See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220663218

Development of real-time visual feedback

assistance in singing training: A review

Article in Journal of Computer Assisted Learning · August 2006

DOI: 10.1111/j.1365-2729.2006.00178.x · Source: DBLP

CITATIONS READS

29 131

3 authors:

David Hoppe Makiko Sadakata

5 PUBLICATIONS 52 CITATIONS University of Amsterdam
40 PUBLICATIONS 217 CITATIONS
SEE PROFILE

SEE PROFILE

Peter Desain
Radboud University Nijmegen
176 PUBLICATIONS 2,449 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Makiko Sadakata
letting you access and read them immediately. Retrieved on: 18 September 2016
 Review article

Development of real-time visual feedback

assistance in singing training: a review
D. Hoppe, M. Sadakata & P. Desain
Music, Mind, and Machine Group, NICI, Radboud University Nijmegen, HE Nijmegen, The Netherlands

Abstract Four real-time visual feedback computer tools for singing lessons (SINGAD, ALBERT, SING &
SEE, and WinSINGAD), and the research carried out to evaluate the usefulness of these
systems are reviewed in this article. We report on the development of user-functions and the
usability of these computer-assisted learning tools. Both quantitative and qualitative studies
confirm the efficiency of real-time visual feedback in improving singing abilities. Having
addressed these findings, we suggest further quantitative investigations of (1) the detailed
effect of visual feedback on performance accuracy and on the learning process, and (2) the
interactions between improvement of musical performance and the type of visual feedback
and the amount of information it presents, the skill level of the user and the teacher’s role.

Keywords real-time visual feedback, review, singing education.

Introduction Welch (1985) developed a schema theory of the

learning of the classical singing class as illustrated in
The classical method of teaching singing is typically
Fig 1a. A model performance is presented by the
based on a master–apprentice model, in which the
teacher, which is followed by the student’s attempt to
teacher gives instructions and feedback on the per-
imitate the model as accurately as possible. After the
formance of the student. This feedback is often given
imitation, the teacher gives feedback on the student’s
either on acoustic quality or a physiological aspect of
performance. The student attempts to improve his/her
singing performance, such as posture or use of the
performance on the basis of the teacher’s feedback.
vocal apparatus. One of the common ways to do this is
The effectiveness of this schema for instruction de-
by providing the student with feedback using imagery,
pends on the student’s ‘knowledge of results’ (KR),
for example: ‘sing as if through the top of your head’.
which is knowledge of the intended outcome of the
However, imagery remains susceptible to ambiguous
sung response (e.g. how much the student knows about
interpretation. Additionally, the time lag between the
using the vocal apparatus to produce the requisite
student’s performance and the teacher’s feedback also
pitch). A drawback in this schematic is that the critical
contributes to the problem. Feedback from the teacher
learning period (CP) is divided over two places in
is hereby dissociated from the online proprioceptive
time. Learning occurs during the student’s response
and auditory sensations accompanying voice produc-
and during the teacher’s feedback. Welch has pointed
tion (Welch 1985).
out that the learning process should be optimal if these
critical periods are reduced, particularly the time be-
tween receiving feedback and applying it. Real-time
Accepted: 14 April 2006
visual feedback (VFB) offers a solution by presenting
Correspondence: David Hoppe, Music, Mind, and Machine Group,
NICI, Radboud University Nijmegen, PO Box 9104 6500, HE the student with feedback on his/her performance
Nijmegen, The Netherlands. E-mail: hoppe.david@gmail.com during the response (see Fig 1b). Aside from reducing

308 & 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd Journal of Computer Assisted Learning 22, pp308–316
Computer-assisted singing training 309

(a) and made use of a peak-picker, originally designed for

Pitch Vocal Feedback Vocal cochlear implants to make an accurate estimate of
model response 1 (KR) response 2
fundamental frequency (F0) in real time. To provide
CP CP
the singer with a musical reference, the system had
(b) libraries of note-sets in ones, threes, or fives. From
Pitch Vocal Vocal
model response 1 response 2
these pre-set libraries, target notes were randomly
picked and presented as a single sound with a fixed
timbre through the audio speaker of the computer. The
program was subdivided into two phases: an assess-
Feedback Feedback
(KR) (KR) ment phase and a development phase. In the assess-
ment phase, a simple note was played to which the
Fig 1 Schema theory of the vocal learning process (adapted subject had to respond by singing into a microphone.
from Welch 1985). (a) Ongoing learning process in classical The mean F0 of the whole sung response was then
singing teaching; (b) interaction of real-time VFB on the compared with the target note. In the development
learning process. KR, knowledge of result; CP, critical learning
phase, a real-time trace of the F0 contour was plotted
period.
against time. Target notes could optionally be placed
on the screen to guide the sung response.
The system was experimentally tested on 32 pri-
CPs to be close to responses, presenting KR during the mary school children aged 7 years (Welch et al. 1989).
vocal response also offers the possibility of im- In a pre-test, the participants had to sing eight ran-
mediately modifying the response and simultaneously domly ordered, relatively low pitches. The initial
observing its effect. pitching abilities of the subjects were recorded using
In fact, the effectiveness of self-monitoring using the SINGAD assessment program. Next, a training
real-time VFB for a better control of physical move- procedure took place in which three matched groups
ments has been demonstrated in many fields. It has were assessed. Group A used the SINGAD system and
been applied for pedagogical use in different domains, received help from a teacher, group B used the SINGAD
such as pronunciation training (Arends & Povel 1991; system without teacher intervention and group C only
Neri et al. 2002) and second-language acquisition received input from conventional singing teaching.
(Dowd et al. 1998; Hirata 2004). Finding a way to The training intervention was followed by a post-test
present feedback in an immediate, understandable, and in which the students again sung the same eight pit-
non-ambiguous way has been the primary drive to ches as in the pre-test. The responses of the partici-
develop real-time VFB technology to assist in singing pants were recorded using the SINGAD assessment
teaching. In this paper, we aim at clarifying the de- program. A significant improvement in pitching abil-
velopments of the VFB systems for singing lessons. ity was found for groups A and B, but not for group C.
Four real-time VFB computer tools for singing in- Especially, group A showed a larger effect. This dif-
struction will be described, including the experimental ference between groups A and B can be explained by
research carried out to evaluate the usefulness of these the role of the teacher in the development phase. The
systems. teacher was instructed to monitor the students’ learn-
ing progress and select new tasks accordingly by
switching to an alternative development program at
Singing assessment and development (SINGAD)
appropriate times. Participants from group B, who did
The pioneer project incorporating real-time VFB into not receive teacher’s assistance, were free to use any
educational software for singing was SINGAD (Howard development program whenever they liked. This
& Welch 1989). The aim of the project was to develop contrast between groups A and B suggests that pro-
a pedagogical tool for the assessment and develop- viding the proper information at the appropriate phase
ment of the singing voices of UK Primary School of the learning process may be crucial for optimal
children using real-time VFB technology. The soft- learning. Nevertheless, a positive effect of real-time
ware package was designed for a BBC microcomputer VFB on the development of pitching ability of primary

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
310 D. Hoppe et al.

three tonal categories (major, minor, and pentatonic).

Each category contained a fixed set of sequences using
SINGAD
permutations of three or five notes within the specific
tonality. By selecting a category, the system would
Atari computer randomly order these sequences per trial. Figure 3
shows an example of the ATARI SINGAD assessment
display. Although not formally evaluated, the ATARI
system was used successfully in schools as a singing
MIDI MIDI
Pitch IN OUT assessment tool (G.F. Welch, personal communica-
MIDI
to
MIDI
synthesizer tion).

Fig 2 Hardware block diagram of the ATARI computer version

of SINGAD. (The figure has been adapted from Howard et al. ALBERT
2003).
ALBERT: Acoustic and Laryngeal Biofeedback En-
hancement in Real Time is another real-time VFB
computer tool, which aimed at providing the devel-
school children, both with and without the guidance of oping singer with a tool to enhance vocal production
a singing teacher when compared with the traditional (Rossiter & Howard 1996). In addition to providing
singing lesson, was a significant finding of this study. VFB on the acoustic output, ALBERT included mon-
In the early 1990s, with the development of more itoring of laryngeal action. The system provided a
powerful computers, the improved version of the greater diversity of VFB displays and parameters: F0,
SINGAD system was introduced (Howard & Welch CQ (larynx closed quotient), spectral ratio, SPL (am-
1993). This second version of the system was made for plitude), shimmer, and jitter. Importantly, the VFB
the ATARI range of machines because it came stan- should not present very detailed information that is at
dard with MIDI hardware. In comparison with the too low a level to be useful to the developing singer.
BBC version, three major changes were made. First, For example: providing VFB on the physiological
the spectral output of the computer was improved. action of the vocal apparatus may distract beginner
Unlike the BBC microcomputer, which had only a singing students from the goal of the singing task:
single sound with a fixed timbre, many different namely, the desired acoustic response. Instead, the
sounds could be selected from an external MIDI information presented by the VFB should be task re-
synthesizer using the new system (see Fig 2). This levant. To meet this criterion, the system was built to
way, the sound of instruments of common use in the enter user-defined parameters by combining any of the
singing studio could serve as stimuli, such as the primary parameters. Through a visualization-control
baroque flute, piano, or guitar. Second, the ATARI had window, the user could configure the displayed in-
a larger memory capacity than the BBC micro- formation. Parameters could either be visualized in
computer, allowing for the storage of complete F0 one, two, or three dimensions. For pedagogical pur-
contours. Owing to the lack of memory capacity, the poses, the labelling attributes of the axes could also be
BBC microcomputer could only store the mean F0, altered or even hidden from the screen. Additionally,
which was taken from the complete sung response. the size and colour of the display could be changed.
Averaging the F0 of the complete sung response ap- ALBERT was tested on two participants (Rossiter
peared to be problematic as it is often subject to pre- et al. 1996). The aim of the study was to identify the
and post-onset noise or random vocalization. With this quality of voice production during VFB implementa-
system, the data could be analysed by strictly mea- tion, measured as the pattern of change during training
suring the performer’s response, thereby omitting any lessons. The parameters used were CQ and spectral
irrelevant data from the analysis. The last problem in ratio (both parameters positively correlate with energy
the former system was that the stimuli were not mu- in the singer’s formant). Each participant had six
sically relevant, as the target notes were randomly lessons in which both speech and singing were as-
picked. This was improved by arranging the stimuli in sessed. One participant received only classical voice

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
Computer-assisted singing training 311

Fig 3 Example of ATARI SINGAD assessment screen for the fourth trial of the set illustrated at the foot of the figure. (The figure has
been adapted from Howard et al. 2003).

training, while the other participant received classical meter condition, as subsequent use actually decreased
voice training and VFB with the following three response accuracy.
parameter conditions: (1) CQ; (2) spectral ratio; and
(3) a combination of CQ and spectral ratio. Each
condition lasted for two lessons successively. For each SING & SEE
condition, the effect of the presented parameter on the
level of both parameters (CQ, spectral ratio) in re- Recently, the SING & SEE project was introduced at the
sponses was measured. An increase in the levels of Conference of Interdisciplinary Musicology (Call-
these parameters indicates the presence of a pro- aghan et al. 2004). The project aimed at developing
nounced singer’s formant. In the single parameter new VFB technology for the singing studio. The main
conditions (1 and 2), the level of the parameter pre- features of research were the investigation of acoustic
sented as VFB increased, while the level of the other analysis techniques, methods of displaying VFB in a
parameter, which was not presented as VFB, remained meaningful way and the pedagogical approaches for
unchanged. For the third condition in which both CQ implementing VFB technology into practice. Three
and spectral ratio were presented, the measured level parameters were distinguished as relevant for usage in
of both these parameters increased. Clearly, the VFB the singing studio: pitch (F0 against time), vowel
was contributing to an improvement in voice quality. identity (R1, R2), and timbre (spectrogram). The
Interestingly, the changes in response accuracy major difference from former studies was that not only
showed the greatest effect when participants used the quantitative, but also qualitative data were of interest
VFB for the first time (e.g. lesson 1, 3, and 5). Sub- in this project.
sequent use only yielded a small effect. This effect Four singing teachers and 21 of their students
was even more pronounced for the combined para- (seven beginning, 11 progressed, three advanced)

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
312 D. Hoppe et al.

Fig 4 (a) Display screen A. SING & SEE pitch grid, F0 against time. Target notes are indicated as dark bars. The pitch trace is indicated
by the line; (b) display screen B. Target notes are indicated as light grey keys. Correct responses turn the target key into bright red.
(The figure has adapted from Wilson et al. 2005).

participated in the study. The study followed a within- students may not fully understand what they see. Also,
subject pre–post-test design in which all students re- teachers commented that the VFB is probably most
ceived the VFB during a 2-week intervention period. efficient during specific stages in development.
Because of different skill levels, the parameters used As for the students, 90% were positive about the use
during the intervention differed between students. of VFB in the singing lessons. They felt that the VFB
Moreover, the parameters used during the intervention improved their understanding of the desired outcome
also differed within groups because the teachers had of the target model that they were required to imitate,
their own teaching style. Each participant was re- as the feedback was immediate and unambiguous.
corded during the baseline, intervention, and follow- Also, the use of the spectrogram received positive
up periods. Both acoustic and perceptual measures feedback. The vowel quadrilateral, on the other hand,
were taken to quantify the changes in singing perfor- met with the same criticism as from the teachers.
mance. After the follow-up period, all participants Recently, another study was conducted to examine a
were interviewed on how the VFB had been integrated novel form of VFB and test its impact on singing using
into the singing lessons, whether the software used had the pitch display from the SING & SEE software
particular strengths or weaknesses and how to make (Wilson et al. 2005). Besides testing the VFB’s impact
possible improvements. on learning, this study also addressed the question of
The paper reported an analysis of the interview data. whether the amount of information presented by the
All teachers were positive about the use of the pitch visual display has an effect on singing performance.
feedback. The spectrogram appeared to be more useful Fifty-six participants took part in this study, with skill
working with more experienced students than with level ranging from non-singer to trained singers. They
beginners. More experienced and advanced students were assigned to three groups. Groups A and B re-
used the spectrogram for training timbre, dynamics, ceived VFB, while group C served as a control. Within
and the singer’s formant, whereas beginners only used each group participants, were divided into three sub-
it for note onset and offsets. The vowel quadrilateral groups according to their skill level. The distribution
appeared not to be appropriate for a high-pitched of skill level between the three groups was kept equal
voice, being only suitable for an adult male voice. as far as was possible.
Importantly, the teachers suggested that the VFB Group A trained with the pitch display from the
displays should be more musically relevant. They also SING & SEE program. This form of visual feedback
warned against the use of VFB without supervision, as presented the participants with a pitch trace, giving the

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
Computer-assisted singing training 313

Fig 5 WinSINGAD. (a) Panel showing a spectogram (upper), spectrum for the final vowel (centre) and ratio (lower); (b) dual-panel
screenshot. Example of a vocal tract area display for a sung/u:/sound. Both the singer and the tutor can monitor posture at the same
time (camera shot). ((a) has been adapted from Welch et al. 2004; (b) has been adapted from Howard et al. 2004).

user constant feedback on pitch history and target lo- VOXed: WinSINGAD
cation (Fig 4a). Group B trained with a keyboard
display (Fig 4b). Only when a note was correctly sung Besides the SING & SEE project, the VOXed project
did the matching key light up. This form of VFB does (Welch et al. 2004) was presented during the same
not provide pitch history. It merely gives the user a Conference of Interdisciplinary Musicology. The
specific right/wrong feedback. Group C was presented project also incorporated real-time VFB for singing
with the same display as group B, but without the pitch education, suggesting that there are growing interests
response feedback. In both a pre-test and a post-test, in computer assistance in musical education. While
all participants were assessed on five test patterns of SING & SEE places emphasis on maximizing VFB
five interval sequences in upward semitone incre- technology itself, VOXed aimed at maximizing the
ments. Before the test recordings began, the pitch collaboration between different fields. Psychologists,
range of the test patterns was adjusted to the partici- voice scientists, singing teachers, and singing students
pant’s pitch range. joined to form an interdisciplinary research team
In addition to the effects of VFB on the learning searching for a better insight on the impact of VFB on
process, the effect of VFB during the training period the learning experience. Importantly, VOXed sought
was analysed in this study. Comparisons were made to work with participants as active agents rather than
between pre- and post-test as well as between pre-test just passive recipients. The goal of the project was to
and intervention. Surprisingly, participants in the ex- investigate possible useful forms of VFB with the use
perimental groups tended to worsen their performance of commercially available visual feedback software.
when they received the VFB. Nevertheless, the com- The windows-compatible software tool WinSINGAD
parison between the pre- and the post-test revealed was developed, which is the successor of the early
that the experimental groups significantly improved SINGAD systems (Howard et al. 2003, 2004). Because
their performance after the intervention compared changes in vocal output over time are of primary in-
with the control group (group C). Comparisons be- terest to singers, the majority of displays are plotted
tween the groups using VFB also indicated that the against time. The parameters available were as fol-
impact of the pitch display was remarkably larger for lows: input waveform; F0 against time; short-term
the non-trained singers, whereas the impact of the spectrum; narrow-band spectrogram; spectral ratio
keyboard display was of greater use to the trained against time; vocal tract (VT) area; and mean/min VT
singers. Apparently, different skill levels require dif- area against time. Any parameter-window could be
ferent forms of VFB. displayed on the screen in combination with another.

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
314 D. Hoppe et al.

Also, a side-view Web cam could be selected for di- any up to three of the six primary parameters could be
rect postural feedback (see Fig 5). combined to form new parameters. In this way, the
Two teachers agreed on using the VFB in their VFB could be altered so that it was most appropriate to
teaching lessons. Each teacher had four students par- the context of the task, be it singing training, pro-
ticipating (eight in total), two of whom were taught nunciation training, or voice therapy. SING & SEE also
with the VFB and the other two serving as controls. focused specifically on singer-related parameters: pitch
The teachers were given full control over the way to (F0 against time), vowel identity (R1, R2), and timbre
use these parameters during their teaching lessons. (spectrogram). The last project we reviewed, VOXed,
Recordings were made during two sample lessons. In introduced the successor of the early SINGAD systems,
the first lesson, VFB was not implemented and served WinSINGAD, and included the widest range of singer-
as a baseline. The second lesson fully established the oriented parameters: input waveform; F0 against time;
use of the VOXed technology. Of each lesson, real- short-term spectrum; narrow-band spectrogram; spec-
time observations as well as video recordings were tral ratio against time; VT area; and mean/min VT
made. Outside the lessons, both teachers and students area against time. A side-view Web cam could be se-
gave semi-structured interviews. Thus far, this study lected for direct information about the student’s pos-
has reported on quantitative analyses of sample lesson ture. In general, VFB features have become more
observational data, supplemented by qualitative com- multifaceted over time. This has allowed the programs
mentary (Welch et al. 2005). to be accessible and useful to a wide range of
The results showed an overall positive appreciation singers. Although SINGAD was designed only for the
of the use of VFB in the singing studio. Teachers were child’s voice development, ALBERT aimed at a broader
so keen on using the VFB that they even started to use application that was not restricted to just singing
it with other students as well, in one case, even with training. Both SING & SEE and VOXed were specifi-
those who were initially assigned as controls. Con- cally designed for singers of all ages and skill levels.
cerning the implementation of the technology in the We can conclude that the usability of these systems
singing lessons, the teachers reported that they found has improved over time by the addition of new
the system to be user-friendly and non-obstructive to functions.
the normal course of teaching events. Of all displays, Interestingly, the effectiveness of VFB on the
the spectrogram was most fully exploited by both learning process seems to depend on the amount of the
teachers. Also, the facility to play back the sung re- student’s musical experience (as initially postulated by
sponse of the student had great advantages, as both Welch 1985). In a controlled experiment, Wilson et al.
teacher and student could now attend to the display (2005) investigated the effect of the amount of in-
and discuss its contents. Furthermore, the time-course formation presented by the visual display on the
analysis of the lessons confirmed the different teach- singing performance of both trained and non-trained
ing strategies between the two teachers. The amount of singers. The keyboard display, which presented spe-
time that each teacher made reference to the VFB also cific right/wrong feedback, was of greater use to the
differed. Overall, the introduction of VFB into the trained singers. The pitch display, which presented
singing lesson was met with great enthusiasm. more detailed and contextualized information, was
found to have a remarkably larger impact on singing in
tune for the non-trained singers. However, Callaghan
Discussion
et al. (2004) found that the greatest use was made of
Four real-time VFB computer tools for professional the spectrogram, especially with progressed students:
voice development were reviewed in this paper. Over the spectrogram appears to be a hard visual display to
time, the original designs have been developed to interpret because it presents much information at once.
provide the user with more information. SINGAD made Although these two studies showed that the amount of
use of a single parameter: fundamental frequency (F0). information displayed in the VFB interacts with the
ALBERT maximally exploited memory capacity offered skill level of the student, exactly how these two factors
by the rapid development of computer hardware in the relate is not yet clear, and therefore requires further
mid 1990s. A unique function of this system was that investigation.

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
Computer-assisted singing training 315

In designing ALBERT, careful consideration was operative group than when working on one’s own
made in constraining the amount of information that (Latane et al. 1979). The observed results of this study
was shown as VFB. To make sure that the VFB might therefore be influenced by the pattern of social
showed information relevant to the task, users were interaction. In contrast, for example, Wilson et al.
given the possibility to enter new parameters by (2005) conducted a well-controlled experiment with
combining any of the six primary parameters. Rossiter many participants. Based on the large number of ob-
et al. (1996) investigated the impact of this feature on servations, they have shown that VFB does, in fact,
the singing process during training lessons. Single and significantly enhance learning to sing in tune. How-
combined parameter display conditions were eval- ever, the number of such controlled experiments is
uated. The combined parameter condition showed the rather small. Many of the valuable qualitative findings
greatest improvement in performance accuracy during should therefore still be evaluated quantitatively.
initial use. However, during subsequent use, the same Another finding from the SINGAD experiment was that
condition yielded a negative effect. Also, greater ap- improvement takes place in pitching ability when VFB
preciation of the feedback was found during initial use is used even without supervision. The program could
of all feedback conditions, in contrast to a general therefore serve as a replacement of singing teachers
decline in performance accuracy during subsequent instead of just being used as a helpful tool to singing
use. Apparently, quality of performance does not al- teachers. However, in observational studies, greater
ways increase during the use of VFB. In fact, Wilson improvement was found when teachers assisted with
et al. (2005) reported that participants tend to worsen VFB. Furthermore, Callaghan et al. (2004) reported that
their performance during the use of VFB, although an their participant teachers were actually against the use of
increase in singing accuracy from pre-test to post-test VFB without supervision, as students may not under-
was observed. These findings are counter intuitive, as stand fully what they see. In their study, the spectrogram
one would expect a linear improvement in performance appeared to work very well and was very informative,
accuracy during the use of VFB. Thus, the actual im- but only with the teacher’s assistance. This real-life
pact of VFB on the singing process and its relation to application in the singing studio suggests that VFB may
the learning process still have to be understood. be more effective when it is properly understood, and
A difficulty in evaluating educational tools is often for this, teacher’s assistance is helpful.
to balance the fundamental research and its application As in other voice development research domains,
to real situations. In most studies that we reviewed, such as pronunciation training (Neri et al. 2002) and
preserving natural singing class conditions seems to second-language acquisition (Dowd et al. 1998; Hirata
overrule controlling experimental factors. For ex- 2004), we addressed several studies showing that VFB
ample, in their study, Welch et al. (1989) trained the helps in learning to sing. It is noteworthy that VFB
experimental groups to sing single notes, whereas serves well as a tool for assistance, rather than a re-
controls had to sing songs, just as in a common pri- placement of the singing teacher. Accordingly, VFB
mary school singing lesson. As the pre- and post-as- technology has been met with great enthusiasm from
sessment procedures used a task similar to the professional singing teachers. For future research,
experimental training procedure, the experimental further quantitative research on the detailed effect of
groups had a clear advantage over the controls during VFB on performance accuracy and on the learning
the final assessment. It might be the case that the process is necessary, as well as a closer investigation
nature of the training task can explain the observed of its interactions with the type of VFB and the
results of this study. Another point is that the pattern amount of information it presents, the skill level of the
of social interaction differed between the experimental user and the presence/absence of a teacher.
groups and the controls. While the experimental Of valuable insight to the field might be the research
groups worked in pairs or threes, the controls had to on feedback and motor skill learning with respect to
sing together as a group. In such group-singing ac- ‘focus of attention’. The effect of the learner’s ‘focus of
tivity, it is very likely that a phenomenon such as attention’ on the learning process was reviewed by Wulf
‘social loafing’ might occur: the tendency to exert less and Prinz (2001). They showed that an internal focus of
effort on a task when working as a part of a co- attention, which is directed to ‘one’s own movements’,

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd
316 D. Hoppe et al.

appears to be less beneficial to the learning process than Howard D.M. & Welch G.F. (1993) Visual displays for the
an external focus of attention, which is directed to ‘the assessment of vocal pitch matching development.
effects of one’s movements’. Accordingly, VFB on Applied Acoustics 39, 235–252.
singing performance that is directed to one’s own Howard D.M., Welch G.F., Brereton J. & Himonides E.
movements (e.g. the vocal tract) may be less effective (2003) Towards a novel real-time visual display for
singing training, http://www.sonustech.com/voxed.
than VFB on the acoustical output (e.g. real-time
Howard D.M., Welch G.F., Brereton J., Himonides E., De-
spectral information). Indeed, taking the phenomena of
Costa M., Williams J. & Howard A.W. (2004) WinSin-
attentional focus into consideration accounts for some
gad: a real-time display for the singing studio.
previous findings, such as the interview data from Logopedics Phoniatrics Vocology 29, 135–144.
Welch et al. (2005), where singing teachers preferred Latane B., Williams K. & Harkins S. (1979) Many hands
the spectrogram to all other feedback options. Further- make light the work: the cause and consequences of so-
more, the option for the ALBERT user to define new cial loafing. Journal of Personality and Social Psychol-
parameters that are most relevant to the task can also be ogy 37, 822–832.
interpreted in terms of an ‘external focus of attention’, Neri A., Cucchiarini C., Strik H. & Boves L. (2002) The
as the goal of a singing task is to reach ‘a desired effect pedagogy–technology interface in computer assisted
from the movement of the vocal apparatus (e.g. the pronunciation training. Computer Assisted Language
desired acoustic output)’. A further understanding of Learning 15, 441–467.
Rossiter D. & Howard D.M. (1996) ALBERT: real-time
internal and external focus of attention in relation to
visual feedback computer tool for professional vocal
real-time VFB learning would therefore enrich the field.
development. Journal of Voice 10, 321–336.
Rossiter D., Howard D.M. & DeCosta M. (1996) Voice
Acknowledgements development under training with and without the influ-
ence of real-time visually presented biofeedback.
This research was supported by the Technology Acoustical Society of America 99, 3253–3256.
Foundation STW, applied science division of NWO Welch G.F. (1985) A schema theory of how children learn to
and the technology program of the Ministry of Eco- sing in tune. Psychology of Music 13, 3–18.
nomic Affairs (NNN6301). We would like to thank Welch G.F., Rush C. & Howard D.M. (1989) Real-
Alex Brandmeyer for English editing. time visual feedback in the development of vocal
pitch accuracy in singing. Psychology of Music 17,
146–157.
References
Welch G.F., Himonides E., Howard D.M. & Brereton J.
Arends N. & Povel D.J. (1991) An evaluation of the visual (2004) VOXed: Technology as a meaningful teaching aid
speech apparatus. Speech communication 10, 405–414. in the singing studio. Proceedings of the conference on
Callaghan J., Thorpe W. & Van Doorn J. (2004) The science interdisciplinary musicology (CIM04), Graz, Australia,
of singing and seeing. Proceedings of the Conference on April 15–18.
Interdisciplinary Musicology (CIM04), Graz, Australia, Welch G.F., Howard D.M., Himonides E. & Brereton J.
April 15–18. (2005) Real-time feedback in the singing studio: an in-
Dowd A., Smith J. & Wolfe J. (1998) Learning to pronounce novatory action-research project using new voice tech-
vowel sounds in a foreign language using acoustic mea- nology. Music Education Research 7, 225–249.
surements of the vocal tract as feedback in real-time. Wilson P.H., Thorpe C.W. & Callaghan J. (2005) Looking at
Language and Speech 41, 1–20. singing: does real-time visual feedback improve the way
Hirata Y. (2004) Computer assisted pronunciation training for native we learn to sing? Second APSCOM Conference: Asia-
English speakers learning Japanese pitch and duration contrasts. Pacific Society for the Cognitive Sciences of Music, South
Computer Assisted Language Learning 17, 357–376. Korea, Seoul, August 4–6.
Howard D.M. & Welch G.F. (1989) Microcomputer-based Wulf G. & Prinz W. (2001) Directing attentionto movements
singing ability assessment and development. Applied effects enhances learning: a review. Psychonomic Bul-
Acoustics 27, 89–102. letin and Review 8, 648–660.

& 2006 The Authors. Journal compilation & 2006 Blackwell Publishing Ltd