Escolar Documentos
Profissional Documentos
Cultura Documentos
V. E.Journal
Caballo
of et
Psychological
al.: Measuring
Assessment
2010
Social
Hogrefe
Anxiety
2010;&Vol.
Huber
in 11
26(2):95107
Co
Publishers
untries
Original Article
Abstract. This paper reports on two studies conducted to develop and validate a new self-report measure of social phobia/anxiety the
Social Anxiety Questionnaire for Adults (SAQ-A) (Cuestionario de ansiedad social para adultos, CASO-A). A diary-item recording
procedure was used to generate the initial pool of items. In Study 1, data from 12,144 participants provided 6 factors with moderate
intercorrelations. Estimates of internal consistency reliability were adequate (range = .86 to .92) for the 6 factors included in the final
confirmatory factor analysis. In Study 2, data provided by 10,118 nonclinical participants were used to explore preliminary reliability
and validity estimates for a revised version of the SAQ-A the Social Anxiety Questionnaire for Adults Revised (SAQ-AR). Approximately 106 researchers from 10 Latin American countries and Spain contributed to this data collection process. Specific comments are
made on the structure of the new questionnaire as regards some commonly-used self-report measures of social phobia/anxiety.
Keywords: social anxiety, social phobia, SAQ-AR, self-report measures, cross-cultural research
Introduction
Once described as a neglected disorder (Liebowitz, 1987),
social anxiety has attracted a great deal of research interest
among psychiatrists and psychologists alike over the past
two decades. Several measures (interviews and inventories) have been developed to tap the social anxiety construct, including the Liebowitz Social Anxiety Scale
(LSAS; Liebowitz, 1987), the Social Phobia and Anxiety
Inventory (SPAI; Turner, Beidel, Dancu, & Stanley, 1989),
the Brief Social Phobia Scale (BSPS; Davidson et al.,
1991), the Social Phobia Scale (SPS; Mattick & Clarke,
1998), the Social Interaction Anxiety Scale (SIAS; Mattick
& Clarke, 1998), the Self-Statements During Public Speaking Scale (Hofmann & DiBartolo, 2000), and the Social
Phobia Inventory (SPIN; Connor et al., 2000). In addition,
a number of older, but still popular, scales exist, such as the
Fear of Negative Evaluation (FNE) and Social Avoidance
and Distress (SAD) Scales (Watson & Friend, 1969).
Although frequently used to assess social anxiety in clinical and research settings, the existing instruments have a
number of limitations. First, items from most of these instruments were not empirically derived. For instance, the
2010 Hogrefe & Huber Publishers
items on the Social Phobia Scale (SPS) and the Social Interaction Anxiety Scale (SIAS) were subjectively derived
mainly from an initial pool of statements comprising 164
items, which themselves were derivatives of other existing
fear survey schedules and social anxiety inventories (Mattick & Clarke, 1998). For example, the Social Phobia and
Anxiety Inventorys (SPAI) initial item pool was generated
by the authors after reviewing available inventories and
DSM-III criteria for social phobia (APA, 1980), and by
compiling a list of complaints from a patient population
(Turner, Beidel et al., 1989). The Social Phobia Inventory
(SPIN; Connor et al., 2000) was based and modeled on a
former inventory, the Brief Social Phobia Scale (BSPS; Davidson et al., 1991), and the Liebowitz Social Anxiety
Scale (LSAS) was generated by the author of the instrument (Liebowitz, 1987). Most of the items in these questionnaires corresponded to those in already existing surveys, and issues regarding content validity were not addressed in greater detail (see Haynes, Richard, & Kubany,
1995, about the importance of this issue).
Not surprisingly, the most popular instruments vary considerably in the number and type of factors underlying the
social anxiety construct; in other words, their factor strucEuropean Journal of Psychological Assessment 2010; Vol. 26(2):95107
DOI: 10.1027/1015-5759/a000014
96
tures appear far from robust. In the case of the LSAS, some
authors have identified four factors (Safren et al., 1999;
Slavkin, Holt, Heimberg, Jaccard, & Liebowitz, 1990),
whereas others have found a 5-factor solution (Baker,
Heinrichs, Kim, & Hofmann, 2002). An additional problem
is that not only the number, but the general content of the
factors differ across studies. Similar inconsistent findings
in factor solutions have been reported for other social anxiety/phobia measures, such as the SPIN (Antony, Coons,
McCabe, Ashbaugh, & Swinson, 2006; Connor et al., 2000;
Johnson, Inderbitzen-Nolan, & Anderson, 2006; Radomsky et al., 2006), the SPAI (Olivares, Garcia-Lopez, Hidalgo, Turner, & Beidel, 1999; Osman, Barrios, Aukes, & Osman, 1995; Turner, Stanley, Beidel, & Bond, 1989), and the
SAD and FNE (Hofmann, DiBartolo, Holaway, & Heimberg, 2004; Olivares, Garca-Lpez, Hidalgo, 2004;
Turner, McCanna, & Beidel, 1987).
In addition to the aforementioned methodological problems with the nonobjective method of social anxiety scale
development is the fact that all of the above measures were
created exclusively for English speakers, primarily from
North America and Australia. The use of these instruments
with Spanish-speaking samples usually involves a somewhat simplistic direct translation of the questionnaires from
English to Spanish (e.g., Olivares et al., 1999, 2004). Unfortunately, this procedure ignores cultural differences in
the expression of social anxiety and social norms (Heinrichs et al., 2006). This is rather ironic when one considers
that social interaction styles and norms are probably among
the most important defining features of a culture and are
often precisely the locus of differences across cultures.
Thus, it remains to be seen whether a questionnaire that
describes a variety of social situations is applicable across
cultures. To address the cultural and methodological limitations of the existing literature, we conducted an extensive
series of studies in order to develop a new social anxiety
questionnaire, without directly relying on items from existing self-report instruments. In contrast to existing measures, we developed the instrument based on items generated by large and very diverse Spanish and Portuguese
speaking samples.
Scale Construction
The experts then paraphrased the 512 social situations into
items. Four additional situations that typically produce
great distress were also selected (stressful life events, such
as suffering an armed attack) and added to control response biases. These 516 items formed the Social Anxiety
Questionnaire for Adults (SAQ-A) (Cuestionario de Ansiedad Social para Adultos; CASO-A), the initial version
of a new self-report instrument intended to assess social
anxiety. The items were randomly ranked and each item
could be answered on a seven-point Likert scale to indicate
the level of uneasiness, stress or nervousness in response
to each situation: 0 = not at all, 1 = very slight, 2 = slight,
3 = moderate, 4 = high, 5 = very high, and 6 = extremely
high. Instructions given to those completing the scale were
as follows:
There follows a series of social situations that may cause you
unease, stress or nervousness to a lesser or greater extent.
Please place an X on the number that best reflects your reaction. If you have never experienced the situation described,
please imagine what your level of unease, stress, or nervousness might be if you were in that situation, placing an X on
the corresponding number.
97
Men
Country
Mean age
(SD)
Mean age
(SD)
Argentina
,499
30.25
(10.89)
,378
29.82
(11.42)
Brazil
,702
26.07
(9.48)
,547
Chile
,376
26.90
(10.86)
Colombia
,852
Costa Rica
Spain
Mexico
Paraguay
All subjects
N
Mean age
(SD)
Women
Men
All subjects
Mean age
(SD)
Mean age
(SD)
Mean
age (SD)
,877 30.05
(11.11)
,329
23.38
(5.42)
,348
24.77
(8.53)
,677 24.09
(1.56)
27.55
(10.79)
1,249 26.76
(10.12)
,405
31.04
(13.06)
,358
30.12
(11.39)
,763 30.61
(12.30)
,308
27.91
(11.52)
, 684 27.36
(11.16)
,310
26.76
(11.65)
,297
26.53
(10.83)
,607 26.64
(11.25)
24,70
(9.60)
,774
25,47
(9.81)
1,626 25.21
(9.78)
,870
26,11
(11.98)
,857
27,80
(13.00)
1,727 26.96
(12.53)
,205
23.23
(9.42)
,122
18.87
(5.82)
,327 21.58
(8.51)
,363
25.87
(9.10)
,186
25.35
(9.68)
,549 25.69
(9.29)
,905
22.80
(8.80)
,668
27.01
(12.00)
1,573 24.58
(10.48)
1,335
23.24
(8.66)
,907
26.21
(11.41)
2,242 24.44
(9.97)
2,377
25.14
(10.34)
1,954
25.29
(9.68)
4,331 25.22
(10.05)
1,258
25.25
(12.16)
1,128
25.55
(9.93)
2,386 25.39
(11.16)
,91
24.62
(8.03)
,77
21.91
(6.82)
,168 23.27
(7.57)
,100
22.48
(5.83)
,100
21.85
(5.85)
,200 22.16
(5.83)
1,002
23.08
(8.37)
,978
23.25
(8.00)
1,980 23.16
(8.18)
,529
21.27
(6.33)
,497
21.71
(6.84)
1,026 21.49
(6.58)
Uruguay
,101
32.39
(12.27)
,100
33.43
(10.93)
,201 32.92
(11.60)
,135
31.30
(12.78)
,114
34.29
(13.11)
Venezuela
,195
27.53
(11.91)
,186
25.56
(9.73)
,381 26.52
(10.88)
,301
19.77
(4.12)
,299
20.53
(4.57)
7,271
25.15
(10.05)
6,126
25.75
(10.22)
13,397 25.43
(10.13)
5,935
24.79
(10.51)
5,091
25.81
(10.74)
Peru
All countries
,249 32.67
(12.99)
600
20.15
(4.36)
11,026 25.65
(10.63)
year and five months. Collaborators used a prepared database in Excel to enter the data.
Participating Subjects
An initial pool of 13,397 participants completed the SAQA (mean age = 25.43; SD = 10.13) (see Table 1 for the
distribution of participating subjects by country). Approximately half (7,271) were women (mean age = 25.15; SD
= 10.05), and 6,126 were men (mean age = 25.76; SD =
10.22). The minimum age for subjects was 16 years. With
regard to age distribution, 5,420 (40.4%) subjects were
younger than 20 years old, 3,029 (22.7%) were between the
ages of 20 and 24, 1674 (12.49) were between 25 and 30,
2225 (16.61) were between 31 and 50, and 1,049 (7.83%)
were 51 years or older. The participants had different levels
of education (students, workers, etc.). Specifically, 17.6%
were university psychology students, 40.6% were university students from other majors, 14% were workers with a
university degree, 13.1% were workers with no university
degree, 9.3% were high school students, and 3.7% could
not be included in any of the former categories (e.g., retired
or unemployed). No data were obtained for the remaining
1.7% of participants.
Missing data were expected, given the size of the partic-
Procedure
98
Results
Factor Analysis of the Initial Version of the
Questionnaire (SAQ-A = 516 Items)
In order to reduce the number of items, we performed an
exploratory principal components factor analysis with varimax rotation, which optimizes complex structures by capturing a small number of large loadings and a large number
of small loadings for each factor. Examination of the scree
plot suggested a 6-factor solution. The hierarchical analysis
of oblique factors gave the same 6-factor solution (Statsoft,
2006). We then performed an oblique principal component
cluster analysis in order to group the items into nonoverlapping clusters, so each cluster could be interpreted as unidimensional. This procedure allowed us to substitute a
group of variables with a smaller one (n-m) with the minimum loss of information in order to maximize the explained variance by the components of the cluster. This procedure is iterative, at each step suppressing those variables
that have the highest ratio values. The smaller these values
are, the greater the evidence that the variable has a strong
relationship with the rest of the components of the cluster
and a weak relationship with the components of the other
clusters. The 512 variables were considered in the analysis
(forcing a solution of 6 clusters). The four control items
were not included in the analysis, but they did allow us to
estimate how many subjects might be filling the questionnaire at random because they were answerable in only one
direction of increasing distress. Given the large sample size
relative to the extremely small number of participants
flagged by the control items, no action was taken. After
successive analyses suppressing variables with the highest
(1 R2own)/(1 R2next)1 ratio values, a solution of 12 items
per cluster was reached. The final distribution of the items
by cluster that were used in the subsequent analyses (exploratory and confirmatory factor analyses) is the same as
that found in Table 2.
Table 2. Item loadings for every factor and correlations item-total score for the SAQ-A
Factor loadings
Items and name of each factor
F1
F2
F3
F4
F5
F6
Item
total
.54
.02
.06
.05
.06
.23
.648
.54
.68
.05
.13
.11
.05
.02
.17
.03
.01
.659
.02
.08
.629
386. Not knowing how to continue a conversation after a topic has been exhausted
.52
.00
.25
.07
.02
.03
.634
.79
.05
388. Proposing an idea to a group of friends and not being taken seriously
.71
.05
.04
.16
.02
.09
.592
.05
.14
.03
.08
.600
.58
.52
.11
.08
.12
.11
.07
.654
.08
.14
.04
.10
.01
.665
.69
.06
.08
.10
.01
.01
.609
.60
.08
.20
.01
.18
.15
.621
.66
.12
.15
.04
.06
.02
.557
487. Being in the home of strangers and not knowing what to say or do
.47
.09
.07
.09
.05
.04
.617
.29
.65
.4
.20
.12
.07
.570
.10
.48
.13
.09
.02
.08
.658
.04
.74
.07
.04
.00
.03
.549
In the formula, R2own represents the determination coefficient of each variable with its own cluster, and R2next the determination coefficient
of each variable with the nearest cluster. Naturally, we would want each component of the cluster to be strongly related with its own cluster
(R2own 1) and less related with the nearest cluster (R2next 0).
99
Factor loadings
Items and name of each factor
F1
F2
F3
F4
F5
F6
Item
total
.26
.45
.05
.02
.03
.03
.656
.01
.73
.09
.02
.01
.02
.640
.25
.50
.02
.03
.05
.05
.601
.20
.67
.06
.03
0.05
.02
.642
397. Being told by someone of the opposite sex that they like me
.11
.72
.01
.02
.01
.02
.636
.19
.52
.11
.03
.03
.03
.616
.13
.74
.05
.01
.06
.03
.643
.02
.71
.16
.06
.06
.02
.642
.05
.61
.17
.06
.05
.04
.611
.06
.03
.56
.07
.09
.05
.569
275. Greeting each person at a social meeting when I dont know most of them
.26
.05
.41
.07
.07
.11
.641
.18
.01
.43
.10
.03
.11
.630
453. Talking about my personal feelings with someone of the opposite sex
F3. Interactions with Strangers
332. Talking on the phone with someone I do not know very well
.03
.03
.68
.07
.05
.04
.554
.03
.02
.76
.07
.00
.00
.563
.04
.13
.58
.04
.00
.09
.542
.07
.09
.70
.02
.05
.02
.594
.07
.09
.78
.00
.02
.00
.567
.02
.33
.37
.06
.10
.00
.545
.11
.20
.54
.03
0.04
.09
.667
501. Looking into the eyes of someone I have just met while we are talking
.07
.22
.44
.01
.07
.03
.523
.12
.04
.67
.00
.15
.01
.470
.05
.14
.08
.56
.19
.08
.479
.04
.02
.09
.55
.00
.04
.456
.12
.08
.21
.50
.04
.15
.458
.00
.05
.20
.57
.10
.07
.521
.05
.08
.12
.48
.11
.08
.455
.12
.05
.17
.61
.02
0.10
.470
.09
.05
.03
.51
.18
.07
.446
.07
.11
.04
.47
.05
.15
.488
.21
.08
.06
.55
.02
.17
.369
.19
.02
.05
.48
.20
.00
.545
.08
.02
.13
.40
.09
.12
.551
.31
.08
.08
.40
.13
.15
.582
.05
.01
.04
.20
.56
.11
.542
.13
.02
.01
.19
.63
.5
.511
.10
.04
.10
.14
.64
.13
.524
222. Asking someone who is speaking loudly at the movies to lower their voice
.08
.03
.03
.14
.63
.09
.549
.07
.14
.19
.02
.46
.05
.578
.15
.01
.15
.06
.54
.15
.464
.26
.06
.09
.02
.52
.16
.472
.27
.03
.06
.01
.53
.01
.597
100
Factor loadings
Items and name of each factor
F1
F2
F3
F4
F5
F6
Item
total
299. Telling a taxi driver that he/she has taken an abnormally long route
.17
.02
.06
.06
.55
.00
.548
.32
.06
.05
.09
.46
.02
.596
482. Telling someone that their behavior bothers me and asking them to stop
.13
.04
.06
.07
.56
.01
.549
.14
.05
.07
.08
.55
.01
.554
.11
.02
.00
.26
.10
.65
.503
.12
.16
.03
.13
.16
.45
.578
.11
.07
.08
.12
.01
.77
.578
.06
.02
.04
.15
.09
.62
.576
.09
.16
.03
.16
.21
.42
.603
.11
.14
.10
.06
.11
.44
.647
.29
.04
.03
.06
.08
.60
.577
.27
.02
.08
.10
.10
.68
.624
.25
.03
.17
.10
.02
.57
.651
.19
.14
.17
.15
.13
.39
.680
.46
.03
.12
.07
.01
.32
.681
.45
.13
.09
.13
.03
.39
.636
of the variables (in fact, they are above the value of .500
which is usually used as a threshold to discard a variable
from analysis). Finally, 60% of communalities were above
.50 (ranging from .35 to .70).
In order to decide the optimal number of factors, a parallel
analysis (Velicer, Eaton, & Fava, 2000; Watkins, 2000) was
implemented using the Monte Carlo procedure with 200 replications to determine the number of eigenvalues with values
above those that could be obtained from the same number of
subjects and variables (i.e., generating a group of random
values with normal distribution, calculating the matrix of correlations and subjecting it to principal components analysis
to calculate the mean eigenvalues). Results show that the 6factor solution is the best fit to our data, given that the size of
randomly generated eigenvalues after factor 6 is higher than
the observed eigenvalues.
This exploratory factor analysis identified 6 factors with
eigenvalues higher than 1.00 explaining 50.24% of the cumulative variance. Item loadings are presented in Table 2.
The first factor (eigenvalue = 25.49) explained 35.42% of
the variance. The 12 items loading highly on this factor
describe Awkward Behaviors in Embarrassing Situations.
The second factor showed an eigenvalue of 3.22 and explained 4.47% of the total variance. The 12 high loading
items describe situations of Interaction with the Opposite
Sex. Factor 3 showed an eigenvalue of 2.32 and explained
3.23% of the variance. The items of this factor refer to situations of Interaction with Strangers. Factor 4, with an eigenvalue of 1.98, explained 2.76% of the variance. The
items refer to situations of Criticism and Embarrassment.
Factor 5, with an eigenvalue of 1.67, explained 2.33% of
the variance and is related to Assertive Expression of Annoyance, Disgust or Displeasure. Factor 6, with an eigenvalue of 1.46, explained 2.03% of variance and was defined
as Speaking/Performing in Public/Talking with People in
Authority. Interfactor correlations were moderate (range =
.33 to .60) (see Table 6).
101
Relative fit
S-B
Model 3
51629.98 12746.49
14706.52
p = .000 p = .000
p = .000
DF
252
237
246
GFI
.70
.91
.89
SRMR
.064
.036
.043
NFI
.95
.99
.99
NNFI
.95
.99
.98
RFI
.94
.99
.98
.95
.99
.99
.12
.063
.066
tors, and (3) measurement errors associated with the indicators are not correlated with each other. The results of the
contrast comparisons of the three models are summarized
in Table 3.
As can be seen in Table 3, Models 2 (6 correlated factors) and 3 (6 first-order factors and one second-order factor) showed a good overall fit, suggesting that the restrictions we specified for the models were correct. However,
the fit of Model 2 was slightly better: the RMSEA index
was .063 in Model 2 and .066 in Model 3; indices SRMR
(.036 vs. .043), GFI (.91 vs. .89), NNFI and RFI (.99 vs.
.98) were also better for Model 2. Other indices comparing
the fit of Models 2 and 3, such as composite reliability and
average variance extracted (AVE) indicated a similar fit for
both models, although again slightly better for Model 2
than Model 3 (see Table 4). The average interitem correla-
Noncentrality CFI
based fit
RMSEA
Table 4. Composite reliability and average variance extracted of the three models
Model 1
Model 2
Model 3
Compos- AVE
ite reliability
Composite AVE
reliability
Compos- AVE
ite reliability
Factor 1 .963
.522
.903
.699
.903
.699
Factor 2
.912
.721
.913
.724
Factor 3
.886
.660
.886
.660
Factor 4
.839
.567
.840
.568
Factor 5
.869
.624
.868
.622
Factor 6
.883
.654
Note. AVE = Average variance extracted.
.883
.654
102
tion was 0.486 for Factor 1, 0.487 for Factor 2, 0.436 for
Factor 3, 0.337 for Factor 4, 0.382 for Factor 5, and 0.442
for Factor 6. The total average interitem correlation was
0.337. Interfactor correlations were from moderate to relatively high (range = .64 to .84) (see Table 6).
In order to determine discriminant validity, the average
variance extracted (AVE) was compared with the coefficient of determination (R2) for each couple of latent variables. All the comparisons (10) carried out showed an AVE
greater than R2. This can be considered as a clear evidence
of discriminant validity since each latent construct must
explain the measures composing it rather than other constructs measures.
Composite reliability of each of the latent variables
(construct reliability) was calculated through the formula:
Procedure
The procedure was similar to the first study. Collaborators
from each country revised each item of the SAQ-AR to fit
the everyday language of their country and culture. There
was no option to add new items. No significant changes
were made to the 72 items composing the CASO-AR. Data
collection took place over a period of 1 year.
In order to calculate additional psychometric properties
of this new questionnaire, such as consistency, validity, and
reliability, we selected some self-report instruments usually
employed to assess social phobia/anxiety, such as the SPAI
Table 5. Correlations (Pearson) among the SAQ-AR and its 6 factors with other self-report measures of social anxiety
Questionnaires for assessing social phobia/anxiety
SAQ-AR and its factors
SPAI
96 items
SPAI
Sp Ag
LSAS
Anxiety
LSAS
Avoidance
SPIN
.64
.59
.59
.43
.59
.62
.58
.58
.45
.58
.75
.75
.62
.44
.64
.69
.64
.62
.51
.60
.49
.44
.50
.39
.48
.62
.55
.55
.44
.56
Table 6. Interfactor correlations for exploratory and confirmatory factor analysis of the abbreviated version
of the SAQ-A
Interfactor correlations for exploratory factor analysis
F1
F2
F3
F4
F5
F1
1.00
F2
0.56
1.00
F3
0.50
0.58
1.00
F4
0.42
0.37
0.33
1.00
F5
0.51
0.49
0.46
0.46
1.00
F6
0.50
0.60
0.51
0.45
0.49
F6
1.00
F2
F3
F4
F1
1.00
F2
0.78
1.00
F3
0.72
0.83
1.00
F4
0.80
0.65
0.64
1.00
F5
0.79
0.72
0.74
0.76
F5
F6
1.00
F6 0.83
0.84
0.81
0.77
0.78
1.00
Note: F1. Awkward behavior in social embarrassing situations; F2.
Interactions with the opposite sex; F3. Interactions with strangers; F4.
Criticism and embarrassment; F5. Assertive expression of annoyance,
disgust or displeasure; F6. Speaking/performing in public/Talking
with people in authority.
103
Results
Confirmatory Factor Analysis
Participating Subjects
A total of 11,026 subjects participated in the second study.
The mean age of the total sample was 25.65 years (SD =
10.63) and consisted of 5,935 women (mean age = 24.79;
SD = 10.51) and 5,091 men (mean age = 25.81; SD =
10.74). The minimum age for subjects was 16 years, but
there was no upper age limit. Table 1 shows the sex, age,
and number of subjects in the participating countries. The
participants had different levels of education (students,
workers, etc.). Specifically, 22% were psychology students, 39.5% were university students with other majors,
14.7% were workers with a university degree, 6.9% were
workers with no university degree, 4.9% were school students, and 6.4% could not be included in any of the former
categories. No data were obtained for the remaining 5.6%
of participants. Missing data were addressed using listwise
deletion, as in the first study, so that the final number of
subjects for factor analysis was 10,118.
Instruments
104
Discussion
This work presents the development and initial psychometric evaluation of a new questionnaire designed to measure
social phobia/anxiety. Although there are already a number
of anxiety scales in existence, they all suffer from several
notable weaknesses, not the least of which is the manner in
which their items were generated. Existing measures contain items that were adapted from other measures, adapted
from DSM-III or DSM-IV criteria, or by generating items
based on the opinions of experts without secondary confirmation of their validity. By contrast, the present research
applied an objective method to diary-generated items by
asking a large group of diverse participants to record any
social situations that elicited social anxiety during their daily lives. Furthermore, in contrast to existing measures, we
recruited a large and culturally diverse sample from Spanish- and Portuguese-speaking countries.
An extensive first version of the questionnaire, the SAQA, was applied to a large sample of people from 10 LatinAmerican countries and Spain. Objective statistical reduction of the scale produced an instrument with 6 factors and
72 items. Given the item generation procedure and the robust factor structure observed in large samples, we believe
that our measure adequately describes the structure of social anxiety among Latin-American and Spanish-speaking
people. Note also that we followed most of the content validation guidelines proposed by some authors (e.g., DeVellis, 2003; Haynes et al., 1995). It remains to be seen whether the same factor structure will be observed among individuals with social anxiety disorder (social phobia) and
among people from other countries (e.g., Europe, North
America, etc.).
When comparing the factor structure of the SAQ-AR
and the most used social anxiety instruments (SPAI, LSAS,
SPIN, SIAS, SPS, and BPS), some surprising findings
emerged. Only 2 factors were consistently identified,
namely, Interactions with Strangers and Speaking/Performing in Public. Interestingly, the factor Interaction with
the Opposite Sex was not identified in the factor structure
of any of the popular existing measures. This seems surprising, given the reported centrality of this problem in the
lives of individuals with social anxiety/phobia. Yet, existing measures include very few, if any, items dealing with
anxiety about social interaction with the opposite sex. For
example, the SPS and SPIN include no items, and the SIAS
and the LSAS include only one item assessing this problem
area. The SPAI includes 17 of the 96 items. However, these
17 items are really subitems grouped in every case with
other three subitems (fear of strangers, people in authority,
and people in general) to give the mean score of 17 higher items. We understand this to be a limitation of the scoring procedure of the SPAI. A better approach might have
been to consider each subitem as an independent item rather than averaging different items prior to calculating a
score. This method would be more likely to result in a sep-
Concurrent Validity
105
This study was made possible by a grant from Spains Ministry of Science and Technology awarded to the research
project with reference BSO2003-07029/PSCE and cofinanced by the European Regional Development Fund
(ERDF). Stefan G. Hofmann is supported by a grant from
the National Institute of Mental Health (MH075889) and
is a consultant for Organon.
We heartily appreciate the collaboration of all the subjects from the different countries who voluntarily participated in this study.
The CISO-A Research Team, co-author of this article, is
composed of the following researchers: Argentina: G. Bra-
Acknowledgments
106
American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders DSM-III (3rd ed.) Washington, DC: Author.
Antony, M. M., Coons, M. J., McCabe, R. E., Ashbaugh, A., &
Swinson, R. P. (2006). Psychometric properties of the Social
Phobia Inventory: Further evaluation. Behavior Research and
Therapy, 44, 11771185.
Baker, S. L., Heinrichs, N., Kim, H., & Hofmann, S. G. (2002).
The Liebowitz Social Anxiety Scale as a self-report instrument: A preliminary psychometric analysis. Behavior Research and Therapy, 40, 701715.
Bandalos, D. L. (2002). The effects of item parceling on goodness-of-fit and parameter estimate bias in structural equation
modeling. Structural Equation Modeling, 9, 78102.
Bentler, P. M. (1995). EQS structural equations software manual.
Encino, CA: Multivariate Software.
Bhogal, K. S., & Baldwin, D. S. (2007). Pharmacological treatment of social phobia. Psychiatry, 6, 217223.
Caballo, V.E., & Nobre, L.A. (2009). Factor analysis, sex differences and psychometric characteristics of the Social Phobia
and Anxiety Inventory (SPAI). Unpublished manuscript, University of Granada, Spain.
Caballo, V. E., Olivares, J., Lpez-Gollonet, C., Irurtia, M. J., &
Rosa, A. I. (2003). Una revisin de los instrumentos para la
evaluacin de la fobia social: Algunos datos empricos [A review of the instruments for the assessment of social phobia:
Some empirical data]. Psicologa Conductual, 11, 539562.
Coffman, D. L., & MacCallum, R. C. (2005). Using parcels to
convert path analysis models into latent variable models. Multivariate Behavioral Research, 40, 235259.
Connor, K. M., Davidson, J. R. T., Churchill, L. E., Sherwood, A.,
Foa, E., & Weisler, R. H. (2000). Psychometric properties of
the Social Phobia Inventory (SPIN): A new self-rating scale.
British Journal of Psychiatry, 176, 379386.
Davidson, J. R., Miner, C. M., De Veaugh-Geiss, J., Tupler, L. A.,
Colket, J. T., & Potts, N. L. (1997). The Brief Social Phobia
Scale: A psychometric evaluation. Psychological Medicine,
27, 161166.
Davidson, J. R., Potts, N. L., Richichi, E. A., Ford, S. M., Krishnan, R. R., Smith, R. D. et al. (1991). The Brief Social Phobia
Scale. Journal of Clinical Psychiatry, 52, 4851.
DeVellis, R. F. (2003). Scale development: Theory and applications (2nd ed.). Thousand Oaks, CA: Sage.
Fresco, D. M., Coles, M. E., Heimberg, R. G., Liebowitz, M. R.,
Hami, S., Stein, M. B. et al. (2001). The Liebowitz Social Anxiety Scale: A comparison of the psychometric properties of
self-report and clinician-administered formats. Psychological
Medicine, 31, 10251035.
Galassi, J. P., DeLo, J. S., Galassi, M. D., & Bastien, S. (1974).
The College Self-Expression Scale: A measure of assertiveness. Behavior Therapy, 5, 165171.
Haynes, S., Richard, D., & Kubany, E. (1995). Content validity
in psychological assessment: A functional approach to concepts and methods. Psychological Assessment, 7, 238247.
Heimberg, R. G., Horner, K. J., Juster, H. R., Safren, S. A., Brown,
E. J., Schneieer, F. R. et al. (1999). Psychometric properties of
the Liebowitz Social Anxiety Scale. Psychological Medicine,
29, 199212.
Heinrichs, N., Rapee, R. M., Alden, L. A., Bgels, S., Hofmann,
S. G., Oh, K. J. et al. (2006). Cultural differences in perceived
social norms and social anxiety. Behavior Research and Therapy, 4, 11871197.
Hofmann, S. G., & DiBartolo, P. M. (2000). An instrument to assess self-statement during public speaking: Scale development
and preliminary psychometric properties. Behavior Therapy,
31, 499515.
Hofmann, S. G., DiBartolo, P. M., Holoway, R. M., & Heimberg,
R. G. (2004). Scoring error of Social Avoidance and Distress
Scale and its psychometric implications. Depression and Anxiety, 19, 197198.
Johnson, H. S., Inderbitzen-Nolan, H. M., & Anderson, E. R.
(2006). The Social Phobia Inventory: Validity and reliability
in an adolescent community sample. Psychological Assessment, 18, 269277.
Kline, R. B. (2005). Principles and practice of structural equation
modeling. New York: Guilford.
Liebowitz, M. R. (1987). Social phobia. Modern Problems in
Pharmacopsychiatry, 22, 141173.
Mattick, R. P., & Clarke, J. C. (1998). Development and validation
of measures of social phobia scrutiny and social interaction
anxiety. Behavior Research and Therapy, 36, 455470.
Nasser-Abu Alhija, F., & Wisenbaker, J. (2006). A Monte Carlo
study investigating the impact of item parceling strategies on parameter estimates and their standard errors in CFA. Structural
Equation Modeling: A Multidisciplinary Journal, 13, 204228.
Olivares, J., Garca-Lpez, L. J., Hidalgo, M. D., & Caballo, V.
(2004). Relationships among social anxiety measures and their
invariance: A confirmatory factor analysis. European Journal
of Psychological Assessment, 20, 172179.
References
107
Vicente E. Caballo
Faculty of Psychology
University of Granada
E-18071 Granada
Spain
Tel./Fax +34 958 125 927
E-mail vcaballo@ugr.es