Characterizing the Musical Phenotype in Individuals With Williams ...

Child Neuropsychology
2004, Vol. 10, No. 4, pp. 223–247
Characterizing the Musical Phenotype in Individuals
With Williams Syndrome
Daniel J. Levitin1, Kristen Cole2, Michael Chiles2, Zona Lai2,
Alan Lincoln3, and Ursula Bellugi2
1Department of Psychology and Faculty of Music, McGill University, Montre´al, Que., Canada,
2Salk Institute for Biological Studies, La Jolla, CA, USA, and 3California School of Professional Psychology,
Alliant International University, San Diego, CA, USA
ABSTRACT
Williams Syndrome (WS), a neurodevelopmental genetic disorder, is characterized by peaks and valleys in
mental function: substantial impairments in cognitive domains such as reasoning, arithmetic ability, and
spatial cognition, alongside relatively preserved skills in social domains, face processing, language, and
music. We report the results of a comprehensive survey on musical behaviors and background administered
to the largest sample of individuals with WS to date (n ¼ 118, mean age ¼ 20.4), and compare the results to
those obtained from a control group of typically developing normal individuals (n ¼ 118, mean age ¼ 20.9)
and two groups of individuals with other neurodevelopmental genetic disorders, Autism (n ¼ 30, mean
age ¼ 18.2) and Down Syndrome (n ¼ 40, mean age ¼ 17.2). Individuals with WS were found to be rated
higher in musical accomplishment, engagement, and interest than either of the comparison groups, and
equivalent on most measures to the control group. Compared to all other groups including the controls, the
WS individuals displayed greater emotional responses to music, manifested interest in music at an earlier
age, and spent more hours per week listening to music. In addition, the effects of music listening (whether
positive or negative) tended to last longer in the WS group. A factor analysis extracted seven principal
components that characterize the musical phenotype in our sample, and discriminant function analysis of
those factors was able to successfully predict group membership for the majority of cases. We discuss the
neurobiological implications of these findings.
INTRODUCTION
2003; Sergent, 1993; Zatorre, 2003). Part of the
reason for this interest is that music is marked by
Two of the most fascinating unsolved puzzles in
its ubiquity and its antiquity – nearly all known
cognitive neuroscience concern the neuroanatomi-
human cultures have music, some of the earliest
cal basis for music cognition and the architecture
human-made artifacts discovered are of musical
of cognitive function in the neurodevelopmentally
instruments (Cross, 2001; Huron, 2001), and in
impaired. Interest in the cognitive neuroscience of
the present day, music plays a central role in the
music perception, cognition, memory, and perfor-
lives of most of us (Sloboda, 1999). Yet funda-
mance has become a central and compelling
mental issues in music cognition research remain
question in recent years (Peretz & Coltheart,
unsolved, and the results of investigations –
Address correspondence to: Dr. Daniel J. Levitin, Department of Psychology, McGill University, 1205
Avenue Penfield, Montreal, Que., Canada H3A 1B1. Tel.: þ1-514-398-8263. Fax: þ1-514-398-4896. E-mail:
levitin@psych.mcgill.ca
Accepted for publication: December 31, 2003.
0929-7049/04/1004-223$16.00 # Taylor & Francis Ltd.
DOI: 10.1080/09297040490909288

---

224
DANIEL J. LEVITIN ET AL.
especially neuropsychological ones – are contra-
strengths and deficits in Williams syndrome, and
dictory (Hodges, 1996). We still cannot say, for
how these compare to members of other popula-
example, whether or not a given individual has
tions. In particular, people with WS present a
‘‘musical talent’’ in a meaningful or quantifiable
mysterious pattern of peaks and valleys in mental
way, both because definitions and assessment
and motor function. The emerging evidence sug-
procedures are elusive. And efforts to fractionate
gests that cognitive and motor function with
musical behaviors into their functional neuroana-
respect to musical activities remain relatively
tomical components have generally not met with
preserved in WS, despite the presence of gross
the success seen in other domains (Marin & Perry,
impairments in cognitive domains that would
1999; Peretz & Coltheart, 2003), such as language
seem to be subserved by the same neural circuitry
(Patel, 2003), mathematical ability (Dehaene,
(Bellugi, Korenberg, & Klima, 2001; Levitin &
1996) or visual perception (Zeki, 1993), and this
Bellugi, 1998; Levitin et al., 2003).
has raised still more questions about the relation
This article reports on the third stage of a
between music and other cognitive processes.
systematic research program to study music cog-
The other puzzle, involving the cognitive ar-
nition, perception, performance, and listening in
chitecture among those with neurodevelopmental
WS. The first stage, a preliminary investigation of
impairments, is also rich with promise yet without
rhythm perception, was published as Levitin &
clear findings thus far. The study of distinct, well-
Bellugi (1998), the second stage, a neuroimaging
defined, and atypical populations is important
study of music and noise perception was pub-
because it offers a unique opportunity to investi-
lished as Levitin et al. (2003). The present work is
gate specific aspects of cognition, and to establish
an attempt to quantify and characterize the WS
the degree to which various cognitive abilities are
musical profile based on a detailed questionnaire
correlated with, or can be decoupled from one
completed by parents of WS individuals. For
another (Burack, 1997). In particular, the study of
comparison purposes, the questionnaire was also
populations with genotypic abnormalities (includ-
administered to parents of typically developing
ing Down’s Syndrome, Autism, and Williams
normal individuals, as well as individuals with
Syndrome) have sparked anew debates regarding
Autism and Down syndrome (two other neuro-
the modularity of brain function, independence
genetic developmental disorders that result in
of mental faculties, and theories of neural organi-
distinctive types of mental impairment). Ques-
zation (Don, Schellenberg, & Rourke, 1999;
tionnaire studies have previously been conducted
Karmiloff-Smith, 1998; Tager-Flusberg & Sullivan,
with the WS population (Don et al., 1999; Klein,
2000). Indeed, one of the newest and most excit-
Armstrong, Greer, & Brown, 1990; Udwin, 1990).
ing theories in cognitive neuroscience, the neuro-
This questionnaire differs from previous ones in
constructivist view, is strongly inspired by these
several significant respects including a larger
special populations, and in particular by Williams
sample, the use of three comparison groups
Syndrome (Karmiloff-Smith, 1998; see also
(Down Syndrome, Autism, and normals), and its
Elman et al., 1996; Quartz & Sejnowski, 1997).
more extensive topic coverage.
Individuals with Williams Syndrome are unique
among those with neurogenetic developmental
Background on Williams Syndrome
disorders in presenting widespread cognitive im-
Over the past decade, research on WS has
pairment which reputedly does not show up in
increased at a rapid rate as this disorder has come
either language or music (Bellugi, Lichtenberger,
to be regarded as one of the most compelling
Jones, Lai, & St. George, 2001; Levitin & Bellugi,
genetic models of human cognition (Korenberg,
1998, 1999; Levitin et al., 2003).
Chen, Hirota, Lai, Bellugi, Burian, Roe, &
Seemingly unrelated, the two topics of music
Matsuoka, 2000). WS – also known as Williams-
cognition and neurogenetic developmental disor-
Beuren Syndrome (WBS; Beuren, Schulze, Eberle,
ders have very recently come together in the study
Harmjanz, & Apitz, 1964) – is a neurogenetic
of individuals with WS. Researchers seek to
developmental disorder occurring in approxi-
better understand the nature of the cognitive
mately 1 in 20,000 live births, and is caused by

---

CHARACTERIZING THE MUSICAL PHENOTYPE IN INDIVIDUALS
225
the hemizygous deletion of approximately 17
Greenberg, & Bellugi, 1997; Levitin & Bellugi,
genes on chromosome 7 (band 7q11.23) between
1998, 1999). Because the deletion is known in WS,
the polymorphic markers D7S1816 and D7S489B
and the phenotypic manifestations are relatively
(Francke, 1999), including the genes for elastin
well-defined and stable among members of the
(ELN), LIM1kinase (LIMK1), Frizzled (FZD9,
group, WS presents a unique opportunity to uncov-
previously
called
FZD3),
and
Syntaxin1A
er the neurobiological basis of complex cognitive
(STXN1A), and representing 1.6–2 million miss-
behaviors, and in particular, to draw out the links
ing base pairs (Francke, 1999; Frangiskakis et al.,
between genes, brain, cognition, and behavior.
1995; Korenberg et al., 2000). A number of
physical abnormalities are believed to be second-
Previous Studies of Williams and Music
ary to the deletion of ELN at 7q11.23 including
Most of the evidence about the musical abilities
infantile hypercalcaemia (Udwin, 1990), supra-
of individuals with WS has come from anecdotal
valvular aortic stenosis (SVAS; Fanconi, 1952;
reports. To date, only three published papers
Williams, Barratt-Boyes, & Lowe, 1961), a
have addressed their musical abilities. In a pre-
particular craniofacial dismorphology (Hovis &
liminary study, Levitin and Bellugi (1998) tested
Butler, 1997), and scoliosis (Hagerman, 1999).
rhythmic skill in an echo clapping task, and found
The gene for DNA replication factor C2 (RFC2)
that WS were commensurate with mental-age-
and STXN1A are now believed to affect the
matched, typically developing controls (CTLs)
release of neurochemicals, and FZD3 affects cell
in their ability to reproduce musical rhythms.
signaling during neurodevelopment (Karmiloff-
One notable difference was that for those trials
Smith, 1998; Korenberg, Bellugi, Salandanan,
on which participants made errors, the WS were
Mills, & Reiss, 2003). In the few reported cases of
far more likely than typically developing nor-
offspring reproduction involving WS individuals,
mals to produce errors that were musically
the heritability of WS appears to be 0.5.
compatible with the example rhythm, what we
WS is characterized by low IQ, ranging from 40
called ‘‘creative completions.’’ Double-blind, pro-
to 100 (mean $ 61, SD 11, Bellugi, Korenberg, &
fessional musician raters had noticed this differ-
Klima, 2001, p. 9; for slightly different estimates
ence without coaching, and we interpreted this
see also Karmiloff-Smith, 1998; Morris & Mervis,
result as one index of the overall musicality of
1999). WS individuals also present deficits in
WS.
key cognitive domains including conceptual rea-
Don et al. (1999) compared the music and
soning (Bellugi, Klima, & Wang, 1996; Bellugi,
language skills in a small group of children with
Lichtenberger et al., 2001), problem solving, arith-
WS (n ¼ 19, ages 8–13) with typically developing
metic, and spatial cognition (Bellugi, Korenberg, &
normal controls (n ¼ 32, ages 5–12). The re-
Klima, 2001; Frangiskakis et al., 1995). As noted
searchers employed a variety of standard instru-
above, the most intriguing aspect of the WS cog-
ments, such as the Peabody Picture Vocabulary
nitive profile is the presence of relatively strong
Test – Revised (PPVT–R; Dunn & Dunn, 1981),
abilities in four specific domains: social drive
the Auditory Closure Test (Kass, 1964), the Digit
(Gosch & Pankau, 1997; Jones et al., 2001; Losh,
Span subtest from the Wechsler Intelligence Scale
Bellugi, Reilly, & Anderson, 2000; Tager-Flusberg,
for Children (WISC-III; Wechsler, 1991), the
Sullivan, Boshart, Guttman, & Levine, 1996;
Primary Measures of Music Audition (PMMA;
Udwin & Yule, 1991), face processing (Bellugi,
Gordon, 1986), and two questionnaires they
Korenberg et al., 2001; Mills et al., 2001; Paul,
designed in order to elicit information from chil-
Stiles, Passorotti, Bavar, & Bellugi, 2002; Pezzini,
dren and parents about the childrens’ musical
Vicari, Volterra, Milani, & Ossella, 1999), lan-
interests, activities, knowledge, and environment.
guage (Bellugi, Korenberg et al., 2001; Mervis,
WS and controls were found to have comparable
Morris, Bertrand, & Robinson, 1999; Volterra,
musical backgrounds and environments, and
Capirci, Pezzini, Sabbadini, & Vicari, 1996),
equivalent histories of creating music. Both groups
and music (Don et al., 1999; Hopyan, Dennis,
reported that music could make them happy,
Weksberg, & Cytrynbaum, 2001; Lenhoff, Wang,
whereas the WS group reported a significantly

---

226
DANIEL J. LEVITIN ET AL.
greater propensity for music to make them feel sad.
the recorded experimenter, yet when those same
Moreover, the WS group were rated to have sig-
items were administered in person by the experi-
nificantly greater interest in music than the CTLs.
menter, the subjects’ performance increased. It
Finally, Don et al. found significant differences
may simply be the case that WS – being unusually
between groups for ratings of both hyperacusis
interested in social interaction (Jones et al.,
(sensitivity to sound) and hypertimbria (unusual
2000) – perform poorly on standardized tests that
attraction to or liking of certain sounds), with the
are administered from a computer or recording.
WS being far more likely to present both. Levitin
Three limitations of the Don et al. and Hopyan
et al. (in press) confirmed the incidence of hyper-
et al. studies therefore were the relatively small
acusis in individuals with WS, and further frac-
sample size, the small number of questionnaire
tionated their auditory experience into three
items, and the fact that they did not employ
additional components or symptoms: auditory allo-
comparison groups of individuals with other
dynia, odynacusis, and auditory fascinations. WS
forms of mental retardation. A first reasonable
were found to suffer from all four of these symp-
null hypothesis would be that the observed results
toms more often than individuals with DS, AUT, or
are in some way related to neurodevelopmental
typically developing normal controls.
impairment or to developmental delay. We
On the standardized tests, Don et al. found that
included individuals from two neurodevelop-
WS and CTLs scored similarly on the PMMA
mentally impaired groups, Autism and Down
tonal tests, whereas the WS group scored signifi-
Syndrome, as well as age-matched controls. We
cantly worse on the rhythm test. Hopyan et al.
used the Don et al. and Hopyan et al. studies as a
(2001) administered the same test to 14 children
starting point, building on them to develop a more
with WBS (mean age 12, SD 3) and 14 chro-
comprehensive and detailed study.
nologically-aged matched controls (mean age 12,
SD 3) and reported that the control group per-
formed significantly better than the WBS group
METHODS
on the rhythm test. While on the surface this may
seem to contradict the earlier findings of Levitin
Subjects
and Bellugi (1998), three explanations present
Questionnaires were administered to the parents of
themselves. First, as Don et al. point out, the
individuals with WS, Autism (AUT), and Down Syn-
rhythm subtest is always administered after
drome (DS), as well as to typically developing normal
the tonal (melody) subtest in accordance with
control subjects (CTLs).
Parents of WS subjects (n ¼ 130) were solicited
Gordon’s (1986) standardized testing procedures.
from William Syndrome Association National Confer-
WS are known to have attentional deficits, and
ences and from the Laboratory for Cognitive Neuro-
Don et al. note that their WS participants found it
science at the Salk Institute for Biological Studies in
increasingly difficult to pay attention on the
California. All WS subjects were positively diagnosed
second of these two tests (an observation we
by the florescent insitu hybridization test (for absence
have confirmed in our own laboratories). Second,
of one copy of the gene for elastin on chromosome 7,
there is evidence that the PMMA contains man-
the FISH test), or by the WS Diagnostic Score Sheet
(DSS; American Academy of Pediatrics, 2001). Of 149
ufacturing defects (Levitin & Bellugi, 1999) that
people with WS that we originally recruited, 118 were
could actually penalize a careful listener. A third
positively diagnosed by the above criteria and thus
and intriguing possibility has to do with the
retained in the present report. The mean full scale IQ of
demand characteristics of the experiment and
our retained sample was 66 (SD 11). Although this is
social facilitation. In a separate study, we admin-
slightly higher than the IQ obtained in some studies, it
istered a modified version of the PMMA to a
is well within the expected range of sampling
group of 20 WS individuals (Levitin & Bellugi,
variability and is not significantly different from that
obtained in previous studies.
1999) and found that they had a great deal of
Parents of DS participants (n ¼ 40) were solicited
difficulty paying attention to the test when it was
from ongoing studies at the Laboratory for Cognitive
administered from a recording. Our subjects
Neuroscience at the Salk Institute in San Diego,
seemed to lack any sense of engagement with
California. The diagnosis of DS was positively

---

CHARACTERIZING THE MUSICAL PHENOTYPE IN INDIVIDUALS
227
confirmed by genetic testing for trisomy 21. Mean full
(4) Musical creativity and reproduction (how often the
scale IQ was 56 (SD 9.1).
child reproduces music, or makes up music, and in
Parents of AUT participants (n ¼ 30) were solicited
the parents’ view, the quality of these)
from the Developmental Neuropsychology Laboratory at
(5) Musical training
the Alliant International University in San Diego,
(6) Age of onset (when the child first began to manifest
California. All AUT participants were positively diag-
certain musical behaviors)
nosed by trained neuropsychologists prior to their
A more detailed description of the questions
participation in the study using a standardized diagnostic
completes this section, and the full questionnaire is
battery that included the Autistic Diagnostic Interview –
included here as Appendix A.
Revised (ADIR), Autistic Diagnostic Observation
The Interest in Music section first assesses the
Schedule (ADOS), and Childhood Autism Rating Scale
subject’s overall level of musical involvement by
(CARS). Mean full scale IQ was 74.5 (SD 27.8).
employing an item from Grison’s (1972) Levels of
CTL participants (n ¼ 118) were obtained from an
Musical Culture, which were originally designed to rate
undergraduate psychology class at San Diego State
premorbid levels of musicality in patients with hemi-
University in California. They were instructed to
spheric brain disease. Subsequent questions summarize
consult with their parents in completing this ques-
the degree of interest the subject has in comparison to
tionnaire and were reminded that the questions referred
their peers, a description of the types of musical
to their childhood history, not their present experience.
activities the subject is involved in, and an estimate
Some younger siblings and children of students
of the hours per week spent on all music related activities.
enrolled in the class completed the questionnaire and
The section on Emotional Responsiveness characterizes
were also included in the study. All subject groups
the subject’s general emotional responsiveness towards
completed questionnaires on a volunteer basis and, in
music, and probes specific emotional qualities within
addition, CTL subjects received extra class credit for
music to determine whether any specific emotional
completing the questionnaire. The age and gender
patterns exist. This section also probes the length of
distribution of the subjects by diagnosis are shown in
time emotional reactions last in subjects. The series of
Table 1. An ANOVA confirmed that there were no
questions on Musical Creativity and Reproduction
statistically significant differences in age between
address complexity, frequency, accuracy, number of
groups, F(3, 302) ¼ 2.1, n.s.
songs or pieces of music reproduced, and specific
patterns of recall. These items are specifically vague in
Materials
order to include the many forms of musical productions
a subject might make. The only criteria for musical
The questionnaire contained 46 items, 33 multiple choice
reproductions are that they must be of a previously
(including Likert scales) and 13 free response. The free-
heard song or piece of music, and they must incorporate
response items were mapped to discrete categories dur-
tonal qualities. Subsequent sections concerning rhyth-
ing data coding by double-blind raters. The types of
mic production and original musical creations consist
questions asked clustered into six categories:
of items rating the frequency and complexity of original
(1) Demographic (diagnosis, age, sex, handedness) and
productions, in addition to a description of typical
physical profile (hearing loss, physical deficits)
musical creations. The series of questions on Music
(2) Interest in music (whether the child plays an instru-
Training address whether an instrument has ever been
ment, hours per week spent listening, playing, etc.)
played, the frequency of playing instruments, and the
(3) Emotional responses to music
amount of formal training in music theory received.
Table 1. Age and Gender Distribution of Subjects in this Study.
AUT
Control
DS
WS
n
30
118
40
118
Chronological age: M (SD)
18.2 (7.7)
20.9 (7.4)
17.2 (9.2)
20.4 (10.4)
Range
9–39
5–44
5–51
5–50
Mental age: M (SD)
74.5 (27.8)
110 (10)Ã
56 (9.1)
66 (11)
Male
24
28
20
61
Female
6
90
20
57
Note. ÃNot measured. We assume the mean of a sample of San Diego State University students is probably 110 and
SD is likely no higher than 10.

---

228
DANIEL J. LEVITIN ET AL.
Age of onset comprised questions addressing when the
p < .001, Reproduction Accuracy, F(3, 294) ¼
child first took an interest in playing an instrument or
9.1, p < .001, Frequency of Spontaneous Rhyth-
listening to music.
mic Productions, F(2, 283) ¼ 8.3, p < .001, Fre-
quency of Original Music Productions, F(3, 277) ¼
Analyses
5.1, p < .002, Playing a Musical Instrument,
The quantitative data obtained from the questionnaire
F(3, 297) ¼ 8, p < .001, Hours Per Week Spent
(e.g., Likert scale scores and multiple-choice scores)
were analyzed using either Chi-square tests (for
Playing a Musical Instrument, F(3, 284) ¼ 5.7,
nominal data such as handedness) or analysis of
p < .001
and
Exposure
to
Music
Theory,
variance (for ordinal data, such as hours spent listening
F(3, 279) ¼ 9.6, p < .001. No other tests were
to music per week), with planned orthogonal contrasts
significant, so we report the planned tests of
to test differences between group means. All tests were
intergroup differences only on these 10 items. [The
adjusted for multiple comparisons. The free response
degrees of freedom are not identical for each
data were transformed into nominal categories by three
comparison due to missing data. For example, if a
double-blind raters who were trained to score the
presence or absence of characteristics in specific
parent felt that their child had shown no particular
domains of interest. We used Kappa as a measure of
interest in music, they would have left the ‘‘age of
interrater agreement setting a threshold of 0.75 (as
onset of musical interest’’ item blank.]
recommended by Landis & Koch, 1977). The Kappa
values obtained for interrater reliability in this study
Planned Comparisons
were all well above this level ( ¼ 0.80 to 1.00 Æ SE
WS and DS individuals were reported to show
() ¼ 0.09) for all items measured.
significantly earlier evidence of first musical inter-
est than controls (p < .001 adjusted;1 Table 2).
WS individuals listened to music an average of
RESULTS
3 hr more per week than controls (p < .001), and
2–3 hr a week more than DS and AUT, respec-
Reliability
tively, although these latter differences did not
First, to assess the reliability of the questionnaire,
reach significance (p $ .2; but see note below un-
we performed a split-subjects reliability analysis
der the next section about binomial comparisons).
(by randomly assigning subjects to one of two
The overall level of musical involvement, as
arbitrary groups) and found no significant dif-
assessed by the Grison scale (which combines
ference between the halves, F(1, 305) ¼ .16,
music listening, instrument playing, and music
p $ .69). We performed additional split-subjects
theory) found WS involvement to be significantly
reliability tests on the seven broad content areas
greater than that of the DS and AUT groups
extracted by factor analysis (the extraction
(p < .05) and similar to CTL (p $ .9). Also sig-
procedure and results are described below in the
nificantly higher than controls in musical involve-
section entitled ‘‘Factor Analysis’’) and again
ment were the AUT (p < .02) and DS (p < .01).
found no significant between the halves (For the
Parents of WS reported a higher level of
seven factors, all with df ¼ (1, 283): F1 ¼ .342,
interest in music-related activities than the CTL
p $ .56; F2 ¼ .812, p $ .37, F3 ¼ .11, p $ .75;
(p < .001) and AUT groups (p < .02) and an
F4 ¼ 2.1, p $ .15, F5 ¼ .09, p $ .76, F6 ¼ 1.8,
equivalent level of interest to DS (p ¼ 1). Individ-
p $ .18, F7 ¼ 1.3, p $ .26).
uals with DS showed somewhat more interest in
music than those with AUT (p $ .08), and sig-
Group Differences
nificantly more interest than controls (p < .05).
An
ANOVA
showed
significant
intergroup
WS also were reported to experience significantly
differences for 10 of the items: the Grison
more emotion when listening to music than CTLs
profile, F(3, 301) ¼ 5.5, p < .001, Musical Interest
Amount, F(3, 298) ¼ 9.9, p < .001, Musical Inter-
est Age of Onset, F(3, 271) ¼ 13, p < .001,
1All p values reported for the contrasts are the adjusted
Emotional Response to Music, F(3, 299) ¼ 12.2,
p values for multiple comparisons.

---

CHARACTERIZING THE MUSICAL PHENOTYPE IN INDIVIDUALS
229
Table 2. Planned Orthogonal Contrasts of Intergroup Differences for 10 Items on the Questionnaire. Values in the
Cells are the Means (and Standard Deviations) for Each Item. Significance Level is p < .05 for all Pairs,
Adjusted for Multiple Comparisons.
Item
CTL
AUT
DS
WS
Grison Profile
3.5a,d (1.1)
2.7c (1.1)
2.9c,w (.74)
3.4d (1.1)
Musical interest amount
4.9d,w (1.3)
4.6w (1.9)
5.7c (1.5)
5.8a,c (1.4)
Musical interest – age of first onset
5.0d,w (3.6)
4.4 (4.0)
2.3c (2.2)
2.7c (2.6)
Emotional response to music
4.8w (1.1)
4.0w (2.0)
5.1 (1.3)
5.7a,c (1.4)
Reproduction accuracy
4.4a,d (1.5)
3.2c (2.2)
3.2c,w (1.5)
4.4d (1.8)
Frequency of spontaneous rhythmic productions
4.0a,d (1.8)
2.1c,w (1.8)
2.9c (2.1)
3.6a (2.2)
Frequency of original music productions
3.1a (1.5)
1.8c,w (1.7)
2.6 (1.7)
3.2a (2.0)
Musical instrument
.83a,d (.38)
.50c,w (.51)
.56c (.50)
.80a (.40)
Hours playing instrument (per week)
1.5w (2.4)
1.2 (2.6)
.82w (1.6)
1.9c,d (3.1)
Music theory
.51a,d (.50)
.12c,w (.33)
.12c,w (.33)
.35a,d (.48)
Note. aSignificantly different from autistic group.
cSignificantly different from control group.
dSignificantly different from DS group.
wSignificantly different from WS group.
or AUT participants (p < .001) and slightly more
DS (p < .02 and p > .06, respectively). The WS
than DS individuals (p $ .08). In spontaneous
played their musical instruments more hours per
reproductions of familiar music, WS and CTLs
week than the CTL (p < .03) or DS groups
both reproduced music similarly well, and more
(p < .001) and marginally more than the AUT
accurately than the DS individuals (p < .001).
group (p < .13) although, it is important to note
CTLs
were
significantly
better
than
AUT
that this latter comparison is driven by a relatively
(p < .05) and WS were better than AUT (p < .06).
small number of AUT individuals in our sample
WS and CTLs tended to play rhythmic patterns
who played an instrument at all (14), and those
as often as one another, and more often than
who did tended to play it a lot. Finally, both the
individuals with AUT (p < .005). In addition,
WS and CTL groups were more likely to have had
the CTLs played rhythms more often than the
music theory than the AUT or DS groups. (WS vs.
DS (p < .04), and although the mean rating for
AUT, p < .03, WS vs. DS, p < .02; CTL vs. AUT
WS versus DS was higher, it did not reach
and CTL vs. DS, p < .001). The CTL individuals
statistical significance (p $ .4). The frequency
were slightly more likely to have had music
with which individuals spontaneously invented
theory than the WS (p < .08).
their own music (melodies and rhythms together)
was significantly greater for WS and CTLs versus
Binomial Comparisons Between WS
the individuals with AUT (p < .01), and the CTL
and Other Neurodevelopmentally
group had greater frequency than the DS group
Impaired Individuals
(p < .03). As before, the mean for the WS was
Of the 10 items in the previous section for which
higher than that for the DS, but not statistically so
intergroup differences were identified by ANOVA,
by the ANOVA (see note below, p $ .3).
post hoc tests revealed that the WS individuals
Both the CTLs and the WS group were sig-
were statistically higher on 4 of the measures than
nificantly more likely to play musical instruments
DS individuals. However, a careful examination
than the individuals with AUT (p < .02 and
of Table 2 reveals that in fact the means for the
p < .04, respectively). In addition, CTLs were
WS group were in the direction of increased
more significantly more likely and WS more
musicality for 9 of the 10 items compared to the
slightly more likely to play an instrument than
DS group (note that with the exception of ‘‘age of

---

230
DANIEL J. LEVITIN ET AL.
onset of first musical interest,’’ higher means
Table 5. Reports of Physical Deficits by Diagnosis
represent more musicality for all of these
(DX) and Statistically Significant Differences
measures, and DS in fact were reported to show
at p < .01.
a lower age of onset for musical interest). We can
Sig
DX
Physical deficits
consider this as a binomial experiment (formally
equivalent in this case to a sign test) and ask what
A, D, W
CTL
1.7 (%)
is the probability that the WS means would all
C
AUT
13.8
C
DS
15.0
show an effect in the same direction compared to
C
WS
15.9
the DS means on at least 9 out of 10 items ? This
is exactly [(10!(.59)(.51)/9!1!) þ (10!(.510)(.50)/
10!0!)], or p $ .01. Thus we can conclude that
(p < .01). The three neurodevelopmentally im-
for these 10 measures, WS were in fact sig-
paired groups had similar levels of reported phys-
nificantly different (higher) than the DS group.
ical deficits to one another, and more than the
The same holds true for a direct comparison of the
CTL group (p < .01; Table 5).
WS and AUT group: by binomial test, the WS
group was significantly different (in the direction
Factor Analysis
of ‘‘more musical’’) than the AUT group at p $ .01.
By design, our questionnaire contained a number
By this same test, there were no significant dif-
of items that attempted to index similar latent
ferences found between DNS and AUT.
qualities of musical behavior and therefore one
would expect some items to be significantly
Physical Characteristics
correlated with others. Factor analysis can reduce
We found differences in handedness with the
the 33 multiple choice items to a smaller, more
parents reporting that DS and WS were both more
manageable number, while at the same time
likely (p < .02) to be left handed or ambidextrous
bringing together items that in fact index the same
than the control group (Table 3). DS were more
underlying attributes. We performed an explor-
likely than other groups to have suffered from
atory factor analysis and settled on seven orthogo-
hearing loss (Table 4), with WS more likely than
nal components which were rotated by the
AUT or CTLs to have also suffered hearing loss
varimax algorithm, and which together account
for 67% of the variance (and use only Eigenvalues
greater than 1, see Fig. 1). The seven factors were
Table 3. Reports of Handedness by Diagnosis (DX) and
truly orthogonal, and intercorrelations among the
Statistically Significant Differences at p < .02.
Sig
DX
Left (%)
Ambi. (%)
Right (%)
D, W
CTL
6.8
0.8
92.4
AUT
10.0
0.0
90.0
C
DS
17.1
1.7
81.2
B
WS
20.0
7.5
72.5
Table 4. Reports of Hearing Loss by Diagnosis (DX)
and Statistically Significant Differences at
p < .01.
Sig
DX
Hearing loss (%)
D, W
CTL
4.2
D, W
AUT
0.0
C, A, W
DS
43.6
Fig. 1. Eigenvalues for the 33 questionnaire scale
C, A, D
WS
17.8
items.

---

Table 6. Factor Analysis Solution of the 26 Multiple Choice Items on the Questionnaire. Seven Principal Components were Extracted Accounting for 67% of the
Variance.
Rotated component matrix
Component
1
2
3
4
5
6
7
I
Frequency of original music production
.81
.15
.15
.08
.00
.03
.06
Spontaneity
CHARA
Frequency of spontaneous rhythmic productions
.81
.12
.18
.13
.03
.14
À.08
Rhythmic complexity
.78
.11
.32
.06
À.08
.02
.09
Original music complexity
.74
.09
.34
À.03
À.12
.03
.20
CT
Original music description
.71
.11
.05
.25
.13
.08
.04
ERIZING
Hours/week playing instrument
.49
.36
.08
.35
.11
À.13
À.05
II
Reproduction length
.16
.77
.09
.05
.01
.06
.09
Reproduction of Music
THE
Reproduction complexity
.25
.77
.16
.11
À.11
.12
.12
Reproduction number
.20
.67
.04
.11
.31
.14
.14
MUSIC
Reproduction frequency
.31
.65
.33
.16
À.08
À.04
.08
Reproduction accuracy
.35
.63
.11
.29
.22
À.02
À.08
AL
III
Emotions and listening
.21
.11
.77
.12
.02
.10
.16
Listening Habits
PHENO
Musical interest amount
.15
.22
.77
.14
.02
.09
À.04
Positive reactions to happy music
.18
.08
.64
À.04
.03
.38
.13
TYPE
Music Listening Hours/week
.18
.33
.59
.10
À.03
.14
À.05
IV
Exposure to music theory
.01
.18
À.05
.80
.03
.10
.12
Theory, Achievement
IN
INDIVI
Playing a musical instrument
.25
.13
.04
.68
.00
.00
.00
Grison Profile
.27
.28
.27
.59
À.18
.01
À.11
DU
V
Reproduction age of onset
À.03
.12
À.02
.06
.79
.06
À.04
Age of Onset
ALS
Interest age of onset
À.02
.13
.37
À.19
.69
À.16
À.07
VI
Negative reactions to sad music
.22
.08
.27
À.02
.13
.82
.13
Negative Reactions
Negative reactions to happy music
À.03
.19
À.07
.04
À.34
.82
À.11
VII
Happy music carryover
.13
.05
.23
À.1
À.06
À.10
.82
Sensitivity
Sensitivity to sound
À.04
À.06
.38
À.02
.19
.07
À.58
Sad music carryover
.02
.15
À.01
.22
.03
.44
.59
Positive reactions to sad music
À.12
.33
.25
À.03
À.49
.11
.49
231

---

232
DANIEL J. LEVITIN ET AL.
factors were all zero. We explored as alternatives
a six factor solution (accounting for 62% of the
variance) and an eight factor solution (accounting
for 71% of the variance) but the rotated com-
ponent matrix made more sense theoretically with
seven factors, and in fact corresponded quite well
to our initial a priori categories (as outlined in the
Methods section). The difference between the
factor analytic solution and our a priori categories
is that the factor analysis split off two meaningful
categories or factors from our initial conception,
one for sensitivity to music and another for
negative reactions to music. The six factor solu-
tion placed, what seemed to us to be concep-
tually separate items, in the same component for
factor number 5 (negative reactions to happy
music, reproduction age of onset, interest age of
onset, and positive reactions to sad music). The
seven factor solution separated these out. The
final rotated components matrix with factor
interpretations are presented in Table 6, and the
hierarchical structure of the factors is presented in
Fig. 2. Hierarchical structure of the seven principal
Figure 2.
components extracted through factor analysis
and their intercorrelations. Letters refer to the
Figure 2 displays the hierarchical structure of
factor names as designated in Table 7. Note
the seven factors extracted, beginning with the
that Factors I (Spontaneity) and III (Listening
first unrotated principal component (FUPC). Note
Habits) break away from the rest relatively
that Factors I (Spontaneity) and III (Listening
early and remain virtually unchanged until the
Habits) split off relatively early, and stay virtually
end of the analysis.
the same all the way to the end of the hierarchy.
Factor II (Reproduction) breaks away at Level 5
and remains virtually unchanged through the next
two iterations.
For Theory/Achievement, the WS and CTLs were
We performed an ANOVA on the seven com-
again similar, with WS showing higher scores
ponents with diagnosis as a factor in the ANOVA
than AUT (p < .005) and DS (p < .001), and
model, and the first six components were sig-
significant differences between AUT and CTL
nificant at p < .01 or less. Planned orthogonal
(p < .001) and DS and CTL (p < .001). The Age
contrasts revealed the following intergroup differ-
of Onset at which various musical behaviors was
ences (Table 7): For Spontaneity, WS and controls
first noticed showed differences between WS
were similar.
and CTL (p < .03) and WS and AUT (p > .03).
The individuals in the AUT and DS groups
Finally, the Sensitivity factor showed differences
showed significantly lower scores on this factor
between the CTLs and all other groups (p < .001).
than controls (p < .02 ). For Listening Habits,
Sensitivity is an especially interesting factor – it
individuals with WS were higher than CTL
comprises items that concern how long the indi-
(p < .001) and AUT (p < .005), and CTL were
vidual remains happy after hearing happy music,
higher than DS (p < .001) who in turn were higher
or sad after hearing sad music. It was designed as
than AUT (p < .03). For Reproduction, WS dif-
an index of the affective response to music in
fered significantly from AUT (p > .02) and AUT
terms of how long music’s effects can last. The
differed from DS (p < .04); WS and CTL and WS
WS individuals scored higher on this factor than
and DS were statistically similar on this factor.
any other group.

---