DEPRESSION AND ANXIETY 23:297–303 (2006)

Research Article
BRIEF FEAR OF NEGATIVE EVALUATION
SCALE—REVISED
R. Nicholas Carleton, M.A.,1 Donald R. McCreary, Ph.D.,2 Peter J. Norton, Ph.D.,3 and
Gordon J. G. Asmundson, Ph.D.1

Rodebaugh et al. [2004: Psychol Assess 2:169–181] recently performed a

confirmatory factor analysis (CFA) on the Brief Fear of Negative Evaluation
scale (BFNE; Leary, 1983: Psychol Bull 9:371–375]. Their study resulted in the
emergence of a two-factor solution comprising straightforwardly worded items
and reverse-worded items. They concluded by recommending use of only the
straightforwardly worded items in the BFNE. Our intent in this study was to
evaluate this recommendation through replication and extension. Participants
included 385 undergraduates from the Universities of Regina and Houston, who
provided responses to a questionnaire battery including either the BFNE or a
revision utilizing straightforwardly worded versions of the reverse-worded items
(BFNE-II). A CFA of the BFNE, using the two-factor model proposed by
Rodebaugh et al., supported their conclusion that the reverse-worded items
comprise a separate, methodologically based factor. However, CFA of the BFNE-
II resulted in an acceptable unitary model that conforms to the theoretical basis
for the BFNE, without risking loss of sensitivity from item removal. Additional
analyses suggest use of the BFNE-II rather than a shortened form. Depression
and Anxiety 23:297–303, 2006. & 2006 Wiley-Liss, Inc.

Key words: social anxiety; FNE; BFNE; revised BFNE; BFNE-II

1998] and nonclinical [e.g., Carleton and Asmundson,

INTRODUCTION 2005; Rodebaugh et al., 2004; Stein et al., 2002]
The fear of negative evaluation is one of three populations. Two popular scales for measuring fear of
fundamental fears thought to give rise to the develop- negative evaluation are the Fear of Negative Evaluation
ment and expression of more general fears, anxiety, and
psychopathologies [Reiss and McNally, 1985]. Specifi- 1
Anxiety and Illness Behaviours Laboratory, University of
cally, the fear of negative evaluation refers to appre- Regina, Regina, Saskatchewan, Canada
hension and distress arising from concerns about being 2
Department of Psychology, York University, Toronto, Ontario,
judged disparagingly or hostilely by others. The other Canada
fundamental fears are anxiety sensitivity (i.e., the fear of 3
Department of Psychology, University of Houston, Houston,
anxiety symptoms or sensations, based on the belief Texas
that they will have harmful consequences) and injury/
Contract grant sponsor: Canadian Institute of Health Research
illness sensitivity (i.e., fears of injury, illness, and death).
Investigator’s Award (to G. J. G. Asmundson).
Taylor [1993] has demonstrated the factorially distinct
Correspondence to: Gordon J. G. Asmundson, Anxiety and
nature of these fundamental fears.
Widespread interest in the fear of negative evalua- Illness Behaviours Laboratory, University of Regina, Regina,
Saskatchewan, S4S 0A2.
tion, stemming in part from clinical imperative (i.e.,
E-mail: [email protected]
expressions of social anxiety and social phobia), has
stimulated development of several related measures Received for publication 28 April 2005; Revised 9 August 2005;
[e.g., Social Interaction Anxiety Scale and Social Accepted 27 September 2005
Phobia Scale; Mattick and Clarke, 1998] and consider- DOI 10.1002/da.20142
able research in clinical [e.g., social anxiety disorder; Published online 10 May 2006 in Wiley InterScience (www.
Brown et al., 1997; Gore et al., 2002; Safren et al., interscience.wiley.com).

r 2006 Wiley-Liss, Inc.

298 Carleton et al.

scale [FNE; Watson and Friend, 1969] and the shorter and discriminant validity. This suggestion is in line
Brief Fear of Negative Evaluation scale [BFNE; Leary, with previous work in which reverse-worded items have
1983]. The FNE comprises 30 true–false items, of formed separate factors [see Marsh, 1996; Spector
which 17 are straightforwardly worded (e.g., ‘‘I am et al., 1997].
afraid people will find fault with me’’) and 13 are The psychometric findings of Rodebaugh and
reverse-worded (e.g., ‘‘I am unconcerned even if I know colleagues [2004] were recently replicated in a study
people are forming an unfavorable impression of me’’). by Weeks and colleagues [2005] as a part of a
Persons with high scores on the FNE are more likely comparative investigation of the BFNE in persons
than persons scoring low on the FNE to catastrophize in with and without social anxiety disorder. They
reaction to even mild negative social events [Stopa and confirmed the two-factor structure of the BFNE as
Clark, 2001]. Although the FNE is highly demonstrative comprising straightforwardly worded and reverse-
of fear of negative evaluation, its length and dichoto- worded items. They went on to chorus the concerns
mous response options make the BFNE an appealing of Rodebaugh and colleagues [2004].
alternative [Leary, 1983; Rodebaugh et al., 2004]. The work of Rodebaugh et al. [2004] and Weeks
The BFNE comprises 12 five-point Likert scale et al. [2005] suggests three things. First, the straight-
items—8 straightforwardly worded and 4 reverse- forwardly worded BFNE items provide a psychome-
worded, 11 of which are verbatim from the FNE. trically sound index of fear of negative evaluation.
The BFNE has been used as a successful short form of Second, the straightforwardly worded BFNE items
the original FNE scale [Taylor, 1993]. Although the correlate highly with other measures of social anxiety
BFNE has correlated with the Social Avoidance and and are sensitive to treatment-based changes in latent
Distress Scale [SADS, Watson and Friend, 1969], it fears of negative evaluation. Third, past use of the
more accurately depicts fear of negative evaluation than reverse-scored items may have resulted in an under-
it does social anxiety [Miller, 1995]. The factor estimation of the effect of fear of negative evaluation.
structure of the BFNE has been assumed to be unitary Given the widespread use of the BFNE in assessment
in previous studies [e.g., Leary, 1983; Stopa and Clark, packages designed for both clinical assessment and
2001; Turner et al., 1987] and, until recently, it has not general social anxiety research, the revisions suggested
been subjected to factor-analytic investigation. by these two studies [Rodebaugh et al., 2004; Weeks
Rodebaugh and colleagues [2004] performed a et al., 2005] warrant replication and further scrutiny
confirmatory factor analysis (CFA) on both the FNE prior to widespread adoption.
and the BFNE. The sample comprised 1,049 under- Our current study had two purposes. First, we sought
graduates from two U.S. universities, approximately to replicate the findings reported in previous CFAs of
63% of whom were women. Their analysis resulted in the BFNE [Rodebaugh et al., 2004; Weeks et al., 2005].
two primary conclusions. First, they concluded that the Second, assuming successful replication, our intent was
BFNE conveyed more information than the FNE due to determine empirically the best course of action for
to increased sensitivity from using Likert-style rather dealing with the reverse-worded items. There are three
than dichotomous response options. This is in line with possible course of action: The four items in question
earlier research indicating that although the FNE could be left as they are, but not be included in scoring;
discriminates between persons with and without the items could be discarded; or the four reverse-
anxiety disorders [Oei et al., 1991; Turner et al., worded items could be reworded to be straightforward.
1987], it fails to discriminate consistently between The determination will be made based on how well the
persons with social phobia and other anxiety disorders. four items correspond to the rest of the BFNE and
Second, they concluded that the assumed unitary factor whether they contribute uniquely and significantly to
structure of both the FNE and the BFNE may be the whole.
inappropriate when compared to a multidimensional
model. This latter conclusion was based on CFA. Their
evaluation of the unitary model of the BFNE yielded METHODS
relatively poor fit indices. A two-factor model, segre-
gating all reverse-scored items from the straightfor- PARTICIPANTS
wardly worded ones, resulted in acceptable fit indices Potential participants for this study included 650
that improved when the two factors were allowed to undergraduate students from the University of Hous-
correlate. A comparison of the unitary and two-factor ton and 254 undergraduate students from the Uni-
models indicated that the two-factor model produced a versity of Regina, all of whom consented to participate
significantly better fit than the one-factor model. This in a large, questionnaire-based study. Participants were
difference, in conjunction with the results from an item either tested en masse in a large classroom setting or
response theory analysis, led Rodebaugh et al. [2004] were allowed to complete questionnaires on their own
to suggest that the reverse-worded items are causing time and return them within 1 week. The questionnaire
confusion and erroneous responding. Their recom- battery from Houston included the original BFNE
mendation was to use only the straightforwardly (i.e., with the reverse-worded items intact), whereas
worded BFNE items to achieve greatest reliability the questionnaire battery from Regina included the
Depression and Anxiety DOI 10.1002/da
 Research Article: BFNE-II 299

TABLE 1. Original BFNE and proposed revisions Regina samples overall [t (1, 383) 5 1.34, P 5.182], nor
a difference between the samples with respect to gender
Item Number Item ratio [w2 (1, N 5 385) 5 0.751; P 5.386]. No significant
1 I worry about what other people will think of me
differences were found based on straightforwardly
even when I know it doesn’t make any difference. worded item total scores between the two samples
2 (original) I am unconcerned even if I know people are forming [F (1, 383) 5 1.072, P 5.301] or within the samples
an unfavorable impression of me. based on gender [Houston, F (1, 199) 5 0.035,
2 (revised) It bothers me when people form an unfavorable P 5.851; Regina, F (1, 182) 5 1.567, P 5.212].
impression of me.
3 I am frequently afraid of other people noticing my MEASURE
shortcomings.
4 (original) I rarely worry about what kind of impression I am The BFNE [Leary, 1983] is a short version of the
making on someone. FNE [Watson and Friend, 1969]. It comprises 12
4 (revised) I worry about what kind of impression I make on items, 8 straightforwardly worded and 4 reverse-
people. worded. Each item is rated on a 5-point Likert scale,
5 I am afraid that others will not approve of me. ranging from 0 (Not at all characteristic of me) to 4
7 (original) Other people’s opinions of me do not bother me. (Extremely characteristic of me). Leary [1983] indicated
7 (revised) I am concerned about other people’s opinions of me. that the psychometric properties of the BFNE were
8 When I am talking to someone, I worry about what
almost identical to those of the FNE, with the total
they may be thinking about me.
9 I am usually worried about what kind of impression
scores correlating at r 5 .96. The BFNE has demon-
I make. strated high internal consistency (a between .90 and
10 (original) If I know someone is judging me, it has little effect .91) and 4-week test–retest reliability (r 5 .75) in
on me. undergraduate samples [Leary, 1983; Miller, 1995].
10 (revised) If I know someone is judging me, it tends to For the purposes of this study, we used revised word-
bother me. ing first suggested by Taylor [1993] to create the
11 Sometimes I think I am too concerned with what BFNE-II. As described earlier, this involved reword-
other people think of me. ing the reverse-worded items to be straightforward
12 I often worry that I will say or do wrong things. (see Table 1).

RESULTS
BFNE-II (i.e., with the reverse-worded items reworded
Descriptive statistics for the Houston and Regina
to be straightforward). The original and revised items
sample responses to the BFNE and BFNE-II items are
are shown in Table 1. shown in Table 2. Indices of univariate skewness and
Because we were comparing two different versions of
kurtosis were included, because non-normality can
the questionnaire, each of which was given only to one
adversely influence multivariate data analyses such
sample, it was important that the two samples be
as CFA. None of these indices were excessively out
matched as closely as possible, so that if any differences
of range.
in the factor structure of the two questionnaires
emerged, it could be said they were a function of the
CONFIRMATORY FACTOR ANALYSIS
question items, not sample differences. Thus, the two
samples were matched based on age, sex, and total OF THE BFNE
scores from the eight straightforwardly worded items CFAs were conducted to assess the degree to which
shared by the BFNE and BFNE-II. the Houston data fit the single- and two-factor BFNE
The final participant sample from Houston included models proposed in the literature. These analyses were
201 undergraduate students, most of whom self- performed using the program EQS [Bentler, 2003], and
declared as Caucasian (59%) or Asian (30%) [64 men followed the procedures outlined by Byrne [1994]. Raw
ages 18–38 (M 5 21.14; SD 5 3.41) and 137 women data were used as input, along with a maximum
ages 18–36 (M 5 20.88; SD 5 2.46)], with no signifi- likelihood estimation procedure. Because the purpose
cant age difference between men and women [F (1, of this CFA was to test two competing models, no item
199) 5 0.370, P 5.544]. The final participant sample cross-loadings were allowed. Following recommenda-
from Regina included 184 undergraduate students [44 tions by Hu and Bentler [1999], model fit was
men ages 19–34 (M 5 22.11; SD 5 2.83) and 140 determined using several indices: (1) w2 (values should
women ages 18–37 (M 5 21.10; SD 5 2.66)], with the not be significant); (2) w2/degrees of freedom (df) ratio
men being slightly older than the women [F (1, (values should beo2.0), (3) comparative fit index (CFI;
182) 5 4.710, P 5.031]. We did not collect racial values should be close to .95), (4) root-mean-square
demographics for the Regina sample; however, based error of approximation (RMSEA; values should be
on the population demographics, they can be assumed close to .06), and (5) the standardized root-mean-
to be primarily Caucasian. There were no significant square residual (SRMR; values should be close to .08).
differences between the ages of the Houston and Emphasis is placed on the latter four fit indices,
Depression and Anxiety DOI 10.1002/da
300 Carleton et al.

TABLE 2. Descriptive statistics for items BFNE (N 5 201) and BFNE-II (N 5 184)

Corrected
Item–Total
M SD Skewness Kurtosis Correlations a-if-item-deleted

Item I II I II I II I II I II I II

1 2.69 2.31 1.08 1.06 0.26 0.51 0.57 0.56 0.77 0.76 0.87 0.95
2r 2.73 3.12 1.04 1.15 0.58 0.22 0.18 0.79 0.26 0.69 0.90 0.95
3 2.35 2.32 1.15 1.06 0.53 0.48 0.58 0.56 0.71 0.76 0.87 0.95
4r 2.77 3.09 1.04 0.94 0.64 0.03 0.23 0.18 0.30 0.8 0.89 0.94
5 2.25 2.43 1.12 1.08 0.50 0.33 0.71 0.71 0.70 0.79 0.87 0.94
6 2.17 2.74 1.15 1.04 0.59 0.03 0.72 0.81 0.74 0.80 0.87 0.94
7r 2.61 2.96 1.11 1.02 0.59 0.05 0.40 0.45 0.46 0.81 0.89 0.94
8 2.42 2.57 1.11 1.06 0.35 0.24 0.77 0.57 0.70 0.79 0.87 0.94
9 2.71 2.78 1.09 1.01 0.28 0.03 0.51 0.54 0.67 0.79 0.87 0.94
10r 2.80 3.04 1.12 1.06 0.77 0.13 0.17 0.49 0.45 0.71 0.89 0.95
11 2.66 2.94 1.30 1.01 0.24 0.06 1.05 0.60 0.70 0.74 0.87 0.95
12 2.56 2.47 1.22 1.03 0.31 0.35 0.90 0.49 0.65 0.70 0.88 0.95

Total 30.7 32.8 9.04 10.04 0.21 0.18 0.12 0.47 – – 0.89 0.95

Note. I, BFNE; II, BFNE-II; r, reverse-worded/revised questions.

Scale a.

TABLE 3. Houston and Regina sample CFA models indices

S-Bw2 df P S-B w2/df CFI-R RMSEA SRMR

Houston—12 items, 1 factor 203.60 54 o.001 3.77 .85 .12 .10

Houston—12 items, 2 factors 86.18 53 o.003 1.63 .98 .06 .05
Regina—12 items, 1 factor 124.29 54 o.001 2.30 .95 .08 .04
Regina—12 items, 2 factors 111.86 53 o.001 2.11 .96 .08 .04

S-Bw2, Satorra–Bentler w2, CFI-R, Robust Comparative Fit Index.

The Houston sample used the BFNE; the Regina sample, the BFNE-II.

because w2 statistics are inflated in larger samples The factor loadings for the one- and two-factor models
[Hu and Bentler, 1999]. are shown in Table 4.
An initial analysis of the Houston data showed the We tested the same two models in the Regina sample
presence of multivariate kurtosis (Mardia’s Normalized using the BFNE-II, where all items were straightfor-
Estimate 5 9.37). Thus, for all analyses with this wardly worded. As with the previous analysis, multi-
sample, the Satorra–Bentler Robust Estimation Proce- variate kurtosis was present (Mardia’s Normalized
dure was employed as a method of controlling for the Estimate 5 10.28) and the Satorra–Bentler Robust
non-normality. The fit indices for the unitary and two- Estimation Procedure was used to control for the
factor models based on the Houston sample are non-normality. In both the one- and two-factor
provided in Table 3. For the unitary factor model, all models, all factor loadings, residual terms, and latent
residual terms and latent factor variances were statis- factor variances were statistically significant (Po.01).
tically significant (Po.01), although one of the factor Furthermore, both models provided good fit to the
loadings (item 2) did not load significantly onto the data (see Table 3). However, the correlation between
single latent factor. Furthermore, the unitary model the two latent factors of the BFNE-II was extremely
showed grossly inadequate fit to the data. As shown in high (r 5 .95), and the revised items actually loaded
Table 3, all of the fit indices were out of range. For the equally well or better on the unitary model. Last, the
two-factor model, all factor loadings, residual terms, two-factor solution does not support prior assumptions
and latent factor variances were statistically significant of BFNE unity [e.g., Leary, 1983; Stopa and Clark,
(Po.01) and the fit indices suggested good model fit. 2001]. Because of the previously mentioned statistical
The two-factor model fit the data significantly better and theoretical reasons, the positively worded two-
than the unitary model [Satorra–Bentler Scaled w2 factor model was dropped from further consideration.
Difference Test (1, N 5 201) 5 186.06; Po.001]. The All told, the two-factor model appears to provide a
correlation between the two latent factors, before any better fit to the data when the four items in question
items were reverse-scored, was moderate (r 5 .41). were negatively worded; however, when the same items
Depression and Anxiety DOI 10.1002/da
 Research Article: BFNE-II 301

TABLE 4. Unitary and two-factor model factor loadings and Houston F (1, 383) 5 14.451, Po.001, Z2 5 .036).
This indicated an increased endorsement of the revised
Regina sample Houston sample BFNE-II items (M 5 12.21, SD 5 3.55) when com-
pared to the reverse-worded BFNE items (M 5 10.91,
Two-factor Two-factor
SD 5 3.19).
Item Unitary Factor 1 Factor 2y Unitary Factor 1 Factor 2
INTERNAL CONSISTENCY
1 0.78 0.78 0.80 0.79
3 0.78 0.80 0.83 0.56 Cronbach’s a is a standard method by which the
5 0.81 0.82 0.80 0.83 coherent nature of a scale can be judged. The higher
6 0.82 0.82 0.85 0.46 the a, the more related items are to one another.
8 0.82 0.82 0.78 0.80 A reliability analysis was performed to evaluate the
9 0.81 0.81 0.75 0.85 item total correlations and the a-if-item-deleted for
11 0.77 0.76 0.72 0.76 both the BFNE and the BFNE-II (see Table 2).
12 0.71 0.71 0.73 0.78 Internal consistency was acceptable for the BFNE
2 0.71 0.68 0.15 0.75
(a 5 .89) but excellent for the BFNE-II (a 5 .95). The
4 0.82 0.54 0.21 0.74
7 0.84 0.50 0.36 0.71
corrected item–total correlations for the revised items
10 0.72 0.73 0.35 0.73 used with the Regina sample (Table 2; M 5 0.75,
SD 5 0.06) were significantly higher than those of the
Note. reverse-worded items from the Houston sample
Straightforwardly worded items, Reverse-worded items,
[M 5 0.36, SD 5 0.11; F (1, 6) 5 39.849, Po.001,
Z2 5 .869]. For the BFNE-II, no conclusive evidence
y
Revised items.
of item redundancy was found in the correlation matrix
and individual item removal did not substantially affect
are positively worded, a unitary model appears to scale a, independent of whether the item was one of the
provide a better fit to theory and data. four revised ones or one of the eight originals (Table 2).

BFNE AND BFNE-II COMPARISONS DISCUSSION

The previously mentioned CFA results indicate that Our intent in this study was twofold. The first
item wording significantly influences responses to the purpose was to replicate the proposed factor structure
BFNE and BFNE-II items. To evaluate this further, a from previous literature. The second purpose was to
series of one-way analyses of variance (ANOVAs) were assess the utility of the four reverse-worded items of
performed comparing participant’s item–responses. the BFNE and determine what course of action should
The total score for the Regina sample with the be taken with them. This required replication of
BFNE-II (M 5 32.77, SD 5 10.04) was significantly previous findings from CFAs of the BFNE [Rodebaugh
higher than the total score for the Houston sample et al., 2004; Weeks et al., 2005]. The two previous
with the original BFNE [M 5 30.71, SD 5 9.04), F (1, CFAs suggest that a two-factor model provides a
383) 5 4.512, Po.05, Z2 5 .012]; however, the effect significantly better fit to the data than the unitary
size was small, accounting for about 1 percent of the model and, importantly, that one factor of the two-
variance. Response means for the BFNE reverse- factor model is the result of a methodological
worded items (M 5 2.73, SD 5 0.79) were compared component of the BFNE scoring procedure. Weeks
to response means for the BFNE straightforwardly et al. [2005] and Rodebaugh et al. [2004] both
worded items (M 5 2.47, SD 5 0.93) using a repeated recommended the use of a shortened BFNE, contain-
measures design. Houston participants endorsed the ing only the straightforwardly worded items.
reverse-worded items significantly more than the Having successfully replicated the previous CFAs—
straightforwardly worded items [F (1, 200) 5 12.535, finding that the reverse-worded items produce a
Po.001, Z2 5 .059]. A second repeated measures ana- second, atheoretical factor—we sought to determine a
lysis was performed comparing the revised (M 5 3.05, more definitive course of action for the BFNE reverse-
SD 5 0.86) and unchanged (M 5 2.57, SD 5 0.89) worded items. Three possibilities existed. First, as
items on the BFNE-II. Regina participants endorsed recommended by Marsh [1996] and Weeks et al.
the revised items significantly more than the un- [2005], reverse-worded items could be left unaltered
changed items [F (1, 183) 5 179.977, Po.001, Z2 5 but not scored. Second, the items could be discarded
.496]. When mean responses to the straightforwardly entirely. Third, the items could be reworded to
worded items from Regina and Houston were com- overcome the associated reverse-wording problem.
pared, the result was not significant [F (1, 383) 5 1.072, The corrected item–total correlations from the Hous-
P 5.301, Z2 5 .003]. Conversely, a significant differ- ton sample BFNE were sufficiently low to warrant
ence was found between the mean responses to the either not scoring or discarding the items, particularly
revised/reverse-worded item responses from Regina when considering their contribution to the atheoretical
Depression and Anxiety DOI 10.1002/da
302 Carleton et al.

factor. Unfortunately, this may result in a reduction of At this time, even though parsimony, along with the
scale sensitivity, particularly if the low corrected item– minimal reduction in scale a, might support removal of
total correlations result from confusion caused by the four items entirely, such a decision may be
wording rather than latent differences in each factor’s premature. The fit indices were acceptable for the
items. Given that previous work with the BFNE has unitary BFNE-II model, without an atheoretical factor;
tended toward either increasing, or applauding there were no substantive problems indicated by the
increases in, the sensitivity of the BFNE [Oei et al., reliability analysis; and there was a significant increase
1991; Rodebaugh et al., 2004; Turner et al., 1987], such in participant item endorsement for the revised items.
a reduction in sensitivity would seem to be regressive. The observation that participant responses to the
Prior to discussing the results of removing or four reverse-worded items increased in comparative
revising the reverse-worded items, it is important to endorsement when the items were revised requires
revisit the reason for their use in the first place. follow-up analyses involving comparable measures
Typically, reverse-worded items are used in longer [e.g., SADS; Watson and Friend, 1969] before the
questionnaires as a means of detecting acquiescent revised items can be definitively ratified or discarded.
response biases [Marsh, 1996]. Although this may have Indeed, after submission of this manuscript, Collins
been necessary in the FNE, having 30 items, the utility et al. [2005] presented research demonstrating good
of this method in the shorter BFNE is suspect. Indeed, reliability, sensitivity, and validity using the full 12
previous research indicates that reverse-worded items items in a clinical sample, also having reworded the
tend to form a separate atheoretical factor resulting in four reverse-worded items in a very similar fashion to
part from participant carelessness and error [Marsh, Taylor [1993].
1996; Spector et al., 1997]. Therefore, whereas the use It is also possible that the different factors found
of reverse-worded questions may be appropriate for reflect related but distinct constructs. Specifically, not
longer questionnaires, it appears to influence adversely caring about being negatively evaluated by others is not
response patterns when used in smaller questionnaires necessarily antonymous with fear of negative evaluation.
like the BFNE. For example, it is plausible that some people who are
The results from the Regina sample indicated that not socially anxious would nonetheless be bothered to
revisions to the reverse-worded items of the BFNE to learn that they have been negatively evaluated, even
make them straightforward had four major effects. though they do not actively fear negative evaluation.
First, the corrected item–total correlations for each Establishing the predictive and discriminant validity of
item increased to levels comparable to those of the the revised scale with socially anxious and other clinical
other items. Second, the CFA analyses resulted in populations would help address these concerns.
acceptable fit indices for a unitary model of the Furthermore, these findings are limited by the use of a
BFNE-II, with the revised items loading comparably convenience sample of undergraduates from two uni-
with other items on a single factor. Third, not only did versities. Although unlikely, it is impossible to deter-
participants’ endorsement of the four revised items mine conclusively that differences in the factor structure
continue to be significantly higher than endorsement of were not due to differences in the sample characteristics.
the other items, but also the difference in endorsement Additional research is required to determine whether
increased significantly. This serves to confirm the or not this factor structure will generalize to older
Weeks et al. [2005] hypothesis that use of the adults or patients with social phobia. Future research
reverse-worded items may actually result in an under- should also explore any remaining method variance in
estimation of fear of negative evaluation. Fourth, a more detail, as a multitrait–multimethod analysis
reliability analysis consisting of an interitem correla- would have required a greater sample size than that
tion and an a-if-item-deleted analysis was performed available in this study [e.g., for a discussion of the
and reviewed. For both statistics, the four revised items limitations of these models, see Tomás and Oliver,
were comparable to the other eight. 1999, or Marsh and Bailey, 1991]. Nevertheless, such
Psychometric changes stemming from revisions to validation could also be achieved by correlating BNFE-
the reverse-worded items of the BFNE do not II items with other independent measures of social
definitively resolve questions regarding their utility. anxiety and fear. Currently, the BFNE-II provides an
Matching the Houston and Regina samples on the total alternative to removing, or not scoring, these items,
scores of the shared items (in addition to demo- allowing researchers to err on the side of caution rather
graphics) allowed for an isolated comparison of the than risk a reduction in sensitivity.
differences between responses to the reverse-worded
items from the Houston sample and responses to
their revised counterparts from the Regina sample.
Still, for a more complete analysis of whether the REFERENCES
reverse-worded items represent a method factor, a Bentler PM. 2003. EQS: Structural equation program manual. Los
multitrait–multimethod analysis could be computed Angeles: Multivariate Software, Inc.
[e.g., Marsh, 1996; Tomás and Oliver, 1999] using a Brown EJ, Turovsky J, Heimberg RG, Juster HR, Brown TA, Barlow
larger sample. DH. 1997. Validation of the social interaction anxiety scale and the

Depression and Anxiety DOI 10.1002/da

 Research Article: BFNE-II 303

social phobia scale across the anxiety disorders. Psychol Assess Reiss S, McNally RJ. 1985. The expectancy model of fear. In: Reiss S,
9:21–27. Bootzin RR, editors. Theoretical issues in behaviour therapy.
Byrne BM. 1994. Structural equation modeling with EQS and EQS/ New York: Academic Press. p 107–121.
Windows: Basic concepts, applications, and programming. Thou- Rodebaugh TL, Woods CM, Thissen DM, Heimberg RG,
sand Oaks, CA: Sage. 287p. Chambless DL, Rapee RM. 2004. More information from fewer
Carleton RN, Asmundson GJG. 2005. Proneness to chronic pain in questions: The factor structure and item properties of the original
people with posttraumatic stress and social anxiety. In: Efekhari A, and brief fear of negative evaluation scale. Psychol Assess 2:
Chair. Next steps in understanding and treatment of PTSD. 169–181.
Symposium presented at 25th annual meeting of the Anxiety Safren SA, Turk CL, Heimberg RG. 1998. Factor structure of the
Disorders Association of America, March 17–20, Seattle, WA. social interaction anxiety scale and the social phobia scale. Behav
Collins KA, Westra HA, Dozois DJA, Stewart SH. 2005. The validity Res Ther 36:443–453.
of the brief version of the fear of negative evaluation scale. Spector PE, Van Katwyk PT, Brannick MT, Chen PY. 1997.
J Anxiety Disord 19:345–359. When two factors don’t reflect two constructs: How item
Gore KL, Carter MM, Parker S. 2002. Predicting anxious response characteristics can produce artifactual factors. J Manage
to a social challenge: The predictive utility of the social interaction 23:659–677.
anxiety scale and the social phobia scale in a college population. Stein MB, Jang KL, Livesley WJ. 2002. Heritability of social anxiety-
Behav Res Ther 40:689–700. related concerns and personality characteristics: A twin study.
Hu L, Bentler PM. 1999. Fit indices in covariance structure J Nerv Ment Dis 190:219–224.
modeling: Sensitivity to underparameterized model mis-specifica- Stopa L, Clark DM. 2001. Social phobia: Comments on the viability
tion. Psychol Methods 3:424–453. and validity of an analogue research strategy and British norms
Leary MR. 1983. A brief version of the Fear of Negative Evaluation for the fear of negative evaluation questionnaire. Behav Cogn
Scale. Pers Soc Psychol Bull 9:371–375. Psychother 29:423–430.
Marsh HW. 1996. Positive and negative global self-esteem: Taylor S. 1993. The structure of fundamental fears. J Behav Ther
A substantively meaningful distinction or artifactors? J Pers Soc Exp Psychiatry 24:289–299.
Psychol 70:810–819. Tomás JM, Oliver A. 1999. Rosenberg’s Self-Esteem Scale: Two
Marsh HW, Bailey M. 1991. Confirmatory factor analyses of factors or method factors. Struct Eq Model 6:84–98.
multitrait-multimethod data: A comparison of alternative models. Turner SM, McCanna M, Beidel DC. 1987. Validity of the social
Appl Psych Meas 15:47–70. avoidance and distress and fear of negative evaluation scales. Behav
Mattick RP, Clarke JC. 1998. Development and validation of Res Ther 25:113–115.
measures of social phobia scrutiny fear and social interaction Watson D, Friend R. 1969. Measurement of social-evaluative anxiety.
anxiety. Behav Res Ther 36:455–470. J Consult Clin Psychol 33:448–457.
Miller RS. 1995. On the nature of embarrassability: Shyness, social Weeks JW, Heimberg RG, Fresco DM, Hart TA, Turk CL,
evaluation, and social skill. J Pers 63:315–339. Schneier FR, Liebowitz MR. 2005. Empirical validation
Oei TPS, Kenna D, Evans L. 1991. The reliability, validity, and and psychometric evaluation of the Brief Fear of Negative
utility of the SAD and FNE scales for anxiety disorder patients. Evaluation Scale in patients with social anxiety disorder. Psychol
Pers Indiv Diff 12:111–116. Assess 17:179–190.

Depression and Anxiety DOI 10.1002/da