A Children’s Sing-Along 1

!

Running Head: Children and Singing

A Children’s Sing-Along: Exploring Familiar Songs, Vocal Model Choice and
Pitch Accuracy
Empirical Thesis
Presented to the Faculty of the
Department of Psychology
Bates College
In partial fulfillment of the
Requirement for the degree of the
Bachelor of Arts
By
Lisa McClellan

Lewiston, Maine
December, 2009

 A Children’s Sing-Along 2

!
Acknowledgements

I would like to sincerely thank my advisor, Helen Boucher, for supporting me
through this process. Thank you to Su Langdon for all the recruitment help and
essentially saving my thesis, and to Gina Fatone for the musical side of things – helping
me remember all the ethnomusicology factors that humble (in a good way) these results.
Thanks to Dr. Annabel Cohen and the AIRS project for the foundation and inspiration.
I would also like to thank my parents and the wonderful companions I’ve found this
most unexpected semester – you made this process smooth and helped put things in
perspective.
Finally and most importantly, thanks to all the parents and wonderful kids that
participated in the study. I was amazed by the generosity and enthusiasm. I hope none
neglect to showcase their talent in the future – I was truly impressed and enjoyed singing
with everyone!

 A Children’s Sing-Along 3

!
Table of Contents

Abstract…………………………………………………………………….…..4
Introduction…………………………………………………………………....5
Understanding Singing: Process and Development………….................6
Pitch-Accuracy Assessment……………………………………………...8
The AIRS Project………………………………………………..………13
Current Study…………………………………………………………...17
Methods…………………………………………………………………….....18
Results………………………………………………………………………...20
Discussion……………………………………………………………………..22
References………………………………………………………………….....26
Appendices…………………………………………………………………....28

 A Children’s Sing-Along 4

!
Abstract

There is currently no standard method for assessing pitch-matching ability, and the
obstacles to achieving standardization include: lack of knowledge about singing
acquisition, analyzing faults and assessment component challenges. Dr. Annabel Cohen
is the director of a research collaboration about singing called AIRS: Advanced
Interdisciplinary Research in Singing. She has been developing an assessment method
called the AIRS Test Battery. After a pilot study conducted last year, there were several
concerns with the methodology. These included song choices, pitch models for
assessment procedures and standardization challenges. The current study helped address
these questions using a sample of 19 five to seven-year-olds. They were tested on their
familiar song preference and vocal performance, performance when using a recorded
vocal model versus a live vocal model, and their overall enjoyment of the procedure.
Results indicate that a recorded vocal model is an effective and enjoyable way to assess
pitch-matching, and that the song Brother John is appropriate to use as a ‘familiar song’
in the test.

 A Children’s Sing-Along 5

!

A Children’s Sing-Along: Exploring Familiar Songs, Vocal Model Choice and Pitch
Accuracy
The voice is a unique instrument and is subjected to the environment, emotions and
personal experiences. Because the singer is so deeply connected with the instrument,
singing can be profoundly expressive and personal, and if shared, can make the singer
vulnerable to the listener (Stohrer, 2006). Singing displays emotion through movement,
facial expressions, words, tone, and the singer’s personal interpretation of the piece. It
can be a personal activity or communal, connecting oneself, the world, and community
(Stohrer, 2006). This ability to invoke complex emotions and connections with others
makes singing a powerful tool that has considerable benefits.
Singing has direct and indirect mental, physical and social health benefits, and
research studies have shown that the act of singing can improve quality of life (Cohen,
2008). Physical benefits include positive effects on breathing and lung function, posture,
body control, relaxation, stress relief, physical activity and energy (Clift, Hancox,
Morrison, Hess, Kreuts & Stewart, 2009). It can also facilitate improvements in cognitive
functioning and academic learning (Cohen, 2008). Research has shown that singing can
greatly improve language ability and speech, and it is often used as a tool to help children
with reading and foreign language learning, as well as with speech therapy (Biggs,
Homan, Dedrick, Minick & Rasinski, 2008). Because it is so beneficial to participate in
singing, it becomes important to understand how people develop the ability, when it is
acquired, and why some are unable to sing ‘accurately.’

 A Children’s Sing-Along 6

!

Those to whom singing comes easily may realize how important and beneficial it
can be, but may not understand how complicated the process is. Vocalizing involves a
series of physiological and psychological responses that allow a person to hear, process a
song or series of pitches, remember these sequences, and decide how to reproduce them
(Welch, 1985).
Welch’s Schema Theory is a way to conceptualize this process (Welch, 1985). First,
a person hears a certain pitch and builds an understanding of where it is coming from, the
space in which it is created and the mechanism, such as the instrument. These define the
initial conditions of the pitch (Welch, 1985). Then the brain develops a desired outcome,
which is an idea of what the singer thinks the sound should be, and this combined with
the initial conditions creates a schema (Welch, 1985). Specifications for a personal voice
program are established, which are based on acquired information, previous responses
and past outcomes of singing attempts. The schema determines an expected sensory
response, also known as feedback, including physical reactions from the body (Welch,
1985). Then the pitch is created, and the actual feedback of bodily responses is compared
to the expected. Discrepancies imply error, and the schema attempts to fix the error
(Welch, 1985).
This complex vocalization and pitch-matching process is fully developed by around
age six. From birth until age five, the process of singing acquisition is taking place
parallel to language development (Aiello, 1994). Infants are immediately exposed to
pitch and tone variations from speech as well as music, but at five-months, infants
possess sensitivity to the sequential structure of melody1. This is supported by studies
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
"!#$%&'(!)*!'$+),$'!-(!./$!01).2/$'!*&3,'*!4554,6$'!),!73*)24%!.)7$!),!422&5'4,2$!8)./!6)9$,!23%.354%!

2&,9$,.)&,*!4,'!2&,*.54),.*!:;<+&5'!=)2.)&,45(!&+!#3*)2!;,%),$>!?@@ABCD!E,!F$*.$5,!73*)2>!8$!5$2&6,)G$!4!

 A Children’s Sing-Along 7

!

showing that a change of melody results in a change of heart rate (Aiello, 1994). At three
years-old, children understand constant modulation2, ups and downs in pitch, pitch and
melody distinction, and they can divide songs into phrases, but with no constant interval
in-between. This means that between phrases of a song, they do not appreciate that there
is time consistent with the song rhythm before the next phrase begins. They also cannot
appreciate stable scale3, meaning that they will often repeat a series of pitches in a
different scale or change the scale mid-song (Aiello, 1994). At four, children can usually
hum the melodic contour correctly, but they have to form words with their lips because
they are dependent on the text to keep their place in the song (Aiello, 1994). Some notes
and intervals are usually inaccurate, but they have achieved greater pitch stability. The
tonality and scale may be inaccurate, and they have stronger rhythmic understanding, but
cannot necessarily maintain a steady beat (Aiello, 1994). At age six, children can
consistently match the model, and have gained a fully developed understanding of
melody.
After understanding the process of singing and grasping the stages of development
and the age at which people have the potential to hear and produce accurate pitches, the
question still remains as to why, at age six and older, some are unable to remember a
series of pitches and produce them. The way to approach this complex question is by first
creating effective ways to assess pitch-matching, pitch memory and overall melodic
understanding. The following studies have addressed questions of both assessment and
pitch-accuracy. These include the ways in which to assess pitch-accuracy in a western
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*1$2)+)2!*$H3$,2$!&+!5/(./7!4,'!1).2/!*$H3$,2$*!25$4.),6!4!-$6),,),6>!7)''%$!4,'!$,'!.&!8/4.!8$!8&3%'!
5$2&6,)G$!4*!4!2&71%$.$!7$%&'(C!!
?!!"#$%&'(")!)*!2/4,6),6!I$(*!'35),6!4!1)$2$!:;<+&5'!=)2.)&,45(!&+!#3*)2!;,%),$>!?@@ABC!E,!74,(!7$%&')$*>!I$(!
2/4,6$*!/411$,!),!2&,*.4,.!14..$5,*C!
J!K!*24%$!)*!4!0*$H3$,2$!&+!,&.$*!),!4*2$,'),6!&5!'$*2$,'),6!&5'$5!&+!1).2/!:;<+&5'!=)2.)&,45(!&+!#3*)2!;,%),$>!
?@@ABCD!L24%$!*.532.35$*!45$!'$1$,'$,.!&,!./$!*1$2)+)2!23%.35$C!!

 A Children’s Sing-Along 8

!

music context, and ideas about cross-cultural assessment, programs to improve pitchaccuracy and theories as to why experience with singing can improve accuracy.
Children Pitch-Matching Assessment
The Singing Development in Early Childhood Project, described by Welch (1994),
was a three-and-a-half year study at the Roehampton Institute in the UK, beginning in
1990. The purpose of this project was to test a newly developed assessment method and
chart the nature of children’s developing singing competences in relation to a range of
singing tasks that are common in elementary classrooms. A group of 200 five-year-olds
were tested then monitored over the next three years.
There were four elements within the testing protocol, highlighting dominant
features in pre-school and early elementary singing. These included: glides, patterns,
single pitches and songs (Welch, 1994). Songs were those familiar to elementary
children, patterns were three-five note sections of these songs without words, single
pitches were also taken from the songs and the glides were abstractions of the musical
contour, presented as glissando4. The participants were taped after hearing a vocal sample
then their renditions were compared.
The responses were recorded onto digital audio tape using a portable recorder. The
data were analyzed by using both a ‘machine-based’ method as well as a ‘human-based’
method. The machine-based analyses was able to provide information about the
underlying physiological bases for children’s singing, and did this by comparing the
participant’s voice with the voice of a trained singer model. Computers were able to
measure wave-lengths and frequencies to assess how accurate each participant’s
performance was. The problem with this method, however, was that the child’s
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
M!N%)**4,'&!)*!4!*%)'),6!7&9$7$,.!+5&7!&,$!,&.!.&!4,&./$5C!

 A Children’s Sing-Along 9

!

musicality and melodic understanding went unnoticed. Also, due to overtones and
inconsistent notes (e.g. singing a single note, but going flat at the end), it was sometimes
hard to tell if the participant got the note correct at all.
The ‘human-based’ analysis consisted of six trained singers assessing and rating
performances. This gave a subjective, yet educated, understanding of pitch accuracy and
melodic qualities that the machine could not distinguish. The panel could also factor in
age, gender and other contextual factors that may alter understanding of a certain
performance.
The article makes the point that with any method of analysis, there will always be a
prominent ‘human’ element. The way that singing, in both a melodic and pitch-accuracy
sense, is approached in these studies is from a western, adult perspective. A cultural
musical model is subjective, created by various groups and monitored by those culturallyspecific rules. The protocol for a test like this would have to look completely different in,
for example, India, where they use a different scale system. Also, because specific
cultural melodies and scales are learned through exposure, infants and young children do
not necessarily have a solid understanding of these musical rules, not because they are
not musical, but because they have not been well-versed in their specific cultural rules.
There is always bias in analysis, so it is important to take this into account when
generalizing children’s pitch-accuracy and ability.
Understanding Human-Based Analysis Methods
As shown in The Singing Development in Early Childhood Project (Welch, 1994),
singing is generally assessed one of two ways, namely: ‘machine-based’ and ‘humanbased’. There are several ways in which to accomplish both, and the human-based

 A Children’s Sing-Along 10

!

method in particular can be difficult to give accurate, objective results. Dr. Bradley
Frankland, a professor at Dalhousie University in Novia Scotia, has a particular research
interest in psychometric assessment and relating these methods to singing pitch analysis.
In a recent correspondence, he described two methods of rating qualitative
analyses: Thurstone’s method of comparative judgment and Thurstone’s method of
categorical judgmen.. Both methods require a system that is based on a rating scale, but
having different methods of analyzing and providing an appropriate way of rating
participant performance (Frankland, 2009). The first method rates each participant in
comparison with all other participants’ performances, and the latter has pre-determined
qualifications for each rating. These methods are not based on musical assessment and
were originally created for educational analyses and questionnaire scaling, but they can
be manipulated and used for singing pitch analysis (Frankland, 2009).
Thurstone’s method of comparative judgment, when applied to singing assessment,
is beneficial because it only has to involve one judge (Frankland, 2009). It is comparative
because each participant’s sample is compared to every other vocalization using a twoby-two comparison. This results in an n(number of participants) X n matrix, and the most
accurate vocalizations are seen by how many times they were rated ‘better’ than the other
renditions (Frankland, 2009). The difficulty with this method is that it is time-consuming
and tedious. Also, because the comparison session takes so long, there may be a shift in
what the judge considers ‘better’. It is also important to having the criteria of what is a
better rendition clearly defined (Frankland, 2009).
The other method, analyzing the singing data based on Thurstone’s method of
comparative judgment, is less time consuming and requires an ‘expert panel’. This

 A Children’s Sing-Along 11

!

consists of rating each vocalization on a Likert scale from accurate to inaccurate
(Frankland, 2009). This method is a more questionable quality of standardization, and is
not guaranteed to be as reliable or valid. The more people you have on the panel,
however, the more reliable it will be. Every judge rates the particular participant’s
vocalization, and then an average of these scores is used (Frankland, 2009).
Another important aspect of this method is the scale criteria. The scale must be
clearly defined, especially the end points (i.e. the lowest score and the highest score). The
judges must know what constitutes a ‘bad’ or ‘inaccurate’ rendition and a ‘good’ or
‘accurate’ rendition (Frankland, 2009). The method choice depends on time, resources
and sample number but both have been used effectively.
Singing Assessment and Improvement with Feedback
A computer-based singing improvement and assessment program called SINGAD
(Singing Assessment and Development) (Howard & Welch, 1987; 1989 from Welch,
1991) was designed to assess pitch-matching and provide insight into the nature and
significance of visual feedback to improve pitch-matching (Welch, 1991). The
participant’s vocal recording was graphed on a computer screen in terms of frequency
(Hz), and this was compared to given samples. There were several options of assessment,
and the choice depends on the developmental stage of the singing participant. The
participant could use a blank screen, with no targets, which allowed a visual perspective
of his/her own singing. Another option was to have a choice of four visual targets, which
could be altered in both horizontal and vertical planes by using arrows on the keyboard.
The final option was to use a single visual target preceded by its auditory pitch
equivalent, allowing the participant to hear the pitch before attempting to match it

 A Children’s Sing-Along 12

!
visually (Welch, 1991).

In a 1991 study, Welch used the SINGAD system to assess singing improvement in
a group of 29 five-year-olds. The study covered the children’s first full year of school,
and the teacher was asked to set up the computer system in the classroom and fit its use
into the daily routine. They ended up spending 10-15 minutes playing with the system on
a weekly basis. The children created pitch patterns on the monitor screen, played with
one or more visual targets and discussed what they saw.
The assessment software monitored students’ vocal pitch accuracy change over the
school year. The children received three formal assessments over the course of these
seven months. In respect to pitch-matching improvement, results indicated significant
improvements in pitch matching, as well as differences in accuracy between varying
target pitches.
Several limitations were mentioned in the discussion including choice of target
pitches and a lack of melodic dimension. It is possible that patterns of pitches, as opposed
to single pitches, may be easier to remember and match. This follows the developmental
stages of musical understanding, considering that melodic patterns or contour are more
easily matched or matched before individual pitches. This suggests that the lack of pitch
matching ability may be due to task difficulty (Welch, 1991). The teacher supported this
notion, offering that the children appeared to be more accurate over the course of the year
when singing songs and when engaging in singing games compared with when they were
being assessed by the computer. Further studies could address the discrepancies between
single pitches and melodic lines, possibly by using both in the pitch matching assessment.

 A Children’s Sing-Along 13

!
AIRS: Advanced Interdisciplinary Research in Singing

A research initiative attempting to fill in research gaps associated with singing
development and acquisition is currently underway. There have already been exciting
findings building off of past research about singing development, and the ultimate goal is
to further understand the ways in which singing can enhance overall health. The project is
called Advanced Interdisciplinary Research in Singing (AIRS).
AIRS is an international research collaboration exploring the properties and
benefits of singing from both theoretical and applied perspectives. The project is based
out of Prince Edward Island, Canada, where the project director, Dr. Annabel J. Cohen,
teaches and conducts research at the University of Prince Edward Island, and 2009 is the
first full year of the project’s seven year duration.
There are three themes within this research and each theme is divided into three
sub-themes. The first is The Development of Singing, containing Singing vs. Speaking,
Multimodal Measures of Singing and Speaking, and the AIRS Test Battery. Theme 2 is
Singing Education divided into Learning to Sing in Natural Settings, Formal Teaching of
Singing and Using Singing to Teach. The final theme is Singing and Well-Being which
explores Cross-cultural Understanding, Intergenerational Understanding and the Mental
and Physical Health Benefits of Singing (AIRS Website, 2009).
Dr. Cohen’s specific research interest is within Theme 1 and development of the
AIRS Test Battery. For the past year she has been developing a battery and protocol for
obtaining audiovisual information assessing people’s ability to listen, process and
produce sung pitches. In a pilot study, six children ages three, five, and seven-years-old
as well as undergraduate students and Alzheimer patients have been tested.

 A Children’s Sing-Along 14

!

In order to analyze the audiovisual material collected from children, Dr. Cohen
envisioned a repository to study singing acquisition which is called CHIMES (Children’s
International Music Exchange System) (Cohen, 2000). This is structurally inspired by a
similar data repository, developed by MacWhinney and Snow in 1984, called the
Children’s Language Data Exchange System (CHILDES). This database is universally
accessible on the Internet to store word utterances collected by researchers and has aided
in the production of over 2,000 research papers (Lannan, 2009 from MacWhinney, 2007).
Dr. Cohen imagines a similar use for CHIMES, hoping that it will be an easily accessible
way for people to store and share data, as well as analyze singing material.
An additional inspiration for the structure of this research and style of analysis is
work done by Stadler Elmer, resulting in an article published in 2000 and research still
being conducted. Stadler Elmer devotes much of her research to children’s musical and
singing behavior, trying to grasp a cross-cultural understanding of how children listen to
singing, their vocal pitch accuracy, as well as melodic understanding, movement and
gestures associated with singing, rhythmic perception and lyrical choices and reactions
(Stadler Elmer & Elmer, 2000). In the 2000 article, the authors described limitations of
the available singing analysis methods and present a new method. They were concerned
that the current systems were too strongly based on bias evaluation because they were
being done by trained musicians and notated in predominantly western music symbols.
They argue that this could work with adult singers of western influence, but will present
profound problems if used with children and infants who have not had much exposure to
popularized music as well as populations that are not western (Stadler Elmer & Elmer,
2000).

 A Children’s Sing-Along 15

!

Stadler Elmer’s goal was to create a system that could analyze and store singing
data in a way that would work with anyone and have it done electronically in order to
avoid musical bias. This would create a ‘singing notation language’ so people across all
cultures could create and exchange data related to singing research (Stadler Elmer, 2009).
Dr. Cohen’s wish is to utilize this analysis method using the AIRS Battery model and the
data extracted from it.
AIRS Test Battery
The battery contains 11 components in order to capture information on verbal
skills, melodic and pitch memory, perception and production, and creativity. The test is
designed to have the researcher play a piano recording and sing a series of songs and
scales and ask the participant to repeat each, as well as using several other components to
assess melodic understanding and creativity. See Appendix A for a detailed description
of the components. Last year, Dr. Cohen and colleagues tested six children once a month
over the course of five months by using the battery and videotaping their interaction with
the children.
The videos were imported onto a computer, audio was extracted from these clips
and then analyzed by a solely ‘machine-based’ method. Each movie was converted into
an audio file by Mpeg stream clip and further edited by using a program called Audacity.
The analysis procedure consisted of two parts: ‘Pitch Analyzer’ and ‘Notation Viewer’.
Each pitch, time of word onset, stability of the word and the song timing all could be
viewed when the audio clip was opened in the pitch analyzer program. This information
was then given a syntax, represented by a textual code written out by the researcher into a
text file using a program called Text Wrangler.

 A Children’s Sing-Along 16

!

Each text file was then opened in Notation viewer in order to create a graph model
of each singing segment. This analysis and graphing procedure was used with component
10, Brother John Reprise. (Lannan, 2009)
Results describe the findings acquired from the qualitative analyses and a
performance overview of the other components. There was consistent variation with age,
specifically older participants performing more accurately and creatively during various
components, and they also had larger vocal ranges. Component 4, Brother John, proved
more difficult for the younger children, and participants of all ages had some trouble with
the words. There was significant improvement, however, over the course of the five
sessions.
To understand the pitch-matching using a computer analysis method, Component
10 – Brother John Reprise was analyzed, and the researchers were able to find the pitch
accuracy of each note. They were only able to do this with the five and seven-year-olds,
because data for the three-year-olds was not complete due to the errors and
incompleteness of songs. By comparing the participants’ rendition and the correct pitches
of the 32 syllables in Brother John, the researcher was able to create a line graph of the
semitones of the five-year-olds and the seven-year-olds, and calculate the mean absolute
pitch error for each age group. The mean absolute error for five-year-olds was higher
than the seven-year-olds. The researcher also found the mean relative pitch error in order
to appreciate if the participants were over or underestimating the pitches. The five-yearolds tended to either over or underestimate the pitches more so than the seven-year-olds,
resulting in a lower mean relative pitch error. (Lannan, 2009)

 A Children’s Sing-Along 17

!
Future Battery Refinement

This pilot study successfully displayed age differences in children’s pitch-matching
and provided the researchers practice with using this type of testing method. It also
offered valuable experience with using complex computer-based methods to chart pitches
and compare participant’s renditions with a given ‘accurate’ model. However, a concern
arose while reviewing the battery components and the videos regarded the sung
components that the children were asked to copy. The way it was done was by using a
piano recording of each musical component, the researcher sang along with the recording
and the child was asked to repeat. There were two main issues with this structure. First,
the researcher was required to sing along with a recording of the correct pitches, but there
was no guarantee that the researcher sang the pitches accurately. The ultimate plan is to
use this battery in many different settings with various people conducting the research. It
is confusing for the participant to hear correct pitches on the piano and incorrect pitches
from the researcher’s singing, and this will affect the participant’s rendition.
Another issue is having two ‘voices’ in general that the participant must listen to:
the piano recording and the researcher’s singing voice. Even if the researcher is singing
the correct pitches, it may be too much cognitive processing listening to two timbres or
‘voices’.
Current Study
In light of these concerns, a plan was developed to record vocal examples of each
sung component so the participant can listen to and repeat pitches from a standardized
recording of a voice. I completed these recordings during my work with the AIRS project
this past summer. The two vocal models, a live voice and a recorded voice, were used in

 A Children’s Sing-Along 18

!

the current study to assess children’s pitch-accuracy and their procedural enjoymentC This
was achieved by using an extension of component 4 and 10 of the AIRS Battery, not the
battery in its entirety.
Another potential battery alteration relates to the choice of song for component 4 –
Familiar Song: Brother John. This component is meant to assess memory, pitch accuracy
and melodic understanding using a familiar song. There was question, however, as to
how familiar this song is and how challenging. Various AIRS research members believe
that it is too complicated and not recognizable enough. Other song suggestions include
Happy Birthday and Twinkle, Twinkle Little Star. It became useful to evaluate this
question, and decide which song would be most appropriate.
Method
Participants
The sample size was 19 five through seven-year-olds. Of these participants, nine
were in the live model group and 10 in the recorded vocal model group. Initially age six
was preferred, but due to recruitment challenges, children who were five and sevenyears-old were included in the sample. They were recruited from Bates College
faculty/staff children and friends and the Boys and Girls Club in Auburn, Maine.
Measures
One measure was an extension of the fourth component in the AIRS Battery:
Familiar Song. There were three songs tested, including Brother John, Twinkle Twinkle
Little Star, and Happy Birthday. See Appendix B for music and lyrics. A vocal recording
was used as the sample for each participant to repeat, as well as a live singing voice and a
pitch pipe during testing. The children were asked which song was most familiar to them,

 A Children’s Sing-Along 19

!

and a series of five smiley faces indicated their reaction to the procedure. This was coded
on a 1-5 scale, with 1 corresponding to a “No fun” frowning face, and 5 indicated the
participant chose the most smiley face, with “Very fun” written underneath. A digital
audio recorder recorded each testing session.
Procedure (see Appendix C for script)
After recruiting participants, gaining IRB and parental consent (See Appendix E for
consent form), participants were assessed one-on-one on their pitch matching ability
using the three songs. The sample included two groups, both with the same procedure,
except one heard vocal samples on a recording and the other heard the live singing during
the actual testing with a pitch pipe indicating the starting note.
First, the child was asked their name, age and birthday. Then he/she heard singing
or the vocal recording of one familiar song and was asked to repeat. Then this was
repeated with the other two songs. The order of songs varied with each participant. The
child was then asked which song was most familiar. Finally, they received a sheet of
paper with five faces on it. The faces were a progression of expressions from frowning to
smiling, and the child was asked to color in the face that best represents how he/she felt
during the singing session, specifically about the model and singing repetition. See
Appendix G for the paper that they were given. The participant was then presented with
small tokens of appreciation (stickers and necklace) and thanked for his/her participation.

 A Children’s Sing-Along 20

!
Results

Of the 19 participants in this study, 18 were assessed on their pitch accuracy. One
participant in the live vocal model group’s vocalizations could not be used because she
did not sing the songs on her own, but sang with the researcher. This participant’s
enjoyment rating and familiar song choice, however, were used. For analyzing pitchaccuracy, an ‘expert panel’ was used as a human-based method. The panel consisted of
two trained singers at Bates College, not including the researcher, and they listened to the
sample and the participant rendition, and rated the participant on their accuracy compared
to the sample. They used a method inspired by Thurstone’s method of categorical
judgment to rate each participant’s vocalization. The experts individually rated each song
on a 1-5 scale then they averaged these scores. A score of 1 meant that the participant’s
rendition did not in any way musically match the model, and a score of 5 indicated a
nearly identical pitch-matching compared to the given model. This average was used as
the rating for each participant. See Appendix D for the form they were given for
instructions and scoring. An alpha test was run across all ratings and songs, ! = .93. This
high alpha value suggests that the experts’ rating were similar to one another and
consistent, therefore, using an average of experts’ ratings was justified.
The data from this project is meant to further refine the AIRS Battery. Results
address which of the three song choices participants perform better on and which they
prefer and recognize. In addition, results indicate which method, singing or a recording,
is more effective while testing.
An independent samples t-test was performed to compare mean ‘fun-rating’ of each
vocal model method, and a separate t-test was run to compare ‘pitch-ratings’ of each

 A Children’s Sing-Along 21

!

vocal model. Results indicate no significant difference between the fun-rating scores of
participants who heard a recorded vocal model (M = 4.50, SD = .97,) and those who
heard a live vocal sample (M = 4.00, SD = 1.00,), t(17) = 1.11, p = .29. There was also no
significant difference between pitch performance of participants who heard a recorded
vocal model (M = 3.71, SD = 1.07) and those who heard a live vocal sample (M = 3.79,
SD = .77), t(16) = !.17, p = .87.
To assess which song was most familiar, a chi-square test was run to compare the
number of ‘familiar song’ votes for each of the songs: Brother John, Twinkle, and Happy
Birthday. Results indicated that Happy Birthday was picked 8 times which was the
highest amount, followed by Brother John with 6 votes and Twinkle was picked the least
number of times with 4 participant picks, "#(3) = 5.63, p = .13. These differences were
not significant, suggesting that songs were equally familiar.
Pitch-ratings of all participants were averaged for each song to give every song a
separate pitch-rating score, and these means were compared using a within-subject
ANOVA. Twinkle received the highest overall pitch-rating score (M = 3.9), followed by
Brother John and Happy Birthday which had identical scores (M = 3.7), F(2,34) < 1.
These differences were not significant, indicating that participants performed all three
songs with equal accuracy.

 A Children’s Sing-Along 22

!
Discussion

Results suggest that participants enjoyed and performed equally well when
presented with both vocal models: vocal live and recorded. Data analysis also revealed
that each of the three songs, Brother John, Twinkle Twinkle Little Star, and Happy
Birthday, were equally as recognizable to participants. Finally, participants did not
perform significantly better on any one song, suggesting equal performance for all. In
summary, these findings suggest that the recorded vocal model method is an effective
way to test pitch accuracy of children participants, and that any of the songs would be
appropriate to use as a “Familiar Song” in a pitch-accuracy assessment test.
These results are meant to refine a current method being developed to assess
singing pitch-accuracy. The AIRS Battery has been undergoing testing in multiple
locations on its reliability, but questions still remained about how best to present singing
models for participants to repeat and which song to use for the ‘Familiar Song’
component of the test. These results suggest that Brother John is an appropriate song to
use in terms of child participant performance and recognition. Also, a recorded vocal
model is an effective way to present participants with standard, consistent models that
they are meant to repeat. The participants responded very well to this manner of testing
and the researcher found it efficient and easy to use. The question of whether or not
model pitches and intervals were accurate no longer was an issue, which will prove
useful when this battery is used in various settings and locations with different
researchers administering it.
There were several study limitations that prevented the findings to be as valid as
they could have been. Recruiting participants was a challenge, considering lack of

 !

A Children’s Sing-Along 23

accessibility to children and time constraints. Having 19 participants was not enough to
get reliable results. Another limitation that was mainly due to recruitment difficulties was
the age range. Age six was ideal for this study, as indicated in previously discussed
literature on singing development and optimal ages for pitch-matching. Having all
participants the same age would have eliminated the age variable, which was not a focus
in this study. An additional limitation was the different locations and circumstances in
which the testing took place. Some participants were tested in a lab which provided a
quiet setting, others were tested in a house with considerable noise and distraction, and
the remainder of participants were tested in an after school program site which was also
rather noisy. Ideally all participants would have been tested in the same quiet location.
A significant limitation was the fact that the pitch analysis was only human-based
and not machine-based. Ideally, the audio of participants singing would have been
imported into a computer program and the pitches would have been compared with the
model in terms of pitch frequencies as described in the AIRS Battery pilot study.
Unfortunately, the researcher did not have access to a program like this. It would have
been helpful to have a human-based analysis with more experts and a machine-based
analysis, providing a more valid assessment of pitch-matching.
Additional concerns pertain to the testing procedure. When participants were asked
how much fun the singing experience was, it seemed difficult for them to distinguish
between the actual task being enjoyable as opposed to the whole experience. If they
found the researcher’s company fun, or if they did not want to hurt the researcher’s
feelings, they may not have given truthful ratings of how much they enjoyed strictly the
singing aspect. The final concern related to the lack of ‘standardization’ between

 A Children’s Sing-Along 24

!

participants. The researcher was focused on making the participants feel comfortable and
enjoy themselves, therefore, conversations, time spent with each participant, and exact
tasks varied slightly between participants.
Despite these limitations, the current study did successfully address the concerns
that arose after the AIRS Battery pilot study. By understanding the appropriate familiar
song choice and which sample works best, the AIRS Battery will be a more effective way
to measure pitch-accuracy and ‘western’ melodic understanding. The purpose of this
study was to improve this particular assessment method, making it as reliable and valid as
possible. By refining one method of analyzing vocal ability, it may become possible to
adapt this basic model to other cultures. There are problems with this goal, however,
because each culture has a unique musical structure. The familiar songs vary, as well as
languages and scales, and the only people who could successfully create a protocol for a
specific cultural musical style are those who are members of that culture or well-versed in
the musical style. Further research could address these issues and find a way to unite
assessment without assuming that strict standardization is necessary.
Singing is a complex progression from interpreting notes and phrases, to mentally
processing these cues, and finally expressing the singer’s own interpretation. The AIRS
vocal assessment, or an interpretation of the test, could help give vocal and
developmental researchers a better sense of singing acquisition. Results from the
assessment could lead to studies exploring why some people develop the ability to sing
and some do not, and what environmental factors contribute to this discrepancy. It could
also address the question of ‘innate’ versus ‘learned’ ability, and how natural ability
factors into musical achievement. By learning more about singing acquisition and

 !

A Children’s Sing-Along 25

development, researchers could better understand how people are able to sing, where
interest and singing motivation originate, and ultimately try to encourage more to
participate. Singing is a valuable resource in promoting healthy living, social
connectedness and personal relaxation, among other things, and many could benefit from
more singing activity.

 A Children’s Sing-Along 26

!
References

Aiello, R. (1994). Musical Perceptions. New York: Oxford University Press.
Biggs, M.C., Homan, S.P., Dedrick, R., Minick, V., & Rasinski, T. (2008). Using an
interactive singing software program: A comparative study of struggling middle
school readers. Reading Psychology, 29, 195-213.
Clift, S., Hancox, G., Skingley, A., Morrison, I., & Bungay, H. (2009, June). Community
Singing for Wellbeing and Health: Report on a Progressive Research Programme
within the Sidney De Haan Research Centre for Arts and Health, UK. Paper
presented at the University of Prince Edward Island, Charlottetown, PEI, Canada.
Cohen, A.J. (2000). Development of tonality induction: Plasticity, exposure and training.
Music Perception, 17, 437-459.
Cohen, A.J. (2008). AIRS Research Environment: Advanced Interdisciplinary Research
in Singing. Retrieved July 15, 2009, from the AIRS: Advanced Interdisciplinary
Research in Singing Website: http://vre.upei.ca/airs/
Cohen, A.J. (2009, June). Developing a test battery of singing abilities with lifespan
application. Paper presented at the University of Prince Edward Island,
Charlottetown, PEI, Canada.
Frankland, Bradley. (2009, October). ‘Standard’ methods of assessment in psychometrics.
Retrieved October 26, 2009, from email correspondence.
Lannan, M.S. (2009). A New Test Battery of Singing Abilities: The Feasibility of a
Longitudinal Study of Children and Young Adults. Unpublished undergraduate
honors thesis, University of Prince Edward Island, Prince Edward Island, Canada.

 A Children’s Sing-Along 27

!

Mithen, S. (2007). The Singing Neanderthals: The Origins of Music, Language, Mind
and Body. Cambridge: Harvard University Press.
Oxford Music Online. (n.d.). Retrieved September 18, 2009, from
http://www.oxfordmusiconline.com/subscriber/
Stadler Elmer, S. (2009, June). A computer based method for analyzing singing. Paper
presented at the University of Prince Edward Island, Charlottetown, PEI, Canada.
Stadler Elmer, S. & Elmer, F.J. (2000). A new method for analyzing and representing
singing. Psychology of Music, 28, 23-42.
Stohrer, S. (2006). The Singer’s Companion. New York: Routledge.
Welch, G.F. (1985). A Schema Theory of How Children Learn to Sing In Tune.
Psychology of Music, 13, 1.
Welch, G.F. (1985). Poor Pitch Singing: A Review of the Literature. Psychology of
Music, 7, 1.
Welch, G.F. (1991). A developmental continuum of singing ability: Evidence from a
study of five-year-old developing singers. Early Child Development and Care, 69,
107-119.
Welch, G.F. (1994). The Assessment of Singing. Psychology of Music, 22, 3-19.

 !

A Children’s Sing-Along 28

Appendix A: AIRS Battery Components
1. Opening Conversation: This consists of a short conversation with the child that enables
the child to become more comfortable with the researcher, and it also gave the researcher
an idea of the child’s verbal skills and speech development.
2. Range Testing: The child hears a glissando while the researcher plays notes on a
xylophone and the child is asked to repeat. The purpose is to appreciate the child’s range.
3. Major Third Song: This warm up name game helps the child practice a major third
scale. The researcher initiates the exchange by singing the phrase, “(Child’s name) where
are you?” and the child is asked to respond, “(Researcher’s name) where are you?” This
involves no accompaniment or recording.
4. Brother John: This involves the child singing the song Brother John after hearing the
piano playing the melody on a recording. The researcher sings with the recording.
5. Favorite Song: The child is asked to sing his/her favorite song. This gives the
researcher an idea of musical preference and pitch accuracy.
6. Musical Elements: Several different musical scales and sequences are played on a
recording and the child is asked to repeat each. This helps evaluate pitch accuracy and
song performance elements. The scales include the first three notes of a major scales
(ascending and descending: do, re, mi, re, do), the first four notes (do, re, mi, fa, mi, re,
do), a major third (do, mi, sol), ascending major scale and descending major scale.
7. Improvise an Ending: This component is designed to assess the child’s grasp of scale
and melodic structure by asking him/her to create an ending to a simple song. It is also a
measure of creativity, and will aid in the comparison with language creativity.
8. Creation of a Song from a Picture Prompt: The child is asked to make up a song with a

 !

A Children’s Sing-Along 29

prompt that they chose from four picture choices. Songs will show the child’s creativity
and comfort with melodic structure.
9. Unfamiliar Song: The researcher sings a song, along with a recording of the piano
melody, by Carolyn MacDode called We Are One from the album My Heart is Moved.
The child is asked to sing back as much as he/she can remember after hearing it once.
This is designed to measure memory of an unfamiliar melody. This song was chosen
because it is an anthem for the natural environment and humanity and is based on the
Earth Charter which has been translated into over 35 languages.
10. Brother John Reprise: The child is asked to sing Brother John again without any
prompts. It is meant to assess memory over a 15 minute time period.
11. Closing Conversation: This is a parting conversation where the researcher thanks the
participant and informs the child when she will be coming again (if it isn’t the last
meeting). (Lannan, 2009)

 !

A Children’s Sing-Along 30

Appendix B: Sheet Music for Songs Used
!
!
!
!

!
!
!
!
!
!
!
!

!

 !

!
!
!
!
*(%+,---./+0/-12'130"$4-5+/6'"4-'7()6%+83(*9

A Children’s Sing-Along 31

!

 !

!
!
!
!
!
!
!
!
!

A Children’s Sing-Along 32

!

 !

A Children’s Sing-Along 33

Appendix C: Procedure Script
1.Conversation and Hello
“Hi, my name is Lisa, what’s yours? And what is your birthday – how old are you?
(Write on back of smiley face sheet) Great, thanks! Well I am a singer and love to sing
whenever I can, do you like to sing? I’d really like to share a few of my favorite songs
with you, you may already know them, and hear you sing! I always love learning more
about people’s favorite songs, what songs they know well and how people can feel more
excited about singing. So today I’m going to ask you to sing three songs: Brother John,
Happy Birthday and Twinkle Twinkle Little Star. Is that okay? You will hear the whole
song first (either by me singing or playing a recording), then parts of the songs, and you
will repeat those parts, and then you will hear the whole song again and repeat it exactly
as you hear it. Just do the best you can, and if you feel scared or you want to stop just let
me know, that’s totally fine. Then after we are done singing, we can talk about which
song was your favorite and how much you enjoyed singing all three songs. Does that
make sense? Are you ready to start? Let’s have some fun!”
2. Brother John (start on C) – Keep track of song order to assure variability
“I’m going to sing (play) a song called Brother John. Do you know that song? I’ll sing
(play) the whole thing, and try to listen carefully. Then I’ll sing (play) it in small parts
and you get to sing the smaller parts back. Then I’ll sing (play) the whole thing again,
and ask you to repeat the whole song exactly as you hear it. Ready?”
(Small parts - Exp: Are you sleeping? Are you sleeping? Brother John? Brother John?
Child: Are you sleeping? Etc. Next part - Exp: Morning bells are ringing, morning bells
are ringing. Ding dang dong, ding dang dong. Child: Repeats)
3. Happy Birthday (start on C)
“Now we will sing the song Happy Birthday. Would you like to pretend it’s your
birthday? Okay, let’s sing Happy Birthday to you, and we’ll do this the same way we did
Brother John. So you’ll hear the whole song, then sing it in parts, then the whole thing.”
(Parts – Exp: Happy Birthday to you, Happy Birthday to you Child: Repeats. Exp: Happy
Birthday to (name), Happy Birthday to you.)
4. Twinkle Twinkle Little Star (start on C)
“Last one is Twinkle Twinkle Little Star. So same deal, first hearing the whole thing,
then sections then singing it all. Because this one is a little longer, we can do it in three
parts. Ready?”
(Parts – Exp: Twinkle twinkle little star, how I wonder what you are. Child: Repeats. Exp:

 !

A Children’s Sing-Along 34

Up above the world so high, like a diamond in the sky. Child: Repeats. Exp: Twinkle,
twinkle little star, how I wonder what you are.)
5. Familiar Song
“Great job! Now can you tell me which of those three songs you recognized the most?
Which did you know the best and feel the most comfortable singing?”
6. Procedure enjoyment
(Give child smiley face sheet and crayon) “Now can you color in the smiley face that best
shows how you felt during the singing? If you thought it was really fun, as fun as it gets,
then you can color in the super smiley face all the way to the left, and if it was no fun at
all, then color in the frowning face. And if it was somewhere in between color in
whatever face you think makes the most sense.”
7. Closing Conversation
“Thank you so much for singing today, you did such an awesome job! I hope you had
some fun, I know I did!”
Materials:
Sheet with smile faces
Crayon
Audio recorder
CD player
CD model
Basic Script
Pitch Pipe
!
!
!
!
!
!

 A Children’s Sing-Along 35

!
Appendix D: Expert Panel Scoring Sheet

Participant
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

18

Brother
John

Twinkle,
Twinkle

Happy
Birthday

Average
Score

Notes

 !

A Children’s Sing-Along 36

Understanding the Rating Scale and Instructions
Thanks for being willing to perform the very important task of rating singing pitch
accuracy! This will help me understand what songs are more familiar and enjoyable for
children to sing, and which they are more likely to perform better on because of this
familiarity. I have not included names of the children for anonymity purposes, but please
rate the participants in order of how they appear on the CD. The participant number on
this rating sheet corresponds with the track name.
You will hear each participant sing three songs all starting on Middle C: Brother John,
Twinkle Twinkle Little Star and Happy Birthday. For each song, they are asked to listen
to the whole thing, then listen to sections and repeat those sections, then hear the whole
song again and repeat it. There may be variation depending on the participant, and there
is dialogue, so you are only meant to score the rendition of them singing the whole song
on their own (the other steps are just meant to familiarize/teach them the song). Please
listen to those parts carefully, and give each song a score between 1 and 5 (see below for
descriptions). Then average those scores.
Here is a description of what the numbers mean. The score is meant to reflect how well
the participant matched the model (you will hear either my voice or a recording of my
voice, but if you need another reference you may use the model recording) so if they sing
the correct melody but it is in a different key, this should be reflected in their score. Do
your best to rate the samples looking for (1) specific pitch mating (i.e. using the same
individual pitches as the model), (2) correct intervals between notes, (3) key, and (4)
overall melodic understanding. I’m not as worried about rhythm or vocabulary. Feel free
to add notes in the provided spaces if you want to explain your score or if it is difficult to
score a particular sample.
Score of 1: does not match the model, incorrect key, no understanding of intervals, many
incorrect notes, no melodic understanding
Score of 2: parts are similar to the model, three of these components are lacking
Score of 3: kind of matches the model, two components are lacking
Score of 4: almost matches the model, one component is lacking
Score of 5: matches the sample, correct key, correct intervals, correct notes, thorough
understanding of the melody
Let me know if you would like more detail about my study and thanks again!!!

 A Children’s Sing-Along 37

!

Appendix E: Informed Consent
Parental Informed Consent Form
I, _________________________, have read the attached letter describing a study to be
conducted under the auspices of Bates College, Lewiston, Maine, Department of
Psychology, to be conducted by Lisa McClellan and Helen Boucher (advisor).
I hereby consent to the participation of my child, _________________________, in this
study.
I understand that I am free to discontinue such minor’s participation at any time without
suffering any disadvantage.
I also understand that the findings of the study will be interpreted and recorded only on a
group basis with no identification of specific individuals. All information about
individuals will be held in confidence, to the extent permitted by law. I also understand
that the sessions will be audio-recorded, but these recordings will remain confidential to
all except the researcher and her advisor, Helen Boucher, as well as Annabel Cohen.
After the data analysis is complete, the tapes will be kept and monitored by the researcher
and Annabel Cohen. Only Annabel Cohen, the researcher and AIRS-affiliated students
working on analysis refinement will have access to these tapes.
I understand that at that time, I have a right to a complete explanation of the nature and
purpose of the study. If I have any questions or wish further information about the study,
I understand that I may call Lisa McClellan at 610-986-5478 or email at
lmcclell@bates.edu. I further understand that if I have additional questions that may not
be answered by the researcher, I may call her advisor, Professor Helen Boucher at 207786-6395, or email hboucher@bates.edu, or contact the chair of the Department of
Psychology, Professor Michael Sargent at 207-786-6277 or email msargent@bates.edu.
In addition to consent for my child’s participation, my signature confirms that I have
received a copy of this consent form together with any attachments which describe the
research to be conducted.

Dated__________

____________________________________
Signature of Parent/Guardian

 !

A Children’s Sing-Along 38

Thank you for considering the possibility of having your son or daughter
participate in this study. Before making a decision whether or not you would like to give
consent, it is important to understand the study, be aware of what the procedure will be,
and know that you are able to decline consent or have the option of discontinuing your
child’s participation at any time.
Study Overview
Singing participation has many health benefits, both mental and physical, and
previous research has shown that varying populations have benefited from more singing
experience in everyday life. In order to explore these benefits and promote participation,
researchers have become interested in how, as children, people develop the ability to
sing, why some are more successful and interested in singing than others, and how
singing ability can be improved to enhance the overall experience. In order to do this,
researchers have used various methods to assess vocal ability by means of pitchmatching, which is the process of hearing various pitches or sequences and repeating
them. There is no definitive reason why only some have the ability to match vocal pitches
accurately, and there is currently no standard method for assessing pitch-matching ability.
Dr. Annabel Cohen is the director of an international research collaboration about singing
called AIRS: Advanced Interdisciplinary Research in Singing. She has been developing
an assessment method called the AIRS Test Battery, and is working to make this an
international model for assessment. The purpose of the current study is to test possible
improvements in administering a shortened version of the AIRS battery. A sample of first
grade children will be tested on their pitch-matching, asked about their song preference
from a choice of three songs, and rated on overall procedural ease when using a recorded
vocal model versus a live vocal model. The final objective is to create one model video
for the battery to promote standardization for future researchers using this testing method.
Procedure
Participants will be first grade students (or between the ages of five and seven),
and they will be split into two groups. One group will be the ‘live vocal model’ sample,
and the other will be the ‘vocal recording model’ sample. Both groups will receive the
same procedure, and the sessions will be taped with an audio recorder. Each participant
will be tested individually, and the session will be approximately 15 minutes long. For
each test, three songs usually familiar to children will be used including Brother John,
Happy Birthday and Twinkle, Twinkle Little Star.
For the ‘live vocal model’ group, each child will hear the researcher sing a song
once, then the researcher will sing the first half of the song and ask the child to repeat the
section. Then the child will hear the second section and be asked to repeat. Finally, the
researcher will sing the whole song and ask the child to repeat. This will be done with
each of the three songs. After this is finished, the researcher will ask the child which song
was his/her favorite and which he/she knew the best. The ‘vocal recording model’ group
will have the same procedure, but instead of hearing the researcher’s voice as the model,
they will hear a recording of the researcher’s voice.

 A Children’s Sing-Along 39

!

Participants will be thanked, and told that they are helping in the process of
finding out first graders’ favorite songs and songs that they are most comfortable with.
They will also be told before the testing that if they do not feel comfortable singing that
they can stop.
Appendix F: Graphs of Results

O37-$5!&+!P45.)2)14,.!Q&.$*!

!"#$%$"&'()*+',-)$./0'

R3,!S4.),6!L2&5$!

1&)./23&/'4*5)6#/*7'"*2'8)2/%'8/7-)2'

 A Children’s Sing-Along 40

!

1$7.-'9":*+'()*+',)#;"&$0)*'

Average Pitch Score

'

K9$546$!P).2/!S4.),6!L2&5$!

1$7.-'9":*+'"*2'8)2/%'8/7-)2'

 !
!
!

Song Pick:

Birthday:

Name:

Appendix G: Sheet Given to Participants to Assess Enjoyment

!

A Children’s Sing-Along 41