Content uploaded by Robert H Jack
All content in this area was uploaded by Robert H Jack on Apr 10, 2018
Content may be subject to copyright.
Democratising DMIs: the relationship of expertise and
Robert H. Jack, Jacob Harrison, Fabio Morreale, Andrew McPherson
Centre for Digital Music
Queen Mary University of London
An oft-cited aspiration of digital musical instrument (DMI)
design is to create instruments, in the words of Wessel and
Wright, with a ‘low entry fee and no ceiling on virtuosity’.
This is a diﬃcult task to achieve: many new instruments
are aimed at either the expert or amateur musician, with
few instruments catering for both. There is often a balance
between learning curve and the nuance of musical control
in DMIs. In this paper we present a study conducted with
non-musicians and guitarists playing guitar-derivative DMIs
with variable levels of control intimacy: how the richness
and nuance of a performer’s movement translates into the
musical output of an instrument. Findings suggest a signif-
icant diﬀerence in preference for levels of control intimacy
between the guitarists and the non-musicians. In particular,
the guitarists unanimously preferred the richest of the two
settings whereas the non-musicians generally preferred the
setting with lower richness. This diﬀerence is notable be-
cause it is often taken as a given that increasing richness is
a way to make instruments more enjoyable to play, however,
this result only seems to be true for expert players.
expertise, learning, control intimacy, richness, sensorimotor
•Applied computing →Sound and music computing;
Within NIME and related research ﬁelds, much discussion
has centred around richness of control, the level of detail
of control that a performer has over an instrument. It has
been proposed for many years that richer mappings between
control input and sound output are better , and much
design eﬀort has gone into this idea. Yet in practice, the
design of digital musical instruments (DMIs) involves bal-
ancing two factors that can often seem at odds with one an-
other: the steepness of the learning curve that a performer
has to climb to make music with an instrument, and con-
trol intimacy, how the richness and nuance of a performer’s
movement translates into musical output.
Licensed under a Creative Commons Attribution
4.0 International License (CC BY 4.0). Copyright
remains with the author(s).
NIME’18, June 3-6, 2018, Blacksburg, Virginia, USA.
In this paper we aim to interrogate the idea that ‘a richer
instrument is a better instrument’. We designed a study in-
volving two diﬀerent groups of players (guitarists and non-
musicians), which compared two levels of richness (audio-
rate coupling between strings and sound model, and note-
based triggers) within the same guitar-derivative DMI. We
found that although the experienced musicians unanimously
preferred the richer instrument, this appears to be only true
for this group. The non-musicians were more ambiguous in
their preference for the instruments, but perhaps surpris-
ingly, tended towards the less rich instrument overall. This
paper aims to clarify the processes that led to this prefer-
ence by analysing interviews conducted with the performers
and their gesture language with diﬀerent levels of richness.
2. RELATED WORK
Theories of sensorimotor skill acquisition generally agree
that the process passes through a number of qualitatively
diﬀerent stages as the learner progresses from novice to
expert [2, 7]. The stage-based approach is assumed to
hold across diﬀerent skill development domains  includ-
ing learning musical instruments . A classic model of
the stage-based approach to skill acquisition is by Fitts 
which consists of three stages:
•cognitive: the performer has a task broken down into
small components that are not related to the whole.
Performance characterised by high error, high vari-
ability and a detachment of the individual components
from the whole task.
•associative: follows an extended period of deliberate
practice. The performer can associate actions with
successful results, makes fewer errors, becomes more
aware of her errors through a (partial) understanding
of the whole task.
•autonomous: reached through further deliberate prac-
tice. The performer can carry out the components of
the skilled action at a faster pace and without con-
scious attention and can focus on higher-level aspects
of the task.
Learning an instrument involves internalising how action
translates to sound, which is initially acquired by explor-
ing and manipulating the instrument with somewhat ar-
bitrary actions that lead to unexpected results , what
Wessel called the ‘babbling’ stage due to its similarities to
the manner in which young children learn to form words
. A process of experimentation, exploration, repetition
and association, builds an internal model that captures the
relationships of the actions the instrument aﬀords and the
resultant sound . Once the musician progresses to a level
of expertise they do not need to focus attention on the in-
dividual operations of manipulating an instrument, instead
focusing on higher-level musical intentions: the instrument
becomes a ‘natural’ extension of the musician’s body and
no longer an obstacle to an embodied interaction with the
Traditional musical instruments are often given as exam-
ples of tools where the relationship of gesture and sound
is both intuitive and complex: Dobrian and Koppelman
 emphasise the importance of relating DMI design to
its acoustic ancestry for this reason. Unlike acoustic in-
struments, DMIs can boost a performer’s ability to achieve
musically complex results by partially shifting the respon-
sibility for sound production away from their sensorimotor
ability and to the digital system. How intuition, complex-
ity and the development of expertise are balanced in DMI
design has been the focus of much research in this ﬁeld.
2.1 Instrument efﬁciency
Eﬃciency in regards to musical instruments can be consid-
ered as the instrument’s ability to transfer input gestures to
musical sound . Jorda posits that the kalimba, at least
in the ﬁrst stages of learning, is a more eﬃcient instrument
than the piano. Whereas a piano has many notes a kalimba
has few and they are all the ‘right notes’; its form intuitively
encourages the performer to play with their thumbs; once
the kalimba is held in both hands it is clear which thumb
should be used to control which notes. So instrument eﬃ-
ciency depends on the relationship between the complexity
of the input gestures and the complexity of the resultant
sound output, and the ease with which a performer can get
from one to the other. Jorda illustrates the balance between
challenge, frustration and boredom by comparing these in-
struments and suggests that new instruments should adhere
to Wessel and Wright’s ‘low entry fee with no ceiling on vir-
Pardue  states ‘it is in learning where DMIs can have
an inherent advantage over traditional instruments’, propos-
ing the term complexity management to describe the notion
that instrument eﬃciency can be progressively managed
over time in order to maintain a rewarding learning expe-
rience at all levels of expertise. By guiding the novice user
towards less complex musical output, certain techniques can
be isolated for technical practice. This approach could also
provide more immediate access to less formal musical prac-
tice such as improvising or ‘jamming’ with other musicians,
activities which McPherson and McCormick  show help
with the level of cognitive engagement during musical prac-
2.2 Designing affordances
When designing DMIs the process is often conceptualised
as the creation of aﬀordances and constraints towards music
making . In a recent paper discussing embodied control
in HCI, Tuuri et al.  distinguish between push eﬀects
which force or guide the user to particular choices, versus
pull eﬀects which relate to an ease of conceiving how ac-
tion relates to a particular output, as characteristics of an
instrument’s aﬀordance structure. Through these eﬀects a
performer is provided with control of an instrument, while
equally the instrument enforces control on the performer.
Jack et al.  propose a model of DMIs that is based
on projection from a high-dimensional space of body move-
ment to a reduced space of sonic and kinematic features
and behaviours. In this model the instrument represents
a bottleneck in the ﬂow of gestural information, which is
positioned through design choices. In this study, we are
investigating the inﬂuence of diﬀerent widths of bottleneck
on a performer’s experience – each of the instruments we
created allow diﬀerent levels of detail to be carried into the
sound output, and we are interested in how this folds back
on the gestures that the performers use.
2.3 Intimate control and expression
Control intimacy, as ﬁrst introduced by Moore , can
be described as a performer’s perceived match of the be-
haviour of an instrument and their psychophysiological ca-
pabilities when controlling the instrument. Intimate instru-
ments (such as the voice, violin, sitar, ﬂute) allow the micro-
gestural movements of the performer to create a wide range
of aﬀective variation in the control of musical sound. Moore
identiﬁes MIDI as suﬀering from a deﬁcit of intimate control
in its discretisation of musical performance into a sequence
of note-on, note-oﬀ with velocity. Wessel and Wright ex-
pand on Moore’s notion of control intimacy and its relation-
ship to virtuosity, stating that ‘a high degree of control in-
timacy can be attained with compelling control metaphors,
reactive low latency variance systems, and proper treatment
of gestures that are continuous functions of time’ [p. 2].
They propose that high control intimacy can encourage per-
formers to continually develop their skill on an instrument
and their personal style: low intimacy implies that the com-
munication between a performer and a device is poor and
there is a barrier to expressive performance, and hence en-
In this study, we aimed to ask ‘does a richer instrument nec-
essarily mean a better instrument?’, and explore how prior
experience and expertise relates to this. Speciﬁcally, we
aimed to investigate the eﬀects of modulating the richness
of a plucked-string guitar-based instrument on the experi-
ence and gestural language of guitarists and non-musicians.
3.1 Instrument design
We designed two guitar derivative DMIs with diﬀerent form
factors but identical sensor topologies, based on real guitar
strings with piezo sensors. One follows the shape and form
of a guitar and is held with a strap, with the right hand
resting over the strings and left hand holding the neck. The
‘tabletop’ version was designed to follow design cues from
boutique music hardware such as modular synth controllers.
Figure 1 shows the two instruments.
Figure 1: Two instruments used for this study: gui-
tar form (L) and tabletop form (R)
3.1.1 Physical construction
Both instruments were designed to be played using simi-
lar techniques to a guitar or other strummed string instru-
ments. They feature six buttons mapped to six chords. The
buttons on the guitar form are placed on the neck, around
where the lower frets would be on a traditional guitar neck.
The tabletop version features the buttons on the lower left
corner, with the strings placed at a 45 degree angle, as this
was found to be the most comfortable for strumming with
the right hand. Both instruments featured a method of
switching between the two sensor mappings described be-
The physical construction of the string modules involves
six short lengths of .040 gauge bass guitar string held loosely
over a ‘strummable area’ of about 10 cm. At one end, the
strings are terminated over a block of felt-covered foam,
with six individual bridge-pieces at the other with inte-
grated piezo disc sensors, and held to a low tension using
adjustable zither pins. This provides a strong acoustic sig-
nal when strummed or plucked, similar to the attack of a
plucked string on a guitar. The thickness of the strings
and low tension provides a short decay and fewer resonant
properties than a typical guitar string held to tension.
3.1.2 Sensor mappings
Both variations of the instrument use the Karplus-Strong
algorithm to simulate six virtual strings which, individually
excited using signals from the individual piezo sensors. We
used a Bela board  for the sensing and string modelling
to create a high-performance, low latency embedded instru-
ment in each case. The only diﬀerence between the two
variations is the mapping of the piezo signal to the exci-
tation of the virtual strings. The two mapping structures
used are ‘audio-rate’ excitation and ‘sample triggering’:
Audio-rate excitation: Excitation of a virtual string
model using a real-time audio signal has been implemented
and documented in previous NIME research, including the
Kalichord , BladeAxe and PlateAxe  and Caress in-
struments . Such instruments allow intuitive control
over the resulting sound by varying the way the physical
model is excited (plucking hard or soft, or with diﬀerent ma-
terials). Our instruments follow a similar principle, but use
dampened strings terminated over piezo sensors to drive the
virtual string models. This allows the use of natural strum-
ming and plucking gestures, as well as less traditional ones
such as tapping, scraping or stroking the strings, which have
musically meaningful results in the resulting audio signal.
Sample triggering: The sample triggering version uses
the same synthesis technique but dramatically reduces the
amount of achievable variation in input signal. Rather than
passing the audio signal directly to the virtual string al-
gorithm, a peak detection algorithm is used to trigger a
pre-recorded pluck recording whenever an amplitude peak
is reached. The recorded pluck was taken directly from the
piezo audio signal, so is directly comparable with the audio-
rate version, however the timbre and dynamics remain static
independent of input gesture.
3.2 Study Design
Participants were asked to compare two versions of the
same instrument. We used a combination of improvisa-
tion/exploration and prescribed musical tasks. We also col-
lected qualitative data using questionnaires and structured
interviews at two points during the study and audio and
video recorded the whole session.
We recruited 32 Participants (16 ‘competent’ guitarists
and 16 non-musicians). Participants were asked to self-
identify at the recruitment stage using the following state-
ments: ‘you are comfortable strumming and playing along
to a tune’ (competent guitarists) and ‘you have no or very
little experience playing an instrument’. We asked par-
ticipants to complete the self-report questionnaire section
of the Goldsmiths Musical Sophistication Index (GoldMSI)
test battery .
The full study was designed to investigate several factors,
some of which are not relevant to this paper. This involved
a comparison of the physical form of the instruments (the
tabletop and guitar-shaped forms described in Section 3.1),
as well as sensor topologies (the string modules described
previously, and a touch sensor variation). The results of
the comparison between form and sensor topology are pre-
sented elsewhere . In this paper, we are concerned only
with those instruments which featured the ‘string’ sensor
topology. The results presented here concern the compar-
ison of these instruments in either ‘sample-triggering’ or
3.2.1 Musical tasks
Participants were instructed ﬁrst to improvise and explore
with the instrument in the sample-triggering setting for
seven minutes. They were then given a further seven min-
utes to rehearse and perform an accompaniment to a record-
ing of a folk song. We recorded a piece taken from the folk-
RNN songbook  for this purpose, arranged for ﬁddle and
electric bass. The chord structure of the song used chords I,
IV and V in the key of G. We added coloured stickers to the
buttons to indicate these chords and printed a colour-coded
score for participants to follow while playing. We also pro-
duced a video ﬁle displaying the chord colours and positions
on screen as they appeared in the score, in a similar manner
to the Guitar Hero games. Participants were allowed to use
either or both of these methods to follow the backing track
but were encouraged to use the printed score if they felt
comfortable to do so. The buttons and score are presented
in Figure 2.
Figure 2: L-R: colour-coded paper score, screenshot
of on-screen chord visualiser, colour-coded buttons
For the ﬁnal musical task, we instructed the participants
to switch to the audio-rate variation of the instrument us-
ing a switch on the instrument’s enclosure. They were then
given ten minutes to improvise and explore with the instru-
ment. No further score-following tasks were given.
3.2.2 Questionnaire and structured interviews
After the score following task we conducted structured in-
terviews with the participants asking questions related to
the techniques they used and their overall impression of
the instrument in the sample-triggering variation. Follow-
ing the ﬁnal musical task with the audio-rate variation, an-
other structured interview took place, this time focusing
speciﬁcally on diﬀerences and similarities between the two
settings. We then asked participants to indicate their pref-
erence for either setting using a horizontal on-screen slider
with ‘setting 1’ (sample-triggering) on the left and ‘setting
2’ (audio-rate) on the right. This produced a value from
0-100, with 0 indicating strong preference for setting 1, and
100 indicating strong preference for setting 2.
Our ﬁndings consist of a quantitative measure from the com-
paritive rating of each setting and a pair of thematic anal-
yses: the ﬁrst on the interviews conducted with the partic-
ipants, the second on their gestural language while playing
4.1 Participant data
We recruited 32 participants, 16 in each group. 19 partic-
ipants were male (13 guitarists and 6 non-musicians), and
13 were female (3 guitarists and 10 non-musicians). Par-
ticipant age ranged from 18 to 62 with an average 32 years
old. The average GoldMSI scores for each group were 89
(SD = 11, minimum = 72) for guitarists and 55 (SD =
11, maximum = 70) for non-musicians. The minimum and
maximum show the proximity of the two groups.
Figure 3 shows the comparative rating of the settings by
each group. A paired t-test on the comparative ratings of
the settings from each group found a signiﬁcant diﬀerence
between groups (t = 5.6833, df = 16.731, p <.01). All
16 guitarists rated the audio-rate setting as better, whereas
there was more disagreement in the non-musician category
with 6 rating audio-rate as better and 10 rating sample-
triggering as better. We also tested for an eﬀect of the dif-
ferent global forms (guitar-shaped vs. tabletop) but found
no signiﬁcant eﬀect, meaning that the diﬀerence in rating
was driven by the diﬀerence between the settings.
Figure 3: Median and IQR ratings of setting 1
(sample-triggering represented by 0 on the y-axis)
and setting 2 (audio-rate represented by 100 on the
y-axis) for all 32 participants.
4.3 Participant reasoning
We performed a thematic analysis on the transcripts from
the structured interview with a focus on reasoning in rela-
tion to the following themes: sound, technique, instrument
behaviour, relation to existing instruments or interfaces.
Table 1 presents some sample quotes that are representative
of the reasoning of each group.
The 6 guitarists were quick to critique the sample-triggering
variation at the end of their ﬁrst session, even without the
knowledge that a richer mapping would be introduced later
in the study. Most of the comments focused on the lack
of timbral expression and the ﬂatness of the articulation
on the instrument in comparison to what they were used
to on a traditional guitar. The addition of these capabili-
ties with the audio-rate setting was mentioned by 12 par-
ticipants in this group. The audio-rate setting’s ability to
support existing technique was also a reoccurring theme,
with particular reference to ﬁngerpicking and articulation.
6 members of this group also made reference to the ‘feel’ of
the instrument as more ‘guitarry’ or ‘natural’ in comparison
to sample-triggering: “setting 2 really uses your knowledge
of guitar. Compared to setting 1 where the strings are not
behaving as strings.”
The non-musicians who preferred sample-triggering gen-
erally reported diﬀerences between the two settings related
to ‘clarity’ and ‘power’ in comparison to the more ‘fragile’
or ‘far away’ audio-rate setting. Sample-triggering was de-
scribed by 9 in this group as easier to generate sound with:
comparisons of the force required to create sound from the
instrument were mentioned, with sample-triggering com-
mended for its ability to create a loud sound with little
eﬀort using a diverse set of playing techniques, whereas the
audio-rate setting was referred to as diﬃcult, hard, or re-
quiring too much pressure. There were also 4 non-musicians
(1 who preferred sample-triggering, 3 who preferred audio-
rate) who stated that they noticed very little or no diﬀer-
ence between the two settings and so just went with their
We performed a further thematic analysis of the video footage,
focusing on identifying the diﬀerent sets of gestures each
participant used in their right hand during the musical tasks.
These observations are presented in Table 2. Our inter-
est was in comparing the diversity of gesture usage in each
group to identify correlations with their given preference.
From Table 2 we can see that there was no clear dis-
tinction between the groups in terms of overall diversity of
gestures: both used a similar variety and spread of gestures
although the more ‘guitar-like’ techniques (strumming as if
holding a plectrum, ﬁnger-style) occurred more frequently
with the guitarists.
Our ﬁndings complicate the notion that ‘a richer instrument
is a better instrument’. The guitarists were unanimous in
their preference for the richer setting, which is unsurprising
as the audio-rate setting more accurately translates the ex-
isting techniques of guitar players to musically meaningful
timbral and dynamic eﬀects. What was less expected was
the ambiguity of response amongst non-musicians, and their
tendency towards the less rich instrument overall.
From the structured interviews we can piece together a
picture of why this diﬀerence in opinion might exist: gui-
tarists were able to speak at length of a lack of detail, ﬂat-
tening of nuance and general shortcomings of the triggering
routine. For this group there was a wasted reserve of ges-
tural potential that had no eﬀect on the musical output of
the instrument. Many of the non-musicians however, were
complimentary of the sample-triggering setting for its clar-
ity and strength of sound. This group frequently spoke of
the audio-rate setting as quieter (which was true for soft
playing but if suﬃcient energy was put into the instrument
then it could be louder than the sample-triggering), more
delicate, and harder to produce a satisfactory sound with.
5.1 Efﬁciency and richness
If we return to Jorda’s notion of instrument eﬃciency 
we could say that the sample-triggering setting is more eﬃ-
cient than the audio-rate setting. The complexity of musi-
cal input is matched for both settings: both have reasonable
coverage of the techniques used to play guitar, and the phys-
ical layout and dynamic material behaviour of the strings
do not change with the settings, supporting the same base
of gestures. It is in the musical output complexity that the
settings diﬀer, with the sample-triggering setting project-
ing a rich and nuanced set of input gestures to a reduced
set of musical features in the sound output. The audio-rate
setting retains the spectral relationship between input and
Table 1: Selected responses from the structured interview.
Timbre Amplitude Technique Realism Behaviour
“There was a def-
inite tonal diﬀer-
“with setting 2
you can play
soft, and you
can play hard”
“all the things that I do
worked on setting 2 but
didn’t necessarily work on
“It really uses your knowledge
of guitar. Compared to setting
1 where the strings are not be-
having like strings”
“Setting 2 was responding
to touch much more deli-
“Setting 1 was like
listening to a guitar
in a concert, but set-
ting 2 was more like
listening to some-
thing on my com-
“Setting 1 was
brighter to me”
“You need to put more
pressure with your hands
with setting 2, it’s harder
to generate the sounds
while with setting 1 you
can just touch the strings
and create the sound”
“There wasn’t much diﬀerence
between the two, but with the
ﬁrst setting everything I did
made a diﬀerence, with the
second one everything sounded
more fragile in a way.”
“My tapping didn’t trig-
ger the string rather, what
I was triggering was the
sound of the string itself”
Table 2: Analysis of gesture occurrence for each group and each setting. Column corresponds to number of
participants who used each gesture under each setting. The gesture analysis of the free improvisation with
setting 1 (sample-triggering) and setting 2 (audio-rate).
Setting: 1 2 1 2
Strumming with a single ﬁnger 14 12 9 4
Strumming with hand like holding plectrum 7 6 13 11
Strumming with multiple ﬁngers 5 5 2 2
Rake / ‘ﬂamenco style’ strum 2 1 1 1
Slow plucking with ﬁngers / thumb 16 12 11 1
Fingerpicking (ﬁnger-style) 8 8 13 11
Plucking with hand like holding plectrum 1 3 2 0
Tapping individual strings 10 6 8 3
Tapping multiple strings with ﬂat ﬁnger 6 2 4 3
Pushing down on strings 6 1 2 —
Scratching / swiping strings 2 8 4 3
Tapping bridge pieces 1 2 6 4
Damping strings / attempting to damp strings with palm 6 4 4 2
Strum / pluck / tap at diﬀerent points along the string 6 3 4 4
Plucking while pressing down/muting strings 1 5 4 —
Observably testing triggering threshold 3 N/A 4 N/A
Observably testing dynamic range N/A 2 N/A 3
output and so requires nuanced control of the strings in or-
der to get a nuanced output, whereas the sample-triggering
setting can work with a much lower level of deﬁnition at the
input: any energy above a certain threshold is transformed
into an impulse and the spectral signature of that gesture is
disregarded. The bottleneck  that the instrument repre-
sents to the interaction can be imagined as wider in this case
in comparison to sample-triggering: more of the performers’
gestural language can be projected through the instrument.
A wider bottleneck might inherently mean lower eﬃciency:
the fact that a greater amount of control input complexity
can aﬀect the output means that the performer becomes
responsible for that extra level of control.
This can also be viewed in terms of required eﬀort. Both
settings transfer energy from input to output (the kinetic
energy of the performer to the sound energy of the instru-
ment). Eﬀort relates to the total amount of energy needed
to achieve a result. The sample-triggering setting required
much less physical eﬀort to achieve an equivalent note to the
audio-rate setting. While eﬀortful interaction is proposed
as a valuable thing in DMIs , in the case of non-musicians
here the relative lack of eﬀort was perceived as attractive.
5.2 Gestural behaviour
The similar spread in gestures between both groups, along-
side the stark diﬀerence in opinion in terms of setting pref-
erence, shows how the aﬀordances and constraints of the
instrument were found in a similar manner for both groups,
but meant diﬀerent things to them depending on experience.
The non-musicians were concerned with eﬃciency of sound
production and the easiest way to generate a note. With
the sample-triggering setting they were quickly drawn to the
pull eﬀects of the instrument – the places where they were
enabled in their control . With the audio-rate setting the
instrument had a diﬀerent aﬀordance structure which, for
many in this group, manifested as push eﬀects – they were
forced to use a particular set of gestures requiring more
nuance and eﬀort than they were capable of to achieve a
satisfactory musical result.
5.3 Learning curves
For the experienced guitarists that are used to navigating
input complexity there is no advantage in reducing it, in fact
many of the guitarists negatively reported the lack of output
complexity from the sample-triggering setting. In the case
of the non-musicians there does however seem to be a use in
reducing the output complexity (and hence increasing the
eﬃciency of the instrument). This could be partly due to
the learning curve that each instrument has. Whereas the
guitarists already know a large amount of techniques that
can be used to play the guitar, and so are able to quickly
make sense of the audio-rate setting, non-musicians want a
more direct route to producing musically satisfying sound
with the instrument and so prefer the faster learning curve
of the more eﬃcient sample-triggering. We can consider the
switch between settings as a kind of ‘complexity manage-
ment’  that provides non-musicians with a shortcut to
taking part in a musical activity as a performer.
5.4 Study design
There is a possible eﬀect of the study design on our obser-
vations from the fact that Setting 1 came ﬁrst. Guitarists
complained about the limitations of Setting 1 before hav-
ing knowledge that a richer setting was coming later in the
study, which is a good support for our ﬁndings. With the
non-musicians, some of the eﬀects could be explained by the
fact that they were already familiar with Setting 1 when
they switched to Setting 2. Diﬀerences in musical expertise
(from GoldMSI) within the two groups is another point that
could beneﬁt from further investigation.
In this paper we explored the notion of control intimacy and
how it relates to instrumental expertise and prior experi-
ence. We set up a study to explore the eﬀects of increasing
the ‘richness’ of an instrument for both experienced mu-
sicians and novices. Our results support the notion that
for experienced musicians, richer instruments are preferable
and more suitable for performance. This is possibly due to
the preservation of the full spectrum of gestures that have
a meaningful eﬀect on the musical output. When designing
for non-musicians, however, the role of richness is less clear.
There was a greater spread in overall preference for the two
instrument variations, with a tendency towards the less rich
version. Viewing this variation as the more musically eﬃ-
cient of the two settings, especially in the hands of a non-
musician, our ﬁndings suggest that an instrument which
requires less physical and mental eﬀort from the player can
lead to more enjoyable experiences for beginners even if it
has a more restricted space of possibilities.
In this paper we have only analysed the ‘ﬁrst contact’ a
performer has with an instrument rather than a longitudinal
evolution of performer and instrument. Going back to the
notion of ‘complexity management’ and the value of learn-
ers taking part in more informal, less technique-oriented
musical situations, we might ﬁnd a compelling argument
for introducing similar methods of adjusting the richness
on future instruments to streamline non-musicians into mu-
sical contexts as performers. Future studies on this subject
might therefore incorporate a longitudinal approach and a
sliding scale of control intimacy.
We would like to extend our gratitude to Ailish Underwood
for crafting the guitar body and to the participants that
took part in this study. This work is supported by EPSRC
under grants EP/N005112/1 (Design for Virtuosity) and
EP/G03723X/1 (DTC in Media and Arts Technology).
 Augmenting the iPad: the BladeAxe, author=Michon,
Romain and Smith, Julius O and Wright, Matthew
and Chafe, Chris, booktitle=Proc. NIME, year=2016.
 J. R. Anderson. Acquisition of cognitive skill.
Psychological review, 89(4):369, 1982.
 P. Bennett, N. Ward, S. O’Modhrain, and P. Rebelo.
Damper: a platform for eﬀortful interface
development. In Proc. NIME, 2007.
 C. Dobrian and D. Koppelman. The ‘E’ in NIME:
musical expression with new computer interfaces. In
Proc. NIME, 2006.
 U. Eversheim and O. Bock. Evidence for processing
stages in skill acquisition: a dual-task study. Learning
& Memory, 8(4):183–189, 2001.
 P. M. Fitts and M. I. Posner. Human performance.
 A. Gentile. Skill acquisition: Action, movement, and
neuromotor processes. Movement science, pages
 J. Harrison, R. H. Jack, F. Morreale, and
A. McPherson. When is a guitar not a guitar?
cultural form, input modality and expertise. In Proc.
 B. Hommel. Acquisition and control of voluntary
action. Voluntary action: Brains, minds, and
sociality, pages 34–48, 2003.
 R. H. Jack, T. Stockman, and A. McPherson. Rich
gesture, reduced control: the inﬂuence of constrained
mappings on performance technique. In Proc. MOCO,
 S. Jord`a. Digital instruments and players: part i –
eﬃciency and apprenticeship. In Proc. NIME.
National University of Singapore, 2004.
 P.-J. Maes, M. Leman, C. Palmer, and M. M.
Wanderley. Action-based eﬀects on music perception.
Frontiers in psychology, 4, 2013.
 T. Magnusson. Designing constraints: Composing and
performing with digital musical systems. Computer
Music Journal, 34(4):62–73, 2010.
 A. McPherson and V. Zappi. An environment for
submillisecond-latency audio and sensor processing on
beaglebone black. In Proc. AES, 2015.
 G. E. McPherson and J. McCormick. Motivational
and self-regulated learning components of musical
practice. Bulletin of the Council for Research in
Music Education, pages 98–102, 1999.
 A. Momeni. Caress: An enactive electro-acoustic
percussive instrument for caressing sound. In Proc.
 F. R. Moore. The dysfunctions of MIDI. Computer
Music Journal, 12(1):19–28, 1988.
 D. M¨
ullensiefen, B. Gingras, J. Musil, and L. Stewart.
The musicality of non-musicians: an index for
assessing musical sophistication in the general
population. PloS one, 9(2):e89642, 2014.
 L. Nijs, M. Lesaﬀre, and M. Leman. The musical
instrument as a natural extension of the musician. In
Proc. Interdisciplinary Musicology, 2009.
 L. S. Pardue. Violin Augmentation Techniques for
Learning Assistance. PhD thesis, Queen Mary
University of London, 2017.
 D. Schlessinger and J. O. Smith. The Kalichord: A
physically modeled electro-acoustic plucked string
instrument. In Proc. NIME, 2009.
 B. Sturm, J. F. Santos, and I. Korshunova. Folk music
style modelling by recurrent neural networks with
long short term memory units. In Proc. ISMIR, 2015.
 K. Tuuri, J. Parviainen, and A. Pirhonen. Who
controls who? embodied control within
human–technology choreographies. Interacting with
Computers, pages 1–18, 2017.
 D. Wessel. An enactive approach to computer music
performance. Le Feedback dans la Creation Musical,
pages 93–98, 2006.
 D. Wessel and M. Wright. Problems and Prospects for
Intimate Musical Control of Computers. Computer
Music Journal, 26(3):11–14, 2002.