Blackwell Science, LtdOxford, UKADDAddiction0965-2140© 2004 Society for the Study of Addiction

99••757769
Original Article
Gambling Related Cognitions Scale (GRCS)
Namrata Raylu & Tian P. S. Oei

RESEARCH REPORT

The Gambling Related Cognitions Scale (GRCS):

development, confirmatory factor validation and
psychometric properties

Namrata Raylu & Tian P. S. Oei*

School of Psychology, University of Queensland, Brisbane, Australia

Correspondence to: ABSTRACT

Professor T. Oei
School of Psychology,
University of Queensland Aims The aims of this study are to develop and validate a measure to screen
Brisbane for a range of gambling-related cognitions (GRC) in gamblers.
Queensland, 4072 Design and participants A total of 968 volunteers were recruited from a
Australia
community-based population. They were divided randomly into two groups.
E-mail: oei@psy.uq.edu.au
Principal axis factoring with varimax rotation was performed on group one and
Submitted 17 October 2003; confirmatory factor analysis (CFA) was used on group two to confirm the best-
initial review completed 2 December 2003; fitted solution.
final version accepted 16 February 2004
Measurements The Gambling Related Cognition Scale (GRCS) was developed
for this study and the South Oaks Gambling Screen (SOGS), the Motivation
Towards Gambling Scale (MTGS) and the Depression Anxiety Stress Scale
(DASS-21) were used for validation.
Findings Exploratory factor analysis performed using half the sample indi-
cated five factors, which included interpretative control/bias (GRCS-IB), illusion
of control (GRCS-IC), predictive control (GRCS-PC), gambling-related expectan-
cies (GRCS-GE) and a perceived inability to stop gambling (GRCS-IS). These
accounted for 70% of the total variance. Using the other half of the sample, CFA
confirmed that the five-factor solution fitted the data most effectively. Cron-
bach’s alpha coefficients for the factors ranged from 0.77 to 0.91, and 0.93 for
the overall scale.
Conclusions This paper demonstrated that the 23-item GRCS has good psy-
chometric properties and thus is a useful instrument for identifying GRC among
non-clinical gamblers. It provides the first step towards devising/adapting sim-
ilar tools for problem gamblers as well as developing more specialized instru-
ments to assess particular domains of GRC.

KEYWORDS Cognition, community, confirmatory factor analyses,

gambling, problem gamblers.

INTRODUCTION commentary on everything that is going on in their

With the advancement of cognitive science and cognitive
behaviour therapy, the role of cognitions in the aetiology
and treatment of problem gambling (PG) has received
much attention in research. The ‘thinking aloud method’
is a technique where gamblers are asked to provide
minds including intentions, urges, ideas and images on
their play as the game is proceeding. Using this method
researchers have identified the existence of gambling-
related cognitions (GRC), such as the ability to predict or
control gambling outcomes among frequent/problem
gamblers for a variety of games (Gilovich & Douglas

*Professor Oei is also the Director, CBT Unit, Toowong Private Hospital.
RESEARCH REPORT
© 2004 Society for the Study of Addiction doi:10.1111/j.1360-0443.2004.00753.x Addiction, 99, 757–769
758 Namrata Raylu & Tian P. S. Oei
1986; Gaboury & Ladouceur 1989; Coulombe et al.
1992; Griffiths 1994, 1996). Many of these studies have
also reported that regular/problem gamblers (especially
those perceived to require skill, e.g. machine gambling)
generally tend to have a greater number of these cogni-
tions than non-regular/non-problem players and non-
gamblers (Walker 1992; Toneatto et al. 1997; Joukhador,
Maccallum & Blaszczynski 2003). The correction of such
cognitions using cognitive therapy have also been found
to reduce monetary risk, frequency of gambling and
urges to gamble (Gaboury & Ladouceur 1989; Sylvain &
Ladouceur 1992; Sylvain, Ladouceur & Boisvert 1997),
adding support to the idea that GRC play a role in the ini-
tiation and maintenance of gambling.
Several authors have tried to identify and categorize
GRC (Wagenaar 1988; Griffiths 1993, 1994; Toneatto
et al. 1997; Toneatto 1999). Toneatto et al. (1997) and
Toneatto (1999) identified three general categories of
GRC. The first category reflected a belief that one could
control gambling outcomes via personal skill, ability or
knowledge. This category included active illusionary con-
trol where individuals had a belief that relying on super-
stitious behaviours, on systems or lucky numbers could
influence gambling outcomes (e.g. possession of certain
objects such as a rabbit’s foot, performing specific rituals
such as a prayer or having certain mental states such as
a positive attitude). This category also included passive
illusionary control, which involved mainly an illusion of
control over luck (e.g. interpreting good luck or success in
other areas of life as signs that they will also succeed at
gambling). Furthermore, this category contained cogni-
tions reflecting magnification of one’s own gambling
skills/ability to win and minimization of other gamblers’
skills/ability to win (Toneatto 1999). The second category
reflected means by which individuals could predict gam-
bling outcomes. This included belief that one had the skill
to make accurate predictions regarding gambling based
on salient cues (e.g. the weather or hunches) or based on
past wins/losses (e.g. expecting that a series of losses will
be corrected by series of wins). It also included probability
errors, such as misunderstanding the nature of probabil-
ity, chasing (believing that continued gambling might
recoup lost money) or obtaining false contingencies
(assigning causal influences to factors that are associated
with winning). The third category reflected interpretative
bias, which consisted of reframing gambling outcomes
that would encourage continued gambling despite losses.
This included attributing successes to one’s own skill and
failures to others’ influences or luck or recalling wins
more easily than losses, and thus expecting to win at
games which they have lost previously.
Other cognitions available in the substance abuse
literature are also relevant to problem gamblers, but have
not yet been discussed thoroughly in the gambling
© 2004 Society for the Study of Addiction
literature. Numerous studies have shown that these cog-
nitions are important in the development and mainte-
nance of substance related problems (Baldwin, Oei &
Young 1993; Beck et al. 1993; Lee & Oei 1993; Lee, Oei &
Greeley 1999). The first of these cognitions relates to indi-
viduals’ perceived expectations about the effects of the
object(s) of interest. Expectancies have found to play a sig-
nificant role in substance abuse (Oei & Baldwin 1994;
Oei, Furgerson & Lee 1998). Very few studies have dis-
cussed gambling expectancies (e.g. Walters & Contri
1998; Oei & Raylu 2004). However, many studies have
researched what motivates one to gamble and continue
gambling, despite persistent losses. These include gam-
bling to: demonstrate one’s worth; receive approval and
social acceptance from others; rebel; relieve negative and
painful emotions; participate due to social reasons; try to
beat the odds; and participate to experience the excite-
ment, reduce boredom or have fun (Murray 1993; Zuck-
erman 1999; Raylu & Oei 2002). This demonstrates that
gamblers may have particular thoughts about how they
expect gambling to affect them. These gambling-related
expectancies (e.g. ‘gambling makes me more relaxed’)
can be developed through exposure to gambling models,
the media and cultural rituals and through one’s experi-
ences. Expecting gambling to be the only way to achieve
these can help to maintain gambling despite continuous
losses. Such expectancies may also be related to rational-
izing why they cannot stop gambling (e.g. ‘without gam-
bling I will be very bored/lonely’). Termination of
gambling is seen as a deprivation of satisfaction or a
threat to wellbeing and functioning. These functions of
gambling expectancies are similar to those put forward
for substance-related expectancies (Beck et al. 1993).
The second cognition relates to the perceived inability
to stop particular addictive behaviours (Beck et al. 1993;
Oei & Burrow 2000). This is similar to drinking refusal
self-efficacy (one’s perceived ability to resist drinking in
high-risk situations) found for alcohol problems (Lee &
Oei 1993; Oei & Baldwin 1994; Lee, Oei & Greeley 1999;
Oei & Burrow 2000). Although little is written specifi-
cally about such perceived inability among gamblers,
many researchers have stated that problem gamblers
often try to stop gambling, especially when they become
aware there might be a problem (Black & Moyer 1998;
APA 1994). These are manifested usually in the thoughts
related to their perceived inability to stop gambling or the
activities they need to complete (e.g. borrowing and theft)
in order to continue gambling (Blaszczynski 1994; Meyer
& Fabian 1996; Sharpe 2002). They may perceive their
feelings of disappointment and distress as intolerable and
feel a sense of helplessness to control their behaviours.
Such thoughts (e.g. ‘my gambling is overpowering’ or ‘I
am not strong enough to stop gambling’) may upset
them further and these beliefs may become self-fulfilling
Addiction, 99, 757–769

prophecies similar to what has been suggested for other
types of addictive problems (Beck et al. 1993). If individ-
uals believe that they are incapable of controlling their
urges, they are less likely to try to control them. Thus, this
will confirm their beliefs about being helpless in overcom-
ing their addictive problem. Such cognitions about their
inability to stop gambling can contribute to relapses and
be responsible for depression, which is common among
problem gamblers (Raylu & Oei 2002).
Despite the significance of GRC among frequent and
problem gamblers, research in this area has been hin-
dered by the lack of adequate measures of GRC. Currently,
two measures that attempt to assess GRC have been intro-
duced in the literature. Both these scales reported good
consistency and validity. However, both of them have sev-
eral limitations. First, although the Video Gaming Device
Inventory (VGDI) developed by Pike (2002) includes
some items to assess GRC, it also includes items that
assess emotions/feelings and behaviours related to PG.
Furthermore, this scale does not have separate subscales
to measure the three domains (i.e. GRC, PG behaviours
and emotions/feelings related to PG). The VGDI is also
restricted in assessing PG related to video gambling. Sec-
ondly, Steenbergh et al. 2002) reported on the develop-
ment and validation of the 21-item Gamblers’ Beliefs
Questionnaire designed to assess GRC. This instrument
consisted of only two closely related subscales (luck and
illusion of control). Reviews (e.g. Toneatto 1999; Raylu &
Oei 2002) suggest that there are more types of GRC. Fur-
thermore, neither of these scales was examined using
confirmatory factor analyses (CFA).
In view of the detrimental effects of GRC on gamblers
and the lack of instruments to assess them, it would be
desirable to have a self-report measure to help screen for
GRC. Thus, the present study focuses on developing a scale
that assesses a wide range of GRC (erroneous cognitions
about success at gambling as well as items related to beliefs
about self in relation to gambling) in the general popula-
tion. Given the link between GRC and PG, such a tool can
help to screen for those individuals in the community that
may be at risk of developing gambling problems and pro-
vide the first step towards developing/adapting such an
instrument for a clinical sample of problem gamblers and
developing specialized instruments to assess particular
domains of GRC. Thus, the purpose of this paper is to
report on the development and psychometric properties of
the Gambling Related Cognitions Scale (GRCS).
METHOD
Participants
Participants were volunteers (n = 968) drawn from a
community-based population. The aim was to ensure that
© 2004 Society for the Study of Addiction
Gambling Related Cognitions Scale (GRCS) 759
the sample contained a range of individuals covering a
wide range of cultural groups, educational backgrounds
and socio-economic status. Thus, the study was advertised
in various cultural (Chinese, Greek, Indian, etc.) clubs and
society magazines, newsletters and notice-boards. Fur-
thermore, the opportunity to participate for course credit
was also advertised on the first-year psychology student
website. Advertisement of the study and request for assis-
tance with the study were also sent to several chosen
employment organizations from the Brisbane business
directory. All participants were between the ages of 16 and
73 years; 50.2% of participants were single, 42.8%
reported living together/married and others (7.0%).
The sample was divided randomly into two groups
using the odds and evens split method. Thus, each group
consisted of 484 participants. Group B (35.7% male;
mean age of 31.73 years, SD = 15.14 years) was used for
the CFA and group A (37.3% male; mean age
31.78 years, SD = 15.19 years) was used for the explor-
atory factor analyses (EFA). Group A consisted of 75.4%
Caucasians, 12.6% Chinese, 4.5% Indians, 3.6% other
Asians (including Vietnamese, Korean, Japanese) and
3.9% other ethnic groups. In group B, 81.8% identified
themselves as Caucasians, 8.1% as Chinese, 4.1% as Indi-
ans, 3% as other Asians and 3% as other ethnic groups.
The South Oaks Gambling Screen (SOGS; Lesieur &
Blume 1987), which assesses extent of PG, identified
4.5% of participants as probable pathological gamblers
(SOGS 5 or greater); 33.1% had SOGS scores between 1
and 4 and 62.4% had SOGS score of 0 (see full description
of SOGS in the next section). The distribution for group B
was 5.4%, 29.3% and 65.3%, respectively. These PG fig-
ures are consistent with reviews published in the gam-
bling literature (Petry & Armentano 1999; Raylu & Oei
2002) that have stated that life-time prevalence rates
ranged from 0.1 to 5.1% (our figure of 4.5% is within the
literature range). Both groups had a similar age range
(16–73 years), and there were no significant differences
in relation to SOGS score, amount gambled, education
levels, marital status, ethnic group, employment status
and income levels.
Measures
South Oaks Gambling Screen (SOGS)
The SOGS is a 20-item questionnaire based on Diagnostic
and Statistical Manual (DSM)-III criteria to screen for life-
time PG. It has been used with patients in a therapeutic
community (Lesieur & Heineman 1988), psychiatric
admissions (Lesieur & Blume 1990) and numerous treat-
ment settings as an aid in diagnostic and forensic screen-
ing (Rosenthal 1989). It has shown to have high validity
(by cross-tabulating patients scores with counsellors’
Addiction, 99, 757–769
760 Namrata Raylu & Tian P. S. Oei
independent assessment scoring Æ r = 0.86, P < 0.001)
as well as high internal consistency reliability, Cron-
bach’s alpha = 0.97, P < 0.001 (Lesieur & Blume 1987).
Motivation Towards Gambling Scale (MTGS; Chantal,
Vallerand & Vallieres, 1994)
This scale contains 28 items evaluating seven types of
motivation. This includes intrinsic motivation toward
knowledge, accomplishment and stimulation from the
gambling activities, extrinsic motivation assessing exter-
nal regulation (external reward), introjected external
regulation (when gambling becomes regulated by self-
imposed pressures), regulation through identification
(where the individual participates in gambling to attain
another important goal such as socializing) and motiva-
tion when there are no perceived relations between
actions and outcomes. Individuals are asked to respond to
each item using a seven-point scale, indicating the degree
to which each statement corresponds to the reasons why
they play their favourite game, ranging from 1 (does not
correspond at all) to 7 (corresponds exactly). Chantal et al.
(1994) reported good reliability and validity for this scale.
Depression Anxiety Stress Scale (DASS-21; Lovibond &
Lovibond, 1995)
This scale assesses levels of depression, anxiety and stress
using 21 items. Individuals are asked to respond to each
item using a four-point scale to indicate the extent to
which each items applied to them in the last 2 weeks,
ranging from 0 (did not apply to me at all) to 4 (applied to
me very much, or most of the time). It has been normed
for Australian populations and been validated against
other Beck Depression and Anxiety Inventories (Lovi-
bond & Lovibond 1995). This scale has been shown to
have high concurrent validity (r = 0.84) and reliability
(i.e. Cronbach’s alpha of 0.94, 0.87 and 0.91 for sub-
scales depression, anxiety and scale, respectively;
Anthony et al. 1998).
A short questionnaire pertaining to demographic
information (e.g. gender, age, employment status, educa-
tion level and ethnicity) was also included. The above
measures used for this study are parts of a larger project
on problem gamblers funded externally by a government
agency.
Procedure
Generation of the original Gambling Related Cognitions Scale
(GRCS) questionnaire
The aim of the study was to develop a brief scale to screen
for gambling cognitions among a community sample.
Items for the questionnaire were generated to cover the
© 2004 Society for the Study of Addiction
wide range of GRC that have been reported in the gam-
bling literature. Consequently, an original 59-item ques-
tionnaire was developed to reflect a range of GRC. Items
were constructed by the authors using the examples and/
or description of the various categories of GRC provided
by previous studies in this area (e.g. Gaboury & Ladou-
ceur 1989; Griffiths 1994; Toneatto et al. 1997; Toneatto
1999). Other items (e.g. those reflecting gambling-
related expectancies and perceived inability to stop gam-
bling) were constructed using examples of these specific
types of cognitions provided by general addictive studies
(e.g. Beck et al. 1993; Lee & Oei 1993; Oei & Baldwin
1994; Lee, Oei & Greeley 1999; Oei & Burrow 2000)
because they have mentioned only recently being in the
gambling literature (e.g. Walters & Contri 1998; Raylu &
Oei 2004). Generally, a minimum of four items reflecting
each category was constructed. Before the questionnaire
was distributed to the participants, the 59 items were
assessed for their clarity and relevance by the two
authors and two other independent evaluators. Minor
revisions (either deletion or rephrasing of items) were
made based upon this process. Consequently, six items
were deleted and this resulted in a 53-item Gambling
Related Cognitions Scale (GRCS).
The participants were asked to use a seven-point Lik-
ert scale (1 = strongly disagree, 2 = moderately disagree,
3 = mildly disagree, 4 = neither agree nor disagree,
5 = mildly agree, 6 = moderately agree, 7 = strongly
agree) to indicate the extent to which they agreed with
the value expressed in each statement. Total score con-
sisted of adding the values gained from each of the 23
items. The higher the total score the higher the number of
GRC displayed.
Distribution of the questionnaires
Participants who agreed to participate in the study were
provided with a set of questionnaires and asked to
return them to the researchers in stamped self-
addressed envelopes. The overall response rate (% of
questionnaires returned) was 90%, while the overall
completion rate (% of returned questionnaires that were
answered completely) was 95%. Responses were confi-
dential to the researchers, and identification codes were
used rather than names. Minor missing data (e.g. 1–2
unanswered item/s per questionnaire) were found for
approximately 2% of individuals and these were
replaced with means.
Generation of the final GRCS questionnaire
Principal axis factoring with varimax rotation using the
Statistical Package for the Social Sciences (SPSS) was
used to develop the final GRCS questionnaire that met the
Addiction, 99, 757–769

following criteria: (a) minimum factor eigenvalues of 1;
(2) minimum factor membership of four items; (c) exclu-
sion of items with factor loadings less than 0.4; and (d)
conceptual coherence of individual factors. All data
(groups A and B combined) were used for this process.
The following process was used to reduce the original
53-item questionnaire to the final 23-item version. First,
all loadings less than 0.4 were suppressed based on
Stevens’ (1992) suggestion that this cut-off point was
appropriate for interpretative purposes. Then, the inter-
item correlations and item-total correlations for all 53
items were calculated. The aim of this was to examine,
through item analysis, which of the items were very sim-
ilar to another. First, any items with double factor loading
were deleted. This included deletion of 10 items. For each
item that correlated significantly highly with another
(greater than correlation of 0.75), the item with the
lower item total correlation was deleted (DeVellis 1991).
This process deleted five items from the 53 items. A fur-
ther eight were deleted based on the fact that they were
items that appeared to measure the same thing. This was
also done to reduce multi-collinearity in the data. Com-
munalities (how much each item correlated with total)
were calculated for the remaining. As suggested by DeV-
ellis 1991), in order to produce a scale measuring rela-
tively specific construct, all items with communalities of
less than 0.3 were deleted. This process was repeated
until deleting more items could not increase further the
internal consistency of the scale (Rapee et al. 1996). This
resulted in a 23-item questionnaire representing five fac-
tors. Three of the factors are consistent with the catego-
ries suggested by previous studies (Toneatto et al. 1997;
Toneatto 1999), including illusion of control (e.g. ‘I have
specific rituals and behaviours that increase my chances
of winning’), predictive control (e.g. ‘Losses when gam-
bling, are bound to be followed by a series of wins’) and
interpretative bias (e.g. ‘Relating my winnings to my skill
and ability makes me continue gambling’). The other two
categories were consistent with gambling-related expect-
ancies (e.g. ‘Having a gamble helps reduce tension and
stress’) and perceived inability to stop gambling (e.g. ‘My
desire to gamble is so overpowering’) discussed earlier in
the paper.
Analyses of the 23-item GRCS
Exploratory factor analyses and CFAs were used to con-
firm the five scales and explore the characteristics of the
23-item GRCS. Cronbach’s alpha analyses were used to
assess the reliability of the GRCS and its five subscales.
Statistical tests such as bivariate correlations, analyses of
variance analyses (ANOVA), discriminant analyses and
multiple regression analyses were used to explore the cri-
terion-related validity, predictive validity and concurrent
© 2004 Society for the Study of Addiction
Gambling Related Cognitions Scale (GRCS) 761
validity of the 23-item GRCS. Finally, ANOVAs were used
to explore gender differences in the GRCS scores as well in
the five subscale scores.
RESULTS
Exploratory factor analysis (EFA)
Principal CFA was performed on the 23 items using group
B data. Kaiser’s Meyer Olkin measure of sampling ade-
quacy was 0.93, showing that the patterns of correlation
are relatively compact, and so factor analysis should pro-
duce distinct and reliable factors (Field 2000). Bartlett’s
Test of Sphericity was significant (P < 0.001), showing
that there were some relationships between the variables.
The factors were subjected to orthogonal (varimax) rota-
tion to maximize the dispersion of the loadings within
factors so that loading a smaller number of variables
highly onto each factor results in more interpretable clus-
ters of factors (Field 2000).
Factor analysis and scree test showed that a five-factor
was most appropriate. All 23 items had factor loadings of
greater than 0.4, communalities of greater than 0.5, no
hyperplane items (items failing to show salient loadings
on any of the factors) and no items with substantial cross-
loadings on other factors. The five factors accounted for
70% of variance in scores. Factor 1 (GRCS-IS) accounted
for 44.00% of the variance (eigenvalue = 10.12), factor II
(GRCS-IB) accounted for 8.29% (eigenvalue = 1.91), fac-
tor III (GRCS-IC) for 6.72% (eigenvalue = 1.55), factor IV
(GRCS-GE) for 5.96% (eigenvalue = 1.37) and factor V
(GRCS-PC) accounted for 4.44% of the variance
(eigenvalue = 1.02). Factor 1 appeared to assess cogni-
tive dysfunctions reflecting their inability to stop gam-
bling, factor II appeared to assess interpretative bias
among the gamblers, factor III taped into illusion of con-
trol such as superstitious beliefs, factor IV appeared to
assess gambling expectations and factor V appeared to
assess predictive control. Factor I had five items, factor 5
had six items and factors 2–4 had four items each. Factor
loadings and communalities are shown in Table 1. A
copy of the scale is included in Appendix I.
CFA
CFA with group B
CFA was conducted on group B data using EQS 5.7b, to
determine which model explained the data set most
fully. The c2 value is a statistic that assesses the fit
between the restricted hypothesized model and the
unrestricted sample data. When the c2 is non-significant
or small, it reflects that there is an acceptable fit to the
Addiction, 99, 757–769
762 Namrata Raylu & Tian P. S. Oei

Table 1 Factor loadings, communalities and Cronbach’s alpha coefficients for the GRCS.

Item GRCS-IS GRCS-IB GRCS-IC GRCS-GE GRCS-PC Communalities

12 0.848 0.821
7 0.821 0.745
2 0.811 0.789
17 0.706 0.617
21 0.704 0.644
10 0.814 0.841
5 0.811 0.852
20 0.783 0.769
15 0.778 0.753
13 0.856 0.844
18 0.816 0.808
3 0.731 0.651
8 0.709 0.664
1 0.803 0.756
6 0.781 0.838
11 0.706 0.688
16 0.702 0.600
22 0.763 0.618
19 0.585 0.505
23 0.536 0.511
9 0.475 0.554
4 0.442 0.562
14 0.430 0.533
% 44.00 8.29 6.72 5.96 4.44
a 0.89 0.91 0.87 0.87 0.77 0.93 (full scale)

% = percentage of variance; a = Cronbach’s alpha coefficient. N = 968.

data. However, for very large sample sizes it is often diffi-
cult to achieve a non-significant result (Marsh, Bella &
McDonald 1988). Thus, the fit of the model should be
interpreted on the basis of a range of other fit indices.
The EQS program produces a range of goodness of fit
indices including outputs for the Bentler Bonett Normed
Fit Index (NFI) and the Bentler Bonnet Non-Normed Fit
Index (NNFI), which takes into account the degrees of
freedom of the model), and the Comparative Fit Index
(CFI). By convention these values are regarded accept-
able if they are generally greater than 0.9 (Bentler
1995). The root mean square residual (RMR) reflects
the proportion of discrepancy between elements in the
sample and the hypothesized covariance matrix. If there
is a good fit between the hypothesized model and the
sample, the RMR will be close to 0.05 or lower. The root
mean square error of approximation (RMSEA), which is
based on population error of approximation measures
‘discrepancy per degree of freedom’ (Joreskog & Sorbom
1993, p. 124). A value of 0.05 or less is recognized as
suggesting a close fit (values up to 0.08 is recognized as
a reasonable error of approximation).
Using maximum likelihood estimation, three models
were tested. The first model to be examined was a one-
factor model in which all items were predicted to load
onto a single factor reflecting a general gambling cogni-
tion factor, with minimal clustering of cognitions. The
analyses showed that the single factor solution was not
a good fit of the data. All fitted indices were less than
the accepted value of 0.9 (i.e. CFI = 0.64; NFI = 0.62;
NNFI = 0.60). Furthermore, the RMR value of 0.09 and
RMSEA value of 0.15 were both outside the accepted
values.
The second model to be examined was the five-factor
model where factors were allowed to intercorrelate.
Although this was a better fit than the one-factor model,
the fit was not adequate as all fit indices were less than
0.9 (CFI = 0.80; NFI = 0.77; NNFI = 0.78). Also, both
the RMR and RMSEA values were outside the accepted
value (RMR = 0.34; RMSEA = 0.12), further suggesting
that the second model was not the best fit to the data.
The third model that was tested using the CFA was a
higher-order model, where the high level of covariation
between the five factors could be accounted for by some
higher-order factor of ‘gambling cognition’. The analyses
showed that this was a good fit of the data as all fit indices
were greater than 0.9 (CFI = 0.93; NFI = 0.91;
NNFI = 0.93). Furthermore, both the RMR and RMSEA
values were in the accepted range (RMR = 0.05;
RMSEA = 0.06).

© 2004 Society for the Study of Addiction Addiction, 99, 757–769


Validation CFA with group A
To validate the five-factor higher-order model, a CFA was
undertaken through EQS7.5b using the data from group
A of the random data split. Results indicated that the five-
factor higher-order model was a stable and a good fit of
the second data. Analyses showed that all fit indices were
over 0.9 (CFI = 0.92; NFI = 0.90; NNFI = 0.91). Further-
more, both the RMR and RMSEA values were acceptable
(0.06 and 0.07, respectively), thus suggesting a good fit.
Reliability and validity of the GRCS
Reliability and validity of the GRCS were examined using
both group A and group B data.
Reliability
Cronbach’s alpha for the overall scale was high
(alpha = 0.93). Moderate to high reliability was also
found for each of the five subscales: GRCS-GE
(alpha = 0.87); GRCS-IS (alpha = 0.89); GRCS-PC
(alpha = 0.77); GRCS-IC (alpha = 0.87); and GRCS-IB
(alpha = 0.91).
Factor intercorrelations for the five-factor model
To explore the correlation between the factors the factor
scores were computed by adding the items for each sub-
scale. The score for each subscale correlated significantly
(P = 0.01) with other subscale scores and the total score,
as shown in Table 2. The factors were found to be strongly
intercorrelated, with all values exceeding 0.4.
Validity
The three types of validity that were explored for the GRCS
using both group A and group B included criterion-related
validity, predictive validity and concurrent validity.
Concurrent validity
At the time of the study, there was an absence of similar
cognitive instruments with which to compare the GRCS.
Table 2 Factor intercorrelations between factors and total score.
Factor GRCS-GE GRCS-IS
GE 1.00
IS 0.53 1.00
PC 0.61 0.49
IB 0.57 0.52
IC 0.50 0.51
TOT 0.81 0.74
n = 968.
© 2004 Society for the Study of Addiction
Gambling Related Cognitions Scale (GRCS) 763
Consequently, to investigate the concurrent validity of the
questionnaire, a range of variables (e.g. anxiety, depres-
sion, stress, gambling behaviour and motivations
towards gambling) that the gambling literature has
shown to be positively correlated with PG were correlated
with the GRCS. Studies in the gambling literature have
shown that individuals that score higher on such vari-
ables tended to have higher GRC. Thus, we expected a
positive correlation between these variables (e.g. anxiety,
depression, stress, gambling behaviour and motivations
towards gambling) and the total number of GRC. Several
questionnaires were used to assess these variables (i.e.
MTGS, SOGS and DASS).
Results showed evidence of concurrent validity with
the total GRCS score correlating significantly with related
variables. Significant positive low correlations were estab-
lished with the DASS subscales (anxiety, depression and
stress) and positive moderate correlations were estab-
lished with the motivation towards gambling subscales
and SOGS score. All correlations were significant at 0.01
level (two-tailed). The correlations are shown in Table 3.
Criterion-related validity
In order to investigate whether the GRCS could discrimi-
nate between non-problem gamblers and probable prob-
lem gamblers, participants were divided into two groups
based on their scores on the SOGS: SOGS group 1 (those
with SOGS = 0, n = 620) and SOGS group 2 (those with
SOGS of 4 or higher, n = 71). To maximize the number of
participants in each SOGS groups, both data sets were
used (i.e. both group A and group B).
Results of an ANOVA showed that there was a signif-
icant difference between the two groups in relation to
their total score (F1,690 = 174.47, P < 0.001). Similar
results were found for each subscale, GRCS-GE
(F1,690 = 174.29, P < 0.001), GRCS-IS (F1,690 = 113.05,
P < 0.001), GRCS-PC (F1,690 = 92.91, P < 0.001),
GRCS-IB (F1,690 = 151.87, P < 0.001) and GRCS-IC
(F1,690 = 28.10, P < 0.001). Means and standard devia-
tions of the GRCS subscale and total scores for each of the
SOGS groups are shown in Table 4.
GRCS-PC GRCS-IB GRCS-IC GRCS-TOT
1.00
0.62 1.00
0.59 0.52 1.00
0.86 0.81 0.76 1.00
Addiction, 99, 757–769
764 Namrata Raylu & Tian P. S. Oei

Table 3 Concurrent validity.

Measures Subscales Correlations GRCS total

GTMS Intrinsic motivation toward knowledge 0.52**

Intrinsic motivation to accomplishment 0.56**
Intrinsic motivation to experience stimulation 0.54**
Extrinsic motivation—regulation through identification 0.51**
Extrinsic motivation—introjected external regulation 0.54**
Extrinsic motivation—external regulation 0.32**
DASS Anxiety 0.20**
Depression 0.15**
Stress 0.12**
SOGS Total 0.43**

**P < 0.01; n = 968.

Table 4 Means (and standard deviations) of subscales and total Predictive validity
GRCS scores for the two SOGS groups.
Multiple regression analyses were conducted to investi-
SOGS group 1 SOGS group 2 gate the extent to which GRCS subscales and GRCS total
Factor (n = 620) (n = 71)
scores could predict level of gambling problem. Age and
GRCS-GE 6.89 (4.14) 14.08 (5.87) gender were controlled for all these multiple regression
GRCS-IS 5.94 (3.45) 11.31 (7.37) analyses. First, a multiple regression analysis was com-
GRCS-PC 11.17 (6.17) 18.72 (6.83) pleted with SOGS as the dependent variable and the
GRCS-IB 5.73 (3.86) 12.07 (5.89) GRCS-TOT as the independent variable. Nineteen per cent
GRCS-IC 5.54 (3.58) 7.99 (4.82) (19% adjusted: 18% by GRCS-TOT and 1% by gender) of
GRCS-TOT 35.28 (16.81) 64.17 (22.31) the variance in SOGS scores was accounted for by the pre-
dictors (r = 0.44, P < 0.01). Participants scoring higher
on SOGS score were more likely to score higher on
Discriminant analyses were completed to explore GRCS-TOT.
the ability of the GRCS subscale scores and GRCS The second multiple regression analyses used SOGS
total scores to classify participants into the two SOGS score as the dependent variable and the five GRCS sub-
groups. The two SOGS groups were regarded as the scales as the independent variables. Twenty-seven per
grouping variables, while the five GRCS subscale cent (27% adjusted) of the variance in SOGS scores was
scores and GRCS total score were regarded as the pre- accounted for by the predictors (r = 0.52, P < 0.01). Both
dictor variables. The first discriminant analyses (using the demographic factors could not significantly predict
the subscales of GRCS as predictors) revealed that the SOGS scores. The proportion of variance in SOGS scores
discriminant function (L = 0.75, c2 = 200, P < 0.001) accounted by GRCS-GE, GRCS-IS, GRCS-IB and GRCS-IC
was significant. The univariate F ratios were also sig- could significantly predict SOGS scores. These four sub-
nificant for all subscales (P < 0.001). Subscales GRCS- scales accounted for 4%, 6%, 3% and 3% of variance,
GE, GRCS-IB, GRCS-IS and GRCS-PC correlated highly respectively. The GRCS-PC could not significantly predict
with discriminant function (r = 0.81, r = 0.76, SOGS scores. Together, the five subscales accounted
r = 0.66 and 0.33, respectively), while subscales approximately 16% of the variance in SOGS scores.
GRCS-IC correlated moderately with the discriminant The results of the second multiple regression analyses
function (r = 0.33). The discriminant functions classi- also demonstrated that the subscale GRCS-IC had a neg-
fied correctly 86% of the participants into the two ative beta and partial correlation in the multiple regres-
SOGS groups (88% of SOGS group 1 and 72% of sion analyses. However, the bivariate correlations of all
SOGS group 2). The second discriminant analyses subscales with the total SOGS score were significantly
(using total GRCS score as the predictor variable) also positive. The bivariate correlation between the SOGS total
revealed a significant discriminant function (L = 0.88, score and each of the GRCS subscales were as follows:
c2 = 185, P < 0.001). The univariate F ratio was also GRCS-GE (0.42), GRCS-IS (0.41), GRCS-PC (0.31), GRCS-
significant (P < 0.001). The discriminant function IB (0.39) and GRCS-IC (0.19). In order to explore
classified correctly approximately 85% of the partici- whether the negative beta for GRCS-IC was due to multi-
pants into the two SOGS groups (87% of SOGS group collinearity, tolerance statistics was calculated. Tolerance
1 and 79% of SOGS group 2). values below 0.1 (Field 2000) or 0.2 (Menard 1995)

© 2004 Society for the Study of Addiction Addiction, 99, 757–769

 Gambling Related Cognitions Scale (GRCS) 765

Table 5 Standard multiple regression analyses results with SOGS total score as dependent variable.

Type of Independent Partial

analysis variables Beta t correlation (R) R2

Analyses Age -0.01 -0.42 -0.01 0.00

with GRCS Gender +0.04 +1.44 +0.05 0.00
subscales GRCS-GE +0.22 +6.04 +0.19 0.04***
GRCS-IS +0.27 +7.88 +0.25 0.06***
GRCS-PC +0.03 +0.73 +0.02 0.00
GRCS-IB +0.23 +5.68 +0.18 0.03***
GRCS-IC -0.21 -5.45 -0.17 0.03***
Analyses Age +0.08 +2.64 +0.09 0.01**
with total Gender +0.05 +0.16 +0.05 0.00
GRCS score GRCS-TOT +0.42 +0.42 +0.42 0.18***

**P < 0.01; ***P < 0.001; n = 968.

indicate serious problems with multi-collinearity. How- Table 6 Means (and standard deviations) of GRCS subscale and
ever, the tolerance values of all subscales were found to be total score for males and females.
above 0.2. One possible reason for obtaining a negative Subscale Female Male
beta and positive bivariate correlation (e.g. for GRCS-IC)
could be the presence of a suppressor variable, where a GRCS-GE** 7.68 (4.66) 10.10 (5.97)
predictor variable appears to have a significant effect only GRCS-IS** 6.23 (3.87) 7.30 (4.77)
GRCS-PC** 11.99 (6.54) 14.42 (7.19)
when another variable is constant (Field 2000). Conse-
GRCS-IB** 6.30 (4.31) 7.99 (5.66)
quently, a further multiple regression analyses was per-
GRCS-IC 5.97 (3.83) 6.20 (3.98)
formed with GRCS-IC as a constant to investigate this GREC-TOT** 38.16 (18.47) 46.01 (22.32)
possibility further. When the GRCS-IC subscale was held
as a constant, the GRCS-PC subscale could predict SOGS **Significant difference at P < 0.01; n = 968.
scores significantly, similar to the other subscales. Results
of the multiple regression analyses are displayed in
Table 5. with the categories provided by Toneatto et al. (1997),
where interpretative bias reflects cognitions relating to
reframing gambling outcomes, predictive control reflects
Gender differences
cognitions relating to ability to predict gambling out-
There was a significant gender difference on four of the comes and illusion of control reflects cognitions relating
GRCS subscales including: GE [F1,483 = 24.43, P < 0.01], to ability to control gambling outcomes. The latter two
IS [F1,483 = 7.28, P < 0.01], PC [F1,483 = 14.35, P < 0.01] are similar to the cognitions found for other addictions
and IB [F1,483 = 13.55, P < 0.01]. The gender difference (Beck et al. 1993; Oei & Baldwin 1994).
for GRCS-IC was not significant [F1,483 = 0.40 (NS)]. CFA confirmed the EFA five-factor model as the best fit
There was also a significant gender difference in the total for the data, suggesting the soundness of the psychomet-
GRCS score [F1,483 = 17.28, P < 0.01]. Males had higher ric properties of the GRCS (Appendix I). However, a high
GRCS scores than females. The means and standard devi- level of covariation was found between factors, which
ations of scores are shown in Table 6. Males also scored could be explained by a single, higher-order model, in
higher on SOGS than females [F1,966 = 20.86, P < 0.001]. which first order factors of cognition subtypes loaded
upon a factor of general cognition. Although the first five
order factors loaded strongly upon the higher order cog-
DISCUSSION nition factor, there was sufficient unique variance (70%)
explained by the five first-order factors to justify
The present study focused on developing and validating a regarding them as dimensions worthy of independent
questionnaire that can screen for a range of GRC in a consideration.
community sample. The five-factor model reflected The high level of internal consistency for the entire
dimensions of interpretative bias/control, illusion of con- scale indicates that computing the total score is appro-
trol, predictive control, perceived inability to stop gam- priate and useful for research and/or clinical screening
bling and expectations of gambling, and provided a good activities and assisting in planning treatment. The
fit of the data. The former three factors are consistent internal consistencies for the five subscales were also

© 2004 Society for the Study of Addiction Addiction, 99, 757–769

766 Namrata Raylu & Tian P. S. Oei
high. However, because the data used in this study were
cross-sectional in nature, it does not assess the sensitiv-
ity of the scale to changes over time. Future studies
need to determine the test–retest reliability of the 23-
item scale, whether or not the scores change over time
and where it is sensitive to changes in GRCS, due espe-
cially to the result of successful cognitive therapy.
Preliminary evidence of the latter has been provided by
a study being conducted currently by the authors using
the GRCS.
Significant positive correlations were demonstrated
with other instruments assessing gambling-related vari-
ables including gambling-related thoughts, motivations
towards gambling scale and gambling problems. Signifi-
cant positive correlations were also established with
DASS, a questionnaire used to assess mood states (levels
of anxiety, depression and stress). These support previous
studies that show that higher levels of GRC are associated
with increased frequency of gambling, PG and with neg-
ative psychological states, such as anxiety, depression
and stress (Raylu & Oei 2002). Additional testing should
be conducted to assess with which variables the GRCS
does not correlate (i.e. divergent validity of the GRCS).
Given the significant correlations between the subscales,
there is a need to explore construct validity separately for
the five subscales. This can be achieved by using collateral
measures that will support the distinct convergent and
divergent validity of each scale.
The GRCS total score and the GRCS subscales have the
ability to discriminate between non-problem gamblers
and problem gamblers. This also supports research that
suggests that GRC can be an important factor in the
development and maintenance of PG (Raylu & Oei 2002).
Higher GRCS scores were associated with higher SOGS
scores. However, given the cross-sectional nature of the
data, further research needs to be conducted to explore
whether GRCS could predict PG at a future date. Further-
more, it appeared that subscale GRCS-IC was masking the
effects of GRCS-PC. Thus, when using subscales to predict
PG, this needs to be taken into consideration. The GRCS
appears to perform somewhat differently than the other
scales. For example, the effect size for this scale in Table 4
is much lower with the SOGS than other scales, and it
produces a negative beta in regression analyses. Future
studies need to explore why this scale is distinct from
other subscales. It may be more appropriate to use total
GRCS scores to predict PG, rather than the subscale
scores.
One of the major strengths of this study is a large sam-
ple size (n = 968). Furthermore, a sample size of 484 was
used for exploratory analyses. A high ratio of participants
to GRCS items (ratio 10 : 1) ensured that stable factors
could be detected in EFA (Kaiser 1970). Also, CFA was
completed with a separate sample, which also had a large
© 2004 Society for the Study of Addiction
sample size (n = 484). Although our findings provide
important support for the GRCS as a measure of GRC
among gamblers, future studies need to assess the psy-
chometric properties with other samples including clini-
cal samples, especially those in treatment groups. Clinical
validity of this sample is also restricted by the fact that the
participants were self-selected into the study. This is
important, as most of the participants in this study were
non-problem gamblers and previous studies/reviews (e.g.
Walker 1992; Murray 1993; Griffiths 1996; Zuckerman
1999; Raylu & Oei 2002) suggest that there are signifi-
cantly more GRC among frequent/problem gamblers
than non-frequent/non-problem gamblers. As the SOGS
scores for community samples are not distributed nor-
mally, there is a strong need to explore the validity of the
GRCS (e.g. predictive and criterion related validity)
among clinical samples, where the SOGS scores generally
tend to be normally distributed. For example, results
showed that 18% of the variance in GRCS-TOT predicted
SOGS scores. Even though this is a significant result, one
would expect this percentage to increase in clinical
samples.
The main aim of this study was to develop and validate
an instrument to assess GRC that have been identified
irrational in nature (e.g. items that concern erroneous
beliefs about success at gambling) as well as items related
to beliefs about self in relation to gambling (i.e. gambling-
related expectancies and perception of inability to control
their gambling behaviour) rather than focusing on par-
ticular subtypes. Consequently, various cognitions were
assessed in a more limited fashion compared to focusing
on only one domain of gambling cognition. For example,
this study did not separate cognitions surrounding per-
sonal skill/ability to influence outcomes, or acts that they
believe may influence outcomes, because both reflect
beliefs/behaviours to control gambling outcomes. How-
ever, an instrument such as the GRCS will provide the
first step towards developing more specialized instru-
ments to assess particular domains/types of GRC (e.g.
subtypes of cognitions within the illusion of control cat-
egory, only erroneous cognitions or beliefs about self in
relation to gambling).
A significant gender difference was found in four of
the five factors as well as the total score, where males
tended to have significantly higher GRCS scores than
females. Most studies in this area have not looked at the
gender differences in GRC (Griffiths 1993, 1990a,b,c).
Toneatto et al. (1997) reported no significant gender dif-
ference in his study. However, it must be noted that this
was based on 38 participants, which were predominantly
males. Consistent with other studies (Volberg & Stead-
man 1988; Volberg 1994; Wood & Griffiths 1998), this
study found higher PG in males than females. Thus, it is
possible that the higher rate of GRCS in males compared
Addiction, 99, 757–769

to females may be accounted for by higher rates of PG
among males. Also, males and females have differences in
the structural as well as functional organization of the
brain, thus leading to differences in responsiveness of the
reward system, susceptibility to addictive problems, dif-
ferences in information processing and different cognitive
strategies utilized in problem-solving, all of which have
significant impacts on cognition and behaviour (Pogun
2001). Because there are limitations associated with
using a sample that is two-thirds female, especially when
male gender is a known risk factor for development of
gambling problems, future studies need to a explore these
gender differences further.
In conclusion, GRCS has good psychometric proper-
ties. It has good validity and reliability, suggesting that
GRCS is a useful tool to assess cognitions among non-
clinical gamblers. It is a much-needed self-report mea-
sure to help screen for individuals with high levels of GRC
among a community group and explore those that may
be more at risk of developing gambling problems. This
will provide a step towards developing similar tools or
adapting this scale for a clinical sample of problem gam-
blers. Consequently, this will assist researchers and clini-
cians not only to identify the types of GRC that occur for a
gambler, but it will also aid in assessing success of cogni-
tive therapy in both clinical and research settings for
problem gamblers. More importantly, it provides the first
step towards developing more specialized instruments to
assess specific domains of GRC.
References
American Psychiatric Association (APA) (1994) Diagnostic and
Statistical Manual of Mental Disorders, 4th edn. Washington,
DC: APA.
Anthony, M. A., Bieling, P. J., Cox, B. J., Murray, W. E. & Swin-
son, R. P. (1998) Psychometric properties of 42 item and 21
item version of the Depression, Anxiety and Stress Scale in
clinical groups and a community sample. Psychological Assess-
ment, 10, 176–181.
Baldwin, A. R., Oei, T. P. S. & Young, R. (1993) To drink or not
to drink: the differential role of alcohol expectancies and
drinking refusal self-efficacy in quantity and frequency of
alcohol consumption. Cognitive Therapy and Research, 17,
511–530.
Beck, A. T., Wright, F. D., Newman, C. F. & Liese, B. S. (1993)
Cognitive Therapy of Substance Abuse. New York: Guilford Press.
Bentler, P. M. (1995) EQS Structural Equations Program Manual.
Encino, CA: Multivariate Software.
Black, D. W. & Moyer, T. (1998) Clinical features and psychiatric
comorbidity of subjects with pathological gambling behav-
iour. Psychiatric Services, 49, 1434–1439.
Blaszczynski, A. (1994) Criminal offences in pathological gam-
blers. Psychiatry, Psychology and Law, 1, 129–138.
Chantal, Y., Vallerand, R. J. & Vallières, E. F. (1994) On the con-
struction and validation of the Gambling Motivation Scale
(GMS). Society and Leisure, 17, 189–212.
© 2004 Society for the Study of Addiction
Gambling Related Cognitions Scale (GRCS) 767
Coulombe, A., Ladouceur, R., Desharnais, R. & Jobin, J. (1992)
Erroneous perceptions and arousal among regular and occa-
sional video poker players. Journal of Gambling Studies, 8, 235–
244.
DeVellis, R. F. (1991) Scale Development. Newbury, CA: Sage.
Field, A. (2000) Discovering Statistics Using SPSS for Windows.
London: Sage.
Gaboury, A. & Ladouceur, R. (1989) Erroneous perceptions and
gambling. Journal of Social Behavior and Personality, 4, 411–
420.
Gilovich, T. & Douglas, C. (1986) Biased evaluations of ran-
domly determined gambling outcomes. Journal of Experimental
Social Psychology, 22, 228–241.
Griffiths, M. D. (1990a) The cognitive psychology of gambling.
Journal of Gambling Studies, 6, 31–42.
Griffiths, M. D. (1990b) Addiction to fruit machines: a prelimi-
nary study among males. Journal of Gambling Studies, 6, 113–
126.
Griffiths, M. D. (1990c) The acquisition, development and main-
tenance of fruit machine gambling in adolescence. Journal of
Gambling Studies, 6, 193–204.
Griffiths, M. D. (1993) Factors in problem adolescent fruit
machine gambling. Journal of Gambling Studies, 9, 31–45.
Griffiths, M. D. (1994) The role of cognitive bias and skill in fruit
machine gambling. British Journal of Psychology, 85, 351–
369.
Griffiths, M. D. (1996) Adolescent Gambling. London: Routledge.
Joreskog, K. G. & Sorbom, D. (1993) LISREL 8: Structural Equa-
tion Modelling with the SIMPLIS Command Language. Chicago,
IL: Scientific Software International.
Joukhador, J., Maccallum, F. & Blaszczynski, A. (2003) Differ-
ences in cognitive distortions between problem and social
gamblers. Psychological Reports, 92, 1203–1214.
Kaiser, H. F. (1970) A second generation little jiffy. Psy-
chometrika, 35, 401–145.
Lee, N. K. & Oei, T. P. S. (1993) The importance of alcohol
expectancies and drinking refusal self-efficacy in the quantity
and frequency of alcohol consumption. Journal of Substance
Abuse, 5, 379–390.
Lee, N. K., Oei, T. P. S. & Greeley, J. D. (1999) The interaction of
alcohol expectancies and drinking refusal self-efficacy in high
and low risk drinkers. Addiction Research, 7, 91–102.
Lesieur, H. R. & Blume, S. B. (1987) The South Oaks Gambling
Screen (SOGS): a new instrument for the identification of
pathological gamblers. American Journal of Psychiatry, 144,
1184–1188.
Lesieur, H. R. & Blume, S. B. (1990) Characteristics of patholog-
ical gamblers identified among patients on a psychiatric
admissions service. Hospital and Community Psychiatry, 41,
1009–1012.
Lesieur, H. R. & Heineman, M. (1988) Pathological gambling
among youthful multiple substance abusers in a therapeutic
community. British Journal of Addiction, 83, 765–771.
Lovibond, S. H. & Lovibond, P. F. (1995) The structure of nega-
tive emotional states: comparison of the Depression Anxiety
Stress Scales (DASS) with the Beck Depression and Anxiety
Inventories. Behavior, Research and Therapy, 33, 335–343.
Marsh, H. W., Balla, J. R. & McDonald, R. P. (1988) Goodness-of-
fit indexes in confirmatory factor analysis: the effect of a sam-
ple size. Psychological Bulletin, 103, 391–401.
Menard, S. (1995) Applied Logistic Regression Analysis. Thousand
Oaks, CA: Sage.
Meyer, G. & Fabian, T. (1996) Pathological gambling and crim-
inal culpability: an analysis of forensic evaluations presented
Addiction, 99, 757–769
768 Namrata Raylu & Tian P. S. Oei
to German penal courts. Journal of Gambling Studies, 12, 33–
47.
Murray, J. B. (1993) Review of research on pathological gam-
bling. Psychological Reports, 72, 791–810.
Oei, T. P. S. & Baldwin, A. (1994) Expectancy theory: a two-
process model of alcohol use and abuse. Journal of Studies on
Alcohol, 55, 525–534.
Oei, T. P. S. & Burrow, R. (2000) Alcohol expectancy and drink-
ing refusal self-efficacy: a test of specificity theory. Addictive
Behaviors, 25, 499–507.
Oei, T. P. S., Fergusson, S. & Lee, N. K. (1998) The differential
role of alcohol expectancies and drinking refusal self-efficacy
in problem and nonproblem drinkers. Journal of Studies on
Alcohol, 59, 704–711.
Oei, T. P. S. & Raylu, N. (2004) Familial influence on offspring
gambling: a cognitive mechanism for transmission of gam-
bling behaviour in families. Psychological Medicine, in press.
Petry, N. M. & Armentano, C. (1999) Prevalence, assessment,
and treatment of pathological gambling: a review. Psychiatric
Services, 50, 1021–1027.
Pike, C. K. (2002) Measuring video gambling: instrument devel-
opment and validation. Research on Social Work Practice, 12,
389–407.
Pogun, S. (2001) Sex differences, in brain and behaviour:
emphasis on nicotine, nitric oxide and place learning. Interna-
tional Journal of Psychophysiology, 42, 195–208.
Rapee, R. M., Craske, M. G., Brown, T. A. & Barlow, D. H. (1996)
Measurement of perceived control over anxiety-related
events. Behavior Therapy, 27, 279–293.
Raylu, N. & Oei, T. P. S. (2002) Pathological gambling: a com-
prehensive review. Clinical Psychology Review, 22, 1009–
1061.
Raylu, N. & Oei, T. P. S. (2004) The Gambling Urge Scale
(GUS): development, confirmatory factor validation and psy-
chometric properties. Psychology of Addictive Behaviors, in
press.
Rosenthal, R. J. (1989) Pathological gambling and problem
gambling: problems of definition and diagnosis. In: Shaffer, H.
J., Stein, S. A. et al., eds. Compulsive Gambling: Theory, Research,
and Practice, pp. 101–125. Lexington, MA: Lexington Books.
Sharpe, L. (2002) A reformulated cognitive-behavioural model
of problem gambling: a biopsychosocial perspective. Clinical
Psychology Review, 22, 1–25.
Steenbergh, T. A., Meyers, A. W., May, R. K. & Whelan, J. P.
(2002) Development and validation of the Gamblers’ Beliefs
Questionnaire. Psychology of Addictive Behaviors, 16, 143–
149.
Stevens, J. P. (1992) Applied Multivariate Statistics for the Social
Sciences, 2nd edn. Hillsdale, NJ: Erlbaum.
Sylvain, C. & Ladouceur, R. (1992) Corrective cognition and
gambling habits of players of video poker. Canadian Journal of
Behavioural Science, 24, 479–489.
Sylvain, C., Ladouceur, R. & Boisvert, J. M. (1997) Cognitive and
behavioural treatment of pathological gambling: a controlled
study. Journal of Consulting and Clinical Psychology, 65, 727–
732.
Toneatto, T. (1999) Cognitive psychopathology of problem gam-
bling. Substance Use and Misuse, 34, 1593–1604.
Toneatto, T., Blitz-Miller, T., Calderwood, K., Dragonetti, R. &
Tsanos, A. (1997) Cognitive distortions in heavy gambling.
Journal of Gambling Studies, 13, 253–266.
Volberg, R. A. (1994) The prevalence and demographics of
pathological gamblers: implications for Public Health. Ameri-
can Journal of Public Health, 84, 237–241.
© 2004 Society for the Study of Addiction
Volberg, R. A. & Steadman, H. J. (1988) Refining prevalence esti-
mates of pathological gambling. American Journal of Psychia-
try, 145, 502–505.
Wagenaar, W. A. (1988) Paradoxes of Gambling Behavior. Hills-
dale, NJ: Erlbaum.
Walker, M. B. (1992) Irrational thinking among slot machine
players. Journal of Gambling Studies, 8, 245–261.
Walters, G. D. & Contri, D. (1998) Outcome expectancies for
gambling: empirical modeling of a memory network in fed-
eral prison inmates. Journal of Gambling Studies, 14, 173–
191.
Wood, R. T. A. & Griffiths, M. D. (1998) The acquisition,
development and maintenance of lottery and scratchcard
gambling in adolescence. Journal of Adolescence, 21, 265–
273.
Zuckerman, M. (1999) Vulnerability to Psychopathology: a
Biosocial Model. Washington, DC: American Psychological
Association.
APPENDIX I
The Gambling Related Cognition Scale (GRCS)
Please indicate (by circling) the extent to which you agree
with the value expressed in each statement (1 = strongly
disagree; 2 = moderately disagree; 3 = mildly disagree;
4 = neither agree or disagree; 5 = mildly agree;
6 = moderately agree; 7 = strongly agree)
1 Gambling makes me happier.
2 I can’t function without gambling.
3 Praying helps me win.
4 Losses when gambling, are bound to be followed by a
series of wins.
5 Relating my winnings to my skill and ability makes
me continue gambling.
6 Gambling makes things seem better.
7 It is difficult to stop gambling as I am so out of control.
8 Specific numbers and colours can help increase my
chances of winning.
9 A series of losses will provide me with a learning expe-
rience that will help me win later.
10 Relating my losses to bad luck and bad circumstances
makes me continue gambling.
11 Gambling makes the future brighter.
12 My desire to gamble is so overpowering.
13 I collect specific objects that help increase my
chances of winning.
14 When I have a win once, I will definitely win again.
15 Relating my losses to probability makes me continue
gambling.
16 Having a gamble helps reduce tension and stress.
17 I’m not strong enough to stop gambling.
18 I have specific rituals and behaviours that increase
my chances of winning.
19 There are times that I feel lucky and thus, gamble
those times only.
Addiction, 99, 757–769

20 Remembering how much money I won last time
makes me continue gambling.
21 I will never be able to stop gambling.
22 I have some control over predicting my gambling
wins.
23 If I keep changing my numbers, I have less chances of
winning than if I keep the same numbers every time.
Scoring
To obtain the raw subscale scores add values of items for
© 2004 Society for the Study of Addiction
Gambling Related Cognitions Scale (GRCS) 769
each subscale. To obtain total raw GRCS score, add the
five raw subscale scores.
To obtain mean subscale scores divide each of the raw
subscale scores by the number of items in each subscale.
To obtain a total mean GRCS score, add the five means
subscale scores. The items that belong to each subscale
are as follows:
Gambling expectancies: 1, 6, 11, 16
Illusion of control: 3, 8, 13, 18
Predictive control: 4, 9, 14, 19, 22, 23
Inability to stop gambling: 2, 7, 12, 17, 21
Interpretive bias: 5, 10, 15, 20
Addiction, 99, 757–769