cover_v3_ manual cover_v2.

qxd 3/15/12 11:06 AM Page 1

Manual

The Social Communication Questionnaire

wps ®
SCQ Michael Rutter, M.D., F.R.S., Anthony Bailey, M.D., Catherine Lord, Ph.D.

Additional copies of this manual ( W- 381C) may be purchased from WPS.

Please contact us at 800-648-8857 or wpspublish.com.
W-381C
01-08_v3_01-08.qxd 3/15/12 11:09 AM Page 1
1
Introduction
The Social Communication Questionnaire (SCQ) is a
40-item, parent-report screening measure that taps the
symptomatology associated with autism spectrum disorder
(ASD). The items are administered in a yes/no response
format and can generally be completed by the parent
(or other primary caregiver) in less than 10 minutes and
scored by the administrator in under 5 minutes. There are
two forms available for the SCQ. The SCQ AutoScore™
Form: Lifetime (WPS Product No. W-381B) is completed
with reference to the individual’s entire developmental
history and produces results that are pertinent to referral for
more complete diagnostic workup. The SCQ AutoScore™
Form: Current (WPS Product No. W-381A) is completed
with reference to the individual’s behavior during the most
recent 3-month period, producing results that are pertinent
to understanding everyday living experiences and
evaluating treatment and educational plans.
The SCQ was originally designed as a companion
screening measure for the Autism Diagnostic Interview-
Revised (ADI-R). The final WPS Edition of the ADI-R
(WPS Product No. W-382; Rutter, Le Couteur, & Lord,
2003) is a 93-item structured interview that takes 11 ⁄ 2 to
2 1 ⁄ 2 hours to complete and elicits an authoritative
account of an individual’s developmental history
relevant to ASD. The SCQ items were deliberately
chosen to match the ADI-R items that were found to
have discriminative diagnostic validity. The relevant
content coverage of the SCQ, while briefer, is parallel
to that of the longer interview, making it suitable for
use with similar populations. The principal caregiver
who is most familiar with both the developmental
history and the current behavior of the individual who
is to be assessed should complete the SCQ. It is
applicable to subjects of any chronological age above
age 4.0 years provided that their mental age is at least
2.0 years. The ADI-R findings (see Rutter et al., 2003)
suggest that the applicability of that instrument may
extend down to a chronological age of 2.0 years as long
as the mental age exceeds 2.0 years. Because this
downward extension had not been tested systematically
with the SCQ at the time of this manual’s publication,
use in ages 2 years, 0 months to 3 years, 11 months
should be undertaken only with appropriate caution or
as new research findings become available.
1
No professional help is required to complete the
SCQ; a parent may be given the instrument and asked to
answer the questions without direct supervision. However,
overall use of the instrument and the development of
interpretations based on SCQ findings should be
undertaken with the supervision of an individual who has
professional training in the care and treatment of
individuals with ASD.
Like the other autism measures to which it is
related, such as the Autism Diagnostic Observation
Schedule (ADOS; Lord, Rutter, DiLavore, & Risi, 2001)
and the ADI-R (Rutter et al., 2003), the SCQ focuses on
behaviors that are rare in nonaffected individuals. For this
reason, the instrument does not provide conventional
scales (dimensions reflecting a continuum of some
unitary underlying skill, trait, or ability), nor can there be
useful norms (numerical estimates based on the
performance of people in the general population).
Instead, the supporting research evidence is presented in
the form of validation studies with clinical populations,
which use statistical analyses such as receiver operating
characteristics (ROC).
General Description
The primary, validated application of the SCQ
results in a single Total Score taken from the Lifetime
form that is then interpreted with reference to cutoff
scores drawn from the research reported on the
instrument. The cutoff identifies individuals who are
likely to suffer from an ASD and for whom more
extended evaluations should be undertaken. In addition,
subscores can be obtained that parallel the basic domains
of the larger ADI-R: Qualitative Abnormalities in
Reciprocal Social Interaction; Qualitative Abnormalities
in Communication; and Restricted, Repetitive, and
Stereotyped Patterns of Behavior. At the time of this
manual’s first publication, these subscores have not been
extensively researched and therefore cannot be used
in clinical application. However, in conjunction
with the Total Score, they may prove useful in the
evaluation of group differences in research usage.
Administration, scoring, and basic interpretation are
discussed in chapter 2 of this manual. Chapter 3 describes
01-08_v3_01-08.qxd 3/15/12 11:09 AM Page 2
2 The Social Communication Questionnaire (SCQ)
the development of the instrument, and the supporting
validation research is outlined in chapter 4. These more
extended discussions can be summarized as follows.
Main Uses
The SCQ provides a dimensional measure of ASD
symptomatology, with a cutoff score that can be used to
indicate the likelihood that an individual has an ASD. The
instrument has three main uses. First, in the primary clinical
application as described briefly above, it can be used as a
screening device with individual children in order to select
(through a focus on scores above the designated cutoff)
those who need a more thorough clinical assessment for a
possible ASD. The ADI-R and the ADOS are appropriate
tools to use for further assessment. It should be appreciated,
however, that (as with any screening questionnaire) there
will be some false negatives (i.e., scores below the cutoff
achieved by children who will prove to have an ASD when
they are assessed in detail). Depending on resources and
clinical or research needs, it may be desirable to undertake
a detailed clinical assessment of a child with a score just
below the cutoff, if other findings suggest that the diagnosis
of ASD needs serious consideration.
Second, SCQ scores can be used on a group basis to
compare overall levels of ASD symptomatology across
different samples. They might be used in this way to assess
ASD symptomatology in children with developmental
language disorders, medical conditions associated with
ASD, or learning difficulties (such as the fragile X
anomaly, tuberose sclerosis, or Down syndrome). Scores
can also potentially be used to chart changes over time in
ASD symptomatology in such groups.
Third, the scores can be used as an indication of the
approximate level of severity of ASD symptomatology,
either across groups or with respect to changes over
time, such as may be required for assessing the possible
benefits of therapeutic or educational interventions
(although this use has yet to be evaluated). If the SCQ
is to be used to measure change it will be necessary to
use the Current form, which applies to a recent
specified period of time (the previous 3 months).
Limitations
As is the case with any screening questionnaire,
the SCQ is not suitable for individual diagnosis. That
is because diagnosis requires information on onset,
course, and context pervasiveness/specificity (as well as
symptom presence). Also, caregiver reports must be
checked against direct clinical observation. Finally,
individual items on the SCQ necessarily rely on
respondent judgments rather than investigator concepts
(other than to the extent that such concepts can shape
questionnaire items). The last consideration is likely to
be particularly important when dealing with unusual
populations. For the same reason, the SCQ is not
suitable for the purpose of providing detailed
descriptions of patterns of behavior in individuals.
It should be noted that the SCQ is not a suitable
screening measure for use with very young children
(below a mental age of 2.0 years) who are in a
developmental phase during which there may be
clinically significant abnormalities that do not take the
full form required to meet the diagnostic criteria for
ASD (see Rutter et al., 2003).
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 13
4
Validity Evidence
In the original standardization study (Berument et al.,
1999), four steps were taken to assess the diagnostic
validity of the SCQ. First, a factor analysis was performed to
determine whether the scale provided a differentiation that
reflected the conceptualization of the three main domains
of abnormality found in autism (Reciprocal Social
Interaction; Communication; and Restricted, Repetitive,
and Stereotyped Patterns of Behavior). Second, the
combination of individual items was assessed by noting
their correlation with the total SCQ score and the extent to
which they differentiated ASDs (including autism) from
other diagnoses. Third, the correlations between the SCQ
and the ADI were calculated. Fourth, receiver operating
characteristic (ROC) curves were applied to determine the
degree to which the SCQ differentiated ASD from other
diagnoses. These analyses were repeated within IQ strata to
check whether the differentiation was affected by IQ level.
Standardization Study Evidence
Factor Analytic Analyses
The SCQ’s factor structure was explored (see Table 4).
Evaluation of three- and four-factor solutions for the 39
items (Items 2 through 40) suggested that a four-factor
model appeared to be the more meaningful. Principal
component factoring with varimax rotation yielded four
factors, which explained 42.4% of the total variation of the
SCQ data, with 24.3% accounted for by a social interaction
factor (Factor 1; eigen value 9.7); 8.7% by a communication
factor (Factor 2; eigen value 3.38); 5% by an abnormal
language factor (Factor 3; eigen value 1.94); and 4.5% by a
stereotyped behavior factor (Factor 4; eigen value 1.74).
The alpha reliability coefficient for the total scale was .90;
for Factor 1 it was .91, for Factor 2 it was .71, for Factor 3
it was .79, and for Factor 4 it was .67. All the individual
item-to-total score correlations were positive and mainly
substantial, in the range of .26 to .73 (23 of the 39 exceeding
.50). The extent to which the four factors mapped onto the
three key domains of autistic phenomena, as operationalized
in the ADI-R algorithm criteria, is indicated by the domain
designation of each item (see Figure 2 in chapter 2). Factor 1
(the social interaction factor) largely coincided with the
13
Reciprocal Social Interaction domain, and Factor 4 (the
stereotyped behavior factor) largely coincided with the
Restricted, Repetitive, and Stereotyped Patterns of Behavior
domain. The Communication domain items (of the ADI-R)
tended to divide between the other two factors, half being in
Factor 2, mainly reflecting communicative deficits, and half
in Factor 3, mainly reflecting abnormal language features,
with some also in Factor 1.
Item Validity
Table 5 shows the extent to which individual SCQ
items differentiated ASDs from other diagnoses, as given
clinically on the basis of standardized interview (ADI-R)
and observation data (ADOS) in the samples from which the
cases were drawn. Of the 39 scored items, 33 showed
statistically significant differentiation. Of the items that did
not, four concerned abnormal language features
(stereotyped utterances, inappropriate questions, pronoun
reversal, and neologisms). Each of these had a relatively
high frequency in non-ASD children but, as shown in
Table 4, had substantial correlations with the Total Score
(.64, .53, .45, and .57, respectively). Two items (self-injury
and unusual attachment to objects) differentiated only
at the 7% significance level; both showed only modest
correlations with the Total Score (.37 and .27, respectively).
Correlations between the ADI and the SCQ were
calculated for the Total Score and the ADI domain
(Reciprocal Social Interaction, Communication, and
Restricted, Repetitive, and Stereotyped Patterns of
Behavior) totals. Correlation coefficients were statistically
significant for all comparisons both within and across
domains (see Table 6).
Receiver -Operating Characteristics (ROC)
To assess the discriminant power of the SCQ, a series
of ROC analyses (Hanley & McNeil, 1982; Fombonne,
1991) and t-tests was carried out. The area under the curve
served as the index of accuracy (see Table 7 for detailed
numerical results). It appears that the discriminant ability of
the SCQ is high in differentiating ASD (including autism)
from non-ASD conditions (including mental retardation).
The SCQ similarly differentiates well between autism and
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 14

Table 4
Factor Loadings
Factor Loading
Correlation
Item Number Between Factor 1: Factor 2: Factor 3: Factor 4:
(Domain Item and Social Communication Abnormal Stereotyped
Designation)a Item Description Total Score Interaction Language Behavior
29 (S) offering to share .73 .73
36 (S) interest in children .60 .73
40 (S) group play .58 .71
37 (S) response to other children’s approaches .63 .70
34 (C) imitative social play .60 .69
31 (S) offering comfort .64 .65
28 (S) showing and directing attention .67 .65
30 (S) seeking to share enjoyment .61 .61
21 (C) imitation .63 .60
39 (S) imaginative play with peers .60 .59
22 (C) pointing to express interest .61 .55
27 (S) social smiling .55 .53
26 (S) eye gaze .58 .52
35 (C) imaginative play .58 .52
33 (S) range of facial expressions .52 .51
38 (-) attention to voice .54 .50
23 (C) gestures .39 .47
32 (S) quality of social overtures .48 .46
19 (S) friends .34 .38
17 (-) self-injury .37 .38
25 (C) head shaking to mean no .61 .64
2 (C) conversation .36 .62
24 (C) nodding to mean yes .60 .57
20 (C) social chat .34 .52
9 (S) inappropriate facial expressions .39 .51
15 (R) hand and finger mannerisms .36 .43
3 (C) stereotyped utterances .64 .80
7 (R) verbal rituals .63 .75
4 (C) inappropriate questions .53 .74
6 (C) neologisms .57 .69
5 (C) pronoun reversal .45 .51
12 (R) repetitive use of objects .51 .65
14 (R) unusual sensory interests .42 .64
8 (R) compulsions and rituals .44 .55
11 (R) unusual preoccupations .41 .50
10 (S) use of other’s body to communicate .30 .50
16 (R) complex body mannerisms .26 .38
18 (-) unusual attachment to objects .27 .35
13 (R) circumscribed interests .28 .32

a
S = Reciprocal Social Interaction domain; C = Communication domain; R = Restricted, Repetitive, and Stereotyped Patterns of Behavior domain;
- = not in algorithm.

14
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 15

Table 5
Item Validity Analysis
Percentage (%) Percentage (%)
Item With Abnormality: With Abnormality:
Number Item χ2 ASDa Other

29 offering to share 27.75**** 76.3 32.5

36 interest in children 24.34**** 79.4 40.0
40 group play 11.20*** 80.5 55.0
37 response to other children’s approaches 26.12**** 78.9 37.5
34 imitative social play 19.20**** 70.0 32.5
31 offering comfort 32.79**** 72.6 23.1
28 showing and directing attention 25.97**** 61.8 15.8
30 seeking to share enjoyment 16.34**** 63.0 27.5
21 imitation 19.94**** 70.7 32.5
39 imaginative play with peers 29.21**** 86.1 46.2
22 pointing to express interest 25.14**** 67.7 25.0
27 social smiling 10.45*** 51.8 23.1
26 eye gaze 19.85**** 65.0 25.6
35 imaginative play 38.91**** 73.4 20.0
33 range of facial expressions 19.64**** 54.1 15.0
38 attention to voice 15.72**** 62.4 27.5
23 gestures 14.67**** 66.9 33.3
32 quality of social overtures 18.30**** 40.8 5.0
19 friends 5.56** 71.1 51.3
17 self-injury 3.18† 40.3 25.0
25 head shaking to mean no 26.35**** 66.0 20.5
2 conversation 5.05* 32.5 13.5
24 nodding to mean yes 26.17**** 69.0 23.7
20 social chat 7.40** 17.4 0.0
9 inappropriate facial expressions 4.03* 27.8 12.5
15 hand and finger mannerisms 33.38**** 76.5 25.0
3 stereotyped utterances 1.57 81.8 72.2
7 verbal rituals 13.89**** 69.2 34.2
4 inappropriate questions .053 56.6 54.1
6 neologisms .97 47.1 37.8
5 pronoun reversal .002 52.5 52.9
12 repetitive use of objects 9.40** 66.5 40.0
14 unusual sensory interests 6.45 53.5 30.8
8 compulsions and rituals 5.24 69.6 50.0
11 unusual preoccupations 20.97**** 67.3 27.5
10 use of other’s body to communicate 8.91** 61.3 35.0
16 complex body mannerisms 12.71**** 61.4 30.0
18 unusual attachment to objects 3.23† 22.8 10.0
13 circumscribed interests 7.62** 54.4 30.0

a
ASD = autism spectrum disorder.
*p < .05. **p < .01. ***p < .001. ****p < .0001. †p = .07.

15
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 16

16 The Social Communication Questionnaire (SCQ)

Table 6
Correlations Between ADI and SCQ Totalsa
SCQ Scores

Repetition/
Reciprocal Social Stereotyped
ADI Scores Total Score Interaction Communication Behaviorb

Total Score .71*** .67*** .61*** .48***

Reciprocal Social Interaction .57*** .59*** .51*** .31***
Communication .61*** .62*** .55*** .36***
Repetition/ Stereotyped Behavior .63*** .48*** .47*** .59***
a
Based on ADI domains. bRestricted, Repetitive, and Stereotyped Patterns of Behavior.
***p < .0005.

Table 7
Discriminative Validity of the SCQ
SCQ Total Score
After Dropping
SCQ Total Score Nonsignificant Items

Diagnosis n Mean t-test AOCa Mean t-test AOC

ASDb (including autism) 160 22.28 8.73*** .86 19.88 9.46*** .88
vs.
non-ASD (including mental retardation) 40 11.18 8.88

Autism 83 25.24 11.01*** .94 22.51 11.99*** .95

vs.
nonautism (excluding mental retardation) 25 10.25 7.96

Autism 83 25.24 7.54*** .92 22.51 8.28*** .93

vs.
mental retardation 15 12.75 10.41

Autism 83 25.24 5.89*** .74 22.51 5.66*** .73

vs.
other ASDs 77 19.09 17.04
a
AOC = area under the curve. bASD = autism spectrum disorder.
***p < .0005.

mental retardation and between autism and non-ASD
diagnoses other than mental retardation. The SCQ also
shows differentiation between autism and other ASDs, but
there is substantial overlap and the differentiation is much
less clear-cut.
The analyses were repeated using an SCQ score that
did not include the six items that failed to differentiate
groups at the 5% level of statistical significance. Some
marginal improvement in discriminative validity
was obtained, but that between autism and other ASDs
was slightly worse. The discriminative validity of the SCQ
was compared with that of the ADI by contrasting the
areas under the ROC curve: for ASD versus non-ASD
they were, respectively, .88 and .87; for autism versus
mental retardation they were .93 and .96; and for autism
versus other ASDs they were .73 and .74. Because the
groups differed in IQ distribution, it was possible that the
SCQ diagnostic differentiation was an artifact, deriving
from the IQ differences. In order to determine whether
that was the case, the analyses were repeated within IQ
bands (see Table 8). Inevitably, as the data derived from
several different studies with different aims and covering
different age ranges, the IQ scores derived from several
different tests. For individuals of school age or older, the
appropriate Wechsler Performance Scale (Wechsler, 1989,
1991) was usually available; most of the younger children
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 17

Chapter 4 Validity Evidence 17

Table 8
ROCa Indexes and t-Tests of the SCQ for IQ-Matched Groups
SCQ Total Score

IQ Diagnosis n Mean SD t-test AOCb SEMc

IQ ≥ 70: ASDd 56 21.18 (8.11) 6.21*** .90 (.04)

Non-ASD 18 8.39 (5.63)

IQ = 50–69: ASD 46 22.53 (6.27) 5.14*** .90 (.05)

Non-ASD 10 11.40 (5.87)

IQ = 30–49: ASD 25 24.70 (5.61) 3.39*** .79 (.12)

Non-ASD 7 14.74 (10.43)

IQ ≥ 70: Autism 27 24.23 (7.11) 7.92*** .95 (.03)

Non-ASD 18 8.39 (5.63)

IQ = 50–69: Autism 28 25.08 (5.64) 6.51*** .96 (.03)

Non-ASD 10 11.40 (5.87)

IQ = 30–49: Autism 18 26.57 (3.93) 4.20*** .84 (.11)

Non-ASD 7 14.74 (10.43)

Note. IQ scores were derived from several different tests including the Wechsler Performance Scale, the Raven’s Matrices, the Mullen Scales
of Early Learning, and the Differential Abilities Scale.
a
ROC = receiver operating characteristic. bAOC = area under the curve. cSEM = standard error of measurement. dASD = autism spectrum disorder.
***p < .0005.

had been tested with Raven’s Matrices (Raven et al., 1986),
Mullen Scales of Early Learning (Mullen, 1995) or the
Differential Abilities Scale (Elliot, 1990). Because of this
test heterogeneity, any detailed assessment of IQ effects
would be inappropriate, but the data were adequate for the
use of broad IQ strata. The findings showed that in the
nonautistic group the mean SCQ score was lowest in the
group with an IQ above 70 (8.39) and highest in the group
with severe retardation (14.74) but did not vary by IQ
within the ASD (including autism) group. The diagnostic
differentiation within all IQ bands (including those with
severe mental retardation) was highly significant,
although it was clearest in the group with an IQ above 70.
Finally, analyses were undertaken to determine
whether individual behavioral domains of the SCQ
provided a better diagnostic differentiation than that
obtained with the Total Score. Individual items on the SCQ
were allocated to the three key domains of autistic
symptoms as determined by the equivalent items on the
ADI-R. The ROC analyses are summarized in Table 9. All
three domains provided satisfactory differentiation of ASD
(including autism) from other diagnoses, with areas under
the curve (AOC) ranging from .79 to .83. However, the
differentiation on the Total Score (AOC = .90) was better,
the difference for both Communication and Restricted,
Repetitive, and Stereotyped Patterns of Behavior domains
being statistically significant (z = 2.75 and z = 2.17,
respectively). The Restricted, Repetitive, and Stereotyped
Patterns of Behavior domain on its own was not very
good at differentiating autism from mental retardation
(AOC = .70) or autism from ASD (AOC = .59). The
results indicate that the most satisfactory differentiation
is provided by the SCQ Total Score.
Examination of the ROCs for the total SCQ suggested
scores of 15 or more as the standard optimal cutoff
for differentiating ASDs (including autism) from other
diagnoses. The sensitivity was .85, specificity was .75,
positive predictive value was .93, and negative predictive
value was .55 in this sample. Other cutoffs may be
preferable for general population samples and the
choice may also vary with the purpose; for example,
screening for case detection versus case collection. The
cutoff of 15 or more gave a sensitivity of .96 and a
specificity of .80 for autism versus other diagnoses
(with mental retardation excluded), and a sensitivity
of .96 and a specificity of .67 for autism versus mental
retardation.
As would be expected, a much higher cutoff (22 or
more) was required to separate autism from other ASDs,
with a sensitivity of .75 and a specificity of .60 at that point.
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 18

18 The Social Communication Questionnaire (SCQ)

Table 9
Diagnostic Differentiation of the Separate Symptom Domain Scores of the SCQ
Diagnosis n Domain Scoresa Mean t-test AOCb SEMc

ASDd 160 S – ASD 10.18 7.55*** .83 (.03)

Non-ASD 40 S – Non-ASD 4.68
C – ASD 6.93 6.49*** .79 (.04)
C – Non-ASD 3.93
R – ASD 7.57 6.56*** .79 (.04)
R – Non-ASD 3.85

Autism 82 S – Autism 11.77 9.16*** .91 (.04)

Non-ASD (excluding mental retardation) 25 S – Non-ASD 4.48
C – Autism 7.89 7.71*** .89 (.04)
C – Non-ASD 3.80
R – Autism 8.10 8.41*** .91 (.04)
R – Non-ASD 2.76

Autism 83 S – Autism 11.77 7.09*** .90 (.04)

Mental retardation 15 S – Mental retardation 5.00
C – Autism 7.89 5.38*** .83 (.07)
C – Mental retardation 4.15
R – Autism 8.10 2.96*** .70 (.08)
R – Mental retardation 5.67

Autism 83 S – Autism 11.77 5.35*** .72 (.04)

ASD (excluding autism) 77 S – ASD (excl. autism) 8.47
C – Autism 7.89 5.17*** .72 (.04)
C – ASD (excl. autism) 5.90
R – Autism 8.10 2.17* .59 (.05)
R – ASD (excl. autism) 7.00
a
S = Reciprocal Social Interaction; C = Communication; R = Restricted, Repetitive, and Stereotyped Patterns of Behavior. bAOC = area under the
curve. cSEM = standard error of measurement. dASD = autism spectrum disorder.
*p < .05. ***p < .0005.

ADI-R to SCQ Concurrent Validity
The associations between the ADI-R and the SCQ
were also examined in the Bishop and Norbury (2002)
sample of children with developmental language disorders.
The ADI-R was scored to provide a designation of autism
meeting the full DSM-IV/ICD-10 criteria (this applied to
8 out of the total sample of 21 and therefore 8 out of the 14
with ASD). It was also scored to provide a broad designation
of ASD, meaning that the criteria for two out of the three
domains were met. The agreement between the ADI-R and
the SCQ was much the same whether the narrow or broad
criteria were used. Of the 8 children meeting the full
ADI-R criteria for autism, 6 scored 15 or more on the SCQ.
Table 10 shows the pattern of intercorrelations between the
ADI-R and the SCQ for the three ADI-R domains.
For the Reciprocal Social Interaction domain, the Pearson
intercorrelation was .92, for the Communication domain
it was .73, and for the Restricted, Repetitive, and
Stereotyped Patterns of Behavior domain it was .89.
It should be noted, however, that within both the ADI-R and
the SCQ the cross-correlations between the Reciprocal
Social Interaction and Communication domains were high
(.77 for the SCQ and .70 for the ADI-R). The Restricted,
Repetitive, and Stereotyped Patterns of Behavior domain, by
contrast, showed only moderate correlation with the other
two domains, this being so for both the SCQ (.48 and .53)
and the ADI-R (.41 and .54).
Despite this good agreement at the domain level, the
item-by-item agreement between the SCQ and the ADI-R
was not impressive. ADI-R items were classified as present
if the score was 1, 2, or 3; these were then compared with
scores of 1 on the SCQ. Agreement ranged from 45% to
85%, with an average of 70.8%. Agreement was much the
same (71.1%) when the ADI-R codes of 0 and 1 were
combined and contrasted with 2.
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 19

Chapter 4 Validity Evidence 19

Table 10
Intercorrelations Among SCQ and ADI-R Domains
SCQ ADI-R
Social Communication Repetitive Social Communication Repetitive

SCQ
Social .77** .48* .82** .84** .35
Communication .53* .68** .73** .37
Repetitive .44* .59** .89**

ADI-R
Social .70** .41
Communication .54*
Repetitive

Note. Data from Bishop & Norbury, 2002.

*p < .05 (two-tailed). **p < .01 (two-tailed).

Table 11
More Intercorrelations Among SCQ and ADI-R Domains
SCQ ADI-R
Social Communication Repetitive Social Communication Repetitive

SCQ
Social .58** .61** .75** .49** .55**
Communication .48** .68** .77** .44**
Repetitive .44** .59** .55**

ADI-R
Social .65** .55**
Communication .53**
Repetitive

Note. Data from Murin, Choomchuay, & Bailey, 2003.

**p < .01 (two-tailed).

Convergent and Discriminant Validation Evidence
Similar analyses (using one randomly selected
individual from each family) were undertaken with the
International Molecular Genetic Study sample (data kindly
provided by Marianne Murin, Jom Choomchuay, and
Anthony Bailey). The Pearson correlation between the
SCQ and ADI-R Total Scores was .78. The intercorrelation
among domains is shown in Table 11, with agreements of
comparable strength to those in the Bishop and Norbury
sample (2002). Because the sample size was large (81 in
the randomly selected subsample and 160 in the total
sample), it was possible to determine whether the intercor-
relations varied by age, gender, language ability, and
performance IQ. It was found that they did not. As in the
Bishop and Norbury sample, the agreements between the
SCQ and the ADI-R were lower at the item level.
Comparing ADI-R scores of 1, 2, and 3 with SCQ scores of
1, agreement ranged from 36.6% to 91.9%, with an
average of 69.8%. The agreements were much the same
when the ADI-R codes of 0 and 1 were collapsed and
contrasted with SCQ scores of 2 (resulting in an average
agreement of 68.5%).
The design of this study, as exhibited in Table 11,
corresponds to a “multitrait-multimethod matrix.” This is
a classic internal validation design, first described by
Campbell and Fiske (1959). The current design represents
a good contrast in method, as it uses data collected in a
multihour interview with coded free responses and data
collected in a questionnaire that takes 10 to 15 minutes
to complete. The findings along the main diagonal of the
center rectangular portion of the matrix tell the important
story and are placed in interpretive context by comparison to
the off-diagonal and tail triangular portions of the matrix.
For the Reciprocal Social Interaction and
Communication domains, the validation support is
moderately strong: the same-domain, different-method
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 20
20
correlation coefficients are the highest values in the entire
matrix at .77 for the former and .75 for the latter (main
diagonal values involving these two domains). Their
contrast coefficients are lower as predicted in this
validation design. The different-domain, same-method
coefficients (the two triangular matrices at the upper left
and lower right) have values of .58, .61, and .48, along with
.65, .55, and .53. The different-domain, different-method
values (the off-diagonal values in the center rectangular
matrix) have values of .49, .55, and .44, along with .68,
.44, and .59. Thus the core domain construct provides a
stronger convergence of variance than does nonconstruct
convergent method variance.
For the Restricted, Repetitive, and Stereotyped
Patterns of Behavior domain, the support is weaker.
The same-domain, different-method correlation
coefficient is .55 and does not contrast strongly with the
different-domain, same-method coefficients or the
off-diagonal different-domain, different-method
coefficients. In this case common method variance is as
strong or stronger than common construct variance.
Whereas the result does not argue against the validity of
the Restricted, Repetitive, and Stereotyped Patterns of
Behavior domain, it does make it clear that this construct
appears less purely measured as evidenced by this design.
Psychometric Properties
At the time that this manual was in preparation, the
SCQ was a comparatively new instrument. Even so, it was
already in use in several large research screening studies.
The authors of one of these studies (Corsello, Leventhal,
& Cook, 2003) made available SCQ data on an early
portion of the data they were in the process of collecting.
This enabled the present authors to analyze the
psychometric properties of the SCQ Total Score
in a relatively large and diverse clinical sample.
The total number of cases analyzed was 214,
comprising 177 males and 37 females. The age range was
Table 12
SCQ Total Score Psychometric Properties by Age
Cases Number of
Group n Items
No language 57 33
2–4 years 37 39
5–6 years 49 39
7–10 years 41 39
11 and older 30 39
Note. n = 214.
a
SEM = standard error of measurement.
The Social Communication Questionnaire (SCQ)
from 2 through 18 years (median = 5 years, mean = 6.4
years, SD = 3.7 years). In the sample, 157 individuals
were credited with language by the parent who did the
rating (Item 1 = yes) while the remaining 57 were not so
credited (Item 1 = no). There was also preliminary
diagnostic information available on 213 of these cases.
Group 1 contained 71 individuals with diagnoses of
autism. Group 2 had 49 individuals with ASD (including
individuals with Asperger syndrome and pervasive
developmental disorder, not otherwise specified
[PDD-NOS] but not Rett syndrome or childhood
disintegrative disorder [CDD]). Group 3 had 37
individuals and included all nonspectrum diagnoses (for
the most part language impairment, mental retardation,
and ADHD). Each group had males and females in roughly
the same proportion as that seen in the whole sample,
and the full age range was also represented in each group.
The checks on the psychometric properties of
the SCQ were conducted in a stepwise fashion. The first
breakdown was essentially a developmental one: the
nonlanguage group was treated separately and the
language group was broken down into four groups by age
(each roughly the same size). The results of this analysis
can be seen in Table 12. It will be noted that the group
without language had the highest SCQ scores, whereas the
four subgroups of children with language had moderate
SCQ Total Scores that did not vary much by age. The total
difference across the four groups with language was only
about half a standard deviation and the difference between
adjacent groups was about one sixth of a standard
deviation. The alpha index of internal consistency was
also fairly uniform across groups, ranging from an
acceptable .84 to a good .93 with increasing age. It will be
noted that, although the standard deviations were higher
in the older age groups, the standard error of measurement
(SEM) was very uniform. All of these results suggest that
individual children are reliably measured by the SCQ
Total Score, which provides an accurate estimate of
underlying true scores.
SCQ Total Score
Mean SD alpha SEMa
18.1 (6.3) .84 2.5
13.3 (6.6) .84 2.6
14.9 (6.8) .84 2.7
15.8 (8.2) .89 2.7
16.5 (9.9) .93 2.6
09-22_v3_09-22.qxd 3/15/12 12:48 PM Page 21
Chapter 4 Validity Evidence
Table 13
SCQ Total Score Psychometric Properties by Diagnostic Classification
Cases Number of
Group n Items
Autism 71 39
ASD 49 39
Nonspectrum 37 39
Note. n = 157. Includes only children with language (SCQ Item 1 = yes).
a
SEM = standard error of measurement.
The second major analysis of psychometric
properties involved the breakdown by diagnostic groups;
in this case only those children with language
(Item 1 = yes) were analyzed, as the number of children
without language was too small to support separate
analysis by diagnostic group. Given the substantially
higher scores achieved by nonlanguage children in the
previously reported age and developmental analyses, it
appeared unwise to report psychometric properties on
groups that mixed subjects across this distinction. Table
13 shows the psychometric properties on the defined
diagnostic groups for the children with language. Here, as
expected, there are substantial differences in group means:
a full standard deviation between extreme groups and over
a half standard deviation separating any two groups. On
the other hand, the measures of internal consistency were
not so variable, again ranging from an acceptable .81 to a
good .92. And again, due to the parallel variation of score
standard deviations, the SEM was very similar from
group to group. These results again suggest that
obtained SCQ Total Score results will be an accurate
reflection of individual differences on the properties
measured by the screener.
Summary
The findings across the four studies may be
succinctly summarized as follows. The agreement
between the SCQ and the ADI-R at both the Total Score
and domain score levels is high, with the agreements
being substantially unaffected by age, gender, language
level, and performance IQ. The findings validate the SCQ
as a screening questionnaire and show that it provides a
21
SCQ Total Score
Mean SD alpha SEM a
18.3 (6.6) .81 2.3
13.7 (7.0) .86 2.6
10.6 (8.4) .92 2.47
reasonable index of symptom severity. The analysis of
psychometric properties indicates that the instrument can
be used as a reliable measure of individual differences on
the characteristics measured by the SCQ Total Score.
Agreement is, however, only moderate at the individual
item level. This means that the SCQ item responses should
not be used as a measure of individual behaviors.
Five main cautions need to be expressed with
respect to aspects of the SCQ that have yet to be
investigated. First, few data are available on its use with
children in the 2.0- to 4.0-year age range. Second, all the
findings concern children who have come to clinical
notice for one reason or another (not necessarily for an
ASD). It is not yet known how the SCQ would fare as a
screening measure in the general population for children
who had not been assessed clinically at all and for whom
the possibility of some form of developmental disorder
had not been raised by parents or professionals. Third,
there are few data on the ADI-R findings for children for
whom the SCQ preceded the ADI-R and for whom the
SCQ score was below 15. Accordingly, more data are
needed in order to determine the frequency of false
negatives on the SCQ. Fourth, there are few data on the
use of the SCQ with individuals who do not have
ASD but who do have social, communication, or
repetitive behavior problems of other types (such as
children with attachment disorders, attention deficit
disorders with hyperactivity, and obsessive-compulsive
disorders). Fifth, no data are available so far on the use
of the Current form of the SCQ to measure changes over
time (as, for example, in the assessment of treatment
efficacy).