P1: KEF

Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004 10:55 Style file version June 25th, 2002

Journal of Psychopathology and Behavioral Assessment, Vol. 26, No. 2, June 2004 (°
C 2004)

Dimensionality of the Obsessive Beliefs Questionnaire (OBQ)

Carol M. Woods,1,4 David F. Tolin,2 and Jonathan S. Abramowitz3

Accepted October 25, 2003

The Obsessive Beliefs Questionnaire (OBQ) is a self-report measure in development by the Obsessive
Compulsive Cognitions Working Group (OCCWG), who studies cognitive aspects of obsessive–
compulsive disorder (OCD). The aim of this paper was to evaluate the dimensionality of the OBQ
in students (n = 995). Two models were tested in a confirmatory framework, corresponding to the
OCCWG’s (a) original six subscales (87 items), and (b) more recent three subscales (44 items). Both
models fit the present data poorly; thus, an exploratory analysis was undertaken. Results revealed one
large factor that is relevant, but not unique, to OCD, and three factors that are conceptually consistent
with the OCCWG’s recent three: (a) distorted beliefs about one’s own thoughts, (b) perfectionism,
and (c) inflated responsibility. Convergent and divergent validity analyses of the present OBQ factors
generally supported the preceding interpretations. Readers are referred to the work of the OCCWG
for revisions to the OBQ.

KEY WORDS: obsessive–compulsive; OCD; OBQ; OCCWG; factor-analysis.

INTRODUCTION However, such strategies paradoxically serve to increase

Contemporary cognitive conceptualizations of ob-
sessional problems posit that obsessions develop from,
and are maintained by, overestimates of the significance
of unwanted upsetting thoughts (e.g., Salkovskis, 1985).
Specifically, this model starts with the idea that most
people experience senseless, intrusive, upsetting thoughts
(e.g., violent, blasphemous) from time to time. Thus, the
difference between normal intrusive cognitions and clini-
cal obsessions lies not in the occurrence or content of the
thought itself, but in the person’s appraisal of the thought’s
significance. Accordingly, whereas most people recognize
the senselessness of such intrusions, an obsessional pattern
occurs if such thoughts are misappraised as highly threat-
ening. Threat appraisals are presumed to lead to anxiety
or distress, as well as to the desire to control or neutralize
the upsetting thought via behavioral or mental ritualizing.
preoccupation with intrusive ideas and prevent the person
from learning that such thoughts are not actually indicative
of harm (for a review see Salkovskis, 1999).
The increasing interest in cognitive models of
obsessive–compulsive disorder (OCD) has led to the de-
velopment of measures to assess cognitive phenomena
thought to underlie this disorder. The Obsessive Beliefs
Questionnaire (OBQ) is one such measure which is cur-
rently in development by the Obsessive Compulsive Cog-
nitions Working Group (OCCWG, 1997, 2001), an inter-
national group of researchers who study cognitive aspects
of OCD. The OBQ was originally designed to measure six
specific cognitive domains identified by the OCCWG and
believed to be particularly important in the development
and maintenance of obsessional problems. These cogni-
tive domains include:

1 Department of Psychology, University of North Carolina, Chapel Hill, Inflated Responsibility

North Carolina.
2 The Institute of Living and University of Connecticut School of
Medicine, Hartford, Connecticut.
3 Mayo Clinic, Rochester, Minnesota.
4 To whom correspondence should be addressed at Department of Psy-
chology, Davie Hall, CB #3270, University of North Carolina, Chapel
Hill, North Carolina 27599; e-mail: woodscm@email.unc.edu.
Persons are thought to have an inflated sense of re-
sponsibility if they believe they have the power to bring
about or prevent negative outcomes perceived as crucial.
Such outcomes may have consequences in the real world
(e.g., occurrence of a car wreck) or exist at a moral level

113
0882-2689/04/0600-0113/0 °
C 2004 Plenum Publishing Corporation
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004
114
(e.g., conclusion that one is a bad person; OCCWG, 1997).
Individuals with OCD are thought to believe themselves to
be personally responsible for the content of their intrusive
thoughts as well as any possible disastrous consequences
that might arise from their thoughts (Foa, Sacks, Tolin,
Przeworski, & Amir, 2002; Salkovskis, 1985). They hold
themselves equally responsible for acts of omission (fail-
ing to prevent harm) as for acts of commission (causing
harm; Wroe & Salkovskis, 2000). Examples of OBQ items
in this domain include: “When I hear about a tragedy, I
can’t stop wondering if I am responsible in some way,”
“For me, not preventing harm is as bad as causing harm,”
and “I often believe I am responsible for things that other
people don’t think are my fault.”
Beliefs About the Importance of Thoughts
OCD sufferers report a belief that the mere pres-
ence of their thoughts makes those thoughts important
and meaningful. One variation of this belief is thought-
action fusion, in which thoughts are believed to influence
the external world (e.g., “I can make an accident hap-
pen by thinking about it”) or thinking about a behavior is
morally equivalent to the behavior itself (e.g., “Thinking
about committing adultery is as bad as actually doing it”;
Abramowitz, Whiteside, Lynam, & Klasy, 2003; Shafran,
Thordarson, & Rachman, 1996). Example OBQ items in-
clude: “The more I think of something horrible, the greater
the risk it will come true,” “Having an unwanted sexual
thought or image means I really want to do it,” and “Hav-
ing nasty thoughts means I am a terrible person.”
Beliefs About the Importance of Controlling
One’s Thoughts
Individuals with OCD often report the belief that
complete control over their thoughts is both necessary
and possible (Purdon & Clark, 1994; Salkovskis, 1985).
Self-report data indicate that OCD patients make fre-
quent attempts to suppress their thoughts (Abramowitz,
Whiteside, Kalsy, & Tolin, 2003). Typically, these indi-
viduals find actual thought control quite difficult (Tolin,
Abramowitz, Przeworski, & Foa, 2002; Tolin, Hamlin, &
Foa, 2002); the resulting thought control failures serve to
increase obsessional anxiety (Tolin, Abramowitz, Hamlin,
Foa, & Synodi, 2002). Example items are: “I would be a
better person if I gained more control over my thoughts,”
“Having control over my thoughts is a sign of good char-
acter,” and “To avoid disasters, I need to control all the
thoughts or images that pop into my mind.”
10:55 Style file version June 25th, 2002
Woods, Tolin, and Abramowitz
Overestimation of Threat
Although this is unlikely unique to individuals with
OCD, such persons may show inflated beliefs about the
probability and cost of aversive events (Carr, 1974; Foa &
Kozak, 1986; Salkovskis, 1985; Woods, Frost, & Steketee,
2002). For example, individuals with contamination fear
may exaggerate the probability of becoming infected, as
well as the severity of illness that will result. Example
items are: “Bad things are more likely to happen to me
than to other people,” “Even ordinary experiences in my
life are full of risk,” and “When anything goes wrong in
my life, it is likely to have terrible effects.”
Intolerance of Uncertainty
Intolerance of uncertainty describes the beliefs that
it is crucial to be certain about everything, that one has
poor capacity to cope with unpredictable change, and that
it is difficult to function in inherently ambiguous situa-
tions (OCCWG, 1997). Persons with OCD often report
difficulty tolerating uncertainty or ambiguity (Carr, 1974;
Tolin, Abramowitz, Brigidi, & Foa, 2003). As a result,
they may experience problems with aspects of decision-
making, such as information gathering and categorization
(Frost, Lahart, Dugas, & Sher, 1988). Example items in-
clude: “If I’m not absolutely sure of something, I’m bound
to make a mistake,” “I need the people around me to be-
have in a predictable way,” and “I often think I will be
overwhelmed by unforeseen events.”
Perfectionism
OCD is thought to be associated with an inability to
tolerate even minor mistakes or imperfection, and beliefs
that it is possible and necessary to find the mistake-free so-
lution to any problem (Frost & Steketee, 1997; OCCWG,
1997). The OBQ assesses perfectionism in relation to ex-
ternal stimuli, such as a need to fill out a form without
making a mistake. Examples items are: “I must work to
my full potential at all times,” “No matter what I do, it
won’t be good enough,” and “There is only one right way
to do things.”
In preliminary work on an 87-item version of the
OBQ, patients with OCD scored higher (more patholog-
ically) than did nonanxious control participants on all
domains. Although test–retest reliability was high, and
convergent validity promising, the discriminant validity
of the six scales was questionable: OBQ scales were as
strongly correlated with measures of anxiety, depression,
and worry as with OCD symptoms. Also, three of the
domains (intolerance of uncertainty, overestimation of
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581
Dimensionality of the OBQ
threat, perfectionism) did not discriminate between peo-
ple with OCD and anxious control participants, suggesting
that they are not unique to OCD (Taylor, Kyrios, Thor-
darson, Steketee, & Frost, 2002). Further detail about
the six domains and the development of the OBQ ap-
pear in Frost and Steketee’s edited book (Frost & Steke-
tee, 2002). In a recent survey of 562 student volunteers,5
several relationships between the rationally derived OBQ
subscales and OCD symptoms were observed, with so-
cial anxiety and depression statistically controlled (Tolin,
Woods, & Abramowitz, 2003). The Threat Overestima-
tion OBQ subscale significantly predicted washing, check-
ing, and performance of mental compulsions (mental neu-
tralizing), OBQ-Control of Thoughts predicted obsessing,
OBQ-Importance of Thoughts predicted neutralizing, and
OBQ-Perfectionism predicted the need to arrange objects
a certain way. These results suggest that, in a sample in-
dependent of data collected by the OCCWG, most of the
rationally derived scales were related to some type of OCD
symptom.
The OCCWG (2001) noted that correlations among
the rationally derived subscales suggested fairly substan-
tial overlap, and have recently reported factor analyses
and further refinement of the 87-item version (OCCWG,
2003). On the basis of exploratory principal axis factor
analyses (oblique rotation) with 410 OCD patients, they
suggest three factors, corresponding to: (a) inflated re-
sponsibility for harm (RH), (b) perfectionism and intoler-
ance of uncertainty (PC), and (c) importance of, and con-
trol over, thoughts (ICT). They also performed exploratory
factor analysis with data from 291 women students, and
concluded that the factor structure overlapped closely with
that obtained with patients. The purpose of this study was
to attempt to cross-validate the factor structure of the OBQ
for students, using a sample distinct from that in use by
the OCCWG.
The use of participants who have not been eval-
uated for OCD diagnosis, in research aimed at illumi-
nating OCD symptomatology, is controversial. There is
not definitive empirical evidence for, or against, the hy-
pothesis that OCD symptoms are analogous in student
and patient samples. Nevertheless, contemporary cogni-
tive models of OCD (e.g., Rachman, 1998; Salkovskis,
1999) posit that obsessions and compulsions develop from
largely normal cognitive and behavioral processes (e.g.,
mistaken beliefs and negative reinforcement). Thus, it is
theorized that OCD symptoms exist on a continuum from
normal intrusive thoughts and undisruptive rituals (e.g.,
5 Thissample is approximately half of the same participants from which
data are used for the present factor analyses. No factor analyses were
included in the Tolin et al. (2003) paper.
January 28, 2004 10:55 Style file version June 25th, 2002
115
superstitions) to extremely frightening obsessional pre-
occupations and time consuming compulsive behaviors
(e.g., Rachman & de Silva, 1978). The present analyses
were undertaken with a student sample, on the basis of the
premise that distinctions between diagnosable and nondi-
agnosable OCD symptoms are a matter of degree. Given
that other researchers share this view, the OBQ is likely to
be administered to students in the future, and it is impor-
tant to evaluate the dimensionality of the measure in this
population.
METHOD
Participants
One thousand four undergraduate students enrolled
in introductory psychology at the University of North
Carolina at Chapel Hill completed a battery of web-
administered questionnaires in exchange for course credit.
Missing data were minimized, but not completely avoided,
by a JavaScript function6 included in the study website
that alerted participants to skipped items. However, a few
missing OBQ item responses led to the casewise dele-
tion of 0.9% of participants (9 people), resulting in a fi-
nal sample size of 995 for the factor analyses. Approxi-
mately 75% of these participants were women (n = 746;
men: n = 249) and their average age was 19.01 years
(SD = 1.99). Participants identified themselves ethnically
as White/Caucasian (n = 751), Black/African American
(n = 137), Asian or Pacific Islander (n = 53), Hispanic
(n = 14), Native American or Alaskan Native (n = 6),
or multiethnic (n = 28; 6 people did not specify their
ethnicity).
Data Analytic Approach
Items on all questionnaires used here (described sub-
sequently) are answered on a Likert-type scale. For ex-
ample, respondents to the OBQ choose an integer be-
tween 1 (disagree very much) and 7 (agree very much).
Accordingly, all factor analyses reported in this paper
were performed using methods designed for categorical
data. The common factor model was fitted to polychoric
correlations among the items, using an estimation system
called WLSMV, implemented in the Mplus program (ver-
sion 2.01; Muthén & Muthén, 2001). WLSMV stands for
weighted least squares estimation with mean and variance
6 The authors thank Johnny Loney for writing the JavaScript function.
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004 10:55 Style file version June 25th, 2002

116 Woods, Tolin, and Abramowitz

adjusted standard errors and chi square statistic. It differs Table I. Descriptive Statistics for All Scales
from usual weighted least squares (WLS) estimation, in
Scale M SD Min Max
part, because the weight matrix includes only variances
(not covariances) among the polychoric correlations. An OBQ
advantage of using only variances in the weight matrix General 65.48 29.04 32 177
is that the procedure converges more quickly than WLS. Thoughts 25.66 10.91 10 62
Perfectionism 43.13 16.10 14 96
Readers are referred to the sizeable methodological lit-
Responsibility 35.35 12.36 10 68
erature for more detail about WLS and related proce- OCI-R
dures (e.g., Christoffersson, 1975; Flora, 2002; Muthén, Washing 1.37 2.13 0 12
1978, 1984; Muthén, du Toit, & Spisic, 1997; Muthén & Checking 1.93 2.20 0 12
Satorra, 1995; Satorra, 1992). Estimated communalities Ordering 3.22 2.93 0 12
Obsessing 1.73 2.09 0 11
(not given in version 2.01 of Mplus) were computed from
Hoarding 3.26 2.44 0 12
rotated loadings and interfactor correlations using the SAS Mental neutralizing 1.25 2.01 0 12
program. CES-D (depressed affect scale) 3.77 3.46 0 15
Four indices were considered to assess fit of the con- SAS (items 1–4) 5.94 1.96 4 16
firmatory models: (a) the incremental fit index of Tucker FNE-S 6.90 5.04 0 17
and Lewis (1973), TLI, (b) the comparative fit index, CFI
Note. SD = standard deviation; OCI-R = Obsessive–Compulsive
(Bentler, 1990), (c) the root mean square error of ap- Inventory-Revised; CES-D = Center for Epidemiological Studies De-
proximation, RMSEA (Steiger & Lind, 1980, as cited pression Scale; SAS = Self-Rating Anxiety Scale; and FNE-S =
in Hu & Bentler, 1998), and (d) the standardized root straightforward-worded items on the Fear of Negative Evaluation Scale.
mean square residual (SRMR; Bentler, 1995; Jöreskog
& Sörbom, 1981). Better fit is indicated as TLI and CFI
approach 1, and as RMSEA and SRMR approach 0. A (d) obsessing, (e) hoarding, and (f) mental neutralizing.7
rough cut-off criterion for good fit is near .95 for TLI and CFA showed that this six-factor model fit the present data
CFI, near .06 for RMSEA, and near .08 for SRMR (Hu & well (CFI = .96, TLI = .99, RMSEA = .05, SRMR =
Bentler, 1999). These guidelines are approximate rules of .04). Cronbach’s internal consistency coefficients for these
thumb, developed in the context of continuous data, but scales (which contain only three items each) were .73, .81,
are currently thought also to be reasonable for categori- .76, .71, .86, and .70 respectively.
cal outcomes (Yu & Muthén, 2001, as cited in Muthén & OCI-R summed scores provide an indication of the
Muthén, 2001). severity and prevalence of OCD symptoms in the present
sample. As shown in Table I, the highest possible score
(12) on each subscale was observed for 5 of the 6 scales
Measures (the exception is Obsessing, for which the highest ob-
served score was 11). Further, 180 participants scored
Included in the battery were the 87-item OBQ (de- 21 or above on the total OCI-R, and 158 participants
scribed in the introduction), the Obsessive–Compulsive scored 4 or above on Obsessing. These cut-off values
Inventory-Revised (OCI-R; Foa et al., 2002), the Cen- were found to provide optimal levels of sensitivity and
ter for Epidemiological Studies-Depression Scale (CES- specificity for distinguishing between OCD patients and
D; Radloff, 1977), the Self-Rating Anxiety Scale (SAS; nonanxious control participants in Foa et al.’s study (2002,
Zung, 1971), and the Fear of Negative Evaluation scale p. 492–493).
(FNE; Watson & Friend, 1969) along with other ques- The CES-D consists of 20 items related to depres-
tionnaires used in separate research (these pertained to sion symptoms experienced in the past month, rated on
non-OCD anxiety symptoms or to disgust). Descriptive a Likert-type scale from 0 (rarely or none of the time;
statistics for all scales are shown in Table I. In each case, less than 1 day) to 4 (most or all of the time; 5–7 days).
scores were computed by summing all items. On the basis of a principle components analysis, Radloff
The OCI-R is an 18-item questionnaire based on the (1977) suggested the use of four subscales, correspond-
earlier 84-item OCI (Foa, Kozak, Salkovskis, Coles, & ing to: (a) depressed affect, (b) positive affect, (c) somatic
Amir, 1998). Participants rate the degree to which they and retarded activity, and (d) interpersonal [symptoms]
are bothered or distressed by OCD symptoms in the past
month on a 5-point scale from 0 (not at all) to 4 (ex- 7 Forbrevity, a detailed description of the results from factor analyses
tremely). The OCI-R is thought to assess six factors, cor- on measures other than the OBQ are not reported here, but they are
responding to (a) washing, (b) checking, (c) ordering, available upon request from the first author.
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581
Dimensionality of the OBQ
(p. 397). For the present purposes, only depressed affect
was of interest; thus items such as: [In the past week,] “I
felt that I could not shake off the blues even with help
from my family or friends,” and “I felt sad,” were used
(items were 3, 6, 14, 17, and 18). Three of four fit in-
dices from a one-factor CFA of these items indicated that
it was reasonable to combine them (CFI = .99, TLI =
.99, RMSEA = .10, SRMR = .04). Cronbach’s internal
consistency coefficient for the depression factor was .87.
The SAS is a 20-item measure of trait anxiety, espe-
cially as expressed through somatic symptoms; for exam-
ple, “I am bothered by headaches, neck, and back pains.”
Participants respond on a Likert-type scale from 1 (none or
a little of the time) to 4 (most or all of the time). The SAS is
usually scored by summing all items (Zung, 1977); how-
ever, a one-factor model fit the present data poorly (CFI =
.80, TLI = .89, RMSEA = .09, SRMR = .08). Thus, on
the basis of an EFA (with oblique promax rotation), one
strong factor (loadings .53, .78, .83, and .92), represent-
ing trait anxiety but not somatization, was selected for use
here. The four included items are: (1) “I feel more nervous
and anxious than usual,” (2) “I feel afraid for no reason
at all,” (3) “I get upset easily or feel panicky,” and (4) “I
feel like I’m falling apart and going to pieces.” Cronbach’s
alpha for these four items was .77.
The FNE is a self-report measure of social anxiety
consisting of 30 true/false items. Although the scale has
been historically treated as unidimensional, it was recently
shown that it may be useful to consider straightforwardly
and reverse-worded items as correlated, but distinct, sub-
scales (Rodebaugh et al., 2003). A two-factor model fit
the present data adequately (TLI = .96, CFI = .89, RM-
SEA = .07, SRMR = .08), and better than a one-factor
model (TLI = .95, CFI = .85, RMSEA = .08, SRMR =
.09). To statistically compare the one- and two-factor mod-
els, they were refitted with WLS estimation (nested model
tests are not possible with WLSMV because of the way
the degrees of freedom for the chi squares are computed,
see Muthén & Muthén, 2001, p. 358). The likelihood ra-
tio difference test showed that the 2-factor model fit the
data significantly better than the 1-factor model, χ 2 (1) =
51.13, p < .001. Thus, only the straightforwardly worded
FNE items (“FNE-S” subscale, Rodebaugh et al., 2003)
are used here. Cronbach’s alpha for these 17 items was .90.
RESULTS
Confirmatory Factor Analyses of the OBQ
The first CFA model represented the 44-item, 3-
factor structure that the OCCWG recently proposed. Taken
January 28, 2004 10:55 Style file version June 25th, 2002
117
together, the indices reflected poor fit (CFI = .72, TLI =
.93, RMSEA = .12, SRMR = .08, n = 995). Because all
participants in the OCCWG’s student sample are women,
the 3-factor model also was fitted to data from women only,
to evaluate the hypothesis that it is applicable to students
who are women. Model fit was poor (CFI = .75, TLI =
.94, RMSEA = .12, SRMR = .08, n = 746 women). Al-
though the OCCWG currently appears to prefer 3 factors
to their original, rationally derived 6, it was desirable to
test all theoretically predicted models in a confirmatory
framework before resorting to exploratory analysis. Thus,
a CFA model was fitted that included all 87 OBQ items,
and represented the six theoretical subscales. This model
fit the data poorly (CFI = .59, TLI = .94, RMSEA = .11,
SRMR = .08, n = 995).
Exploratory Factor Analyses of the OBQ
Given the poor fit of the confirmatory models, an
EFA was undertaken. Promax rotation, the only oblique
rotation available in Mplus at the time of this analysis,
was performed to allow for nonzero correlations among
the factors. First, in accordance with the OCCWG’s pro-
cedure (OCCWG, 2003) item 71 was eliminated because
the wording was nearly identical to that of item 23. Thus,
the EFA was performed with 86 items. The number of
factors was selected based on substantive interpretabil-
ity, and three fit indices: the TLI, RMSEA (both described
earlier), and the root mean square residual (RMSR), an un-
standardized version of the SRMR. Fit improves as RMSR
approaches 0. A 4-factor solution was interpretable and fit
the data well (TLI = .98, RMSEA = .07, RMSR = .04).
All items and loadings are shown in Table II. Sixty-
six items had a loading of .50 or greater, and were as-
signed to a factor. Items whose loadings fell short of this
(arbitrary) criterion appear with the items with which they
would aggregate if a less stringent criterion were used for
factor assignment. The letter printed in the far left col-
umn indicates which of the six theoretical concepts (re-
viewed in the introduction) the OCCWG intended the item
to measure. Letters in the second-to-left column indicate
the factor to which the item was assigned in the OCCWG’s
recent EFA. A number in this column is the item’s load-
ing, from the OCCWG’s analysis (OCCWG, 2003), on
the factor that is analogous to the factor on which it loads
most saliently in the present analysis. As explained below,
factor 2 is analogous to ICT, factor 3 is analogous to PC
and factor 4 is analogous to RH.
As shown in Table II, 32 items loaded .50 or above on
the first factor, including some items from each of the six
theoretical domains (threat: 11, importance of thoughts: 6,
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004 10:55 Style file version June 25th, 2002

118 Woods, Tolin, and Abramowitz

Table II. Exploratory Factor Analysis of the 86-Itema OBQ (n = 995)

Factors
Previous
classification OBQ item 1 2 3 4 h2

T RH 61. I am more likely than other people to accidentally cause harm to myself or to others. .73 −.05 −.05 .12 .55
T — 30. Bad things are more likely to happen to me than to other people. .71 −.08 .14 .07 .59
R — 7. When I hear about a tragedy, I can’t stop wondering if I am responsible in some way. .70 .14 −.12 .10 .59
T RH 50. If I do not take extra precautions, I am more likely than others to have or cause a serious disaster. .69 .02 .00 .18 .62
U — 26. If an unexpected change occurs in my daily life, something bad will happen. .69 .07 .06 .09 .63
T — 80. When things go too well for me, something bad will follow. .69 −.17 .12 .14 .55
T — 40. Small problems always seem to turn into big ones in my life. .65 −.19 .23 .20 .61
P PC 84. No matter what I do, it won’t be good enough. .65 −.16 .39 −.01 .64
C — 37. If I have an intrusive thought while I’m doing something, what I’m doing will be ruined. .64 .27 .07 −.05 .65
U — 63. If something unexpected happens, I will not be able to cope with it. .64 .01 .10 .14 .58
I ICT 64. Having bad thoughts means I am weird or abnormal. .63 .39 .04 −.21 .66
T RH 68. Even when I am careful, I often think that bad things will happen. .63 −.06 −.01 .35 .63
C ICT 69. Having intrusive thoughts means I’m out of control. .62 .40 .00 −.07 .72
T — 82. When anything goes wrong in my life, it is likely to have terrible effects. .62 −.05 .14 .26 .65
T — 9. I am much more likely to be punished than are others. .62 .15 .04 −.01 .50
T — 16. Things that are minor annoyances for most people seem like disasters for me. .62 −.08 .24 .04 .52
C — 75. Being unable to control unwanted thoughts will make me physically ill. .61 .36 −.06 −.11 .59
U — 85. I often think that I will be overwhelmed by unforeseen events. .61 −.06 .10 .24 .58
I — 14. The more distressing my thoughts are, the greater the risk that they will come true. .60 .35 −.03 .01 .64
I — 49. Thinking about a good thing happening can prevent it from happening. .59 .06 .03 .08 .45
P — 51. If I don’t do as well as other people, that means I am an inferior person. .59 −.12 .55 −.14 .68
I — 18. The more I think of something horrible, the greater the risk it will come true. .58 .32 −.07 .04 .59
I ICT 76. Having violent thoughts means I will lose control and become violent. .58 .32 −.02 −.11 .52
R — 4. If I imagine something bad happening, then I am responsible for making sure it doesn’t happen. .57 .14 −.11 .17 .48
I ICT 34. If I have aggressive thoughts or impulses about my loved ones, this means I may secretly want .57 .29 −.02 −.13 .46
to hurt them.
U — 3. If I am uncertain, there is something wrong with me. .56 .25 .07 −.14 .47
R — 47. I often believe I am responsible for things that other people don’t think are my fault. .54 .05 .03 .25 .51
U — 70. It is terrible to be surprised. .54 .14 .08 −.02 .42
T RH 6. I often think things around me are unsafe. .53 −.02 −.03 .28 .39
C — 5. If I don’t control my unwanted thoughts, something bad is bound to happen. .51 .48 −.13 −.01 .63
C — 54. To avoid disasters, I need to control all the thoughts or images that pop into my mind. .50 .35 .01 .15 .65
T RH 72. Harmful events will happen unless I am very careful. .50 .06 .03 .36 .58
C — 8. Whenever I lose control of my thoughts, I must struggle to regain control. .47 .35 −.02 .03 .51
C — 15. I can have no peace of mind as long as I have intrusive thoughts. .48 .34 −.03 .12 .56
T RH 39. Avoiding serious problems (e.g., illness or accidents) requires constant effort on my part. .45 .03 .03 .34 .49
U PC 10. If I’m not absolutely sure of something, I’m bound to make a mistake. .42 .04 .20 .20 .44
T RH 79. Even ordinary experiences in my life are full of risk. .42 −.06 −.04 .39 .39
U — 87. I need the people around me to behave in a predictable way. .40 .02 .23 .18 .42
I — 48. If an intrusive thought pops into my mind, it must be important. .31 .22 −.06 .23 .32
I ICT 58. Having a blasphemous thought is as sinful as committing a sacrilegious act. .00 .79 −.05 .02 .62
I ICT 83. Having a bad thought is morally no different than doing a bad deed. .12 .76 −.09 .03 .64
I ICT 24. For me, having bad urges is as bad as actually carrying them out. .24 .67 −.04 .02 .63
C .45 29. I must be ready to regain control of my thinking whenever an intrusive thought or image occurs. .11 .60 .02 .19 .59
I ICT 55. I should not have bizarre or disgusting thoughts. .14 .64 .00 .00 .51
C .39 2. Having control over my thoughts is a sign of good character. −.20 .59 .14 .03 .36
C ICT 59. I should be able to rid my mind of unwanted thoughts. −.17 .59 .04 .22 .44
C ICT 86. If I don’t control my thoughts, I’ll be punished. .38 .59 −.04 −.03 .66
I ICT 46. Having nasty thoughts means I am a terrible person. .41 .54 .14 −.17 .64
C .34 12. I would be a better person if I gained more control over my thoughts. .28 .52 .02 .04 .52
C .40 17. I must know what is going on in my mind at all times so I can control my thoughts. .26 .44 .14 .09 .52
I .33 1. Having bad thoughts or urges means I’m likely to act on them. .36 .41 −.09 −.07 .36
C .38 44. If I exercise enough will-power, I should be able to gain complete control over my mind. −.11 .40 .22 .24 .41
I ICT 66. Having an unwanted sexual thought or image means I really want to do it. .25 .38 .05 .01 .32
P PC 45. For me, things are not right if they are not perfect. .27 .00 .74 −.07 .69
P PC 19. In order to be a worthwhile person, I must be perfect at everything I do. .45 −.09 .70 −.17 .72
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004 10:55 Style file version June 25th, 2002

Dimensionality of the OBQ 119

Table II. (Continued)

Factors
Previous
classification OBQ item 1 2 3 4 h2

U PC 32. It is essential for me to consider all possible outcomes of a situation. −.15 .13 .70 .21 .69
P .49 22. If I fail at something, I am a failure as a person. .58 −.07 .63 −.22 .75
P .47 36. If someone does a task better than I do, that means I failed the whole task. .58 −.02 .60 −.27 .73
U PC 57. It is essential for everything to be clear cut, even in minor matters. −.07 .12 .59 .29 .65
P PC 28. If I can’t do something perfectly, I shouldn’t do it at all. .48 −.07 .59 −.14 .62
P PC 74. I must keep working at something until it’s done exactly right. −.06 .01 .57 .33 .59
P PC 78. If I don’t do a job perfectly, people won’t respect me. .44 −.11 .57 .04 .65
P PC 56. For me, making a mistake is as bad as failing completely. .51 .03 .56 −.12 .71
P PC 65. I must be the best at things that are important to me. .00 −.02 .55 .20 .45
P PC 33. Even minor mistakes mean a job is not complete. .05 .05 .55 .20 .52
P PC 13. Things should be perfect according to my own standards. .20 −.10 .53 .04 .41
P PC 31. I must work to my full potential at all times. −.35 .10 .53 .25 .44
U PC 35. I must be certain of my decisions. −.26 .09 .42 .34 .41
R .45 21. It is ultimately my responsibility to ensure that everything is in order. .24 .01 .40 .22 .46
P PC 42. I should be upset if I make a mistake. .27 −.03 .38 .21 .42
P .46 11. There is only one right way to do things. .31 .23 .35 −.17 .37
R RH 41. For me, not preventing harm is as bad as causing harm. .23 −.10 −.21 .87 .71
R RH 38. In all kinds of daily situations, failing to prevent harm is just as bad as deliberately .23 −.11 −.22 .85 .67
causing harm.
R RH 77. To me, failing to prevent a disaster is as bad as causing it. .24 −.05 −.11 .72 .58
R RH 43. I should make sure others are protected from any negative consequences of my decisions −.18 .10 .08 .66 .48
or actions.
R RH 20. When I see any opportunity to do so, I must act to prevent bad things from happening. −.12 .00 −.02 .61 .32
R RH 23. Even if harm is very unlikely, I should try to prevent it at any cost. −.18 .06 .11 .59 .46
R .47 62. For me, even slight carelessness is inexcusable when it might affect other people. .02 .13 .13 .55 .49
R RH 27. If I don’t act when I foresee danger, then I am to blame for any consequences. .15 .04 .03 .54 .42
U .41 73. I should go to great lengths to get all the relevant information before I make a decision. −.08 .01 .24 .54 .44
R .33 25. I must think through the consequences of even my smallest actions. −.05 .17 .15 .50 .44
U RH 53. In order to feel safe, I have to be as prepared as possible for anything that could go wrong. .27 .01 .10 .49 .49
R RH 67. If my actions could have even a small effect on a potential misfortune, I am responsible .18 .16 .04 .49 .49
for the outcome.
T .42 52. I believe that the world is a dangerous place. .23 −.10 −.01 .40 .23
R .40 81. If I take sufficient care, I can prevent any harmful accident from occurring. .20 .11 .10 .30 .28
U .41 60. I should be 100% certain that everything around me is safe. .17 .17 .21 .26 .37

Note. h 2 = communality. Items are ordered by factor, then by magnitude of loading within factor. Bolded loadings are .50 or larger. Letters in the far left
column represent the factor the item was theoretically expected to measure: U = tolerance for uncertainty; T = threat estimation; C = control of thoughts;
I = importance of thoughts; R = inflated responsibility; P = perfectionism. Letters in the second-to-left column represent the factor the item was assigned
to in the OCCWG’s recent EFA: RH = inflated responsibility for harm; PC = perfectionism and intolerance of uncertainty; ICT = importance of, and
control over, thoughts,— = item with its highest loading on the present general factor did not load above .49 on any of the OCCWG’s factors. A number
in the second-to-left column is the item’s loading (in the OCCWG’s analysis; OCCWG, 2003) on the factor that is analogous to the factor on which it
loads most saliently in the present analysis. For example, in the OCCWG’s analysis, item 29 loaded .45 on the ICT factor, which is analogous to the
present thought factor, on which its loading was .60.
a 86, rather than 87 items were analyzed in accordance with the OCCWG’s procedure—item 71 was dropped.

control of thoughts: 5, tolerance of uncertainty: 5, respon-
sibility: 3, perfectionism: 2), and each of the OCCWG’s
more recent factors (RH: 5, PC: 1, ICT: 4). An additional
seven items with all loadings <.50 were more related to
this factor than the to other factors (threat: 2, control of
thoughts: 2, tolerance of uncertainty: 2, importance of
thoughts: 1, RH: 2, PC: 1). The construct that explains
correlations among these items appears to be negative be-
liefs and expectations, that are relevant, but not specific,
to OCD. The two items with the largest loadings were:
“I am more likely than other people to accidentally cause
harm to myself or others,” and “Bad things are more likely
to happen to me than to other people.” Cronbach’s in-
ternal consistency coefficient (α) for the scale was .95,
and descriptive statistics for summed scores are shown in
Table I.
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004
120
The second factor in Table II related to beliefs about
thoughts. Ten items loaded .50 or above: Five items about
importance of thoughts, and five pertaining to control over
thoughts. Seven of these were assigned to the OCCWG’s
ICT factor, and the remaining three loaded .34, .39, and
.45 (respectively) on the ICT factor in the OCCWG’s anal-
ysis. An additional four items (control: 2, importance: 2)
were more related to the present factor 2 than to other
factors. One of these was assigned to the ICT factor, and
the other three loaded .33, .38, and .40 on the ICT fac-
tor. Example items are: “Having a blasphemous thought
is as sinful as committing a sacrilegious act,” and “Hav-
ing a bad thought is morally no different than doing a bad
deed.” Cronbach’s α was .86 for this factor, and descriptive
statistics are shown in Table I.
The third factor (Cronbach’s α = .92, descriptive
statistics in Table I) was dominated by perfectionism
items. Fourteen items had loadings ≥.50; 12 of these were
from the perfectionism domain and 2 from tolerance of un-
certainty. The OCCWG assigned 12 of these items to the
PC factor, and the remaining 2 loaded .47 and .49 on the
PC factor, respectively. Four items were more related to
the present factor 3 than to other factors (perfectionism: 2,
tolerance of uncertainty: 1, responsibility: 1). Two of these
were assigned to the PC factor, and the others loaded .45
and .46 on the PC factor. Items with large loadings were,
“For me, things are not right if they are not perfect,” and
“In order to be a worthwhile person, I must be perfect at
everything I do.”
The fourth factor (Cronbach’s α = .87, descriptives
in Table I) reflected inflated responsibility. Ten items had
loadings ≥.50, nine of which were related to responsi-
bility (1 was about tolerance of uncertainty). Seven of
these were assigned to the OCCWG’s RH factor, and the
remaining items loaded .33, .41, and .47 (respectively)
on the RH factor. In the present analysis, five additional
items were more related to factor 4 than to other factors
(threat: 1, uncertainty: 2, responsibility: 2). Of these, two
were assigned to the RH factor, and the others loaded
.40, .41, and .42 on the RH factor. Items with large load-
ings included, “For me, not preventing harm is as bad as
causing harm,” and “In all kinds of daily situations, fail-
ing to prevent harm is just as bad as deliberately causing
harm.”
Communalities (in Table II) ranged from .23 to .75,
with a mean of .53 (standard deviation = .12), indicat-
ing that a fair amount of variance in most OBQ items
could be explained by the factors. The moderate correla-
tions among factors (ranging from .36 to .49), shown in
Table III, suggest that the four dimensions are correlated
but distinct.
10:55 Style file version June 25th, 2002
Woods, Tolin, and Abramowitz
Table III. Correlations Among OBQ Factors
General Thoughts Perfectionism
Thoughts .44
Perfectionism .36 .36
Responsibility .38 .45 .49
Relationship Between OBQ Factors and Other Scales
To assess the relationships between OBQ factors and
other measures of psychopathology, a structural equa-
tion model (SEM) was fitted with the four OBQ fac-
tors as latent dependent variables, predicted by two
demographic observed variables (sex, ethnicity), six la-
tent variables corresponding to subscales of the OCI-R
(checking, washing, obsessing, mental neutralizing, or-
dering, and hoarding), and three additional latent vari-
ables representing SAS (items 1–4), CESD-depressed af-
fect, and FNE-S. The OBQ-general factor was expected
to be best predicted by SAS, CESD-depressed affect, and
FNE-S, and less strongly by OCD symptoms, whereas
the other three OBQ factors were expected to be best
predicted by at least some types of OCD symptoms,
and less strongly by SAS, CESD-depressed affect and
FNE-S.
Indicators of the OBQ factors were the 66 items that
had loaded .50 or above on a factor in the EFA, and indi-
cators of other latent variables were the items described
in the Method section. Demographic variables were in-
cluded in the model to potentially reduce error variability;
no specific relationships were expected. To avoid cate-
gories with small numbers of people, ethnic groups were
dichotomized into White/non-White for this analysis. Be-
cause of casewise deletion on multiple variables, the sam-
ple size for the SEM model was 959. The WLSMV es-
timator in the Mplus program was used to fit the model.
Global model fit was evaluated with the TLI and RMSEA
described earlier, as well as the weighted root mean square
residual (WRMR), a statistic similar to SRMR except that
it is weighted by the variance of the parameter estimates.
WRMR ranges from 0 to infinite, with smaller values in-
dicating better fit. Global fit was acceptable (TLI = .96,
RMSEA = .07, WRMR = 1.92).
Regression relationships are reported in Table IV. The
general OBQ factor was the only one significantly pre-
dicted by CES-D and SAS. Standardized regression coef-
ficients (b∗ s) were .09 and .13, respectively. OBQ-general
was also significantly predicted by FNE-S, washing, and
obsessing, with b∗ s ranging from .11 to .35. Other vari-
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004 10:55 Style file version June 25th, 2002

Dimensionality of the OBQ 121

Table IV. Relationships Between OBQ Factors and Other Measures of Psychopathology

Obsessive beliefs factor

General Thoughts Perfectionism Responsibility
Predictor b (SE) b∗ p b (SE) b∗ p b (SE) b∗ p b (SE) b∗ p

Sex −0.04 (.07) −.08 .515 −0.14 (.06) −.08 .022 0.05 (.05) .03 .316 0.08 (.06) .04 .206
Ethnicity 0.02 (.06) .01 .717 −0.12 (.06) −.07 .044 0.07 (.05) .04 .174 <0.01 (.06) <.01 .981
CES-D 0.08 (.04) .09 .021 −0.08 (.05) −.08 .123 0.07 (.04) .08 .094 0.09 (.05) .09 .113
SAS 0.13 (.05) .13 .009 −0.07 (.06) −.08 .245 0.07 (.05) .08 .192 −0.08 (.07) −.08 .273
FNE-S 0.18 (.03) .20 <.0001 0.06 (.04) .07 .101 0.18 (.03) .22 <.0001 0.05 (.04) .06 .219
Washing 0.10 (.04) .11 .023 0.06 (.06) .07 .291 0.04 (.05) .05 .436 0.12 (.07) .13 .076
Checking −0.01 (.06) −.01 .861 −0.01 (.08) −.01 .924 −0.03 (.07) −.03 .681 0.02 (.09) .02 .834
Ordering −0.02 (.04) −.02 .634 0.04 (.05) .05 .394 0.28 (.04) .33 <.0001 0.14 (.06) .14 .014
Obsessing 0.34 (.05) .35 <.0001 0.70 (.07) .75 <.0001 0.07 (.05) .07 .227 0.26 (.07) .26 .0002
Hoarding 0.03 (.05) .03 .551 −0.01 (.06) −.01 .811 0.02 (.05) .02 .737 0.21 (.06) .18 .0006
Neutralizing 0.12 (.07) .11 .083 −0.17 (.09) −.18 .065 0.04 (.08) .04 .642 −0.18 (.11) −.17 .101

Note. b = unstandardized regression coefficient; SE = standard error; b∗ = standardized regression coefficient; CES-D = Center for Epidemiological
Studies Depression Scale (depressed affect items); SAS = Self-Rating Anxiety Scale (items 1–4); FNE-S = Fear of Negative Evaluation, straightfor-
wardly worded subscale, the last six predictors are factors of the Obsessive–Compulsive Inventory-Revised.

ables were not significant predictors of OBQ-general, DISCUSSION

with b∗ s between −.01 and .11. OBQ-thoughts was
the only factor significantly predicted by demographic
variables. In particular, non-White students, and men,
scored a little bit higher on the OBQ-thoughts subscale
than White students, and women, respectively (b∗ = .08
and .07, respectively). Controlling for sex and ethnic-
ity, OBQ-thoughts was strongly and significantly pre-
dicted by obsessing (b∗ = .75). Other variables were
not significantly predictive, with b∗ s between −.18 and
.07. OBQ-perfectionism was significantly predicted by
FNE-S and ordering, with b∗ s of .22 and .33, respec-
tively. Other predictors were nonsignificant, with b∗ s
between −.03 and .08. OBQ-responsibility was sig-
nificantly predicted by obsessing (b∗ = .26), ordering
(b∗ = .14), and hoarding (b∗ = .18). Coefficients for non-
significant predictors of OBQ-responsibility ranged from
−.17 to .13.
Measurement-model paths (i.e., those from latent
variables to items) were not of primary interest here, be-
cause factor structure was evaluated in separate analyses.
However, all such paths were statistically significant (ra-
tios of regression estimates to their standard error ranged
from 11.77 to 42.45), with standardized regression co-
efficients ranging from .35 to .93. Pairwise correlations
ranged from .23 to .77 among predictor factors, and from
.27 to .37 among OBQ factors. Multiple R 2 s showed that
predictors explained 58% of variance in OBQ-general,
44% in OBQ-threat, 40% in OBQ-perfectionism, and 24%
in OBQ-responsibility.
Recently proposed conceptual models of OCD posit
that cognitive phenomena underlie the etiology and main-
tenance of this disorder (e.g., Rachman, 1998; Salkovskis,
1999). The OBQ represents an ongoing effort to identify
and assess cognitive domains important in OCD. Authors
of the OBQ continue to revise and examine its factor struc-
ture, and affirm that this process will be aided by additional
research with independent samples, both clinical and non-
clinical (OCCWG, 2001). The purpose of the present study
was to contribute to the scale validation process by eval-
uating the dimensionality of the OBQ using data distinct
from that in use by the OCCWG. Dimensionality was eval-
uated with data from a large student sample that included
both men and women. Evaluation of the factor struc-
ture in this population is important because researchers
are likely to administer the OBQ to students in the
future.
A confirmatory model specifying six theoretical sub-
scales for the 87-item OBQ fit the data poorly. However,
the OCCWG (2003) has reduced the number of items from
87 to 44, on the basis of exploratory analyses with pa-
tients, which they (basically) replicated with women stu-
dents. They hypothesize three factors: (a) inflated respon-
sibility for harm (RH), (b) perfectionism and intolerance
of uncertainty (PC), and (c) importance of, and control
over, thoughts (ICT). A confirmatory model specifying
this structure with 44-items also fit the present data poorly,
including when it was fitted to data from women only.
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581
122
Possible explanations for the misfit of the confir-
matory models relate to the mode of administration and
methodological differences between confirmatory and
exploratory factor analysis. Although every effort was
made to replicate the paper version of the OBQ with the
web version used for this study, it is possible that the factor
structure could be influenced by the method of administra-
tion (i.e., computer versus paper–pencil). Whether, or how,
the tendency to respond honestly or thoughtfully to ques-
tions about OCD symptoms is influenced by computer-
based assessment is unknown. Students in this study did
not interact with a researcher in person at all. This may
have created a sense of anonymity that facilitated honest
responding, or it may have created a sense of exonera-
tion from any possible negative consequences of respond-
ing haphazardly. The environment in which participants
completed the questionnaires was not controlled, and may
have influenced their responses. Additionally, the 44-item
model may not fit well in a confirmatory framework, in
part, because items tend to load nontrivially on more than
one factor in EFA, but then items are assigned to a single
factor (with paths to all other factors set to 0) in CFA.
Reduction of item overlap among subscales may be an
important goal of future research.
Given the poor fit of the theory-driven models, an
EFA of 86 items8 was undertaken. A four-factor solution
was extracted, consisting of one large factor and three
smaller factors. The large factor is referred to as “OBQ-
general” because it included some items from each of
the six theoretical domains, and may reflect some gen-
eral aspect of anxiety disorders (e.g., negative beliefs and
expectations) that is less specific to OCD than domains
such as importance of and control over thoughts, inflated
responsibility and perfectionism. About one third of the
items on the general factor were from the overestimation
of threat domain, which the OCCWG suggested was rele-
vant, but not specific, to OCD (Sookman & Pinard, 2002).
Salkovskis and Forrester (2002) explained that the intoler-
ance of uncertainty and overestimation of threat domains
may be general vulnerability factors, rather than facets
that underlie OCD symptomatology in particular. Consis-
tently, there was no difference between people with OCD
and anxious controls on either the Threat Estimation or
the Intolerance of Uncertainty scales in the OCCWG’s
recent sample (Taylor et al., 2002). In the current paper,
OBQ-general was the only factor significantly predicted
by trait anxiety and depressed affect, and the effect size
for anxiety was quite a bit larger than that for other fac-
tors. It was also significantly predicted by social anxiety.
8 Eighty-six,
rather than 87, items were factor analyzed because item 71
was eliminated due to redundant wording with item 23.
January 28, 2004 10:55 Style file version June 25th, 2002
Woods, Tolin, and Abramowitz
Thus, the factor appears nonspecific to OCD. However,
OBQ-general is relevant to at least some aspects of OCD,
because it was significantly predicted by washing and ob-
sessing, and obsessing was the factor’s strongest predictor.
The three smaller factors are consistent with the OC-
CWG’s theoretical domains (OCCWG, 1997, 2001), and
similar to their recently proposed three factors (OCCWG,
2003). The thought factor represents the synthesis of
two of the OCCWG’s constructs: importance of thoughts
and control of thoughts. A merging of these domains is
expected because believing that an intrusive thought is
important, and may cause a feared event to occur, likely
motivates a person to control their thoughts to reduce dis-
tress or the likelihood of the feared event (Thordarson &
Shafran, 2002). A connection between the thought do-
mains also is in accord with empirical work indicating
that control of thoughts may be temporally secondary to
importance of thoughts (e.g., Rassin, Muris, Schmidt, &
Merckelbach, 2000). Furthermore, there was substantial
overlap between the present OBQ-thoughts factor and the
OCCWG’s ICT factor (OCCWG, 2003). Ignoring the ar-
bitrary criterion for assigning items to factors, 14 thought-
related items aggregated together (i.e., had their highest
loading on the same factor) in both the present analy-
sis and the OCCWG’s analysis (OCCWG, 2003) (items
1, 2, 12, 17, 24, 29, 44, 46, 55, 58, 59, 66, 83, and 86).
With demographic variables and other scales controlled in
an SEM model, there was a strong relationship between
OBQ-thoughts and obsessing.
A third factor, OBQ-perfectionism, included most
of the items from the rationally derived perfectionism
subscale of the OBQ-87, and 12 of the 16 items on the
OCCWG’s PC factor. Only two of the items originally
thought to measure perfectionism were not included in
the present perfectionism factor (items 51 and 84 loaded
more highly on the general factor than the first factor, but
also loaded .55 and .39, respectively, on the perfectionism
factor). The perfectionism factor also included two items
previously thought to measure intolerance of uncertainty
(items 32 and 57). As explained by Frost, Novara, and
Rhéaume (2002), a relation between perfectionism and
intolerance of uncertainty is expected.
OBQ-perfectionism was significantly predicted by
both social anxiety and ordering. A relation between so-
cial anxiety and perfectionism is consistent with research
showing that participants with social phobia score signif-
icantly higher in perfectionism compared to nonclinical
controls (e.g., Purdon, Antony, & Swinson, 1999). Order-
ing may be a type of sensory perfectionism, or “not-just-
right experience” (Coles, Frost, Heimberg, & Rhéaume,
2003), wherein an individual has a vague sense that an
experience (like getting dressed) was not quite right and
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581
Dimensionality of the OBQ
needs to be perfected. Sensory perfectionism is a fairly
unstudied construct that warrants further investigation.
The final factor, OBQ-responsibility, was dominated
by items from the rationally derived domain of inflated
responsibility, and included 7 of the 16 items assigned to
the OCCWG’s RH factor. Only four items thought to mea-
sure responsibility loaded more highly on a factor other
than OBQ-responsibility (items 4, 7, 21, and 47). Items
not thought to measure responsibility that were assigned
to the OBQ-responsibility factor (or that loaded highest on
this factor, but had a loading less than .50) were related to
tolerance of uncertainty or overestimation of threat. This
is consistent with the OCCWG’s recent RH factor, which
included responsibility, threat, and uncertainty items. The
present OBQ-responsibility factor differs from the OC-
CWG’s RH factor in that no threat items loaded .50 or
above (item 52 loaded .40), whereas seven threat items
loaded .50 or above on the RH factor. In the present anal-
ysis, most of the threat-related items were highly related
to the OBQ-general factor, rather than to other factors.
On the basis of SEM analyses, OBQ-responsibility
was significantly predicted by obsessing, ordering, and
hoarding. Given the coherence of theories about respon-
sibility and OCD symptoms (e.g., Salkovskis, 1985), one
might expect responsibility to relate to the other three types
of OCD symptoms as well. It is possible that such relation-
ships did not emerge because each type of OCD symptom
was measured by only three items. OBQ-responsibility
was not significantly predicted by depressed affect, trait
anxiety, or social anxiety, consistent with the suggestion
that inflated responsibility is fairly specific to OCD (e.g.,
Salkovkis & Forrester, 2002).
Overall, there are two primary differences between
the present and previous (OCCWG, 2003) EFA solutions.
One is the presence of a general factor. Given the sub-
jective nature of factor extraction, it is possible that two
reasonable, but different, sets of criteria for selecting the
number of factors could lead to different decisions. How-
ever, if it were merely the case that the OCCWG ignored
a factor that the present authors interpreted, the extraction
of three factors with the present data should lead to a solu-
tion that (nearly) matches the OCCWG’s. That is not the
case (details available from the first author).
An alterative explanation is that the presence of the
general factor is related to differences in the factor-analytic
methods. The OCCWG performed principal axis factor
analysis (PFA) of Pearson correlations, with oblique rota-
tion (unspecified), whereas the present authors used a type
of weighted least squares estimation of polychoric cor-
relations and oblique, promax rotation. However, if this
accounted for the presence of the general factor, principal
axis factor analysis (with each of several plausible types
January 28, 2004 10:55 Style file version June 25th, 2002
123
of oblique rotation) should produce a solution similar to
that obtained by the OCCWG. This is not the case (de-
tails available from the first author). Thus, the emergence
of a general factor in the present analysis, but not in the
OCCWG’s analysis, may relate to sampling variability,
differing methods of questionnaire administration in the
two studies, or a complex interaction of these differences,
along with differences in estimation and factor extraction
procedures.
There also is a discrepancy in the particular items
assigned to the nongeneral factors, which arises from dif-
ferences in the magnitude of the factor loadings. Items
were assigned to the factor on which their largest loading
was ≥.50, an arbitrary criterion used by both the OCCWG
and the present authors. Although such a criterion is useful
for creating subscales from factor analysis, the particular
items assigned to a factor in this fashion often vary from
one study to another, in part, because of sampling error in
the magnitude of the factor loadings. Additionally, load-
ings are influenced by the presence of the general factor
and the factor-analytic method. For some items, loadings
from the OCCWG’s analysis are smaller than the present
loadings. As was shown in Table II, every item that loaded
on one of the nongeneral factors in the current analysis
could have been assigned to the analogous OCCWG fac-
tor in their analysis, had the loading been just a little bit
larger (loadings for excluded items ranged from .33 to
.49). Similarly, loadings for some items, such as 35 and
42, were larger in the OCCWG’s analysis (.68 and .60,
respectively) than the present analysis (.42 and .38, re-
spectively), and were assigned to the previous PC factor,
but not to the present perfectionism factor.
Although the particular items assigned to each non-
general factor in the current analysis are not identical to
those assigned to the analogous factor in the OCCWG’s
analysis (OCCWG, 2003), the “analogous” factors are,
in fact, conceptually identical. Thus, it is reasonable to
conclude that, in students, subsets of OBQ items measure
three distinct factors corresponding to inflated responsibil-
ity for harm, perfectionism, and intolerance of uncertainty,
and importance of, and control over, thoughts. Results par-
tially cross-validate the factor structure proposed by the
OCCWG (2003) because evidence for the same three un-
derlying constructs was found in both analyses. However,
further research is needed to better understand the emer-
gence of a general factor when the entire OBQ was ana-
lyzed, as well as reasons for the misfit of the three-factor,
44-item model.
It would be useful if future research improved upon
the limitations of this study. The four factors suggested
by the present data were found by exploratory meth-
ods and should be cross-validated with other samples of
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581 January 28, 2004
124
students. Continued evaluation of the OBQ’s factor struc-
ture in patients with OCD is of course important, and a
direct comparison between patients and students (as in
multiple-group CFA) could help clarify whether it is jus-
tifiable to assume that the structure is the same in both
groups. Further, an examination of the relation between
OBQ factors and multiple measures of OCD symptoms
would permit a more thorough evaluation of the types of
symptoms to which OBQ factors relate than was possible
here. Assessment of relations between OBQ factors and
additional, or different, measures of anxiety, depression, or
other constructs also would strengthen conclusions about
the construct validity of the factors. Finally, a controlled
experiment in which participants are randomly assigned
to complete the OBQ either on paper, or on the web, would
be informative.
The broad aim of this research was to inform the
ongoing scale-validation process led by the OCCWG,
thereby helping to identify cognitive phenomena relevant
to OCD. Findings are consistent with theoretical prepo-
sitions and empirical data that have accumulated in the
literature on cognition in this complex and heterogeneous
disorder. The ultimate goal of such research is to enhance
understanding of the development and persistence of ob-
sessions and compulsions, and lead to progress in devel-
oping effective therapeutic procedures for this condition.
ACKNOWLEDGMENTS
This research was supported, in part, by NIMH
NRSA grant number 5 F31 MH67334-02 to Carol M.
Woods. The authors are grateful to Gail Steketee for com-
ments on drafts of this paper and information about the
OCCWG’s factor analysis prior to its publication.
REFERENCES
Abramowitz, J. S., Whiteside, S., Kalsy, S., & Tolin, D. F. (2003).
Thought control strategies in obsessive–compulsive disorder: A
replication and extension. Behaviour Research and Therapy, 41,
529–540.
Abramowitz, J. S., Whiteside, S. P., Lynam, D., & Klasy, S. (2003). Is
thought-action fusion specific to obsessive–compulsive disorder?:
A mediating role of negative affect. Behaviour Research and Ther-
apy, 41, 1069–1079.
Bentler, P. M. (1990). Comparative fit indices in structural models. Psy-
chological Bulletin, 107, 238–246.
Bentler, P. M. (1995). EQS structural equations program manual.
Encino, CA: Multivariate Software.
Carr, A. T. (1974). Compulsive neurosis: A review of the literature.
Psychological Bulletin, 81, 311–318.
Christoffersson, A. (1975). Factor analysis of dichotomized variables.
Psychometrika, 40, 5–32.
Coles, M. E., Frost, R. O., Heimberg, R. G., & Rhéaume, J. (2003). “Not
just right experiences”: Perfectionism, obsessive–compulsive fea-
10:55 Style file version June 25th, 2002
Woods, Tolin, and Abramowitz
tures and general psychopathology. Behaviour Research and Ther-
apy, 41, 681–700.
Foa, E. B., Huppert, J. D., Leiberg, S., Langner, R., Kichic, R., &
Salkovskis, P. M. (2002). The Obsessive-Compulsive Inventory:
Development and Validation of a Short Version. Psychological As-
sessment, 14, 485–496.
Foa, E. B., & Kozak, M. J. (1986). Emotional processing of fear: Expo-
sure to corrective information. Psychological Bulletin, 99, 20–35.
Foa, E. B., Kozak, M. J., Salkovskis, P. M., Coles, M. E., & Amir,
N. (1998). The validation of a new obsessive compulsive disorder
scale: The Obsessive Compulsive Inventory. Psychological Assess-
ment, 10, 206–214.
Foa, E. B., Sacks, M. B., Tolin, D. F., Przeworski, A., & Amir, N. (2002).
Perception of responsibility for harm in OCD with checking and
non-checking compulsions: A replication and extension. Journal
of Anxiety Disorders, 16, 443–453.
Flora, D. B. (2002, June). Evaluation of categorical variable method-
ology for confirmatory factor analysis with ordinal data. Paper
presented at the annual meeting of the Psychometric Society,
Chapel Hill, NC.
Frost, R. O., Lahart, C. M., Dugas, K. M., & Sher, K. J. (1988). In-
formation processing among non-clinical compulsives. Behaviour
Research and Therapy, 26, 275–277.
Frost, R. O., Novara, C., & Rhéaume, J. (2002). Perfectionism in ob-
sessive compulsive disorder. In R. O. Frost & G. Steketee (Eds.),
Cognitive approaches to obsessions and compulsions: Theory, as-
sessment, and treatment (pp. 91–105). New York: Pergamon.
Frost, R. O., & Steketee, G. (1997). Perfectionism in obsessive–
compulsive disorder patients. Behaviour Research and Therapy,
35, 291–296.
Frost, R. O., & Steketee, G. (Eds.). (2002). Cognitive approaches to
obsessions and compulsions: Theory, assessment, and treatment.
Boston: Pergamon.
Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure mod-
eling: Sensitivity to underparameterized model misspecification.
Psychological Methods, 3, 424–453.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covari-
ance structure analysis: Conventional criteria versus new alterna-
tives. Structural Equation Modeling, 6, 1–55.
Jöreskog, K. G., & Sörbom, D. (1981). LISREL V: Analysis of lin-
ear structural relationships by the method of maximum likelihood.
Chicago: National Educational Resources.
Muthén, B. O. (1978). Contributions to factor analysis of dichotomous
variables. Psychometrika, 43, 551–560.
Muthén, B. O. (1984). A general structural equation model with di-
chotomous, ordered categorical, and continuous latent variable in-
dicators. Psychometrika, 49, 115–132.
Muthén, B. O., du Toit, S. H. C., & Spisic, D. (1997). Robust inference
using weighted least squares and quadratic estimating equations in
latent variable modeling with categorical and continuous outcomes.
Manuscript obtained from B. Muthén.
Muthén, B. O., & Satorra, A. (1995). Technical aspects of Muthén’s
LISCOMP approach to estimation of latent variable relations with a
comprehensive measurement model. Psychometrika, 60, 489–503.
Muthén, L. K., & Muthén, B. O. (2001). Mplus user’s guide. Los Angeles,
CA: Muthén & Muthén.
Obsessive Compulsive Cognitions Working Group. (1997). Cognitive
assessment of obsessive–compulsive disorder. Behaviour Research
and Therapy, 35, 667–681.
Obsessive Compulsive Cognitions Working Group. (2001). Develop-
ment and initial validation of the obsessive beliefs questionnaire
and the interpretation of intrusions inventory. Behaviour Research
and Therapy, 39, 987–1006.
Obsessive Compulsive Cognitions Working Group. (2003). Psychome-
tric validation of the Obsessive Belief Questionnaire and the In-
terpretation of Intrusions Inventory: Part 2. Factor analyses and
testing of a brief version. Manuscript submitted for publication.
Purdon, C., Antony, M. M., & Swinson, R. P. (1999). Psychometric
properties of the Frost multidimensional perfectionism scale in a
P1: KEF
Journal of Psychopathology and Behavioral Assessment (JOBA) pp1116-joba-480581
Dimensionality of the OBQ
clinical anxiety disorders sample. Journal of Clinical Psychology,
55, 1271–1286.
Purdon, C., & Clark, D. A. (1994). Obsessive intrusive thoughts in non-
clinical subjects. Part II. Cognitive appraisal, emotional response
and thought control strategies. Behaviour Research and Therapy,
32, 403–410.
Rachman, S. (1998). A cognitive theory of obsessions: elaborations.
Behaviour Research and Therapy, 36, 385–401.
Rachman, S., & de Silva, P. (1978). Abnormal and normal obsessions.
Behaviour Research and Therapy, 16, 233–248.
Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale
for research in the general population. Applied Psychological Mea-
surement, 1, 385–401.
Rassin, E., Muris, P., Schmidt, H., & Merckelbach, H. (2000). Rela-
tionships between thought-action fusion, thought suppression, and
obsessive–compulsive symptoms: A structural equation modeling
approach. Behaviour Research and Therapy, 38, 889–897.
Rodebaugh, T. L., Woods, C. M., Thissen, D. M., Heimberg, R. G.,
Chambless, D. L., & Rapee, R. M. (2003). More information from
fewer questions: The factor structure and item properties of the
original and brief fear of negative evaluation scale. Manuscript
submitted for publication.
Salkovskis, P. M. (1985). Obsessional-compulsive problems: A
cognitive-behavioural analysis. Behaviour Research and Therapy,
23, 571–583.
Salkovskis, P. M. (1999). Understanding and treating obsessive–
compulsive disorder. Behaviour Research and Therapy, 37, S29–
S52.
Salkovskis, P. M., & Forrester E. (2002). Responsibility. In R. O. Frost &
G. Steketee (Eds.), Cognitive approaches to obsessions and compul-
sions: Theory, assessment, and treatment (pp. 45–61). New York:
Pergamon.
Satorra, A. (1992). Asymptotic robust inferences in the analysis of mean
and covariance structures. In P. V. Marsden (Ed.), Sociological
methodology (pp. 249–278). New York: Plenum.
Shafran, R., Thordarson, D. S., & Rachman, S. (1996). Thought-action
fusion in obsessive compulsive disorder. Journal of Anxiety Disor-
ders, 10, 379–391.
Sookman, D., & Pinard, G. (2002). Overestimation of threat and intol-
erance of uncertainty in obsessive compulsive disorder. In R. O.
Frost & G. Steketee (Eds.), Cognitive approaches to obsessions
January 28, 2004 10:55 Style file version June 25th, 2002
125
and compulsions: Theory, assessment, and treatment (pp. 63–89).
New York: Pergamon.
Taylor, S., Kyrios, M., Thordarson, D. S., Steketee, G., & Frost, R. O.
(2002). Development and validation of instruments for measuring
intrusions and beliefs in obsessive compulsive disorder. In R. O.
Frost & G. Steketee (Eds.), Cognitive approaches to obsessions
and compulsions: Theory, assessment, and treatment (pp. 117–138).
New York: Pergamon.
Thordarson, D. S., & Shafran, R. (2002). Importance of thoughts. In R. O.
Frost & G. Steketee (Eds.), Cognitive approaches to obsessions
and compulsions: Theory, assessment, and treatment (pp. 15–28).
New York: Pergamon.
Tolin, D. F., Abramowitz, J. S., & Brigidi, B. D., & Foa, E. B. (2003). In-
tolerance of uncertainty in obsessive–compulsive disorder. Journal
of Anxiety Disorders, 17, 233–242.
Tolin, D. F., Abramowitz, J. S., Hamlin, C., Foa, E. B., & Synodi, D. S.
(2002). Attributions for thought suppression failure in obsessive–
compulsive disorder. Cognitive Therapy and Research, 26, 505–
517.
Tolin, D. F., Abramowitz, J. S., Przeworski, A., & Foa, E. B. (2002).
Thought suppression in obsessive–compulsive disorder. Behaviour
Research and Therapy, 40, 1251–1270.
Tolin, D. F., Hamlin, C., & Foa, E. B. (2002). Directed forgetting
in obsessive–compulsive disorder: Replication and extension. Be-
haviour Research and Therapy, 40, 793–803.
Tolin, D. F., Woods, C. M., & Abramowitz, J. S. (2003). Relationship
between obsessional beliefs and obsessive–compulsive symptoms.
Cognitive Therapy and Research, 27, 657–669.
Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum
likelihood factor analysis. Psychometrika, 38, 1–10.
Watson, D., & Friend, R. (1969). Measurement of social-evaluative anx-
iety. Journal of Consulting and Clinical Psychology, 33, 448–457.
Woods, C. M., Frost, R. O., & Steketee, G. (2002). Obsessive–
compulsive (OC) symptoms and subjective severity, probability,
and coping ability estimations of future negative events. Clinical
Psychology and Psychotherapy, 9, 104–111.
Wroe, A. L., & Salkovskis, P. M. (2000). Causing harm and allowing
harm: A study of beliefs in obsessional problems. Behaviour Re-
search and Therapy, 38, 1141–1162.
Zung, W. W. K. (1971). A rating instrument for anxiety disorders. Psy-
chosomatics, 12, 371–379.