the threat and get on with
the tasks of living could be equally adaptive. Still, the weight
of the evidence in the present studies shows that HRG har-
diness is associated with positive problem-solving coping
strategies, and negatively associated with avoidance coping
styles. Future research is needed to identify under what
conditions different coping styles are most effective in deal-
ing with stressful situations.
Study 2also showed, in a large International sample, that
persons high in HRG hardiness tend to make more positive
stress appraisals, seeing potential benefits even in difficult
situations. These individuals also rate themselves as more
engaged and successful in their work. While hardiness the-
ory clearly indicates that hardiness should facilitate growth
from stress, there is to date only modest empirical support
for this. One recent study addressing this issue examined
US Army recruiters, who face considerable pressure in their
jobs to meet monthly production quotas (Bartone & Bowles,
2021). The authors report that hardiness in a large national
sample of recruiters was related to post-traumatic growth
and increased well-being. Further study is needed to
address the question of when stress exposure can lead to
actual growth and improvement and what role hardiness
may play in this process.
In comparing the new HRG hardiness scale with the pre-
vious short DRS-15 version, the correlations obtained here
were all quite high (.84 or above), with the exception of
the challenge scale, which was .61. This was expected con-
sidering that the previous challenge scale had weak con-
struct validity, which the new scale attempts to correct
with better coverage of the challenge dimension. For exam-
ple, the old challenge scale did not address the aspect of
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 235
learning from experience and failures, which the HRG chal-
lenge scale now includes. So the lower correlation between
the HRG and DRS-15 challengescalesisexpected,and
reflects the improved construct validity of the new scale.
A final important question is: does the new HRG hardi-
ness measure provide incremental predictive validity over
the earlier version? Based on a series of regression analyses
performed with the Canadian sample, the answer appears
to be yes. The HRG scale showed significantly increased
power to predict active, problem-solving coping and posi-
tive reframing. The HRG challenge scale, in particular,
showed incremental validity in predicting anxiety, depres-
sion, and positive and negative coping approaches. The
HRG thus shows improved predictive power over the most
recent earlier hardiness test.
Limitations
Several limitations of this research should be mentioned. In
assessing predictive validity, the present study relies on
cross-sectional data in which relevant outcome indicators
were collected at the same time point as HRG hardiness
scores. A preferred design would be a longitudinal one in
which outcome measures are collected at a later point in
time, thus allowing for more clear inferences regarding cau-
sal directionality.
Another potential limitation in the present work concerns
our reliance on an online survey methodology using
Mechanical Turk (MTurk). While this methodology has
become quite common in the social sciences, some ques-
tions have been raised about the representativeness of sam-
ples obtained in this way. For example, some researchers
have pointed out that MTurk samples tend to be younger,
more unemployed, and more depressed (Ophir et al.,
2020; Paolacci & Chandler, 2014). However, these same
studies indicate that these limitations can be largely over-
come by the use of careful screening methods. In the pre-
sent study, we took steps to exclude any respondents with
mental health problems, and any who showed spurious
response patterns or lack of attention to the task. Further,
in the normative sample (Study 1), we deliberately sampled
particular demographic groups for age, gender, employ-
ment status, education, and geographic region, thus
providing greater assurance regarding generalizability of
results.
In Study 2, the anonymized nature of the International
sample data imposed some limitations due to the lack of
demographic data. Lacking information on age, we were
unable to test for HRG measurement invariance across
age groups. In addition, because all participants in the Inter-
national sample completed the survey in English, and we do
not know if English was their primary language, it is possi-
ble that there were differences in English proficiency within
this sample which could have impacted how they
responded to the questionnaire. The relatively small size
of the Canadian sample also prevented testing for measure-
ment invariance across age. However, the HRG did prove
to be invariant across age groups in the larger, normative
sample. Future studies should explore this issue further in
other groups, and also assess the applicability of the HRG
in teenagers and adolescents.
Conclusion
While hardiness has generally been conceived as a fairly
stable trait, there is evidence that it can also be increased
through training programs and life experiences (Bartone,
Eid, et al., 2016; Stein & Bartone, 2020). Considering the
growing interest in programs that aim specifically to
increase the stress resilience qualities associated with hardi-
ness, it is all the more important that a highly valid and reli-
able measure of hardiness be available to researchers and
practitioners. The HRG was developed to help meet this
need.
The HRG is fairly brief at 28 items and easy to adminis-
ter. It was designed to be easily adaptable for use in other
languages and free of cultural bias and idiomatic expres-
sions. The HRG shows appropriate convergent and discrim-
inant validity, as well as a good fitting hierarchical model.
Likewise, it demonstrates good generalizability across mul-
tiple and diverse samples and is valid for women and men
and different age groups. The HRG thus provides an impor-
tant advance in measuring the hardiness construct, and
should prove to be a useful tool in research, clinical and
consulting settings.
References
Andrew, M. E., Mnatsakanova, A., Howsare, J. L., Hartley, T. A.,
Charles, L. E., Burchfiel, C. M., McCanlies, E. C., & Violanti,
J. M. (2013). Associations between protective factors and
psychological distress vary by gender: The Buffalo Cardio-
Metabolic Occupational Police Stress Study. International
Journal of Emergency Mental Health, 15(4), 277288.
Bartone, P. T. (1984). Stress and health in Chicago Transit
Authority bus drivers (Unpublished doctoral dissertation).
Department of Behavioral Sciences, University of Chicago.
Bartone, P. T. (1989). Predictors of stress-related illness in city
bus drivers. Journal of Occupational Medicine, 31, 657663.
https://doi.org/10.1097/00043764-198908000-00008
Bartone, P. T. (1991, June). Development and validation of a short
hardiness measure. Paper presented at the American Psycho-
logical Society Annual Convention, Washington, DC. http://
www.hardiness-resilience.com/docs/aps91b.pdf
Bartone, P. T. (1995, June). A Short Hardiness Scale. Poster
presented at the American Psychological Society Annual
Convention, New York, NY. https://www.researchgate.net/pub-
lication/331875947_A_Short_Hardiness_Scale_updated
European Journal of Psychological Assessment (2023), 39(3), 222239 Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
236 P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure
Bartone, P. T. (1999). Hardiness protects against war-related
stress in Army reserve forces. Consulting Psychology Journal,
51,7282. https://doi.org/10.1037/1061-4087.51.2.72
Bartone, P. T. (2006). Resilience under military operational stress:
Can leaders influence hardiness? Military Psychology, 18(Suppl.),
S131S148. https://doi.org/10.1207/s15327876mp1803s_10
Bartone, P. T. (2013). Cross-cultural adaptation of the DRS-15
Dispositional Resilience Scale: A short hardiness measure. Final
report on Fulbright Research Fellowship. http://www.hardi-
ness-resilience.com/docs/Crosscultural adaptation of DRS
Fulbright.pdf
Bartone, P. T., & Bowles, S. V. (2021). Hardiness predicts post-
traumatic growth and well-being in severely wounded service-
men and their spouses. Military Medicine, 186(56), 500504.
https://doi.org/10.1093/milmed/usaa250
Bartone, P. T., Eid, J., & Hystad, S. W. (2016). Training hardiness
for stress resilience. In N. Maheshwari & V. Kumar (Eds.),
Military psychology: Concepts, trends and interventions (pp.
231248). Sage.
Bartone, P. T., Eid, J., Hystad, S. W., Jocoy, K., Laberg, J. C., &
Johnsen, B. H. (2015). Psychological hardiness and avoidance
coping are related to risky alcohol use in returning combat
veterans. Military Behavioral Health, 3, 274282. https://doi.
org/10.1080/21635781.2015.1085931
Bartone, P. T., Eid, J., Johnsen, B. H., Laberg, J. C., & Snook, S. A.
(2009). Big Five personality factors, hardiness, and social
judgment as predictors of leader performance. Leadership &
Organization Development Journal, 30(6), 498521. https://doi.
org/10.1108/01437730910981908
Bartone, P. T., & Homish, G. G. (2020). Influence of hardiness,
avoidance coping, and combat exposure on depression in
returning war veterans: A moderated-mediation study. Journal
of Affective Disorders, 265, 511518. https://doi.org/10.1016/
j.jad.2020.01.127
Bartone, P. T., Roland, R. R., Picano, J. J., & Williams, T. J. (2008).
Personality hardiness predicts success in US Army Special
Forces candidates. International Journal of Selection and
Assessment, 16,7881. https://doi.org/10.1111/j.1468-2389.
2008.00412.x
Bartone, P. T., Ursano, R. J., Wright, K. M., & Ingraham, L. H.
(1989). The impact of a military air disaster on the health of
assistance workers: A prospective study. Journal of Nervous
and Mental Disease, 177(6), 317328. https://doi.org/10.1097/
00005053-198906000-00001
Bartone, P. T., Valdes, J. J., & Sandvik, A. (2016). Psychological
hardiness predicts cardiovascular health. Psychology, Health &
Medicine, 21, 743749. https://doi.org/10.1080/13548506.
2015.1120323
Bentler, P. M. (1990). Comparative fit indexes in structural
models. Psychological Bulletin, 107, 238246. https://doi.org/
10.1037/0033-2909.107.2.238
Britt, T. W., Adler, A. B., & Bartone, P. T. (2001). Deriving benefits
from stressful events: The role of engagement in meaningful
work and hardiness. Journal of Occupational Health Psychology,
6,5363. https://doi.org/10.1037/1076-8998.6.1.53
Brown, T. A. (2015). Confirmatory factor analysis for applied
research (2nd ed.). The Guilford Press.
Carver, C. S. (1997). You want to measure coping but your
protocols too long: Consider the brief COPE. International
Journal of Behavioral Medicine, 4(1), 92100. https://doi.org/
10.1207/s15327558ijbm0401_6
Chalmers, R. P. (2012). mirt: A multidimensional item response
theory package for the R environment. Journal of Statistical
Software, 48(6), 129. https://doi.org/10.18637/jss.v048.i06
Choi, S. W., Gibbons, L. E., & Crane, P. K. (2011). lordif: An R
package for detecting differential item functioning using
iterative hybrid ordinal logistic regression/item response theory
and monte carlo simulations. Journal of Statistical Software,
39(8), 130. https://doi.org/10.18637/jss.v039.i08
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences (1st ed.). Erlbaum.
Contrada, R. J. (1989). Type A behavior, personality hardiness, and
cardiovascular responses to stress. Journal of Personality and
Social Psychology, 57, 895903. https://doi.org/10.1037/0022-
3514.57.5.895
Cronbach, L. J. (1951). Coefficient alpha and the internal structure
of tests. Psychometrika, 16, 297335. https://doi.org/10.1007/
BF02310555
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The
Satisfaction With Life Scale. Journal of Personality Assessment,
49(1), 7175. https://doi.org/10.1207/s15327752jpa4901_13
Eid, J., Johnsen, B. H., Bartone, P. T., & Nissestad, O. A. (2008).
Growing transformational leaders: Exploring the role of person-
ality hardiness. Leadership & Organization Development Journal,
29(1), 423. https://doi.org/10.1108/01437730810845270
Eisinga, R., Grotenhuis, M., & Pelzer, B. (2013). The reliability of a
two-item scale: Pearson, Cronbach, or Spearman-Brown?
International Journal of Public Health, 58(4), 637642. https://
doi.org/10.1007/s00038-012-0416-3
Endler, N. S., & Parker, J. D. A. (1994). Assessment of multidi-
mensional coping: Task, emotion, and avoidance strategies.
Psychological Assessment, 6(1), 5060. https://doi.org/
10.1037/1040-3590.6.1.50
Eschleman, K. J., Bowling, N. A., & Alarcon, G. M. (2010). A meta-
analytic examination of hardiness. International Journal of
Stress Management, 17(4), 277307. https://doi.org/10.1037/
a0020476
Escolas, S. M., Pitts, B. L., Safer, M. A., & Bartone, P. T. (2013).
The protective value of hardiness on military posttraumatic
stress symptoms. Military Psychology, 25(2), 116123. https://
doi.org/10.1037/h0094953
Florian, V., Mikulincer, M., & Taubman, O. (1995). Does hardiness
contribute to mental health during a stressful real-life situa-
tion? The roles of appraisal and coping. Journal of Personality
and Social Psychology, 68, 687695. https://doi.org/10.1037/
0022-3514.68.4.687
Funk, S. C. (1992). Hardiness: A review of theory and research.
Health Psychology, 11, 335345. https://doi.org/10.1037/0278-
6133.11.5.335
Fülöp, F., Bőthe, B., Gál, E., Cachia, J., Demetrovics, Z., & Orosz, G.
(2020). A two-study validation of a single-item measure of
relationship satisfaction: RAS-1. Current Psychology. https://
doi.org/10.1007/s12144-020-00727-y
Ganellen, R. J., & Blaney, P. H. (1984). Hardiness and social
support as moderators of the effects of life stress. Journal of
Personality and Social Psychology, 47(1), 156163. https://doi.
org/10.1037/0022-3514.47.1.156
Goldammer, P., Annen, H., Stöckli, P. L., & Jonas, K. (2020).
Careless responding in questionnaire measures: Detection,
impact, and remedies. The Leadership Quarterly, 31(4), Article
101384. https://doi.org/10.1016/j.leaqua.2020.101384
Hayes, A. F., & Coutts, J. J. (2020). Use omega rather than
Cronbachs alpha for estimating reliability. But....Communi-
cation Methods and Measures, 14(1), 124. https://doi.org/
10.1080/19312458.2020.1718629
Horn, J. L., McArdle, J., & Mason, R. (1983). When is invariance not
invariant: A practical scientists look at the ethereal concept of
factor invariance. Southern Psychologist, 1(4), 179188.
Horowitz, M. J. (2001). Stress response syndromes: Personality
styles and intervention (4th ed.). Aronson.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in
covariance structure analysis: Conventional criteria versus new
Ó2022 The Author(s). Distributed as a Hogrefe OpenMind article European Journal of Psychological Assessment (2023), 39(3), 222239
under the license CC BY-ND 4.0 (https://creativecommons.org/licenses/by-nd/4.0)
P. T. Bartone et al., Development and Validation of an Improved Hardiness Measure 237
alternatives. Structural Equation Modeling, 6(1), 155. https://
doi.org/10.1080/10705519909540118
Hystad, S. W. (2012). Exploring gender equivalence and bias in a
measure of psychological hardiness. International Journal of
Psychological Studies, 4(4), 6979. https://doi.org/10.5539/ijps.
v4n4p69
Hystad, S. W., Eid, J., Johnsen, B. H., Laberg, J. C., & Bartone,
P. T. (2010). Psychometric properties of the Revised Norwegian
Dispositional Resilience (Hardiness) Scale. Scandinavian Jour-
nal of Psychology, 51, 237245. https://doi.org/10.1111/j.1467-
9450.2009.00759.x
Hystad, S. W., Eid, J., Laberg, J. C., Johnsen, B. H., & Bartone,
P. T. (2009). Academic stress and health: Exploring the mod-
erating role of personality hardiness. Scandinavian Journal of
Educational Research, 53(5), 421429. https://doi.org/10.1080/
00313830903180349
Hystad, S. W., Olsen, O. K., Espevik, R., & Säfvenbom, R. (2015). On
the stability of psychological hardiness: A three-year longitu-
dinal study. Military Psychology, 27, 155168. https://doi.org/
10.1037/mil0000069
Johnsen, B. H., Bartone, P., Sandvik, A. M., Gjeldnes, R., Morken,
A. M., Hystad, S. W., & Stornæs, A. (2013). Psychological hardi-
ness predicts success in a Norwegian Armed Forces border
patrol selection course. International Journal of Selection and
Assessment, 21, 368375. https://doi.org/10.1111/ijsa.12046
Johnsen, B. H., Eid, J., Pallesen, S., Bartone, P. T., & Nissestad,
O. A. (2009). Predicting transformational leadership in naval
cadets: Effects of personality hardiness and training. Journal of
Applied Social Psychology, 39(9), 22132235. https://doi.org/
10.1111/j.1559-1816.2009.00522.x
Johnsen, B. J., Hystad, S. W., Bartone, P. T., Laberg, J. C., & Eid, J.
(2014). Hardiness profiles: Defining clusters of the Disposi-
tional Resilience Scale and their relation to soldiershealth.
Military Behavioral Health, 2(2), 123128. https://doi.org/
10.1080/21635781.2013.867789
Jöreskog, K. G. (1971). Simultaneous factor analysis in several
populations. Psychometrika, 36(4), 409426. https://doi.org/
10.1007/BF02291366
Kardum, I., Hudek-Knežević, J., & Krapić, N. (2012). The structure
of hardiness, its measurement invariance across gender and
relationships with personality traits and mental health out-
comes. Psihologijske Teme, 21(3), 487507.
Klag, S., & Bradley, G. (2004). The role of hardiness in stress and
illness: An exploration of the effect of negative affectivity and
gender. British Journal of Health Psychology, 9, 137161.
https://doi.org/10.1348/135910704773891014
Kobasa, S. C. (1979). Personality and resistance to illness.
American Journal of Community Psychology, 7(4), 413423.
https://doi.org/10.1007/BF00894383
Kobasa, S. C., & Maddi, S. R. (1977). Existential personality theory.
In R. J. Corsini (Ed.), Current personality theories (pp. 243276).
Peacock.
Kobasa, S. C., & Puccetti, M. A. (1983). Personality and social resources
in stress resistance. Journal of Personality and Social Psychology,
44(4), 839850. https://doi.org/10.1037/0022-3514.45.4.839
Kovács, I. K., & Borcsa, M. (2017). The relationship between
anxiety, somatic symptoms and hardiness in adolescence.
Romanian Journal of Applied Psychology, 19(2), 4249.
Kowalski, C. M., & Schermer, J. A. (2019). Hardiness, perseverative
cognition, anxiety, and health-related outcomes: A case for and
against psychological hardiness. Psychological Reports, 122(6),
20962118. https://doi.org/10.1177/0033294118800444
Kroenke, K., Spitzer, R. L., & Williams, J. B. (2001). The PHQ-9:
Validity of a brief depression severity measure. Journal of
General Internal Medicine, 16(9), 606613. https://doi.org/
10.1046/j.1525-1497.2001.016009606.x
Lakens, D. (2013). Calculating and reporting effect sizes to
facilitate cumulative science: A practical primer for t-tests
and ANOVAs. Frontiers in Psychology, 4, Article 863. https://doi.
org/10.3389/fpsyg.2013.00863
Lambert,V.A.,Lambert,C.E.Jr.,Klipple,G.L.,&Mewshaw,E.A.
(1989). Social support, hardiness and psychological well-being in
womenwitharthritis.Image: The Journal of Nursing Scholarship,
21, 128131. https://doi.org/10.1111/j.1547-5069. 1989.tb00116.x
Maddi, S. R. (2002). The story of hardiness: Twenty </