2.1. Representationalism in music technology

As Barad (Reference Barad2003) observes, a central tenet of both positivist and constructivist worldviews is representationalism: ‘the power of words to mirror preexisting phenomena’. Barad elaborates: ‘The idea that beings exist as individuals with inherent attributes, anterior to their representation, is a metaphysical presupposition that underlies the belief in political, linguistic, and epistemological forms of representationalism. Or, to put the point the other way around, representationalism is the belief in the ontological distinction between representations and that which they purport to represent; in particular, that which is represented is held to be independent of all practices of representing’. In the representationalist worldview, words and symbols enjoy special status by pointing to things that already exist in the world, and coming to terms with what they mean and how they relate to one another gets us to closer to understanding objective truth.

Representationalism is widely accepted in music technology research, especially in music informatics (Born Reference Born2020, sec. 2.3). This includes the premise that music can be considered an objective ‘thing’ rather than a process or activity – for example, that datasets of recorded audio tracks are music for all practical purposes.Footnote ³ Representationalism also underpins the ‘retrieval’ of conceptual information from signals, including transcription from audio to scores of discrete events characterised by pitch, loudness and timing, or the classification of audio tracks by genre, emotion or other factors, as if these words and numbers hold stable, universally agreed meanings across contexts and cultures. Instead, as Born (Reference Born2020) argues, purported universals in music informatics bear traces of an ‘I-methodology’ (Oudshoorn et al. Reference Oudshoorn, Rommes and Stienstra2004) where engineers who are disproportionately White, male and versed in mainstream Western pop music seek to create widely applicable tools by imagining their perspectives to represent a universal musical subject. Born sees ‘a risk of teleology in beginning with existing tools derived from the analysis of Western pop and seeing if they “work” for some fragmentary, decontextualised trait extracted from the total socio-musical existence’ of another musical tradition (Born Reference Born2020, 197).

Much has been written in recent years about the resultant challenges of diversity and ethics in music informatics and related fields (Huang et al. Reference Huang, Holzapfel, Sturm and Kaila2023; Gómez-Cañón et al. Reference Gómez-Cañón, Siavichay, Cavdir, Kaneshiro and Porcaro2025; García-Benito Reference García-Benito2025), particularly with the increasing use of AI systems for music analysis and generation at scale (Morreale Reference Morreale2021; Sturm et al. Reference Sturm, Déguernel and Huang2024; Morreale et al. Reference Morreale, Martinez-Ramirez, Masu, Liao and Mitsufuji2025; Morrison and McPherson Reference Morrison and McPherson2026). Here, it suffices to question (as I have at greater length in Morrison and McPherson Reference Morrison and McPherson2024; McPherson et al. Reference McPherson, Morrison, Davison and Wanderley2024) whether these words and concepts were ever truly embedded in the music to begin with, or whether such analyses have the effect of reifying post hoc descriptors about music into its purported generative basis (Van der Schyff Reference Van der Schyff2015). To put it a different way: if an automated music transcription system turns an audio file into MIDI or sheet music without obvious errors, has it therefore rediscovered at least part of the true and neutral basis for that music? Transcription is undeniably useful for practical analytical tasks, including performance-related tasks such as automatic score following; however, the question of whether words and symbols faithfully represent (rather than co-create) reality has important implications for the design of tools to create music.

2.2. Entanglement should mean more than just interaction

Before continuing, I would like to lay down a personal marker on my use of terminology from Barad’s agential realism (HCI-and NIME-adjacent summaries of which can be found in Frauenberger Reference Frauenberger2019; Reed et al. Reference Reed, Benito, Caspe and McPherson2024). Barad reads insights from Niels Bohr’s quantum ‘philosophy-physics’ through the feminist scholarship of Judith Butler and Donna Haraway (and vice versa) to challenge existing systems of scientific knowledge and representation. Barad’s project does not reject all science or the idea of a knowable reality; rather, agential realism proposes that the fundamental units of existence are not things (‘independent objects with inherent boundaries and properties’) but phenomena: ‘relations without preexisting relata’ (Barad Reference Barad2003: 815). For this reason, agential realism is often described as a ‘relational ontology’ (Frauenberger Reference Frauenberger2019), and putting relations at the foundation of existence leads Barad to a reconfiguring of terminology: intra-action (rather than interaction); diffraction (rather than reflection); agential cuts (bringing into existence localised patterns of cause and effect) and many more.

While the application of the Baradian lexicon to HCI and music technology can take many valid forms, there is a risk of researchers adopting these new terms as drop-in replacements for more familiar ones, leading to confusion on what entanglement theories might have to offer. For example, it isn’t necessary or helpful to use entanglement (or intra-action, or any other Barad-ism) to describe the movement and collisions of billiard balls on a table; classical physics is perfectly adequate for this.Footnote ⁴ Similarly, plenty of insightful literature exists on mutually interacting musical stakeholders (O’Modhrain Reference O’Modhrain2011; Marquez-Borbon and Stapleton Reference Marquez-Borbon and Stapleton2015) and on the relationships between performers and instruments (Jack et al. Reference Jack, Stockman and McPherson2017; Nijs Reference Nijs2017; Tuuri et al. Reference Tuuri, Parviainen and Pirhonen2017). Insofar as these sources assume that ‘human’ and ‘instrument’ are categories with well-defined boundaries and identities, entanglement thinking is not needed to make such points.

My interest in entanglement theories is rather in how they can destabilise concepts we take for granted, by proposing that boundaries are enacted rather than inherent. Barad draws out this point using Niels Bohr’s famous proposition that the apparatus of measurement is inseparably entangled with the thing that it measures – observer and observed mutually produce the phenomenon (and each other). As Barad puts it, ‘concepts are specific material configurations’ (Juelskjær and Schwennesen Reference Juelskjær and Schwennesen2012: 12) rather than representations mirroring an underlying reality. This suggests that words like ‘instrument’, ‘player’, ‘gesture’, ‘composition’ and even ‘music’ do not refer to stable, context-independent concepts; instead, the concepts are enacted through specific and localised materials, practices and relationships (Sergeant Reference Sergeant2016; Reed et al. Reference Reed, Benito, Caspe and McPherson2024).

With this in mind, when I say ‘X isn’t real’, more precisely I mean that X does not have an inherent identity independent of its relations, and conversely that the word for ‘X’ doesn’t have a stable meaning representing something that already exists in the world. Instead, the act of describing or measuring X is what brings X into being in the first place, in a process that should always remain open to scrutiny.

3.1. Reification and additive representations

I advocate for a greater awareness amongst designers of how building tools with notation (of any sort) acts as a reification: not a descriptive representation of what was already there so much as an inscription (Akrich Reference Akrich1992) of which parameters will become valued, which then serves to shape future practice through the things it makes easier or harder to do (McPherson et al. Reference McPherson and Lepri2020). Arguably, this is the idea behind a number of digital instruments which self-consciously merge the concepts of ‘instrument’ and ‘score’: composed instruments (Schnell and Battier Reference Schnell and Battier2002; Murray-Browne et al. Reference Murray-Browne, Mainstone, Bryan-Kinns and Plumbley2011), tangible scores (Tomás and Kaltenbrunner Reference Tomás and Kaltenbrunner2014), magnetic scores (Privato et al. Reference Privato, Magnusson and Torfi Einarsson2023), agential scores (Armitage and Magnusson Reference Armitage and Magnusson2023) and so forth. Here, the ideological non-neutrality of the notational system itself becomes part of the give-and-take between performer and instrument, and I suspect that few of these designers would suggest that the notation bound up in their instrument should be used as a representation of a broad swathe of other musical practices.

However, it remains common across DAWs and music programming languages alike to propose that music should be generated using one or more systems of symbolic notation. These notation systems might encode relatively specific conventions (such as MIDI or shorthand names for specific sounds found in live coding languages like Tidal Cycles or Sonic Pi (McLean and Wiggins Reference McLean and Wiggins2010; Aaron Reference Aaron2016)), or they might be more abstract (such as OSC or the programmer-defined parameter values that follow the time fields in a Csound score). In practice, systems with strong conventions are rarely completely restrictive (e.g., many keyboard-like conventions can be challenged with the careful use of MIDI Pitch Wheel or Control Change messages), while supposedly open-ended systems are often still used in conventional ways (McPherson and Tahıroğlu Reference McPherson and Tahıroğlu2020).Footnote ⁶ But we see in all of these systems either an additive property, where music is produced through combinations of smaller encapsulated elements, or a factorial property, where musical elements can be described by their position within a mathematical space which encourages the deployment of geometric metaphors (McPherson et al. Reference McPherson, Morrison, Davison and Wanderley2024). Even textural music without conventional ‘notes’ can be generated from additive processes such as granular or concatenative synthesis (Schwarz Reference Schwarz2007), while spatial audio might be produced from an accumulation of sound objects given discrete locations and radiation patterns (Wolstanholme et al. Reference Wolstanholme, Vahidi and McPherson2023).

To be clear, I do not claim that any of these tools, languages or notation systems are better or worse than any other. But any assumption that the individual quasi-atomic or factorial elements of notation are a ‘real’ representation of pre-existing musical phenomena risks inverting cause and effect. To further explore this point, I will turn to my experience in the highly traditional, notated music practice of classical string playing.

3.2. Why are MIDI string synths so disappointing?

Like many classically trained string players, I have a poor opinion of most MIDI bowed string synths, especially solo strings (as opposed to orchestral sections). Why, to my ears anyway, do so many of these systems sound like a pale imitation of good string technique? In early MIDI synthesis, the situation might have been attributed to the challenges of producing a realistic audio facsimile of a violin using the synthesis methods of the time (e.g., FM, subtractive or low-bandwidth sampling). Audio recording is clearly no longer the driving limitation in modern sample libraries. Later, in my own DMI design (McPherson et al. Reference McPherson, Gierakowski and Stark2013), I attributed it to a lack of sufficient real-time ‘expressive’ control over parameters such as vibrato (often relegated to mechanical-sounding LFOs) or dynamics (often chopped up into note-level modulations rather than unfolding naturally over a phrase). I thought that putting those parameters under the fingertips of the DMI player might yield a more realistic emulation.

I now believe the problem speaks to a deeper incompatibility between assuming MIDI messages to be the generative substrate of music, versus what I do when I play the viola. As an orchestral and chamber player, almost all of my playing is from notation. However, as musicians have known for centuries but technologists still sometimes forget, standard Western notation acts as more of a culturally and historically situated aide-mémoire than a literal description of precisely what acoustic events will take place.Footnote ⁷ Notwithstanding Magnusson’s description of notation as choosing what is valued (Magnusson Reference Magnusson2019), there are many aspects of string playing that are highly valued but not assigned symbols.

One of these aspects is articulation, the complex acoustic-tactile transients that occur between (and sometimes within) notated events. Contrary to the MIDI worldview, these are neither instantaneous events (e.g., note velocity), nor are they amenable to manipulation at a ‘control rate’ suitable for human fingers turning plastic knobs. Sample libraries sometimes attempt to classify articulation into discrete linguistic categories (legato, staccato, spiccato, martélé, etc.) as if these words have stable, well-defined meanings across different instruments and players (or even across a single performance). The MIDI 2.0 specification even includes a Note On with Orchestral Articulation messageFootnote ⁸ featuring an 8-bit classification for articulation type and various subclasses covering commonly used words about string playing, including bowing, left-hand techniques and which string a note should be played on. It might be a step forward from mere note number and velocity, but the proliferation of sub-categories and attributes also suggests an attempt to shoehorn string playing into a protocol that fundamentally is not suited for it.

Ultimately, the practice of bowed string playing isn’t additive from small atomic units, nor is auditory–motor imagery a linguistic or conceptual process. Its musical language is both bigger (phrasing) and smaller (articulation) than notes, and it emerges from a process of 4E cognitionFootnote ⁹ deep and reciprocal enough to challenge the very notion of the separation of ‘performer’ and ‘instrument’ into well-defined entities (Nijs Reference Nijs2017). Notation can guide this process, and it can be transcribed post hoc from a performance, but to think of it as the ‘real’ generative basis of the practice will always leave the technologist playing catch-up.

4.1. Frequency doesn’t (inherently) exist

To put the argument into clearer relief, I will leave aside the complexities of human pitch perception; my argument here is not that frequency is objective but pitch is subjective. Instead, I will take fundamental frequency (periodicity) as a direct correlate to pitch and argue that frequency doesn’t exist as an inherent and independent entity. Rather, the phenomenon of frequency only comes to exist in context through an entanglement with the apparatus of measurement, the frequency detector. Fourier theory tells us that a basic prerequisite for any frequency detector is a time window long enough to cover at least one period of the signal, and that precision in the frequency domain comes at the direct cost of imprecision in the time domain (i.e., longer time windows are needed to resolve finer details in frequency). The most obvious implication, as is well known in audio engineering, is that frequency detection on real-time signals necessarily incurs a significant latency, and that the lower the frequency to be detected, the longer this latency becomes.

Frequency is a useful analytical frame, especially for offline or latency-insensitive applications. But does that make it a fundamental, context-independent basis for musical sounds? Even leaving aside inharmonic (‘unpitched’) instruments like drums and noise-based musical practices, audio signals that we take to have well-defined pitch often exhibit significant regions of ambiguous periodicity or non-periodicity. This is particularly true for transients, where establishing regularity of oscillation after an input of energy can take tens or hundreds of milliseconds. Frequency detectors are easily stymied by this behaviour, such that designers often resort to waiting for even longer time periods to elapse before taking measurements of the frequency – even as they purport that frequency is a fundamental generative property of the audio signal. Put another way, instantaneous temporal onsets and deterministic frequency depend on mutually incompatible apparatuses.

Of course, myriad digital and analogue instruments feature frequency generators (oscillators) alongside some means of producing transient signals (envelope generators). Practically, even a simple oscillator and voltage-controlled amplifier can avoid the latency problem that afflicts more complex instruments (like MIDI guitars) which use audio feature analysis to control synthesis processes. Analytically, every audio system faces the same mathematical constraints around frequency measurement and window sizes. Does this mean frequency and onset should indeed be considered inherent properties of musical audio, independent of any representation?

Reconciliation can be found once again in Barad’s statement that ‘concepts are specific material configurations’ (Juelskjær and Schwennesen Reference Juelskjær and Schwennesen2012). They continue: ‘the details of the apparatus – like the bolts fixing one part of the apparatus to another, or springs that enable parts of the apparatus to move and be responsive – are of fundamental importance’. Here, both the oscillator and the frequency detector are specific material configurations that enact a concept of ‘frequency’, but despite the shared word, the two concepts are not necessarily identical. To examine the implications in more detail, I will return to the case of viola playing.

4.2. Where is ‘pitch’ located on the viola?

This sounds like the set-up to a bad viola joke, but I intend it as a serious question. What are the material properties I should inspect on a viola to discover the pitch it produces? The obvious answer from musical acoustics is to examine the length, tension and mass density of the strings to calculate their resonant modes. In this idealised account, the rest of the instrument (bow, bridge, body, air, any part of the player’s body other than the immediate contact point with the string) might alter the spectrum or dynamics of the sound, but to a close approximation will be irrelevant to determining frequency.

However, when I begin to draw the bow on the string, ‘frequency’ doesn’t immediately jump out. Rather, patterns of stick-slip friction emerge at the region of bow-string contact and propagate in both directions along the string, chaotically at first, and eventually (with the right playing technique) reinforcing one another in a periodic process known as Helmholtz motion. Acousticians have shown the establishment of Helmholtz motion on a bowed string to have chaotic properties, dependent on particular combinations and micro-details of bow speed and pressure (Woodhouse Reference Woodhouse1993). In my own playing – largely as a result of my imperfect technique – some transients unpredictably settle into other vibration regimes such as multiple flyback (a harsh, buzzy sound) or multiple slip (producing airy, flautando sounds or harmonics). Acoustically, there is no a priori physical manifestation of pitch within the instrument, only complex spatial-temporal phenomena that mostly resolve into periodic oscillation in between periods of instability.

Lest this seem like so much (bow)hair-splitting, there is a further wrinkle. My viola, like many larger violin-family instruments, has a wolf tone, a high-Q resonant mode within the instrument body. Vibrations in acoustic instruments are bidirectionally coupled: string vibrations couple through the bridge to the body of the instrument, and the same is true in reverse. Wolf tones occur when the back-propagated energy from a resonance in the instrument’s body interferes with the establishment of Helmholtz motion on the string (Gourc et al. Reference Gourc, Vergez, Mattei and Missoum2022). On my instrument, the resonance occurs near the pitch notated as F above middle C (349Hz). As a result, under certain playing conditions – quiet dynamics, light bow pressure, high up the lowest (C) string – it is nearly impossible to get the instrument to speak at this pitch at all. Instead of clear periodicity that would match an analytical concept of frequency, I get a strange warbling, with concomitant changes in tone quality and tactile sensations through the bow. If ‘frequency’ is actually a material configuration, it is an idiosyncratic one, and those idiosyncrasies become part of my playing.

4.3. Virtual violas: seeking refuge in physics

None of the aforementioned properties of my viola, even its idiosyncratic wolf tone, are inherently impossible to model numerically. Physical modelling of acoustic instruments, including string instruments (Bilbao and Ducceschi Reference Bilbao and Ducceschi2023), is a mature enough technique to feature in high-quality commercial audio plug-ins.Footnote ¹⁰ This research extends to how such simulations might be musically controlled (Demoucron and Rasamimanana Reference Demoucron and Rasamimanana2009; Carrillo et al. Reference Carrillo, Bonada, Maestre, Guaus and Blaauw2011) and what kinds of sensors might capture the necessary physical measurements (Néill et al. Reference Néill, Van Walstijn and Ortiz2025). In an important way, physical modelling offers a refuge from prescriptive musical concepts: for many forms of numerical simulation, nothing in the mathematics assumes the prior existence of stable concepts of ‘frequency’, ‘note onset’ or any other musical idea. Rather, these concepts gain meaning through contextual interpretation of what the models do, analogous to how musicians interpret acoustic instrument behaviours. Hypothetically, given a sufficiently comprehensive model of an instrument and acoustic space, one could produce a credible enough facsimile that a human listener couldn’t tell the difference, even as the computation would remain agnostic to any concept of ‘notes’.

But can such a simulation really remain note-agnostic? With physical modelling, one can easily produce an immaculate ship in a tightly sealed bottle, impervious to human manipulation. To render a numerical simulation playable, the problem of physics quickly shifts to the problem of human interfacing – or more controversially, simulating humans. Interfaces used to control bowed string simulations, such as graphics tablets (Demoucron and Rasamimanana Reference Demoucron and Rasamimanana2009), load cells (Néill et al. Reference Néill, Van Walstijn and Ortiz2025) or 3D motion capture (Carrillo et al. Reference Carrillo, Bonada, Maestre, Guaus and Blaauw2011), are each in their own way distinct from the high-bandwidth intercoupled audio-tactile dynamics of acoustic bowing. Invariably, assumptions must be encoded to render the numerical simulation useful to a performer. These might include MIDI control of pitch or, more expansively, machine interpretation of scores (Demoucron and Rasamimanana Reference Demoucron and Rasamimanana2009), or the classification of bowing technique (Maestre et al. Reference Maestre, Blaauw, Bonada, Guaus and Pérez2010) into discrete categories similar to those found in MIDI 2.0 (Section 3.2). The musical assumptions haven’t gone away, and they have just moved upstream from the numerical simulation.

5.1. Notes are enacted, not extracted

The upshot of a relational framing of instrument identity is that the Baradian apparatuses which give meaning to concepts like notes and frequencies must be drawn widely enough to incorporate (more-than-)human cognition and cultural systems. This view presents severe challenges to the digital instrument designer. We have no technical means to observe the intentions or sensorimotor imagery of a performer, nor any robust way for a digital instrument to be aware of and adapt to the cultural situation of its use. Even worse for the designer, intentions and cultural situations are never fixed or stable enough to underpin a design specification without changing in the process. As a result, the technical ‘solution’ comes to define the problem rather than the other way around: relying on narrow mathematical or engineering-driven definitions of concepts ends up reconfiguring both human thinking and musical cultures around those definitions. As Morreale et al. (Reference Morreale, Martinez-Ramirez, Masu, Liao and Mitsufuji2025) put it, ‘epistemology (our knowledge about music) determines ontology (what music is)’.

Consider a practical example: guitar-to-MIDI converters (Section 4.1) will typically take audio as input (possibly with separate channels per string) and produce MIDI note messages as output. The converter encodes rules about what constitutes a ‘note’ and how the pitch and velocity of the symbolic message will be derived from a set of signal processing elements. As Reed et al. (Reference Reed, Benito, Caspe and McPherson2024) argue, the device becomes a locus of knowledge about music which will inevitably diverge from the human musician’s own knowledge of the same musical situation. One result is a set of frictions. Reed et al. describe as an ‘ambiguity shift’, in which the performer experiences these competing loci of knowledge as glitches or technical failures. Indeed, online reviews of guitar-to-MIDI converters sometimes discuss challenges with perceived tracking accuracy.

However, the actual situation is more insidious than discussions of tracking accuracy let on. As Nijs (Reference Nijs2017) and Ingold’s firsthand account attest, the boundary between player and instrument is fluid and contingent. A guitar-to-MIDI converter is thus better analysed as a guitarist-to-MIDI converter, accounting for the reciprocal effects of the guitar on the guitarist’s ideation, perception and action. The guitarist, playing without such a device, would have had a certain set of sounds and techniques in mind. They could choose to hold firm to those ideas when playing with the MIDI converter, which will bring all of the aforementioned frictions to the fore. Instead, what is far more likely to happen is that the guitarist will change their expectations and techniques to match what the MIDI system wants to produce. The effect of players adapting their techniques to the affordances of an instrument is well established (Tuuri et al. Reference Tuuri, Parviainen and Pirhonen2017; Jack et al. Reference Jack, Stockman and McPherson2017). By adapting to think in terms of note onsets, the distance between the two loci of knowledge is reduced and the player experiences greater conformance with their expectations, with more satisfying results. This, too, can be seen in online reviews and discussions of such technologies.Footnote ¹²

This shift of technique and intention might go unremarked in the majority of cases. The vagueness of the words and concepts we use to talk about music could easily sustain the fiction that nothing substantial changed with the introduction of MIDI conversion. Guitar playing was already about notes, we might tell ourselves – just look at tablature! We’ve merely built a machine that understands our notes! But this elides the fundamentally different role of the ‘note’ as an (incomplete) cue for performance versus a generative element of the instrument itself. Concepts, humans, materials and discourses are still entangled, but technology-led decision-making ends up enacting what Barad calls an agential cut, giving meaning to the boundaries between entities and opening up certain possibilities while foreclosing others (Barad Reference Barad2003). Notes become real through our own designerly actions, and we believe they were there all along.

5.2. Is it still okay to like notes?

There exists a considerable body of DMI literature on instruments as co-creative agents and performance as exploration of the material properties of instruments (Mudd Reference Mudd2023; Melbye Reference Melbye2023; Stapleton and Davis Reference Stapleton and Davis2021; Tahiroğlu Reference Tahiroğlu2024; Redhead Reference Redhead2025). That literature is helpful in unsticking ossified conventional wisdom and narrow historical discourses of what good music ought to be. This paper shares with that literature some degree of revolutionary fist-shaking against the real or imagined hegemony of Western tonal music with its normative ideas of expressivity and virtuosity (see Gurevich and Treviño (Reference Gurevich and Treviño2007) on the latter point) or against mass-market popular music with its tightly circumscribed musical conventions. Battle lines (or maybe just taxonomies?) are drawn, with contrasts thrown into sharp relief: Mudd (Reference Mudd2019) distinguishes between communication-oriented interaction (traditional music) versus material-oriented interaction (experimental music); Magnusson (Reference Magnusson2009) distinguishes instruments into extensions of the body (acoustic) or extensions of the mind (digital/live coding).

It’s important to acknowledge the very real power imbalances between large corporate interests ingrained in every part of the popular music ecosystem, the major institutional support and conservatoire education system supporting Western art music, versus the limited scale and resource of experimental practices (though as Hayes and Marquez-Borbon (Reference Hayes and Marquez-Borbon2020) show, institutional support has its own influences on DMI research). But I would be cautious of any effort to canonise experimental practices by codifying their essential elements, assigning them new words and symbols, and then using those words and symbols to guide future design processes, or relatedly, to enact new rigid binaries (even with respect to ‘old’ ways of doing things) when their avoidance is one of the very things entanglement theories try to teach us.

Many otherwise inspiring and well-conceived new instruments would liberate us from the hegemony of certain musical concepts by precluding their existence in the first place. Plenty of new DMIs which are capable of vast abstract timbral palettes or complex generative behaviour are also entirely unsuitable for playing ‘Happy Birthday’, even if the player was so inclined. This is neither good nor bad in any absolute sense, but it represents a different agential cut and calls for careful attention to what practices are excluded by it and why (Giraud Reference Giraud2019).

So, what happens if I intend to play notes? Is that forbidden?Footnote ¹³ As a classically trained acoustic instrumentalist, am I stuck between accepting reductive and caricatured representations of music in mass-market digital tools (Pardue and Astrid Bin Reference Pardue and Bin2022), parting ways with core elements of my musical language or rejecting the use of digital technology entirely to stay within my comfort zone? The problems begin when such a decision has to be categorical rather than nuanced. Musical intentions enact concepts and categories rather than simply following them, but intentions are also technologically mediated (holding a viola, I think of certain musical ideas far more readily than others; sitting at a piano, a different set of ideas is ready to hand). This situation puts the instrument designer in a position of considerable responsibility.