Psychological Assessment © 2011 American Psychological Association

2012, Vol. 24, No. 1, 66 –76 1040-3590/12/$12.00 DOI: 10.1037/a0024544

Development of a Short Form Social Interaction Anxiety (SIAS) and

Social Phobia Scale (SPS) Using Nonparametric Item Response Theory:
The SIAS-6 and the SPS-6
Lorna Peters Matthew Sunderland and Gavin Andrews
Macquarie University University of New South Wales

Ronald M. Rapee Richard P. Mattick

Macquarie University University of New South Wales

Shortened forms of the Social Interaction Anxiety Scale (SIAS) and the Social Phobia Scale (SPS) were
developed using nonparametric item response theory methods. Using data from socially phobic partic-
ipants enrolled in 5 treatment trials (N 456), 2 six-item scales (the SIAS-6 and the SPS-6) were
developed. The validity of the scores on the SIAS-6 and the SPS-6 was then tested using traditional
methods for their convergent validity in an independent clinical sample and a student sample, as well as
for their sensitivity to change and diagnostic sensitivity in the clinical sample. The scores on the SIAS-6
and the SPS-6 correlated as well as the scores on the original SIAS and SPS, with scores on measures
of related constructs, discriminated well between those with and without a diagnosis of social phobia,
providing cutoffs for diagnosis and were as sensitive to measuring change associated with treatment as
were the SIAS and SPS. Together, the SIAS-6 and the SPS-6 appear to be an efficient method of
measuring symptoms of social phobia and provide a brief screening tool.

Keywords: social phobia, social interaction anxiety, short screening tools, item response theory, validity

The Social Interaction Anxiety Scale (SIAS) and the Social (1998) were the first to show high internal consistency (Cron-
Phobia Scale (SPS; Mattick & Clarke, 1998) are a companion bach’s .94) and good test–retest reliability (r .91) across
set of measures that aim to assess two related yet distinct facets 4-week and 12-week intervals. Both scales have excellent con-
of social phobia. The scales were developed in response to the struct validity with high and significant correlations displayed
apparent lack of comprehensive and valid measures designed to between scores on the SIAS and SPS and scores on scales that
assess specific scrutiny fears (i.e., eating, drinking, writing, and measure similar features of social phobia (Mattick & Clarke,
using public toilets in the presence of others) and the more 1998; Ries et al., 1998). Furthermore, multivariate analyses of
generalized social interaction anxieties (i.e., talking to friends variance have shown scores on the SIAS and SPS can ade-
and strangers, attending a social gathering/party), which are quately discriminate between patients with social and other
now recognized as the core features of social phobia in the anxiety disorders (Peters, 2000) and can reliably measure
Diagnostic and Statistical Manual of Mental Disorders (4th ed.; change throughout cognitive behavioral therapy (Mattick &
DSM–IV; American Psychiatric Association, 1994). The SIAS Peters, 1988; Mattick, Peters, & Clarke, 1989). As a result, the
and SPS have since displayed favorable psychometric proper- SIAS and SPS have been used extensively since their develop-
ties in both clinical and nonclinical samples. Mattick and Clarke ment in both research and clinical practice.
One disadvantage, which has the potential to limit the use of the
SIAS and SPS, is the length of administration. The SIAS and SPS
This article was published Online First July 11, 2011. take on average 10 to 15 min to complete. The SIAS/SPS are too
Lorna Peters and Ronald M. Rapee, Centre for Emotional Health, long to ever be used in epidemiological research where there is
Department of Psychology, Macquarie University, North Ryde, New South pressure to reduce respondent burden. Even in treatment outcome
Wales, Australia; Matthew Sunderland and Gavin Andrews, Clinical Re- research, there is increasing pressure to assess a broader range of
search Unit for Anxiety and Depression (CRUfAD), University of New outcome and mechanism constructs and hence assessment of each
South Wales at St. Vincent’s Hospital, Sydney, New South Wales, Aus- construct needs to be as short as possible. Similarly, in experimen-
tralia; Richard P. Mattick, National Drug and Alcohol Research Centre,
tal work, including a measure that takes 10 min to complete
University of New South Wales, Sydney, New South Wales, Australia.
This research was supported in part by grants awarded to Gavin An-
reduces the number of additional constructs and experimental
drews and Ron Rapee by the National Health and Medical Research procedures that can reasonably be included. In the clinical setting,
Council. the pressures of shorter consultation times, particularly in primary
Correspondence concerning this article should be addressed to Lorna care settings, it seems likely that clinicians will benefit from more
Peters, Centre for Emotional Health, Department of Psychology, Macqua- economical methods to regularly monitor and screen for social
rie University NSW 2019, Australia. E-mail: [email protected] phobia.
66
67 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 67

The need for more economical forms of assessment in both Parametric (e.g., Bock, 1972; Rasch, 1960; Samejima, 1979)
research and clinical practice has been confirmed by the prolifer- and nonparametric (e.g., Ramsay, 1991; Samejima, 1998) item
ation of many short-form screening instruments in the recent response models can be applied when plotting the response func-
literature. The design of a short-form instrument is characterized tions of individual items. The primary difference between them is
by shortening or refining a more comprehensive measurement in that parametric methods place a stronger assumption on the spe-
order to generate a smaller scale, with virtually the same psycho- cific mathematical form of the item response function (e.g., a
metric properties but with the advantage of taking on average 2–3 logistic and monotonic structure), while nonparametric procedures
min to administer and score (Coste, Guillemin, Pouchot, & Fer- hold no such a priori assumptions (Ramsay, 2000). Instead, non-
manian, 1997). At the present time there appears to be a lack of parametric models obtain information about the item response
short-form instruments available to researchers and clinicians that function directly from the data without forcing the data to conform
validly and reliably measure the severity of both social interaction to a specific model. Furthermore, nonparametric methods offer the
anxiety and scrutiny fears associated with social phobia. Like the advantage of identifying items with nonmonotonic item response
majority of contemporary short-form instruments, the functions (Ramsay, 2000), for example, nonparametric curves
development of a short form of the SIAS and the SPS can be have the ability to decrease and increase and can change rapidly
assisted by the use of item response theory (IRT). Unlike classical along different levels of the latent trait. Although it has been
test theory, IRT requires a stronger set of assumptions when shown that both methods tend to result in similar item selection
evaluating test items, namely, that a single common latent factor (Lei, Dunbar, & Kolen, 2004), it was decided to apply a nonpara-
can account for all covariance amongst items (unidimensionality) metric approach for the current study because we had no prior
and that items and persons can be measured in relation to their knowledge regarding the mathematical form of the item response
position on this latent factor (Embretson & Reise, 2000). functions within the measurement of social phobia. Meijer and
IRT provides a statistical framework for evaluating Likert-type Baneke (2004) recommended a nonparametric approach when
test items by graphically plotting the probability of selecting each routinely investigating psychopathology data, particularly during
response option of a particular item as a mathematical function of test construction when the researcher requires more information
the underlying latent trait. The graphs that are produced are known about how the items are functioning, as is the case with the current
as option characteristic curves (OCC), and the performance of a study. Previously, nonparametric approaches have been success-
given item is measured by interpreting the slope and shape of the fully applied to estimate the item characteristics and construct
curves. For example, the SIAS and the SPS each contain 20 items short forms of the Social Phobia and Anxiety Inventory
rated on a scale from 0 to 4, where 0 indicates that the item is not (Roberson-Nay, Strong, Nay, & Beidel, 2007) and the Dysfunc-
characteristic of the individual and 4 indicates that the item is tional Attitude Scale (Beevers, Strong, Meyer, & Pilkonis, 2007).
extremely characteristic of the individual. That is, there are five The current study aims to apply nonparametric IRT to shorten or
possible response options for each item that indicate the severity refine the SIAS and the SPS in an attempt to create short, more
of social anxiety. The OCCs for a single item represent the economical screening scales for social interaction anxiety and
probabil- ity of a specific response option (0 – 4) being endorsed at performance scrutiny fears. To maintain the two psychological
any level of the construct of social anxiety; that is, an OCC is constructs measured by the SIAS and SPS, it was decided to
plotted for each of the five response options for each item. An conduct the IRT analyses separately on both measures and select
item with desirable properties should fully capture the construct the most discriminating items from both scales. As a means to
that it in- tends to measure and should adequately make multiple clarify that the good psychometric properties of the long forms are
discrimi- nations along the full continuum of the latent trait. Thus, carried over to the short forms, this study also reports results
a “good” item will have OCCs that change rapidly with changes in examining construct validity, sensitivity to diagnostic status and
the level of the social anxiety construct (Santor & Ramsay, 1998) sensitivity to change in an independent clinical sample of partic-
and that reveal that the probability of selecting a lower response ipants presenting to an anxiety disorders clinic and construct
option (e.g., 0 in the case of the SPS or SIAS) peaks at lower validity in a sample of undergraduate psychology students.
levels of the latent trait and systematically decreases at higher
levels of the latent trait, while the probability of selecting higher Method
options (e.g.,
5 in the case of the SPS or SIAS) systematically increases at
higher levels of the latent trait (Embretson & Reise, 2000). A Participants
poorly performing item fails to make multiple discriminations Archival data sets were used to allow development and valida-
amongst individuals along the latent trait, and therefore, the tion of the short forms of the SIAS and SPS. Participants were
likelihood of selecting the higher response options does not drawn from four independent sources. The first sample (Sample
systematically in- crease with higher levels of the latent trait. 1), used to develop the short forms, consisted of 456 patients
Thus, the OCCs tend to have a flat appearance and rarely intersect admitted to one of five randomized controlled trials (RCTs)
with each other sug- gesting that cases with high and low levels of examining an online clinician-assisted cognitive behavioral
the underlying latent trait tend to have a similar likelihood of therapy program for the treatment of social phobia, known as The
endorsing the same options (Ramsay, 2000). Poorly performing Shyness Program (Aydos, Titov, & Andrews, 2009; Titov,
items that do not make multiple discriminations across the full Andrews, Choi, Sch- wencke, & Johnston, 2009; Titov, Andrews,
continuum fail to provide an adequate level of information relating Choi, Schwencke, & Mahoney, 2008; Titov, Andrews, &
each individual to their underlying latent trait; thus, these items Schwencke, 2008; Titov, Andrews, Schwencke, & Drobny, 2008).
could be removed when constructing a short form without a The RCTs were a joint research initiative of the University of
substantial loss of test infor- mation. New South Wales and St.
 SIAS and SPS;
vali

confirmatory factor analysis; IRT

of theconvergent
68 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 68

Diagnostic
thelong sensitivity
SPSforms
and SIAS;
Vincent’s Hospital in Sydney, Australia. The sample contained
slightly more female participants (58%) than male participants,

and
with an average age of 39 years (SD 11.7, range 18 –72). The

shortof

Participants in the clinical samples all met diagnostic criteria for Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association) social phobia.
selection criteria for all RCTs required that each participant have

theforms

RCT randomized controlled trial; CBT cognitive behavioral therapy; CFA

a diagnosis of DSM–IV social phobia, as confirmed by the Mini
International Neuropsychiatric Interview (MINI; Sheehan et al.,

long
and on
1998). All five RCTs were approved by the St. Vincent’s Hospital

scores
Analyses
ethical review board, and each participant was required to com-

the short
plete informed written consent. Data were provided in anonymous

between
format for the current analysis.

CFA; IRT modeling

The second clinical sample (Sample 2) consisted of 446 (216

scores on
female) patients with an average age of 34.15 years (SD 10.78;

Relationship
range 18 –75) treated in one of several treatment trials for social
phobia at the Emotional Health Clinic (EHC; formerly known as

Relationship between
the Macquarie University Anxiety Research Unit), Macquarie Uni-
versity, Sydney, Australia. The patients from this clinical sample
all met inclusion criteria for the trials into which they were
enrolled which included that they had a primary diagnosis of
social phobia as measured by the Anxiety Disorders Interview
Schedule–IV (ADIS–IV; Di Nardo, Brown, & Barlow, 1994) and

16.45

15.64

5.82
SD

13.50
a Clinician Severity Rating of 4 or above (i.e., there was at least
moderate impairment caused by social phobia). The description of

SPS
the trials is given in more detail in Gaston, Abbott, Rapee, and

34.45

34.40

6.04
19.10
Neary (2007); Rapee, Abbott, Baillie, and Gaston (2007); and

M
Rapee, Gaston, and Abbott (2009). All of the participants were
randomly allocated to one of several treatments that were being

11.59

13.69

8.16
compared, with 370 receiving an active CBT treatment. All trials
SD

13.36 14.26
SIAS

received approval from the Macquarie University Human Re-

search Ethics Committee. 54.79

52.96
The third sample (Sample 3) were 137 (79 female) participants

13.99 24.77
M

who were community volunteers with a mean age of 35.76 years

(SD 13.99; range 18 –75) for a number of research projects at
Age (years)

11.72

10.78

the EHC. The diagnostic status of these participants was ascer-

Phobia Scale; SIAS Social Interaction Anxiety Scale.

assisted CBT
SD

tained via the ADIS–IV. None had a diagnosis of social phobia. 4.83
The fourth sample (Sample 4) were 164 (115 female) under-
credit
at the 39.09

446 (230) 34.15

35.76

graduate psychology students with a mean age of 20.54 years

20.54
N (n male) M

(SD 4.83; range 17– 47) who completed the SIAS and SPS as
trialsclinician

course

part of a number of different research projects conducted by Lorna

(191)

(58)

Peters. All of the students participated in return for course credit.

volunteers
for
(49)

The samples and their functions in the article are described in

456
online

164137
participating

Table1.
; Treatment
sampleexamining
Table 1

Community

Materials
EHC
Description of the Samples Used in the Analyses

Undergraduate students
a

SIAS/SPS (Mattick & Clarke, 1998). The SIAS and SPS are
; RCTs

companion scales designed to measure two related facets of social

fears/anxiety. The SIAS is a self-report measure consisting of 20
Clinical
Clinical sample a

statements that each respondent rates as characteristic of them

Source

using a 5-point scale, ranging from 0 to 4 (e.g., not at all charac-

teristic or true of me to extremely characteristic or true of me).
The statements reflect the level of general anxiety associated with
the initiation and maintenance of social interactions (e.g., meeting
and talking to strangers, friends, or members of the opposite sex).
Likewise, the SPS is a self-report measure consisting of 20 state-
ments using the same 5-point scale. The statements reflect the
experience of anxiety associated with the performance of various
Note.
Sample

tasks while being scrutinized by others (e.g., working, eating,

a

drinking, writing, using public toilets). The scores on the SIAS

and SPS have been shown to possess desirable psychometric
properties
69 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 69

with a high level of internal consistency (Cronbach’s 0.90), to graphically express the probability of endorsing each option of
high test–retest reliability (r 0.91; Mattick & Clarke, 1998; an SIAS or an SPS item as an ordinal monotonic function of the
Osman, Gutierrez, Barrios, Kopper, & Chiros, 1998), and adequate underlying dimension of social anxiety (either interaction fears in
discriminant and construct validity (Brown et al., 1997; Heimberg, the case of the SIAS or scrutiny fears in the case of the SPS),
Mueller, Holt, Hope, & Liebowitz, 1992; Mattick & Clarke, 1998; presented in the OCCs. Briefly, the kernel smoothing approach
Peters, 2000). Furthermore, studies of the factor structure have first ranks each patient on the basis of their summed raw scores on
consistently revealed that the SIAS and SPS tap two distinct yet the items of the measure (in this case, the SIAS or the SPS). Each
related constructs of social anxiety (Olivares, Garcia-Lopez, & rank is then converted to quantile scores of the standard normal
Hidalgo, 2001; Osman et al., 1998). distribution; these values are used as the values on the latent trait
Brief Fear of Negative Evaluation (bFNE; Leary, 1983). ( ; in this case, social anxiety). A new variable is then defined,
The bFNE is a 12-item self-report scale measuring fear of negative called an indicator variable, which assigns a value of 1 to each
evaluation. Participants rate how characteristic a statement is of patient if they choose a particular response option from a
them on a 5-point scale (1 not at all characteristic to 5
particular item and a value of 0 if they do not. Finally, a
extremely characteristic). The scores on the bFNE have good
smoothing kernel, a form of local averaging, is applied to
concurrent validity with the scores on the Fear of Negative Eval-
estimate the probability of selecting a particular option from a
uation Scale (Watson & Friend, 1969) and demonstrate adequate
particular item within different regions of the underlying level of
test–retest reliability (Leary, 1983).
social anxiety, by smoothing the relationship between the binary
Depression Anxiety Stress Scales—21-Item Version (DASS-
indicator vector and the latent trait vector (Ramsay, 2000;
21; Lovibond & Lovibond, 1995). The DASS-21 is a self-
Ramsay, 1991; Santor & Ramsay, 1998). To achieve the aim of
report measure of depression, anxiety, and stress/tension. Partici-
shortening the SIAS and the SPS, any items that failed to
pants rate the extent to which they feel each of 21 statements about
adequately discriminate cases along the full range of the
common symptoms of depression, anxiety, and stress apply to
particular underlying social anxiety construct (inter- action fears
them on a 4-point scale, from 0 (does not apply at all) to 3
or scrutiny fears) were removed, therefore improving the
(applies very much or most of the time). Total scores on the three
efficiency of the scales while maintaining the desirable psy-
seven- item subscales range from 0 to 21, with higher scores
chometric properties. To assist with the selection of items, we
indicating greater symptomatology. The DASS-21 scores
inspected the item characteristic curves (ICCs), which plot the
demonstrate an excellent factor structure, distinguishing clearly
overall performance of each item rather than each individual
between features of depression, physical arousal and psychological
option, as is the case with the OCC. The ICCs are interpreted by
tension, and agitation. Scores on the DASS-21 have also been
examining the expected response option on the item along the
shown to have good internal consistency and concurrent validity
y-axis, in this case there are five response options, ranging from 0
(Antony, Biel- ing, Cox, Enns, & Swinson, 1998).
to 4, with the standardized normal latent trait score on the x-axis,
in this case severity of the underlying social anxiety construct
Statistical Analysis
(either interaction fears or scrutiny fears). The ICCs for each item
Unidimensionality. One of the assumptions implicit in the were judged to have poor discriminating properties if the likeli-
item response model used in the current study is that all items load hood of endorsing the more severe options did not systematically
sufficiently on to one underlying construct (Hays, Morales, & increase with higher levels of the underlying social anxiety con-
Reise, 2000), often referred to as essential unidimensionality struct; thus, the ICC takes a flat appearance. Furthermore, we
(Stout, 1990). Both scales were originally designed to assess two required that each item successfully make multiple discriminations
separate yet related facets of social phobia (social interaction fears (the point where the probability of selecting one option becomes
in the case of the SIAS and scrutiny fears in the case of the SPS). more likely than another or the difficulty thresholds) across the
To assess this assumption, data from both scales were combined full range of underlying social anxiety observed in the
and analyzed by fitting a two-factor confirmatory factory analysis majority of patients under investigation (i.e., between the fifth
(CFA) of polychoric correlations using a robust diagonally and 95th per- centile). Graphically, this requires the ICC of each
weighted least squares method of estimation (Jöreskog, 1990). The item to steadily increase from the lowest response option (0) to
choice of model fit statistics was informed by previous studies that the highest re- sponse option (4) along increasing levels of the
suggest the chi-square goodness-of-fit statistic is overly sensitive underlying social anxiety construct, thus forming a steep slope.
to minor differences in large samples (Browne, MacCallum, Kim, The selection pro- cess proceeded in an iterative manner,
Anderson, & Glaser, 2002). As an alternative, Hu and Bentler meaning that items with poor discriminating ability were
(1998) suggested that good model fit is evidenced by a combina- removed before resubmitting the remaining items for further
tion of the Tucker–Lewis fit index (TLI 0.95), the comparative analysis. This process continued until all items exhibited
fit index (CFI 0.95), and the standardized root-mean-square favorable properties, and these items were
residual (SRMR 0.08).1 The statistical models were fitted to the selected as the final item set.
data using the LISREL computer software package Version 8.80 As a means to assist the item selection process further, it was
(Jöreskog & Sörbom, 2006). decided, prior to the analysis, to aim for a selection of items
totaling in the range of five to seven for each scale; therefore, the
Item Response Modeling pair of short measures, when administered together, would contain
A nonparametric Gaussian kernel smoothing approach, incor-
porated in the TESTGRAF software (Ramsay, 2000), was applied 1
Additional information, including the factor loadings and residual
variances for each item in the CFA, is available from the authors on
request.
70 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 70

10 to 14 items. This decision was based on the observed trend in upper end of social interaction anxiety severity, where the proba-
the current literature suggesting that useful, efficient, and accurate bility of endorsing the higher options suddenly increases. In the
short screening scales for the common mental disorders contain on analysis of the full set of SPS items, the ICCs indicated that Items
average 10 to 15 items and take 2 to 3 min to administer (Coste et 4, 7, 8, 15, 16, and 17 (see Items 7–12 in the Appendix) possessed
al., 1997). favorable discriminating properties across the full scrutiny fear
spectrum. Once again, the remaining SPS items did not appear to
Results adequately discriminate between respondents at the lower levels of
the latent trait. The ICCs for two items are presented in Figure 1
Confirmatory Factor Analysis to demonstrate the selection criteria and the overall differences
between curves of a selected and unselected item. For example,
The results from the CFA conducted using data from Sample 1 Item 10 of the SPS (“I would find it difficult to drink something if
indicated that a two-factor model provided an adequate level of fit
in a group of people”) does not appear to discriminate between the
(SRMR 0.08, CFI 0.95, TFI 0.94). The correlation between
item’s response options (y-axis), until the 75th percentile as indi-
the two latent factors was 0.61 (SE 0.04). Items from each scale
cated by the flat slope and sharp increase toward the upper end of
loaded on their respective factor, with loadings above 0.3 for all
the latent trait (x-axis), while Item 15 (“I worry I might do
items, except for the three reverse scored SIAS items (Item 5, “I
something to attract the attention of other people”) makes multiple
find it easy to make friends of my own age”; Item 9, “I am at ease
discriminations between the fifth and 95th percentile, as indicated
meeting people at parties, etc.”; and Item 11, “I find it easy to
by the steady increase from the first response option to the last
think of things to talk about”), with factor loadings of 0.12,
across the full social anxiety spectrum. As mentioned previously,
0.13, and
the analyses continued in an iterative process so that the most
0.16, respectively. This suggests that the three reverse scored
discriminating items were selected and then resubmitted for anal-
items possibly confused the respondents and may form a separate
factor (cf. Rodebaugh, Woods, & Heimberg, 2007). The data were ysis. The selected items from both scales continued to exhibit
rean- alyzed with the three reverse-scored items forming a third adequate discriminating properties and were chosen as the final
factor; however, the results did not offer a substantial item sets for the short forms of the SIAS and SPS. The six items
improvement in model fit (SRMR 0.076, CFI 0.95, TFI from the SIAS and the six items from the SPS form two scales—
0.95). Therefore, applying the principle of parsimony, the two- the SIAS-6 and the SPS-6 (see Appendix). The ICCs for the final
factor model was selected as the best fitting model. The results of items taken from the SPS and SIAS are presented in Figures 2 and
the CFA provide a reasonable justification for analyzing the two 3, respectively.
scales separately. The test information curves for the short forms and the long
forms were also produced and are presented in Figure 4 to provide
a greater comparison between the long and short form in regard to
IRT Modeling
test performance. Briefly, the test information curves describe the
Consistent with the findings from the CFA, two separate anal- amount of information the overall test provides at various points
yses were conducted, one for the SIAS and one for the SPS. In the on the latent trait (social interaction anxiety in the case of the
analysis, using the data from Sample 1, of the full set of SIAS SIAS and scrutiny fears in the case of the SPS). Higher
items, inspection of the ICCs indicated that Items 2, 4, 6, 8, 10, information values at a particular trait level indicates that the
and test provides more information and in turn is more precise
13 (see Items 1– 6 in the Appendix) possessed favorable discrim- regarding an indi- vidual’s latent trait at that level. Typically,
inating properties across the full social interaction anxiety spec- scales of psychopathol- ogy have a flat test information curve so
trum. The remaining SIAS items did not appear to adequately that the scale provides
discriminate between respondents at the lower levels of the latent
trait. In these items the ICCs remained relatively flat until the

Figure 1. Nonparametric item characteristic curves of Social Phobia Scale (SPS) Item 10 (unselected item: “I
would find it difficult to drink something if in a group of people”) and SPS Item 15 (selected item: “I worry I
might do something to attract the attention of other people”).
71 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 71

Figure 2. Nonparametric item characteristic curves for the six-item Social Phobia Scale (SPS-6) items taken
from the SPS. The standard (Std.) normal latent trait is scrutiny fears operationalized by the SPS. The item score
(y-axis) is the response option on a scale from 0 to 4, where 0 indicates that the item is endorsed as not
characteristic of the individual and 4 indicates that the item is endorsed as extremely characteristic of the
individual.

similar information across the full range of the latent trait contin- Further comparison of the SIAS-6 and SPS-6 with the SIAS and
uum (Ramsay, 2000). As can be seen in Figure 4, the full form SPS was conducted in the student sample (Sample 4; n 164).
scales tend to provide more information across each full latent trait The scores on the SIAS-6 and on the SIAS correlated significantly
continuum as indicated by relatively flat curves. On the other (r .89, p .001), as did the scores on the SPS-6 and the SPS
hand, the short form test information curves indicate that the short (r .93, p .001).
scales are most informative for individuals within the fifth and
95th percentiles, as evidenced by a loss of information at the
extreme low and high ends of each latent trait. Convergent Validity
In the EHC clinical sample (Sample 2), the correlations of
Relationship Between the Short and Long Forms of scores on the SIAS 6 and on the SPS-6 with scores on measures of
the SPS and SIAS related constructs (fear of negative evaluation [bFNE], general
anxiety [DASS–A], stress [DASS–S], and depression [DASS–D])
The correlations between each shortened scale (SIAS-6 and were compared with the correlations of scores on the SPS and
SPS-6) and the long forms of the SIAS and SPS were examined at SIAS with scores on the same related constructs at pre-treatment.
three time points (pre-treatment, post-treatment, and at 3-month The correlations are presented in Table 2. As shown in Table 2,
follow-up) in an independent clinical sample (the EHC sample; scores on the SIAS-6, the SPS-6, the SIAS and the SPS had
Sample 2). Scores on the SIAS-6 correlated strongly and signifi- significant and moderately strong relationships with scores on the
cantly with scores on the SIAS at pre-treatment (r .88, p measures of related constructs. In addition, the correlations of the
.001), at post-treatment (r .92, p .001), and at 3-month SIAS-6 and SPS-6 with the related variables did not differ from
follow-up (r .89, p .001). Similarly, scores on the SPS-6 the correlations of the SIAS with the related variables or the SPS
correlated strongly and significantly with scores on the SPS at with the related variables ( ps .05)
pre-treatment (r .92, p .001), at post-treatment (r .94, p In the student sample (Sample 4), the correlations of scores on
.001), and at 3-month follow-up (r .93, p .001). These the SIAS-6 and the SPS-6 with scores on the SIAS and the SPS
correlations between the short forms and their counterpart long and the bFNE were examined. The SIAS-6 correlated significantly
forms suggest that the short forms are tapping similar constructs to with the bFNE (r .57, p .001), as did the SPS-6 (r .60, p
that captured in the original SIAS and SPS. .0001). The correlation with the bFNE did not differ between the
72 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 72

Figure 3. Nonparametric item characteristic curves for the final six-item Social Interaction Anxiety Scale
(SIAS-6) items taken from the SIAS. The standard (Std.) normal latent trait is social interaction anxiety
operationalized by the SIAS. The item score (y-axis) is the response option on a scale from 0 to 4, where 0
indicates that the item is endorsed as not characteristic of the individual and 4 indicates that the item is endorsed
as extremely characteristic of the individual.

SIAS-6 and the SIAS (r .57 vs. r .65), t(161) 2.84, p .05, (SE .007; 95% CI .95–.98). Since none of the 95%
nor between the SPS-6 and the SPS (r .60 vs. r .66), t(159) confidence intervals around the AUC values include a value of
2.69, p .05. 0.5, scores on all of the tests are performing better than chance in
discriminating between those who have a diagnosis of social
Diagnostic Sensitivity phobia and those who do not have a diagnosis of social phobia. A
comparison between the SIAS-6 and the SIAS indicates that
To examine whether the SIAS-6 and the SPS-6 were equivalent
the AUCs are significantly different (difference between AUCs
to the SIAS and the SPS in terms of their ability to detect diag- .09; 95% CI
nostic status, a receiver operating characteristic (ROC) analysis
.009 –.03; p .001). Similarly, a comparison between the SPS-6
was carried out. ROC analyses compared the ability of the SIAS-6, and the SPS indicates that the AUCs for the two forms are also
the SPS-6, the SIAS, and the SPS to discriminate between those
significantly different (difference between AUCs .04; 95% CI
with an ADIS–IV diagnosis of social phobia (EHC clinical .02–.05; p .001). Thus, while the scores on the SIAS-6 and the
sample; Sample 2 n 446) and those without an ADIS–IV
SPS-6 significantly discriminate between those who meet diagnos-
diagnosis of social phobia (Sample 3; n 137 control tic criteria for social phobia and those who do not, the scores on
participants). The area under the ROC curve (AUC) summarizes
the short scales do not discriminate between the diagnostic groups
the ability of the instru- ment to discriminate between those with as well as the SIAS. The optimum cutoff score for discriminating
the disorder of interest and those who do not have the disorder
between those with and without a diagnosis of social phobia is 7 or
(Hanley & McNeil, 1982). The AUC value can also be used to higher on the SIAS-6 (sensitivity 84.86; 95% CI 81.2– 88.0;
compare the diagnostic performance of two or more tests (Griner,
specificity 97.67; 95% CI 93.4 –99.5) and 2 or higher on the
Mayewski, Mushlin, & Greenland, 1981). MedCalc for Windows, SPS-6 (sensitivity 87.97; 95% CI 84.6 –90.8; specificity
Version 11 (MedCalc Software, Mariakerke, Belgium) was used to
86.05; 95% CI 78.8 –91.5).
calculate and com- pare the AUCs for the tests. The AUC for
the SIAS-6 was .97 (SE .006; 95% confidence interval [CI]
.96 –.98), for the SPS-6 was .93 (SE .01; 95% CI .91–.95), Sensitivity to Change
for the SIAS was The change in the SIAS-6 and the SPS-6 scores and in SIAS and
.99 (SE .003; 95% CI .98 –1.0), and for the SPS was .97 SPS scores from pre-treatment to post-treatment and from pre-
73 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 73

Figure 4. Test information curves for the long Social Interaction Anxiety Scale (SIAS; top left panel), short
SIAS (top right panel), long Social Phobia Scale (SPS; bottom left panel), and short SPS (bottom right panel).
The standard (Std.) normal latent trait is social interaction anxiety in the case of the SIAS and scrutiny fears in
the case of the SPS.

treatment to 3-month follow-up was examined in the EHC clinical follow-up scores for the SIAS-6, the SPS-6, the SIAS and the SPS.
sample. There were significant reductions from pre-treatment to The standardized residualized change scores for pre- to post-
post-treatment and pre-treatment to 3-month follow-up in scores treatment were strongly and significantly correlated (r .90, p
on the SIAS-6, the SPS-6, the SIAS, and the SPS (all ts 9.4, p .001) for the SIAS-6 and the SIAS and (r .93, p .001) for the
.001). To examine whether the change measured by the SIAS-6, SPS-6 and the SPS. The standardized residualized change scores
the SPS-6, the SIAS and the SPS was similar, standardized residu- for pre-treatment to 3-month follow-up were strongly and signif-
alized change scores were created using pre-treatment scores to icantly correlated (r .89, p .001) for the SIAS-6 and the SIAS
predict, first, the post-treatment scores and, second, the 3-month and (r .92, p .001) for the SPS-6 and the SPS. These

Table 2
Correlations Between the SIAS-6, SPS-6, SIAS, and SPS and Theoretically Related Constructs at Pre-Treatment in the EHC Sample

Measure SIAS-6 SPS-6 SIAS SPS DASS-D DASS-A DASS-S bFNE

SIAS-6 — .51 .88 .54 .48 .32 .38 .42

SPS-6 — .54 .92 .33 .50 .43 .42
SIAS — .56 .54 .33 .43 .55
SPS — .35 .56 .45 .45
DASS-D — .43 .57 .38
DASS-A — .59 .37
DASS-S — .42
bFNE —
M 13.16 10.51 52.96 34.4 18.48 15.82 20.93 50.64
SD 4.88 6.36 13.69 15.64 10.70 8.02 8.95 7.61

Note. SIAS Social Interaction Anxiety Scale; SPS Social Phobia Scale; DASS-D Depression scale from the Depression Anxiety Stress Scales;
DASS-A Anxiety scale from the Depression Anxiety Stress Scales; DASS-S Stress scale from the Depression Anxiety Stress Scales; bFNE Brief
Fear of Negative Evaluation scale. All correlations significant ( p .001).
74 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 74

correlations suggest that the change in SIAS scores and in SPS validity from a variety of samples, which were independent from
scores did not differ between the original measures and their the sample used to develop the shortened scale. This suggests that
corresponding short forms. the SIAS-6 and the SPS-6 can generalize to other populations of
interest (samples in clinical trials, nonclinical volunteer and stu-
Discussion dent samples) and is not specific to the sample from which it was
developed. Second, on the basis of the CFA, which suggested that
The aim of this research was to develop short forms of two a two-factor solution fit the data best, it was decided that items
widely used measures of social phobia, the SIAS and the SPS from the SIAS and from the SPS would be examined separately in
(Mattick & Clarke, 1998), without sacrificing the validity of the the IRT analysis. Thus, the validity results are also for two short
measures. Using data pooled from a number of randomized con- six-item scales (the SIAS-short and the SPS-short), rather than for
trolled trials of online treatment for social anxiety, a a combined short form. In order to determine whether the two
nonparametric IRT approach was adopted to develop two short scales can be combined into a single score, examination of
shortened scales con- taining six items from the SIAS and six the structure of the items in the SIAS-6 and SPS-6, by CFA, is
items from the SPS. Thus, the SIAS-6 and the SPS-6 retain the warranted in future research. Third, the evaluation of the SIAS-6
breadth of measurement across both interaction anxiety and and the SPS-6 in a student sample was limited to correlations
scrutiny fears afforded by the original measures. Using classical between the shortened forms, the SIAS and SPS, and the bFNE.
test theory methods, the SIAS-6 and SPS-6 scores were tested for Further examination of the psychometric properties of the short-
their validity in three samples (clinical, control, and student) ened forms in a nonclinical sample is required to assess its utility
independent from the sample used in the IRT analysis. The high across the continuum of social anxiety scores.
correlations between scores on the SIAS-6 and the SPS-6 and The loss of clinical utility of the briefer SIAS-6 and SPS-6
scores on the original SIAS and SPS suggest that the shortened compared with the original scales should be recognized. By re-
forms tap similar constructs to the original forms. In addition, the ducing the original SPS to six items, some of the more unusual
scores on the SIAS-6 and SPS-6 appear to function at least as well and specific feared situations that present in some cases of social
as scores on the original SIAS and SPS in terms of sensitivity phobia are discarded. These include eating, drinking, or writing in
to change over time in social phobia symptoms in a sample public and, for a small subset of often quite disabled/distressed
of people treated with CBT for social phobia. The correlation of men, the use of public toilets. That is, the content representative-
the shortened forms with related constructs (specifically general ness of the scales has been narrowed. (It is likely that their specific
anxiety, stress, and depression, as measured by the DASS, and fear nature has made them less discriminating in the samples herein
of negative evaluation, as mea- sured by the bFNE) was as high, or and led to their dropping from the reduced scales.) However, these
higher, than the correlation of the SIAS and SPS with those related are important components of the presenting picture for many
constructs, suggesting that the shortened forms have good patients and render the clinician less aware of the nature of
convergent validity. Of particular relevance to settings where a specific fears. While this observation does not take away from the
brief tool screening for social anxiety is required, the ability of the utility of the SIAS-6 and the SPS-6 as brief and time-efficient
SIAS-6 and the SPS-6 to discriminate between those who have a measures, the interested clinician will need to enquire further in
diagnosis of social phobia and those who don’t is excellent, cases where a patient scores highly to understand the nature of the
although there is a small loss in diagnostic sensitivity when fear experi- enced. This problem is less relevant to the SIAS items,
compared with the longer SPS and SIAS. Taken together, the as social interaction anxiety is a more general set of fears. On a
results from the traditional methods of testing validity converge to related clinical note, it should be made clear that the shortened
suggest that the SIAS-6 and SPS-6 provide an accurate and forms or, indeed, the full-length versions of the SIAS and SPS are
efficient measure of social anxiety. not designed to provide a diagnosis of social phobia according to
Despite the good showing of the shortened forms a number of DSM–IV criteria. Neither the shortened forms nor the full-length
limitations in the research must temper the conclusions. First, it is versions address the full set of diagnostic criteria that require not
likely that the high correlations between the SIAS-6 and the SPS-6 only fear and avoidance of situations in which the person may be
and the SIAS and SPS are attributable to the fact that the partic- embarrassed but also that the individual is significantly impaired
ipants’ scores on the shortened forms were calculated using items and/or distressed by their fear. Where it is essential for a full
that were embedded in the longer SIAS and SPS, rather than using diagnosis to be made according to DSM–IV criteria, the shortened
an independent administration of the shortened scales. By corre- forms provide a brief and efficient method for identifying those
lating a score from part of an instrument (in this case the SIAS-6 who may have social phobia prior to a more costly and time-
and the SPS-6) with the score for the whole instrument (in this consuming comprehensive diagnostic assessment that assesses im-
case, the SIAS and the SPS), we have not taken into account any pairment and distress associated with fear and avoidance of social
systematic errors that may occur when administering the two situations.
scales independently from each other. Thus, future examination of In conclusion, using several large clinical samples, short ver-
the SIAS-6 and the SPS-6 might focus on whether independent sions of two commonly used measures of social anxiety have been
administrations of the shortened scales and the SIAS and SPS developed. The development of this efficient pair of scales, which
result in attenuation of the relationship between the measures. In retain the sound psychometric properties of the original long
addition, future examination might more carefully examine what forms, has the potential to improve the assessment of the severity
effect, if any, the apparently small loss in diagnostic sensitivity of social phobia by reducing respondent burden as well as improv-
that comes with the shorter forms has when compared with an ing clinical utility at a relatively minor loss in validity. Given that
independent administration of the full-length scale. Despite this mental health clinics and general practitioners often experience
limitation, the current study was able to provide some evidence of
75 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 75

reduced amounts of patient time, we envisage that the SIAS-6 and Jöreskog, K. G., & Sörbom, D. (2006). LISREL 8.8 for Windows [Com-
the SPS-6 would be ideal for use as screening tools as well as for puter software]. Lincolnwood, IL: Scientific International.
measuring treatment outcome in clinical samples. In addition, the Leary, M. (1983). A brief version of the Fear of Negative Evaluation Scale.
shorter administration time will enhance the assessment of severity Personality and Social Psychology Bulletin, 9, 371–375. doi:10.1177/
of social phobia in epidemiological studies, in treatment outcome 0146167283093007
research, and in experimental studies. Lei, P.-W., Dunbar, S. B., & Kolen, M. J. (2004). A comparison of
parametric and nonparametric approaches to item analysis for multiple-
choice tests. Educational and Psychological Measurement, 64, 565–587.
References doi:10.1177/0013164403261760
Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emo-
American Psychiatric Association. (1994). Diagnostic and statistical man-
tional states: Comparison of the Depression Anxiety Stress Scales (DASS)
ual of mental disorders (4th ed.). Washington, DC: Author.
with the Beck Depression and Anxiety Inventories. Behaviour Research
Antony, M. M., Bieling, P. J., Cox, B. J., Enns, M. W., & Swinson, R. P.
and Therapy, 33, 335–343. doi:10.1016/0005-7967(94)00075-U
(1998). Psychometric properties of the 42-item and 21-item versions of
Mattick, R. P., & Clarke, J. C. (1998). Development and validation of
the Depression Anxiety Stress Scales (DASS) in clinical groups and a
measures of social phobia scrutiny fear and social interaction anxiety.
community sample. Psychological Assessment, 10, 176 –181. doi:
10.1037/1040-3590.10.2.176 Behaviour Research and Therapy, 36, 455– 470. doi:10.1016/S0005-
Aydos, L., Titov, N., & Andrews, G. (2009). Shyness 5: The clinical effec- 7967(97)10031-6
tiveness of Internet-based clinician-assisted treatment of social phobia. Mattick, R. P., & Peters, L. (1988). Treatment of severe social phobia:
Australasian Psychiatry, 17, 488 – 492. doi:10.1080/10398560903284943 Effects of guided exposure with and without cognitive restructuring.
Beevers, C. G., Strong, D. R., Meyer, B., & Pikonis, P. A. (2007). Journal of Consulting and Clinical Psychology, 56, 251–260. doi:
Efficiently assessing negative cognition in depression: An item response 10.1037/0022-006X.56.2.251
theory analysis of the Dysfunctional Attitude Scale. Psychological As- Mattick, R. P., Peters, L., & Clarke, J. C. (1989). Exposure and cognitive
sessment, 19, 199 –209. doi:10.1037/1040-3590.19.2.199 restructuring for severe social phobia: A controlled study. Behavior
Bock, R. D. (1972). Estimating item parameters and latent ability when Therapy, 20, 3–23. doi:10.1016/S0005-7894(89)80115-7
responses are scored in two or more nominal categories. Psychometrika, Meijer, R. R., & Baneke, J. J. (2004). Analyzing psychopathology items: A
37, 29 –51. doi:10.1007/BF02291411 case for nonparametric item response theory modeling. Psychological
Brown, E. J., Turovsky, J., Heimberg, R. G., Juster, H. R., Brown, T. A., Methods, 9, 354 –368. doi:10.1037/1082-989X.9.3.354
& Barlow, D. H. (1997). Validation of the Social Interaction and Olivares, J., Garcia-Lopez, L. J., & Hidalgo, M. D. (2001). The Social
Anxiety Scale and the Social Phobia Scale across the anxiety disorders. Phobia Scale and the Social Interaction Anxiety Scale: Factor structure
Psychological Assessment, 9, 21–27. doi:10.1037/1040-3590.9.1.21 and reliability in a Spanish-speaking population. Journal of Psychoedu-
Browne, M. W., MacCallum, R. C., Kim, C. T., Anderson, B. L., & cational Assessment, 19, 69 – 80. doi:10.1177/073428290101900105
Glaser, R. (2002). When fit indices and the residuals are incompatible. Osman, A., Gutierrez, P. M., Barrios, F. X., Kopper, B. A., & Chiros, C. E.
Psycho- logical Methods, 7, 403– 421. doi:10.1037/1082-989X.7.4.403 (1998). The Social Phobia and Social Interaction Anxiety Scale: Eval-
Coste, J., Guillemin, F., Pouchot, J., & Fermanian, J. (1997). Methodological uation of psychometric properties. Journal of Psychopathology and
approaches to shortening composite measurement scales. Journal of Clini- Behavioral Assessment, 20, 249 –264. doi:10.1023/A:1023067302227
cal Epidemiology, 50, 247–252. doi:10.1016/S0895-4356(96)00363-0 Peters, L. (2000). Discriminant validity of the Social Phobia and Anxiety
Di Nardo, P., Brown, T., & Barlow, D. (1994). Anxiety Disorders Inventory (SPAI), the Social Phobia Scale (SPS) and the Social Inter-
Interview action Anxiety Scale (SIAS). Behaviour Research and Therapy, 38,
Schedule for DSM–IV—Lifetime version. Albany, NY: Graywind. 943–950. doi:10.1016/S0005-7967(99)00131-X
Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychol- Ramsay, J. O. (1991). Kernel-smoothing approaches to nonparametric item
ogists. Mahwah, NJ: Erlbaum. characteristic curve estimation. Psychometrika, 56, 611– 630. doi:
Gaston, J. E., Abbott, M. J., Rapee, R. M., & Neary, S. A. (2006). Do 10.1007/BF02294494
empirically supported treatments generalize to private practice? A Ramsay, J. O. (2000). Testgraf: A program for the graphical analysis of
benchmark study of a cognitive– behavioural group treatment multiple choice test and questionnaire data. Unpublished manuscript,
pro- gramme for social phobia. British Journal of Clinical Department of Psychology, McGill University.
Psychology, 45, Rapee, R., Abbott, M., Baillie, A., & Gaston, J. (2007). Treatment of
33– 48. doi:10.1348/014466505X35146 social phobia through pure self-help and therapist-augmented self-help.
Griner, P. F., Mayewski, R. J., Mushlin, A. I., & Greenland, P. (1981). British Journal of Psychiatry, 191, 246 –252.
Selection and interpretation of diagnostic tests and procedures. Annals doi:10.1192/bjp.bp.106.028167
of Internal Medicine, 94, 555– 600. Rapee, R., Gaston, J., & Abbott, M. (2009). Testing the efficacy of
Hanley, J. A., & McNeil, B. J. (1982). The meaning and use of the area under theoretically-derived improvements in the treatment of social phobia.
a receiver operating characteristic (ROC) curve. Radiology, 143, 29–36. Journal of Consulting and Clinical Psychology, 77, 317–327. doi:
Hays, R. D., Morales, L. S., & Reise, S. P. (2000). Item response theory 10.1037/a0014800
and health outcomes measurement in the 21st century. Medical Care, Rasch, G. (1960). Probabilistic models for some intelligence and attain-
38(9, Suppl. 2), II28 –II42.
ment tests. Copenhagen, Denmark: Paedagogiske Institut.
Heimberg, R. G. Mueller, G. P., Holt, C. S., Hope, D. A., & Liebowitz,
Ries, B. J., McNeil, D. W., Boone, M. L., Turk, C. L., Carter, L. E., &
M. R. (1992). Assessment of anxiety in social situations and being
Heimberg, R. G. (1998). Assessment of contemporary social phobia
observed by others: The Social Interaction Anxiety Scale and the Social
verbal report instruments. Behaviour Research and Therapy, 36, 983–
Phobia Scale. Behavior Therapy, 23, 53–73. doi:10.1016/S0005-
994. doi:10.1016/S0005-7967(98)00078-3
7894(05)80308-9
Roberson-Nay, R., Strong, D. R., Nay, W. T., & Beidel, D. C. (2007).
Hu, L. T., & Bentler, P. M. (1998). Fit indices in covariance structure
Development of an abbreviated Social Phobia and Anxiety Inventory
modeling: Sensitivity to underparameterized model misspecification.
(SPAI) using item response theory: The SPAI-23. Psychological Assess-
Psychological Methods, 3, 424 – 453. doi:10.1037/1082-989X.3.4.424
ment, 19, 133–145. doi:10.1037/1040-3590.19.1.133
Jöreskog, K. G. (1990). New developments in LISREL: Analysis of
Rodebaugh, T. L., Woods, C. M., & Heimberg, R. G. (2007). The reverse
ordinal variables using polychoric correlations and weighted least
of social anxiety is not always the opposite: The reverse-scored items of
squares. Quality and Quantity, 24, 387– 404. doi:10.1007/BF00152012
76 PETERS,
DEVELOPMENT
SUNDERLAND,
OFANDREWS,
THE SIAS-6RAPEE,
AND THE
ANDSPS-6
MATTICK 76

the social interaction anxiety scale do not belong. Behavior Therapy, 38, Titov, N., Andrews, G., Choi, I., Schwencke, G., & Johnston, L. (2009).
192–206. doi:10.1016/j.beth.2006.08.001 Randomized controlled trial of web-based treatment of social phobia
Samejima, F. (1979). A new family of models for the multiple-choice item without clinician guidance. Australian and New Zealand Journal of
(Research Report 79 – 4). Knoxville, TN: Department of Psychology, Psychiatry, 43, 913–919. doi:10.1080/00048670903179160
University of Tennessee. Titov, N., Andrews, G., Choi, I., Schwencke, G., & Mahoney, A. (2008).
Samejima, F. (1998). Efficient nonparametric approaches for estimating Shyness 3: An RCT of guided vs. unguided Internet based CBT for
the operating characteristics of discrete item responses. Psychometrika, social phobia. Australian and New Zealand Journal of Psychiatry, 42,
63, 111–130. doi:10.1007/BF02294770 1030 –1040. doi:10.1080/00048670802512107
Santor, D. A., & Ramsay, J. O. (1998). Progress in the technology of Titov, N., Andrews, G., & Schwencke, G. (2008). Shyness 2: Treating
measurement: Application of item response theory. Psychological As- social phobia online: Replication and extension. Australian and New
sessment, 10, 345–359. doi:10.1037/1040-3590.10.4.345 Zealand Journal of Psychiatry, 42, 595– 605. doi:10.1080/
Sheehan, D. V., Lecrubier, Y., Sheehan, K. H., Amorim, P., Janavs, J., 00048670802119820
Weiller, E., . . . Dunbar, G. C. (1998). The Mini-International Neu- Titov, N., Andrews, G., Schwencke, G., & Drobny, J. (2008). Shyness 1:
ropsychiatric Interview (M.I.N.I.): The development and validation Distance treatment for social over the Internet. A randomized controlled
of a structured diagnostic psychiatric interview for DSM–IV and trial. Australian and New Zealand Journal of Psychiatry, 42, 585–594.
ICD-10. Journal of Clinical Psychiatry, 59(Suppl. 20), 22–33. doi:10.1080/00048670802119762
Stout, W. F. (1990). A new item response theory modeling approach with Watson, D., & Friend, R. (1969). Measurement of social-evaluative
applications to unidimensionality assessment and ability estimation. anxiety. Journal of Consulting and Clinical Psychology, 33, 448 –
Psychometrika, 55, 293–325. doi:10.1007/BF02295289 457.

Appendix

SIAS-6 and SPS-6

Instructions: For each question, please circle a number to indicate the degree to which you feel the statement
is characteristic or true of you. The rating scale is as follows:

0 1 2 3 4

Not at all Slightly Moderately Very Extremely

characteristic or characteristic or characteristic or characteristic or characteristic or
true of me true of me true of me true of me true of me

1. I have difficulty making eye contact with others. 0 1 2 3 4

2. I find it difficult mixing comfortably with the people I work with. 0 1 2 3 4
3. I tense up if I meet an acquaintance on the street. 0 1 2 3 4
4. I feel tense if I am alone with just one person. 0 1 2 3 4
5. I have difficulty talking with other people. 0 1 2 3 4
6. I find it difficult to disagree with another’s point of view. 0 1 2 3 4
7. I get nervous that people are staring at me as I walk down the street. 0 1 2 3 4
8. I worry about shaking or trembling when I’m watched by other people. 0 1 2 3 4
9. I would get tense if I had to sit facing other people on a bus or train. 0 1 2 3 4
10. I worry I might do something to attract the attention of other people. 0 1 2 3 4
11. When in an elevator, I am tense if people look at me. 0 1 2 3 4
12. I can feel conspicuous standing in a line. 0 1 2 3 4

Note. Items 1– 6 are from the Social Interaction Anxiety Scale (SIAS), and Iitems 7–12 are from the Social Phobia Scale
(SPS).

Received June 22, 2010

Revision received May 16, 2011
Accepted May 17, 2011