Assessment of factor structure of Ryff’s Psychological Well-Being Scales in Italian adolescents

Abstract

Introduction: The multidimensional psychological well-being model proposed by Ryff (RPWB) represents the widely used instrument to assess the human positive functioning by six interrelated dimensions. Previous studies have shown controversial results regarding the latent structure and factorial validity of the model. It was emphasized that, it is necessary to use particular methodological procedures that are appropriate for the treatment of ordinal variables. The aim of the research was to explore the dimensions of psychological well-being of RPWB, within the Italian context. Method: 602 adolescents attending high schools in the Florence area took part in the study. The data gathered with the Italian version of RPWB, were analyzed using polychoric correlations. Exploratory Factor Analyses, on the 54-, 42-, and 18-item versions, were conducted using the Maximum likelihood method for the extraction of factors, the Direct Oblimin for the rotation along with hierarchical analysis of oblique factors. Confirmatory Factor Analyses were conducted on the 18 items version. Results: Comparisons of 9 concurrent models through relative and absolute Goodness of Fit Indices, suggested that the most appropriate solutions
were those with one or two second order factors and five or six first order factors correlated between them. Conclusions: Although, these results confirmed the multidimensionality of the instrument, in particular the substantial validity of the Ryff model; nevertheless further analyses are required in order to better clarify its nature and structure.

Full text

Page 1

Sirigatti et al. (Ryff) 1
Published in final edited form as:
Sirigatti, S., Stefanile, C., Giannetti, E., Iani, L., Penzo, I., & Mazzeschi, A. (2009). Assessment of factor structure of
Ryff’s Psychological Well-Being Scales in Italian adolescents. Bollettino di Psicologia Applicata, 259, 30-50.

Assessment of factor structure of Ryff’s Psychological Well-Being Scales
in Italian adolescents

Saulo Sirigatti*, Cristina Stefanile**, Enrichetta Giannetti**,
Luca Iani*, Ilaria Penzo*, Annamaria Mazzeschi***
* Cognitive and Clinical Psychology Laboratory – European University of Rome
** Department of Psychology – University of Florence
*** Institute of Education – University of London

SUMMARY. Introduction. The multidimensional psychological well-being model proposed by Ryff (RPWB)
represents the widely used instrument to assess the human positive functioning by six interrelated
dimensions. Previous studies have shown controversial results regarding the latent structure and factorial
validity of the model. It was emphasized that, it is necessary to use particular methodological procedures
that are appropriate for the treatment of ordinal variables. The aim of the research was to explore the
dimensions of psychological well-being of RPWB, within the Italian context. Method: 602 adolescents
attending high schools in the Florence area took part in the study. The data gathered with the Italian version
of RPWB, were analyzed using polychoric correlations. Exploratory Factor Analyses, on the 54-, 42-, and
18-item versions, were conducted using the Maximum likelihood method for the extraction of factors, the
Direct Oblimin for the rotation along with hierarchical analysis of oblique factors. Confirmatory Factor
Analyses were conducted on the 18 items version. Results: Comparisons of 9 concurrent models through
relative and absolute Goodness of Fit Indices, suggested that the most appropriate solutions were those with
one or two second order factors and five or six first order factors correlated between them. Conclusions:
Although, these results confirmed the multidimensionality of the instrument, in particular the substantial
validity of the Ryff model; nevertheless further analyses are required in order to better clarify its nature and
structure.

RIASSUNTO. Introduzione: Il modello multidimensionale di benessere psicologico elaborato da Ryff
(RPWB) rappresenta lo strumento maggiormente impiegato per valutare il funzionamento umano positivo
mediante sei dimensioni tra loro correlate. Gli studi precedenti hanno fornito risultati controversi circa la
struttura latente e la validità fattoriale del modello. E’ stata sottolineata la necessità di impiegare procedure
statistiche appropriate per il trattamento di variabili ordinali. Lo scopo del presente lavoro è stato di
esplorare le dimensioni del benessere psicologico del RPWB all’interno del contesto italiano, analizzandone
le proprietà psicometriche. Metodo: Hanno partecipato 602 adolescenti fiorentini frequentanti scuole medie
superiori. I dati, raccolti con la versione italiana del RPWB, sono stati analizzati basandosi su correlazioni
policoriche. Analisi fattoriali esplorative – su versioni di 54, 42 e 18 item – sono state condotte utilizzando
per l’estrazione dei fattori il metodo “Maximum Likelihood” e per la rotazione il “Direct Oblimin”, nonché
l’analisi gerarchica dei fattori obliqui. Le analisi fattoriali confermative sono state eseguite sulla versione di
18 item. Risultati: La comparazione di nove modelli concorrenti, attraverso indici relativi e assoluti di
adeguatezza, ha indicato come più adeguate le soluzioni che prevedevano uno o due fattori di secondo
ordine e cinque o sei fattori di primo livello tra loro correlati. Conclusioni: Sebbene questi risultati
confermino la multidimensionalità dello strumento e, in particolare, la sostanziale validità del modello
proposto da Ryff, ulteriori approfondimenti sono necessari per chiarirne maggiormente la natura e la
struttura.

Keywords: Psychological well-being; Structural validity; Measurement

Page 2

Sirigatti et al. (Ryff) 2
1. Introduction
In the last four decades research on well-being has been focused on two different traditions (Ryan &
Deci, 2001; Waterman, 1993): the hedonic well-being (Kahneman, Diener & Schwarz, 1999) or Subjective
Well-Being (SWB), and the eudaimonic well-being (Ryff, 1989; Ryff & Keyes, 1995), or Psychological
Well-Being (PWB). Both models are not new and were originally grounded in ancient philosophical works.
The foundations of hedonic tradition can be found in ancient Greece, where Aristippus urged to experience
the greatest possible pleasure. Instead, the concept of eudaimonia was formulated by Aristotle that
denigrated the fulfillment of pleasures in and of itself, and believed that happiness was based on the
expression of virtues, and on living in accordance with one’s daimon (Waterman, 1993), namely unique
capacities, which has to be recognized and achieved during life.
SWB has been defined as a person’s cognitive and affective evaluation of his or her own life
(Diener, Lucas & Oishi, 2002). The cognitive component refers to the individual’s evaluation of life
satisfaction, whereas the affective component refers to the presence of positive affect and relative lack of
negative emotions over time. In order to assess SWB, scientists have created several scales that represent
different conceptualizations of well-being, such as the Life Satisfaction Index (Neugarten, Havighurst &
Tobin, 1961), the Bradburn’s Affect Balance Scale (Bradburn, 1969), the Satisfaction With Life Scale
(Diener, Emmons, Larsen & Griffin, 1985), the Positive and Negative Affect Schedule (Watson, Clark &
Tellegen, 1988) and the more recent Subjective Happiness Scale (Lyubomirsky & Lepper, 1999), which
encompasses both affective and cognitive components of hedonic well-being.
Alternatively, PWB model refers to the human potentials and resources to reach optimal functioning
(Ryff, 1989; Ryff & Keyes, 1995). The growing interest in PWB has required the theoretical development of
a new construct as well as valid and reliable instruments to operationalize this concept in both clinical and
population samples (Abbott Ploubidis, Huppert, Kuh, Wadsworth & Croudace, 2006; Keyes & Lopez, 2002).
A widely used instrument in this field is represented by the Ryff’s Psychological Well-Being Scales
(RPWB), developed on the basis of an extensive literature review, and on the integration of mental health,
clinical and life span developmental theories (Ryff, 1989; Ryff & Keyes, 1995). In particular, the process of
individuation by Jung (1933), the model of psychosocial stages by Erikson (1959), the formulation of
maturity by Allport (1961), the description of the fully functioning person by Rogers (1961), and the theory
of self-actualization by Maslow (1968) served as theoretical frameworks to generate a model of positive
psychological functioning. The multidimensional model included six dimensions: a positive attitude toward
the self and one’s past life (Self-Acceptance), the ability to have an open and satisfying relationship with
others (Positive Relations with Others), a sense of independence and self-determination (Autonomy), the
competence to manage the environment and external activities (Environmental Mastery), believing that there
is a meaning to one’s life (Purpose in Life), and a positive attitude to new experiences (Personal Growth)
(Ryff, 1989; Ryff & Keyes, 1995).
Ryff’s PWB has been widely studied across different cultural contexts and settings with samples of
different sizes and socio demographic features. Furthermore, while most studies have used a 18-item version,

Page 3

Sirigatti et al. (Ryff) 3
a number of surveys have also employed the versions with 4-items, 7-items, 9-items, or 14-items. Other
studies have also used versions with different numbers of items such as the one proposed by Van
Dierendonck (2004) with 39 items (Abbott, et al., 2006; Abbott, Ploubidis, Huppert, Kuh & Croudace, 2009;
Burns & Machin, 2009; Cheng & Chan, 2005; Clarke, Marshall, Ryff & Wheaton, 2001; Kafka & Kozma,
2002; Ryff, 1989; Ryff & Keyes, 1995; Springer & Hauser, 2006; Van Dierendonck, Diaz, Rodriguez-
Carvajal, Blanco & Moreno-Jiménez, 2008).
Although earlier studies confirmed the proposed structure (Ryff, 1989; Ryff & Keyes, 1995), more
recent studies showed that its latent structure and factorial validity remain scarce and contentious (Abbott et
al., 2006; Kafka & Kozma, 2002). To date, six factor analytic studies of RPWB have been published and
only one (Kafka & Kozma, 2002) failed to provide support for the six factor model, but that study is not
credible, given the size of the sample relative to the number of RPWB item factor analyzed and the use of
principal components analysis with Varimax rotation (Ryff & Singer, 2006). All remaining five studies used
confirmatory factor analytic procedures (CFA) to test the factorial validity of the RPWB. All findings
supported the evidence that the six factor model is the best fitting model. Other studies have been published
since 2006, in addition to those mentioned by Ryff and Singer (2006): several psychometric studies of the
RPWB – ranging from 1989 to 2009 – are summarized in Table 1.
Among the recent studies those carried out by Van Dierendonck and colleagues (2008), and by
Burns and Machin (2009) might be mentioned. In the first one, utilizing confirmatory factor analysis, one-,
two-, three- and six-factor models were examined. Notwithstanding that four out the six dimensions
overlapped to a great extent, the six factor model, with a single second order factor, was the best fitting
model. Therefore, obtained results confirmed a six dimension model identified by Ryff and Singer (2006),
while they did not support Springer and Hauser’s position (2006).
Burns and Machin (2009) investigated the 6-factor RPWB model factor structure using data from a
life events study and an organizational climate study. Confirmatory Factor Analysis (CFA) of the item
identified in the initial Exploratory Factor Analysis (EFA) was undertaken to assess whether the RPWB
model identified in the EFA reported better Goodness of Fit Indices than the a-priori factor model, and a
number of alternative models identified in the previous literature. Despite the encouraging results, the
complexity of the problem prevented any satisfactory conclusion about the structural validity of RPWB. The
Authors believed that further development of RPWB measures is needed to reflect its hierarchical and multi-
dimensional nature, as much as with the scales’ current form, the construct validation of RPWB factors
would continue being problematic.
In recent years, the focus on positive human functioning has allowed to carry out several
investigations in the field, including Italy. Ryff’s PWB was also widely used in both clinical and general
populations as well as in different item versions and Likert response scale. The RPWB is often used in a
battery of self-assessment questionnaires to measure different variables, and other times this is part of a
broader research aimed at assessing the psychological well-being and its related and with the purpose of
identifying risk and protective factors, useful information for developing interventions for prevention and

Page 4

Sirigatti et al. (Ryff) 4
health promotion (Rafanelli, Park, Ruini, Ottolini, Cazzaro & Fava, 2000; Fava, Rafanelli, Ottolini, Ruini,
Cazzaro & Grandi, 2001; Steca, Ryff, D’Alessandro & Delle Fratte, 2002; Ruini, Ottolini, Rafanelli, Ryff &
Fava, 2003; Giannetti, Penzo, Rigoli & Sirigatti, 2006; Giannetti, Penzo, Hatalskaya & Sirigatti, 2008; Ruini
et al., 2009). Regarding statistical analysis, the authors made correlations between the six dimensions of the
model proposed by Ryff and other indicators of well-being, and the coefficients of internal consistency of the
scales. In particular, in the Italian validation of RPWB by Ruini et al. (2003), test-retest and internal
correlation coefficients resulted significant for the six scales, replicating prior studies. To date, neither EFA
nor CFA have been conducted with Italian samples and versions.
The majority of the studies concerning the structure of Ryff’s PWB has been based on the English
version of the questionnaire, therefore it is uncertain as to whether such findings are applicable to other
populations using translated versions of the scales. Moreover, several models have been evaluated using
different methodological and estimation procedures, making any comparison of results rather hard to
manage. The contradictory results suggest the deepening of the construct examination through further
psychometric investigations in order to explore the dimensionalities of Ryff’s PWB, taking into specific
attention Italian versions and samples. On the base of previous studies and recommendations, some
methodological choices appeared necessary. Given the modalities of response to RPWB questionnaire, the
measured variables were ordinal, and could not be treated as they were continuous; for this reason the
estimation of interitem polychoric correlations was preferred (Holgado, Chacón, Barbero & Vila-Abad,
2010). The indications obtained from EFA contributed to the identification of the items to be included in the
Italian versions of RPWB, while CFA allowed to compare a range of structural models on the basis of
several Goodness of Fit indices.

2. Method
2.1. Participants
A total of 602 adolescents took part in the study voluntarily, and they were not provided with any
incentives, monetary or otherwise, for their participation. There were 17% of male and 83% of female
adolescents. 14% of the participants was between the ages of 13 and 14, 52% between 15 and 16, 30%
between 17 and 18 and the remaining 4% was 18 or older. The majority (87%) was born in Florence or in its
surrounding areas and almost the totality of participants was residing in Florence or in its vicinity; all
participants attended high schools in Florence.

2.2. Measures
The Italian version of Carol Ryff’s (1989) Psychological Well-Being Scales was administered to
participants. The Italian adaptation, carried out by Ruini and colleagues (2003), consisted in a 84 item self-
rating inventory assessing six dimensions: Self-Acceptance (SA), Positive Relations with Others (PR),
Autonomy (AU), Environmental Mastery (EM), Purpose in Life (PL), and Personal Growth (PG). Items
illustrating each dimension are, e.g., as follows: ‘‘In general, I feel confident and positive about myself”

Page 5

Sirigatti et al. (Ryff) 5
(SA), “People would describe me as a giving person, willing to share my time with others” (PR), ‘‘I tend to
be influenced by people with strong opinions’’ (AU), ‘‘The demands of everyday life often get me down”
(EM), ‘‘Some people wander aimlessly through life but I am not one of them” (PL), and ‘‘I do not enjoy
being in new situations that require me to change my old familiar ways of doing things” (PG).
Participants indicated their response on a four point Likert-type scale from 1 (disagree strongly) to 4
(agree strongly), with higher scores on each scale indicating greater well-being on each dimension. Ruini and
colleagues (2003) administered the questionnaire to a sample of 415 people, and reported the following test-
retest correlation coefficients: SA: r = .82; PR: r = .81; AU: r = .21; EM: r = .31; PL: r = .81; PG: r = .78.

2.3. Procedure
These data were collected as a part of a larger study that examined possible influence of personal and
socio-cultural factors on psychological well-being. The data were gathered collectively during normal class
time; beforehand the aims of the investigation and a brief explanation of appropriate response procedures
were thoroughly explained. Confidentiality was guaranteed to the students as far as the test contents were
concerned. Consent was obtained from participants, parents or guardians.

2.4. Data Analysis
Even though the original Ryff’s PWB included 120 items (20 per dimension), shorter versions
comprising 84 items (14 per dimension), 54 items (9 per dimension), 42 items (7 per dimension), and 18
items (3 per dimension) are now widely used. Such suggestions, integrated with the results obtained by the
EFA of the responses given to the 84 item Italian version of RPWB, allowed the building of three shorter
versions – comprising of 54, 42, and 18 items – utilized for the further statistical computations.
The sample was randomly divided into two groups, using an odd-even split procedure. Data
inspection revealed missing data in both sub-samples (first one: .095%; second one: .036%). Therefore, the
method suggested in the LISREL program (Jöreskog & Sörbom, 1996a; 1996b), namely to impute real
values from another case with similar observed values, was applied. When the search for similar response
pattern imputation produced no results, the listwise deletion was used. At the end, the first sub-sample –
utilized for the subsequent EFA – included 296 participants; the second one – used for the CFA – comprised
of 298 people.
Evidence of construct validity was sought by carrying out factor analysis, both EFA and CFA, on a
matrix of interitem polychoric correlations, on the account that items might be considered as categorized
continuous variables from a normal multivariate distribution. Once the matrix of polychoric correlations had
been estimated, the supposition of bivariate normality was tested, by calculating the percentage of tests that
rejected the null hypothesis of bivariate normality for each pair of correlations.
With the aim of exploring how the RPWB items empirically clustered, an EFA was carried out,
entering the matrix of polychoric correlations into SPSS and STATISTICA programs, using Maximum
Likelihood as the estimation method, along with Oblimin rotation with Kaiser normalization, and

Page 6

Sirigatti et al. (Ryff) 6
hierarchical factor analysis. These EFA, which were the first analysis conducted in the Italian adaptation of
the RPWB, took into consideration 54, 42, 18 item versions. These versions were suggested by the
examination of the pertinent literature (Abbott, et al., 2006; Burns & Machin, 2009; Kafka & Kozma, 2002;
Lindfors, Berntsson & Lundberg, 2006; Springer & Hauser, 2006; Springer, Hauser & Freese, 2006; Van
Dierendonck et al., 2008), and by the indications obtained with exploratory factor analyses. Following the
hints of prior investigations, a six factor model was initially examined, other solutions were subsequently
examined.
Several CFA were performed on the second trial randomly selecting of the participants, in order to
test the viability of the hypothesized structure that had been formulated from theoretical considerations, and
results of the EFA. The goodness of fit of confirmatory factor models with polychoric correlations and
asymptotic covariance matrix was assessed by means of the weighted least squares method, implemented in
the LISREL program (Jöreskog & Sörbom, 1996b). The goodness of fit indices were computed only for the
18-item version, as much as the minimum sample size for estimating asymptotic covariance matrix required
for the other versions was not reached. In particular, the following models were assessed:
(1) one-factor model, which assumes that all items load on a general well-being factor;
(2) a two-factor model, in which Self-Acceptance, Autonomy, Environmental Mastery, and Purpose
in Life load on one factor, and Positive Relations with Others and Personal Growth load on a second factor;
(3) a three-factor model, in which Self-Acceptance, Environmental Mastery, Purpose in Life, and
Personal Growth load on one factor, Personal Relations on the second factor, and Autonomy on the third
factor;
(4) a five-factor model, in which Environmental Mastery, and Purpose in Life load on the same
factor;
(5) the six-factor structure as described by Ryff;
(6) a six-factor structure, and one second-order latent construct;
(7) a five-factor structure, as formulated in (4), and one second-order latent construct;
(8) a six-factor structure, and two second-order latent constructs, one of them comprising Self-
Acceptance, Autonomy, Environmental Mastery and Purpose in Life, and the other one Positive Relations
with Others and Personal Growth;
(9) a five-factor structure, as formulated in (4), and two second-order latent constructs, one of them
comprising Self-Acceptance, Autonomy, Environmental Mastery and Purpose in Life, and the other one
Positive Relations with Others and Personal Growth.

3. Results
Of the estimated 3486 polychoric correlations in RPWB, the test of bivariate normality produced the
results shown in Table 2.

Page 7

Sirigatti et al. (Ryff) 7
Therefore, we concluded that the model of categorized multivariate normality is appropriate for these
data, and that the use of polychoric correlations is justified.

3.1. Exploratory Factor Analysis
In order to explore the latent structure of Ryff’s Psychological Well-Being model, EFA has been
conducted using Maximum Likelihood as the estimation method, along with Oblimin rotation with Kaiser
normalization, and hierarchical factor analysis. The 54-, 42-, and 18- item versions were assessed through
first– and higher- order factor analysis.
First-order factor analysis. Analyses, concerning 6- and 5-factor solutions for the 54 item version
(see Appendix A), showed a cumulative explained variance, equal, respectively, to 47.48% and to 45.49%.
The 6-factor analysis allowed to identify the following factors: factor 1, comprising, among others, all items
relating to Purpose in Life (9 loadings out of 22); factor 2, included 8 items from Positive Relations with
Others (8 out of 12); on factor 4, 5 and 6 loaded items concerning Autonomy (7 out of 8), Self-Acceptance (9
out of 14), and Personal Growth (9 out of 13). Only 2 items loaded on factor 3. With regard to the
Environmental Mastery dimension, there was a high dispersion of saturations on various factors . Moreover,
item #14 “Non sto bene con la gente e nell’ambiente che mi circonda (I do not fit very well with the people
and community around me)” loaded on Positive Relations with Others rather than on the proposed dimension
of Environmental Mastery (see Table 3).
Similar results were obtained when the 5-factor solution was taken into consideration, but items
loaded rather often accordingly to hypothesized dimensions, exhibiting a low scattering of saturations along
the five factors (see Table 4). Of the 18
items that seemed to represent both Purpose in Life and Environmental Mastery dimensions, 15 out of 22
loaded on the first factor. Factor 2 comprised items relating to Positive Relations with Others (8 out of 12),
while items concerning Personal Growth (7 out of 8), Autonomy (8 out of 10) and Self-Acceptance (9 out of
14) loaded, respectively, on factor 3, 4 and 5. With the 42 item version (see Appendix A), the cumulative
percentage of explained variance was equal to 47.75 (see Table 5).
The EFA carried out on this version yielded the following 5-factor solution: a first factor included 14
out of 20 items from both Environmental Mastery and Purpose in Life, while the items loaded on factor 2 (7
out of 9) were related to hypothesized Personal Growth dimension. Moreover, the items loading on factors 3,
4 and 5 concerned Positive Relations with Others (7 out of 8), Autonomy (7 out of 8), and Self-Acceptance (7
out of 9). This solution appeared to be more adequate then the previous ones, as much as the items belonging
to different scales tended to load on specific factors, and items being part of the same scale less frequently
loaded on different factors.
With 18 item version (see Appendix A and B), the cumulative percentage of explained variance was
equal to 50.93 (see Table 6).
The EFA performed on this version output the following 5-factor solution: a first factor included 6
out of 6 items from both Environmental Mastery and Purpose in Life, while the items loaded on factor 2

Page 8

Sirigatti et al. (Ryff) 8
concerned Positive Relations with Others (3 out of 3). Moreover, the items loading on factors 3, 4, and 5
were related to hypothesized dimensions: Autonomy (3 out of 3), Personal Growth (3 out of 4) and Self-
Acceptance (3 out of 5). This solution appeared to be rather appropriate, as much as the items belonging to
different scales tended to load on distinct factors, and items being part of the same scale rarely loaded on
different factors.
Hierarchical factor analysis. The multidimensional psychological well-being model proposed by
Ryff (1989) has been subsequently investigated using EFA with hierarchical factor analysis. The 54-, 42-,
and 18-item versions were tested. Results obtained with 54- and 42- items showed a diffuse item cross-
loading, making rather difficult and unclear the interpretation of the results. In the 18-item version, asking
for the extraction of 5 first-order factors, a two second-order factors solution resulted; the cumulative
percentage of explained variance was equal to 51.60 (see Table 7). Two second-order general factors
emerged quite clearly; one concerning mainly Self-Acceptance, Autonomy, Environmental Mastery, and
Purpose in Life, another regarding Positive Relations with Others, and Personal Growth.
The five unique factors showed the following specificities: items concerning Environmental Mastery
and Purpose in Life loaded on Primary 1, Positive Relations with Others loaded on Primary 2, Autonomy on
Primary 3, Self-Acceptance on Primary 4, and Personal Growth on Primary 5.

3.2. Confirmatory factor analysis
Using the second sub-sample, confirmatory factor analysis of the 18 items identified by examining
the pertinent literature, and by the previous EFA, was carried out in order to test the nine different competing
models, formerly mentioned. Table 8 presents the goodness of fit indices concerning models from 1 to 5.
A clear trend appeared: as the number of factors included in the factorial structure increased the
adequacy of the solution to the empirical data bettered. Therefore the six factor model (model 5) came out as
the best fitting model, even though differences between models 4 and 5 were minimal, as far as some
goodness of fit indices were concerned. Notwithstanding that the AIC and the ECVI, corresponding to model
5, reached the lowest values of the first group of models, and the GFI, AGFI, but not the CFI, were higher
than .90, and PGFI was higher than .65, the fit could not be considered completely satisfying, particularly
when models from 6 to 9 were taken into account (see Figures 1 and 2).
Models 6-9 assessed the goodness of fit of a hierarchical structure comprising five or six factors at
the first level, and one or two second-order latent variables (see Table 9). The four models proved to be a
rather satisfactory representation of the observed data (see Figures 3 and 4). Although there were
discrepancies between estimates and observations, the NFIs of about .94, the AGFIs of .95, the GFIs of .96,
RMSEAs of .035, the RMRs of .065, and CFIs of .97 indicated adequate fit for all four models, showing fits
more close than the previous five factorial structures.
Also the indices AIC (ranging from 258 to 260) and ECVI of about .87 indicated that the fit of
models 6-9 was better in comparison of those 1-5, and the goodness of fit of the structural models 6-9 was
substantially equivalent. In the attempt to identify, among these good fitting models, the best solution, the

Page 9

Sirigatti et al. (Ryff) 9
Parsimonious Goodness-of-Fit Index (PGFI; Mulaik, James, Van Alstine, Bennett, Lind & Stilwell, 1989)
was calculated; however the value of such an index resulted for all these four models equal to about .73.
Only theoretical considerations might help selecting the most adequate structural solution.

4. Discussion and conclusions
The factorial structure of the RPWB scales has been explored across different population samples,
cultural contexts and versions. The original English version has been translated into several other languages,
including Italian (Ruini et al., 2003). However, to date, neither EFA nor CFA have been conducted with
Italian samples and versions. The aim of this study was to make available the first exploration of the
proposed multidimensional model of psychological well-being by Ryff (1989) in the Italian context.
So far, controversial results were obtained regarding the dimensionality of the theoretical models of
RPWB. As the data was collected by administering a questionnaire, part of such disagreements might be
ascribed to the different statistical procedures adopted for data analysis. Following leads derived from recent
investigations, on this occasion polychoric correlation estimates were preferred to the more used – at least in
the recent past – calculation of Bravais-Pearson correlation coefficients. As a matter of fact, recent analyses
of the RPWB scales, carried out in diverse studies (Abbott et al., 2006; Springer & Hauser, 2006; Van
Dierendonck et al., 2008), in calculating polychoric correlation estimates, because previous methods did not
take into account, perhaps incorrectly, that RPWB responses were categorical, and obtained from Likert
scales. In any case, when polychoric correlations were estimated, the factor results – obtained from both an
EFA and a CFA – reproduced better the measurement model that is present in the data, irrespective of the
number of factors (Holgado et al., 2010).
The critical comment, that Ryff and Singer (2006) directed at the investigation conducted by Kafka
and Kozma (2002), suggested that one should avoid performing EFA using principal components analysis
with Varimax rotation, in consideration that aim is the identification of the latent constructs, and that the six
dimensions are intercorrelated. For this reason, the latent structure was searched by means of Maximum
likelihood technique, and the best solution was looked for by means of an oblique rotation.
Using an Italian version of the RPWB scales, the present investigation provided some further
confirmation to the previously reported factorial structure of the original version of the scales. This assertion
particularly suited the 18 item form, built on the base of the literature suggestions, and of the results of
preliminary EFAs. Solutions implying second-order factor and correlated first-order factor models were
superior to other estimated models. The selection of the best among these four models (from 6 to 9) was not
possible basing on the goodness of fit indices; even the parsimony criterion offered no assistance. The results
of previous studies might orient the choice towards the model admitting a second-order single factor model,
and six primary factors representing the Ryff’s PWB scales. This model showed dramatic improvement in fit
over suggested alternatives, especially the single-factor model.
These results appeared to be in substantial agreement with the studies conducted by others, such as
Ryff and Keyes (1995), Clarke et al. (2001), Van Dierendonck (2004), Chen and Chan (2005), Springer and

Page 10

Sirigatti et al. (Ryff) 10
Hauser (2006), and Lindfors et al. (2006). Using CFA procedures, provided evidence that theory-guided six-
factor model was the best-fitting model. Moreover, Ryff and Singer (2006) mentioned other studies outlining
remarkable results.
Other studies have brought this model under serious debate. For instance, Springer and Hauser
(2006) addressed serious doubts about the tenability of the six-factor model. Springer, Hauser and Freese
(2006) added that there was compelling empirical evidence of theoretical and measurement problems with
RPWB six-factor model of psychological well-being, and that four of the six RPWB factors were virtually
indistinguishable. Their results suggested that four of the six dimensions (such as: Personal Growth, Purpose
in Life, Self-Acceptance, and Environmental Mastery) empirically might be only one dimension. The high
correlation between these four factors was also observed in a subsequent study (Burns & Machin, 2009).
This study presented several limitations. The model was not assessed employing a nationally
representative sample, but a convenience sample of Italian high school students. The socio-cultural
characteristics of the sample – as far as age, education, occupation are concerned – were quite homogeneous,
therefore the generalization to other groups was difficult. The size of the sample did not allow the CFA of
54-, and 42-item versions. A four point Likert scale was used, instead of the six point one, as traditionally
done, making arduous comparisons with the results of other researches.
In conclusion, confirmatory factor analyses with data from a sample of Italian adolescents supported
the multidimensional structure of the Ryff’s PWB. It seemed that several versions might hold across
different formats, languages, countries and cultures. The present investigation supported a second-order
factor and correlated first-order factor model. Nevertheless some further analyses in order to better clarify
the nature and the dimensions of RPWB are required. As Abbott et al. (2009) noticed, interest in the
assessment of well-being is increasing among researchers, practitioners and policy makers; hence, it would
be essential to have appropriate instruments to measure well-being, effectively and across its full spectrum.

References
ABBOTT, R.A., PLOUBIDIS, G.B., HUPPERT, F.A., KUH, D. & CROUDACE, T.J. (2009). An evaluation
of the precision of measurement of Ryff’s Psychological Well-Being Scales in a population sample.
Social Indicators Research, DOI 10.1007/s11205-009-9506-x.
ABBOTT, R.A., PLOUBIDIS, G.B., HUPPERT, F.A., KUH, D., WADSWORTH, M.E.J. & CROUDACE,
T.J. (2006). Psychometric evaluation and predictive validity of Ryff's psychological well-being items
in a UK birth cohort sample of women. Health and Quality of Life Outcomes, 4, 76 (October).
ALLPORT, G.W. (1961). Pattern and growth in personality. New York: Holt, Rinehart & Winston.
BRADBURN, N.M. (1969). The structure of psychological well-being. Chicago, IL: Aldine Publishing.
BURNS, R.A. & MACHIN, M.A. (2009). Investigating the structural validity of Ryff’s psychological well-
being scales across two samples. Social Indicators Research, 93(2), 359-375.
CHENG, S.T. & CHAN, A.C. (2005). Measuring psychological well-being in the Chinese. Personality and
Individual Differences, 38(6), 1307-1316.
CLARKE, P.J., MARSHALL, V.W., RYFF, C.D. & WHEATON, B. (2001). Measuring psychological well-
being in the Canadian study of health and aging. International Psychogeriatrics, 13(supp 1), 79–90.

Page 11

Sirigatti et al. (Ryff) 11
DIENER, E., LUCAS R.R. & OISHI S. (2002). Subjective Well-Being. The science of happiness and life
satisfaction (pp. 63-73). In C.R. SNYDER & S.J. LOPEZ (Eds.), Handbook of Positive Psychology.
Oxford: Oxford University Press.
DIENER, E., EMMONS, R., LARSEN, R. & GRIFFIN, S. (1985). The Satisfaction with Life Scale. Journal
of Personality Assessment, 49(1), 71-75
ERIKSON, E. (1959). Identity and the life cycle. Psychological Issues, 1, 18-164.
FAVA, G. A., RAFANELLI, C., OTTOLINI, F., RUINI, C., CAZZARO, M. & GRANDI, S. (2001).
Psychological well-being and residual symptoms in remitted patients with panic disorder and
agoraphobia. Journal of Affective Disorders, 65(2), 185–190.
GIANNETTI, E., PENZO, I., HATALSKAJA, H. & SIRIGATTI, S. (2008). Psichologiceskoe blagopolucie,
sozialnaja i ekonomiceskaja podderchka i udovletvorennost shisniu italianskich i belorusskich
iunosheiji i devuschek. [Psychological well-being, life satisfaction, economical and social support in
Belarus and Italian adolescents]. In V.V. Grizenko, L.L. Dikevich, O.N. Kapirenkova, O.A. Lapshova,
N.V. Molchanova, I.M. Osipenko, N.N. Pletnevskaja, E.M. Turok & N.V. Zelueva (Eds.)
Teoreticeskie problemi etniceskoji i kross-kulturnoji psichologii [Theoretical problems of
etnopsychology and cross-cultural psychology] (pp. 182-188). Smolensk (Russia): Universum.
GIANNETTI, E., PENZO, I., RIGOLI, S. & SIRIGATTI, S. (2006). Emotional intelligence and college
maladjustment as correlates of psychological well-being: preliminary research. In V.V. SEMENOVA
& G.V. LOSIK (Eds.). Sovremennie podchodi k prodvijeniju zdorovija. [Modern approach to health
promotion.] (pp. 38-42). Gomel (Belarus): Gomel State Medical University.
HOLGADO, F.P., CHACÓN, S., BARBERO, I. & VILA-ABAD, E. (2010). Polychoric versus Pearson
correlations in exploratory and confirmatory factor analysis of ordinal variables. Quality & Quantity:
International Journal of Methodology, 44(1), 153-166.
JÖRESKOG, K.G. & SÖRBOM, D. (1996a). PRELIS 2: User’s reference guide. Chicago, IL: Scientific
Software International.
JÖRESKOG, K.G. & SÖRBOM, D. (1996b). LISREL 8. User’s reference guide. Chicago, IL: Scientific
Software International.
JUNG, C.G. (1933). Modern man in search of a soul (W.S. Dell & C.F. Baynes, Trans.). New York:
Hartcourt, Brace & World.
KAFKA, G.J. & KOZMA, A. (2002). The construct validity of Ryff’s scales of psychological well-being
(SPWB) and their relationship to measures of subjective well-being. Social Indicators Research,
57(2), 171–190.
KAHNEMAN, D., DIENER, E. & SCHWARZ, N. (Eds.) (1999). Well-being: the foundations of hedonic
psychology. New York: Russell Sage Foundation.
KEYES, C.L.M. & LOPEZ, S.J. (2002). Toward a science of mental health. Positive directions in diagnosis
and interventions. In C.R. Snyder e S.J. Lopez (Eds.), Handbook of positive psychology. New York:
Oxford University Press.
LINDFORS, P., BERNTSSON, L. & LUNDBERG, U. (2006). Factor structure of Ryff’s psychological well-
being scales in Swedish female and male white-collar workers. Personality and Individual
Differences, 40(6), 1213-1222.
LYUBOMIRSKY, S. & LEPPER, H. (1999). A measure of subjective happiness: Preliminary reliability and
construct validation. Social Indicators Research, 46(2), 137-155.
MASLOW, A.H. (1968). Toward a psychology of being (2nd ed.). New York: Van Nostrand.
MULAIK, S.A., JAMES, L.R., VAN ALSTINE, J., BENNETT, N., LIND, S. & STILWELL, C.D. (1989).
Evaluation of goodness-of-fit indices for structural equation models. Psychological Bulletin, 105(3),
430-445.
NEUGARTEN, B.L., HAVIGHURST, R. & TOBIN, S. (1961).The measurement of life satisfaction.
Journal of Gerontology, 16, 134-143.

Page 12

Sirigatti et al. (Ryff) 12
RAFANELLI, C., PARK, S. K., RUINI, C., OTTOLINI, F., CAZZARO, M. & FAVA, G.A. (2000). Rating
well-being and distress. Stress Medicine, 16(1), 55-61.
ROGERS, C.R. (1961). On becoming a person. Boston: Houghton Mifflin.
RUINI, C., OTTOLINI, F., RAFANELLI, C., RYFF, C. & FAVA, G.A. (2003). La validazione italiana delle
Psychological Well-Being Scales (PWB). Rivista di Psichiatria, 38(3), 117-130.
RUINI, C., OTTOLINI, F., TOMBA, E., BELAISE, C., ALBIERI, E., VISANI, D., OFFIDANI, E.,
CAFFO, E. & FAVA, G.A. (2009). School intervention for promoting psychological well-being in
adolescence. Journal of Behavior Therapy and Experimental Psychiatry, 40(4), 522-532.
RYAN, R.M. & DECI, E.L. (2001). On happiness and human potentials: a review of research on hedonic
and eudaimonic well-being. Annual Review of Psychology, 52, 41-66.
RYFF, C.D. (1989). Happiness is everything, or is it? Explorations on the meaning of psychological well-
being. Journal of Personality and Social Psychology, 57(6), 1069-1081.
RYFF, C.D. & KEYES, C.L. (1995). The structure of psychological well-being revisited. Journal of
Personality and Social Psychology, 69(4), 719-727.
RYFF, C.D. & SINGER, B.H. (2006). Best news yet on the six-factor model of well-being. Social Science
Research, 35(4), 1103-1119.
SCHERMELLEH-ENGEL, K., MOOSBRUGGER, H. & MÜLLER, H. (2003). Evaluating the fit of
structural equation models: Tests of significance and descriptive goodness-of-fit measures. Methods of
Psychological Research Online, 8(2), 23-74.
SPRINGER, K.W. & HAUSER, R.M. (2006). An assessment of the construct validity of Ryff’s scales of
psychological well-being: Method, mode, and measurement effects. Social Science Research, 35(4),
1080-1102.
SPRINGER, K. W., HAUSER, R. M. & FREESE, J. (2006). Bad news indeed for Ryff’s six-factor model of
wellbeing. Social Science Research, 35(4), 1120-1131.
STECA, P., RYFF, C.D., D’ALESSANDRO, S. & DELLE FRATTE, A. (2002). Il benessere psicologico:
Differenze di genere e di età nel contesto italiano. Psicologia della Salute, 2, 121-143.
VAN DIERENDONCK, D. (2004). The construct validity of Ryff’s scales of psychological well-being and
its extension with spiritual well-being. Personality and Individual Differences, 36(3), 629-643.
VAN DIERENDONCK, D., DIAZ, D., RODRIGUEZ-CARVAJAL, R., BLANCO, A. & MORENO-
JIMÉNEZ, B. (2008). Ryff’s six-factor model of psychological well-being. A Spanish exploration.
Social Indicators Research, 87(3), 473-479.
WATERMAN, A.S. (1993). Two conceptions of happiness: Contrasts of personal expressiveness
(eudaimonia) and hedonic enjoyment. Journal of Personality and Social Psychology, 64(4), 678-691
WATSON, D., CLARCK, L.A. & TELLEGEN, A . (1988). Development and validation of brief measures of
positive and negative affect: the PANAS scales. Journal of Personality and Social Psychology, 54(6),
1063-1070.

Page 13

Sirigatti et al. (Ryff) 13
Table 1
Summary of psychometric studies of RPWB: sample characteristics and item-version (derived from original sources)

Author N. of item Sample
Ryff (1989)
120
(20-item scale)
Sample size: 321

Multiple groups selected from educational institution and community volunteers (191 women and 129 men), divided into
three age groups:
- young adults (N = 133); mean age = 19.53 years (SD = 1.57);
- middle-aged adults (N = 180); mean age = 49.85 years (SD = 9.35);
- older adults (N = 80); mean age = 74.96 years (SD = 7.11).
Ryff & Keyes (1995)
18
(3-item scale)
Sample size: 1108

Women = 59%; mean age = 45.6 years (SD = 14.8)

Nationally representative sample non institutionalized, who participated in telephone interviews selected with random digit
dialing procedures, divided into three age groups:
- young adults (N = 133) were between the ages of 25 and 29;
- midlife-aged adults (N = 805) were between the ages of 30 and 64;
- older adults (N = 160) were 65 or older.
Clarke, Marshall, Ryff &
Wheaton (2001)
18
(3-item scale)
Sample size: 4960

Canadian Study of Health and Aging (CHSA)
Seniors without severe cognitive impairment or dementia
Women = 58%; mean age = 75.5 years (SD = 5.2)
Kafka & Kozma (2002)
120
(20-item scale)
Sample size: 277

Canadian students (Female = 186; Male = 91)
Mean age = 21.3 years (SD = 3.8), range: 18-48 years
(continued on next page)

Page 14

Sirigatti et al. (Ryff) 14
Table 1 (continued)
Author N. of item Sample
Van Dierendonck (2004)
84
54
18
(14-, 9-, 3-item scale)
Study 1
Sample size: 233

Dutch first year undergraduate psychology students (Female = 156; Male = 77) who participated with the research for
course credit
Mean age = 22.0 years (SD = 6.0)
54
18
(9-, 3-item scale)
Study 2
Sample size: 420

Dutch professional (Female = 130; Male = 290)
Mean age = 36.0 years (SD = 8.0)
Cheng & Chan (2005)
24
(4-item scale)
Sample size: 1259

Chinese volunteers at 20 hospitals randomly sampled from 40 public hospitals all over Hong Kong
Mean age = 44.7 years (SD = 16.6), range 18-86 years; predominantly female (81.6%)
Abbott, Ploubidis,
Huppert, Kuh,
Wadsworth, & Croudace
(2006)
42
(7-item scale)
Sample size: 1179

Women aged 52
The sample comprised participants from the Medical Research Council's National Survey of Health and Development
(NSHD), the 1946 British birth cohort study
Postal questionnaire
Sample size: 1270

Swedish white-collar workers; women = 55%
Mean age = 45.3 years (SD = 7.2)
Mail survey

Lindfors, Berntsson &
Lundberg (2006)
18
(3-item scale)
Springer & Hauser (2006)
42
12
18
18
(7-, 2-, 3-item scale)
Wisconsin Longitudinal Study (WLS) – graduate Mail Survey. Sample size: 6282
Wisconsin Longitudinal Study (WLS) - graduate Telephone Interview. Sample size: 6038
National Survey of Midlife in the United States (MIDUS). Sample size: 2731 (25-74 years)
National Survey of Families and Households (NSFH II). Sample size: 9240
(continued on next page)

Page 15

Sirigatti et al. (Ryff) 15
Table 1 (continued)
Author N. of item Sample
Van Dierendonck, Diaz,
Rodriguez-Carvajal,
Blanco & Moreno-
Jimenez (2008)
39 (version by Van
Dierendonck, 2004):

6-item (SA, PR, EM,
PL); 7-item (PG);
8-item (AU)
Sample size: 919

592 individuals from Spain; 327 individuals from Columbia
Female = 417; Male = 525

Mean age = 30.0 years (SD = 14.0), range 16-74 years
Burns & Machin (2009)
84
(14-item scale)
Study 1
Sample size: 401

Life Events Study comprising undergraduate students from the Department of Psychology at the University of Southern
Queensland (USQ).
Female = 333; Male = 68
54
(9-item scale)
Study 2
Sample size: 679

Cross-national organizational climate study comprising three samples of schoolteachers:
- International teacher cohort (N = 176)
- Norwegian teacher cohort (N = 250)
- Australian teacher cohort (N = 253)

Female = 427; Male = 252
Most participants (46.2%) were aged between 30 to 55 years of age, though 63.2% of the Norwegian sample was aged
45 years and older