Contents lists available at ScienceDirect

Addictive Behaviors
journal homepage: www.elsevier.com/locate/addictbeh

An Assessment of the Validity of the Gamblers Belief Questionnaire☆
Kahlil S. Philander
a
b

a,b

b,⁎

, Sally M. Gainsbury

b

, Georgia Grattan

T

Carson College of Business, School of Hospitality Business Management, Washington State University, 915 North Broadway, Everett, WA 98201, United States of America
Science Faculty, Brain and Mind Centre, School of Psychology, University of Sydney, Australia

HIGHLIGHTS

beliefs can lead people to gamble beyond limits of affordability and contribute to harms.
• Irrational
Gamblers Belief Questionnaire is a 20-item test that can be self-administered.
• The
US adults completed an online survey asesssing perception of skill and chance.
• 184
items have validity in predicting distorted thoughts related to illusion of control and the gambler's fallacy.
• GBQ
There
is an opportunity to improve the parsimony of GBQ questions and the discriminant ability of its subscales.
•
ARTICLE INFO

ABSTRACT

Keywords:
Gamblers belief questionnaire
Cognitive distortions
Problem gambling
Illusion of control
Gamblers fallacy

Cognitive distortions in gambling are irrational thoughts that cause an individual to overestimate their level of
control over the outcome of the game and diminish the role of chance. Due to their strong relation to gambling
disorders, they are a particularly important characteristic to assess and understand in gamblers. Although numerous measures of gambling-related cognitive distortions exist, studies assessing criterion validity are scarce. In
this study, we develop several tests of the Gamblers Belief Questionnaire (GBQ), a versatile and widely used
scale. A sample of 184 U.S. adults was recruited through Amazon Mechanical Turk to complete an online study
that included measurement of the GBQ and an assessment of the perceived role of skill and chance in various
gambling and non-gambling activities. In addition to a confirmatory factor analysis of the scale, three novel
validation tests were developed to understand whether the GBQ subscales can identify and discriminate measures of illusion of control and gambler's fallacy distortions. Our validation tests demonstrate that the scale does
measure both distortions, providing information about gamblers' cognition that is unexplained by gambling
problems, frequency of play, and demographics. Conversely, our analysis of the factor structure does not show
good fit. We conclude that the GBQ measures gambling-related cognitive distortions, but there may be an opportunity to reduce the number of scale items and further refine precision of the two subscales.

1. Introduction
Cognitive distortions in gambling are irrational thoughts that cause
individuals to overestimate their level of control over the outcome of
the game and diminish the role of chance (Barrault & Varescon, 2013).
Research on gamblers' cognitive distortions suggests that they are an
important component to understand both normative and disordered
gambling behavior (Hodgins, Stea, & Grant, 2011). Notably, evidence
suggests that cognitive distortions lead to continued gambling despite
significant financial loss (Goodie & Fortune, 2013; Walker, 1992), and
play a causal role in the maintenance and development of gambling

disorders (e.g. Blaszczynski & Nower, 2002; Goodie & Fortune, 2013;
Hodgins et al., 2011; Jacobsen, Knudsen, Krogh, Pallesen, & Molde,
2007; Xian et al., 2008).
Identification of cognitive distortions is important to clinical practice in the treatment of gambling problems. Correcting distorted
thoughts is often part of treatment protocols (Hodgins et al., 2011;
Sharpe & Tarrier, 1993), and gamblers with more distorted thoughts are
more likely to relapse from their recovery goals (Oei & Gordon, 2008).
In a review of the role of cognitive distortions in treatment, Fortune and
Goodie (2012) find that strategies focused on the correction of cognitive
distortions, either alone or in conjunction with cognitive behavioral

This work was supported by an Australian Research Council Discovery Early Career Research Award [DE1060100459] awarded to Dr. Sally Gainsbury.
This research complied with APA ethical standards and was approved by the University of Sydney Human Research Ethics Committee.
⁎
Corresponding author.
E-mail address: sally.gainsbury@sydney.edu.au (S.M. Gainsbury).
☆

https://doi.org/10.1016/j.addbeh.2019.05.029
Received 5 March 2019; Received in revised form 25 May 2019; Accepted 25 May 2019
Available online 31 May 2019
0306-4603/ © 2019 Elsevier Ltd. All rights reserved.

K.S. Philander, et al.

therapies, generally show therapeutic success that is sustained across
multiple follow-up periods.
Although several measures of gambling-related cognitive distortions
are available, construct validation has tended to focus on face and
content validity, rather than tests of criterion validity after scale development. Where studies have explored other dimensions of validity,
analysis tends to be limited to correlations with measures of gambling
problems (Goodie & Fortune, 2013), and covariance analyses of factor
structure (e.g. Raylu & Oei, 2004; Smith, Woodman, Drummond, &
Battersby, 2016). Despite cognitive distortions being well studied in
gamblers, the related measures would benefit from more diverse assessments of validity.
In this study, we develop several novel tests to evaluate the criterion
validity of the Gamblers Belief Questionnaire (GBQ) (Steenbergh,
Meyers, May, & Whelan, 2002), and its associated Illusion of Control
(GBQ-IoC) and Luck/Perseverance (GBQ-LP) subscales. We focus on the
GBQ due to its popularity in the literature and the versatility of its use.
In a review of gambling cognitive distortions scales, only one other
scale had more published studies linking the measure to gambling
disorders, and that scale was designed for use with video lottery
terminal players only (Goodie & Fortune, 2013).1 Our tests examine
how GBQ scores are related to differences in perceived skill and chance
involved in dissimilar game types and lengths of play. As part of our
tests, we control for gambling problems and frequency of play to demonstrate that the scale provides non-redundant information about
gamblers' cognition. We also test the validity of the GBQ's factor
structure through a confirmatory factor analysis.
Due in large part to role that distorted thinking is believed to play in
the etiology of gambling disorder (Blaszczynski & Nower, 2002), improving psychometric understanding of cognitive distortions would be a
valuable contribution for treatment, prevention and research. A wellvalidated tool would help inform related modifications and refinements
to treatment, product designs, public health communications, and other
topics of individual and societal interest. In addition, emerging technologies like virtual reality, mobile gambling, and video game-like
products are changing the nature of gambling products, creating an
increasing need to validate a broader psychometric measure that can
assess cognitive distortions across a range of applications (Gainsbury &
Blaszczynski, 2017; King, Gainsbury, Delfabbro, Hing, & Abarbanel,
2015).

2000; Dunn & Wilson, 1990; Ladouceur & Maryland, 1987; Toneatto,
Blitz-Miller, Calderwood, Dragonetti, & Tsanos, 1997). For a thorough
review of gambling-related cognitive distortions and related measures,
see related works by Fortune & Goodie (Fortune & Goodie, 2012;
Goodie & Fortune, 2013).
Langer (1977) discusses two skill and luck concepts that are important to understand cognitive distortions in the context of games. The
first is that individuals may not make distinctions between skill and
chance elements because of their simultaneous presence in many activities. Skill and luck are often quite closely associated in subjective
experiences, which makes them hard to distinguish and reinforces illusions of control. For example, a winning hand in poker that may be
attributable to luck, skill, or a combination of the two factors. Second,
Langer states that individuals have a general incentive in their lives to
develop a mastery over their environment, and that complete mastery
would include an ability to ‘beat the odds’. She notes that individuals'
attempts to achieve competency are incompatible with viewing chance
events as uncontrollable. This may partially explain why individuals
rely on heuristics for prediction, which subsequently reinforce cognitive
distortions like the gambler's fallacy.
1.2. Gamblers belief questionnaire
The GBQ is a self-administered scale used to assess gambling-related
cognitive distortions. The scale was originally validated through five
studies described in Steenbergh et al. (2002). The first study included a
review of literature to create a list of items to assess general gamblingrelated cognitive distortions, which was then reviewed by three experts
in related fields. The second study used responses from a sample of
community members and undergraduate students to conduct an exploratory factor analysis on the items, leading to a 21-item scale with
two factors. The third study assessed reliability with an undergraduate
sample, whose test/re-test correlations were examined after a two-week
period, showing a correlation of 0.77. The fourth study demonstrated
convergence of GBQ scores with measures of gambling disorders in
community and student samples, and the fifth study demonstrated that
the scale was unrelated to a measure of social desirability in the second
study's student sample. Validated translations of the scale were subsequently produced in Spanish (Winfree, Meyers, & Whelan, 2013), Chinese (Wong & Tsang, 2012), and Italian (Marchetti et al., 2016). The
scale was revised to 20-items in a later study of treatment-seeking
disordered gamblers, based on item content consideration (Winfree,
Ginley, Whelan, & Meyers, 2015). We evaluate the 20-item version of
the scale.
Items are measured on a 7-point scale (1 = strongly agree,
7 = strongly disagree), which are reverse coded and summed. Higher
scores indicate higher levels of cognitive distortions. There are two
subscales: 1) An 8-item illusion of control construct; and 2) a 12-item
luck/perseverance construct. The GBQ-IoC is broadly composed of illusion of control related questions and the GBQ-LP is broadly composed
of gambler's fallacy related questions, although some questions crossloaded at typical threshold scores (Hair, Black, Babin, Anderson, &
Tatham, 2010).
Cronbach's alpha was not reported in Steenbergh et al. (2002), but
Mattson, MacKillop, Castelda, Anderson, and Donovick (2008) estimate
GBQ reliability with an α of 0.93 in a sample of college undergraduates.
The GBQ-IoC α was 0.89 and the GBQ-LP α was 0.94 in the same study.
Winfree et al. (2015) estimated the GBQ α at 0.87 in a clinical sample of
treatment seeking gamblers. The Spanish-translated version of the GBQ
shows similar psychometric properties, with a valid factor structure
with some cross-loadings, and high reliability scores with a GBQ α of
0.95, a GBQ-IoC α of 0.86 and a GBQ-LP α of 0.96 (Winfree et al.,
2013).
The only confirmatory factor analysis (CFA) was conducted by

1.1. Cognitive distortions in gambling
Distortions present themselves in gamblers through more than one
phenomenon. One source of bias in cognition is the use of heuristics,
such as availability or representativeness. The frequently cited “gambler's fallacy” is an error in representativeness, where a gambler believes that independent events are correlated as part of reaching a longrun average. For example, a gambler may believe an odd number is
more likely to land on a roulette wheel after several consecutive even
numbers have occurred, despite the events being statistically independent. The gambler's fallacy is demonstrated broadly across a
range of games (e.g. Clotfelter & Cook, 1993; Sundali & Croson, 2006;
Xu & Harvey, 2014), and is found to be predictive of gambling disorder
risk (Holtgraves, 2009).
Non-heuristic cognitive distortions also present themselves in
gamblers. The most noteworthy example is the illusion of control. An
illusion of control occurs when a player believes that her probability of
personal success is higher than the objective probability of success
(Langer, 1975). Experimental studies have demonstrated these thoughts
among gamblers in many different scenarios (Davis, Sundahl, Lesbo,
The authors of the review also note, “[the GBQ] has been investigated in the
most extensive number of studies, with a large collective N coming from a large
diversity of laboratories, a large effect size, and a narrow 95% CI.”

K.S. Philander, et al.

Pilatti, Cupani, Tuzinkievich, and Winfree (2016) on the 20-item scale.
They report acceptable fit, but the results are difficult to interpret as no
likelihood ratio statistics were reported. Based on the factor structure of
the GBQ, we propose our first hypothesis:

2. Methodology
2.1. Participants
A sample was recruited using Amazon Mechanical Turk, an online
web-based platform for human executed tasks. Ethics clearance was
granted by [REDACTED] Human Research Ethics Committee.
Participants were restricted to those with an MTurk approval rating of
at least 95%, consistent with practices adopted in previous research
(Goodman, Cryder, & Cheema, 2013). Participation was restricted to
English speaking North Americans of the legal gambling age (21 years
of age or older) that had lived in or visited the jurisdictions that contain
SBGMs (Nevada, New Jersey, Connecticut, and California) in the past
12 months in order to recruit participants who may have had experience with this gambling activity.
Sample size is an important feature of this study. A sufficiently large
sample size is a necessary condition to reject some null hypotheses in
support validation arguments. Conversely, some tests of model fit are
sensitive to sample size and will produce Type I errors if samples are too
large. For example, the likelihood ratio test is noted to almost always be
statistically significant if sample sizes are larger than 400 observations
(Kenny, 2012; Satorra & Saris, 1985). To inform sample size selection,
we use the Computing power and minimum sample size for RMSEA tool
(Preacher & Coffman, 2006) to calculate model power assuming an
alpha of 0.05 and desired power of 0.8. We use the root mean squared
error of association (RMSEA) from Winfree et al. (2015) as our null
RMSEA and MacCallum, Browne, and Sugawara's (1996) value of 0.05
as ‘good’ fit as our alternative RMSEA. Based on those figures, the tool
recommends a minimum sample of 169 observations. A total of 232
respondents were recruited; 47 respondents were removed from analysis due to failing at least one of two attention checks, and one was
removed for completing the survey in an unfeasibly short time period.
There was no missing data as all questions required a response. In total,
responses from 184 individuals were used in this study.
Respondents were disproportionally male (68%); all had a high
school diploma or equivalent, and 63.59% had a bachelors degree or
higher; and most were employed full-time (78.26%), with a small
number of part-time/casually employed (9.24%), unemployed (4.89%),
student (3.26%), retired (1.63%), or other employment status (2.72%)
respondents. They were diverse in their reported gambling problems:
non-problem (45.11%), low-risk (26.09%), moderate-risk (7.07%), and
problem (21.74%). In Table 1, we summarize respondent characteristics and their assessment of relative skill in games related to this
study.

H1. A two-factor structure described by the GBQ-IoC and the GBQ-LP
fits gambler responses to the 20-item questionnaire.
Our subsequent hypotheses are based on the theoretical and empirical evidence of cognitive distortions in gambling and the presence of
illusion of control and gambler's fallacy questions on the GBQ. To
provide a point of comparison, the study included a focus on a new
form of gaming machine, skill-based gaming machines (SBGMs). In
contrast to slot machines, whose outcomes are completely determined
by chance, SBGMs are games of mixed skill and chance that incorporate
an element of skill into traditional EGM mechanics. Currently only
available in a few gambling jurisdictions within the U.S., SBGMs allow
players to increase their chances of winning or the size of the payout
depending on their performance in the game (Fisher, 2016). We derive
the following hypotheses through a deductive process to assess the
criterion validity of the GBQ-IoC and the GBQ-LP:
H2a. The GBQ-IoC is positively related to the difference in perceived
skill of games of only chance versus the perceived skill of games of only
skill.
H2b. The GBQ-LP is unrelated to the difference in perceived skill of
games of only chance versus the perceived skill of games of only skill,
after controlling for the GBQ-IoC.
Explanation: Individuals who view games of only chance as closer in
perceived skill to games of only skill will have higher estimated levels of
illusion of control, relative to other respondents. After controlling for
shared variance, there should be no relationship with measures of
gambler's fallacy since attributions to repeated play would be removed
by differencing perceptions.
H3a. The GBQ-IoC is positively related to the difference in perceived
skill of games of only chance versus the average perceived skill of
games of mixed skill and chance.
H3b. The GBQ-LP is unrelated to the difference in perceived skill of
games of only chance versus the average perceived skill of games of
mixed skill and chance, after controlling for the GBQ-IoC.
Explanation: Individuals who view games of only chance as closer in
perceived skill to games of mixed skill and chance, will have higher
estimated levels of illusion of control, relative to other respondents.
After controlling for shared variance, there should be no relationship
with measures of gambler's fallacy since attributions to repeated play
would be removed by differencing perceptions of the two games.

Table 1
Respondent summary statistics.

H4a. The GBQ-LP is positively related to the difference in perceived
likelihood of winning on a game determined by chance over relatively
long period of time versus a relatively short period of time.
H4b. The GBQ-IoC is unrelated to the difference in perceived likelihood
of winning on a game determined by chance over relatively long period
of time versus a relatively short period of time, after controlling for the
GBQ-LP.
Explanation: In a game of chance with negative expected value, the
probability of winning will fall with more wagers over time. Individuals
who view skilled players as more likely to win money over a long period
of time than a short period of time will therefore have higher levels of
gambler's fallacy related cognitive distortions. After controlling for
shared variance, there should be no difference in illusion of control, as
the comparison is within the same game.

Variable

Count

Mean

Std. dev.

Min

Max

Age
Slot Play Frequency
SBGM Play Frequency
Perceived Skill in Chess
Perceived Skill in SBGMs
Perceived Skill in Slots
PGSI Score
GBQ
GBQ-IoC
GBQ-LP

34.02
2.43
2.04
80.23
48.52
10.07
4.29
71.10
31.73
39.37

9.29
1.33
1.51
30.67
27.28
14.65
6.34
29.48
11.59
19.08

Note. Frequency variables are labeled responses from ‘not at all’ to ‘daily’ regarding, “In the past 12 months, how often have you typically gambled on
[Game]?”

K.S. Philander, et al.

2.2. Design

We estimate a series of ordinary least squares (OLS) models that
regress ∆Slot−Game onto GBQ-LP, GBQ-IoC, and other controls. We hypothesize respondents with higher illusions of control will view slots as
closer-to, or potentially greater than, SBGMs in skill. We hypothesize no
relationship between ∆Slot−Game and GBQ-LP, after controlling for GBQIoC, since there is no time dimension in the comparison to implicate the
gambler's fallacy. This set of models tests the criterion validity of the
GBQ-IoC and the discriminant validity of the GBQ-LP by assessing their
explanatory power against two activities that can objectively be view as
high in chance (slots machines) and mixed-skill and chance (SBGMs).
To test H4a and H4b, we ask respondents about their perceptions of
SBGM outcomes over relatively short (one hour) and long (50 h) periods of time. After being shown the SBGM demo video, respondents
were asked to rate the extent to which they agreed that, “A player of
greater skill is more likely to win money on the skilled game gambling
machine over [1 or 50] hours” on a five-point Likert scale from
‘strongly disagree’ (1) to ‘strongly agree’ (5). We compute differences
reported scores over one-hour (m = 4.19, sd = 0.88) and fifty hours
(m = 4.28, sd = 0.85), and from mean responses. Formally,

Through an online survey supplemented with media, respondents
were shown videos of electronic gaming machines (EGMs) to facilitate a
baseline understanding of slot machines and SBGMs. They were then
asked a series of questions about perceived skill and chance in games
(both gambling and non-gambling).
Respondents were asked to simultaneously order eight different
gambling and non-gambling games (e.g. chess) along a 100-point scale
from all chance (0) to all skill (100) (Perceived Skill). They were also
asked questions about the extent to which they agreed that players of
greater skill would be more likely to win money playing specific gambling games. Respondents were then asked several demographic questions, gambling frequency questions, and were administered the GBQ
and the Problem Gambler Severity Index (PGSI, Ferris & Wynne, 2001).
We use classification categories from Currie, Hodgins, and Casey
(2013). Two attention check items were distributed in separate sections
of the survey to identify non-conscientious or random responders (e.g.,
“please choose ‘somewhat disagree’ as your response to this question”)
(Marjanovic, Struthers, Cribbie, & Greenglass, 2014).

1h 50h = (x slot1h
i

2.3. Analysis

x islot )

(x ichess

x ichess )

(1)

where, xij refers to the score given by respondent ‘i’ to activity ‘j’ and x i j

refers to the sample mean of that activity. Intuitively, we measure
whether respondents view chess and slots as close (smaller values) or
far (larger values) in relative skill. Larger values of ∆Slot−Chess are interpreted as higher levels of cognitive distortions.2 We then estimate a
set of ordinary least squares (OLS) models that successively regress
∆Slot−Chess onto GBQ-IoC, GBQ-LP, and other controls. This set of models
tests the criterion validity of the GBQ-IoC and the discriminant validity
of the GBQ-LP by assessing their explanatory power against two activities that can objectively be view as high in chance (slots machines)
and high in skill (chess).
To test H3a and H3b, we exploit similarities and differences in two
forms of EGMs, slot machines and SBGMs. After respondents were
shown sample demo videos of a representative slot machine and a representative SBGM, which depicted the user experience of playing the
games, they were asked questions on perceived skill and chance for the
respective game shown. The SBGMs game has an actual element of skill
that is shown in the video, while the slot machine has no element of
skill and is not described with any skill element. The order of viewing
for the videos was randomized for respondents. On a five-point Likert
scale ranging from ‘strongly disagree’ (1) to ‘strongly agree’ (5), respondents are asked whether, “A player of greater skill is more likely to
win money on the [slot machine | skilled game gambling machines]
over one hour”.
We compute differences in responses regarding slot machines
(m = 2.24, sd = 1.33) and responses regarding SBGMs (m = 4.19,
sd = 0.88), from mean responses and each other. Formally,
Slot SBGM = (x slot
i

xislot )

(x igame

x igame )

(x islot 50h

x islot50h)

(3)

We estimate a series of ordinary least squares (OLS) models that
regress ∆1h−50h onto GBQ-LP, GBQ-IoC, and other controls. We hypothesize respondents with higher gambler fallacies to view longer
periods of play as involving more skill, which would be indicated by a
positive and significant coefficient on GBQ-LP. We hypothesize no relationship between ∆50h/1h and GBQ-IoC in the second stage model,
after controlling for GBQ-LP. Differencing of one-hour and fifty-hour
responses should remove effects of illusion of control, as they are
questions relating to the same game, leaving only differences in the
related time dimension, which should reveal cognitive distortions
around luck and the role of persistence.

To test the H1, we estimate a confirmatory factor analysis (CFA)
model of the GBQ factors (Steenbergh et al., 2002) using Stata/MP
15.1. We assess model fit using a likelihood ratio test, the RMSEA the
comparative fit index, and the Tucker-Lewis Index.
To test H2a and H2b, we compute the difference in ratings for chess
and slot machines using the 100-point chance/skill rating scale, in
comparison to mean responses:
Slot Chess = (x slot
i

x islot1h )

3. Results
Fig. 1 illustrates the CFA model. We find the GBQ factors do not fit
this data well. From our likelihood ratio test, we reject the assumption
that the model fits as well as the saturated model, χ2(169) = 518.14,
p < 0.001. Our measure of population error fails to reach cutoff values
for mediocre fit, RMSEA = 0.11, > 0.08 (MacCallum et al., 1996). The
lower bound rejects the hypothesis that the fit is close 90% CI, lower
bound = 0.10, > 0.05, and the upper bound fails to reject the assumption of poor fit, 90% CI, upper bound = 0.12, > 0.10 (Browne &
Cudeck, 1992; StataCorp, 2017). The baseline fit indices are not close to
the desired value of 1.00 (Bentler, 1990; StataCorp, 2017), Comparative
Fit Index = 0.89, Tucker-Lewis Index = 0.88.
Our constructed tests show better validity. As shown in Table 2, we
find evidence supporting H2a. Perceived differences in chess and slot
skill are found to be predicted by GBQ-IoC scores, even after controlling
for gambling problems, play frequency, and demographics (model 3).
We also find PGSI group membership to be related to the difference
score. However, we fail to find support for H2b, as the GBQ-LP is noted
to have a statistically significant effect in all models (4–6).
Our tests of H3a and H3b are similar. As shown in Table 3, we find
evidence supporting H3a but not H3b. Differences in perceived slot and
SBGM skill are differentiated by relative scores on the GBQ-IoC, but
GBQ-LP are also significant when the GBQ-IoC and other controls are
included in the model. Notably, the GBQ-IoC is not significant in those
models (4–6).
In our tests of H4a and H4b, we find evidence in support of both
hypotheses. After controlling for gambling problems and play frequency
(models 2 and 3), we find evidence that GBQ-LP is related to perceived
differences in outcomes on SBGM over one hour versus 50 h. In models
4–6, we fail to find evidence that GBQ-IoC is related to the same outcome measure when we control for GBQ-LP, supporting H4b. We also

(2)

While the mean differencing in Eq. (1) only impacts the constant term in the
regression models, we use it in our design as it allows for a more intuitively
understandable formulation.

K.S. Philander, et al.

Fig. 1. Confirmatory factor analysis model results (n = 184).

find PGSI group membership to be related to the difference score
(Table 4).

it was hypothesized the opposite may be the case. It is worth considering whether the perceived latent constructs represented by the two
subscales are meaningfully independent, or are related to a wider
cognitive distortion attribute.3 Ejova, Delfabbro, and Navarro (2015)
provide some evidence that there is an underlying process or belief
structure that is connected to vulnerability of specific gambling-related
beliefs, which would help explain these results. Relatedly, in
Steenbergh et al. (2002) and Winfree et al. (2013), multiple items crossloaded onto both factors at the ‘rule-of-thumb’ value of 0.3 or more
(Hair et al., 2010), suggesting that aspects of the subscales may be
conflated. Again, this could lead in part to the discrimination challenges
we observed in our some of tests.
Based on our study and others, there is some evidence that there is
an opportunity to improve the parsimony of GBQ questions and the
discriminant ability of its related subscales. The GBQ generally performed poorly across a range of test statistics in our CFA, despite our
efforts to identify an appropriate sample size of regular gamblers. This
is similar to the results from Winfree et al. (2015), which found poor
test statistics in a CFA of the GBQ in a clinical sample of respondents. A
simplification of the model by the removal of some items and

4. Discussion
As a feature of both the normative and disordered gambling experience, cognitive distortions are an important phenomenon to understand motivations, behaviors, and appropriate treatment protocols.
In this study, we designed three validation tests and used one statisticalbased approach to evaluate one of the most widely used measures of
gambling-related cognitive distortions, the GBQ. Overall, our results
suggest that the GBQ items have validity in predicting distorted
thoughts around gambling activities related to illusion of control and
the gambler's fallacy. This relationship is generally robust, showing a
measurable effect with our measures even after controlling for frequency of play, sex, age, and education levels. It further demonstrates
to be non-redundant to measures of gambling problems, as the both
subscales show a measurable relationship with our dependent variables
when we control for PGSI categories.
The absence of evidence supporting H2b and H3b showed similar
patterns and may be related to the same phenomenon. In both cases,
GBQ-LP effects were significant and GBQ-IoC were generally not, when

We thank an anonymous reviewer for this insight.

K.S. Philander, et al.

Table 2
OLS regressions of perceived skill difference in slots and Chess (DV : ∆Slot−Chess).
(1)
GBQ-IoC
GBQ-LP

(2)

(4)

0.46
(0.17)

0.45
(0.18)

No
No
No
No
Yes
0.24
0.23

−1.69
(4.15)
6.12
(10.9)
68.8⁎⁎⁎
(12.9)
Yes
No
No
No
Yes
0.55
0.53

−2.59
(4.22)
7.01
(11.8)
66.1⁎⁎⁎
(12.9)
Yes
Yes
Yes
Yes
Yes
0.57
0.53

PGSI Low
PGSI Moderate
PGSI Problem
Slot Play Freq.
Sex
Age Categories
Education
Constant
N
R2
Adjusted R2

(3)

1.70
(0.21)

⁎⁎⁎

⁎⁎

⁎

−0.54
(0.27)
1.62⁎⁎⁎
(0.21)

⁎

No
No
No
No
Yes
0.41
0.40

(5)

(6)

−0.13
(0.24)
0.65⁎⁎
(0.23)
−3.99
(4.14)
2.72
(11.4)
60.5⁎⁎⁎
(13.7)
Yes
No
No
No
Yes
0.57
0.54

−0.11
(0.25)
0.61⁎⁎
(0.23)
−4.77
(4.20)
3.99
(12.2)
58.4⁎⁎⁎
(13.6)
Yes
Yes
Yes
Yes
Yes
0.59
0.54

Table 4
OLS regressions of perceived impact of skill on SBGM outcomes over time
(DV : ∆1h−50h).
(1)

GBQ-IoC
GBQ-LP

of

perceived

difference

in

slots

and

GBQ-LP
PGSI Low

(2)

(3)

(4)

(5)

(6)

0.08⁎⁎⁎
(0.01)

0.05⁎⁎⁎
(0.01)

0.06⁎⁎⁎
(0.01)

No
No
No
No
No
Yes
0.34
0.34

−0.08
(0.26)
0.30
(0.40)
0.37
(0.40)
Yes
Yes
No
No
No
Yes
0.44
0.40

−0.14⁎⁎
(0.26)
0.32
(0.40)⁎
0.55
(0.43)
Yes
Yes
Yes
Yes
Yes
Yes
0.47
0.40

0.01
(0.01)
0.05⁎⁎⁎
(0.01)

0.01
(0.01)
0.05⁎⁎⁎
(0.01)
−0.27
(0.25)
0.04
(0.40)
−0.30
(0.49)
Yes
Yes
No
No
No
Yes
0.49
0.44

0.02
(0.02)
0.04⁎⁎⁎
(0.01)
−0.33
(0.26)
0.06
(0.41)
−0.13
(0.50)
Yes
Yes
Yes
Yes
Yes
Yes
0.51
0.44

PGSI Moderate
PGSI Problem

No
No
No
No
No
Yes
0.45
0.45

0.004
(0.003)

PGSI Moderate
PGSI Problem
SBGM Play Freq.
Sex
Age Categories
Education
Constant
N
R2
Adjusted R2

No
No
No
No
Yes
0.01
0.01

0.009⁎
(0.004)
−0.10
(0.11)
−0.37⁎
(0.16)
−0.82⁎⁎
(0.25)
Yes
No
No
No
Yes
0.11
0.06

0.010⁎⁎
(0.004)
−0.12
(0.11)
−0.37⁎
(0.16)
−0.72⁎⁎
(0.27)
Yes
Yes
Yes
Yes
Yes
0.17
0.09

(4)

(5)

(6)

−0.002
(0.006)
0.005
(0.005)

−0.006
(0.006)
0.012⁎
(0.005)
−0.100
(0.11)
−0.38⁎
(0.17)
−0.84⁎⁎⁎
(0.25)
Yes
No
No
No
Yes
0.11
0.06

−0.001
(0.006)
0.011⁎
(0.005)
−0.12
(0.11)
−0.37⁎
(0.16)
−0.73⁎⁎
(0.27)
Yes
Yes
Yes
Yes
Yes
0.17
0.08

No
No
No
No
Yes
0.01
0.00

Note. PGSI Low are scores from 1–4; PGSI Moderate are scores from 5 to 7; and
PGSI Problem are scores from 8+. Slot Play Freq. is slot play frequency. Age
categories are ten-year bands beginning at ‘20’. Heteroskedasticity robust
standard errors in parentheses.
⁎
p < 0.05.
⁎⁎
p < 0.01.
⁎⁎⁎
p < 0.001.

4.1. Limitations & future research

SBGMs

(1)

PGSI Low

Slot Play Freq.
SBGM Play Freq.
Sex
Age Categories
Education
Constant
N
R2
Adjusted R2

skill

(3)

GBQ-IoC

Note. PGSI Low are scores from 1–4; PGSI Moderate are scores from 5 to 7; and
PGSI Problem are scores from 8+. Slot Play Freq. is slot play frequency. Age
categories are ten-year bands beginning at ‘20’. Heteroskedasticity robust
standard errors in parentheses.
⁎
p < 0.05.
⁎⁎
p < 0.01.
⁎⁎⁎
p < 0.001.
Table 3
OLS regressions
(DV : ∆Slot−SBGM).

(2)

Our study design provides validation where intended, however, it
bears emphasizing that there is no definitive clinical or external objective measure by which to evaluate our tests, the GBQ, or gamblingrelated cognitive distortions more generally. As such, the extent to
which our tests can be viewed as useful is dependent on the validity of
the tests themselves. Despite our attempts to build those measures deductively rather than subjectively, the absence of the intended results
could be a function of a poorly designed test, as opposed to an invalid
scale. While the CFA in this study is a more objective measure of construct validity, it also provides less direction in terms of future adaptations of the scale. Also, this study used an internet-based sample from
MTurk. Although we followed research best practices, the typical cautions apply.
Future research related to the GBQ or other measures of cognitive
distortions should generally be focused in two areas. First, new external
tests should be developed to assess the validity of the scales and their
component items. Because there is no objective measure of gamblingrelated cognitive distortions, a rigorous approach to validating the
measure is warranted, which should include multiple external validity
tests and samples. These could be done using behavioral measures,
rather than self-reported ratings. Second, both statistical and experimental methods should be used to refine the GBQ to improve question
parsimony and the uniqueness of the subscales. As originally observed
by Steenbergh et al. (2002) the GBQ-IoC and the GBQ-LP are closely
correlated, with items cross-loaded on both subscales.

Note. PGSI Low are scores from 1–4; PGSI Moderate are scores from 5 to 7; and
PGSI Problem are scores from 8+. Slot Play Freq. is slot play frequency. Age
categories are ten-year bands beginning at ‘20’. Heteroskedasticity robust
standard errors in parentheses.
⁎
p < 0.05.
⁎⁎
p < 0.01.
⁎⁎⁎
p < 0.001.

Declarations of interest

reexamination of the factor structure may actually improve related
performance, in addition to the benefit of reducing questionnaire
length. Reliability statistics from past studies further support the notion
that the GBQ could be made shorter. For example, Winfree et al. (2013)
and Mattson et al. (2008) both found reliability values above α > 0.9
for the GBQ. Values that high suggests there may be redundancies in the
scales, and that fewer items could be used (Tavakol & Dennick, 2011).

None
Role of funding sources
Funding for this study was provided by an Australian Research
Council Discovery Early Career Research Award [DE1060100459]
awarded to Dr. Sally Gainsbury. The ARC had no role in the study design, collection, analysis or interpretation of the data, writing the

K.S. Philander, et al.

manuscript, or the decision to submit the paper for publication.

of Behavioral Addictions. https://doi.org/10.1556/2006.4.2015.045.
Ladouceur, R., & Mayrand, M. (1987). The level of involvement and the timing of betting
in roulette. Journal of Psychology: Interdisciplinary and Applied.. https://doi.org/10.
1080/00223980.1987.9712654.
Langer, E. J. (1975). The illusion of control. Journal of Personality and Social Psychology..
https://doi.org/10.1037//0022-3514.32.2.311.
Langer, E. J. (1977). The psychology of chance. Journal for the Theory of Social Behaviour.
https://doi.org/10.1111/j.1468-5914.1977.tb00384.x.
MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods..
https://doi.org/10.1037/1082-989X.1.2.130.
Marchetti, D., Whelan, J. P., Verrocchio, M. C., Ginley, M. K., Fulcheri, M., Relyea, G. E.,
& Meyers, A. W. (2016). Psychometric evaluation of the Italian translation of the
Gamblers' Beliefs Questionnaire. International Gambling Studies, 16(1), 17–30.
Marjanovic, Z., Struthers, C. W., Cribbie, R., & Greenglass, E. R. (2014). The conscientious
responders scale: A new tool for discriminating between conscientious and random
responders. SAGE Open. https://doi.org/10.1177/2158244014545964.
Mattson, R. E., MacKillop, J., Castelda, B. A., Anderson, E. J., & Donovick, P. J. (2008).
The factor structure of gambling-related cognitions in an undergraduate university
sample. Journal of Psychopathology and Behavioral Assessment. https://doi.org/10.
1007/s10862-007-9063-z.
Oei, T. P. S., & Gordon, L. M. (2008). Psychosocial factors related to gambling abstinence
and relapse in members of gamblers anonymous. Journal of Gambling Studies.. https://
doi.org/10.1007/s10899-007-9071-7.
Pilatti, A., Cupani, M., Tuzinkievich, F., & Winfree, W. (2016). Confirmatory factor
analysis of the Spanish version of the Gamblers' Beliefs Questionnaire in a sample of
Argentinean gamblers. Addictive Behaviors Reports.. https://doi.org/10.1016/j.abrep.
2016.09.001.
Preacher, K. J., & Coffman, D. L. (2006). Computing power and minimum sample size for
RMSEA.
Raylu, N., & Oei, T. P. S. (2004). The Gambling Related Cognitions Scale (GRCS):
Development, confirmatory factor validation and psychometric properties. Addiction..
https://doi.org/10.1111/j.1360-0443.2004.00753.x.
Satorra, A., & Saris, W. E. (1985). Power of the likelihood ratio test in covariance
structure analysis. Psychometrika.. https://doi.org/10.1007/BF02294150.
Sharpe, L., & Tarrier, N. (1993). Towards a cognitive-behavioural theory of problem
gambling. British Journal of Psychiatry.. https://doi.org/10.1192/bjp.162.3.407.
Smith, D., Woodman, R., Drummond, A., & Battersby, M. (2016). Exploring the measurement structure of the Gambling Related Cognitions Scale (GRCS) in treatmentseekers: A Bayesian structural equation modelling approach. Psychiatry Research.
https://doi.org/10.1016/j.psychres.2016.02.002.
StataCorp (2017). Stata structural equation modeling Reference Manual Release 15. College
Station, Texas: Stata Press.
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(2), 143–149. https://doi.org/10.1037/0893-164X.16.2.143.
Sundali, J., & Croson, R. (2006). Biases in casino betting : The hot hand and the gambler ’
s fallacy. Judgment and Decision Making. retrieved from: http://journal.sjdm.org/
jdm06001.pdf.
Tavakol, M., & Dennick, R. (2011). Making sense of Cronbach's alpha. International
Journal of Medical Education.. https://doi.org/10.5116/ijme.4dfb.8dfd.
Toneatto, T., Blitz-Miller, T., Calderwood, K., Dragonetti, R., & Tsanos, A. (1997).
Cognitive distortions in heavy gambling. Journal of Gambling Studies.. https://doi.
org/10.1023/A:1024983300428.
Walker, M. B. (1992). The psychology of gambling. Elmsford, NY, USA: Pergamon Press.
Winfree, W. R., Ginley, M. K., Whelan, J. P., & Meyers, A. W. (2015). Psychometric
evaluation of the Gamblers' Beliefs Questionnaire with treatment-seeking disordered
gamblers. Addictive Behaviors, 43, 97–102. https://doi.org/10.1016/j.addbeh.2014.
12.016.
Winfree, W. R., Meyers, A. W., & Whelan, J. P. (2013). Validation of a Spanish translation
of the Gamblers' Beliefs Questionnaire. Psychology of Addictive Behaviors. https://doi.
org/10.1037/a0030824.
Wong, S. S. K., & Tsang, S. K. M. (2012). Validation of the Chinese version of the
Gamblers' Belief Questionnaire (GBQ-C). Journal of Gambling Studies.. https://doi.
org/10.1007/s10899-011-9286-5.
Xian, H., Shah, K. R., Phillips, S. M., Scherrer, J. F., Volberg, R., & Eisen, S. A. (2008).
Association of cognitive distortions with problem and pathological gambling in adult
male twins. Psychiatry Research. https://doi.org/10.1016/j.psychres.2007.08.007.
Xu, J., & Harvey, N. (2014). Carry on winning: The gamblers' fallacy creates hot hand
effects in online gambling. Cognition.. https://doi.org/10.1016/j.cognition.2014.01.
002.

Contributors
Authors SG and KP designed the study and wrote the protocol,
Author GG conducted literature searches. Author KP conducted the
statistical analysis and wrote the first draft of the manuscript. All authors contributed to and have approved the final manuscript.
Conflicts of interests
All authors declare that they have no conflicts of interest.
References
Barrault, S., & Varescon, I. (2013). Cognitive distortions, anxiety, and depression among
regular and pathological gambling online poker players. Cyberpsychology, Behavior
and Social Networking. https://doi.org/10.1089/cyber.2012.0150.
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin,
107(2), 238–246. https://doi.org/10.1037/0033-2909.107.2.238.
Blaszczynski, A., & Nower, L. (2002). A pathways model of problem and pathological
gambling. Addiction.. https://doi.org/10.1046/j.1360-0443.2002.00015.x.
Browne, M. W., & Cudeck, R. (1992). Alternative ways of assessing model fit. Sociological
Methods & Research, 21(2), 230–258. https://doi.org/10.1177/
0049124192021002005.
Clotfelter, C. T., & Cook, P. J. (1993). Notes: The “Gambler's Fallacy” in Lottery Play.
Management Science.. https://doi.org/10.1287/mnsc.39.12.1521.
Currie, S. R., Hodgins, D. C., & Casey, D. M. (2013). Validity of the problem gambling
severity index interpretive categories. Journal of Gambling Studies.. https://doi.org/
10.1007/s10899-012-9300-6.
Davis, D., Sundahl, I., & Lesbo, M. (2000). Illusory personal control as a determinant of
bet size and type in casino craps games. Journal of Applied Social Psychology.. https://
doi.org/10.1111/j.1559-1816.2000.tb02518.x.
Dunn, D. S., & Wilson, T. D. (1990). When the stakes are high: A limit to the illusion-ofcontrol effect. Social Cognition.. https://doi.org/10.1521/soco.1990.8.3.305.
Ejova, A., Delfabbro, P. H., & Navarro, D. J. (2015). Erroneous gambling-related beliefs as
illusions of primary and secondary control: A confirmatory factor analysis. Journal of
Gambling Studies.. https://doi.org/10.1007/s10899-013-9402-9.
Ferris, J., & Wynne, H. (2001). The Canadian problem gambling index : User manual.
Ottawa, ON: Canadian Centre on Substance Abusehttps://doi.org/10.1007/s10899010-9224-y.
Fisher, C. J. (2016). The rise of skill-based gaming regulation. American Gaming Lawyer,
28–31 Autumn.
Fortune, E. E., & Goodie, A. S. (2012). Cognitive distortions as a component and treatment
focus of pathological gambling: A review. Psychology of Addictive Behaviors. https://
doi.org/10.1037/a0026422.
Gainsbury, S. M., & Blaszczynski, A. (2017). Virtual reality gambling: Implications for
regulation and gambling disorders. Gaming Law Review, 21(4), 314–322.
Goodie, A. S., & Fortune, E. E. (2013). Measuring cognitive distortions in pathological
gambling: Review and meta-analyses. Psychology of Addictive Behaviors, 27(3),
730–743. https://doi.org/10.1037/a0031892.
Goodman, J. K., Cryder, C. E., & Cheema, A. (2013). Data collection in a flat world: The
strengths and weaknesses of mechanical Turk samples. Journal of Behavioral Decision
Making.. https://doi.org/10.1002/bdm.1753.
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2010). Multivariate
Data Analysis. Prentice Hall816. https://doi.org/10.1016/j.ijpharm.2011.02.019.
Hodgins, D. C., Stea, J. N., & Grant, J. E. (2011). Gambling disorders. The Lancet,
378(9806), 1874–1884. https://doi.org/10.1016/S0140-6736(10)62185-X.
Holtgraves, T. (2009). Evaluating the problem gambling severity index. Journal of
Gambling Studies.. https://doi.org/10.1007/s10899-008-9107-7.
Jacobsen, L. H., Knudsen, A. K., Krogh, E., Pallesen, S., & Molde, H. (2007). An overview
of cognitive mechanisms in pathological gambling. Nordic Psychology.. https://doi.
org/10.1027/1901-2276.59.4.347.
Kenny, D.a. (2012). Measuring model fit. Davidakenny.Nethttps://doi.org/10.1186/17485908-6-64.
King, D. L., Gainsbury, S. M., Delfabbro, P. H., Hing, N., & Abarbanel, B. L. L. (2015).
Distinguishing between gaming and gambling activities in addiction research. Journal