2.1. Liveness

What is a live performance? In acoustic music, this question appears to be straightforward: a concert is live, but a recording is not. However, looking closely, a complex of ambiguous cases arises. For example, recorded music that is edited, mastered and produced in a studio differs from so-called live recordings. The live recording captures spontaneity, risk, an acceptance of error and fault, and an experience closer to a live concert than the studio recording. Consequently, the inclination to assume all live performances exhibit the same degree of liveness contravenes the different experiences that grow from different live music-making. Performances of precisely notated scores, scores with indeterminate procedures, performances of symbolic scores open to individual interpretation and completely improvised music (a kind of real-time composition) all position works within degrees of liveness.

In electroacoustic music, the range of possible human interaction compounds the issue. Some electroacoustic music is fixed as a recording, and others change with each realisation. Further, these facets are uncorrelated to the extent to which humans input or interact with the music. These characteristics introduce more variables to the multidimensional space of liveness.

Stroppa only lists two conditions for a performance to be experienced as ‘live’. The first is the presence of a human performer, but equally important, they must be playing an instrument ‘that is accepted as such by the musical community’ (Stroppa Reference Stroppa1999: 50). He points out that a live performer on a stage working a mixer would not be as live for an audience as a performer on a traditional instrument. Yet, there is a growing audience who recognise equally impenetrable laptop performances or no-input mixer ensembles as live. This may be the acceptance of technology after 16 years as bona fide instruments, or it could be due to other factors discussed later.

Emmerson (Reference Emmerson2007) wrote comprehensively on what makes electroacoustic music ‘live’. He problematises shades of difference, arguing quite convincingly that even fixed medium works create a living presence through human reception in contradiction to Stroppa. In his discussion, he acknowledges tricky ‘borderline’ cases, one of which is the example of machine-generated music without human interaction. In this article, this borderline case is narrowed specifically to real-time computer-generated music, but broadened to include levels of human interaction, human control and computer autonomy. This focus illuminates the problems with nomenclature as well as directly addresses the relationship between live and real-time.

The range of recording contexts (studio versus live) and the multiple generative/interpretive strategies (from score to improvisation) uncover other aspects of what one considers when thinking of live music: spontaneity, indeterminate influences, risk of failure and acceptance of error, as well as the living presence and audience experience. There are likely many more unconscious or smaller factors, but these rank among the most significant. Though individual listeners may rate the musical encounters differently, these factors clearly transcribe a multidimensional space conferring liveness.

In summary, liveness is attributed by:

• ∙ spontaneity – where musical decisions, from minor interpretive differences to larger formal choices, are made;

• ∙ risk of failure – some musical gambits may not be entirely successful or convincing and errors can be made;

• ∙ an acceptance of error and fault – mistakes, failures, errors and other unwanted consequences cannot be ‘cut out’ before the audience receives them, a condition that is accepted if not necessarily welcomed;

• ∙ audience/performer contract – the performer or composer asserts liveness and the audience agrees to the assertion, impacting the audience experience;

• ∙ indeterminacy – this may be a compositional decision, but it is also a result of spontaneity, risk and error;

• ∙ living presence – a complicated combination of agency, action and human reception thoroughly examined and defined by Emmerson (Reference Emmerson2007);

• ∙ audience experience – though also significantly impacted by the audience contract, this includes the social and psychological conditions of attending a performance.

2.2. Real-time

Nearly every discussion of real-time defines it as a binary. However, the different terminology authors utilise reflects their own aesthetic and philosophical stance on the issue. When comparing the authors’ definitions, a wide range of what is considered to be real-time emerges. This article argues, through these definitions, that real-time is a complicated scale, not a categorical qualification.

Miller Puckette does not explain his dialectic of ‘compositional’ and ‘performative’, but his description of challenges in developing Pd gives rise to his intentions. He approaches the categories from a programmer’s point of view: things that must be scheduled in a pre-composed, timed order versus things that happen freely. He refers to the former as ‘compositional’ and the latter as ‘performative’ (Puckette Reference Puckette2004). The performative seems to rely on human input or triggers, as well. Yet, composed elements often trigger real-time processes.Footnote ² The boundary between compositional and performative is not clearly distinguishable as non-real-time or real-time.

Carl Seleborg also tackles the difference from a software perspective. His words, real-time and time-deferred (or time-delayed depending on one’s translation of temps-différé), reflect the original purpose of such tools as Max (real-time) and OpenMusic (time-deferred). OpenMusic facilitates computer-aided composition (Seleborg Reference Seleborg2004). CAC is traditionally considered to be a non-real-time process: the composer programs the computer to generate materials, which he or she then translates into a symbolic score for human performance.

From this, Andrea Agostini and Daniele Ghisi attempted to bridge the gap between offline CAC and real-time processes. Their software, bach, allows for Max to generate symbolic scores in real-time. This allows for the architecture of Max, software intended to be performative in Puckette’s and their terms, to be a source of CAC, resulting in human-readable scores. Agostini and Ghisi refer to traditional CAC as ‘speculative’ (Agostini and Ghisi Reference Agostini and Ghisi2013). This word suggests that composers working with CAC do not know what their results will be with each program. However, most seasoned composers have some idea of their algorithm’s output, with only occasional surprises. Therefore, perhaps ‘notional’ should replace speculative.

The phrase ‘algorithmic music’ carries with it enormous baggage. In some instances, a speaker may be referring to completely computer-generated, highly formalised music. In other instances, a composer may simply refer to anything generated by an algorithm. Nick Collins is careful to use ‘generative music’ for the materials used in real-time laptop performance, although he occasionally refers to algorithmic composition (Collins Reference Collins2003). Perhaps this is a nod to the fact that real-time generative music must rely on algorithms, while accommodating the connotation that algorithmic music is autonomous and non-human. He elaborates various approaches to live laptop performance and advocates for audience education. An educated audience would have a different contract with the performer and a different experience from a naïve audience. Collins makes a subtle distinction between performers using off-the-shelf software and performers using programming environments such as Pd or SuperCollider (another real-time music tool). Those who use off-the-shelf software typically rely on pre-programmed actions more than those who do not. This reflects the reality that, though live, laptop performers can still be using software to simply trigger pre-programmed sounds, passages, gestures, loops, etc. This suggests that live laptop performance, including live coding, can display varying degrees of real-time processes. In his discussion, Collins acknowledges that performance can consist solely of a mouse click or key press, but that the audience needs to learn, somehow, that these are performative actions (to use Puckette’s words).

Guy Garnett carefully delineates what music is interactive by its aesthetic implications. In the broader genre of performance-oriented computer music, where there is at least one live performer in the mix, he contends with the subgenre of interactive computer music. His argument is that any human interaction:

For Garnett, interactivity is the key to live performance. Still, a button-push tends to stretch the term ‘interactive’; after all, even fixed-medium works without live diffusion require someone to press a button.

There is consistency between Garnett and Collins in what minimal actions a human performer may take to make a live performance. Garnett never refers specifically to real-time computer music, though material that responds to human interaction can be as fixed, indeterminate or real-time as any laptop performance. So, Garnett’s discussion does not differ wildly from Collins. Though interactivity here is a circumstance of real-time, it requires more discussion. An elaboration follows in section 2.4.

Similar to Garnett, Robert Rowe asserts that interactivity assures liveness, that ‘the musical values evinced in interactive compositions are ultimately the same as those underlying a string quartet’ (Rowe Reference Rowe1999: 87). Unlike Garnett, Rowe does not make a constitutional stand against real-time computer music without a human performer. Though he distinguishes between real-time computer music with and without human interaction, he does not privilege one over the other as more live, simply more interactive. Rowe also alludes to the notion of a computer as performer. It follows that, once a computer becomes performer, it becomes a live manifestation, a kind of artificial human presence. Whether or not that artificial presence is sophisticated, if the audience perceives an agent with will, then a living presence emerges.

Perhaps, then, an extreme version of this concept can be found with George Lewis. In his work, the computer-as-performer is elevated to the equivalent status of the human performer. He argues that this multidominant approach exists outside the pan-European experimental ethos (Lewis Reference Lewis2000). There is definitely a notion of liveness in his work, in that both computer and human performers are improvising: responding to each other, listening to responses, accompanying or initiating new material. As the computer is improvising, it is necessarily composing in real-time with a significant degree of autonomy.

From these authors, real-time and non-real-time are identified as performative versus compositional, real-time versus time-deferred, performative versus speculative (or notional), generative versus algorithmic, interactive versus non-interactive, improvised versus notated, indeterminate versus fixed. By examining these dialectics from a distance, however, each implies diverging interpretations of real-time and live. Some of these positions overlap, while others cross orthogonally, forming complex intersections. In section 3, concrete musical examples will differentiate these concepts and how they operate in the discourse.

2.3. Live electronics

In the first uses of the phrase ‘live electronics’, composers referred to analogue circuits that synthesised or modified sounds immediately. This stood in stark contrast to the studio composition on fixed media. Even in the earliest days of computer music, compositions had to be coded, compiled, run and recorded, so there was no more liveness to computer music than tape music.Footnote ³ However, digital audio technologies developed at an astounding rate. Though computers took some time before they could process sound in real time, dedicated digital hardware designed for specific purposes (e.g. reverb, filters, oscillators) fell into the category of live electronics, most likely because they functioned as black boxes much like analogue components. The crucial distinction was that these devices operated in real time. However, ‘real-time’ as an adjective appeared only when computers became fast enough to behave like live electronics and probably as a consequence of computer science concerns. This may account for early ambiguity in applying ‘live electronics’ or ‘real-time’ to a performance. At least, this may be true for Anglophones, as the French used temps réel in studio contexts (Emmerson Reference Emmerson2007).

In 1991, Barry Schrader asserted that audiences struggled with performances of tape music because it had no performer. He identified live electronics as the first solution to this unease (Schrader Reference Schrader1991). Therefore, another aspect of ‘live electronics’ implies there is a human performer of an instrument or otherwise. Real-time computer music is agnostic on that account. Emmerson’s definition of ‘live electronics’ requires a live performer, whether on electronic or acoustic instruments, but also more inclusively someone ‘who may cause, form or influence [the sound] through electronically mediated interfaces under their immediate control’ (Emmerson Reference Emmerson2007: 90). This does not fix the issue, either, as Emmerson is quick to point out borderline cases that prove the rule.

In my experience today, people reserve ‘live electronics’ to refer to devices or systems that synthesise or process sounds in real time but do not qualify as stand-alone computers. The language may or may not be sorting itself out. However, this means that there are musical practices descending from compositional uses for live electronics that result in significantly different approaches to those practices descending from computer music, even if eventually all works get ported to computer software. So what to make of the computer re-creations of analogue electronic works, such as Christopher Burns’s (Reference Burns2002) import of Stockhausen’s Mikrophonie I, or Miller Puckette’s Pd Repertory Project (2001)? The easy, yet somewhat assailable, answer is that today’s audiences do not see a computer on a stage and call it ‘live electronics’ regardless of ancestry or human presence. So, live electronics do not factor into the discussion of liveness in real-time computer music.

2.4. Interactive music

Depending on a musician’s purpose, ‘interactive’ can be quite broad or quite narrow. Garnett defines interactive as works ‘wherein the performer in some way controls the electronics or the electronics affect the performer’s sounds’ (Garnett Reference Garnett2001: 21). In this definition, control can flow in one of either direction: performer to electronics or electronics to performer. But, some musicians seem to prefer when control flows in both directions. Philippe Manoury singles out Arshia Cont’s innovations with Antescofo (2007) that ‘allow a fluidity of “dominance”, between musician and computer: each may, at a given moment, listen and respond to the other’ (Manoury Reference Manoury2013: 85). Lewis and Rowe both created pieces where the computer’s role was increasingly dominant even before Cont’s software (Rowe Reference Rowe1999; Lewis Reference Lewis2000).

The first efforts to integrate live performers with tape music might have allayed audience’s needs for human stage presence, but the performers themselves had mixed opinions about performing with immutable accompaniment. Perhaps more commonly, performers agree with Elizabeth McNutt, a veteran performer of flute and electroacoustic music. She puts it eloquently:

At the same time, Stroppa felt that his performance of Kontakte by Stockhausen (for percussionist, pianist and 4-channel tape) was as nuanced as any acoustic ensemble work, despite McNutt’s and others’ criticism of tape’s rigidity (Stroppa Reference Stroppa1999: 43). However, McNutt’s opinion seems to be representative of the majority of performers’ experiences, and interactive electronics are the solution.

The main qualifying distinction here is that at least one component has an effect on another, and the degree of interactivity relies on the degree to which both components affect each other. Therefore, this can be applied to live electronics or real-time computer music; it is a separate territory.

3.1. Determinate vs indeterminate

The works most accurately described as indeterminate are works that utilise statistical probabilities to generate material. Many contemporary examples use these algorithms to create music in real-time. This accommodates Iannis Xenakis’s message regarding statistical works: multiple versions must be experienced for truly stochastic music (Xenakis Reference Xenakis1992). Limited by technology, Xenakis had to create canonical examples of works that otherwise would be ideally realised multiple times. These fixed-medium works, though created by indeterminate processes, are no more or less live than other electroacoustic, fixed-medium works. Their liveness relies on Emmerson’s ‘living presence’. However, Peter Hoffmann re-programmed the Gendyn system created by Xenakis, allowing for one such work, Gendy3, to be realised in real-time (Hoffmann Reference Hoffmann2000). Thus, Xenakis’s original intention for multiple versions can be realised for an audience in real-time.

But, Xenakis’s work shows that real-time works are not the only way to make indeterminate music live. Pieces such as Pithoprakta (1956) and Syrmos (1959), among others, are early instances of CAC. The processes themselves are not created in real-time, and these indeterminate works were actually translated into standard musical notation for human performance. The materials may be constructed offline and the indeterminate processes fixed, but the performances are as live as any acoustic, notated work. In fact, some indeterminate works by Boulez or Stockhausen may be more ‘live’ because they require real-time decision making for acoustic, notated scores.

So it seems, though indeterminate processes can now be realised in real-time and potentially live, they are not clearly so.

3.2. Recorded vs real-time generation

When it comes to electroacoustic music, what is recorded versus live becomes increasingly complex. Of particular interest are older works created for analogue technology that are translated to the real-time digital domain. Puckette (Reference Puckette2001) translated four works into real-time digital versions, but the works he translated were already considered live (see Pd Repertory Project 2007). They were works that in their original forms used live performers (though the electronics involved could be live or recorded).

Burns realised two early analogue works in real-time, as well. Though his treatment of Stockhausen’s Mikrophonie I resonates with Puckette’s examples, Burns also recreated I Am Sitting in a Room by Alvin Lucier in real-time (Burns Reference Burns2002). To some, the authentic version of this work is Lucier’s original recording in 1970. The original ‘score’ is a paragraph written by Lucier. In his original instructions, Lucier did not actually refer to a real-time version (Tarantino Reference Tarantinond), perhaps due to the technology available in 1969. Later, Lucier included real-time realisations as a potential version of the work (Lucier Reference Lucier1995). At the very least, this undermines the notion that his original recording is the most authentic. Burns’s discussion of the realisation opens the door to many interesting insights into Lucier’s work. The real-time versions ‘offer opportunities which can only be implicit in fixed-media versions of the piece. A live realization tends to increase our sense of wonder at the piece’ (Burns Reference Burns2002: 61). He asserts that Nicolas Collins’s argument for the necessity of the tape version is weakened by this experience, despite its value in bringing a private space into a public space (Collins Reference Collins1990; Burns Reference Burns2002).

Perhaps Burns’s most interesting point is that the real-time version makes the piece more live, because (contradicting the original text) we are no longer sitting in a room ‘different to the one you are in now’. Like Hoffmann’s recreation of Gendy3, each real-time version of I Am Sitting in a Room is different; it depends on the acoustic factors of the inhabited space. The very make-up of the audience is a living presence in the work. This version feels more live than Hoffmann’s Gendy3, though. Yet, the real-time version is a process that runs with practically no intentional human interaction, just like Gendy3. In many ways, the real-time version of I Am Sitting in a Room includes factors that make the well-known tape piece more live: risk of failure, indeterminate influences, a living presence and a different audience contract with the performer/composer. Gendy3 does not have the audience contract, the indeterminacy is less living, and the risk of failure may be smaller.

3.3. Notated vs improvised

If liveness in acoustic music can be qualified by spontaneity, living presence, indeterminate influences, risk of failure, acceptance of error, audience experience and so on, then this implies that improvised music could be more live than notated music. This implication is maintained in the case of electroacoustic music.

One example is Manoury’s recent work, Tensio (2010). In this work, a notated string quartet interacts with real-time computer processes to create material. The string quartet part is composed and notated precisely, allowing for no changes beyond small normative interpretations. The computer, however, has a variety of different roles. Although the synthesised sounds are all generated in real-time, the structure, pitches, gestures and so on of the sounds come from a variety of indeterminate and determinate procedures (Manoury Reference Manoury2013). On the one hand, there are nearly recorded passages where timbres are generated live; on the other hand, there are completely live, indeterminate procedures within highly limiting parameters. On a straightforward glance, the inclusion of live performers would situate this work as live, but it is unclear how live the real-time computer material is.

In comparison, in radical opposition both musically and philosophically, Lewis’s work Voyager (1987 [1993]) is completely improvised by both a human and a computer performer. Both the human and computer performers listen, respond and can initiate new material. As a freely improvised work, a recording is not representative of the piece. Furthermore, the computer is an additional living presence.

Between these two extremes could be Rowe’s Maritime (1992). Two-thirds of the work is notated and one-third is improvised. It also requires a human violinist to interact with the work, but there are moments when the computer has its ‘own personality’ (Rowe Reference Rowe1999: 86). So, like Manoury and Lewis, this work is live. But it could be argued that it resides between the two pieces in degrees of liveness, if it could be argued that improvised music is more live than notated music and that computer autonomy also factors in liveness.

If computer autonomy is a generator for living presence through the notion of computer performer, then the discussion must revisit real-time computer music that does not require a human performer. In that case, the computer not only has autonomy, but is also the only performer in the work. So, what kind of real-time material engenders the role of computer as performer? There may be no prescriptive answer. However, the conditions that lead to whether a computer becomes a performer rely on the audience/performer contract. If the audience is willing to believe that the computer is more than an instrument, in fact the agent with will, the same real-time material becomes more live. If the computer performed a solo version of Voyager, or if it was creating a real-time version of Gendy3, the comparative liveness of the two pieces ultimately resides with the experience of the listener.

Manoury, Rowe and Lewis illuminate shades of liveness through the subtle distinctions of notated versus improvised material. But, there is another improvised genre that problematises this discussion: laptop performance. In this case, unlike these others, the human performer is not using a traditional instrument and the computer is not a performer. Rather, the computer is the instrument. Like anyone performing on an instrument, material can be composed and decided in advance or improvised extemporaneously. Almost all material generated by the laptop is happening in real-time, but there could also be recorded material that is triggered at liberty.

Collins’s description of the human interactions in live laptop performance can look rather similar to real-time computer music without human intervention, because audiences cannot relate the minimal physical gestures to the resulting music. At some basic level, audiences participating in live laptop performances know that it is live, even if human contribution is impenetrable. In the case of live-coding concerts, where the human action is made visible or understood by the educated audience, the experience itself may be more live. The audience may also assume that the material can be indeterminate and/or improvised. So, liveness in laptop performances is more an audience construct, not derived from the creation of the music itself. At the same time, there is risk associated with improvisation; Collins points out that ‘in the design of interactive system [sic] for performance the trade-off between improvisation/adaptability and preparation/quality/seamlessness must be confronted. Risk taking can be made an important part of the work, or the work will demonstrate measured pre-composition’ (Collins Reference Collins2003: 71) This consideration enforces the audience’s acceptance of the liveness of a laptop performance.

The discussions of liveness and real-time music with these particular musical examples raise more questions than answers. In problematic distinctions, this may be more desirable than quantitative, definitive labels. Therefore, it is rather more interesting to compare examples and determine relative degrees of liveness. This article proposes the following scales as a starting point.

First, liveness in acoustic music spans the range from studio recording to a live, improvised concert performance (see Figure 1).

Figure 1 Liveness of acoustic music.

Liveness of electroacoustic music also spans a range, where real-time computer music exists in many degrees of liveness (see Figure 2).

Figure 2 Liveness of electroacoustic music.

Given these axes, one could potentially situate the examples here as shown in Figure 3.

Figure 3 Comparative liveness of musical examples and its problems. Where does the real-time version of Gendy3 belong?

These scales are by no means definitive. Boundary cases overlap, and some examples can be located in multiple spaces, especially given the ephemeral nature of the audience’s experience. The spectrum serves only as a springboard for debate.