• Home
• About CHARM
• Discography
• Sound Files
• Research Projects
• Outputs
  • Publications by Project
  • Publications by Year
  • Publications by Researcher
  • Presentations
  • Changing Sound of Music
    • 1. Introduction
    • 2. Performances
    • 3. Recordings
    • 4. Singing
    • 5. Violin playing
    • 6. Piano playing
    • 7. Style change
    • 8. Expressivity
    • 9. Conclusion
    • Updates
    • Bibliography
    • Captions
    • Indexes
• Analysing Recordings
• History of Recordings
• Download Area

2. Understanding the sources: Musical works and musical performances

2.1. Musical works

¶1 There are a number of issues we have to think about, and there’s quite a lot we need to know, before we can talk sensibly about music in recordings. First we need to take a view on the relationship between music and performance, and then between performance and recorded performance. Among other things, we need to know how recordings were made—by no means a small topic, since techniques changed throughout the twentieth century and continue to do so. Then we shall be in a better position to determine the extent to which recordings can safely be used as sources in projects that seek to study performance.

¶2 What we mean by ‘music’ is going to be crucial to our view of recordings and their relation to performance and so I have to begin by looking at this more fundamental question. Let’s agree that we’re going to talk about western classical music. We don’t have to, but since most of this book is about that it seems unnecessary for these purposes to decide what is music for all places and times. Nor do I intend to get bogged down arguing the difference between music and other sound. Since John Cage it has become impossible to be categorical about the distinction: so much depends on how one chooses to think about it. Even so, deciding what, or perhaps better, where music is—in the notation, or in performance, or in our minds—is not easy.

¶3 Naturally, we all feel that we know what music is; that is to say, we can think about music, we can imagine it, and we know it when we hear it; but can we explain what it is to anyone else? It’s surprisingly difficult to do: music theorists, philosophers and lexicographers have been trying for at least 2000 years. For a while, after about 1800, the problem became simpler, at least for those who studied what we now call classical music. For it was around then that people began to think of music not just as something one did but also as a collection of compositions that had the potential to live indefinitely as musical ‘works’ that, if they were good, would be studied, read, played and admired long into the future, even though their composers and original social contexts might be long since gone. Music, therefore, became something one could imagine as a body of work, one that was growing but one whose existing members had a stable form: one could go and find them, get them off the shelf, read them, think and write about them. Musical scores, which had previously been seen as simply a partial record of some aspects of composition, a convenient aid to music-making, became the image of music itself. ¹

¶4 It is easy to see how much this benefited composers. As their stock rose they became more like painters, sculptors and architects, makers of objects that would live indefinitely. Once one starts to think of a work of music as an object it naturally begins to need a more fixed form, and composers began to specify in notation details that had previously been left to performers to determine. The score gradually became authoritative, and ‘the music’ was to be found within it. Performances, as a consequence, were not so much the music any more, but rather were attempts to do the music justice: a performance could never be as good as the piece being performed. ² The performer did his best, but the work could never be as fully realised in sound as it had been in the composer’s imagination or (and this too was often said) as it could be in the imagination of a musically intelligent reader. ³

¶5 What saved performers from servitude was that, for all but the most expert readers, they remained absolutely necessary, and in fact the more composers demanded of them the more impressive they became. This in itself should have been enough, one might now think, for a reassessment of the relative importance of composer and performer; but of course it was also true that much of the credit for what performers achieved returned to the composer. Consequently, by the end of the nineteenth century, a new equilibrium was more or less achieved. Composers were the real geniuses, but performers could be spectacularly impressive showmen, using composers’ scores for displays of virtuosity and imagination that thrilled audiences every bit as much as the compositions they played. ⁴ It was an uneasy alliance, and one from which performers, in time, might still have broken away. But at this point something quite unexpected happened. In 1877, Thomas Edison showed that it was possible to reproduce sound.

¶6 It is difficult for us today to imagine how extraordinary that was: and yet, if one measures its significance by its novelty—the extent to which it introduced an ability inconceivable before—it seems one of the most remarkable things that has ever happened. A phenomenon that had previously been the most ephemeral of all suddenly became repeatable, indeed fixed: from now on, one could listen to the same sounds again and again. How this would affect the power relations between composers and performers would have been hard to predict, and in fact it took a long time for a trend to become obvious. In some respects performers benefited greatly. The stars became far better known and far better off. A page from the Victor royalty accounts for Caruso from June 1906 to June 1909 shows Caruso earning £9,922 0s 4d on sales of 67,785 discs, ⁵ and there can be little doubt that what he was singing was much less important to buyers than the fact that it was he who was singing it. Moreover, performers became in some respects very much more skilled. Recording discouraged mistakes; increased accuracy allowed composers to write more complex music; and so on. Recording is not the only factor in this, but any conservatoire teacher with a long memory will tell you today that standards among young musicians struggling to gain a foothold in the profession have never been so high. Much of this is a long-term result of the perfection increasingly demanded by the ever-greater accuracy of sound reproduction.

¶7 The downside to this must already be obvious. If accuracy comes first, spontaneity and originality are pushed into second place. As has been repeatedly claimed, recording seems to have encouraged conformity and predictability. ⁶ But is this really true? Hearing what is characteristic of the performance style of one’s own time is far more difficult than picking out the defining features of past styles, styles no longer inextricably bound up with our own way of seeing the world. One of the things that (I predict) is going to become increasingly obvious as time goes by—and also as we get better at studying performances—is that there is just as much variety of approach now as ever there was. What I think has happened, in addition to the unavoidable difficulty I’ve just mentioned of standing back from the performance style around one, is that editing, in removing the slips made by musicians (just as film has removed the verbal and action slips that characterise real life) has removed a ‘vital’ aspect of human musical performance. ⁷ And musicians, in learning to play as more nearly perfectly as recordings, have done the same. It’s not that performance is any less varied now, simply that it is more unblemished. It’s hard to argue with the complaint that national styles of performing have all but disappeared (though the benefits of nationalism in any form ought to be continually open to question). But in another important respect the rather gloomy view of current commentators has merit. Individuality, still prized in theory, has been discouraged in practice, partly for the reasons I’ve just mentioned, but also—and less defensibly—because for so many musicologists creativity is not the performer’s job: only the composer is considered to be a creator of music. The most accurate possible performance of the score is all that is required. None of this has been helped by developments in the academy. At just the same time that performance was becoming stricter, in the universities musicology was insisting that strict methodology could solve music-historical puzzles. By examining every bit of evidence from the past, and by thinking clearly about it, musicologists would be able to say what a composer had originally intended: they could specify the original notes he had written down, removing corruptions allowed to aggregate through carelessness or ill-advised interpretation; and through the new discipline of analysis they could reveal a work’s fundamental structure. It became ever clearer that the performer’s ‘duty’ was to reproduce the composer’s text and point up the music’s underlying form. ⁸

¶8 So performers were disciplined not only by recording but also by academia, and both played an ever-more important part in their training. It’s easy to see how, when its time came, the historically informed performance movement developed such a stranglehold on so much music-making. Here were historical facts to be respected, instruments that tended to produce cleaner sounds than before (a balance of harmonics favouring upper partials), a modern environment in which precision was valued higher than ever. What could result if not a sense that accuracy, faithfulness to the past, servitude of performer to composer, were not just desirable but morally right? Living under an ideology like this, music could not possibly be thought to reside anywhere but in the composer’s score.

¶9 Recording, therefore, has tended to reinforce rather than to undermine the notion that music exists in works that exist in scores that represent composers’ wishes that are to be observed. At least, it has reinforced that notion until recently. ⁹ But what I aim to show in this book is that to study recordings can lead us to conclusions that are quite the opposite. Music doesn’t exist in works, works don’t exist in scores, and neither does music, nor do scores represent composers’ wishes, nor should composers’ wishes necessarily be observed. ¹⁰ Fortunately, perhaps, all these conclusions have been or are being reached in other, perhaps more theoretically inclined branches of musicology; the point to be made here is that this is not simply a theoretical issue, but a conclusion that arises out of the way music has been practised and is perceived. How this is so will gradually become clear as we proceed.

¶10 Another route to this conclusion comes through philosophy, although it has to be said that not many philosophers of music have yet reached it. Nevertheless some account of what has been happening in the philosophy of music will be useful before we go any further. As will become apparent, most philosophers of music are not first and foremost musicians, and tend to see music as a philosophical category before they see it as a sensation or as the output of a set of musical skills. ¹¹ While from philosophy’s point of view that is as it should be, it remains generally the case (though there are notable exceptions) that the practicalities of music-making, and the realities of educated music perception, are not always very well understood. Their counterview, of course, is that musicians don’t think straight.

The philosophers’ view

¶11 The modern philosophy of musical works and performances, from musicology’s point of view, begins with Roman Ingarden, whose The Work of Music and the Problem of its Identity appeared in 1962 (and in English only as recently as 1986). ¹² Ingarden’s approach makes philosophical sense of traditional cultured understandings of the relationship between composer, performer and listener. For him, a performance ‘is not in any sense a constitutive element of the work’ any more than pigments constitute an element of a painting,[53] because the work is not its physical components but rather the thing that they make, not a physical object—whether paint on canvas or sound in time—but rather an ‘intentional object’, the piece as made in the composer’s mind at the point of its creation. The composer’s intentions for the work define its identity.[117] ‘[A]t the moment of its creation all its phases exist simultaneously’; only in performance are they taken apart and performed sequentially; thus the work does not fully exist in performance.[70, 65] Nor does it exist fully in score: the composer cannot specify all the work’s details because notation is not adequate for the task: ‘both the fixed and the open elements have been conceived by the composer as fully defined and fixed, but he does not command a musical notation that would do them justice.’[116] Consequently we cannot know which performance best represents the work as an ideal aesthetic object: we can’t know enough about the ideal work to know that.[143] ‘The composer’s artistic achievement is ... the creation of the work as a schema subject to musical notation ... that displays a variety of potential profiles’ through its performances.[157]

¶12 Ingarden’s notion of an intentional object is useful because it offers a way of dealing with the evident but conceptually very difficult facts—difficult because nothing else is quite like this—that a piece of music is 1) composed by an individual and appears to exist in some sense thereafter, 2) is something that can be discussed in detail independently of any performance, and 3) is recognisable and yet at the same time different whenever it is performed. In a nutshell the problem is that musical works exist without a fixed or fixable form. What the philosopher of music needs is a rationalisation for this that makes a work of music as like any other artwork as possible. Ingarden’s intentional object offers a way of doing this, but at the cost of a view of music that may not be fully grounded in historical or social reality. It is far from clear that composers have always, or even often, believed that their works were fully conceived by them, having a single ideal form, which they would specify fully if only notation were adequate to the task. There may have been a tendency towards such a belief increasing during the nineteenth and (in some quarters) the twentieth century, but as I’ve already suggested that is intimately bound up with a change towards a view of music as text rather than as practice, and with the particular construction of the notion of a musical work identified by Lydia Goehr. ¹³ It is socially contingent, in other words, and not a safe basis for a philosophy of music. By tying our view of a work to a composer’s intentions we are putting ourselves in a position in which we can only fail to perform the work. ¹⁴ The composer’s intentions can never be known sufficiently. And while it is certainly true that many performers pay lip-service to the idea that their aim is to realise the composer’s intentions, few if any believe that those intentions are to be discerned by asking or reading about the composer. On the contrary, for almost all performers a sense of the composer’s intentions emerges from time spent playing the work. What feels right comes to seem as if it must be what the composer intended; and the more right it feels the more inclined one is to believe that one has reached the heart of the work. This is all unproblematic and to be expected; we simply have to accept that there is an element of well-meant self-delusion in reading from one’s performance back to the composer’s wishes.

¶13 Just as relevant, in assessing the plausibility of a philosophy of music grounded in intentionality, is a realistic assessment of the audience’s experience. Who among an audience of listeners knows or cares about the composer’s wishes? Historians may well know something, and care a lot; but in being outraged by performances that ignore their knowledge and beliefs they are responding not to inadequacies in the musical experience, but rather to the performer’s making personal rather than historical judgements about what to play. Their outrage comes from a moral judgement about the performance, not a musical one. Most listeners in a typical audience, however, bring few moral criteria to bear on what they hear; for most, what matters is how the sound of the music makes them feel. And to that extent they are behaving much more like performers than like historians. Just as the performer places each note in relation to the last so that it feels right, so most listeners evaluate their experience of the performance according to the rightness of the sequence of sounds they have heard. Thus—returning to philosophies of music based on notions of intention—by shifting the identity of the work away from performance towards the composer’s imagination we place it somewhere where we can never find it, and somewhere where it has no power to account for the experience of music in reality.

¶14 Jerrold Levinson also identifies the work inextricably with the composer and the context in which it was composed, but at the same time accepts performance as essential to it. Thus a work is a performed sound structure specified by a composer at a given time. ¹⁵ What this means is that a performance that departed from the sound structure specified by the composer would not produce the same work. ¹⁶ Philosophically that seems unproblematic, but of course it bears only a partial relation to the experience of musicians. It is perfectly possible to change quite a lot of notes in a score, to play more than are indicated, or fewer, or other, and still produce a recognisable and persuasive performance that most listeners will find it sensible to consider as a performance of that work. They may be wrong from a philosophical point of view; but on the whole it seems reasonable to accept as reliable the judgement of experienced listeners. One could leave it at that and simply accept that musicians habitually think things about musical works that cannot be strictly true. But I want to argue that they are in fact correct, not simply because they ought to know best, though I think that a significant consideration, but also because incomplete specification is a normative feature of musical works, and it is by no means reasonable to draw a firm line between those details that the composer has chosen to specify and those she has not, or to argue that the first category of details is fundamentally different in kind from the second.

¶15 Let’s look at a real case. A composer notates three crotchets in a bar. A performer may play those durations with proportional lengths of 2:5:4, as in Chopin’s Mazurka Op. 24 no. 4 recorded by Ignacy Friedman in 1930. ¹⁷ There is no way of knowing Chopin’s intentions. If one’s interested in them, one may wish to know that Friedman danced mazurkas in Poland as a boy, so his views on rhythmic stress may have some value in terms of the work’s historical context (though Chopin wrote to his family of his mazurkas that ‘they are not written for dancing’). ¹⁸ We can’t know what Chopin imagined here. All we can say is, first, that Friedman is most definitely not playing three equal-length notes, or anything like them, and second, that the artistic results are highly satisfactory. Either the sound structure specified by the composer is not being performed, or—if it is to include what the composer wanted, regardless of what notational convention led him to write—we do not know what that sound structure was. It might have been 5:2:4, or 5:4:2, or anything else that a performer can make work according to the judgement of listeners. If the work is anything more than its notation, but other than a particular performance, we cannot know precisely what it is.

¶16 We accept relatively easily the idea that notes may be longer or shorter than specified. We’re less easy with the thought that the pitches might be different. Yet there are any number of examples on record of performers changing pitches deliberately—for instance doubling them at the octave to produce a richer or more brilliant effect, or filling in a diatonic scale with semitones to make it chromatic, or even making it a glissando. These examples come from a period (before the Second World War) in which such changes were entirely normal, and were regarded much as ornamentation was in the seventeenth and eighteenth centuries. For those performers they lie well within the identity boundaries of the musical work. They may or may not for us, and they might or might not have for the composer. These things change. Like so many things to do with music, views of the performer’s role in relation to the score are not fixed. ¹⁹ It seems to follow that statements about the identity of the musical work, if they are to be true beyond their own historical context, need to be so general as to be barely worth the bother of making. What performers do with scores is far more interesting and, indeed, more determinative of what music is.

¶17 It won’t be surprising, then, that I’m not much in sympathy with either Stephen Davies’s or Peter Kivy’s elegant attempts to make historically informed performance a necessary ingredient in the identity of the musical work. For them, as for so many musicologists, the historical informing is to be done by the composer’s intentions. For Davies, pinning down Levinson’s definition still further, ‘a musical work is a performed sound structure as made normative in a musico-historical setting’. ²⁰ Consequently, ‘Scores implicate the historical and social contexts in which they are generated, for the instructions they encode can be understood only by the person aware of the conventions by which they are to be read.’[159] From which it follows that ‘Because it is essentially implicated in a work’s performance [in other words, it’s not the work if this doesn’t happen], authenticity is an ontological requirement, not an interpretive option.’[207, his italics] Of course there are practical limits to this. Although logically he should, Davies doesn’t require us to reproduce the composer’s mental performance, or a performance of his time, since he recognises that that is now impossible; we simply have to do our best to be authentic. There are, nevertheless, some extraordinary let-out clauses which hint, beyond his willingness to compromise on grounds of practicality, at a philosophy driven more by preference than by consistency. We needn’t use boys in performing Shakespeare, he tells us, because the prohibition of women from the Elizabethan stage was a moral and not a theatrical judgement,[213] which is identical to the argument that we needn’t use boys in Bach cantatas, even though boys’ voices were surely integral to Bach’s conception of his church music, and it’s only one step away from the old argument that we needn’t use harpsichords because Bach would have used a Steinway had it been available.

¶18 Peter Kivy argues precisely that. ²¹ ‘[S]ince we want to perform their music now, we want to ask questions about what their performing intentions would be, given the conditions under which their music would now be performed, not under the conditions then prevailing.’[37] Our duty, in pursuing the ideal of performance authentic to the composer’s intentions is to reproduce in us the impressions received by his original audiences. An authentic performance of the opening chorus of the St Matthew Passion, therefore, is unlikely to be achieved using the forces of Bach’s time, since the impression they make now is less grand (Kivy proposing that for us musical grandeur is measured against performances of the Berlioz Requiem and Carmina Burana) than Bach intended.[53] I leave the reader to decide whether Bach might have responded to the opening of the St Matthew Passion in the way that we respond to Carmina Burana. To judge by his emphasis on the poor intonation of Bach-period flutes and oboes it looks as if at the time he wrote his book Kivy had not heard many decent performances using original instruments,[52] and one may be forgiven for suspecting that part of the motivation for his study is a desire to defend us all from the obligation to go on listening to them.

¶19 What I find most interesting about these arguments is that, despite the claims that historically informed performance is, one way or another, essential to the identity of the musical work, this idea seems never to have occurred in the philosophy of music until individual philosophers had heard historically informed performances. In other words, performance changed philosophy and not vice versa. It’s a neat illustration of the very evident fact that most music philosophy is nothing more than a reasoned justification for current musical taste. For us it’s another respect in which performance leads thinking about music, indeed, determines what music is thought to be.

¶20 Another common, and more interesting idea discussed by Kivy is that performances are a kind of arrangement of a work. He finds this argument relatively easy to make for baroque music, where the performer is intended to bring material of his own to the performance via his realisation of a figured bass or his ornamentation of a melodic line.[131] But by seeing the things performers do with later music in order to shape phrases (differences in note-lengths and so on) Kivy is able to argue that these personal interventions are akin to composition and thus amount to arrangements of the composer’s score.[132-8] This appears at first glance to offer a neat way of preserving intact the notion that the work resides in its composer’s intentions, while allowing that different performances are still performances of the work. But what does it really achieve, other than to permit—because we all do permit in practice—a certain amount of difference between an imagined work and its multiple performances? Is it anything more than another play on words intended to save the notion that the work exists independently of performances? We still can’t get at that imagined performance; it’s still lost for ever; it still doesn’t exist. Like Davies, Levinson, Ingarden and other recent philosophers of music, what Kivy cannot find a way to accept is that works exist only in performances. Why not? Perhaps because works, to them, seem too important to be accessible only through performers. If that’s why—and it certainly seems to be a subtext running through most writing on performance whether by philosophers or musicians—then it’s nothing more than intellectual snobbery of a sort that arises all too inevitably once one begins to think of musical works, in Goehr’s sense, as creations of genius that exist comparably and with comparable status to works of painting, building, and sculpture. The performer cannot be allowed to function as anything more than the artist’s loyal servant, preparing the colours, perhaps filling in the background, adding the varnish. But music is different, and it requires different conceptualising.

Philosophy and recordings

¶21 It may be easier to see, now, why recordings are so problematic for philosophy. The job of recordings is to sound as like performances as possible. Illusion, which philosophy tries so hard to remove, is actually part of the very nature of recorded music. Stan Godlovich, possibly following Fred Gaisberg, ²² but probably Theodor Adorno, ²³ compares recordings to photographs, traces or records of something that happened. ‘The temptation to neglect this’, he says, ‘likely stems from the stunning excellence of the resemblance.’ ²⁴ I’d like to leave to one side, for the moment, the question of how recordings differ from concert performances as a result of the kinds of things that happen in recording studios, because we’ll examine that in more detail later. But here we need just to consider, for a moment, the philosophical objections to recording in the light of what I’ve argued are philosophy’s unrealistic notions of musical performance. Are recordings really comparable to photographs? ²⁵ Not very directly. Photographs freeze one moment, and in such a way as to produce something never experienced in life. ²⁶ In life, time does not freeze. Video or film might have offered a better analogy. ²⁷ But it’s not at all clear that it helps us understand the relationship between recording and performance any more precisely. Film need be no truer to life than an audio recording to a concert performance, despite offering pictures along with the sound. Both recordings and films allow us to experience as a continuity a collection of takes that were performed for the recording machinery separately. One difference is that a film is much more obviously a confection. Watching a film we all understand that (save only in continuous recordings of live events) we are seeing only a selection of scenes chosen from a continuity that we imagine must have happened in real time but much of which would have been too uneventful to be presented in full with any kind of dramatic satisfaction. But in music we expect and wish to experience the whole, which has already been constructed in order to be dramatically satisfying without cuts. There are two kinds of editing going on, in other words: editing for drama, and editing for performance. A film normally uses both, a music recording only the second, but there can be all sorts of exceptions: opera is more like film; theatre more like music performance.; and there are always the exceptional cases, for example a recording of a musical performance so long and undifferentiated as to be editable for drama in any way you like without it making the slightest difference. (I have a recording of a BBC broadcast of Cage’s HPSCHD which begins by fading into a glorious cacophony over which the announcer says, ‘Welcome to the Round House where tonight’s Prom has already begun…’) In any case, I’m not sure that there is a great deal to learn from the comparison except that music recordings are typically much more like music performances than films are like life. Recordings of concert performances do everything, in fact, that concert performance does, except to provide a hall with performers and listeners working around one. One can, of course, argue that recordings provide the illusion of being performances but are truthfully something else. However, music is not concerned with truth. Music is about how it makes us feel.

¶22 Let me be completely clear about this. Music’s fundamental purpose is to generate a pleasurable response in the listener. Anything one might want to say about it has to follow from that. Therefore what it sounds and feels like is the most important thing about it. If recordings of music sound like performances that is all that is required of them.

¶23 By ‘pleasurable’, of course, I don’t just mean ‘easy to listen to’; I intend to include every kind of positive experience that music can generate. I want also, in glossing that last paragraph, to emphasise the importance of listening in performing and in reading scores. ²⁸ Performers are, unless on autopilot, the closest listeners of all. It’s impossible to place a note effectively, unless through habit, except by listening intently to what has just sounded and sensing how the next note will best follow from it. At the same time the performer takes a longer-range view, if only, in any detail, of the past few and the next few moments. Godlovitch is therefore wrong, I think, to insist that performance takes two, and that ‘In performance full and proper, the listener and player must be distinct.’[42] Performers communicate with themselves as listeners because it’s essential to performing. ²⁹ The performer playing to herself has an audience, and a very discriminating one. Reading a score one is, of course, performing it in one’s imagination: it’s impossible to read music otherwise. One can look at a score, observe patterns of notes on the page, but one cannot read it except in imagined sound. ³⁰ There is, in fact, no access to music except through performance. It follows that everything necessary for performance underlies and shapes our perception and understanding of any piece of music. The importance of this will become clearer in a moment.

¶24 So when Godlovitch claims that to place listeners on a par with composers and performers is ‘just crazy’,[45] while I’m happy to laugh with him I have to answer that it would be crazy not to. What Godlovitch has overlooked is that while listeners can be separated from composers and performers, listening cannot; so that what listeners do is what composers and performers do before they do anything else. It is through listening that music is made.

¶25 I hope it’s becoming clearer, now, why I think that the notion of the work as primary, and of performance and listening as secondary (or in the case of listening, even tertiary) is mistaken. The three are much more nearly identical than that. The work is something of which we have a sense when we listen to a performance. ³¹ The rest of the time it’s just an idea. A work isn’t shown to exist by virtue of our ability to talk about it. That we can say, ‘Beethoven’s 5th begins (the present tense implying its present existence) with three quavers followed by a crotchet’ doesn’t mean that it exists now, as we’re discussing it. It simply means that if we heard a performance, or imagined that passage, those are the durations we would hear (more or less). But they don’t exist out there, only in our minds and only when we think about them. It’s a mental construct, nothing more. Their existence in our minds depends, as we have seen, on performance, so that you can’t think anything very specific about a piece except within the context of a manner of performance. This is the crucial point. We imagine those four notes differently according to where and when we imagine them. In Berlin in the 1900s (conducted by Arthur Nikisch) they were drawn out, admonitory, almost a warning of what was to come; in London in 1955 (Klemperer) they were heavier, more portentous; by 1988 (Norrington) they had become urgent and brash. Their notation may have remained the same, but that notation is nothing more until converted into sound. And the sound that results is unavoidably shaped by performers’ tastes in performance style. It follows that period style and personal taste lie beneath, and play a necessary role in determining, the identity of the musical work. Recordings, in turn, give us access to one hundred years of changing tastes in performance style, and so of changing musical works. If we want to know what music is, they are of the greatest possible interest.

¶26 So, we need to reconceive performance not as an add-on, a necessary evil which the work survives only in part, but rather as integral to the identity of the work. The work, which began life only as what the composer imagined, is now the whole recoverable history of what performers have produced when they have performed it, ³² and of what listeners have experienced when they perceived those performances. Consequently when we come to consider recordings we are not considering something many stages removed from the work—concretisations (Ingarden) or traces (Godlovich) of instantiations (Levinson) or arrangements (Kivy) of musical works—but events during which we create the work in partnership with musicians from the past. If we want to know about the history of the work, they offer us the fullest evidence we shall ever have.

Further considerations

¶27 Before we go on to look at what recordings are, two supplementary points, hinted at above, need to be made explicit. The first has to do with music’s large-scale structure; the second with our obligation to the composer’s notation. Both arise out of this reassessment of the role of performance, now apparently an essential role, in all our thinking about pieces of music.

¶28 Music happens. It doesn’t exist as an object, but only as a process. A piece of music is thus not like a painting or—despite the attractive but misleading metaphor ‘architecture is frozen music’ ³³ —music is not liquid architecture, it doesn’t assemble a structure as it goes along, so that the completion of the piece leaves us with a magnificent building to sit and contemplate. It’s a lovely idea, and one that has had a whole approach to thinking about music (the analysis of form) built upon it, but it’s not a particularly appropriate metaphor. Music exists only as it happens, and each moment is gone as soon as it’s sounded. ³⁴ That is not to say that a performer has no concern for larger-scale form, though whether she does is a matter of choice. It is certainly not to say that composers pay no attention to the structure of longer spans than the few seconds around the ‘now’ moment of which performers and listeners are fairly fully aware. ³⁵ On the contrary, composers have always been concerned to find ways of organising large spans of musical time through patterned form; and the more so since musicology began to publish analyses of it. Quite how larger-scale form is perceived remains to be discovered. It is an article of faith among analysts that it can be perceived, and it seems reasonable to assume that composers thought so too. Presumably in some way it is. But we shall have to wait for experimental psychology to show how. ³⁶

¶29 Nevertheless, it is worth remembering, however much we may think of form as analysts, that most of our listening is concerned with the musical sounds happening just now. We assess them mainly in relation to what has just happened, and in expectation of what is coming next. It’s not hard to sense the relation of a phrase that is nearing completion to the phrase that came before: one remembers in enough detail at that level to have a reasonably good sense of local form: at any rate, one does with a bit of musical training. But how much any of us is aware of the span of a development in relation to that of the exposition, or of a movement in relation to the movement that preceded it, is another matter: clearly one is less aware the longer the spans of time. In any case, the important point is that the musical surface matters much more when one is listening to a piece than the musicological study of pieces leads us to believe. Most meaning is produced from moment to moment, by gestures in sound (whether composed or contributed by the performer) that take seconds, or less, rather than minutes. And there is therefore a case for an analytical view of music that pays far more attention to events at the surface than to longer-term patterning. This is something that the analysis of performance can do rather well, taking the performance as the music, and examining the kinds of things that performers do to give meaning to sounds. It’s one respect in which a musicology of performance tends to produce more listener-oriented findings, that is to say findings both derived from things heard and informative for listeners.

¶30 The second supplementary point to emphasise, concerning our obligation to the score, is more at odds with traditional beliefs. I’ve already stressed that we can have little knowledge of a composer’s intentions. Composers may speak about what they want, especially in relatively modern times when there is an audience interested in what they have to say—which was hardly the case before the nineteenth century. But words can convey little of how one wants a piece to be experienced, which is, after all, the most important thing. Anything one may specify, in a score or in words or by requiring certain instruments, can be turned into a wide range of different sounds by performers without any contradiction of those instructions. Sometimes composers can perform their own music; and some commentators, though by no means all, will maintain that these performances are definitive. Robert Philip has shown some of the problems that arise, especially when performers make more than one recording of a piece, and when, as happens frequently, they don’t follow their own written instructions. ³⁷ Whatever the composer’s intentions when he performs, the music will come out different, changed by his abilities and tastes as a performer (though actually what he writes will be shaped by at least his tastes in performance style, if not his abilities, so that may be less of an issue, except that deficiencies as a performer may prevent him from producing the performance style he assumed). So we have to be very wary of inferring anything special from composers’ performances: ³⁸ was she technically able to play the piece as she’d imagined it or as she’d like to hear it? Was she mentally able to? Did playing it lead her to change her conception, or to do things she’d not intended, things that made sense while she was playing but that she wouldn’t consider part of her conception of the piece, or that she wouldn’t think others need do? Performing isn’t like imagining-while-composing, and it leads to different results. Faced with the notation, then, the performer starts again from scratch, working not from the composer’s conception of the music but just from the notation, making a new conception which becomes the beginning of the process culminating in sounds perceived by listeners. The only constraint is that performers—for career and ideological reasons—tend to work well within the performing traditions for the piece in their time. Performers who don’t are very rare indeed (Glenn Gould is the example that always springs to mind). But in principle the performer starts afresh.

¶31 So once the composer has done his job, has written down the advice to performers—which is perhaps how we should now consider notation ³⁹ —and has passed it on, the performer takes over, in effect, as the source of a piece of music. The question then becomes, ‘How does this performer make the music happen?’. And as we know, there is a huge number of possible answers. An awful lot can be done differently by different performers and still leave us with something we recognise as a particular ‘work’. The work, as I’ve hinted already, is the collection of all possible recognisable performances. Beyond that there is no way of saying what is a particular piece and what is not. A lot can be changed and still leave the same work recognisable. It is impossible to say how much. With a well-known piece having a strong profile in one dimension (for example, the rhythms in parts of The Rite of Spring, or the scoring of the Prélude a l’après-midi d’un faune) it might be possible to change almost everything in other dimensions (for example, all the pitches) and still have some people able to recognise it. It would be interesting to do some experiments to find out; but whatever the results, it’s clear that the identity of the work is much looser than the score, or a text-based view of the work, suggests. Consequently, performances that change many details of the notation may still be acceptable performances of a work, provided only that they are musically satisfying. ⁴⁰ There seems no need to insist that the pitches, durations and so on be exactly those notated, since it is perfectly evident to anyone who listens that a wonderful performance can alter at least a proportion of these and still be wonderful. To offer just two examples from among thousands that might be cited, consider Paderewski at the end of Chopin’s Mazurka Op. 59 no. 2 where Chopin specifies no pedal, a rest before each pair of chords, and the chords played staccato; whereas Paderewski plays the exact opposite, with pedal and legato (Sound File 1) (wav file). ⁴¹ The result is musically entirely convincing. Or consider Harry Plunket Greene singing (in English) ‘The Hurdy-Gurdy Man’, Schubert’s ‘Der Leiermann’. Here the vocal durations are changed throughout, yet, though furthest from Schubert’s text, it is surely one of the most moving performances of this piece ever recorded (Sound File 2) (wav file). ⁴²

¶32 Performance is a collaboration between composer and performer, a negotiation between what the composer put down and what the performer wants to make of it. Performer and composer are therefore on a much more equal footing than musicology has supposed, which is why the equal billing on advertisements and discs, often decried by critics, is actually quite appropriate. As the philosopher R G Collingwood wrote as long ago as 1938, ‘Every performer is the co-author of the work he performs’. ⁴³ It helps to think of both composing and performing as improvisations: composer and performer both improvise a performance, deciding what happens next in the light of how what they just did turned out. ⁴⁴ There are always alternatives open to them, and one can't assume that what they decide on is the only good solution, even if it’s the best (which one can’t be certain of). Anyhow, what gets notated is so partial a representation of something that could so easily have been different, and of something that will be different in important ways every time it’s performed, that one can’t sensibly see the notation as having the force of unalterable law. We must ask ourselves, therefore, how real is the obligation so many have always supposed that we have to stick to the letter of the score.

¶33 The other common constraint on performers, as I’ve suggested, is recent performing tradition. But I think it follows from what’s been said here that a decision to work within that is one of convenience—easing career advancement—or a decision made on principle, a moral decision, nothing to do with sound. So I’m not sure that a strong case can be made, on musical grounds, for sticking to any particular traditions of performance any more than to the notation. The only test is how well something sounds. There can’t be any absolute judgement as to what is best, because it’s always possible that another performance will be better, better for the moment at which it happens. The composer’s instructions may be made to work well, but so may others. What matters, ultimately, is how the brain processes sounds, not which sounds are the proper ones. ⁴⁵

2.2. Musical performances

¶34 In looking at musical works we’ve already come to a number of conclusions about performances, and I’ve argued that the two phenomena are much more nearly identical than has traditionally been supposed. What remains to be said about performances in general, then, can take us in a new direction, away from the notated text and towards perception.

¶35 In order to understand what performance does we need to understand something about how we perceive music through our ears and brains. This will be a recurring theme of the book, so for now some general points will suffice. The way our brains process musical sounds is in many (but not all) respects the same as the way they process any other sounds, including speech. ⁴⁶ Incoming sounds are analysed by the ear (the basilar membrane) into their component frequency bands, and signals are sent to the brain representing each frequency that’s present and its amplitude. The ear does a spectrum analysis, in other words, and sends the resulting data on to the brain. The brain’s job, then, is to decide which of these frequencies belong together and to construct a representation of the sound in the mind and to make a best guess at its source. ⁴⁷ It does this very quickly, simply because animals that identified sounds fastest and most accurately lived longer and reproduced more, passing on their ability. So when we listen to music we can construct in our minds a representation of the sound sources and follow the progress of a sequence of related sounds in real time with ease; and that in turn leaves us with spare mental capacity to think about how these sounds relate and the patterns that they form. Within this process the earliest reaction to incoming sounds is automatic, with very fast onset, and causes involuntary changes in the autonomic nervous system. ⁴⁸ These changes we feel as an emotional response. While the evolutionary purpose of these responses may be to cause us to fight or flee, we don’t need to make that choice while listening to music (at least, not that quickly). But precisely because it is (and has to be) an automatic response we can’t simply switch it off. It works when we hear music just as it does when we hear any kind of sound. ⁴⁹ What this means is that when we listen to music, we respond to its sounds emotionally (pre-cortically) before we have time to think about it in any detail. ⁵⁰ As the incoming sound data passes on to the cortex, however, we are able to respond to it in more complex ways, to consider, for example, the contexts of sounds and the structure of longer-term sequences. ⁵¹ Emotion before form, in order words. ⁵² It follows that the emotional content of a sound is absolutely fundamental to meaningful perception of it. Performance, making sounds that make sense to us, therefore inevitably deals heavily with emotional content.

¶36 This helps to give a useful answer to the question, What do performers do? Essentially two things: 1) they make audible aspects of musical structure; 2) they give it emotional force through expressivity. ⁵³ In other words, they are delineating the structure of music with the shapes of (the changes of feeling involved in) emotional experience, reinforcing the cognitive appraisal of musical sounds with the shaped feelings that for listeners momentarily precede that. To put it another way, what the listener experiences is a feeling (pre-cognitive) followed almost instantly by the explanation for it (cognitive); what the performer does is to emphasise the explanation by applying a sounding representation of appropriate feeling to it. ⁵⁴

¶37 The relationship between making structure audible and giving it emotional force seems on the face of it clear: performers make structure audible by giving it emotional force through expressivity; but it’s also likely that those events that are given emotional force tend to be perceived as important in terms of musical structure. Thus while every performer today will agree that gross aspects of local musical structure (phrases, cadences, harmonic ebb and flow) require some expressive response unless a performance is to seem mechanical, it’s also the case that performers can draw attention to some moments and away from others so as to construct an individual reading of a score. That there is such widespread agreement as to which moments have to be expressively shaped may suggest that composition is more determinative of perceived structure than is expressivity. And one might go on to suggest that, historically speaking, habits of expressivity therefore developed in order to point up details of compositions. But in music-evolutionary terms it seems highly likely that musical form developed its particular characteristics (arch-shape lines, dissonance between consonances, cadences) precisely because those features cross-domain mapped easily between sound and emotional response. In other words, composers found ways of composing out emotional responses to sound which performers intensify by shaping the composition as they play.

¶38 That expressivity works just as well without much in the way of compositional structure is suggested by the results of a recent study in which participants improvised the expression of different emotions using a single note. ⁵⁵ Musicians and non-musicians generated similar musical effects, and (other) musicians and non-musicians agreed substantially on what was being expressed. The study needs to be replicated among members of a variety of musical cultures, but if the findings prove robust they will strongly suggest that expressive performance depends on associations between sound and feeling that long pre-date musical composition.

¶39 Patrik Juslin (2003) has helpfully defined musical performance expression as a multi-dimensional phenomenon consisting of five components, the first two of which are those we’re discussing at the moment: ⁵⁶

¶40 Juslin’s remaining three components are to do with the means of expression, and as we go on to look at the realisation of the first two in performances their role will become clearer.

¶41 Juslin proposes that it will become possible to distinguish between these through identifying different kinds of patterns for each and different neural networks involved in their processing. The latter is highly probable, and may be enough to explain our ability to make these distinctions. Separation in sound patterns seems less intuitively likely. Each component is signaled by the same means―namely, fluctuations in whichever of the three dimensions of sound are available to the performer via their instrument―and consequently all five are intimately bound up together in the perceived qualities of the music.

¶42 At any rate, making musical structure audible and giving it emotional force through expressivity are very closely related, without seeming identical: expression signals structure, certainly, but it’s not just a sign of structure; it’s also emotionally affective. If they’re identical, then expression cannot operate without structure, yet we've just seen that it needs none in the sense in which a composer would understand the concept. Similarly, if they're identical then structure cannot be separable from expression, yet we always assume it is. In fact, the problem we have in separating out these two things, and in being clear about how they differ, if at all, arises precisely from this: we insist that the musical substance—notes arranged so as to form a structure—is separate and (this is the crucial point) has expressive content in itself that is separate from its expression in performance. I’ve already suggested that that is a mistake, an understandable one that arises out of our need to simplify the world and to examine one aspect of a problem at a time, but nevertheless a crippling mistake that prevents us from seeing music for what it is, something whose substance cannot be conceived or perceived except as expressive performance, that is, as meaningful sound. Nevertheless, and aware now of the artificiality of the procedure, let’s explore these two functions of performance in turn.

Making structure audible

¶43 As shown so effectively by Gestalt psychology, we try to simplify the world, so as to make it easier to perceive, by organising things and grouping events together in our minds. ⁵⁷ This is true at a micro-level (recognising the integrity of things: sounds, objects) and at the macro-level (grouping ideas into subjects: history, psychiatry). The same happens in perceiving music, most obviously at the phrase level, but also (less perceptibly, I’ve suggested above) at a larger formal level and at a note-to-note streaming level. And composers are doing the same thing as they put notes together in their minds and their scores. In listening we understand this partly according to natural laws of grouping (to the extent that there are any), partly (much more) according to our understanding of musical components and their relatedness through form and continuity (obviously greater for expert listeners), and partly through what performers do to make these kinds of groupings audible (which we shall examine much more closely later on).

¶44 Evolution has selected the way we group sounds as being the most conducive to survival. It assumes that sounds close in time or in content are more likely to belong together and to represent the same source, because in nature sounds do not change haphazardly. ⁵⁸ This is partly why timing is so important in performance: precisely how two sounds relate in time determines precisely how we feel about them. Our perception of sounds earlier in the evolutionary process developed using timing as information with survival value, information that required a very fast, almost instinctive response. Consequently timing differences mean a lot to us, and they mean it at a fast-reacting emotional level. In addition, we call on our memory of past correct identifications. The two processes together are sufficient to explain why and how we group sounds in music and in musical performance.

¶45 Composers work with this, but also try to fool the brain into hearing as related things that outside a musical context it would identify as separate; what Bregman calls ‘chimeras’. ⁵⁹ Performers then try to overcome this by separating out the strands to which they wish to call particular attention, which in the case of soloists performing with an ensemble is likely to be their own part. Thus when singers, for example, or pianists emphasising one or other strand, come in early, or alter the start of a note to draw attention to it, or sing fractionally sharp, or modify their spectrum, they’re in a sense trying to escape from that ‘collaborative intention’ and be perceived as separate—and thus prominent—again. A literal reading of the notation tries to insist upon that collective hearing; while a text-led view of Lieder almost requires that the singer try to break away and dominate. ⁶⁰

¶46 So, ‘what is a performance’ can be a philosophical question if we want it to be, but as a practical question it must be answered in relation to the way the brain processes sound, and how music works with that. It’s about grouping—how sounds are grouped or separated in frequency, amplitude, and time (and especially time)—and it’s about what sounds signal, what they represent, what they evoke in our minds, especially in our memories, which means that their meanings depend on learning, culture, experience; but the process is physiological and inherited.

Expressing structure

¶47 But none of this is perceptible except through real or imagined performance (performance in either case). A clue to what this means, and how it works, is given by the impossibility of a human performer giving an inexpressive performance of a score. Because, thanks to MIDI, we now know what an inexpressive performance is like, we know that no musician can achieve one. (Sound File 3) (wav file) ⁶¹ Two things prevent it: first it is counter-intuitive to such an extent that we can’t do it any more than we can speak on an unvarying pitch and at an unvarying volume; secondly, we (even the most expert) don’t have the physical control that would make it possible. Whether these two things are related deserves investigation. (Is expressivity through irregularity in any way related to the imprecision of our motor control, a virtue arising out of a necessity? Presumably not, since it’s fine motor control that allows us to decide how to be expressive with our bodies. But equally we proverbially recognise imperfection as an index of humanity.)

¶48 Similarly, it is impossible to read a score without responding to it on an emotional level, without becoming engaged with the musical material and feeling something of the excitement (or whatever) that the notes were arranged in order to evoke. And that is an experience of performance, as is any response to the sound of music, wherever that sound may be, whether in one’s head or coming in through one’s ears, whether produced by oneself or by somebody else. That being so, music cannot be separated from a response to it. It may be convenient to think of it as having some existence on the page or as a concept, but as we’ve already seen, that is not music: it’s notation or an idea. Music by definition sounds. And sounds evoke responses. Music and response are part of the same system.

¶49 In that system, each listener is involved differently. Not very differently, within a single cultural group, but differently enough for opinions about the music to differ. And ‘the music’ here cannot mean just ‘the score’; it has also to mean a performance of the score. This raises interesting questions about the relationship between changes in performance styles and composition styles. If the notes and their performance are inextricably bound up together, then must not the same be true for composition styles and styles of playing? And yet we know from a century of recorded evidence that they are not inextricable. It’s possible that the belief that there is some special relationship between the notes and the style of performance imagined by their composer is a belief based in the sense that somehow music and performance do have the intimate relationship I’m suggesting here. But it’s also obvious that neither musical substance nor musical performance is fixed in any necessary way; in other words, you can change some of the notes, and you can change the way you play the notes, and still have a workable, convincing result. Thus while music and performance may be inextricably bound up, specific styles of pieces and specific manners of performing are not: at that level, the level of individual instances, things are much more fluid.

¶50 Does music need performers to make it? The historical development from live performance to recorded performance to computer generated performance is thought-provoking. We’ve moved fairly comfortably from stage 1 to stage 2, and it seems more than likely that in due course we’ll move comfortably into stage 3.This in turn will put philosophical arguments about intention into a new light. Imagine a performance that was a thrilling musical performance, unexpected but convincing, deeply involving. Would it make it any less involving to know that it was made by a computer program? Surely not, unless we deliberately intend to be disappointed. If we accept that what matters is how it affects us, then there’s no problem. The program originated in work by humans, of course, though it may produce results they never imagined. But regardless of the source of this wonderful performance, its qualities are the measure of its value. ⁶²

¶51 This helps to explain why the notion of ‘musical’ also works on different levels. Fundamentally, musicality is the ability to make a convincing relationship (convincingly expressive of some meaning), expressed in notes composed or played, between notes and performance. But exactly how that relationship is made is subject to an enormous amount of variation. And so what is musical changes over time, because at this more detailed level things are much more fluid. To understand what music is we have to understand something of that variety and of the ways in which styles of musicality change. That’s why studying performances on record is so important.