Towards a Four-Dimensional Model of Burnout: A Multigroup Factor-Analytic Study Including Depersonalization and Cynicism

Abstract

This article investigated whether cynicism and depersonalization are two different dimensions of burnout or whether they may be collapsed into one construct of mental distance. Using confirmatory factor analyses in two samples of teachers (n = 483) and blue-collar workers (n = 474), a superior fit was found for the four-factor model that contained cynicism, depersonalization, exhaustion, and professional efficacy as dimensions of burnout. In particular, cynicism and depersonalization emerged as unique burnout dimensions. Moreover, it appeared from multigroup analyses that this four-dimensional structure of burnout is partially invariant across both samples. Cynicism and depersonalization seemed to play a different role in both samples, particularly as far as their relationship with professional efficacy is concerned. It is recommended that future research on burnout should include the cynicism and depersonalization constructs.

Full text

TOWARD A FOUR-DIMENSIONAL MODEL OF BURNOUT
A MULTIGROUP FACTOR-ANALYTIC STUDY
INCLUDING DEPERSONALIZATION AND CYNICISM
MARISA SALANOVA, SUSANA LLORENS,
MÓNICA GARCÍA-RENEDO, RAÜL BURRIEL, AND EDGAR BRESé
Universitat Jaume 1, Castellón
WILMAR B. SCHAUFELI
Utrecht University
This article investigated whether cynicism and depersonalization are two different
dimensions ofbumout or whether they may be collapsed into one construct of mental dis-
tance. Using confinnatory factor analyses in two samples ofteachers (n = 483) and blue-
collarworkers (n =474), a superiorfit was foundforthe four-factormodel thatcontained
cynicism, depersonalization, exhaustion, and professional efficacy as dimensions of
burnout. In particular, cynicism and depersonalization emerged as unique bumout
dimensions. Moreover, it appeared from multigroup analyses that this four-dimensional
structure ofbumout is partially invariant across both samples. Cynicism and depersonal-
ization seemed to playa different Tole in both samples, particularly as far as their relation-
ship with professional efficacy is concerned. It is recornrnended that future research on
burnout should include the cynicism and depersonalization constructs.
cynicism; depersonalization; bumout; mental distanc
Kevwords;
Originally, bumout was defined as a syndrome of emotional exhaus-
tion, depersonalization, and reduced personal accomplishment that occurs
exclusively among professionals who deal directly with recipients such as
This research was supported by grants from fue Ministerio de Ciencia y Tecnología
(SEC2000-1031), Jaume I University (Castellón, Spain), and Bancaixa Foundation (11232.011
1). Correspondence concerning this article should be addressed to Marisa Salanova, Ph.D.,
Universitat Jaume 1, Department of Psychology, Campus de Riu Sec, s/n. 12071 Castel1ón,
Spain; e-mail: salanova@psi.uji.es.
Educational and Psychologicai Measurement, Vol. 65 No. 5, October 2005 901-913
DOr: 10.1177/0013164405275662
rg 2005 Sage Publications

EDUCAll0NAL ANO PSYCHOLOGICAL MEASUREMENT
902
students, pupils, clients, patients, or delinquents (Maslach, 1982, 1993).
Emotional exhaustion refers to fue depletion or draining of emotional re-
sources caused by interpersonal demands. Depersonalization refers to an
impersonal and dehumanized perception of recipients, characterized by a
callous, negative, and detached attitude. Finally, lack of personal accom-
plishment is fue tendency to evaluate one's work with recipients negatively.
These three components represent fue energetic (e.g., feeling used up), attitu-
dinal (e.g., being excessively detached), and self-evaluative (e.g., doubting
one's competence) nature of burnout, respectively (Maslach, Schaufeli, &
Leiter,2001).
Emotional exhaustion is regarded as fue basic individual stress compo-
nent of the syndrome that comes close to an orthodox job strain variable
(Maslach, 1993), whereas personal accomplishment is akin to fue concept of
efficacy beliefs (Bandura, 1999). This leaves depersonalization as fue most
innovative component of burnout. However, from fue outset, fue validity of
depersonalization has been questioned, for instance, by Garden (1987), who
argued that this dimension of burnout gauges several distinct attitudes,
including distancing, hostility, rejection, and unconcem. This might also be
fue reason why fue intemal consistency of fue depersonalization scale of fue
Maslach Burnout Inventory (MBI; Maslach, Jackson, & Leiter, 1996)-the
most popular instrument to assess burnout-is often found to be relatively
low when compared to both other scales that measure emotional exhaustion
and personal accomplishment (Lee & Ashforth, 1996).
Although initially bumout was restricted to fue helping professions, it was
later broadened and defined as a crisis in one' s relationship with work in gen-
eral and not necessarily as a crisis in one's relationship with people at work
(Maslach et al., 2001). The three original bumout dimensions were rede-
fined, and an altemative version of the MBI-the MBI-General Survey
(MBI-GS)-was developed that can also be used outside human services
occupations (Schaufeli, Leiter, Maslach, & Jackson, 1996). That is, exhaus-
tion as operationalized in fue MBI-GS refers to severe fatigue irrespective of
its cause, cynicism reflects an indifferent or distant attitude toward one's
work instead of other people, and lack of professional efficacy encompasses
both social and nonsocial aspects of occupational accomplishment.
However, in contrast to exhaustion and personal accomplishment, fue
broadening of fue bumout concept changed fue meaning of depersonaliza-
tion in a rather fundamental way. The reason is that, by definition, deperson-
alization involves other people so that its meaning cannot be broadened
beyond fue social relationships in which it occurs. This problem was solved
in fue case of fue MBI-GS by considering depersonalization as a special case
of "mental distancing." That is, where depersonalized human services pro-
fessionals exhibit a psychological distance toward their recipients, cynical
non-human services employees show a similar psychological distance
regarding their work in a more general sense. In other words, fue target of fue

SALANOVA ET AL. 903
mental distancing differs. In fue case of bullan services employees, fue tar-
gets are their recipients, whereas for employees who work with things or with
information, fue target is the job itself. This agrees with Dean, Brandes, and
Dharwadkar (1988), who argued that organizational cynicism has tour dif-
ferent targets: fue work organization at large, organizational change, fue
work environment, and fue persons at fue job (i.e., other employees and
recipients). These last two types of cynicism correspond with MBI deperson-
alization and MBI cynicism, respectively. Recently, in a study among cus-
taller services representatives, Abraham (2000) discriminated empirically
among fuese tour types of organizational cynicism and showed that each was
related in a slightly different way to several outcomes such as job satisfaction,
organizational commitment, and organizational citizenship behavior. In con-
clusion, it seems that depersonalization and cynicism, as measured with fue
MBI, are distinct constructs, yet they can be considered manifestations of
fue broader concept of organizational cynicism. To date, fueTe has been no
empirical test of fue distinctiveness of MBI depersonalization and MBI
cynicism in relation to both other burnout dimensions.
The purpose of this article, therefore, is to examine whether MBI deper-
sonalization and MBI cynicism can be considered two different components
of burnout or whether they merge together into one component of mental dis-
tancing. In other words, can two different targets of mental distancing be dis-
tinguished (i.e., people and work) within fue broader concept ofburnout? We
sought to answer this question by studying two different samples consisting
of bullan services professionals (secondary school teachers) and non-
bullan services employees (blue-collar workers). Similar results across both
samples would support fue robustness of OUT findings. We used the MBI-GS
to assess burnout (Schaufeli et al., 1996) as well as fue depersonalization
subscale of fue original MBI (Maslach et al., 1996). U sing confirmatory fac-
tor analysis, fue three-factor structure of fue original MBI (e.g., Byrne, 1993;
Gold, Bachelor, & Michael, 1989; Gorter, Albrecht, Hoogstraten, &
Eijkman, 1999; Lee & Ashforth, 1990; Schaufeli & Van Dierendonck, 2000)
as well as fue MBI-GS (e.g., Bakker, Demerouti, & Schaufeli, 2002; Leiter &
Schaufeli, 1996; Schutte, Toppinen, Kalimo, & Schaufeli, 2000; Taris,
Schreurs, & Schaufeli, 1999) has been convincingly demonstrated.
Method
Participants and Procedure
The total sample consisted of 957 Spanish employees (499 woman,
52.1 %, and 453 men, 47.3%). About half of fue sample (n = 483,49%) were
secondary school teachers (Sample 1), and the other half (n = 474, 51 %) were
hllll'--~n113T wnTkeT~ frnm the tile industrv (Samole 2). The mean ae:e of fue

904 EDUCATlONAL AND PSYCHOLOGICAL MEASUREMENT
.
l
total sample was 36 years, 8 months (SD = 9.0), and ages ranged from 18 tc
62 years. The teachers worked in 34 schools, most of which were public
(83%), whereas fue tile workers were employed in three private companies
In Sample 1 (43.6% men, 56.4% women), ages ranged from 23 to 60 years
(M = 40.2, SD = 8 years, 2 months). In Sample 2 (51.7% men, 48.3%
woman), ages ranged from 18 to 62 years (M = 33.2, SD = 8 years, 4 months).
Statistically, teachers were signiticantly older than blue-collar workers,
t(928) = -12.73, p < .001, with a large effect (d = .84; Cohen, 1988).
Participants were asked to till out fue MBI-GS as part of an occupational
health and safety audit. Human resources officers and school managers dis-
tributed fue questionnaires in fue tile companies and fue secondary schools,
respectively. A covering letter explained fue purpose ofthe study, explained
that fue participation was voluntary, and guaranteed contidentiality. Respon-
dents were asked to return the completed questionnaires in a sealed envelope,
either to fue person who had distributed them or directly to fue research team.
~
.
~
~
Instrumenti
Exhaustion (EX), cynicism (CY), and professional efficacy (PE) were
assessed with fue Spanish version (Salanova & Schaufeli, 2000) ofthe MBI-
GS (Schaufeli et al., 1996). Depersonalization (DP) was measured with fue
corresponding scale afilie original MBI-HSS (Maslach et al., 1996). In the
case of fue blue-colIar workers, recipients was replaced by coworkers in fue
items tapping depersonalization. Exhaustion was measured with 5 items
(e.g., "1 feel emotionalIy drained by my work"), cynicism was measured with
4 items (e.g., "1 have become more cynical about whether my work contrib-
utes anything"), depersonalization was measured by 5 items (e.g., "1 deal
with people with whom I work like objects"), and professional efficacy was
measured with 6 items (e.g., "In my opinion, I am good at my job"). High
scores on EX, DP, and CY and low scores on PE are indicative ofburnout. AlI
items were scored on a 7-point frequency scale, ranging from O (never) to 6
(everyday). As shownin Table 1, exceptforDPin theteacher's sample, inter-
nal consistencies (Cronbach's a) of scores on rol scales satisfied fue criterion
of .70 (NunnalIy & Bernstein, 1994), and in at least in the teachers sample,
most also satisfied fue more stringent criterion of .80 (Henson, 2001). As was
noted in fue introduction, for DP, slightly lower a values have been found
more often.
Data Analyses
Confirmatory factor analyses (CFA), as implemented by AMOS
(Arbuckle, 1997), was used to test fue fit of various models to fue data ofboth
samples. First, the fit of fue tour-factor burnout model (EX, CY, DP, FE) was

905
SALANOVA ET AL
Table 1
Means, Standard Deviations, Internal Consistencies (Cronbach's a), and Inten:orrelations
Blue-Collar Workers
(n = 474)
~
.lntercorrelation
EX CY DP PE
MSD IX
1.00
.54**
.37**
-.34**
1.08
1.20
.83
.80
.88
.84
.68
.81
.86
.83
.71
.78
EX
Cy
DP
PE
2.03
1.71
1.05
4.27
~
,
,
t
~
,
+
Note. EX = exhaustion; CY = cynicism; DP = depersonalization; PE = professional efficacy. Teachers are be
low !he diagonal, and blue-collar workers are above !he diagonal.
*p < .05. **p < .001.
.
t
compared in each sample with that of fue three-factor model (EX, CY + DP,
FE) and with fue one-factor model that assumes that all items weight on one
single underlying dimension (i.e., burnout). Next, using fue so-caIled multiple-
group method, fue factoriaI invariance of fue best -fitting model was examined
across both samples simultaneously (Byrne, 2001).
The goodness of fit of fue models was evaluated using absolute and rela-
tive indices. The four absolute goodness-of-fit indices calculated were (cf.
Joreskog & Sorbom, 1986) (a) fue X2 goodness-of-fit statistic, (b) fue good-
ness-of-fit index (GFI), (c) fue adjusted goodness-of-fit index (AGFI), and
(d) fue root mean square error of approximation (RMSEA). Following fue
recommendations ofMarsh, BaIla, and Hau (1996), we also computed three
relative indices: (a) Tucker-Lewis index (TLI), (b) comparative fit index
(CFI), and (c) fue incremental fit index (IFI). Because fue distributions of fue
GFI and fue AGFI are unknown, no statistical test or critical vaIue is available
(Joreskog & Sorbom, 1986). Values smaller than .06 for fue RMSEA are
indicative of an acceptable fit (Hu & Bentler, 1999), whereas a cutoff value
clase to .90 for fue IFI is suggested for a good fit (Hoy le, 1995). For fue
remaining fit indices (TLI, CFI), as a rule of thumb, values greater than .95
are considered as indicating an adequate model fit (Hu & Bentler, 1999).
i.
~
.
.
1Results
Descriptive Analyses
Table 1 shows fue means, standard deviations, internal consistencies,
and intercorrelations of fue tour burnout dimensions in both samples. As
expected, EX, DP, and CY are significantly and positively related, whereas
PE is negatively related with fue other burnout dimensions. However, the
.
.55**
1.00
.45**
-.45**
.33**
.48**
1.00
-.29**
-.10*
-.19**
-.07
1.00
2.03 1.19
1.24 1.22
.90 .96
4.41 .93

:DUCAnONAL AND PSYCHOLOGICAL MEASUREMENT
906
negative correlation between PE and DP was not statistically significant in
fue blue-collar sample.
Four analysis of variance tests were carried out to assess differences in
bumout levels among both samples (Huberty & Morris, 1989). Results
showed that compared to blue-collar workers, teachers felt more cynical,
F(I, 956) = 34.26,p < .001, and depersonalized, F(I, 956) = 7 .06,p < .01, and
they experienced less professional efficacy, F(I, 956) = 6.93, p < .01. How-
ever, a statistically nonsignificant difference was obtained for exhaustion
between fue samples, F( 1, 956) = .004, n.s. In terms of effect sizes, fue differ-
ences between both samples were "medium" for cynicism (d = .38) and
"small" for depersonalization and personal efficacy (both ds = .16; cf. Cohen,
1988).
Model Testing
Next, three altemative models were tested for each sample separately
using CFA: a Que-factor model (MI) that assumes one latent factor, a three-
factor model (M2) that assumes three latent and correlated factors (EX, CY +
DP, FE), and a four-factormodel (M3) that assumes four latent and correlated
factors (EX, CY, DP, FE).
MI and M2 fit very poorly to fue data of fue teachers (see Table 2), with
Done of fue fit indices meeting its criterion. The fit of M3 was somewhat
better and superior to that of M2 and M 1 as indicated by fue statisticalIy sig-
nificant values of LlX2 (see Table 2). The so-called modification indices indi-
cated that fue fit of M3 could be further improved by allowing three pairs of
errors (exl-ex2, cy3-cy4, and pe4-pe5) to correlate (see fue Discussion sec-
tion for a rationale). Indeed, a subsequent test of fue revised model (M4) that
included fuese three correlated error terms revealed a statistically significant
improvementoverM3, LlX2(3) = 206.13,p < .001, withmostfitindices meet-
ing or approaching their critical values.
As is shown in Table 3, a similar pattem of results was obtained in fue
blue-collar sample. Again, neither MI flor M2 fit well to fue data, and fue fit
ofM3 was superior to that ofMI and M2. Like fue teacher sample, fue fit of
M3 could be further improved (M4) by including fue same correlated errors
(exl-ex2, cy3-cy4, and pe4-pe5; see fue Discussion section for a rationale.
Next, fue best-fitting model (M4) was simultaneously fit to both samples
using multigroup analyses to test fue invariance of fue factor loadings, corre-
lated errors, and correlations between factors across both samples. As
expected, M4 provided reasonable fit to fue data across both samples, with
most fit indices meeting their corresponding critical values or at least
approaching them (see Table 4). However, the fit deteriorated somewhat
when all factor loadings and correlations (between factors and between
errors) were constrained to be equal in both samples (M4j. This means that

.
.
.
~
t
~
"'"
11
-=-
~
-8
'3
~
..
~
.,
..
.,
¿.
~
'<
..
So
'"
~
c-
So
N ~
Z'?
~d
11 11
~~
;;-"
" "
]]
":: s
1.: ~
~ '¡j
.= ..
..'"'
~ S
~oE
O 11
~~
!1: ..
u-
.. .g
.g S
c ..
.- O
~ ~
.~ ~
~ '"
j ,
~~
.1!; 11
FN
11 ~
~~
§ S
.= ~
.. ~
S .~
'¡< ...
O "
.. C
c.O
g. 11
'""-
O~
..
O .
t: -
" 8
" .
~ v
" *'"
'#:*
á*-
" ..
s ~
~ á
~5
11 c
00
-( .~
¡I¡ ~
"'.::-
~]
;; ,E
.g 5
c ~
~ ~
~ ~
o ~
~i
c s
8,6
00 c
'g ~
~ ~
" ~
';g'E
11 ~
!1: g
o ~
~~
~'6
-oM
c ~
.~ "
oE -5 .
~- .g-
0-
, -(
~ " s
" " ..
c -o o
~ c ~
o ":: ~
ool.:,+;,
11 3 ~
!1:coE
~e'g
~ ~ .~
c g ~
~ - ~
~
~
11
-=.
t:
~
~
~
8"
.::
~
...
~
~
.C;-
~
"(
...
~
~
~
~
'" ~
;z'?
~8
11 11
¡¡:...
-~
;<.;:
.g.g
o: O
.= e
!Q ..
" Si.
> "
.'" ..
..'"
~ ~
~.p,
o 11
~'"~
¡¡: ..
u"ü
.. "Ó
~ ~
o: ..
.- o
"' ~
.- ~
~ '"
j ,
~ ~
]-5
" 11
~'"
11 ~
~~
;; e
o ..
.'" o
.. ~
,5 ~
x '
~§
g. 11
o~
g..;
,,8
" .
~ V
Q *""
g'*
§*..
" ..
§ '§
§!:E
11 ~
.( .-
~ "'
~~
~~
;< .~
.g ~
o: "'
.- ¡¡
oE "
.,!. ~
o "'
cb"ü
¡:¡ "8
o: e
"8 "
OD,s
"Ó o:
~ 3
"' ~
Q ~
'g'B
11 ¡:¡
¡¡: ~
~ ~
..~
~'6
"ÓN
o: ~
.- "
oE,s .
.,!.-"ü
9<"8
~ " e
g .g ..
"Ó o: o
o .- ~
o,~ ~
OD- ,
11 ] ~
¡¡:o:.p,
~~-g
~ ~ .~
~.s ~
907

QOR
.,..\0\0""""
N - - 11 ] ~
~ o> 5
¡(!!,s
.g ] o~
oS c -
- o.. "
lC..'8
" '" e,. c
0= o -
~ u o
Q,:]]
e 11 ~
o "=
u-o
11 ~ ...
~ ~ 11
U '"ii 'O"~
¡(~~
.g e o,
c 0=]
o~ '" C
'" " O
~o~ ""
~-",
~ ~ §
,,~ u
'"
u~bQ
F ¡s oS
11 e-g
- - o
~o- - -
~ o
o, "" -
c ~ ~
og = ""
2,s
e ." o~
o~ .. ~
.."'-
8 o~ "
Q,~'8
g. 11 e
'"-'0"-
~ ~ s
o o u
t:-.g
"8 ~
!! o =
lO v..2 o
50 *"'" 11 ~
"'* ~."
2* '0""
" "'" ~ e
e - o, Q,
- 2] "
2 u c ,.
¡: lC o.. 0=
11 02 .. e
bQ '" "
< o~ C ~
~ '" o o~
~""u~ -~..,
~]se
o, '" u P
.. o:; .g .t=
.gSCbQ
C '" " C
o~ !! " 0=
~,,~'3
..!. ~ 2, '"
o",!!
, - "'-
'" ~ C "
~ ¡s og '8
~e.!Se
¡s " !! "
bQ,s~u
~ C 8 2
~ H ,s O¡
= ~ o~ ,.
'6' '¡¡ ! oS
.. .c " -
11 ~ '8 s
~ge~
9 ~ ~ ~
""",,-=
o U o
.. o~ .. ...
.g;;"Z"
C ~ = '"
o~ " o 'O"
~,s~~
9~ 8-g
'" o-.t C
'" .. ~ o~
1" .g .;: ~
oS .g ~
- o o
bQlC e u
11 3 - ~
~ C S o
" " ~ o~
o e",,-
~ !! ~ ~
~ oS ..§ 8

909
SALANOVA ET AL.
although fue underlying factor structure is apparently similar in both sam-
pIes, fue size of fue factor coefficients and interfactor correlations mar differ
somewhat.
To assess invariance in greater detail, three additional models were tested
to fue data: (a) a model that as sumes only the correlations between factors to
be invariant (M4co)' (b) a model that assumes only fue factor loadings to be
invariant (M4ca)' and (c) a model that assumes only the correlations between
fue error terms to be invariant (M4er). As can be seen from Table 4, fue fit of
all three models was slightly inferior compared to that of M4. This suggests
that neither fue correlations between factors, nor fue factor coefficients, nor
fue correlations between error terms are completely invariant across both
samples.
In the final step, an iterative process was used as recommended by Byme
(2001) to assess fue invariance of each estimate separately. That is, fue
invariance of each factor loading, correlated error term, and correlation
between factors was assessed individually by comparing fue fit of fue model
in which a particular estimate was constrained to be equal across both sam-
pIes with that of fue previous model in which this was not the case. When the
fit did not deteriorate, this constrained element was included in fue next
model in which another constrained estimate was added, and so oo. The final
model (M4fi) showed that fue correlation between EX and DP, and between
CY and DP, as well as fue correlations between the two errors (ex1-ex2, cy3-
cy4) are invariant across both samples. In addition, fue factor coefficients, of
allEX items, 3 of 4 CY items (cy1, cy2, cy3), as well as 2 of5 DP items (dep1,
dep4) and 3 of 6 PE items (pe1, pe2, pe6), tumed out to be invariant across
both samples as well. Thus, it appeared that EX has fue highest proportion of
invariant items (100%), followed by CY (75%), PE (50%), and DP (40%),
respectively.
Discussion
The aim of ibis artic1e was to investigate whether cynicism and deperson-
alization mar be considered two different dimensions of burnout or whether
they can be col1apsed into one construct of mental distance. We studied two
samp1es: teachers who work with other peop1e (i.e., students) and b1ue-collar
workers from a ti1e factory who work with objects and who process data
using advanced computerized machinery. According to fue traditional view,
bumout occurs on1y in human services professionals such as teachers, but
recent1y, it has been acknow1edged that burnout also might occur in other
occupational groups such as b1ue-collar workers (Mas1ach et al., 2001). In
fue former case, it is assumed that depersonalization is an essential dimen-
sion of burnout, whereas in fue latter case, depersonalization is substituted by
cynicism. Conceptually speaking, both cynicism and depersonalization are
manifestations of mental distancing. Por depersonalization, this distancing is

910 EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT
directed toward the people with whom one is working (i.e., students in teach-
ing and colleagues in blue-collar work), whereas with cynicism, fue dis-
tancing is directed toward fue broader context of fue job itself.
Our results show that instead of one mental distance construct, burnout
can be characterized in both samples by separate depersonalization and cyni-
cism dimensions, which along with exhaustion and reduced professional
efficacy, constitute the burnout syndrome. That is, in both samples, fue four-
factor model with separate depersonalization and cynicism dimensions fit
better to the data than fue three-factor model with depersonalization and cyn-
icism collapsed into one factor. However, to increase the fit ofthe four-factor
model, we had to allow three pairs of error terms to correlate. Although this
might increase the risk of chance capitalization (Curdeck & Brown, 1993),
this procedure is thought to be justified because similar correlated error terms
were observed previously in other samples: exl-ex2 among South Afrlcan
police officers (Storm & Rothmann, 2003); cy3-cy4 among students from
Portugal, Spain, and the Netherlands (Schaufeli, Martínez, Marques-Pinto,
Salanova, & Bakker, 2002) and blue- and white-collar workers from Sweden,
Finland, and fue Netherlands (Schutte et al., 2000); and pe4-pe5 among
Spanish information and communication technology users (Salanova,
Schaufeli, Llorens, Peiró, & Grau, 2000) and white-collar workers (Schutte
et al., 2000). Hence, it seems that instead of being sample specific, fue corre-
lated errors refIect cornmon variance between items that is independent from
country and occupation. For instance, the items "1 feel emotionally drained
from my work" (exl) and "1 feel used up at fue end of fue workday" (ex2)
share some unique variance, probably because both refer to extreme tired-
ness after work, whereas fue other EX items refer to tiredness in fue moming
or less often to intensive fatigue.
Although fue four-factor model (including fuese three correlated errors)
fit well to fue data in both samples, it was not entirely invariant across both
samples. However, an interesting pattem emerged from an iterative proce-
dure that was followed to assess invariance in greater detail. It appeared that
all EX items, three of four CY items, and fue correlations between fuese two
scales-and fue included error terms-were invariant across both samples.
This means that the core of bumout-namely, EX and CY (Maslach et al.,
2001)-was invariant across both samples. Different factor coefficients and
correlations were obtained for DP and FE, though. This can be explained by
the fact that depersonalization has a quite different meaning in both samples.
For teachers, relationships with students are critical for their job performance
and hence for their feelings of professional efficacy, whereas for blue-collar
workers, colleagues do not play such an essential role in this respecto This is
exemplified by fue negative statistically significant correlation between DP
and PE in fue teacher sample as compared to fue statistically nonsignificant

SALANOVA ET AL. 911
corresponding correlation in fue blue-collar sample(see Table 1). In other
words, teachers who depersonalize their students feel inefficacious because
students are, after all, fue "essence" of their work (i.e., teaching), whereas
blue-collar workers who depersonalize their coworkers do not feel ineffica-
cious necessarily. For teachers, good relationships with students are essential
for being successful, but blue-collar workers can do their work well even
when relationships with colleagues are poor.
In conclusion, our study suggests that cynicism and depersonalization
each contribute in a distinct way to fue bumout syndrome. However, we
focused only on fue relationships among bumout dimensions. Future
research should establish whether cynicism and depersonalization are differ-
ently related to particular job characteristics (i.e., job demands and
resources) and outcome variables (i.e., workers performance, absenteeism).
Hence, we recornmend including cynicism in addition to fue three traditional
MBI bumout dimensions when studying human services and to include
depersonalization in addition to fue three MBI-GS dimensions when study-
ing non-human services occupations.
References
Abraham, R. (2000). Organizational cynicism: Bases and consequences. Genetic, Social and
General Psychology Monographs, 126, 269-292.
Arbuckle, J. L. (1997). Amos users' guide version 4.0. Chicago: Smallwaters Corporation.
Bakker, A. B., Demerouti, E., & Schaufeli, W. B. (2002). Validation of the Maslach Burnout
Inventory-General Survey: An Internet study across occupations. Anxiely, Stress and Cop-
ing, 15, 245-260.
Bandura, A. (1999). Social cognitive theoryofpersonality. InL. Perwin & O. John (Eds.),Hand-
book ofpersonalily (2nd ed., pp. 154-196). New York: Guildford.
Byrne, B. M. (1993). The Maslach Inventory: Testing for factorial validity and invariance across
elementary, intennediate, and seconctary teachers. Journal of Occupational and Organiza-
tional Psychology, 66, 197-212.
Byrne, B. M. (2001). Structural equation modeling with AMOS: Basic concepts, applications
and programming. Mahwah, NJ: Lawrence Erlbaum.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Lawrence Erlbaum.
Cudeck, R., &Browne,M. W. (1993). Alternativewaysofassessingmodelfit. InK. A. Bollen&
J. Scott Long (Eds.), Testing structural equation models (pp. 1-9). Newbury Park, CA: Sage.
Dean, J. W., Brandes, P., & Dharwadkar, R. (1988). Organizational cynicism. Academy of Man-
agement Review, 23, 314-352.
Garden, A. M. (1987). Depersonalization: A valid dimension ofburnout? Human Relations, 140,
545-560.
Gold, Y., Bachelor, P., & Michael, W. B. (1989). The dimensionally of a modified fonn of the
Maslach Burnout Inventory for university students in a teacher-training programo Educa-
tional and Psychological Measurement, 49, 549-561.
Gorter, R., Albrecht, W., Hoogstraten, J., & Eijkman, M. (1999). Pactorial validity of the
Maslach burnout inventory-Dutch version (MBI-NL) among dentists. Joumal ofOrganiza-
tional Behaviol; 20, 209-217.

912 EDUCAnONAL AND PSYCHOLOGICAL MEASUREMENT
Henson, R. K. (2001). Understanding internal consistency re1iabi1ity estirnates: A conceptual
primer on coefficient alpha. Measurement and Evaluation in Counseling and Development,
34,177-189.
Hoy1e, R. H. (1995). The structura1 equation mode1ing approach: Basic concepts and fundamen-
tal issues. In R. H. Hoy1e (Ed.), Structural equation modeling: Concepts, issues and applica-
tions (pp. 1-15). Thousand Oaks, CA: Sage.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for lit indexes in covariance structure analysis:
Conventiona1 criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
Huberty, C. J., & Morris, J. D. (1989). Mu1tivariare analysis versus mu1tip1e univariate analyses.
Psychological Bulletin, 105, 302-308.
Joreskog, K. G., & Sorbom, D. (1986). USRELuser guide version VI (4th ed.). Mooresvi11e, IL:
Scientific Software International.
Lee, R. T., & Ashforth, B. E. (1990). On fue meaning ofMas1ach's three dimensions ofburnout.
Joumal of Applied Psychology, 75, 743-747.
Lee, R. T., & Ashforth, B. E. (1996). A meta-ana1ytic examinationofthe corre1ates ofthe three
dimensions ofjob burnout. Joumal of Applied Psychology, 81,123-133.
Leiter, M. P., & Schaufe1i, W. B. (1996). Consistency of fue burnout construct across occupa-
tions. Anxiety, Stress and Coping, 9, 229-243.
Marsh, H. W., Bal1a,J. R., & Hau, K. T. (1996). Anevaluationofincrementalfitindices: Ac1ari-
fication ofmathernatical and empirical properties. In G. A. Marcou1ides & R. E. Schumacker
(Eds.), Advanced structural equation modeling, issues and techniques (pp. 315-353). Mah-
wah, NJ: Lawrence Er1baum.
Mas1ach, C. (1982). Understanding burnout: Definitional issues in analyzing a comp1ex phe-
nomenon. In W. S. Paine (Ed.),Job stress and bumout(pp. 29-40). Bever1y Hil1s, CA: Sage.
Maslach, C. (1993). Bumout: A mu1tidimensional perspective. In W. B. Schaufe1i, C. Maslach,
& T. Marek (Eds.), Professional bumout: Recent developments in theory and research
(pp. 19-32). Washington, DC: Tay1or & Francis.
Mas1ach, C., Jackson, S. E., & Leiter, M. P. (1996). Maslach Bumout 1nventory: Manual (3rd
ed.). Palo Alto, CA: Consulting Psychologists Press.
Maslach, C., Schaufe1i, W. B., & Leiter, M. P. (2001).Job burnout.AnnuaIReviewofPsychology,
52,397-422.
Nunnally, J. C., & Bernstein, l. H. (1994). Psychometric theory (3rd ed). New York: McGraw-
Hil1.
Salanova, M., & Schaufe1i, W. B. (2000). Exposure to technology and its re1ationship to bumout.
Behovior and 1nfonnation Technology, 19, 385-392.
Salanova, M., Schaufe1i, W. B., L1orens, S., Peiró, J. M., & Grau, R. (2000). Desde el "burnout"
al "engagement": una nueva perspectiva? [From burnout to engagement: A new perspec-
tive?] Revista de Psicología del Trabajo y las Organizaciones, 16, 117-134.
Schaufe1i, W. B" Leiter, M. P" Maslach, C" & Jackson, S. E. (1996). Mas1ach Burnout Inven-
tory-General Survey. In C. Mas1ach, S. E. Jackson, & M. P. Leiter (Eds.), The MaslachBum-
out 1nventory: Test manual (3rd ed.). Palo Alto, CA: Consulting Psycho1ogists Press.
Schaufe1i, W. B., Martínez, l. M., Marques-Pinto, A., Salanova, M., & Bakker, A. (2002). Burn-
out and engagement in university students: Across-national study. Joumal ofCross-Cultural
Psychology, 33, 464-481.
Schaufeli, W. B., & Van Dierendonck, D. (2000). UBOS: Utrechtse Bumout Schaal-Test
Handleiding [UBOS-TestManua1: Dutch version ofthe Maslach Bumout Inventory]. Lisse,
fue Nether1ands: Swets and Zeitlinger.
Schutte, N" Toppinen, S" Kalimo, R., & Schaufeli, W. B. (2000). The factorial validity ofthe
Mas1ach Bumout Inventory-General Survey across occupational groups and nations. Jour-
nal ofOccupational and OIRanizational Psychology, 73,53-66.

913
SALANOVA ET Al
Storm, K., & Rothmann, l. (2003). A psychometric analysis of fue Utrecht Work Engagement
Scale in fue South African police service. South African Journal of Industrial Psychology,
29,62-70.
Taris, T. W., Schreurs, P. J. G., & Schaufeli, W. B. (1999). Construct validity of fue Maslach
Bumout Inventory-General Survey: A two-sample examination of its factor structure and
correlates. Work and Stress, I3, 223-237.