BEHAVIOUR

RESEARCH AND
THERAPY
PERGAMON Behaviour Research and Therapy 36 (1998) 1143±1154

Dimensions of perfectionism across the anxiety disordersp

Martin M. Antony a, b, *, Christine L. Purdon a, c, Veronika Huta d,
Richard P. Swinson b
a
Department of Psychology, St. Joseph's Hospital, Hamilton, Ont., Canada
b
Department of Psychiatry and Behavioural Neurosciences, McMaster University, Hamilton, Canada
c
Department of Psychology, University of Waterloo, Waterloo, Canada
d
Department of Psychology, McGill University, Montreal, Canada
Received 20 January 1998

Abstract

To explore the role of perfectionism across anxiety disorders, 175 patients with either panic disorder
(PD), obsessive compulsive disorder (OCD), social phobia, or speci®c phobia, as well as 49 nonclinical
volunteers, completed two measures [Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R., (1990).
The dimensions of perfectionism. Cognitive Therapy and Research, 14, 449-468; Hewitt, P. L., & Flett,
G. L., (1991). Perfectionism in the self and social contexts: Conceptualization, assessment and
association with psychopathology. Journal of Personality and Social Psychology, 60, 456-470.] that assess
a total of nine di€erent dimensions of perfectionism. Relative to the other groups, social phobia was
associated with greater concern about mistakes (CM), doubts about actions (DA), and parental criticism
(PC) on one measure and more socially prescribed perfectionism (SP) on the other measure. OCD was
associated with elevated DA scores relative to the other groups. PD was associated with moderate
elevations on the CM and DA subscales. The remaining dimensions of perfectionism failed to
di€erentiate among groups. The clinical implications of these ®ndings are discussed. # 1998 Elsevier
Science Ltd. All rights reserved.

1. Introduction

The term `perfectionism' refers to the desire to achieve the highest standards of performance,
in combination with unduly critical evaluations of one's performance (Frost et al., 1990).

p
An earlier version of this paper was presented in November 1996 at the meeting of the Association for
Advancement of Behavior Therapy in New York City.
* Corresponding author: Department of Psychology, St. Joseph's Hospital, 50 Charlton Ave East., Hamilton,
Ontario, Canada L8N 4A6. Tel.: +1-905-522-1155, ext. 3048; fax: +1-905-521-6120; e-mail
mantony@stjosham.on.ca

0005-7967/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved.

PII: S 0 0 0 5 - 7 9 6 7 ( 9 8 ) 0 0 0 8 3 - 7
1144 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

Perfectionistic individuals, then, are individuals who believe that they can and should achieve
perfect performance, perceive anything less than perfect performance as unsatisfactory, and
selectively attend to cues that their standards have not been met (Hamacheck, 1978; Burns,
1980; Pacht, 1984). Thus, perfectionistic individuals are likely to be unsatis®ed with their
performance, as they consistently set demands that they are unable to meet.
Although the detrimental e€ects of this paradox are obvious, and clinicians have long
proposed a relationship between perfectionism and psychopathology, only recently have
investigators attempted to de®ne the construct of perfectionism precisely enough for its role in
psychopathology to be examined empirically. Furthermore, whereas earlier theorists had
described perfectionism as a unidimensional construct (see Burns, 1980; Pacht, 1984),
investigators have only recently begun to consider the multidimensional nature of
perfectionism.
Frost et al. (1990) were the ®rst to develop a measure designed speci®cally to assess
dimensions of perfectionism in clinical and nonclinical groups. Through review of the existing
literature on perfectionism, this group of researchers hypothesized that the construct of
perfectionism is comprised of six dimensions: (a) a tendency to react negatively to mistakes and
to equate mistakes with failure (concern over mistakes), (b) a tendency to doubt the quality of
one's performance (doubts about actions), (c) a tendency to set very high standards and place
excessive importance on these for self-evaluation (personal standards), (d) a tendency to
perceive one's parents as having high expectations (parental expectations), (e) a tendency to
perceive one's parents as being overly critical (parental criticism) and (f) a tendency to
emphasize the importance of order and organization (organization). The scale based on these
dimensions was referred to as the multidimensional perfectionism scale (MPS-F).
At about the same time, Hewitt and Flett (1991a) also developed a multidimensional
measure of perfectionism. These researchers argued that the existing views of perfectionism
were too narrow, focusing only on self-criticism and ignoring interpersonal situations in which
perfectionistic standards might be activated. They argued that perfectionism consists of three
dimensions: (a) the tendency to set exacting standards for oneself as well as to evaluate one's
own behavior stringently (self-oriented perfectionism), (b) the tendency to have unrealistically
high standards for the behavior of signi®cant others (other-oriented perfectionism) and (c) the
tendency to believe both that signi®cant others have unrealistically high standards for oneself,
and that they engage in stringent evaluation of one's behavior (socially prescribed
perfectionism). Hewitt and Flett (1991a) also titled their scale the multidimensional
perfectionism scale (MPS-H). Although there is overlap in the constructs measured by the
MPS-F and MPS-H (e.g. the socially prescribed perfectionism scale from the MPS-H is
correlated with the parental criticism and parental expectations scales on the MPS-F, the
dimensions from the two measures do not overlap entirely (Frost et al., 1993).
Perfectionism, as measured by these scales, is related to general symptoms of anxiety in
nonclinical samples (Minarik and Ahrens, 1996), mixed groups of psychiatric patients (Hewitt
and Flett, 1993), and samples of individuals with depression and anxiety disorders (Hewitt and
Flett, 1991b). Furthermore, perfectionism is implicated in the development and maintenance of
speci®c anxiety disorders. For instance, cognitive theories of obsessive±compulsive disorder
(OCD) have suggested that perfectionistic thinking contributes to certain types of obsessions
(e.g. doubts about whether a task was completed correctly) and compulsive activity (e.g.
 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154 1145

washing until it feels `just right') (McFall and Wollersheim, 1979; Obsessive Compulsive
Cognitions Working Group, 1997). This hypothesis is supported in a study by Frost and
Steketee (1997), in which perfectionism (as measured by the MPS-F) was compared across
clinical samples of individuals with OCD and panic disorder and a sample of nonanxious
volunteers. Patients with OCD had higher overall scores on the MPS-F than did nonanxious
volunteers, which was accounted for by elevations on the `concern over mistakes' and `doubts
about actions' subscales. Additionally, the OCD sample had higher `doubts about actions'
scores than did a sample of patients with panic disorder. Interestingly, although psychological
models of panic disorder have not included perfectionistic beliefs as a relevant precipitating
factor (McNally, 1990; Antony et al., 1992; Antony and Barlow, 1996), those with panic
disorder also had higher scores on the `concern over mistakes', `parental criticism', and total
perfectionism scores than the nonanxious controls.
At least two additional studies support the relationship between perfectionism and OCD
symptomatology. RheÂaume et al. (1995) found that scores on each of the MPS-F subscales
except organization were signi®cantly correlated with scores on the Padua inventory, a measure
of OCD severity. In a study by Frost and Shows (1993), concern over mistakes and doubts
about actions were signi®cantly correlated with compulsive indecisiveness (a symptom often
reported by people with OCD) in undergraduate students.
Perfectionistic beliefs are also considered important to the development and maintenance of
social phobia. In their comprehensive model of social phobia, Heimberg et al. (1995) described
three types of beliefs held by individuals with social phobia: (a) social situations are potentially
dangerous because they may lead to humiliation, (b) meeting a very high standard of social
performance is the only way to avoid or prevent humiliation in social situations and (c) this
standard is never met. Juster et al. (1996) indeed found that compared to nonanxious controls,
individuals with social phobia had higher scores on the `concern over mistakes', `doubts about
actions' and `parental criticism' scales of the MPS-F.
At this time, then, speci®c dimensions of perfectionism have been identi®ed, and existing
models of mood and anxiety disorders have implicated particular types of perfectionistic beliefs
in their development and maintenance. Furthermore, empirical support for the role of various
types of perfectionistic beliefs in these disorders has been obtained. Although these data may
help to forward our understanding of the mechanisms and treatment of psychopathology, more
research investigating dimensions of perfectionism and their relationship to various disorders is
necessary. For instance, to date, no published studies have compared individuals with a broad
range of anxiety disorders to nonclinical volunteers, nor have any studies directly compared
individuals with OCD to those with social phobia. This comparison is of interest because of
the anxiety disorders, these two have been hypothesized to be the most closely linked to
perfectionism. Finally, of the studies that have examined the role of perfectionism in various
anxiety disorders, only the MPS-F dimensions have been studied. No published study has yet
examined the Hewitt and Flett (1991a) dimensions in patients with particular anxiety disorders.
The purpose of the present paper was to compare dimensions of perfectionism across anxiety
disorders. Speci®cally, patients with social phobia, OCD, panic disorder and speci®c phobia,
and a sample of nonanxious volunteers were compared on the six dimensions of perfectionism
identi®ed by Frost et al. (1990) and the three dimensions of perfectionism identi®ed by Hewitt
and Flett (1991a). All patient groups were diagnosed using a structured interview based on
1146 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

DSM-IV criteria. Based on existing models for speci®c anxiety disorders, and current empirical
®ndings, it was hypothesized that, relative to other groups, patients with OCD and social
phobia would show higher levels of perfectionism on the MPS-F and MPS-H scales. It was
expected that the social phobia group would have higher scores on the `concern over mistakes',
`doubts about action', and `parental criticism' scales than the panic disorder, speci®c phobia
and nonclinical groups. We also expected that social phobia would be associated with higher
levels of self-oriented and socially prescribed perfectionism than other clinical and nonclinical
groups. Furthermore, we expected that OCD would be associated with higher scores on the
`doubts about action' than other anxiety disorders. Finally, panic disorder patients were
expected to report levels of perfectionism intermediate between patients with social phobia or
OCD on the one hand, and individuals with speci®c phobias or no anxiety disorder on the
other hand. All procedures in the present study were approved by the Ethics Review
Committee in the Department of Psychiatry at the University of Toronto.

2. Method

2.1. Participants

Participants were patients diagnosed with panic disorder with or without agoraphobia (PD;
n = 44), OCD (n = 45), social phobia (SOC; n = 70) or speci®c phobia (SPC; n = 15). Other
diagnostic groups (e.g. generalized anxiety disorder, posttraumatic stress disorder) were not
included due to small numbers of these patients in our clinic sample. In addition, a group of
nonclinical volunteers (NCV; n = 49) served as a comparison group. To be included in the
present study, participants in the anxiety disorders groups had to have a principal DSM-IV
diagnosis from one the four diagnostic groups mentioned above. In addition, patients who met
criteria for more than one of these diagnostic groups were excluded from the present study. All
participants were between the ages of 18 and 65. Patients with current diagnoses of substance
abuse/dependence, psychotic disorder or bipolar disorder were excluded from our clinic
database and therefore not included in this study.
Participants from the four anxiety disorders groups were all referred for an evaluation in the
Anxiety Disorders Clinic at the Clarke Institute of Psychiatry and completed this study as part
of their evaluation. Each individual was interviewed using the structured clinical interview for
DSM-IV (SCID-IV; First et al., 1996). In addition, a second clinical interview was conducted
by a sta€ psychiatrist. For cases in which the two interviewers disagreed, a diagnosis was
reached by consensus of the two interviewers.
Participants in the nonclinical comparison group were recruited by advertisements posted in
the community, seeking individuals without a history of mental health problems. Participants
in this group received a telephone interview based on the screening questions from the SCID-
IV to ensure that they did not have a history including any of the major forms of
psychopathology. Individuals for whom this telephone interview was inconclusive (e.g. for
which there was some indication of a possible problem) were excluded from the study.
Participants in the nonclinical comparison group were paid for their participation in this study.
The demographic characteristics of each group are summarized in Table 1.
 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154 1147

Table 1
Demographic characteristics of study participants
PD OCD SOC SPC NCV
Female (%) 54.8 44.4 38.6 80.0 61.2
Mean age (yr) 36.05 35.20 36.18 33.40 28.41

Marital status
Married or cohabiting (%) 53.5 35.5 35.7 33.3 22.4
Never married (%) 37.2 60.0 57.1 53.4 75.5
Divorced, separated, widowed (%) 9.4 4.4 7.1 13.3 0.0

Ethnicity
Caucasian (%) 87.5 86.7 87.7 92.9 51.0
East Asian (%) 5.0 2.2 7.7 7.1 26.5
Other (%) 7.5 11.1 4.6 0.0 22.4

Religion
Roman Catholic (%) 26.2 30.2 31.7 35.7 26.5
Protestant (%) 31.0 25.6 33.3 21.4 22.4
Jewish (%) 19.0 23.3 5.0 21.4 0.0
Other (%) 23.8 21.0 25.4 21.4 51.1

Education
Who did not complete high school (%) 13.7 9.1 7.1 13.4 0.0
Completed high school (%) 50.0 43.2 42.8 40.0 42.9
Completed college (%) 29.6 36.4 38.6 26.7 47.0
Completed graduate studies (%) 4.5 11.4 11.4 20.0 10.2

Annual family income (Canadian $)

Earning less than $20,000 (%) 17.5 32.5 22.7 16.7 26.1
Earning $20,000±$60,000 (%) 51.5 35.0 42.5 41.7 43.4
Earning $60,000 or more (%) 30.0 32.5 34.9 41.7 30.4

PD is panic disorder with or without agoraphobia, OCD obsessive compulsive disorder, SOC social phobia, SPC
speci®c phobia and NCV nonclinical volunteers.

2.2. Measures

2.2.1. Multidimensional perfectionism scale (MPS-F; Frost et al., 1990)

This 35-item questionnaire generates an overall perfectionism score as well as scores for six
subscales that re¯ect speci®c domains of perfectionism: concern over mistakes (CM), doubts
about actions (DA), personal standards (PS), parental expectations (PE), parental criticism
(PC) and organization (OR). The total perfectionism score is the sum of all subscales except
OR, which tends not to correlate highly with the other subscales or with total perfectionism
(Frost et al., 1990).
According to Frost et al. (1990), the CM subscale re¯ects a tendency to be overly self-critical and
self-evaluative and is central to the construct of perfectionism. Sample items from this scale include
``If I fail at work/school, I am a failure as a person'' and ``I hate being less than best at things''. This
1148 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

subscale accounts for the largest portion of the variance when the scale is subjected to factor
analysis. The DA subscale re¯ects a tendency to be uncertain about the correctness of one's
decisions or behaviors. Sample items include ``I usually have doubts about the simple everyday
things I do'' and ``It takes me a long time to do something right''. The PS subscale re¯ects a
tendency to set high standards for performance as re¯ected in the sample items, ``I set higher goals
than most people'' and ``I am very good at focusing my e€orts on attaining a goal''. Because
growing up with parents who are highly critical or have high expectations is believed to contribute
to perfectionism, the MPS-F includes subscales to measure parental expectations and parental
criticism. Sample items from the PE and PC subscales include ``My parents set very high standards
for me'' and ``As a child, I was punished for doing things less than perfectly'', respectively. Finally,
the OR scale measures a tendency to be overly orderly, organized, and tidy, as re¯ected by the item
``organization is very important to me''.
With respect to its psychometric properties, the MPS-F has good internal consistency and
appears to be a reliable and valid measure of perfectionism (Frost et al., 1990; Frost et al., 1993).

2.2.2. Multidimensional perfectionism scale (MPS-H; Hewitt and Flett, 1991a)

This 45-item questionnaire generates scores based on three subscales, each with 15 items.
Self-oriented perfectionism (SO) re¯ects a tendency to be overly perfectionistic with oneself, as
re¯ected in the item, ``I demand nothing less than perfection from myself''. The second
subscale, other-oriented perfectionism (OO) re¯ects a tendency to expect perfection from other
people, as re¯ected by the sample item, ``The people who matter to me should never let me
down''. The third subscale, called socially prescribed perfectionism (SP), is a measure of a
person's beliefs regarding other people's expectations of him or her. A sample item is ``Those
around me readily accept that I can make mistakes too''. This questionnaire does not have a
total perfectionism score. Like the Frost et al. (1990) scale, this measure is a reliable and valid
measure of perfectionism (Hewitt et al., 1991).

2.2.3. Beck depression inventory (BDI)

The BDI is a widely used measure of depressive symptomology, created in 1961 (Beck et al.,
1961), and revised in 1979 (Beck et al., 1979) to its current form. It is a 21-item scale derived
from clinical observations and appears to be a reliable and valid measure of depression (Beck
et al., 1988).

3. Results

3.1. Preliminary analyses

Analysis of variance (ANOVA) on age revealed a signi®cant e€ect of diagnostic group, F(4,
211) = 5.16, p < 0.01. Post hoc tests using the Tukey HSD revealed that the nonclinical group
had a signi®cantly lower mean age than did the PD, OCD and SOC groups. Chi-squared tests
of demographic characteristics revealed that all groups tended to have more members who
were single than married, cohabitating, separated or divorced (w2=32.36, p < 0.04), and that
whereas the NCV and SPC groups had more women then men, the PD, OCD and SOC groups
had more men then women (w2=12.45, p < 0.05).
 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154 1149

Although there was no theoretical reason to assume that marital status might be an important
mediating factor in perfectionism scores, it was thought that age and sex could potentially
in¯uence perfectionism scores, either on their own or in an additive e€ect with diagnostic group.
In order to determine whether or not these di€erences needed to be taken into account in the
main analysis of group di€erences in the di€erent perfectionism measures, a series of
multivariate analyses of variance was performed on the Frost et al. MPS subscales and the
Hewitt and Flett MPS subscales separately, entering age and diagnostic group, and sex and
diagnostic group, respectively, as predictors. No signi®cant main e€ect of age on either set of
subscales was detected, nor was there a signi®cant age by diagnostic group interaction.
Similarly, there was no signi®cant main e€ect of sex on either measures' subscales, nor was there
a signi®cant interaction of sex with diagnostic group. Thus, it was concluded that pre-existing
di€erences in sex and age across the diagnostic groups did not in¯uence perfectionism scores on
either measure, and that analyses could proceed without taking these factors into account.
To assess whether group di€erences in perfectionism could be related to di€erences in
depression, groups were compared with respect to the BDI using a one-way ANOVA. There
was a signi®cant e€ect of group, F(4, 217) = 20.35, p < 0.001. Post hoc tests using Tukey's
HSD revealed that the PD (M = 15.80, S.D. = 9.72), OCD (M = 15.78, S.D. = 9.64) and
SOC (M = 16.70, S.D. = 10.73) groups had signi®cantly higher mean scores on the BDI than
did the NCV (M = 3.29, S.D. = 3.65) group, and that the SOC group had a higher mean than
did the SPC (M = 9.67, S.D. = 5.97) group. Thus, there were some di€erences in BDI scores
across groups. However, because depression and anxiety are highly correlated, it was thought
that statistical control for depression could result in the loss of `interesting' variance, and so it
was decided that analyses should proceed without controlling for pre-existing group di€erences
in depression.

3.2. Analyses of perfectionism across groups

A single multivariate analysis of variance (MANOVA) was conducted on the six subscales of
the MPS-F with diagnostic group entered as the independent variable. Means and standard
deviations on each subscale across each group are presented in Table 2. There was a signi®cant
main e€ect of diagnostic group, F(24, 840) = 6.67, p < 0.001. Two discriminant functions
emerged, accounting for 64.82 and 32.76% of the total variance respectively. In order to
understand speci®c di€erences across groups, a series of post hoc analyses was performed
comparing each group to the other. These contrasts followed the principle of Fisher's protected
tests, in which post hoc comparisons are made using the same level of alpha as was used to
examine the main e€ect (see Cohen and Cohen, 1983).
First, the PD group was compared to the other groups. The PD group was found to di€er
from the NCV group, F(6, 207) = 5.21, p < 0.001. The loading matrix of correlations between
predictors and the discriminant function suggests that the best predictors of group di€erences
were the concern over mistakes and doubt about action subscales. According to the univariate
Fs, the PD group had higher scores on both measures. Items from these two scales together can
be interpreted as a general tendency to fear possible negative social consequences from making
mistakes. The PD group also di€ered from the OCD group, F(6, 207) = 10.65, p < 0.001. The
discriminant function loadings suggested that the doubt over action scales best distinguished the
1150 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

Table 2
Means and (standard deviations) across groups on the various subscales of the Frost et al. (1990) multidimensional
perfectionism scale
Subscale PD OCD SOC SPC NCV
a a,b c b,d
CM 24.14 (8.21) 21.53 (7.87) 27.48 (8.35) 18.21 (5.66) 17.43d (5.25)
DA 10.00a (3.33) 14.44b (4.34) 13.03b (4.03) 8.14a,c (2.85) 7.74c (3.15)
PS 22.67a (4.34) 21.93a (6.44) 22.42a (5.89) 21.86a (6.32) 22.74a (6.07)
PE 14.14a (5.52) 11.82b (4.99) 13.49a,b (5.40) 13.50a,b (6.31) 13.63a,b (4.31)
PC 9.71a,b (4.55) 9.29b (4.39) 11.33a (4.54) 9.71a,b (5.69) 8.69b (3.65)
OR 23.69a (4.35) 23.33a (5.30) 22.03a (5.19) 22.57a (5.26) 22.57a (4.38)

PD is panic disorder with or without agoraphobia, OCD obsessive compulsive disorder, SOC social phobia, SPC
speci®c phobia and NCV nonclinical volunteers. CM is concern over mistakes, DA doubts about actions,
PS = personal standards, PE parental expectations, PC parental criticism, OR organization and Total is the sum of
all subscales except OR.
a,b
Means sharing superscripts across rows do not di€er signi®cantly at p < 0.05, according to univariate F tests.

groups, and interpretation of the univariate Fs revealed that the PD group had lower doubt
over action and higher parental expectation scores. This latter subscale, then, shares variance
with the other subscales and did not contribute signi®cantly to group di€erences when all were
entered as predictors. The PD group was also found to di€er from the SOC group, F(6,
207) = 3.95, p < 0.01. The discriminant function loadings revealed that the concern over
mistakes and doubt about action subscales best predicted group di€erences, and this was
re¯ected in the univariate Fs, where the PD group had lower scores on both scales. The PD
group thus exhibited a lower overall tendency to fear the social consequences of mistakes than
did the SOC group. The PD group did not di€er signi®cantly from the SPC group on the initial
analysis, although univariate Fs revealed that the PD group scored higher on the CM scale. In
sum, the PD group was distinguished from the NCV by a greater fear of mistakes, from the
OCD group by less doubt over action and higher parental expectations, from the SOC group by
less concern over mistakes and doubts about actions, and from the SPC gap by higher concerns
over mistakes.
The OCD group was then compared to the other groups. The OCD group's pro®le of scores
on the six perfectionism subscales was found to di€er from that of the NCV group, F(6,
207) = 15.23, p < 0.001. Discriminant function loadings suggested that the concern over
mistakes subscale was the best predictor of group di€erences, and the univariate Fs revealed
that the OCD group had lower scores on the concern over mistakes and parental criticism
subscales. Examination of the univariate Fs revealed that the OCD group had signi®cantly
higher scores on this measure and the concern over mistakes subscale, although as evidenced
by the discriminant function loadings, this subscale was not important to distinguishing groups
when included with the other ®ve subscales. The OCD group also di€ered from the SOC
group, F(6, 207) = 8.58, p < 0.0001. Discriminant function loadings suggested that the concern
over mistakes subscale was the best predictor of group di€erences, although the univariate Fs
revealed that the OCD group had lower scores on the concern over mistakes and parental
criticism subscales but higher scores on the doubt over action subscale. Again, the latter two
subscales lost their ability to distinguish between groups when all subscales were used as
 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154 1151

predictors. The OCD group also di€ered from the SPC group, F(6, 207) = 6.11, p < 0.001.
The only subscale to load on the discriminant function was doubt about action, and
according to the univariate Fs, this was also the only subscale to show signi®cant mean
di€erences, with the OCD group having higher mean scores. Thus, the OCD group was
distinguished from the NCV, PD and SPC groups by stronger doubts about having
performed activities `correctly', and from the SOC group by having less tendency to interpret
mistakes as catastrophic.
Group di€erences between the SOC and NCV groups were examined next. There was a
signi®cant group di€erence, F(6, 207) = 17.11, p < 0.001. According to the discriminant
function loadings, the concern over mistakes, doubt over actions and parental criticism
subscales best predicted group di€erences. This was also re¯ected by the univariate Fs, which
revealed that the SOC group had higher scores on all three subscales. Together, the items in
these three subscales suggest a fear of the consequences of mistakes in combination with a
feeling of external pressure that one's performance is not good enough. The SOC and SPC
groups were examined next. The pattern of scores on the six subscales was di€erent across the
two groups, F(6, 207) = 5.16, p < 0.001. Concern over mistakes and doubt about actions were
found to be the best predictors of group di€erences, according to the discriminant function
loadings, and this was re¯ected in the univariate Fs, where the SOC group was found to have
higher scores on both subscales. Finally, di€erences between the SPC and NCV groups were
examined. No di€erence in the pattern of scores across groups was observed.
The Hewitt and Flett MPS scales were examined across groups next. Means and standard
deviations on each subscale across each group are presented in Table 3. A MANOVA was
performed on the three H-MPS subscales with diagnostic group entered as the independent
variable. There was a signi®cant main e€ect of group, F(12, 645) = 3.71, p < 0.001.
Interpretation of the discriminant function loadings revealed that the self-oriented and socially
prescribed scales were the most important in distinguishing groups, and this was re¯ected in
the univariate Fs, which revealed signi®cant group di€erences on these two subscales. In order
to examine group di€erences, a series of post hoc analyses was conducted in the same manner
as described earlier. The PD group was ®rst examined in relation to each of the other groups.
The PD group's pattern of scores on the H-MPS scales di€ered from that of the NCV group's,
F(3, 213) = 5.85, p < 0.01. The self-oriented and socially prescribed scales loaded signi®cantly
on the discriminant function, and univariate Fs revealed that the PD group had higher scores
on the latter scale. This suggested that the PD group was distinguished from the NCV group
by a general tendency to believe that others hold them to a high standard. The PD group was
also found to di€er from the SPC group, F(3, 213) = 3.22, p < 0.05. Both the discriminant
function loading and the univariate Fs revealed that the PD group had higher scores on the
Socially Prescribed perfectionism scale, revealing a stronger tendency to believe others have
unrealistically high standards for them. The PD group was not found to di€er from either the
OCD or the SOC groups.
The OCD group di€ered from the NCV group (F(3, 213) = 3.56, p < 0.05. Interpretation of
the discriminant function loadings suggested that the OCD group was distinguished from the
nonclinical group by scores on the self-oriented and socially prescribed perfectionism scales.
The univariate Fs suggested that the OCD group had higher scores on the latter measure, but
1152 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

Table 3
Means and (standard deviations) across groups on the various subscales of the Hewitt and Flett (1991a)
multidimensional perfectionism scale
Subscale PD OCD SOC SPC NCV
a a a a
SO 71.61 (13.67) 69.48 (18.87) 70.53 (17.59) 63.13 (21.89) 64.08a (15.21)
OO 60.32a (12.34) 55.93a (13.87) 56.34a (14.93) 56.87a (15.94) 57.06a (11.04)
SP 55.71a,b (13.00) 51.30a,c (15.15) 57.75b (15.55) 43.47c,d (13.11) 44.12d (11.24)

PD is panic disorder with or without agoraphobia, OCD obsessive compulsive disorder, SOC social phobia, SPC
speci®c phobia and NCV nonclinical volunteers. SO is selforiented perfectionism, OO other-oriented perfectionism
and SP socially prescribed perfectionism.
a,b
Means sharing superscripts across rows do not di€er signi®cantly at p < 0.05, according to univariate F tests.

no signi®cant di€erence on the former, although scores were also higher for the OCD group.
The OCD group did not di€er from the SPC or SOC groups.
The SOC group was found to di€er from the NCV group, F(3, 213) = 12.13, p < 0.001.
Once again, according to the discriminant function loadings, the group di€erence was best
predicted by the self-oriented and socially prescribed measures of perfectionism. The univariate
Fs revealed that the SOC group had higher scores on the latter scale. The SOC group also
di€ered from the SPC group, F(3, 213) = 5.73, p < 0.01. The discriminant function loadings
again suggested that the di€erence was best predicted by the self-oriented and socially
prescribed measures of perfectionism. The univariate Fs revealed that the SOC group had
higher scores on the latter measure. The SPC group was found to be no di€erent from the
NCV group.

4. Discussion

The pattern of ®ndings on the MPS-F for social phobic patients was similar to results from
other studies, in which individuals with social phobia had higher scores than nonanxious
volunteers on the CM, DA, and PC scales (Juster et al., 1996). In addition, the pattern of
results on the MPS-F for individuals with OCD also replicated previous ®ndings (Frost and
Steketee, 1997) with OCD being associated with higher scores relative to nonpatient controls
for CM and DA, and higher scores than PD on the DA scale. The heightened doubts about
actions seen in OCD may be related to the tendency for OCD to be associated with particular
types of rituals, such as checking.
On the MPS-H measure, the original hypotheses were not con®rmed, in that groups did
not di€er with respect to self-oriented perfectionism. However, as hypothesized, participants
in the PD, OCD and social phobia groups did report more socially prescribed perfectionism
compared to individuals in the nonclinical volunteer groups. In addition, those in the PD
and social phobia groups scored higher on socially prescribed perfectionism than those in the
speci®c phobia group. The relatively high levels of socially prescribed perfectionism in the
PD, social phobia and OCD groups suggests that individuals with these anxiety disorders
 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154 1153

perceive others as having high expectations. It is somewhat surprising that patients with
social phobia did not report even higher levels of socially prescribed perfectionism (compared
to those with OCD and PD), given that a hallmark of this disorder is concern about
criticism from others.
Overall, the ®ndings from this study are consistent with the hypotheses that social phobia
(Heimberg et al., 1995) and OCD (Obsessive Compulsive Cognitions Working Group, 1997)
are each associated with perfectionistic thinking. However, the fact that individuals in the PD
group reported elevated concern over mistakes and doubts about actions, relative to
nonanxious individuals, suggests that perfectionism may be a feature of PD as well. Perhaps all
the anxiety disorders (speci®c phobia may be an exception) are associated with a general
underlying need to control events in order to prevent unexpected dangers. It is possible that
this need for control is related to perfectionistic thinking (e.g. the belief that making mistakes
may lead to a loss of control over particular outcomes).
The fact that di€erences between groups were stronger for the MPS-F measure than the
MPS-H measure is consistent with the possibility that the dimensions of the MPS-F are more
directly related to the psychopathology underlying the anxiety disorders. However, because this
is the ®rst study to examine di€erences among speci®c anxiety disorders using the MPS-H,
these results remain to be replicated.
Additionally, it is important that investigators begin to examine behavioral correlates of
perfectionism in these groups. Although one study (Frost et al., 1997) has examined the
relationship between self-monitoring of mistakes and concern over mistakes in a nonclinical
sample, no studies have yet examined the relationship between domains of perfectionism and
speci®c behaviors in clinical groups. For example, it would be helpful to know whether
increased doubts about actions are associated with more frequent checking in OCD. In
addition, we have yet to determine whether concern about mistakes predicts any behaviors of
interest in social phobic individuals (e.g. a tendency to avoid taking social risks).
A further area for future research is the relationship between scale elevations on these
perfectionism measures and cognitive features of anxiety disorders and related conditions. For
example, it is possible that scores on particular subscales may be elevated for di€erent reasons
in di€erent clinical groups. Concern over mistakes and doubts about actions may have been
elevated in OCD because of anxiety over possible danger that might occur if actions were not
completed in particular ways, rather than a desire to things perfectly or to meet strict
standards. In contrast, elevated concern over mistakes in people with social phobia may have
been more related to concern about meeting particular standards.
The present ®ndings have important implications for the treatment of anxiety disorders.
Because speci®c domains of perfectionism are associated with particular anxiety disorders, it
might be important for clinicians to address the relevant types of perfectionistic thoughts in
treatment using cognitive and behavioral strategies. Currently, there are no published studies in
the anxiety disorders that have investigated the e€ect of perfectionism on treatment outcome,
nor the e€ect of treatment on perfectionistic thinking. Ultimately, the importance of the
relationship between perfectionism and psychopathology will be determined by the relevance of
perfectionism to the etiology and treatment of particular disorders.
1154 M.M. Antony et al. / Behaviour Research and Therapy 36 (1998) 1143±1154

References

Antony, M. M., & Barlow, D. H. (1996). Emotion theory as a framework for explaining panic attacks and panic disorder. In R. M.
Rapee (Ed.), Current controversies in the anxiety disorders (pp. 55±76). New York, NY: Guilford Publications.
Antony, M. M., Brown, T. A., & Barlow, D. H. (1992). Current perspectives on panic and panic disorder. Current Directions in
Psychological Science, 1, 79±82.
Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression: A treatment manual. New York: Guilford.
Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the Beck depression inventory: twenty-®ve years of
evaluation. Clinical Psychology Review, 8, 77±100.
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory measuring depression. Archives of General
Psychiatry, 4, 561±571.
Burns, D. D. (1980). The perfectionist's script for self-defeat. Psychology Today, November, 34±51.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences. Hillsdale: Lawrence Erlbaum.
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1996). Structured clinical interview for DSM-IV axis I disorders Ð
patient edition (SCID-I/P, Version 2.0). New York: Biometrics Research Department, New York State Psychiatric Institute.
Frost, R. O., & Shows, D. L. (1993). The nature and measurement of compulsive indecisiveness. Behaviour Research and Therapy, 31,
683±692.
Frost, R. O., & Steketee, G. (1997). Perfectionism in obsessive±compulsive disorder patients. Behaviour Research and Therapy, 35, 291±
296.
Frost, R. O., Heimberg, R. G., Holt, C. S., Mattia, J. L., & Neubauer, A. L. (1993). A comparison of two measures of perfectionism.
Personality and Individual Di€erences, 14, 119±126.
Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimensions of perfectionism. Cognitive Therapy and Research, 14,
449±468.
Frost, R. O., Trepanier, K. L., Brown, E. J., Heimberg, R. G., Juster, H. R., Makris, G. S., & Leung, A. W. (1997). Self-monitoring
of mistakes among subjects high and low in perfectionistic concern over mistakes. Cognitive Therapy and Research, 21, 209±222.
Hamacheck, D. E. (1978). Psychodynamics of normal and neurotic perfectionism. Psychology, 15, 27±33.
Heimberg, R. G., Juster, H. R., Hope, D. A., & Mattia, J. I. (1995). Cognitive-behavioral group treatment for social phobia: descrip-
tion, case presentation, and empirical support. In M. Stein (Ed.), Social phobia: clinical and research perspectives (pp. 293±321).
Washington, DC: American Psychiatric Press.
Hewitt, P. L., & Flett, G. L. (1991a). Perfectionism in the self and social contexts: conceptualization, assessment and association with
psychopathology. Journal of Personality and Social Psychology, 60, 456±470.
Hewitt, P. L., & Flett, G. L. (1991b). Dimensions of perfectionism in unipolar depression. Journal of Abnormal Psychology, 100, 98±
101.
Hewitt, P. L., & Flett, G. L. (1993). Dimensions of perfectionism, daily stress and depression: a test of a speci®c vulnerability hypoth-
esis. Journal of Abnormal Psychology, 102, 58±65.
Hewitt, P. L., Flett, G. L., Turnbull-Donovan, W., & Mikail, S. F. (1991). The multidimensional perfectionism scale: reliability,
validity, and psychometric properties in psychiatric samples. Psychological Assessment: A Journal of Consulting and Clinical
Psychology, 3, 464±468.
Juster, H. R., Heimberg, R. G., Frost, R. O., Holt, C. S., Mattia, J. I., & Faccenda, K. (1996). Social phobia and perfectionism.
Personality and Individual Di€erences, 21, 403±410.
McFall, M. E., & Wollersheim, J. P. (1979). Obsessive±compulsive neurosis: a cognitive-behavioral formulation and approach to
treatment. Cognitive Therapy and Research, 3, 333±348.
McNally, R. J. (1990). Psychological approaches to panic disorder: a review. Psychological Bulletin, 108, 403±419.
Minarik, M. L., & Ahrens, A. H. (1996). Relations of eating behavior and symptoms of depression and anxiety to the dimensions of
perfectionism among undergraduate women. Cognitive Therapy and Research, 20, 155±169.
Obsessive Compulsive Cognitions Working Group (1997). Cognitive assessment of obsessive±compulsive disorder. Behaviour Research
and Therapy, 35, 667±681.
Pacht, A. R. (1984). Re¯ections on perfectionism. American Psychologist, 39, 386±390.
RheÂaume, J., Freeston, M. H., Dugas, M. J., Letarte, H., & Ladouceur, R. (1995). Perfectionism, responsibility and obsessive±
compulsive symptoms. Behaviour Research and Therapy, 33, 785±794.