Sounds of melody--pitch patterns of speech in autism.

Abstract

The objective of this study was to find a pattern in vocalizations of children with Autism Spectrum Disorder (ASD). We compared the intonational features of 15 children with ASD who showed speech, aged 4-10 years, with 10 age-matched typically developing controls. Exaggerated pitch, pitch range, pitch excursion and pitch contours were observed in speech of children with autism, but absent in age-matched controls. These exaggerated features, which are distinctive characteristics of motherese, were also seen in interactions of an independent group of 8 mothers of typical infants using child-directed speech. Our findings provide the first evidence of a distinct pattern in vocal output from children with autism. They also demonstrate that speech patterns might follow a delayed developmental trajectory in these children.

Full text

Page 1

Neuroscience Letters 478 (2010) 42–45
Contents lists available at ScienceDirect
Neuroscience Letters
journal homepage: www.elsevier.com/locate/neulet
Sounds of melody—Pitch patterns of speech in autism
Megha Shardaa, T. Padma Subhadraa, Sanchita Sahaya, Chetan Nagarajaa, Latika Singha,
Ramesh Mishraa, Amit Senb, Nidhi Singhalc, Donna Ericksond, Nandini C. Singha,∗
aNational Brain Research Centre, NH-8, Nainwal Mode, Manesar 122 050, India
bSitaram Bhartia Institute for Science and Research, B-16 Qutub Institutional Area, New Delhi 110 016, India
cAction for Autism, The National Centre for Autism, Pocket 7 & 8, Jasola Vihar, New Delhi 110 025, India
dShowa Music University, Kawasaki 215-8558, Japan
a r t i c l ei n f o
Article history:
Received 24 February 2010
Received in revised form 14 April 2010
Accepted 28 April 2010
Keywords:
Autism
Speech
Prosody
Pitch
Intonation contours
Motherese
a b s t r a c t
The objective of this study was to find a pattern in vocalizations of children with Autism Spectrum
Disorder (ASD). We compared the intonational features of 15 children with ASD who showed speech,
aged 4–10 years, with 10 age-matched typically developing controls. Exaggerated pitch, pitch range,
pitch excursion and pitch contours were observed in speech of children with autism, but absent in age-
matched controls. These exaggerated features, which are distinctive characteristics of motherese, were
also seen in interactions of an independent group of 8 mothers of typical infants using child-directed
speech. Our findings provide the first evidence of a distinct pattern in vocal output from children with
autism. They also demonstrate that speech patterns might follow a delayed developmental trajectory in
these children.
© 2010 Elsevier Ireland Ltd. All rights reserved.
Autism Spectrum Disorder (ASD) is a pervasive developmental
brain disorder that manifests in the first 3 years of life. It is char-
acterized by abnormal social behaviour, reduced ability or interest
in communicating with others, language dysfunction and rigidi-
ties of behaviour and thought [2]. Communicative dysfunction is
a characteristic feature of ASD [2] and in many cases, manifests
through absent or deviant vocalizations. Although past research on
vocalatypicalitieshasshedsomelightonthenatureofcommunica-
tive abnormalities in autism [3], no clear patterns have emerged to
characterise vocalizations from ASD populations due to the hetero-
geneity of the disorder and variability in tasks and measurements
used. As a result, speech therapists continue to struggle with meth-
ods and approaches to improve and induce spoken language in
children with autistic. It is clinically important to develop easy-to-
implement interventions that might facilitate verbal development
in autism.
Earlier statistics report more than 50% of the ASD population to
be nonverbal [8]. It has been established, by advances in the past
decade, that there is an early brain dysfunction in autism which
intensive intervention can alter, leading to optimistic outcomes.
Additionally, it has now been shown that more than 75–95% chil-
dren, who receive early intervention, develop useful speech [8].
∗Corresponding author.
Tel.: +91 124 233 8920/28x333; fax: +91 124 233 8927/28.
E-mail address: nandini@nbrc.ac.in (N.C. Singh).
Given this premise of abnormal early development, it is possible
that children with ASD have a delayed developmental trajectory.
Asaresult,communicationmilestonesfollowedbytypicalchildren
might appear at later time points manifesting as atypical develop-
mentaltraits.Amongthemostcommonlyreportedatypicalfeature
of speech and communication in autism is unusual prosody [17].
Prosodyisabroadtermthatencompassesaffective,pragmatic,syn-
tactic as well as intonational functions in communication and is
usedtorefertothemannerofspeakingratherthanthecontent.The
acoustic measures of prosody include mean voice pitch (percep-
tual correlate of fundamental frequency, F0), pitch range, intensity,
speaking rate, pause duration and intonation contours [4].
Distinctiveintonationalpatternsareknowntoaidmaturationof
speech and communication skills in typically developing children
[10]. Speech directed to infants commonly known as ‘motherese’
hasdistinctprosodicpatternscharacterizedbyhigherpitch,slower
tempo and exaggerated intonation. Early intonation patterns of
typically developing children also exhibit motherese-like features,
which they outgrow by 2–3 years of age [9]. We hypothesize that
children with ASD in the age group of 4–10 years may exhibit
delays in their verbal communication strategies, when compared
with age-matched typically developing children. This hypothesis is
motivatedbysomeevidencefrompastresearchthathasshownthat
receptive prosodic skills develop late in children with ASD as com-
pared to age-matched controls [19]. In addition, the DSM and ICD
criteria cite delayed development of speech as a key feature of lan-
guage in autism, and delay is acknowledged in many descriptions
0304-3940/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.neulet.2010.04.066

Page 2

M. Sharda et al. / Neuroscience Letters 478 (2010) 42–45
43
Fig. 1. (a–c) Fundamental frequency contours.
Fig. 2. (a–c) Box plots of mean pitch, pitch range and pitch excursion.
of the language characteristics of individuals with autism [20]. We
predict that the motherese-like intonation patterns, that typically
developing children outgrow by 2–3 years of age [9], will be exhib-
ited in children with ASD in the age-group of 4–10 years due to a
developmental delay in their verbal communication strategies. We
tested the manifestation of this delay using methods of acoustic
analysis to compare characteristic intonation patterns of vocal-
izations from three groups: children with ASD aged 4–10 years,
age-matched typical controls and vocalizations from mothers of
typical infants.
Thirty-three subjects were recruited for this study: 15 chil-
dren (14M, 1F) with ASD (AUT), 10 (9M, 1F) age-matched typically
developing (TD) control children and 8 mothers (MOT) of typi-
cally developing infants aged 6–18 months. The AUT group was
recruited from child and adolescent psychiatry units of hospitals
and met the diagnostic criteria of Autism/PDD-NOS with speech in
the Diagnostic and Statistical Manual for Mental Disorders, DSM
IV [2] and assessment using CARS (Childhood Autism Rating Scale)
[21] in some cases. Since the incidence of ASD is much higher in
boys it was not possible to balance the groups for gender [24]. The
AUT group had mean age 6.25 years (SD 1.5) and the TD group
had mean age 7.3 years (SD 2.0), the age range being 4–10 years
for both groups. All subjects were English-Hindi bilinguals. The
verbal ability of the ASD group was confirmed by parents report
and psychological assessment by clinicians. The criteria for select-
ing children was a minimum vocabulary of 20 words by age 4 and
ability to elicit a minimum of 2min of vocalizations in interactive
speech conditions. No quantitative judgements were made about
the verbal IQ of these children as the parameters measured in the
study looked only at the qualitative aspects of the utterances inde-
pendent of content of vocalizations. Members of all groups had no
known speech or hearing deficits as confirmed by a thorough clin-
ical workout. Informed consent was obtained for each individual’s
participation as approved by the Human Ethics Committee of the
Institute.
Speech samples were elicited using a spontaneous speech task.
In preparation for the task, children from both the groups, AUT and
TD, were instructed to name twenty pictures presented sequen-
tially on a computer screen. Some pictures in the task were of a
car, a television and an ice cream. The interviewer made a note
of the objects, which most of the AUT group appeared to display
an interest in (ice-cream, television and car). Those pictures were
then displayed on the computer screen and discussed with the
children using opening questions like ‘Do you like ice cream?’ and
‘Whichisyourfavouritecar?’.Thespontaneousresponsesobtained
were simultaneously recorded and were used to carry out all the
acousticanalysis.Allthemothersofthetypicallydevelopinginfants
(MOT) who participated in the study were also instructed to talk to
their infants (between 6 and 18 months) describing the objects ice
cream, television and car. This was done to elicit comparable sam-
ples of conversational speech revolving around a common topic.
All recordings were made using a noise cancellation microphone
attached to a laptop computer. Extraneous noise was manually
edited out.
Thelongestuninterruptedsampleinspontaneousspeechforthe
AUT group was 80s, thus 80s long sequences were also excerpted
from the TD and MOT group for acoustic analysis. Each sample
was 16-bit digitized at a sampling rate of 22,050Hz using Gold-
Wave software. We used a pragmatic approach of chunking the
speech into speech segments which were demarcated by pauses
no longer than 300ms. The following acoustic measurements were
made using the cross-correlation algorithm in Praat [5]:
1. Pitch (F0) Measures: Measurements of mean, maximum and
minimum pitch were made.
2. Pitch Range: In order to quantify the dynamics of pitch change,
a pitch range measure was calculated as the difference between
maximum and minimum pitch used by the subject measured in
Hertz.
3. Pitch Excursion: An expanded pitch contour is described in
terms of a long, continuous pitch glide, with pitch excursions
of 13 semitones/s or more [10]. Since the perception of changes
in pitch is more proportional than absolute, the conversion from
absolute frequency to semitone ratio allows more meaningful
comparisons across talkers with different mean F0 [12]. Pitch
excursion was calculated using the formula proposed by de
Pijper [7]. Each spontaneous speech sample was broken down
into speech chunks separated by pauses of 300ms.
Table 1
Regression matrix.
GroupPitchPitch rangePitch excursion
p value
Regression effects with age (r2)
ASD0.197
TD0.481
0.452
0.03
0.043
0.531
>0.05
>0.05

Page 3

44
M. Sharda et al. / Neuroscience Letters 478 (2010) 42–45
Pitch Excursion
(semitones/s)
=
⎡
⎣
39.863 log
?
F0max
F0min
?
speech chunk duration
⎤
⎦
Fig. 1(a–c) shows fundamental frequency contours. The AUT
contours are similar to MOT speech and show expanded prosodic
features, evidenced by more dynamic contours.
Fig. 2(a–c) shows box plots representing quantitative measure-
ments of mean pitch, pitch range and pitch excursion.
Table 1 shows a regression matrix to rule out effects of age on
all measured parameters for both AUT and the TD groups.
Fig. 2(a) shows mean pitch for each of the three groups:
AUT=355.8Hz SD 61.7, MOT=326.3Hz SD 31.2Hz, TD=275.4Hz
SD 22.5Hz. One-way ANOVA showed significant differences
between the groups (F(1,32)=9.013, p<0.001). A post-hoc SNK
(Student–Neuman–Keuls) test of multiple pair-wise comparisons
between groups showed significant differences between AUT and
TD (p<0.001), and also between MOT and TD (p=0.028) but no sig-
nificant difference between mean pitch of AUT and MOT (p=0.157)
although the effect was medium (Cohen’s d=0.6, r=.288).
The pitch range for the three groups is presented in
Fig. 2(b), with AUT=550.6Hz SD 84.9, MOT=466.07Hz SD 70.7,
TD=465.7Hz SD 41.2. As the pitch range data were not part of
a normal distribution, a non-parametric Kruskal–Wallis one-way
ANOVA on ranks was used, and showed significant differences
betweengroups(H(1,2)=9.947,p=0.007).Apost-hocDunn’stestof
multiple comparisons showed significant difference between AUT
andTD(p<0.05),butnodifferencebetweenAUTandMOT(p>0.05)
with a large effect size (Cohen’s d=1.08, r=0.475).
Fig. 2(c) shows the average pitch excursion for each group
(−AUT=13 semitones/s, SD 7.8, MOT=18 semitones/s SD 5.88,
TD=6.8semitones/sSD2.4).Aone-wayANOVAshowedsignificant
differences between groups (F(1,32)=7.314, p=0.003). A post-hoc
SNK (Student–Neuman–Keuls) test of multiple pair-wise compar-
isons between groups showed significant differences between AUT
and TD (p=0.002), and also between MOT and TD (p=0.022), but
no significant difference between mean pitch excursion of AUT
and MOT (p=0.076), with a medium-sized effect (Cohen’s d=−.72,
r=−0.3).
Our findings show that pitch patterns of verbal children with
ASD in the age group of 4–10 years are different from age-matched
typically developing children. Pitch patterns of the AUT group are
characterized by exaggerated intonation contours manifested in
terms of elevated pitch, higher pitch range and pitch excursion as
comparedtoTD.Nosignificanteffectsofagewereseenonanyofthe
measured parameters in both groups of children as confirmed by
theregressionanalysis(Table1).Thisdemonstratedthatthediffer-
encesinthepatternsofspeechintheAUTandTDgroupsreflecteda
developmentaldelayofverbalskillsintheAUTgroup.Itisnotewor-
thy,thatexaggeratedpitchpatternsarealsodistinguishingfeatures
of motherese as demonstrated by analysis of child-directed vocal-
izations of mothers in the MOT group [10]. Our results showed that
pitch patterns of the AUT and MOT group had no significant dif-
ferences. Although there was a difference in group sizes (AUT=15,
MOT=8), we see a medium-sized effect. This might be due to two
reasons. The AUT group demonstrated considerable variability in
their pitch patterns. Second, since we compare vocalizations from
mothers and children, it is possible that there might be other con-
founding factors like age and experience, which might have led to a
larger effect. However, despite these factors, our preliminary find-
ings show interesting pitch patterns in ASD speech, which can be
explored further.
Longitudinal studies on the development of speech have shown
that in typically developing children between 0 and 5 years, mean
pitch decreases as a function of age [1]. Additionally, in typically
developing children, pitch range decreases with age and pitch
variabilityapproacheszerobetween7and9years[9].Pitchcharac-
teristicsoftheAUTgroupintheagegroup4–10years,areexhibited
by typically developing children between 2 and 3 years, suggesting
that these features might follow a delayed developmental trajec-
tory [9].
For typically developing infants, mothers use specific tonal and
temporal patterns of infant-directed speech in order to enhance
language development [11]. It is through melodic qualities (higher
pitch, pitch range, and expanded intonation), embedded in the
prominent intonation contours of motherese, that speech first
becomes meaningful to infants [16]. Our finding that the pitch pat-
terns of the AUT group are similar to the MOT group suggests a
role for motherese-like speech in verbal development of children
with ASD. The close relationship between receptive and expres-
sivelanguageskillsformsabasisforunderstandingtheimportance
of motherese-like speech in environments of older children with
ASD [18]. Even when verbal development is delayed, the role of
parent-child interaction in enhancing communication cannot be
ignored and motherese provides one such platform of interaction.
Literatureonparentalresponsivenesshasshownthattheimpactof
mothersonchildren’sdevelopmentisdeterminedbythe‘style’and
persistent nature of mother-child interaction [6]. It has also been
shown that children are more responsive to conversational rather
than monotonic interactions [15]. The use of motherese might not
induce speech in children who never develop language, but might
aid verbal language in children who suffer developmental delays.
The relation between motherese and expression of affect may be
anothercrucialaspectforchildrenwithautism.Affectoremotional
content in speech is a well-known characteristic of exaggerated
contours also seen in motherese. The classic view of autism [14]
places a “biological disturbance of affective contact” as a primary
deficit. Yet, this does not necessarily imply a prosodic disturbance.
A recent study [13] investigated the production of affective speech
in children with autism and found that they are probably able to
process large pitch and intensity variations, but they tend to over-
shoot the intonational target and are deficient in correlating the
pitch and intensity variations with the appropriate emotions. This
is further corroborated by our findings of dynamic pitch ability in
children with ASD.
Although not many studies on vocalizations of children with
autism have found consistent results, there is a hypothesis that
these individuals display atypical phonological and prosodic fea-
turesinearlyvocalizations[22].Pastliteratureonprosodyinautism
has primarily been focussed on affective and pragmatic aspects.
Studies by Baltaxe et al. [4] analysed frequency range patterns in
spontaneous utterances of children with autism compared TD con-
trols but did not find significant differences suggesting that the
mean of the ranges did not adequately capture the deviances in
speech of the autism group. Other studies by Paul et al. [18] and
Shriberg et al. [23] have looked at stress patterns in autism and
have reported atypicalities. While all these studies report atypical
prosodic patterns in children with autism, their findings have been
variablemakingitdifficulttodrawconsistentconclusionsaboutthe
nature of prosodic deficits in ASD productions. Our study is the first
attempt to identify a distinct pattern in vocalizations of individuals
with ASD using a simple speech-eliciting task. The identification
of exaggerated speech in older children with ASD might provide
an impetus for designing intervention strategies that might work
more effectively in these high-functioning subgroups.
Although we look at a small proportion of the ASD spectrum,
the results of our study provide the first evidence of a consistent
pitch pattern in a group of individuals with ASD. Further research
willaimatstudyingmorecontrolpopulationsofchildrenwhohave
not been exposed to motherese and correlate expressive language
measures with motherese exposure.

Page 4

M. Sharda et al. / Neuroscience Letters 478 (2010) 42–45
45
Preliminaryasourfindingsare,theycouldtobeexploredfurther
to establish a basis for continued use of motherese-like speech in
enhancing verbal communication in sub-populations of ASD indi-
viduals.
Acknowledgments
This research was funded by the National Brain Research Cen-
tre and the Department of Information Technology, Government of
India. The authors would like to thank the children and mothers
who participated in the study; Shobha Srinath and N. Shivshankar,
from the National Institute for Mental Health and Neurosciences,
India for providing data.
References
[1] S. Amano, T. Nakatani, T. Kondo, Fundamental frequency of infants’ and par-
ents’ utterances in longitudinal recordings, J. Acoust. Soc. Am. 119 (2006)
1636–1647.
[2] American Psychiatric Association, Diagnostic and Statistical Manual of Mental
Disorders,4thed.(DSMIV),AmericanPsychiatricAssociation,Washington,DC,
1994.
[3] C.A.M. Baltaxe, J.Q. Simmons, E. Zee, Intonational patterns in normal, autistic
and aphasic children. In A. Cohen, M. van de Broecke (Eds.), Proceedings of the
10th International Congress of Phonetic Sciences, Dordrecht, Foris, 1984.
[4] R.Banse,K.R.Scherer,K.R.Acoustic,Profilesinvocalemotionexpression,J.Pers.
Soc. Psy. 70 (1996) 614–636.
[5] P. Boersma, P. Weenink, Praat: Doing phonetics by computer, Version 4.3.14,
2005.
[6] M.H. Bornstein, Maternal responsiveness: characteristics and consequences,
New Dir. Child Dev. 43 (1989) 75–87.
[7] J.R. De Pijper, Modelling British English Intonation, Dordrecht, Foris,
1983.
[8] G. Dawson, S. Faja, Autism spectrum disorders: A developmental perspective.
In:T.P.Beauchaine,S.P.Hinshaw(Eds.),ChildandAdolescentPsychopathology.
Wiley, Hoboken, NJ, in press.
[9] S. Eguchi, I.J. Hirsh, Development of speech sounds in children, Acta Oto-
Laryngol., Suppl. 257 (1969) 1–51.
[10] A. Fernald, T. Simon, Expanded intonation contours in mothers’ speech to new-
borns, Dev. Psych. 20 (1984) 104–113.
[11] A. Fernald, P.K. Kuhl, Acoustic determinants of infant preference for motherese
speech, Infant Behav. Dev. 10 (1987) 279–293.
[12] D.L. Grieser, P.K. Kuhl, Maternal speech to infants in a tonal language: support
for universal prosodic features in motherese, Dev. Psych. 24 (1988) 14–20.
[13] K. Hubbard, D.A. Trauner, Intonation and emotion in autism spectrum disor-
ders, J. Psycholing. Res. 36 (2007) 159–173.
[14] L. Kanner, Autistic disturbances of affective contact, Nervous Child 2 (1943)
217–250.
[15] K. Lamers, L.J. Hall, The response of children with autism to preferred prosody
during instruction, Focus Autism Other Dev. Disab. 18 (2) (2003) 95–104.
[16] M.M. Lewis, Infant Speech: A Study of the Beginnings of Language, Routledge
and Kegan Paul, London, 1951.
[17] J. McCann, S. Peppé, Prosody in autism spectrum disorders: a critical review,
Int. J. Lang. Commun. Dis. 38 (2003) 325–350.
[18] R. Paul, A. Augustyn, A. Klin, F. Volkmar, Perception and production of prosody
by speakers with autism spectrum disorders, J. Autism Dev. Dis. 35 (2005)
205–220.
[19] S. Peppe, J. McCann, F. Gibbon, A. O’Hare, M. Rutherford, M. Receptive, Expres-
sive prosodic ability in children with high-functioning Autism, J. Speech Lang.
Hear. Res. 50 (2007) 1015–1028.
[20] I. Rapin, M. Dunn, Update on the language disorders of individuals on the
autistic spectrum, Brain Dev. 25 (2008) 166–172.
[21] E. Schopler, R.J. Reichler, B.R. Reiner, The Childhood Autism Rating Scale for
Diagnostic Screening and Classification of Autism, Irvington, New York, 1986.
[22] S.J. Sheinkopf, Peter Mundy, D. Kimbrough Oller, Michele Steffens, Vocal atyp-
icalities of preverbal autistic children, J. Autism Dev. Dis., 30, 2000.
[23] L.D. Shriberg, R. Paul, J.L. McSweeny, A.M. Klin, D.J. Cohen, F.R. Volkmar, Speech
and prosody characteristics of adolescents and adults with high-functioning
autismandAspergersyndrome,J.SpeechLang.Hear.Res.44(2001)1097–1115.
[24] P.E.Tanguay,Pervasivedevelopmentaldisorders:a10-yearreview,J.Am.Acad.
Child Adol. Psych. 39 (9) (2000) 1079–1095.