Artigo 6

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Journal of Personality Assessment

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/hjpa20

Do Strengths Converge into Virtues? An Item-,


Virtue-, and Scale-Level Analysis of the Italian
Values in Action Inventory of Strengths-120

Tommaso Feraco, Nicole Casali & Chiara Meneghetti

To cite this article: Tommaso Feraco, Nicole Casali & Chiara Meneghetti (2021): Do
Strengths Converge into Virtues? An Item-, Virtue-, and Scale-Level Analysis of the Italian
Values in Action Inventory of Strengths-120, Journal of Personality Assessment, DOI:
10.1080/00223891.2021.1934481

To link to this article: https://doi.org/10.1080/00223891.2021.1934481

View supplementary material Published online: 06 Jul 2021.

Submit your article to this journal Article views: 33

View related articles View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=hjpa20
JOURNAL OF PERSONALITY ASSESSMENT
https://doi.org/10.1080/00223891.2021.1934481

Do Strengths Converge into Virtues? An Item-, Virtue-, and Scale-Level Analysis


of the Italian Values in Action Inventory of Strengths-120
Tommaso Feraco1,2, Nicole Casali1, and Chiara Meneghetti1
1
Department of General Psychology, University of Padova, Padua, Italy; 2Pentathlon Srl, Napoli, Italy

ABSTRACT ARTICLE HISTORY


The Values in Action Inventory of Strengths (VIA-IS) is a widely-used measure for character. Its fac- Received 23 February 2021
torial structure is still debated, however, and previous validation studies usually failed to examine Accepted 9 May 2021
the unidimensionality of the single character strengths. In addition, no studies to date have exam-
ined its Italian version. We validated the structure of the Italian short form of the VIA-IS in a sam-
ple of 16722 participants. Using confirmatory factor analysis and treating items as ordinal
variables, we followed three simple, but too often neglected, steps: we studied the unidimension-
ality of the single strengths first, then their convergence into second-order virtues, and then fitted
a hierarchical model that includes items, strengths and virtues, as originally proposed by Peterson
and Seligman. All strengths except “love of learning” were unidimensional, and both the virtues
and the final hierarchical models showed acceptable fit indices, unlike three models derived from
an exploratory factor analysis. The same findings emerged for a smaller sample of 1035 partici-
pants. Finally, both character strengths and virtues showed positive relations with general mental
health and negative relations with psychological distress. These results are discussed considering
previous studies on the factorial structure of the VIA-IS.

Introduction defined as “the psychological ingredients – processes or


mechanisms – that define the virtues” (Peterson & Seligman,
The conceptualization of character proposed by Peterson
2004, p. 13). Strengths of the wisdom and knowledge virtue,
and Seligman (2004) radically shaped the field of positive
for example, are distinct but they all have in common the
psychology (Niemiec, 2013), but its measurement is still
acquisition and the use of knowledge. For this reason, vir-
debated. The self-report questionnaire originally proposed
tues may be statistically conceptualized as second-order fac-
by Peterson and Seligman (2004) has often showed sub-opti-
tors describing the common variance shared by the
mal structural results, with previous validations mainly rely-
ing on reliability indexes only (i.e., Cronbach’s alpha) corresponding strengths (Ng et al., 2017). Strengths were
ignoring, for example, scales’ unidimensionality (Ng et al., derived from a literature review of positive psychological
2017). For these reasons, in this study we analyze the valid- traits. To guide their research, the authors initially identified
ity of the Italian version of the VIA-IS-120 adopting a dif- 7 criteria (that subsequently became 10, and now there are
ferent and more complete statistical approach that we 12) that character strengths must meet to qualify as such
believe can result in a more reliable version of the measure, (e.g., they must be measurable and trait-like, and contribute
that could be adopted in future studies using the instrument. to individual and others’ fulfilment). On the lowest level of
Peterson and Seligman (2004) define good character on two abstraction there are situational themes, which are specific
dimensions: a vertical one comprising three hierarchically- habits that enable people to display a certain strength in a
organized conceptual levels (virtues, character strengths, and certain situation, thus explaining the great deal of variation
situational themes); and a horizontal dimension, as each we can observe in individual behavior. Horizontally, the
conceptual level consists of several elements (six virtues, 24 high six are: wisdom and knowledge (knowledge used for
strengths). On the highest level of abstraction, there are six good); courage (willful exercise to reach goals despite adver-
core virtues, identified after extensive cross-cultural reviews sity); humanity (fairness in relations with others); justice
as the most highly valued moral characteristics in history (fairness in community life); temperance (control over
and philosophy. These virtues, also called “the high six”, are excess); and transcendence (belief in a greater meaning
defined as “core characteristics valued by moral philosophers beyond oneself). The character strengths include 24 positive
and religious thinkers” (Peterson & Seligman, 2004, p. 13) trait-like features connected to specific virtues: cognitive
and recur within and between different cultures. Character strengths (e.g., creativity, curiosity) relate to the virtue of
strengths lie on the intermediate level of abstraction and are wisdom and knowledge; emotional strengths (e.g., bravery,

CONTACT Tommaso Feraco tommaso.feraco@phd.unipd.it Department of General Psychology, University of Padova, via Venezia 8, Padova, Italy.
Supplemental data for this article is available online at https://doi.org/10.1080/00223891.2021.1934481
ß 2021 Taylor & Francis Group, LLC
2 FERACO, CASALI, MENEGHETTI

perseverance) to the virtue of courage; interpersonal strengths Table 1. The VIA classification. Adapted from Peterson and Seligman (2004,
pp. 29–30).
(e.g., love, kindness) to the virtue of humanity; civic strengths
Core virtues Character strengths
(e.g., teamwork, fairness) to the virtue of justice; strengths such
Wisdom & Creativity [originality, ingenuity]
as forgiveness and modesty to the virtue of temperance; and Knowledge Curiosity [interest, novelty-seeking, openness to experience]
strengths like appreciation of beauty and spirituality to the vir- Judgment [open-mindedness, critical thinking]
tue of transcendence. Table 1 shows the six core virtues and Love of learning [systematically adding knowledge]
Perspective [wisdom]
their related character strengths. After establishing their tax- Courage Bravery [valor]
onomy, both vertically and horizontally, Peterson and Perseverance [persistence, industriousness]
Seligman (2004) operationalized their model and developed a Honesty [authenticity, integrity]
Zest [vitality, enthusiasm, vigor, energy]
measure for strengths (the Values in Action Inventory of Humanity Love [closeness, intimacy]
Strengths, VIA-IS), which contains 240 positively-keyed items Kindness [generosity, nurturance, care, compassion,
(10 for each strength). As a last step, by means of factor ana- altruistic love]
Social intelligence [emotional intelligence, personal
lysis, they identified five factors that reflect, but do not per- intelligence]
fectly replicate the six core virtues, which they called Justice Teamwork [citizenship, social responsibility, loyalty]
emotional, intellectual, interpersonal, restraint, and theological. Fairness
Leadership
Nevertheless, the authors underline the importance of applying Temperance Forgiveness [mercy]
the explanatory power of the psychological analysis and meas- Humility [modesty]
urement to enrich the philosophical nature of their taxonomy Prudence [cautiousness]
Self-regulation [self-control]
(Peterson & Seligman, 2004, p. 13). Transcendence Appreciation of beauty and excellence [awe, wonder,
elevation]
Gratitude [thankfulness]
Factorial structure of the via-is: unidimensionality issues Hope [optimism, future-mindedness, future orientation]
Humor [playfulness]
Since the first factorial analysis on the VIA-IS (Peterson & Spirituality [religiousness, faith, purpose]
Seligman, 2004), its structure has been empirically investi-
gated time and again in validation or theoretical studies exploratory approaches, however. As we have shown, this
across different countries. We identified and reviewed 19 led to a plethora of models that risked being unreliable and
such publications (see Table 2) on the structure of the VIA- were consequently not adopted in subsequent studies. Each
IS, finding that they mainly relied on principal component new study involving the VIA-IS started with a new explora-
analysis (PCA) and principal axis factoring (PAF). None of tory analysis of the tool’s internal structure, basically ignor-
these studies reported the same results, however (see Table ing all the models that had gone before, generating, in our
S1 in the Supplemental materials). Each study yielded a dif- opinion, high confusion as well as low replicability of the
ferent model of the latent structure of the 24 strengths, with findings (e.g., Casali et al., 2020; Heintz & Ruch, 2020;
differences in the number of factors extracted, in the content Martınez-Martı & Ruch, 2017; Peterson et al., 2008; Petkari
of those factors (in terms of the strengths composing them), & Ortiz-Tallo, 2018; Weber et al., 2013). Interestingly, CFA
and even in the labels used to name the factors. Scholars was also very rarely used (e.g., Anjum & Amjad, 2020;
found evidence to support grouping the 24 strengths into McGrath, 2014; Ng et al., 2017) in direct efforts to replicate
three (e.g., Duan et al., 2012; McGrath, 2015; Seibel et al., the structure originally theorized by Peterson and Seligman
2015; Shryack et al., 2010), five (e.g., Aza~
nedo et al., 2017; (2004); and, when it was, the fit indices were poor and the
Peterson & Park, 2004; Ruch et al., 2010; Singh & Choubisa, authors opted to investigate other models. Even when alter-
2010) or, less frequently, four (Anjum & Amjad, 2020; Brdar native models were fitted using a data-driven approach
& Kashdan, 2010; Macdonald et al., 2008) or two factors (based on the results of EFAs), the fit indices of CFAs were
(Khumalo et al., 2008). As summarized graphically in Table always barely acceptable (see Table 2). In fact, this approach
S1 in the Supplemental materials, an intellectual second- leads to a significant loss in terms of theory, since any asser-
order factor usually comprising strengths such as creativity, tion made on the instrument and its relations relies on cur-
curiosity or love of learning emerged in several studies. On rent data rather than theoretically-driven hypotheses and
the other hand, fairness, forgiveness, kindness, leadership interpretations, making it hard to generalize results and rep-
and teamwork often converged into an “interpersonal” licate findings. Instead, authors have to completely rethink
second-order factor. A third typical second-order factor was the theory behind the instrument (as in McGrath, 2014,
labeled “restraint” (or “conative”, or “cautiousness”), and 2018) or the instrument itself (as in Duan et al., 2012; or Ng
frequently included the strengths of honesty, perseverance, et al., 2017). Most of the previous studies on the factorial
prudence and self-regulation. There is more variation as structure of the VIA-IS also based their exploratory results
concerns “transcendence”, or “theological” and “emotional” on measures of character strengths based only on
second-order factors. Table S2 in the Supplemental materials Cronbach’s alpha for reliability, without checking their uni-
further elucidates these models by showing the co-occur- dimensionality, which was not confirmed in the two studies
rence of each pair of strengths within a given second-order that did analyze it (Khumalo et al., 2008; Ng et al., 2017). In
factor for the different models reviewed. Authors (e.g., Duan other words, aggregate measures of character strengths were
et al., 2012) very rarely used confirmatory factor analysis used, without any prior assessment of whether or not they
(CFA) to corroborate the findings they obtained with could be aggregated in a single factor.
DO STRENGTHS CONVERGE INTO VIRTUES? 3

Table 2. Methods used in previous studies on the VIA-IS structure, and results obtained.
VIA-IS
Study Version Language N Analyses Item-level # Fit indices (when available)
Peterson & Park, 2004 240 English FA a 5
Peterson & 240 English PCA (varimax) a 5
Seligman, 2004
Macdonald et al., 2008 240 English 123 PCA (varimax) a 4
Khumalo et al., 2008 240 English 256 CPCA (obliminal) – item level PCA 24
CPCA (obliminal) – virtue level a 6
CPCA (obliminal) – virtue level a 6
EPCA (obliminal) a 2
Ruch et al., 2010 240 German 1674 PCA (varimax) a 5
Singh & Choubisa, 2010 240 English 123 PCA (varimax) a 5
Brdar & Kashdan, 2010 240 Croatian 881 ML (promax) a 4
Shryack et al., 2010 240 English 332 PCA (varimax) a 3
Duan et al., 2012 240 Chinese 839 EFA item-level EFA
96 420 PCA (varimax) 3
96 419 CFA 1-factor 1 CFI ¼ .80, GFI ¼ .71, IFI ¼
.80, TLI ¼ .78, RMSEA
¼ .11
CFA 2-factor (a) 2 CFI ¼ .82, GFI ¼ .74, IFI ¼
.82, TLI ¼ .81, RMSEA
¼ .10
CFA 2-factor (b) 2 CFI ¼ .83, GFI ¼ .80, IFI ¼
.83, TLI ¼ .81, RMSEA
¼ .10
CFA 2-factor (c) 2 CFI ¼ .83, GFI ¼ .77, IFI ¼
.83, TLI ¼ .81, RMSEA
¼ .10
CFA 3-factor 3 CFI ¼ .90, GFI ¼ .85, IFI ¼
.90, TLI ¼ .89, RMSEA
¼ .08
Littman-Ovadia & 240 Hebrew 635 PCA (varimax) a 5
Lavy, 2012
McGrath, 2014 240 English 458.998 a: PAF (promax, varimax) þ No 5
PCA (varimax)
b: same as (a) – Item-level PAF 24
c: same as (a) with 24 a 5
new scales
CFA based on Peterson and No 6 CFI ¼.69, GFI ¼ .67, AGFI ¼
Seligman (2004) .60, RMSEA ¼.14,
SRMR ¼.10
CFA based on (a) a 5 CFI ¼.89, GFI ¼ .87, AGFI ¼
.78, RMSEA ¼.10,
SRMR ¼.05
CFA based on (c) a 5 CFI ¼.90, GFI ¼ .90, AGFI ¼
.84, RMSEA ¼.08,
SRMR ¼.05
Azan~edo et al., 2014 240 Spanish 1060 PCA (varimax) a 5
Littman-Ovadia, 2015 120 English 726.771 PAF (promax) a 5
Seibel et al., 2015 240 Portuguese 1975 PCA (varimax) a 1
PCA (varimax) a 3
PCA (varimax) a 4
Ng et al., 2017 240 English 447.577 CFA based on Peterson and No 6 CFI ¼ .69, TLI ¼ .69, RMSEA
Seligman (2004) ¼.04, SRMR ¼ .06
CESEM-bi No 6 CFI ¼ .80, TLI ¼ .79, RMSEA
¼.03, SRMR ¼ .07
CESEM No 6 CFI ¼ .78, TLI ¼ .77, RMSEA
¼.03, SRMR ¼ .04
107 CFA 24 strengths CFA, x 6 CFI ¼ .85, TLI ¼ .84, RMSEA
¼.03, SRMR ¼ .05
Hierarchical CFA 24-6 6 CFI ¼ .79, TLI ¼ .79, RMSEA
¼.04, SRMR ¼ .07
CESEM 24 strengths 6 CFI ¼ .90, TLI ¼ .89, RMSEA
¼.03, SRMR ¼ .03
CESEM global bifactor 6 CFI ¼ .92, TLI ¼ .90, RMSEA
¼.03, SRMR ¼ .03
McGrath, 2016 240 15.540 CFA based on McGrath (2014) No 5 CFI ¼.75, RMSEA ¼ .12, SRMR
¼ .08
CFA þ MI based on No 5 CFI ¼.96, RMSEA ¼ .07, SRMR
McGrath (2014) ¼ .05
Multigroup CFA based on No 5 CFI ¼ .90, RMSEA ¼ .11,
McGrath (2014) SRMR ¼ .09
Azan~edo et al., 2017 120 Spanish 2143 PAF (promax) a 5 –
H€ofer et al., 2019 240 German 1073 PAF (promax) a 5 –
120 685 PAF (promax) a 5
(continued)
4 FERACO, CASALI, MENEGHETTI

Table 2. Continued.
VIA-IS
Study Version Language N Analyses Item-level # Fit indices (when available)
Anjum & Amjad, 2020 240 Urdu 542 CFA based on Peterson and a 6 CFI ¼ .85, NFI ¼.83, IFI ¼ .85,
Seligman (2004) RMSEA ¼.11, SRMR ¼ .23
PCA (varimax) a 4
CFA a 4 CFI ¼ .92, NFI ¼.91, IFI ¼ .92,
RMSEA ¼.06, SRMR ¼ .05
Note. VIA-IS ¼ Values in Action Inventory of Strengths; FA ¼ factor analysis; PCA ¼ principal component analysis; CPCA ¼ confirmatory principal component ana-
lysis; EPCA ¼ exploratory principal component analysis; ML ¼ maximum likelihood; EFA ¼ exploratory factor analysis; CFA ¼ confirmatory factor analysis;
CFI ¼ comparative fit index; GFI ¼ goodness of fit; IFI ¼ incremental fit index; TLI ¼ Tucker Lewis index; RMSEA ¼ root mean square error of approximation;
PAF ¼ principal axis factoring; AGFI ¼ adjusted goodness of fit; SRMR ¼ standardized root mean square residual; CESEM ¼ clustered exploratory structural equa-
tion modeling; NFI ¼ normed fit index.

Ignoring the unidimensionality of the single character statistically converge. Statistical convergence, however, is not the
strengths raises issues in empirical studies that investigate: (i) only way to study how strengths and virtues are conceptually
the associations of virtues with variables of interest (e.g., related. To this end, other approaches may be considered, for
Martınez-Martı & Ruch, 2017; Peterson et al., 2008; Petkari & instance analyzing how prototypical a strength is of a certain
Ortiz-Tallo, 2018; Weber et al., 2013) because the virtues might virtue (Ruch & Proyer, 2015). For the purpose of this validation
be composed of unreliable strengths; or (ii) the associations of study, we relied on the procedure outlined above to generate a
single character strengths with variables of interest (e.g., more reliable and valid estimate of the validity of the VIA-IS-
Gander et al., 2020; Karris Bachik et al., 2020; Martınez-Martı 120 questionnaire (Beaujean, 2014) compared to previous EFA
& Ruch, 2014) because the scales for the strengths could be approaches. The results of this procedure were also compared
composed of more than one factor (i.e., they might be multidi- with three models obtained using an EFA approach.
mensional). Regarding the former approach, studies generally Finally, character strengths should contribute to life fulfill-
either used the VIA-IS virtues as they are (as in Harzer & ments and help individuals coping with adversity (Peterson &
Ruch, 2015; Shoshani & Slone, 2016) or they sought the factor- Seligman, 2004). Previous research confirmed this conceptual-
ial structure of the strengths best fitting their data before pro- ization (Niemiec, 2013) and consistent findings support the
ceeding with their analyses, leading again to unreliability of the association of character strengths and virtues with a higher
results and replication issues. Therefore, we believe that a dif- level of life satisfaction, as also shown in a recent meta-ana-
ferent statistical approach must be followed to ensure higher lysis of 22 studies (Bruna et al., 2019), general mental health
validity and reliability of such a widespread instrument. (e.g., depression, Huta & Hawley, 2010), resilience (Martınez-
Martı & Ruch, 2017), self-efficacy (Weber et al., 2013), fewer
psychological problems (e.g., distress, Duan & Wang, 2018),
Rationale of the study post-traumatic growth (Peterson et al. 2008), and academic
In the present study we examine the structure of the Italian achievement (Lounsbury et al., 2009). Thus, to further analyze
version of the VIA-IS-120. Previous validation studies, or stud- the validity of the Italian version of the VIA-IS-120, we also
present the correlations of character strengths and virtues
ies focusing on its factorial structure, have some methodological
with both general mental health and psychological distress in
issues – as also highlighted by Ng et al. (2017) – that led, in
terms of anxiety, depression, and stress symptoms. This
our opinion, to a possible misuse of the measure. We identified
choice of variables, instead of positive factors like life satisfac-
three main problems. The first, and probably most important,
tion or happiness, was driven by the contingent pandemic
refers to the unidimensionality of the single character strength
situation. In such difficult times, we felt it was important to
items. This is usually ignored and only reliability indices are
see whether character strengths might have sustained people’s
considered. Unidimensionality problems on the strength level
mental health by being relate to milder symptoms of distress
could then lead to unreliability on the virtue level, making it
and better general mental health. We investigated this by test-
impossible to replicate the second-order structure twice. This
ing the theoretical “adversity functions” of character
second issue is probably rooted in the first unidimensionality
strengths, whereby character strengths should sustain individ-
issue. Then the numerous unreplicated, unconfirmed structures
uals before, during, and after an adverse situation through
of virtues probably induced scholars to run exploratory analyses
their buffering, reappraisal, and resilience functions, respect-
on structures emerging from their own studies, thereby lower-
ively (Niemiec, 2020). These data were previously collected
ing the generalizability of their results. In our study, we tried to for other purposes and published (Casali et al., 2020).
deal with all of these problems by basing our analysis first on
the unidimensionality of the single character strengths, then on
their possible convergence into the virtues proposed by Method
Peterson and Seligman (2004), before going on to examine a
Participants
hierarchical model inclusive of the whole theoretical structure.
We considered character strengths as converging into second- All participants completed the VIA-IS-120 online. We con-
order latent factors (i.e., the six core virtues), and used the term sidered a total of 17757 participants. Data for 16722 partici-
“second-order virtues” to indicate that these virtues represent pants (Sample 1) were kindly provided by the VIA Institute
second-order factors into which character strengths should on Character that authorized their use for research aims.
DO STRENGTHS CONVERGE INTO VIRTUES? 5

These participants completed the inventory directly on the score was calculated according to the Italian validation
Institute’s website between 2016 and 2020. Information on (Cronbach’s alpha ¼ .85; Giorgi et al., 2014).
gender was available for 12874 participants (4567 males;
8237 females; 70 other). Unfortunately, information on age
Procedure
was only available in terms of age range for 173 participants
(125 were between 25 and 54 years old, 10 were over 54, The VIA Institute on Character translated the VIA-IS-120
and 38 were under 25 years old). A second sample of 1035 into various languages, including Italian, and gives anybody
participants1 (Sample 2; 271 males, 764 females; mean age ¼ the opportunity to register and complete the questionnaire
37.88, SD ¼ 14.66; age range: 18–81) was collected online for free. We contacted the Institute to ask for permission to
using Qualtrics. Participants in the second sample were con- access the data collected on their site by people completing
tacted by email, social media or personal contact, and com- the Italian version of the questionnaire online. We also
pleted the questionnaires on a voluntary basis. Data from requested information on age and gender, where available.
Sample 1 were used for the first round of analyses, which The VIA Institute kindly provided their data and gave us
included an exploratory use of modification indices (MI), permission to use them for research purposes. For the
and an EFA. Sample 2 was used to confirm the results of second sample, the VIA-IS-120, the GHQ-12, and the
the analyses run on the first sample and to evaluate conver- DASS-21 were implemented in Qualtrics and a link to
gent validity of character strengths with mental health and the questionnaires was sent to researchers’ personal contacts
distress symptoms. and posted on social media. Respondents voluntarily
answered the questionnaires independently and without
any incentive.
Measures
The Values in Action Inventory of Strengths-120 (VIA-IS- Data analysis
120; Littman-Ovadia, 2015; Italian version provided by the
VIA Institute) is a short form of the original VIA-IS To examine the unidimensionality of the 24 character
(Peterson & Seligman, 2004). The already-translated version strengths, we conducted a series of CFAs, analyzing the local
of the instrument was provided by the VIA Institute and dependence of the items, and the monotonicity of the
checked by the authors for typos, clarity, and consistency responses for each scale. To examine the strengths’ conver-
with the original version. It is a 120-item questionnaire for gence into a second-order virtue, and the general factorial
measuring character strengths (see the VIA Institute https:// structure of the VIA-IS-120, we conducted a series of CFA.
Given that Likert scale responses are ordinal, not continuous
www.viacharacter.org/ for a complete description). Each
data, all 120 items were always treated as ordinal variables
strength is measured by 5 items scored on a 5-point Likert
(Shi et al., 2020), so diagonally-weighted least squares
scale (1 ¼ “not at all like me” to 5 ¼ “very much like me”).
(DWLS) was consequently preferred as estimator over max-
For instance, curiosity and hope are measured with items
imum likelihood (ML). Such a procedure is statistically
such as “I am always curious about the world” and “I always
advisable, and should produce more reliable results in the
look on the bright side”. The original measure showed a
case of Likert-scale type variables (Flora & Curran, 2004;
high internal consistency for every strength (Cronbach’s
Pastore & Lombardi, 2014). The fit of each model was
alpha range: .67–.90 for the VIA-IS, Peterson & Seligman, assessed using multiple indices: the comparative fit index
2004, and .75-.91 for the VIA-IS-120, Littman- (CFI); the Tucker Lewis index (TLI); and the standardized
Ovadia, 2015). root mean squared residual (SRMR). This last index was
The Depression, Anxiety, and Stress Scales-21 (DASS-21, chosen instead of the more commonly used root mean
Lovibond & Lovibond, 1995; validated in Italian by Bottesi squared error of approximation (RMSEA) because recent
et al., 2015). It is a 21-item questionnaire for measuring simulation studies found the former more reliable when
three distress factors: depression (e.g., “I could not feel any using ordinal data in large samples (Shi et al., 2020), and
positive emotion); anxiety (e.g., “I felt I was having a panic when the number of parameters considered is large
attack”); and stress (e.g., “I felt stressed”). Respondents indi- (Maydeu-Olivares et al., 2018). Chi-square (v2) was not con-
cate how often they felt as stated during the last seven days sidered because it is very sensitive to sample size and model
on a 4-point Likert scale (0 ¼ “never happened” to 3 “it complexity, and less informative than the other indices
happened almost every day”). Both the scores for the three (Schermelleh-Engel et al., 2003). As there are no absolute
subscales (Cronbach’s alpha between .74 and .85; Bottesi standards for assessing a model’s fit, and previous CFA on
et al., 2015) and a total general distress factor (Cronbach’s the VIA-IS showed barely acceptable fit indices according to
alpha ¼ .90; Bottesi et al., 2015) were computed. strict cutoffs (Hu & Bentler, 1999), we adopted a non-strin-
The General Health Questionnaire-12 (GHQ-12, gent cutoff rule. We considered models with CFI and TLI
Goldberg, 1978; validated in Italian by Giorgi et al., 2014). It values of .90 or more (Bentler & Bonett, 1980), and SRMR
is a 12-item questionnaire measuring general psychological values of .09 or less (Schermelleh-Engel et al., 2003)
health. Respondents must indicate how often they felt as as adequate.
stated during the last two weeks on a 4-point Likert scale (0 The analyses included six steps, and Sample 1 was used for
¼ “more than usual” to 4 “much less than usual”). A total the first four. We inspected the measurement models of the 24
6 FERACO, CASALI, MENEGHETTI

character strengths in step one, and of the six virtues originally 2004, p. 13). To our knowledge, no studies have since exam-
proposed by Peterson and Seligman (2004) in step two. In step ined whether it is true that the 24 strengths converge statis-
three, a hierarchical model was fitted to measure the validity of tically into the hypothesized virtues. In the present study we
the factorial structure of the VIA-IS-120 model, including all ran a CFA for each virtue, including the hypothesized
the items, the single character strengths and the six virtues. In a strengths that should connect to it and the items related to
fourth step, we ran an EFA and assessed the goodness of fit of said strengths, adding six separate second-order models of
three alternative models (four-, five, and –six factor models) to virtues to the 24 first-order measurement models.
compare them with the CFA model. The final CFA model was
then run on a smaller sample representative of a typical study
on character strengths (Sample 2, n ¼ 1035). This enabled us to The hierarchical structure of the via-is-120
confirm the findings obtained in the first three steps. In the first
three steps, Modification indices were calculated whenever the After confirming the goodness of the first- and second-order
models showed slightly poor fit indices in an effort to better models, we integrated them in a single hierarchical model,
understand the fit issue and try to overcome it (Beaujean, which included: i) all the items (treated as ordinal variables)
2014). If the fit indices were distinctly poor, the model was con- loading on a given related strength; ii) all the strengths load-
sidered unsuitable for inclusion in successive analyses and dis- ing on a given virtue; iii) a correlation between each virtue
carded. All analyses were also run without considering the because they are assumed to be correlated. See Figure S1 in
results of previous measurement models. Finally, in the sixth the Supplemental materials for a graphical representation of
step, we calculated the correlations of the 24 character strengths the hierarchical structure of the VIA-IS-120.
and the six virtues with general mental health and dis-
tress symptoms.
The “lavaan” package (Rosseel, 2012) in the R software Exploratory factor analysis
(R Development Core Team, 2019) was used for the
To accomplish this step, we randomly split the dataset into
CFA analyses.
a calibration sample (N ¼ 8361) and a validation sample
(N ¼ 8361). The calibration sample was used to run the
Measurement model of the 24 character strengths EFA, while the validation sample was used to fit a CFA
based on the results of the EFA. Following the suggestions
The unidimensionality of each character strength was
of Velicer et al. (2000), we adopted two criteria to identify
assessed with CFA. We ran a CFA for each strength, includ-
the number of second-order factors to extract from the 23
ing the five related items considered as ordinal. This step
character strengths (love of learning was excluded): min-
was not usually included in other studies on the factorial
imum average partial analysis (MAP, Velicer et al., 2000);
structure of the VIA-IS, but it is fundamentally important
because it reveals whether there might be a measurement and parallel analysis (PA, Horn, 1965). In accordance with
problem, and on which level, or else it confirms unidimen- the number of eigenvalues higher than one (i.e., 7.91, 2.15,
sionality. Directly assessing the structure of the complete 1.67, 1.42, 1.13, 0.87), the results of the PA identified five
model – without testing the item-level structure – would factors to be extracted, while the MAP indicated four. We
probably result in the inclusion of certain measures (single then extracted two different models using promax oblique
strengths) that could prove problematic. Unidimensionality rotation to account for the correlation between the virtues.
was also studied using the “aisp” function of the “mokken” Similarly, a six-factor model was extracted in line with the
package in R (Van der Ark, 2007), which provides an auto- number of virtues proposed by Peterson and Seligman
matic algorithm based on Mokken’s item selection proced- (2004). Finally, three CFA were run for the three models
ure (Mokken, 1971) and examines monotonicity based on using the same specifics as for the previous CFA models.
Loevinger’s H. Items were considered satisfactory when their
H value was at least .30 (Mokken, 1971). As for local
dependence (LD), the r transformed variance-covariance Replication analysis
matrices of the residuals of the single CFA models were In the fifth step, we replicated the final model using data
examined. If the pairwise correlations between the residuals
from Sample 2 to see if our previous results still held when
of two items exceeded .20, we considered them as possible
a smaller, independent sample was considered, as done in
indicators of local dependence (Terwee et al., 2019). EFA
other studies (e.g., Duan et al., 2012; H€
ofer et al., 2019).
models were also considered to make sure they did not yield
better fit indices than the theoretical model.
Convergent validity
Measurement model of the six virtues
As a last step, using data from Sample 2, we calculated the
The original model proposed by Peterson and Seligman correlations between both character strengths and virtues
(2004) suggested that the 24 strengths reflect six second- and the five convergent measures: GHQ-12, DASS-21 (total
order virtues, as they are “distinguishable routes to display- score), anxiety, depression, and stress. Only correlations
ing one or another of the virtues” (Peterson & Seligman, with p < .001 were considered significant.
DO STRENGTHS CONVERGE INTO VIRTUES? 7

Table 3. Descriptive statistics, reliability coefficients, fit indices and mean of the factor loadings of each character strength.
M SD Skew Kurt a x CFI TLI SRMR MFL
Appreciation of beauty 4.04 .63 .69 .56 .80 .75 .99 .98 .04 .67
Bravery 3.66 .69 .42 .06 .81 .78 .99 .98 .04 .68
Creativity 3.74 .73 .40 .08 .88 .86 .99 .98 .06 .79
Curiosity 3.61 .72 .44 .01 .82 .80 .99 .98 .04 .70
Fairness 3.91 .56 .53 .73 .74 .68 .99 .98 .03 .60
Forgiveness 3.58 .79 .40 .15 .83 .82 .99 .99 .04 .71
Gratitude 3.63 .75 .36 .12 .86 .86 1.00 .99 .03 .75
Honesty 4.23 .53 .77 1.22 .79 .76 .99 .98 .05 .66
Hope 3.45 .79 .42 .17 .81 .79 .97 .95 .07 .69
Humility 3.49 .71 .39 .11 .76 .74 .96 .92 .07 .64
Humor 3.69 .74 .41 .13 .83 .84 .99 .99 .06 .71
Judgment 4.17 .52 .69 1.16 .75 .69 .99 .98 .03 .62
Kindness 4.06 .57 .66 .82 .79 .73 1.00 1.00 .02 .66
Leadership 3.57 .63 .20 .14 .77 .75 .95 .91 .08 .64
Love 3.79 .73 .59 .18 .79 .77 .96 .92 .08 .67
Love of learning 3.63 .75 .21 .55 .79 .85 .95 .90 .19 .73
Perseverance 3.64 .78 .48 .12 .88 .86 .99 .99 .05 .78
Perspective 3.69 .68 .38 .12 .82 .80 1.00 .99 .03 .69
Prudence 3.57 .69 .35 .07 .79 .78 .99 .98 .04 .66
Self  regulation 3.29 .76 .23 .30 .74 .73 .99 .98 .05 .61
Social intelligence 3.76 .62 .44 .28 .75 .72 .97 .94 .06 .62
Spirituality 3.07 .81 .03 .45 .78 .79 .98 .97 .07 .65
Teamwork 3.63 .62 .37 .32 .74 .69 .99 .99 .03 .61
Zest 3.39 .80 .25 .27 .85 .83 .99 .98 .05 .73
Note. M ¼ mean; SD ¼ standard deviation; Kurt ¼ kurtosis; CFI ¼ comparative fit index; TLI ¼ Tucker Lewis index; SRMR ¼ standardized root mean square residual;
MFL ¼ mean of the factor loadings.

Results On the virtues level, all but one of the virtues hypothe-
sized showed good fit indices (see Table 4), while transcend-
The results of the character strengths’ measurement models
ence reached a slightly inadequate SRMR (.11), despite good
showed that all the strengths except for love of learning can
CFI and TLI values (.93 and .92, respectively). This would
be considered unidimensional (see Table 3 for details). The
suggest that strengths could define virtues in much the same
fit indices were always acceptable, with the CFI ranging
way as originally hypothesized by Peterson and Seligman
from .95 to 1, the TLI from .91 to 1, and the SRMR ranging
(2004). Each model of a virtue included its related strengths,
from .03 to .08. On the other hand, love of learning showed
and the items loading on each strength. In particular, the
a very high SRMR (.19), and a barely acceptable TLI (.90).
virtue of wisdom and knowledge included creativity, curios-
Love of learning should connect to the virtue of wisdom
ity, judgment, and perspective. Love of learning was
and knowledge, but it was not considered in subsequent
analyses because of its unacceptable factorial structure. The excluded from the analyses; adding it to the model at this
results were the same using Mokken’s automated item selec- point again gave rise to unacceptable fit indices (SRMR ¼
tion method, which showed that the first two items on the .10). The second-order loadings of the virtue of wisdom and
love of learning scale (items 17 and 48) loaded on a second knowledge ranged between .64 and .78 (mean ¼ .72). The
factor. Similarly, local dependence was only found between virtue of courage included bravery, honesty, perseverance
the items for love of learning and the first two items’ resid- and zest, and its second-order loadings ranged between .74
uals, which showed strong correlations with each other (r ¼ and .85 (mean ¼ .72). The virtue of humanity included
.37), and negative correlations with the third and fourth love, kindness and social intelligence, and its second-order
items’ residuals (r > .26). Monotonicity was also generally loadings ranged between .67 and .89 (mean ¼ .79). The vir-
confirmed: of the 120 items, only seven showed H values tue of justice included fairness, leadership and teamwork,
lower than .30. In particular, one item for the gratitude, and its second-order loadings ranged between .83 and .98
honesty, prudence, self-regulation, and social intelligence (mean ¼ .90). The virtue of temperance included forgive-
scales, and two items for the spirituality scale did not meet ness, humility, prudence and self-regulation, and its second-
the criteria (see Table S3 in the Supplemental materials for order loadings ranged between .39 and .69 (mean ¼ .57).
the complete set of values). The loadings of each strength The virtue of transcendence included appreciation of beauty
were acceptable, however, their means varied from .60 (for and excellence, gratitude, hope, humor and spirituality. As
fairness) to .79 (for creativity), and all strengths showed an the SRMR for this virtue was too high (.11), we only
acceptable reliability in terms of Cronbach’s alpha (.74 < a inspected the modification indices for this model, which
< .88), and McDonald’s ordinal omega (.68 < x < .86). See showed that some items had correlations between residuals.
Table 3 for details and Table S4 in the Supplemental materi- In particular, the correlations between residuals for items 14
als for the correlations between the strengths. To sum up, (“I practice my religion”, spirituality) and 71 (“My faith
with the exception of love of learning, the scales for all the makes me who I am”, spirituality), for items 29 (“Despite
single strengths showed good psychometric properties, challenges, I always remain hopeful about the future”, hope)
though attention should be paid to the monotonicity of and 30 (“My faith never deserts me during hard times”, spir-
some items. ituality), and for items 87 (“I have a great sense of humor”,
8 FERACO, CASALI, MENEGHETTI

Table 4. Descriptive statistics, fit indices, mean of the factor loadings mean, and correlations (all significant at the .001 level) of each virtue.
M SD CFI TLI SRMR MFL .1 .2 .3 .4 .5 .6
1.Wisdom and knowledge 3.80 .49 .95 .94 .08 .72 – .59 .51 .41 .34 .58
2.Courage 3.73 .53 .95 .94 .08 .72 .59 – .57 .49 .37 .66
3.Humanity 3.87 .50 .96 .96 .06 .79 .51 .57 – .60 .38 .67
4.Justice 3.70 .50 .95 .94 .07 .90 .41 .49 .60 – .52 .48
5.Temperance 3.48 .49 .95 .94 .07 .57 .34 .37 .38 .52 – .40
6.Transcendence 3.58 .53 .95 .95 .09 .73 .58 .66 .67 .48 .40 –
Transcendence without modification indices – – .93 .92 .11 .71
Note. CFI ¼ comparative fit index; TLI ¼ Tucker Lewis index; SRMR ¼ standardized root mean square residual; MFL ¼ mean of the factor loadings.

humor) and 119 (“I am known for my good sense of humor”, We then fitted the best model (m2) again using our
humor) showed high modification indices (MI ¼ 13301.80, experimentally-collected Sample 2 (1035 participants). In
6615.56, and 6509.93, respectively). After specifying these this case (m6), the fit indices mainly overlapped with those
correlations and fitting the model again, the fit indices were obtained from the larger sample (CFI ¼ .90, TLI ¼ .89,
acceptable (SRMR ¼ .09). The second-order loadings of the SRMR ¼ .08). The fit indices of all the hierarchical models
virtue of transcendence then ranged between .45 and .91 are listed in Table 5.
(mean ¼ .73). All the virtues correlated significantly with Data from Sample 2 were also used for studying VIA-IS-
each other, with transcendence showing generally stronger 120 convergent validity. Results (see Table 6 for complete
correlations (especially with humanity and courage), and correlations) show that 9 strengths and five virtues (i.e.,
temperance showing generally weaker correlations, while excluding justice) correlated significantly with GHQ-12
justice correlated particularly strongly with humanity (see scores; 10 strengths and five virtues (i.e., excluding justice)
Table 4). correlated significantly with DASS-21 total scores; in par-
A first hierarchical model (m1) including all the models ticular, only three strengths and two virtues correlated with
previously tested, and a correlation between each pair of vir- the anxiety subscale scores; 19 strengths and all the six vir-
tues was fitted (see Figure S1 in the Supplemental materials tues correlated with the depression subscale scores; nine
for a graphical representation). The model included 115 strengths and four virtues (i.e., excluding justice and wisdom
items treated as ordinal variables (love of learning items and knowledge) correlated significantly with the stress sub-
were excluded) loading on 23 strengths, loading in turn on scale scores. It should be noted that 10 out of 24 character
strengths – appreciation of beauty, fairness, humility,
six virtues. The results showed slightly inadequate fit indices
humor, judgment, kindness, leadership, love of learning
(CFI ¼ .89, TLI ¼ .89, SRMR ¼ .08). Modification indices
(that was excluded from the analyses due to unidimensional-
were inspected again and the model was rerun, specifying a
ity issues), perspective, and prudence – did not show any
correlation between the residuals of four strengths (perspec-
significant relation with the outcome variables. For what
tive and prudence, MI ¼ 11912.15; perseverance and self-
concerns the virtues, transcendence, courage, and temper-
regulation, MI ¼ 10387.97) and two items (5 “I have no trou-
ance significantly correlated with all the outcome variables,
ble eating healthy foods”, self-regulation; and 68 “I can humanity with all the outcome variables except anxiety, and
always stay on a diet”, self-regulation, MI ¼ 16230.77). This wisdom and knowledge with three of the outcome variables
second model (m2) showed acceptable fit indices (CFI ¼ (i.e., GHQ-12, DASS-21 total score, and depression). Justice
.90, TLI ¼ .90, SRMR ¼ .08), so we decided not to specify only significantly related with depression.
any other modification index. Complete loadings of the final
model are given in Table S5 in the Supplemental materials.
The same model was also fitted without considering the Discussion
results of previous measurement models (e.g., also including The VIA-IS is a well-known and widely used measure of
love of learning), but it showed lower fit indices (CFI ¼ .88, character derived from the huge theoretical work of
TLI ¼.88, SRMR ¼ .08). To ensure that a model derived Peterson and Seligman (2004), however its psychometric
from EFA did not yield remarkably better fit indices, we properties and factorial structure are still to be defined and
extracted three models from the calibration sample, one an Italian validation is unavailable. Here, for the first time,
with four and the other with five second-order factors we examine its validity in the Italian version on a large sam-
(based on the results of the MAP and PA), and one with six ple of individuals and try to deal with some of the issues
second-order factors, without including data for love of identified in previous validation and factorial studies. To
learning (see Table S6, S7, and S8 in the supplemental mate- our knowledge, this is the first study to investigate the
rials for the complete set of results). The three models (m3, internal structure of the VIA-IS-120 at both item- and
m4, and m5) were then fitted using CFA on the validation strength-level and to treat items as ordinal, which is strongly
sample. They both included the 115 ordinal items loading recommended for Likert-type responses if the sample size is
on 23 strengths, which loaded on four (m3), five (m4), or large enough for threshold estimation (Flora & Curran,
six (m5) second-order factors, respectively. The correlation 2004; Pastore & Lombardi, 2014). It is important to bear in
between each pair of factors was also specified. The results mind, however, that: we had no details of the age of partici-
show lower fit indices for these models than for m2 (CFI ¼ pants in the first sample (though data provided by the VIA
89, .88, .89; TLI ¼ .89, .88, .89; SRMR ¼ .08, .08, .08). Institute of Character generally cover the whole population;
DO STRENGTHS CONVERGE INTO VIRTUES? 9

Table 5. Fit indices of the hierarchical models. The final model is in bold. Table 6. Correlations of character strengths and virtues with general mental
CFI TLI SRMR health and distress symptoms (anxiety, depression, and stress).
m0 .88 .88 .08 GHQ DASS Anxiety Depression Stress
m1 .89 .89 .08 Appreciation of beauty .04 .03 .10 .02 .01
m2 .90 .90 .08 Bravery .08 .09 .01 .18 .03
m3 .89 .89 .08 Creativity .16 .10 .00 .16 .08
m4 .88 .88 .08 Curiosity .23 .22 .09 .31 .14
m5 .89 .89 .08 Fairness .00 .02 .01 .01 .05
m6 .90 .89 .08 Forgiveness .10 .21 .13 .20 .22
Note. CFI ¼ comparative fit index; TLI ¼ Tucker Lewis index; Gratitude .24 .27 .10 .36 .21
SRMR ¼ standardized root mean square residual. m0 ¼ original model of 24 Honesty .07 .13 .08 .16 .09
strengths and six virtues (Peterson & Seligman, 2004); m1 ¼ m0 excluding Hope .31 .40 .20 .50 .31
love of learning; m2 ¼ m1 with modification indices; m3 ¼ four-factor solu- Humility .03 .07 .01 .04 .11
tion; m4 ¼ five-factor solution; m5 ¼ six-factor solution; m6 ¼ m2 fitted on Humor .14 .07 .02 .11 .05
Sample 2. Judgment .02 .02 .04 .02 .01
Kindness .03 .06 .01 .09 .06
Leadership .09 .06 .02 .11 .05
see Ng et al., 2017 for an example); we mostly focused on Love .17 .22 .10 .31 .15
Love of earning .11 .05 .01 .11 .02
the factorial validity of the measure and did not analyze Perseverance .16 .25 .14 .34 .15
convergent validity by using peer/other-ratings or life satis- Perspective .04 .03 .02 .07 .02
faction, as usually done (e.g., Ruch et al., 2010); and we con- Prudence .04 .11 .09 .09 .11
Self-regulation .17 .22 .13 .23 .19
sidered the Italian version of the short form of the Social intelligence .10 .09 .01 .15 .07
questionnaire (i.e., VIA-IS-120). As a consequence, our find- Spirituality .21 .24 .09 .31 .20
ings might not apply to the original VIA-IS, or to versions Teamwork .10 .15 .06 .18 .13
Zest .30 .34 .17 .45 .25
in other languages. However, we strongly believe that an Wisdom and knowledge .16 .13 .03 .20 .09
item-level analysis and the unidimensionality of each of Courage .21 .28 .14 .39 .18
the 24 character strengths’ subscales should be tested Humanity .13 .17 .06 .24 .12
Justice .08 .09 .01 .12 .10
systematically also in the other versions and that relying Temperance .13 .23 .14 .22 .24
only on reliability information, as previously done, should Transcendence .27 .28 .10 .38 .23
be deprecated.  ¼ p < .001.
Unidimensionality of the character strengths subscales
went untested in almost every previous study. To our know- look precisely at the most problematic items. Two items
ledge, the first authors to test it with a CFA were Ng et al. regarding spirituality (item 14, “I practice my religion”; and
(2017): they found that the subscales of the long form of the item 71, “My faith makes me who I am”) showed high corre-
instrument (VIA-IS) were not unidimensional (i.e., the fit lations between residuals. This may be because these items
indices were poor), so they reduced the number of items refer more to being religious than to spirituality in general,
based on their factor loadings until unidimensionality was while the opposite is true of the other items comprising this
achieved. We found the subscales of the Italian short form character strength (e.g., item 7, “I am a spiritual person”;
of the VIA-IS unidimensional (except for love of learning item 110, “My beliefs make my life important”). In the same
scale, in which two items [17, 48] differed in content from vein, item 30 regarding spirituality (“My faith never deserts
the other three, and there was a high correlation between me during hard times”) describes a sense of faith in the
residuals), however, and reliable in terms of both future that might also be intended as hope that things will
Cronbach’s alphas and McDonald’s ordinal omega. Our get better. This might explain why its residuals correlated
findings would suggest that single character strength sub- with items from the character strength of hope, such as
scales are unique factors (at least in the Italian short form) “Despite challenges, I always remain hopeful about the
and, with the exception of love of learning, can be used in future” (item 29). Modification indices also pinpointed two
further studies. It should be noted nevertheless that seven items regarding humor (item 87, “I have a great sense of
items did not meet the monotonicity assumptions, and humor”; and item 119, “I am known for my good sense of
might be revised in future studies to further increase the val- humor”). Compared with the other items relating to this
idity of the measure. The unidimensionality of the strengths character strength (e.g., item 57, “I try to add some humor
was still not enough, however, to validate their use from a to whatever I do”), these two items refer to how people see
multivariate perspective, i.e., also including the other themselves, and could consequently overlap with each other
strengths, which are strongly correlated with each other and much more than with the others, which are more behavioral
may overlap. This aspect was ascertained in subsequent and refer to the use of humor in different situations. Such
analyses, by answering the question first of their conver- small similarities could prove particularly important in a
gence into common virtues and then of the convergence of smaller set of items as in short forms of the VIA-IS (like the
both strengths and virtues into a complete model of charac- one adopted in the present case), because they derive from
ter strengths. the aggregation of items correlating more strongly with the
Our results largely confirmed the measurement models of latent factor, not from a meaningful analysis of the items’
each of the six virtues, the only exception being transcend- content (Littman-Ovadia, 2015). Keeping these modification
ence, for which the SRMR was poor. We inspected modifi- indices in mind, we specified the correlations between the
cation indices to shed light on the issue, this enabled us to above-mentioned items in a second model and found better
10 FERACO, CASALI, MENEGHETTI

fit indices, although the use of modification indices does not practical (e.g., McGrath et al., 2018; Ruch et al., 2021; Ruch
entitle us to conclude that the virtue of transcendence per- & Proyer, 2015) standpoint, and came up with theory-driven
fectly fits our data. The character strength of spirituality (as in Miller, 2019) or data-driven ( as in McGrath et al.,
strength in particular is more difficult to distinguish as a 2018; Ng et al., 2017 ) alternatives. For instance, they pro-
unique strength when considering also the other strengths posed vises as counterparts of virtues (Miller, 2019), polyto-
converging into transcendence, even though it emerged as mous assignment of strengths to virtues (Ruch et al., 2021),
unidimensional in a separate analysis. All the other virtues a unique super-ordered factor expressing dispositional posi-
showed good fit indices, which means that their strengths tivity (Ng et al., 2017), or a reduced tripartite taxonomy
can be grouped together as single factors or, in other words, (Duan et al., 2012; McGrath, 2015; McGrath et al., 2018).
as a unique second-order variable. That said, the primary aim of the present study was to valid-
The last level of our analysis involved including all the ate the Italian version of the VIA-IS-120 and, in doing so,
virtues and strengths in a single hierarchical model. we attempted to address some of the methodological issues
Although previous studies had always found very poor fit we identified in the use currently made of the VIA-IS, with-
indices when studying the theoretical model originally pro- out venturing into new theoretical proposals. To answer
posed by Peterson and Seligman (2004), our hierarchical such theoretical questions, further studies must be run and
model resulted in an acceptable SRMR and only slightly different empirical approaches should be taken into account
inadequate CFI and TLI values. Following the same proced- (e.g., Ruch & Proyer, 2015). Overall, what emerged from
ure as for the virtue of transcendence, we identified three our analyses is that (with the exception of love of learning)
main problems in the model: a correlation between two the inventory’s character strengths can be considered unidi-
items of self-regulation that could be described as “diet mensional and reliable in Italian samples, and mainly con-
items” (i.e., item 5, “I have no trouble eating healthy foods”; necting to the hypothesized virtues (again with some
and item 68, “I can always stay on a diet”), and between exceptions, as discussed earlier), in line with results obtained
two pairs of strengths converging into different virtues (per- using other methods by Ruch and Proyer (2015), and by
spective and prudence, which belong respectively to wisdom Ruch et al. (2021). On the basis of the only other study ana-
and to temperance; and perseverance and self-regulation, lyzing the factorial structure of the single character strengths
which belong respectively to courage and temperance). One (Ng et al., 2017), we can extend our results to other VIA-IS
way to clarify these associations could be to test alternative versions (in other languages). We can also tentatively rec-
models in which the strengths that showed to be correlated ommend using the short form (VIA-IS-120) because, unlike
are placed together as converging into the same virtue, and the original form, it showed unidimensionality on the char-
then check whether the model fit improves as a result. In acter strength level. This should be tested in more depth,
McGrath et al. (2018), for instance, perseverance and self- however. The use of non-unidimensional scales should
regulation were both found to load on the “self-control” fac- be avoided.
tor. It could also be argued, however, that - even if they Moving forward, as some strengths emerged as partially
hypothetically connect to different virtues - strengths like overlapping, and possibly pointing to multiple virtues, one
perseverance (or grit in Duckworth’s work) and self-regula- solution could be to revise the items on the basis not only
tion are known to be related, but separate constructs (see of correlations (as in Littman-Ovadia, 2015), but also of
Duckworth & Gross, 2014, for instance), so it may be that content, in order to measure strengths more accurately, as
the items used in the VIA-IS are not precise enough to tear done more recently by McGrath (2017), but would demand
them apart. Notwithstanding these flaws, the hierarchical to a complete revision of the VIA-IS short form.
model resulted in an acceptable fit, in line with the results Be that as it may, we strongly believe that confirmatory
obtained using other methods, such as EFA (Anjum & analyses, guided by solid theoretical assumptions, are the
Amjad, 2020; McGrath, 2014) or CFA on reduced versions best way to make the VIA-IS uniform, replicable, and usable
(Duan et al., 2012; McGrath, 2016; Ng et al., 2017). Such fit across studies, and thereby strengthen the value of its associ-
indices were also slightly better than those of the two EFA ations with other outcomes.
models fitted on our data, suggesting that the CFA model Finally, the VIA-IS-120 proved reasonably reliable in
might be preferable because it also has the advantage of the both a large sample and a second, smaller independent sam-
strong theoretical basis proposed by Peterson and Seligman ple, albeit with a few issues relating to a poor fit for love of
(2004). Finally, the hierarchical model could also be repli- learning and an overlap for some strengths. Future studies
cated in a smaller independent sample (Sample 2), indicat- should adopt confirmatory approaches to validate existing
ing that it may be reliable even in studies on smaller or new theoretical proposals and investigate the structure of
samples, such as those usually collected for experimen- the VIA-IS hierarchically, considering both item and
tal research. strength levels, and treating items as ordinal variables.
Peterson and Seligman (2004) conceptualized character Convergent validity was also examined in the second sample
on the grounds of a commendable amount of theoretical including measures of mental health (GHQ-12) and distress,
research and their proposal was designed to be open to in terms of anxiety, depression, and stress (DASS-21). Such
adjustments. It was not meant to be exhaustive or flawless. correlations showed to be small-to-medium and in line with
Several other authors have questioned their taxonomy from previous studies investigating the relation between character
both a theoretical (e.g., Miller, 2019; Snow, 2019) and a strengths and distress in different populations (e.g., Antebi-
DO STRENGTHS CONVERGE INTO VIRTUES? 11

Gruszka et al., 2021; Haridas et al., 2017; Huta & Hawley, References
2010; Lee et al., 2019; Lim, 2015) with transcendence
Anjum, A., & Amjad, N. (2020). Values in action inventory of
strengths such as gratitude and hope usually resulting the strengths (VIA-IS): Translation and validation in Urdu language.
most related to higher mental health status (Huta & Hawley, Pakistan Journal of Psychological Research, 35(1), 163–189. https://
2010; Petkari & Ortiz-Tallo, 2018). These results confirm the doi.org/10.33824/PJPR.2020.35.1.10
importance of precisely measure character strengths to sus- Antebi-Gruszka, N., Friedman, A. A., & Schrimshaw, E. W. (2021).
Character strengths and their associations with well-being and men-
tain positive psychology interventions to nurture wellbeing, tal distress among lesbian, gay, bisexual, and queer individuals.
mental health, and help people flourishing (Niemiec, 2020). Journal of Gay & Lesbian Social Services, 33(2), 157–179. https://doi.
org/10.1080/10538720.2020.1859424
Aza~nedo, C. M., Fernandez-Abascal, E. G., & Barraca, J. (2014).
Conclusions Character strengths in Spain: validation of the values in action
inventory of strengths (VIA-IS) in a Spanish sample. Clınica y
Using CFA, we confirmed the validity of the Italian version Salud, 25(2), 123–130. https://doi.org/10.1016/j.clysa.2014.06.002
of the VIA-IS-120 on character strengths, virtues and hier- Aza~nedo, C. M., Fernandez-Abascal, E. G., & Barraca, J. (2017).
archical levels. Our results showed that all the strength fac- Version corta del cuestionario VIA de fortalezas personales.
tors except love of learning are unidimensional and can be Psicothema, 29, 254–260. https://doi.org/10.7334/psicothema2016.225
Beaujean, A. A. (2014). Latent variable modeling using R: A step-by-step
considered separately in future studies. On a more general guide. Routledge. https://doi.org/10.4324/9781315869780
level, with few exceptions, the items of the 23 strengths did Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness
not appear to load on other strengths, and the latter con- of fit in the analysis of covariance structures. Psychological Bulletin,
verged well into the virtues originally hypothesized by 88(3), 588–606. https://doi.org/10.1037/0033-2909.88.3.588
Bottesi, G., Ghisi, M., Altoe, G., Conforti, E., Melli, G., & Sica, C.
Peterson and Seligman (2004). With the exception of two (2015). The Italian version of the depression anxiety stress scales-21:
pairs – self-regulation and perseverance, and of prudence Factor structure and psychometric properties on community and
and perspective – the different strengths did not load on vir- clinical samples. Comprehensive Psychiatry, 60, 170–181. https://doi.
tues other than the one hypothesized nor did they correlate org/10.1016/j.comppsych.2015.04.005
Brdar, I., & Kashdan, T. B. (2010). Character strengths and well-being
too strongly with other strengths. This evidence leads us to
in Croatia: An empirical investigation of structure and correlates.
conclude that, after excluding love of learning, the VIA-IS- Journal of Research in Personality, 44(1), 151–154. https://doi.org/10.
120 can be used to study single character strengths and 1016/j.jrp.2009.12.001
aggregate scores of the original virtues (Peterson & Bruna, M. O., Brabete, A. C., & Izquierdo, J. M. A. (2019). Reliability
Seligman, 2004), without exploring the convergence of generalization as a seal of quality of substantive meta-analyses: The
case of the VIA inventory of strengths (VIA-IS) and their relation-
strengths into new components.
ships to life satisfaction. Psychological Reports, 122, 1167–1188.
https://doi.org/10.1177/0033294118779198
Casali, N., Feraco, T., Ghisi, M., & Meneghetti, C. (2020). Andra tutto
Notes bene”: Associations between character strengths, psychological dis-
1. These data were collected for a study already published: tress and self-efficacy during Covid-19 lockdown. Journal of
Casali et al., 2020 Happiness Studies, https://doi.org/10.1007/s10902-020-00321-w
Duan, W., Ho, S. M. Y., Yu, B., Tang, X., Zhang, Y., Li, T., & Yuen, T.
(2012). Factor structure of the Chinese virtues questionnaire.
Research on Social Work Practice, 22(6), 680–688. https://doi.org/10.
Acknowledgments 1177/1049731512450074
Duan, W., & Wang, Y. (2018). Latent profile analysis of the three-
The authors would like to acknowledge the VIA Institute for giving us
dimensional model of character strengths to distinguish at-strengths
access to the data used in this study and for granting us permission to
and at-risk populations. Quality of Life Research, 27(11), 2983–2990.
use them for this publication. https://doi.org/10.1007/s11136-018-1933-1
Duckworth, A., & Gross, J. J. (2014). Self-control and grit: Related but
separable determinants of success. Current Directions in
Declaration of interest statement Psychological Science: A Journal of the American Psychological
This research did not receive any specific grant from funding agencies Society, 23, 319–325. https://doi.org/10.1177/0963721414541462
in the public, commercial, or not-for-profit sectors. The authors have Flora, D. B., & Curran, P. J. (2004). An empirical evaluation of alterna-
tive methods of estimation for confirmatory factor analysis with
no potential conflicts of interest to report.
ordinal data. Psychological Methods, 9(4), 466–491. https://doi.org/
10.1037/1082-989X.9.4.466
Gander, F., Hofmann, J., Proyer, R. T., & Ruch, W. (2020). Character
Funding
strengths – Stability, change, and relationships with well-being
The present work was conducted as part of the Dipartimenti di changes. Applied Research in Quality of Life, 15(2), 349–367. https://
Eccellenza research program (DM 11/05/2017 n. 262), supported by a doi.org/10.1007/s11482-018-9690-4
grant from MIUR to the Department of General Psychology, Giorgi, G., Perez, J. M. L., D’Antonio, A. C., Perez, F. J. F., Arcangeli,
University of Padua. G., Cupelli, V., & Mucci, N. (2014). The general health question-
naire (GHQ-12) in a sample of Italian workers: mental health at
individual and organizational level. World Journal of Medical
Data availability statement Sciences, 11, 47–56. https://doi.org/10.5829/idosi.wjms.2014.11.1.
83295
Data for Sample 2 are available on Figshare. Doi: 10.6084/ Goldberg, D. P. (1978). Manual of the general health questionnaire.
m9.figshare.13353539 Nfer-Nelson Publishers.
12 FERACO, CASALI, MENEGHETTI

Haridas, S., Bhullar, N., & Dunstan, D. A. (2017). What’s in character Martınez-Martı, M. L., & Ruch, W. (2017). Character strengths predict
strengths? Profiling strengths of the heart and mind in a community resilience over and above positive affect, self-efficacy, optimism,
sample. Personality and Individual Differences, 113, 32–37. https:// social support, self-esteem, and life satisfaction. The Journal of
doi.org/10.1016/j.paid.2017.03.006 Positive Psychology, 12(2), 110–119. https://doi.org/10.1080/
Harzer, C., & Ruch, W. (2015). The relationships of character strengths 17439760.2016.1163403
with coping, work-related stress, and job satisfaction. Frontiers in Maydeu-Olivares, A., Shi, D., & Rosseel, Y. (2018). Assessing fit in
Psychology, 6, 1–12. https://doi.org/10.3389/fpsyg.2015.00165 structural equation models: A Monte-Carlo evaluation of RMSEA
Heintz, S., & Ruch, W. (2020). Character strengths and job satisfaction: versus SRMR confidence intervals and tests of close fit. Structural
Differential relationships across occupational groups and adulthood. Equation Modeling: A Multidisciplinary Journal, 25(3), 389–402.
Applied Research in Quality of Life, 15(2), 503–527. https://doi.org/ https://doi.org/10.1080/10705511.2017.1389611
10.1007/s11482-018-9691-3 McGrath, R. E. (2014). Scale- and item-level factor analyses of the VIA
H€ofer, S., Hausler, M., Huber, A., Strecker, C., Renn, D., & H€ oge, T. inventory of strengths. Assessment, 21, 4–14. https://doi.org/10.1177/
(2019). Psychometric characteristics of the German values in action 1073191112450612
inventory of strengths 120-item short form. Applied Research in McGrath, R. E. (2015). Integrating psychological and cultural perspec-
Quality of Life, 15, 597–611. https://doi.org/10.1007/s11482-018- tives on virtue: The hierarchical structure of character strengths. The
9696-y Journal of Positive Psychology, 10(5), 407–424. https://doi.org/10.
Horn, J. L. (1965). A rationale and test for the number of factors in 1080/17439760.2014.994222
factor analysis. Psychometrika, 30, 179–185. https://doi.org/10.1007/ McGrath, R. E. (2016). Measurement invariance in translations of the
BF02289447 VIA inventory of strengths. European Journal of Psychological
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covari- Assessment, 32(3), 187–194. https://doi.org/10.1027/1015-5759/
ance structure analysis: Conventional criteria versus new alterna- a000248
tives. Structural Equation Modeling: A Multidisciplinary Journal, McGrath, R. E. (2017). Technical report: The VIA assessment suite for
6(1), 1–55. https://doi.org/10.1080/10705519909540118 adults: Development and evaluation. VIA Institute on Character.
Huta, V., & Hawley, L. (2010). Psychological strengths and cognitive McGrath, R. E., Greenberg, M. J., & Hall-Simmonds, A. (2018).
vulnerabilities: Are they two ends of the same continuum or do they Scarecrow, tin woodsman, and cowardly lion: The three-factor
have independent relationships with well-being and ill-being? model of virtue. The Journal of Positive Psychology, 13(4), 373–392.
Journal of Happiness Studies, 11(1), 71–93. https://doi.org/10.1007/ https://doi.org/10.1080/17439760.2017.1326518
s10902-008-9123-4 Miller, C. B. (2019). Some philosophical concerns about how the VIA
Karris Bachik, M. A., Carey, G., & Craighead, W. E. (2020). VIA char- classifies character traits and the VIA-IS measures them. The
acter strengths among U.S. college students and their associations Journal of Positive Psychology, 14(1), 6–19. https://doi.org/10.1080/
17439760.2018.1528377
with happiness, well-being, resiliency, academic success and psycho-
Mokken, R. J. (1971). A Theory and Procedure of Scale Analysis. De
pathology. The Journal of Positive Psychology, 1–14. https://doi.org/
Gruyter.
10.1080/17439760.2020.1752785
Ng, V., Cao, M., Marsh, H. W., Tay, L., & Seligman, M. E. P. (2017).
Khumalo, I. P., Wissing, M. P., & Temane, Q. M. (2008). Exploring the
The factor structure of the Values in Action Inventory of Strengths
validity of the values-in-action inventory of strengths (VIA-IS) in an
(VIA-IS): An item-level exploratory structural equation modeling
African context. Journal of Psychology in Africa, 18(1), 133–142.
(ESEM) bifactor analysis. Psychological Assessment, 29, 1053–1058.
https://doi.org/10.1080/14330237.2008.10820180
https://doi.org/10.1037/pas0000396
Lee, B., Kaya, C., Chen, X., Wu, J. R., Iwanaga, K., Umucu, E., Bezyak,
Niemiec, R. M. (2013). VIA character strengths: Research and practice
J., Tansey, T. N., & Chan, F. (2019). The buffering effect of charac-
(The first 10 years). In Well-being and cultures. (pp. 11–29).
ter strengths on depression. European Journal of Health Psychology,
Springer. https://doi.org/10.1007/978-94-007-4611-4_2
26(3), 101–109. https://doi.org/10.1027/2512-8442/a000036 Niemiec, R. M. (2020). Six functions of character strengths for thriving
Lim, Y. J. (2015). Relations between virtues and positive mental health at times of adversity and opportunity: A theoretical perspective.
in a Korean population: A multiple indicators multiple causes Applied Research in Quality of Life, 15(2), 551–572. https://doi.org/
(MIMIC) model approach. International Journal of Psychology, 50, 10.1007/s11482-018-9692-2
272–278. https://doi.org/10.1002/ijop.12096 Pastore, M., & Lombardi, L. (2014). The impact of faking on
Littman-Ovadia, H. (2015). Short form of the VIA inventory of Cronbach’s alpha for dichotomous and ordered rating scores.
strengths: Construction and initial tests of reliability and validity. Quality & Quantity, 48(3), 1191–1211. https://doi.org/10.1007/
International Journal of Humanities Social Sciences and Education, 2, s11135-013-9829-1
229–237. Retrieved from www.arcjournals.org Peterson, C., & Park, N. (2004). Classification and measurement of
Littman-Ovadia, H., & Lavy, S. (2012). Character strengths in Israel: character strengths: Implications for practice. In Positive psychology
Hebrew adaptation of the VIA inventory of strengths. European in practice. (pp. 433–446). John Wiley & Sons, Inc. https://doi.org/
Journal of Psychological Assessment, 28(1), 41–50. https://doi.org/10. 10.1002/9780470939338.ch27
1027/1015-5759/a000089 Peterson, C., Park, N., Pole, N., D’Andrea, W., & Seligman, M. E. P.
Lounsbury, J. W., Fisher, L. A., Levy, J. J., & Welsh, D. P. (2009). An (2008). Strengths of character and posttraumatic growth. Journal of
investigation of character strengths in relation to the academic suc- Traumatic Stress, 21, 214–217. https://doi.org/10.1002/jts.20332
cess of college students. Individual Differences Research, 7, 52–69. Peterson, C., & Seligman, M. E. P. (2004). Character strengths and vir-
Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative tues: A handbook and classification. Oxford University Press and
emotional states: Comparison of the depression anxiety stress scales American Psychological Association.
(DASS) with the beck depression and anxiety inventories. Behaviour Petkari, E., & Ortiz-Tallo, M. (2018). Towards youth happiness and
Research and Therapy, 33, 335–343. https://doi.org/10.1016/0005- mental health in the United Arab Emirates: The path of character
7967(94)00075-U strengths in a multicultural population. Journal of Happiness Studies,
Macdonald, C., Bore, M., & Munro, D. (2008). Values in action scale 19, 333–350. https://doi.org/10.1007/s10902-016-9820-3
and the Big 5: An empirical indication of structure. Journal of R Development Core Team. (2019). R: A language and environment for
Research in Personality, 42(4), 787–799. https://doi.org/10.1016/j.jrp. statistical computing. R Foundation for Statistical Computing.
2007.10.003 Rosseel, Y. (2012). lavaan: An R package for structural equation model-
Martınez-Martı, M. L., & Ruch, W. (2014). Character strengths and ing. Journal of Statistical Software, 48(2), 1–36. https://doi.org/10.
well-being across the life span: Data from a representative sample of 18637/jss.v048.i02
German-speaking adults living in Switzerland. Frontiers in Ruch, G., Gander, F., Wagner, L., & Giuliani, F. (2021). The structure
Psychology, 5, 1–10. https://doi.org/10.3389/fpsyg.2014.01253 of character: On the relationships between character strengths and
DO STRENGTHS CONVERGE INTO VIRTUES? 13

virtues. The Journal of Positive Psychology, 16(1), 116–113. https:// Individual Differences, 48(6), 714–719. https://doi.org/10.1016/j.paid.
doi.org/10.1080/17439760.2019.1689418 2010.01.007
Ruch, W., & Proyer, R. T. (2015). Mapping strengths into virtues: The Singh, K., & Choubisa, R. (2010). Empirical validation of values in
relation of the 24 VIA-strengths to six ubiquitous virtues. Frontiers action-inventory of strengths (VIA-IS) in Indian context.
in Psychology, 6, 460–460. https://doi.org/10.3389/fpsyg.2015.00460 Psychological Studies, 55(2), 151–158. https://doi.org/10.1007/s12646-
Ruch, W., Proyer, R. T., Harzer, C., Park, N., Peterson, C., & Seligman, 010-0015-4
M. E. P. (2010). Values in action inventory of strengths (VIA-IS): Snow, N. (2019). Positive psychology, the classification of character
Adaptation and validation of the German version and the develop- strengths and virtues, and issues of measurement. The Journal of
ment of a peer-rating form. Journal of Individual Differences, 31(3), Positive Psychology, 14(1), 20–31. https://doi.org/10.1080/17439760.
138–149. https://doi.org/10.1027/1614-0001/a000022 2018.1528376
Schermelleh-Engel, K., Moosbrugger, H., & M€ uller, H. (2003). Terwee, C. B., Crins, M. H. P., Boers, M., de Vet, H. C. W., & Roorda,
Evaluating the fit of structural equation models: Tests of significance L. D. (2019). Validation of two PROMIS item banks for measuring
and descriptive goodness-of-fit measures. Methods of Psychological social participation in the Dutch general population. Quality of Life
Research Online, 8, 23–74. Research, 28(1), 211–220. https://doi.org/10.1007/s11136-018-1995-0
Seibel, B. L., DeSousa, D., & Koller, S. H. (2015). Adaptaç~ao Brasileira Van der Ark, L. A. (2007). Mokken scale analysis in R. Journal of
e Estrutura Fatorial da Escala 240-item VIA Inventory of Strengths. Statistical Software, 20(11), 1–19. https://www.jstatsoft.org/article/
Psico-USF, 20(3), 371–383. https://doi.org/10.1590/1413- view/v020i11
82712015200301 Velicer, W. F., Eaton, C. A., & Fava, J. L. (2000). Construct explication
Shi, D., Maydeu-Olivares, A., & Rosseel, Y. (2020). Assessing fit in through factor or component analysis: A review and evaluation of
ordinal factor analysis models: SRMR vs. RMSEA. Structural alternative procedures for determining the number of factors or
Equation Modeling, 27(1), 1–15. https://doi.org/10.1080/10705511. components. In R. D. Goffin & E. Helms (Eds.), Problems and solu-
2019.1611434 tions in human assessment: Honoring Douglas N. Jackson at seventy.
Shoshani, A., & Slone, M. (2016). The resilience function of character (pp. 41–71). Springer.
strengths in the face of war and protracted conflict. Frontiers in Weber, M., Ruch, W., Littman-Ovadia, H., Lavy, S., & Gai, O. (2013).
Psychology, 6. https://doi.org/10.3389/fpsyg.2015.02006 Relationships among higher-order strengths factors, subjective well-
Shryack, J., Steger, M. F., Krueger, R. F., & Kallie, C. S. (2010). The being, and general self-efficacy - The case of Israeli adolescents.
structure of virtue: An empirical investigation of the dimensionality Personality and Individual Differences, 55(3), 322–327. https://doi.
of the virtues in action inventory of strengths. Personality and org/10.1016/j.paid.2013.03.006

You might also like