The Sounds of Silence: Language, Cognition,

and Anxiety in Selective Mutism

KATHARINA MANASSIS, M.D., ROSEMARY TANNOCK, PH.D., E. JANE GARLAND, M.D.,
KLAUS MINDE, M.D., ALISON McINNES, PH.D., AND SANDRA CLARK, PH.D.

ABSTRACT
Objectives: To determine whether oral language, working memory, and social anxiety differentiate children with selective
mutism (SM), children with anxiety disorders (ANX), and normal controls (NCs) and explore predictors of mutism severity.
Method: Children ages 6 to 10 years with SM (n = 44) were compared with children with ANX (n = 28) and NCs (n = 19) of
similar age on standardized measures of language, nonverbal working memory, and social anxiety. Variables correlating
with mutism severity were entered in stepwise regressions to determine predictors of mute behavior in SM. Results:
Children with SM scored significantly lower on standardized language measures than children with ANX and NCs and
showed greater visual memory deficits and social anxiety relative to these two groups. Age and receptive grammar ability
predicted less severe mutism, whereas social anxiety predicted more severe mutism. These factors accounted for 38% of
the variance in mutism severity. Conclusions: Social anxiety and language deficits are evident in SM, may predict mutism
severity, and should be evaluated in clinical assessment. Replication is indicated, as are further studies of cognition and of
intervention in SM, using large, diverse samples. J. Am. Acad. Child Adolesc. Psychiatry, 2007;46(9):1187Y1195. Key
Words: selective mutism, anxiety disorders, social anxiety, language impairment, working memory.

Selective mutism (SM) is an intriguing condition found
at the interface between child psychopathology and
disorders of speech and language. SM occurs when
children fail to speak in certain social environments
(often at school) despite speaking in others (usually
home) (American Psychiatric Association, 1994).
Although considered relatively uncommon (Bergman
et al., 2002), this condition can be debilitating both
socially and academically and shows a high persistence
over time (Steinhausen and Juzi, 1996). Treatment
strategies have been developed pragmatically (reviewed
Accepted March 23, 2007.
Drs. Manassis and Tannock are with The Hospital for Sick Children and the
University of Toronto; Drs. Garland and Clark are with the University of
British Columbia, Vancouver; Dr. Minde is with McGill University, Montreal;
and Dr. McInnes is with the University of Windsor, Windsor, Ontario, Canada.
This work was supported by a grant from the Ontario Mental Health
Foundation.
Reprint requests to Dr. Katharina Manassis, Department of Psychiatry, The
Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, M5G
1X8, Canada; e-mail: katharina.manassis@sickkids.ca.
0890-8567/07/4609-1187Ó2007 by the American Academy of Child
and Adolescent Psychiatry.
DOI: 10.1097/chi.0b013e318076b7ab
in McInnes and Manassis, 2005), without reference to
an etiological understanding of SM. Improving this
understanding could lead to empirically based, poten-
tially effective treatments.
Two main theoretical accounts have been proposed
for important etiological factors in children with SM
who are being schooled in their native language:
anxiety, especially social anxiety, and language impair-
ment or broader developmental delays (Kristensen,
2000). Recent findings also suggest a deficit in the
auditory efferent system as an etiological factor (Bar-
Haim et al., 2004). It is postulated that this deficit
prevents children from desensitizing to their own
vocalizations and anxious children cope with this
problem through selective mutism.
Other theories are proposed for children who are
recent immigrants and may be particularly vulnerable to
SM (Bradley and Sloman, 1975; Elizur and Perednik,
2003; Toppelberg et al., 2005), but these children have
been the subject of little research. Elizur and Perednik
(2003), however, reported that children of recent
immigrants with SM had higher levels of social anxiety
and lower rates of neurodevelopmental problems than
children with SM from nonimmigrant families

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:9, SEPTEMBER 2007 1187

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
MANASSIS ET AL.
(N = 19). Recently immigrated non-English speakers
were not included in the present study because our
language measures would not be applicable to them.
Our group designed a unique protocol to system-
atically assess the language and learning abilities of
children with SM as well as their anxiety (Manassis
et al., 2003). Due to limited verbal expression in SM,
assessing such factors can be challenging. In a previous
pilot study comparing a small group of children with
SM with socially phobic children of similar age on this
protocol, we found greater language impairment in SM,
impairment in SM relative to normative scores on a task
requiring nonverbal working memory (NVWM), and
comparable levels of anxiety in the SM and socially
phobic groups (Manassis et al., 2003; McInnes et al.,
2004). The present study sought to replicate and
expand these findings by studying a larger group of
children with SM and including both normal and
anxious comparison groups. To do so, we included
academic centers in three metropolitan sites (Toronto,
Vancouver, and Montreal). No children from the pilot
study were included in the present report.
We further sought to determine whether oral
language, working memory, and social anxiety could
predict mutism severity. If so, then group differences
could be related to clinical impairment in SM and
intervention in these areas may be helpful for affected
children. Our study is unique in that no previous
studies of SM have systematically examined these three
factors within the same sample.
Hypotheses included children with SM will have
language deficits relative to nonmute anxious children
and controls without a psychiatric disorder, children
with SM will have NVWM deficits relative to nonmute
anxious children and controls without a psychiatric
disorder, children with SM and nonmute anxious
children will have higher levels of social anxiety than
controls without a psychiatric disorder, and language
ability, NVWM scores, and social anxiety level will
predict children`s degree of mutism as reported by
parents on a standardized inventory.
METHOD
Sample
After obtaining approval from research ethics boards at all of the
participating sites, the sample was recruited through specialized
anxiety disorders clinics in three urban areas. Children were
1188
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
physician referred and from all socioeconomic classes, with some
overrepresentation of more affluent relative to less affluent families
(mean Hollingshead level II; minor professional) and of white
families relative to other racial groups. There were no group
differences in socioeconomic status or racial composition.
The sample included 44 children with SM (32 girls, 12 boys), 28
children with Axis I anxiety disorders (ANX; 13 girls, 15 boys; 17
generalized anxiety disorder, 7 social phobia, 4 separation anxiety
disorder), 19 normal controls (NCs; 14 female, 5 male). There was a
trend toward a group gender difference on #2 ( p = .05; female
preponderance in mute and normal groups).
Inclusion Criteria for SM
Children ages 6 to 10 with a primary diagnosis of SM on
structured parent and (where possible) child interviews (Anxiety
Disorders Interview Schedule [ADIS; Silverman and Albano,
1996]), where primary was defined as the diagnosis associated
with the greatest impairment. All of the sites had already been
trained on this interview and been using it routinely for several
years. All of the interviewers were experienced child and adolescent
psychiatrists or child and adolescent psychologists. The duration of
the mutism had to be at least 1 month, but not the first month of
school. Children with comorbid conditions were included in the
study as long as the conditions did not account for the mutism.
Comparison children had to have a primary diagnosis of ANX on
ADIS and not have SM or (in NCs) did not meet criteria for any
Axis I disorder on this interview. NCs were recruited through
advertisements in the same geographic areas as subjects, and a
subsequent snowball technique (i.e., all of the control subjects were
asked to name a friend who may be interested in participating).
Exclusion Criteria
Children were excluded from the study if they were not primarily
English speaking because we could administer measures only in
English, and testing in the child`s nonnative tongue may confound
language results. We required a minimum of 2 years in an English
school or daycare setting to consider them primarily English
speaking. Our Montreal site was located in a predominantly
English-speaking area of the city. There were other languages
spoken in some homes in addition to English, including Cantonese
(8.3%), Portuguese (4.7%), Spanish (4.2%), and Tagalog (2.3%).
The proportion of homes where other languages were spoken did
not differ significantly between groups.
Children with an abnormal audiometric screen at 500, 2,000,
and 4,000 Hz (hearing impairment can contribute to mutism); with
mutism related to trauma, major depression, psychosis, or a
neurological condition; with nonverbal intelligence scores <80 on
the WISC-III; or children who were unable to complete measures
were excluded. Children already taking serotonin reuptake
inhibitors or other psychotropic medication were also excluded
because these medications may affect some of the cognitive measures
administered.
In practice, exclusions occurred only for children with SM: two
were already taking serotonergic medication and three were unable
to complete key measures (unresponsive to assessor, even with
nonverbal administration). Numerous non-English speakers with
SM were screened out at intake, so we do not have exact data on
these individuals.
J. AM . ACAD. CHILD ADOLESC. PSYCH IATRY, 46:9, SE PTE MBER 2007

Sample Size
Because this is the first study comparing anxiety, language, and
cognition in these diagnostic groups, we were interested in large
effects that would merit further investigation and clinical attention.
To detect a large effect ( f = 0.40), a minimum of 21 participants
per group was needed to maintain a power of 0.80 and ! of .05
(Cohen, 1988). More children in the clinical groups than in the
control group were seen, however, likely reflecting their families`
high motivation to access our centers.
Measures
After families provided informed consent, an assessment visit was
scheduled for study participation. In most cases, all of the measures
and diagnostic interviews were completed in 1 day, but some
children required a second visit to complete measures due to fatigue.
Measure administration was counterbalanced to minimize order
effects and selection of measures considered both psychometric
quality (well-validated, reliable measures preferred) and the need for
the measure to be completed without requiring verbal responses.
We attempted to keep measure administrators blind to group
assignment, but this was often not possible with SM, given the
obvious lack of oral communication.
Language. The Peabody Picture Vocabulary Test-III (PPVT-III;
Dunn and Dunn, 1997) is an extensively norm-based measure that
estimates children`s receptive vocabulary skills. The child is asked to
point to one of four pictures that matches a spoken word.
The Lindamood Auditory Conceptualization Test (LACT;
Lindamood and Lindamood, 1971) is used to assess children`s
phonemic awareness ability, which is a strong predictor of later
reading achievement. Two related abilities (one task each) are tested:
the ability to discriminate one speech sound from another and the
ability to perceive and compare the number and order of sounds
within spoken patterns. Children respond by selecting and
manipulating colored blocks that represent sounds and sound
patterns. Converted scores are calculated for each task, but not
scaled or standardized for age. The LACT has high test-retest
reliability (0.96) in the age group of interest and is highly predictive
of standardized reading and spelling scores.
The Test of Reception of Grammar (TROG; Bishop, 2003) is a
well-validated test of grammar that requires a pointing response to
select which of four pictures best represents a spoken sentence. For
example, the child is asked to choose which of four pictures best
matches the sentence BThey are sitting on the table.^ The pictures
include two children standing at a table, two children sitting on the
floor (lexical distractors), one child sitting on a table (grammatical
distractor), and two children sitting on a table (correct answer). The
TROG has been standardized on 2000 British children ages 4 to 12
years. It has demonstrated good discriminant validity in the age
group of interest, concurrent validity with measures of similar
linguistic concepts, and test-retest reliability (0.74).
NVWM. To allow a more detailed yet not overly taxing
examination of NVWM, we administered the spatial span and spatial
working memory subtests from the WISC-III-PI (also known as Corsi
Blocks Forward and Backward Test). This measure provides scaled
scores for both subtests, for ease of comparison to normative samples,
and was the main NVWM measure examined in the analyses.
The NVWM measure from our preliminary study (finger
windows forward and backward [Manassis et al., 2003]) was
repeated for replication and supplemented with a test of pure visual
memory to distinguish this function from NVWM.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:9, SEPTEMBER 2007
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
SELECTIVE MUTISM AND LANGUAGE
The latter measure, the Visual Patterns Test (Della Sala et al.,
1997), requires the subject to memorize a series of black and white
squares in grids of varying sizes. The subject then attempts to
reproduce each pattern by coloring a blank grid. The test does not
require the visual locations to be sequenced (as in the Corsi Blocks
Forward and Backward Test) and thus provides a purer test of visual
memory. Good test-retest reliability has been shown (0.73Y0.75).
Distinction from the Corsi Blocks Forward and Backward Test in
both normal and clinical subjects has been demonstrated, suggesting
it is a measure of visual as opposed to spatial nonverbal short-term
memory.
Anxiety. The ADIS (Silverman and Albano, 1996) is a
semistructured diagnostic interview focusing on childhood anxiety
disorders as described in DSM-IV, but screening for all of the
major child psychiatric diagnoses. It includes clinician severity
ratings, reflecting the impact of the disorder on the child`s life.
It has shown high concurrent validity with established anxiety
measures such as the Multidimensional Anxiety Scale for Children
(MASC; Rynn et al., 2006) and is used extensively in research
on childhood ANX. The ADIS was attempted with children
and parents, but given the nature of SM, most information was
parent derived.
Reports from two informants were obtained for each ques-
tionnaire measure because correspondence between informants is
usually only fair for internalizing psychopathology (reviewed in
Barbosa et al., 2002). The questions were read aloud to the child by
the examiner if the child needed help with reading, and children
were allowed to point to the response. Children in all age groups
under 8 years generally required help with reading. Data from the
two informants were analyzed separately for both measures.
The Social Anxiety Scale for Children-R (SASC, parent and
child; La Greca and Stone, 1993) was included because SM has been
linked most closely to social anxiety. The SASC consists of 22 items
using a 5-point Likert scale (from not at all to all of the time)
assessing children`s anxiety in relation to peers and social situations.
The items are summed to yield three subscales: Fear of Negative
Evaluation, Social Anxiety Specific, and Social Anxiety General.
Acceptable concurrent validity, internal consistencies (0.60Y0.90)
and test-retest reliability over 4 months (0.70) have been established
(La Greca and Stone, 1993).
The Multidimensional Anxiety Scale for Children (MASC, parent
and child; March, 1997) is a 39-item, 4-point Likert scale to assess a
broad spectrum of anxiety symptoms in children. Four robust factors
have been found by factor analysis: physical symptoms, social anxiety,
harm avoidance, and separation anxiety. High internal reliability
(itemYtotal correlations of 0.4Y0.8) and test-retest reliability at 3
weeks and 3 months have been demonstrated. Convergent validity
specific to other anxiety measures (versus measures of depression or
externalizing behavior) has been established. Parents completed the
measure as they understood their child`s symptoms (consistent with
March, 1997), but results were analyzed separately by informant
because previous studies have found only marginally significant
parentYchild concordance (March, 1997).
The Selective Mutism Questionnaire (SMQ; Bergman et al.,
2001) was also included. This is a specific parent- and teacher-report
measure of mutism severity. Parents or teachers report the child`s
speaking behaviors and difficulties associated with lack of speech in
three domains (at school, at home/with family, and in public) on a 4-
point rating scale. They indicate the speaking behavior occurred
never/not at all to always/extremely in a given domain (i.e., lower
scores indicated less speech). The three factors have satisfactory
internal consistency and significant correlations with corresponding
1189
MANASSIS ET AL.
interference items, and the sum of all factors correlates significantly
with overall interference rating (Bergman et al., 2001). Cronbach !
values in our preliminary study were SMQ (total) = .87, SMQ
(school) = .85, SMQ (home) = .76, and SMQ (public) = .82.
Maternal report was most consistently available in this sample, and
thus it was used in all of the analyses.
Data were analyzed using the Statistical Package for the Social
Sciences (SPSS, version 12). Groups were compared using analyses
of covariance for continuous measures (see below) and # 2 for
categorical measures.
RESULTS
Group Differences in Comorbidity, Severity, and Age
Twenty-seven of the children with SM also met
criteria for social phobia, and these children had more
severe social anxiety by maternal report (t = 2.94,
p < .01), but not self-report, than those without social
phobia. Additional comorbidities in this group included
oppositional defiant disorder (three children) and specific
phobia (one child). Seven children in the anxiety group
had more than one disorder (all were anxiety disorders).
The role of comorbidity was only examined in selectively
mute children with and without social phobia, given the
small number of other comorbidities.
Mean clinician severity ratings on ADIS were higher
for children with SM than children with ANX (t = 2.34,
p < .05) and consistent with a markedly disturbing/
disabling (severe) problem in children with SM versus a
definitely disturbing/disabling (moderate) problem in
children with ANX.
The mean age was 8.48 years (SD 1.51), with
children with SM significantly younger than the other
groups (Table 1); mean performance IQ was 107.80
(SD 15.59) with a trend toward the performance IQ in
NCs being significantly higher than other groups (p =
.09; Table 1). To account for group differences, all of
the analyses for measures that were not age standardized
included age as a covariate. Sex differences were also
examined for all of the analyses. None were found, but
this may have been due to sample size constraints.
Testing Main Hypotheses
To test the first hypothesis concerning language
deficits in children with SM relative to children with
ANX and NCs, language measures were compared
across groups using analyses of variance (all measures
age standardized), and significant group differences
were probed with post hoc Tukey tests (Table 1). For all
three language measures, children with SM performed
1190
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
more poorly compared to children with ANX and NCs
( p = .001 for PPVT post hoc comparisons; p < .05 for
LACT and TROG post hoc comparisons). The
percentage of children with SM with clinically sig-
nificant scores (i.e., >1 SD below the mean) was also
greater than for the ANX and NC groups on all three
measures (PPVT: 18%, 10%, 12%, respectively; LACT
(total): 34%, 20%, 12%, respectively; TROG: 43%,
20%; 12%, respectively) on # 2 analyses.
To test the second hypothesis concerning NVWM
deficits in children with SM relative to children with
ANX and NCs, measures of nonverbal function were
compared across groups using analyses of variance
(Table 1). Analysis of covariance was used (controlling
for age) for Backward Finger Windows, the only
measure that is not age standardized. Post hoc Tukey
tests for measures where groups differed significantly
revealed that children with SM had deficits relative to
both children with ANX and NCs on the visual
patterns test (p = .001 for both) and spatial span
forward (p = .025; p = .001, respectively). These are
both measures of the ability to hold (temporarily
store) spatial information in memory; the former does
not include a sequencing requirement, whereas the
latter does. Children with SM also had deficits relative
to NCs but not relative to children with ANX for
spatial span backward (p = .005), a measure of
NVWM. They also had a higher percentage of
children with clinically significant scores (i.e., >1 SD
below the mean) for the visual patterns test and the
spatial span backward on # 2 analyses (20% and 34%,
respectively for children with SM, with other groups
<12%), but not the spatial span forward.
To test the third hypothesis concerning increased
social anxiety in SM and children with ANX relative to
NCs, groups were compared on a general measure of
anxiety (MASC) and on a specific measure of social
anxiety (SASC) on both mother and child report
(Table 1). Significant group differences were found. On
post hoc analyses, children with SM self-reported
greater social anxiety than the other two groups (p =
.003 for children with ANX; p = .001 for NCs), and
their mothers perceived more social anxiety than
mothers of NCs (p = .001) but not more than mothers
of children with ANX. Self-reported anxiety on the
MASC was greatest in children with ANX relative to
the other two groups (p = .020 for children with SM;
p = .025 for NCs). Mothers also perceived more anxiety
J. AM . ACAD. CHILD ADOLESC. PSYCH IATRY, 46:9, SE PTE MBER 2007
 SELECTIVE MUTISM AND LANGUAGE

TABLE 1
Means and Standard Deviations by Diagnostic Group
Measure Selective Mutism (n = 44) Anxiety Disorder (n = 28) Normal Controls (n = 19)
Age, y 7.87 (1.57) 9.09 (1.10) 9.07 (1.35)***
Performance IQ 106.77 (15.46) 105.23 (15.89) 115.00 (15.32)
Percentage nonnative English households 20.5% 21.4% 15.8%
SMQ (raw score)
School 1.45 (0.45) 3.08 (0.56) 3.56 (0.49)***
Family 3.03 (0.61) 3.69 (0.35) 3.94 (0.12)***
Other 1.80 (0.63) 2.86 (0.48) 3.57 (0.54)***
Total 6.28 (1.31) 9.63 (1.21) 11.08 (1.04)***
Language measures
PPVT-III (scaled score) 97.35 (14.76) 111.00 (17.86) 110.95 (13.64)***
LACT
1 (converted) 24.26 (5.64) 26.79 (2.30) 26.25 (1.95)
2 (converted) 41.10 (20.94) 52.97 (19.78) 54.75 (17.78)a
Total converted score 66.08 (24.80) 80.11 (20.94) 79.56 (17.73)*
TROG (standard score) 89.86 (19.14) 100.86 (14.20) 102.26 (15.92)**
Working memory
Sp. Span (scaled scores):
Forward 8.86 (3.04) 10.79 (3.57) 11.78 (2.76)**
Backward 9.34 (3.58) 9.93 (3.02) 11.89 (2.40)*
Finger Windows
Forward (scaled score) 10.00 (3.03) 9.61 (3.46) 11.58 (3.50)
Backward (raw score) 7.16 (4.10) 10.71 (4.12) 10.26 (3.81)a
Visual Patterns Test (scaled score) 67.89 (9.83) 77.18 (11.22) 81.74 (20.53)***
Anxiety measures
SASC (raw score)
Mother 56.06 (14.15) 52.58 (12.48) 36.44 (8.89)***
Child 48.80 (10.65) 40.71 (11.75) 39.79 (9.50)*
MASC (t score)
Mother 57.97 (18.08) 68.56 (11.67) 48.59 (15.60)***
Child 50.26 (11.39) 58.50 (9.97) 49.62 (11.70)*
Note: Direction of significant effects: lowest age, SMQ scores, language scores, and cognitive scores for SM versus other groups (except
spatial span backward where children with SM were comparable to children with ANX); highest MASC scores for ANX versus other groups;
higher SASC scores for SM and ANX (maternal report) groups than normal controls. SMQ = Selective Mutism Questionnaire; PPVT =
Peabody Picture Vocabulary Test; LACT = Lindamood Auditory Conceptualization Test; TROG = Test of Reception of Grammar; Sp. Span =
Spatial Span; SASC = Social Anxiety Scale for Children; MASC = Multidimensional Anxiety Scale for Children.
a
Significant, but not when controlled for age.
* p < .05; **p < .01; ***p < .001.

in children with ANX than in the other two groups (p =
.009 for children with SM; p = .001 for NC).
Exploration of Possible Predictors of SM
Children with SM differed from both comparison
groups on the SMQ, showing greater evidence of
mutism in all environments (p = .001 for all post hoc
comparisons; Table 1). This provides some evidence of
the construct validity of the measure. Because results
were so consistent across environments, only the total
SMQ was used in subsequent analyses. Bivariate
correlations between the SMQ and all other measures
that differentiated groups were examined (Table 2).
Significant negative correlations were found between
the SMQ and social anxiety by both mother and child
report (i.e., higher social anxiety associated with less
speech), and significant positive correlations were
found between the SMQ and measures of language
and NVWM (i.e., better performance on these
measures associated with more speech). These measures
were also highly intercorrelated and negatively corre-
lated with child report (but not mother report) of social
anxiety. Age was positively correlated with SMQ (i.e.,
increased speech with increased age).

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:9, SEPTEMBER 2007 1191

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
MANASSIS ET AL.

TABLE 2
Bivariate Correlations Between SMQ and Variables Differentiating Groups in Children With SM (n = 44)
Variable Age SASC (M) SASC (C) Sp. Span(f ) VPT LACT TROG PPVT-3 SMQ
Age 1
SASC (M) j0.110 1
SASC (C) j0.305** 0.335** 1
Sp. Span(f) 0.003 j0.211 j0.417** 1
VPT 0.525** j0.235 j0.377** 0.379** 1
LACT 0.425** .099 j0.354** 0.316** 0.474** 1
TROG 0.081 j0.073 j0.247* 0.317** 0.216* 0.502** 1
PPVT-3 0.154 j0.193 j0.325** 0.250* 0.302* 0.421** 0.664** 1
SMQ 0.403** j0.463** j0.401** 0.260* 0.394** 0.331** 0.401** 0.397** 1
Note: SASC = Social Anxiety Scale for Children; M = maternal; C = child; Sp. Span(f) = Spatial Span (forward); VPT = Visual Patterns Test;
LACT = Lindamood Auditory Conceptualization Test; TROG = Test of Reception of Grammar; PPVT = Peabody Picture Vocabulary Test;
SMQ = Selective Mutism Questionnaire.
* p < .05; **p < .01; ***p < .001.

Because we did not have any a priori hypotheses as to
which specific measures would be the best predictors of
SMQ, a series of stepwise regression analyses was done
to explore these predictors for the SM group only
(Table 3). No more than five predictors were entered in
a given analysis, given the statistical need to have at least
eight SM subjects per predictor (total number of
children with SM = 44). Entering age, mother-reported
social anxiety, and the TROG accounted for a larger
proportion of the variance than any other stepwise
regression, with an adjusted R 2 of 0.38.
Secondary Analyses
We were interested in the role of comorbid social
phobia in cognitive and linguistic aspects of SM,
recognizing that sample size was not adequate to test a
specific hypothesis in this area. Analyses of covariance
(controlling for age) for all cognitive and language
measures revealed that children with SM with social
phobia did better on the Backward Finger Windows
test for NVWM (F3,38 = 4.19, p < .05) than children
with SM without social phobia, but this finding was no
longer significant when applying the Bonferroni
correction for multiple comparisons.
We also reran all language and cognitive group
comparisons covarying for performance IQ. Significant
differences remained on two linguistic measures (PPVT
and TROG: F4,84 = 9.04, p = .001 and F4,84 = 4.79, p =
.011, respectively) with children with SM performing
worse than the comparison groups, and one NVWM
measure (spatial span forward F4,84 = 6.46, p = .002)
with SM performing worse than children with ANX
and NCs.

TABLE 3
Stepwise Regression Models for SMQ in Children With Selective Mutism (N = 44)
Model " t p F (Model) R 2 (Model) Adjusted R 2 SE
1. Constant 2.12 .037 16.41 0.19 0.18 2.03
Age 0.436 4.05 .000
2. Constant 3.99 .000 16.60 0.33 0.31 1.87
Age 0.396 3.97 .000
SASC (M) j0.370 j3.71 .000
3. Constant 1.70 .094 15.51 0.41 0.38 1.76
Age 0.362 3.82 .000
SASC (M) j0.361 j3.84 .000
TROG 0.288 3.05 .003
Note: SMQ = Selective Mutism Questionnaire; SASC (M) = Social Anxiety Scale for Children-Maternal; TROG = Test of Reception of
Grammar.

1192 J. AM . ACAD. CHILD ADOLESC. PSYCH IATRY, 46:9, SE PTE MBER 2007

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.

Finally, we re-ran all language and cognitive group
comparisons covarying for socioeconomic status. Sig-
nificant differences remained on two linguistic mea-
sures (LACT total converted and PPVT; F4,84 = 4.01,
p = .020 and F4,84 = 4.35, p = .018, respectively) and on
the visual patterns test (F4,84 = 4.67, p = .014), with
children with SM performing worse than comparison
groups.
DISCUSSION
Although our findings are limited by small compar-
ison groups and variability in age, the consistently
poorer performance of the SM group on linguistic
measures is striking. All three linguistic measures
differentiated groups with children with SM performing
worst. This difference remained significant for two of
three measures even when covarying for performance IQ
and socioeconomic level. This finding is consistent with
data from our previous study and other studies reporting
subtle language impairments in SM (Kristensen, 2000;
Kristensen and Oerbeck, 2006; Manassis et al., 2003;
McInnes et al., 2004). Detailed nonverbal linguistic
assessment may be indicated in children with SM (see
McInnes and Manassis, 2005 for further description).
The nature and origin of language deficits in SM also
merit further study. For example, auditory processing
deficits have been recently linked to SM (Bar-Haim
et al., 2004). It is unclear whether such deficits
represent the biological substrate for SM-related
language impairment or a separate problem. If, as
Bar-Haim and colleagues suggest, auditory efferent
activity is impaired in SM such that these children fail
to desensitize to their own vocalizations (and are
therefore overstimulated by them), this deficit would be
particularly relevant in processing extended speech
(versus single sounds or words). Consistent with this
hypothesis, we found increased proportions of children
with SM in the clinical range as the requirement to
process spoken language (and presumably repeat it
internally while selecting and executing the nonverbal
response) in each of the linguistic tasks increased (i.e.,
most for the TROG, least for the PPVT).
Group differences on cognitive measures were less
consistent. Measures of visual memory appeared to
differentiate children with SM from other anxious
children and normals but measures of NVWM did not
do so consistently. A specific deficit in visual memory
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:9, SEPTEMBER 2007
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
SELECTIVE MUTISM AND LANGUAGE
may contribute to social anxiety, for example, by
impairing recognition of faces or recognition of other
visual social cues. The high intercorrelation found
between visual memory and child-reported social anxiety
measures is consistent with this hypothesis. Another
possibility is that the motor demands of the tasks
(coloring a grid for the Visual Patterns test; manipulating
blocks for the Spatial Span) are problematic in SM (and
possibly in ANX as well), consistent with the finding of
Kristensen and Oerbeck (2006) that motor function was
a strong predictor of visual memory in SM. They found
reduced verbal memory span, but not reduced visual
memory span once results were controlled for motor
function. The lack of group differences on the Finger
Windows test, which makes fewer motoric demands, is
consistent with this hypothesis. There was also a trend
toward worse performance on this task for children with
SM without (rather than with) comorbid social phobia
that merits further investigation. Further studies of
language, cognition, auditory processing, and the
relationships among them are clearly needed to under-
stand etiological mechanisms of SM.
Our findings replicate the specific link proposed
between social anxiety (versus other forms of anxiety) in
SM. Consistent with previous studies (Kristensen,
2000), selectively mute children showed greater social
anxiety both by self- and parent report than normal
children or children with ANX. Overall anxiety as
reported on the MASC, however, was greatest for
children with ANX. Unfortunately, the number of
children with social phobia in the sample was too small
to determine whether SM represents an extreme of
social phobia as some have suggested (Anstendig,
1999), possibly because diagnosis was of necessity
based mainly on parent report. Our previous study
(Manassis et al., 2003) suggested that children with SM
were comparable to nonmute socially phobic children
in their level of social anxiety.
In this study the SMQ appeared promising in
discriminating SM from other childhood anxieties.
Therefore, it was used as a rough, quantitative measure
of the child`s degree of mutism, to examine potential
predictors of SM severity. Age, social anxiety (mother
report), and one language measure (TROG) accounted
for 38% of the variance in SMQ score among the
children with SM. This finding supports the need to
assess both social anxiety and language when such
children present clinically. It is possible that cognitive
1193
MANASSIS ET AL.
factors contributed to severity as well, but their
contribution was likely obscured by their correlation
with social anxiety or age.
The fact that age was predictive of greater speech
supports previous findings of gradual improvement of
SM over time but with infrequent complete remission
(Kolvin and Fundudis, 1981; Steinhausen and Juzi,
1996). Alternatively, families of older children with
severe SM or younger children with mild SM may have
been less motivated to participate in the study (i.e.,
sampling bias). Families of children with severe,
persistent SM may have given up on the possibility of
successful intervention, whereas families of young,
mildly affected children may have assumed that the
condition could still be outgrown. Consistent with this
idea, six of nine children with SM who did not follow
through on physician referral to the study were older
than the mean for SM study participants. It is also
possible that speech quantity in certain environments
increases with development in nonmute children as well
as in children with SM. Only prospective, longitudinal
studies of the development of children with SM relative
to their nonmute peers will determine the natural history
of this disorder and its response to various interventions.
Limitations
The small sample size and age differences between
groups constitute limitations of this study. A larger
sample would not only strengthen our findings but also
would allow relationships between various predictors of
SM to be examined using more rigorous methods (e.g.,
path analysis). The use of stepwise regression is a
limitation of the present study, although consistent
with the exploratory nature of this analysis. The
underrepresentation of boys and small number of
NCs constitute further sampling limitations.
The availability of questionnaires only in English also
eliminated a large segment of the SM population from
this study: those who are not attending school in their
native language (Bradley and Sloman, 1975). Alter-
natively, factors accounting for SM may differ in this
population. Elizur and Perednik (2003), for example,
found that SM in immigrant children was associated
with fewer developmental problems than SM in
nonimmigrant children, suggesting different pathways
to SM in these populations. Nonnative speakers
attending school for extended periods in English
(included in this study) may also differ from children
1194
Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.
with English as a first language. Teasing apart
symptoms of social phobia affecting speech from
symptoms of SM is also a measurement challenge in
this type of study.
Studies with more economically and racially diverse
samples are also indicated as well as children in rural
settings (all of the participating sites were urban).
Assessing the degree of mutism by multiple informants
and including more detailed linguistic testing would also
strengthen future studies.
Clinical Implications
Children with SM are often highly impaired, perhaps
even more so than those with ANX (as rated by
clinicians in our study). Speech and language assess-
ment appears important in the assessment of these
children. Collaboration between child and adolescent
psychiatrists and speech-language pathologists is there-
fore likely to serve them well (McInnes and Manassis,
2005). There may be a therapeutic role for language
remediation in SM, although this intervention awaits
further study.
Social anxiety must also be examined and addressed
in SM including reports from multiple informants.
There is already some evidence for anxiolytic medica-
tions in SM (Black and Uhde, 1994; Dummit et al.,
1996), particularly selective serotonin reuptake inhibi-
tors, and psychological interventions targeting social
anxiety are being developed and evaluated (reviewed in
McInnes and Manassis, 2005). The SMQ appears to be
a promising measure of SM severity linked to both social
anxiety and language. This measure may also serve as a
useful baseline for examining treatment outcomes.
Finally, clinicians must keep abreast of new findings
in relation to SM because a growing number of research
groups are examining potential etiological factors in
and potential interventions for this highly debilitating
condition.
Disclosure: Dr. Manassis obtains royalties from her book Keys to
Parenting Your Anxious Child, which was used in the clinical care of
some participating children and families. The other authors have no
financial relationships to disclose.
REFERENCES
American Psychiatric Association (1994), Diagnostic and Statistical Manual
of Mental Disorders, 4th edition (DSM-IV). Washington, DC: American
Psychiatric Association
J. AM . ACAD. CHILD ADOLESC. PSYCH IATRY, 46:9, SE PTE MBER 2007
 SELECTIVE MUTISM AND LANGUAGE

Anstendig KD (1999), Is selective mutism an anxiety disorder? Rethinking
DSM-IV classification. J Anxiety Disord 13:417Y434
Bar-Haim Y, Henkin Y, Ari-Even-Roth D, Tetin-Schneider S, Hildesheimer
M, Muchnik C (2004), Reduced auditory efferent activity in childhood
selective mutism. Biol Psychiatry 55:1061Y1068
Barbosa J, Manassis K, Tannock R ( 2002), Measuring anxiety: Parent-
child reporting differences in clinical samples. Depress Anxiety 15:
61Y65
Bergman RL, Keller M, Wood J, Piacentini J, McCracken J (2001), Selective
Mutism Questionnaire (SMQ): development and findings. Presented at
the Annual Meeting of the American Academy of Child and Adolescent
Psychiatry, Honolulu, October 23Y28
Bergman RL, Piacentini J, McCracken JT (2002), Prevalence and
description of selective mutism in a school-based sample. J Am Acad
Child Adolesc Psychiatry 41:938Y946
Bishop DVM (2003), Test for Reception of Grammar (Version 2): TROG-2
Manual. London: Psychological Corporation
Black B, Uhde T (1994), Treatment of elective mutism with fluoxetine: a
double blind placebo-controlled study. J Am Acad Child Adolesc
Psychiatry 33:1000Y1006
Bradley S, Sloman L (1975), Elective mutism in immigrant families. J Am
Acad Child Adolesc Psychiatry 14:510Y514
Cohen J (1988), Statistical Power Analysis for the Behavioral Sciences, 2nd ed.
Hillsdale, NJ: Erlbaum
Della Sala S, Gray C, Baddeley A, Wilson L (1997), Visual Patterns Test:
A Test of Short-Term Visual Recall. Suffolk, UK: Thames Valley Test
Dummit ES III, Klein R, Tancer NK, Ashe B, Martin J (1996), Fluoxetine
treatment of children with selective mutism: an open trial. J Am Acad
Child Adolesc Psychiatry 35:615Y621
Dunn LM, Dunn LM (1997), Peabody Picture Vocabulary Test III-Revised.
Circle Pines, MN: American Guidance Services
Elizur Y, Perednik R (2003), Prevalence and description of selective mutism
in immigrant and nonimmigrant families: a controlled study. J Am Acad
Child Adolesc Psychiatry 42:1451Y1459
Kolvin I, Fundudis T (1981), Elective mute children: psychological,
development, and background factors. J Child Psychol Psychiatry 22:
219Y232
Kristensen H (2000), Selective mutism and comorbidity with developmental
disorder/delay, anxiety disorder, and elimination disorder. J Am Acad
Child Adolesc Psychiatry 39:249Y256
Kristensen H, Oerbeck B (2006), Is selective mutism associated with deficits
in memory span or visual memory? An exploratory case-control study.
Depress Anxiety 23:71Y76
La Greca AM, Stone WL (1993), Social Anxiety Scale for Children-Revised:
factor structure and concurrent validity. J Clin Child Psychol 22:17Y27
Lindamood C, Lindamood P (1971), Lindamood Auditory Conceptualisation
Test Manual. R. Allen, TX: DLM Teaching Resources
Manassis K, Fung D, Tannock R, Sloman L, Fiksenbaum L, McInnes A
(2003), Characterizing selective mutism: is it more than social anxiety?
Depress Anxiety 18:153Y161
March J (1997), The Multidimensional Anxiety Scale for Children (MASC):
factor structure, reliability, and validity. J Am Acad Child Adolesc
Psychiatry 36:554Y565
McInnes A, Fung D, Manassis K, Fiksenbaum L, Tannock R (2004),
Narrative skills in children with selective mutism: an exploratory study.
Am J Speech Lang Pathol 13:304Y315
McInnes A, Manassis K (2005), When silence is not golden: an integrated
approach to selective mutism. Semin Speech Lang 26:201Y210
Rynn MA, Barber JP, Khalid-Khan S, et al. (2006), The psychometric
properties of the MASC in a pediatric psychiatric sample. Anxiety Dis
20:139Y157
Silverman WK, Albano AM (1996), Anxiety Disorders Interview Schedule for
DSM-IV. San Antonio, TX: Graywind
Steinhausen HC, Juzi C (1996), Elective mutism: an analysis of 100 cases.
J Am Acad Child Adolesc Psychiatry 35:606Y614
Toppelberg CO, Tabors P, Coggins A, Lum K, Burger C (2005),
Differential diagnosis of selective mutism in bilingual children. J Am
Acad Child Adolesc Psychiatry 44:592Y595

Sleep Arrangements and Behavior of Bed-Sharing Families in the Home Setting Sally A. Baddock, PhD, Barbara C.
Galland, PhD, Barry J. Taylor, MBChB, FRACP, David P.G. Bolton, MRCP, PhD

Objectives: We aimed to provide a quantitative analysis of the sleep arrangements and behaviors of bed-sharing families to further
understand the risks and benefits as well as the effects of infant age and room temperature on bed-sharing behaviors. Methods:
Forty infants who regularly bed shared with Q1 parent Q5 hours per night were recruited. Overnight video of the family and
physiological monitoring of the infant was conducted in infants` homes. Infant sleep position, potential for exposure to expired
air, head covering and uncovering, breastfeeding, movements, family sleep arrangements, responses to the infant, and
interactions were logged. Results: All infants slept with their mother. Fathers were included in 18 studies and siblings in 4. Infants
usually slept beside the mother, separated from the father/siblings (if present), facing the mother, with head at mothers` breast
level, touching, or with mother cradling. Median overnight breastfeeding duration was 40.5 minutes. Mothers commonly faced
their infant, but infants were rarely in a position that potentially exposed them to maternal expired air. Fathers were seldom in
contact with the infant during sleep. Of the 102 head-covering episodes observed in 22 infants, 80% were because of changes in
adult sleep position. Sixty-eight percent of head uncovering was facilitated by the mother; half of these events were prompted by
the infant. A 1-C increase in room temperature decreased infant head covering by 0.2 hours. Conclusions: The mother-infant
relationship is of prime importance during bed sharing, whether the father is present or not. The focus around breastfeeding
often dictates the sleep position of the infant and mother, though room temperature may also influence this. In colder rooms
infants tend to spend more time with their face covered by bedding. Frequent maternal interactions rely on the ability of the
mother to arouse with little stimulation. Mothers, perhaps impaired by alcohol, smoking, or overtiredness, may not be able to
respond appropriately. Pediatrics 2007;119:e200Ye207.

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:9, SEPTEMBER 2007 1195

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.