I

I begin with Nelson Goodman, whose Languages of Art did much to spark recent enthusiasm for the project that I have mentioned. His basic idea, it will be recalled, is that a musical work is ‘defined’ by the score in which it is notated: ‘complete compliance with the score’, he says, ‘is the only requirement for a genuine instance of the work’, so that even ‘the most miserable performance without actual mistakes [counts] as such an instance, while the most brilliant performance with one wrong note does not’.Footnote ² Goodman's reason for making this hugely counter-intuitive claim is that, if we allow the brilliant performance with one wrong note to count as a genuine or legitimate instance of the work, then, by transitivity of identity – through a succession of further wrong notes – we can get all the way to the thought that ‘Three Blind Mice’ counts as a genuine or legitimate instance of, say, Beethoven's Fifth Symphony; which is, as he remarks, absurd.Footnote ³

To most of Goodman's readers, however, the absurdity of counting ‘Three Blind Mice’ as a legitimate instance of Beethoven's Fifth has not seemed self-evidently greater than that of refusing to count, say, Furtwängler's 1937 performance of it, fluffed horn-notes and all, as such an instance; and so the tendency, post-Goodman, has been to try to build enough wiggle-room into the initial identity conditions to allow the latter, but not the former, to count as a genuine instance of Beethoven's work.Footnote ⁴ (I'll return to the issue of wiggle-room shortly.) The tendency, then, has been to accept that while Goodman's position does perhaps state a sufficient condition of a performance's being a genuine instance of a work – i.e. that it be in complete compliance with the score – it does not state a necessary condition. And, for the time being, I think we should go along with this.

II

Before moving on, though, I want to take a closer look at Goodman's position. The first thing to notice about it is how admirably explicit he is in distinguishing between identity-related issues – or, as one might call them, ontological issues – and evaluative issues. A performance can be a fully legitimate instance of a work despite being ‘the most miserable’ imaginable; whereas, on the other side, a performance can be altogether ‘brilliant’ while failing to be such an instance. This insistence, that ontological and evaluative issues can be treated completely separately, is an article of faith that Goodman's successors have endorsed, perhaps for the good- or obvious-seeming reason that one wouldn't, with one's ontological hat on, want to appear to be ruling putative instances of works in or out merely because one admired or thought ill of them.

The second thing to notice is how asymmetrical Goodman's apparently very symmetrical formulation is likely to strike one as being. The ‘most miserable performance without actual mistakes’, he says, counts as a genuine instance of a work, ‘while the most brilliant performance with one wrong note does not’. Purely intuitively, it seems to me, we are likely to accept the first half of this without demur. We are all familiar with workaday, competent renditions of works – performances that linger in the memory not one jot – while never once having had it occur to us that they might not in fact be performances or instances of the works in question. (Our reasons, here, are closely related to the ones that we have for not resisting the application of the term ‘music lessons’ to what are given in most primary schools, despite the recognition that nothing especially musical is likely to feature in or result from them.) Whereas the second half of Goodman's formulation is apt to strike us as outrageous. Surely, we will protest, if anything is a legitimate instance of Beethoven's Fifth Symphony then Furtwängler's 1937 performance is, the odd wrong note notwithstanding. It's quite brilliant! How can the occasional stumble affect that?! And so on.

But we don't need to appeal to intuition in order to grasp an asymmetry here. For there is a conceptual asymmetry in Goodman's formulation too. The first half of it – the half about miserable performances – appears to make perfect sense. There seems to be nothing problematic in understanding the claim that an uninspired and uninspiring, but accurate, performance of Beethoven's Fifth counts as a performance, as a genuine instance, of it, however drab. But the second half – that ‘the most brilliant performance with one wrong note’ does not count as such an instance – is far thornier; is, indeed, or so I suggest, strictly unintelligible. We cannot, that is, attach any real sense to the claim that Furtwängler's 1937 performance doesn't count as a genuine instance of Beethoven's Fifth Symphony, as Goodman would have us do. And this is because we cannot say what it counts as instead.

Perhaps, it might be suggested, it counts as an illegitimate instance of the work. But, if so, Goodman's initial problem simply resurfaces: for, by transitivity of identity, we can get all the way from there to the thought that ‘Three Blind Mice’ is an illegitimate instance of Beethoven's Fifth; which is absurd, since ‘Three Blind Mice’ is no sort of instance of that work at all. So it appears that Furtwängler's 1937 performance can be regarded as neither a legitimate nor an illegitimate instance of Beethoven's Fifth, since either alternative would license, it seems, counting ‘Three Blind Mice’ as an instance of that work too. It appears, that is, that the Furtwängler cannot be any kind of instance of that work. But then in what sense might it be said to be a ‘brilliant performance’? Of what?

There would seem to be two possibilities here. One is that Furtwängler's performance is a brilliant performance of something else – of a work that is very like, but is distinct from, Beethoven's Fifth Symphony. But this seems wildly implausible. On Goodman's schema, after all, it would appear to be a condition of a performance's counting as a legitimate instance of a work that it be produced in (complete) compliance with the score that defines that work. Yet the score that Furtwängler and his orchestra were attempting to comply with was that of Beethoven's Fifth Symphony, not of some other work similar to it.

We might, perhaps, meet this objection by claiming that a performance can be of a work without having been meant to be a performance of it. So here, we might say, Furtwängler and the Berlin Philharmonic, although intending to comply with the score that defines Beethoven's Fifth, actually produced a performance that conformed with another, closely similar score – and so was in fact a performance of that instead. But we must resist this suggestion, for at least two reasons. The first is rather Kantian in flavour. Just as Kant insisted that, in order to be moral, an action must be done out of respect for – or, as we might say, in compliance with – and not merely in conformity with, the moral law, so we should insist that, in order to be recognizably a performance at all, a given sound-event must be produced in compliance with, and not merely in conformity with, a score. For, in its ordinary sense, ‘performance’ is an intentional notion – a fact that Goodman's use of the term ‘compliance’ appropriately picks out; and we should be reluctant, in the absence of compelling reasons to the contrary, to abandon that sense, and to allow that performances can be given by accident.Footnote ⁵ The second reason to resist the present suggestion is that, if we were to grant that what Furtwängler's performance is a performance of is a work that is very like, but is distinct from, Beethoven's Fifth, we could no longer make sense of the idea that it contains ‘wrong note[s]’. For, by hypothesis, the score that Furtwängler's performance in fact conforms to is perfectly realized in the sound-event that he and the Berlin Philharmonic produced.

The other possible way out might be to suggest that the term ‘performance’ should be understood here in one of its other ordinary senses, perhaps as meaning, simply, ‘something done’ – as when we speak of agents ‘performing’ actions – or as meaning ‘something done showily’ – as when we might describe someone's way of leaving the room as a ‘performance’. Construed in one or the other of these ways, we might say that Furtwängler's 1937 performance was a performance in the sense, merely, that it was something done, perhaps showily, with an orchestra – maybe indeed something brilliant (certainly the sounds that we hear are thrilling enough), but not to be understood in terms of its compliance or non-compliance, or even its conformity or non-conformity, with any score. But this is unconvincing. First, it breaks the rule of charity that states that agents should, wherever possible, be understood as in fact doing what they take themselves to be doing – and there is no question here but that Furtwängler and his orchestra took themselves to be performing Beethoven's Fifth Symphony, and, indeed, that they took themselves not to be, merely, making orchestral noises, showily or otherwise. And, second, it makes it wholly obscure what might be meant by the idea that their ‘performance’ contains wrong notes. Wrong by what standards? Why pick out just these notes as mistakes? I can see no way of answering such questions.

The upshot of the foregoing is that there really is no plausible sense that can be attached to Goodman's invocation of the ‘brilliant performance with one wrong note’: either the performance turns out not really to be a performance, brilliant or otherwise, or else its notes turn out not to be wrong. So there is, I think, no properly strict interpretation of Goodman's formulation that makes sense of it – and it does need to be strict, if we are not, by transitivity of identity, to find ourselves in ‘Three Blind Mice’ territory. Which suggests that there is something deeply the matter with what he is trying to do. I shall try to say presently what I think that is.

III

As a first step in that direction, let's look more closely at the idea that legitimate performances are those produced in complete compliance with a score. This idea is, in one obvious sense, very exacting – a feature of it that, as we have seen, leaves Goodman with nothing sensible to say about whatever it was that Furtwängler and his orchestra brought off in 1937.Footnote ⁶ But it should also be noted that, in another sense, the idea is actually quite accommodating, for it allows that a potentially indefinitely large number of readily distinguishable sound-events should all count as legitimate instances of a work. In part, we can see this from Goodman's own almost gleeful admission that ‘the most miserable performance’ might, according to him, count as such an instance: for so, evidently, might a less miserable performance, or an average performance, or indeed a brilliant one, so long as none of them contains ‘actual mistakes’; and these differences in quality are respects in which the relevant sound-events differ from and are distinguishable from one another. But it can also be seen from the familiar observation that scores underdetermine performances.Footnote ⁷

Depending on the style and period of the music in question, scores prescribe more or less closely what a performer is to do. Some scores (from, e.g., the early baroque) give only a figured bass, leaving it to the performer to decide what melodic material to place on top; some scores specify the instruments to be used, while others do not; some scores contain detailed prescriptions concerning dynamics (Alban Berg's, notoriously, distinguish f not only from ff but also from fff and ffff), while others contain none; some scores are very specific about tempo, and changes in tempo, while others say nothing; and so on. But no matter how prescriptive a particular score might be, and along however many dimensions, there will always be some latitude, and so always an indefinitely wide range of ways of complying with it. So, for example, your performance of Berg's piano sonata might, at its loudest, be much louder than mine, and someone else's might be louder still. But so long as we are all playing our ffffs louder than our fffs, and are playing both of these louder than our ffs (let alone our mere fs), we are all complying with the dynamics that the score prescribes.

It follows from this that Goodman's account, although it demands that a performance be completely compliant with the score, also permits much latitude, whether that latitude is taken voluntarily – as by a performer determined to play his pianissimos as softly as possible – or not – as by someone whose andante is at the slow end simply because he cannot play the piece any faster. Thus Goodman construes scores as sets of instructions, where those instructions underdetermine the performances that comply with them. And, to this extent at least, what he says seems unproblematic.

IV

Things only become problematic, it seems, when a performer exceeds the latitude permitted by Goodman's demand for complete compliance with the score – when the performance includes ‘actual mistakes’ or the odd ‘wrong note’.

Goodman speaks as if all fallings short of complete compliance consisted either in playing non-prescribed notes by accident or in accidentally failing to play the notes prescribed. And of course many departures from a score have precisely this character – for example, Furtwängler's 1937 performance of Beethoven's Fifth: the fluffed horn-notes are clearly both non-prescribed and inadvertently produced. But, just as complete compliance with a score can be achieved by deliberate as well as by non-deliberate occupation of the latitude that a score permits, so departures from a score can be deliberate as well as not. Such departures can, moreover, involve many more things than playing the wrong notes or failing to play the right ones. Depending on the period and style of the work in question, scores can prescribe – and hence departures from them are possible – along at least six different dimensions: these are, crudely, the dimensions of melody (or pitch), harmony, rhythm, dynamics, tempo and timbre. So we must suppose that, for Goodman, departures along any of these, whether inadvertent or otherwise, count as non-compliance: the second half of his formulation, that is, should state – albeit less punchily – that ‘the most brilliant performance with one departure from the score’ does not count as a legitimate instance of a work.

We have been assuming, for the sake of argument at least, that Goodman's demand for complete compliance with a score represents at most a sufficient – but not a necessary – condition of work-identity. And its non-necessity is shown, we have been assuming, by a performance such as Furtwängler's, which is ‘brilliant’ but contains notes not prescribed by the score. Furtwängler's performance departs, inadvertently, from Beethoven's score along the melodic dimension, the one dimension that Goodman explicitly acknowledges. And we can probably agree with Goodman that, even if this does not disqualify it as a performance of Beethoven's Fifth, it does constitute a blemish in it. We can probably agree, that is, that Furtwängler's performance would have been in some sense better had it been in complete compliance with the score – had the brass section not had its dodgy moments. For those moments, we can probably conclude, add nothing to the performance that we should actively prize.

There are other cases, however, in which we might want to say something different – even when these involve, as in the Furtwängler case, nothing but non-deliberate departures from the melodic dimension of what a score prescribes. I have in mind examples such as Sviatoslav Richter's 1960 performance of the Liszt piano sonata, in which, as a more or less inevitable side-effect of the driven, demonic quality of its delivery, the occasional wrong note features; or Jacqueline du Pré's accidental pizzicato in the second of the big ascending scales in the first movement of Elgar's cello concerto (the version conducted by Barbirolli). In these cases, it seems to me, we might well prize the performers' departures from the score, resulting from and testifying to, as both do, the quite extraordinary intensity of their playing – and, in the Richter case at least, testifying also to the work's ultra-virtuousic demands coupled with his own determination to take no prisoners in meeting them. We might not want to say that these performances are better for departing from the relevant scores in their respective ways; but nor, it seems to me, would we have much inclination to say that either performance would have been better for not departing from the score in the way that it does.

Two things are worth noting about examples of this kind. The first is that they strengthen the assumption that Goodman's identity-condition is at most sufficient, rather than necessary. For, if what I have said is plausible, it seems that some departures from a score are not even to be regarded as blemishes in the performances that include them – i.e. are not to be regarded as moments that, unlike those provided by Furtwängler's errant horns, we might, ceteris paribus, wish were otherwise.Footnote ⁸ The second thing to note is that what I have just said is true only on the assumption that the departures in question are accidental. If we were to find, for instance, that Du Pré had planned her very distinctive glitch, or had decided, once she had inadvertently made it, to include it on future occasions, we would regard the performances in question not as thrillingly intense, but as contrived, as merely affecting intensity, and would think the worse of them for it.

V

But now consider another sort of case, in which non-compliance with the score is perfectly deliberate. A good example can be found in performances of the closing bars of the passacaglia finale of Brahms's Fourth Symphony, whose score indicates no deviation at this point from the tempo already prescribed. Brahms could have indicated such a deviation: he was fully cognizant of accelerandos and rallentandos, and was not averse to using them. So the absence from his score of any instruction to speed up or to slow down is, in effect, the instruction to do neither, but to play these closing bars in tempo. And yet: two of the most celebrated performances that have been given of Brahms's symphony depart, and depart entirely intentionally, from the score in just this respect. The ones that I have in mind are Toscanini's 1935 performance, which decelerates markedly in its final bars, and Furtwängler's 1943 performance, which does the exact opposite. And both are brilliant.

It is quite difficult, from a perspective at all like Goodman's, to make sense of this. One can imagine that a follower of Goodman might attempt to accommodate the Richter and Du Pré examples by saying that the intention to comply completely with a score is a condition of the resultant performance's counting as a legitimate instance of a work: this would be one way of trying to build in the sort of ‘wiggle room’ that I mentioned earlier. But that move clearly cannot work here. For it is quite evident that what the Toscanini and Furtwängler performances evince is a fully premeditated intention to depart from Brahms's score, to not comply with it.

One possibility here would be to suggest that the two conductors were proposing revisions to Brahms's score, or even corrections to it. But neither suggestion is attractive. The latter implies that Brahms's Fourth Symphony is something independently and antecedently existent which Brahms, in his score, failed to transcribe accurately: and even if one were in the grip of the sort of metaphysical picture that would make this suggestion so much as passingly plausible – as no one, surely, should be – it cannot account for both performances' being corrections to Brahms's score, since they ‘correct’ it in mutually exclusive ways. Nor is the other suggestion, that Toscanini and Furtwängler were proposing revisions to Brahms's score, any better. Its effect is just to multiply the number of works at issue, so that we have Brahms's Fourth Symphony, Brahms's Fourth-as-revised-by-Toscanini and Brahms's Fourth-as-revised-by-Furtwängler. Neither conductor, on this conception, is actually performing Brahms's Fourth Symphony – although both took themselves to be doing so; and neither is departing in any way from the score of the work that, in fact, he is performing, since, by hypothesis, the score of Brahms's Fourth-as-revised-by-Toscanini prescribes a deceleration in the final bars, while that of Brahms's Fourth-as-revised-by-Furtwängler prescribes the reverse.

Another possibility would be to draw on the distinction that is sometimes made between interpretation and performance.Footnote ⁹ According to this, an interpretation is a way of performing a work, so that many different performances can count as instances of the same interpretation. Here, then, we can think of Toscanini as performing his interpretation of Brahms's Fourth and of Furtwängler as performing his. This has the advantages of introducing nothing metaphysically extravagant; of construing Toscanini and Furtwängler as doing nothing at odds with what they took themselves to be doing; and of keeping the number of works at issue down to one – Brahms's Fourth Symphony – which is where it should be. But it secures these advantages only at the cost of not moving anything on. For it just reintroduces the original problem at the level of interpretation rather than performance. How are we to make sense of the idea that an interpretation which deliberately departs from the score of the work that it interprets is, still, an interpretation of that work? Or, to put things in a more overtly Goodman-like way, how are we to block the move which might, by transitivity of identity, take us all the way from the thought that Toscanini's interpretation of Brahms's Fourth, say, is a legitimate instance of it to the thought that so-and-so's interpretation of Brahms's Fourth, a performance of which is identical to ‘Three Blind Mice’, is also a legitimate instance of it?

VI

But these suggestions – that we're dealing with revisions, or with corrections, or with interpretations – are actually only red herrings, prompted by our willingness, or so I'll suggest shortly, to follow Goodman even as far as we have done. The question we should really be asking is what we should say about Toscanini's and Furtwängler's performances of Brahms's Fourth if we set Goodman aside entirely for the moment, and wear only our listeners' hats.

We would say, I think, that both are brilliant, that both depart from the score, albeit in opposite ways – but, also, that both depart from the score to very similar effect. For both bring out the fact that Brahms's symphony ends tragically, that it ends in a way that is both fated and fateful. In the Toscanini, the effect is like that of watching a train hurtle into a rock-face; in the Furtwängler, the hurtling is also headlong, but is into the abyss. In both we hear the catastrophe. And this, I think, would make us want to say that these performances are just better than any performance which, while wholly compliant with the score, did not allow us to hear this. Indeed, for what it's worth, I have never heard the close of the passacaglia played to such devastating effect in any performance that does stay in tempo. Barbirolli's 1967 performance (with the Vienna Philharmonic) comes close; so, too, perhaps, does Celibidache's 1959 Milan performance: both have the right air of ineluctability about them. But neither conductor, to my ear at least, does full justice to the ending of Brahms's symphony, as Toscanini and Furtwängler do.

What this brings out is the unsurprising fact that there is more to a musical work than can be indicated in a score. And this, in turn, brings out the inadequacy of saying, as I did earlier, that scores underdetermine the performances that comply with them. For, if we take seriously the idea that a score is a set of instructions, we should at once note that, in the case of almost all sets of instructions, there is a point to them that goes beyond any mere requirement that they be complied with. Very rarely is instruction-following an end in itself. Rather, instructions tend to be means to the realization of ends that are, in the relevant sense, external to them. Flat-pack assembly-instructions are means to the end of pieces of furniture; recipes are means to the end of meals; battlefield orders are means to the end of victories; and so on. And this entails that, except in the rarest of cases – and even in those mostly only by luck – successfully following instructions requires a grasp of the end to which they are directed, of their point, so that one sees beyond the instructions, as it were, to their realization. (Just try making sense of, for example, the pictograms that come as instructions with IKEA flat-packs without keeping firmly in mind what the finished product is to be, and indeed what it should look like.)

There are of course exceptions to this – as, for example, when one is instructed to fold a piece of paper into eight, say, and to cut all sorts of intricate little holes in it. Here, there is a point to the instructions that lies beyond them – perhaps that the result, when opened out, should be surprising or delightful – but this is not an end of which one need have any grasp in order successfully to follow them.

It is unlikely, however, that musical scores are helpfully to be thought of in this way. Rather, it seems reasonable to suppose that they are central cases of the kind of sets of instructions which, if they are to be followed successfully, require a grasp of their point. And the natural suggestion is that Toscanini and Furtwängler, in their respective performances of Brahms's symphony, showed that they had indeed grasped one aspect of that point – namely, that the symphony's ending is tragic – an understanding that they may have gained through study of the score itself, through previous experiences of attempting to perform the work, or from hearing the performances of others. That they took the point that they grasped to transcend the authority of (the letter of) the score is shown in the performances of the work that they gave. And that we agree with them about this is shown in our judgement that their performances are brilliant.Footnote ¹⁰

If the foregoing is correct, we can conclude that instructions whose point lies beyond them, and whose successful implementation requires a grasp of that point, are never indefeasible. And we can note that this is why, to return briefly to Goodman, the demand that a performance comply completely with the set of instructions that a score comprises could never be a necessary condition of that performance's being a legitimate instance of the work notated there. For this demand is always open to being flouted by a ‘brilliant’ performance, by a performance which evinces a grasp of the point of the score that licenses – i.e. renders legitimate – a departure from it.

VII

There are some quite interesting consequences that follow from this, and I'll turn to several of these in a moment. But first it is necessary to address an objection that is sure to be made to the argument that I have just offered.

It will be objected that I have made things very easy for myself by picking an example in which a brilliant performance departs from the score along the dimension of tempo, rather, say, than melodically or harmonically. For, the objection continues, not everything that a score prescribes is prescribed with equal rigidity; and, in particular, prescriptions concerning tempo, as perhaps also dynamics, are not as rigid as prescriptions concerning melody or harmony – i.e. prescriptions concerning the actual notes that are to be played, however loudly and at whatever speed. Which is why – the objection may well conclude – Goodman chose to highlight ‘mistakes’ and ‘wrong notes’ rather, merely, than unprescribed accelerations and slowings down.

One possible response to this objection would be to suggest that it argues backwards. From the fact that we are very pleased to regard a performance of Brahms's Fourth, say, as ‘brilliant’, despite its departure from the score along the dimension of tempo, it seems that we are to conclude that prescriptions concerning tempo are somehow intrinsically less rigid than prescriptions along other dimensions; and so that the notion of complete compliance with the score does not (really or fully) include complete compliance with the tempo prescribed in it. Which is, surely, just to beg the question.

Moreover, the response might continue, we are also very happy to regard a performance that departs from the allegedly more rigid prescriptions concerning melody as ‘brilliant’, as is shown, for example, by Richter's performance of the Liszt sonata. So what of differences in rigidity now? And if the idea is supposed to be, simply, that Richter's departures are condonable because accidental rather than deliberate, we can easily change the example. Reinvoking a distinction mentioned a moment ago, we can describe Busoni's transcription for piano of the Chaconne from Bach's D minor violin partita as an interpretation of it – indeed, as an undeniably brilliant interpretation of it. And yet this interpretation makes melodic prescriptions quite at odds with those to be found in Bach's score: it prescribes notes other than Bach's, and fails to prescribe all of Bach's notes at the moments that Bach prescribes them. But it is ‘brilliant’ nonetheless because it gets and captures the point of the Chaconne.

I think that both parts of this response to the objection are correct, as far as they go; but also that they miss a deeper thought that the objection draws upon. This is the idea that, irrespective of what deviations we will tolerate, the actual notes, and the actual harmonic and rhythmic relations between them, that a score comprises are simply more fundamental to a work's being what it is than are any considerations about how it is to be played – for instance, loudly, softly, increasingly quickly or increasingly slowly, or, indeed, on what instruments.

There is, no doubt, a strong intuitive appeal to this idea – not least because something like it may be roughly true of much of the music that many of us listen to. And something like it is true, surely, of music in the classical tradition between the early eighteenth and the early twentieth centuries, from which all of my examples have been drawn. So I need to say more.

I noted earlier that, depending on period and style, a score can be prescriptive along a number of different dimensions, and that it can prescribe a greater or lesser amount along each of these. This indicates at once that there is no settled generalization to be made about which dimensions are intrinsically more germane or significant than others: their significance in a given score is wholly relative to the period and style of the work notated there. So, for example, in the case of a jazz standard, the most germane dimension will tend to be the harmonic; which is why suitably acculturated listeners are so much better at spotting them when they are played than are those whose listening habits direct them more automatically to, say, the melodic dimension of the work; for in this style of music prescriptions along the latter are ordinarily to be understood rather flexibly. But note that even here we can only talk of tendencies and of what is ordinarily likely to be the case: there are, as it were, some more or less reasonable stylistic expectations in play; but even these are, in the end, open to revision in light of the peculiarities of particular, individual works, and of the significance, relative to them, of this or that dimension of prescription.

So we can say that, in general, there is a reasonable expectation that in classical music of the relevant period the dimensions of melody, harmony and rhythm will be more significant than that of tempo. And this is, no doubt, why it is relatively easy to find examples of ‘brilliant’ performances of such music that depart from what the score prescribes along the latter dimension. But – as I suggested a moment ago – it is not as if there are no examples of brilliant performances that make other kinds of departures: Busoni's transcription of Bach's Chaconne is just one such instance, an instance that will be detected by any suitably acculturated listener.Footnote ¹¹ Rather, in this as in all cases, we can talk only of tendencies, of what is generally likely to be true of works of this or that sort. And, of course, the fact remains that the work that I have focussed on – Brahms's Fourth Symphony – does contain explicit prescriptions concerning tempo in its score, prescriptions which, even if they are to be regarded as less rigid than some of the others that the score contains, are nonetheless roundly, and even productively, disregarded in the ‘brilliant’ performances that I have mentioned.

One might in fact be tempted to make a stronger claim in the present context, and say that, because the score contains relatively few prescriptions concerning tempo, as opposed to those concerning rhythm, say, any one of the former should be regarded as less defeasible than any one of the latter (if, as Goodman and others assume, the score in some sense ‘defines’ the work). But this temptation should be resisted. It is true that there is one reasonably robust generalization that lies in this sort of direction – namely, that the less, overall, that a score prescribes, the less defeasible are its individual prescriptions (consider the limiting case of John Cage's 4'33'': the playing of any note on the piano at all would be sufficient to disqualify a performance as being of that work). But we can't get from there to any useful generalizations concerning individual dimensions of prescription, partly for reasons having to do with a well know informal fallacy, but mostly because, as I have insisted, the significance of this or that dimension in a given score is wholly relative to the period and style of the work notated there, and may even be relative to the work itself.Footnote ¹²

Thus, to return to the objection, it is true that it is easier to find examples of classical performance that are ‘brilliant’ despite departing from prescriptions concerning tempo than it is to find them among performances which depart from prescriptions concerning, e.g., melody; and it is true that this is, in part, a function of the reasonable stylistic expectations that an acculturated listener will have in the context of music of this sort. But none of this shows that the idea of complete compliance with a score somehow excludes complete compliance with prescribed tempi; nor does it show that the departures from a score of a performer who disregards the latter are somehow different in kind from those of a performer who disregards prescriptions along any other dimension with which a score concerns itself. So the objection fails.

VIII

I have argued, then, that instructions whose point lies beyond them, and whose successful implementation requires a grasp of that point, are never indefeasible. And I have suggested that this shows why, contra Goodman, the demand that a performance comply completely with the set of instructions that a score comprises could never be a necessary condition of that performance's being a legitimate instance of the work notated there. There is always the possibility that a performance will be given – a ‘brilliant’ performance – which evinces a grasp of the point of the score in a way that essentially involves, and hence licenses, a departure from it.

I said a moment ago that, if these conclusions are correct, some quite interesting consequences follow. One of these is that the ways in which we can usefully think about musical content are circumscribed in respects that some philosophers have missed. So, for example, Andrew Kania has recently warned us of the dangers of equivocating between two senses of ‘musical content’, according to one of which we can specify the content of a work in advance of any attempt to grasp or understand its point, according to the other of which we cannot give any such advance specification. He attempts to explain the distinction that he has in mind through an analogy with literature:

– just as, he suggests, ‘anyone with access to to a copy of [a] score and the ability to read it can tell you to a large extent the content a performance would need to have if it were to be a performance’ of a given work.Footnote ¹³

But the analogy with literature is unhelpful. First, a literary text does not stand in the same relation to a literary work as a score stands in to a musical work. Scores, we have followed Goodman in supposing, are instructions for producing performances of musical works, whereas texts – notwithstanding some heroic argumentation to the contrary by Peter Kivy – are not instructions for producing performances of literary works.Footnote ¹⁴ So while it is true that one might check word by word that two copies of Finnegan's Wake are identical, without an understanding of what those words signify; and while it is perhaps true that there is a sense in which one might be said to be checking thereby that both copies have the same literary content; nothing at all follows from this about the issue of musical content. For the content of a set of instructions is not equivalent to the content of its realization. The content of a score is no more the content of the work notated there, of which performances might be given, than the content of a recipe is identical to the content of the meal prepared in its light, or than the content of a set of flat-pack assembly-instructions is a table, or a day-bed. And this point is evidently unaffected by a properly completed version of Kania's would-be analogy. For it is clearly true that someone without an understanding of a given work could check, note by note, whether two putative scores of it are identical: but this would not be thereby to check the musical content of the work for whose realization they provide instructions. However musical content is to be understood, in other words, it cannot be on the assumption that, in terms of content, scores and works are identical.Footnote ¹⁵

Another consequence of my argument is that we should perhaps reconsider our willingness to concede to Goodman that complete compliance with the score might at least be a sufficient condition of a performance's counting as a legitimate instance of a work – that, as he has it, ‘the most miserable performance without actual mistakes’ – i.e. without actual departures from the score – counts as such an instance. I said at the beginning that we were likely to accept this without demur; and perhaps we should accept it. But we should also note that a performance might be ‘miserable’ in Goodman's sense in virtue of competely missing the point of the work performed – by failing to register, for instance, that the work ends in tragedy, by being leaden where light-footedness is called for, by remaining oblivious to the need to be emphatic just here, by missing the deep pulse that should animate the work and substituting for it something merely akin to a beat; and so on. We might well think that any performance with these shortcomings is not just miserable, because point-missing and aesthetically inert, but is not really an instance of the work at all. And this, in turn, might make us wonder whether ontological and evaluative issues can, after all, be kept as distinct as Goodman and his followers would have us believe.

Perhaps. I'll suggest in a moment that the latter worry is in fact warranted. But as to the former – whether an aesthetically wretched performance of a work can really count as a legitimate instance of it – it seems to me that the proper response is a shrug of the shoulders. Who cares? Count it if you want to – just don't expect me ever to listen to it again. Why should the question whether it is legitimate as an instance stir me?

IX

The reason why it might feel as if it should stir me is that the corresponding question at the other end of the evaluative scale certainly does have, and has proved to have, real purchase. For, as I noted at the outset, no one at all has been willing to go along with Goodman's thought that ‘the most brilliant performance with one wrong note’ does not count as a legitimate instance of a work – and this, not for the already good enough reason that it is impossible to make sense of the claim, as I tried to show earlier that it isn't, but for the less philosophical and better, more musical, reason that if anything counts as a legitimate instance of, say, Beethoven's Fifth Symphony, then Furtwängler's 1937 performance does. We seem to regard the ‘brilliant’ performance as a trump card in this context.

I think that we are right to do this, on grounds that I have attempted to defend. I also think that we are right, on much the same grounds, to be more or less indifferent to the question whether the wretched performance should be regarded as a sort of anti-trump card (who cares?). But the next move that has standardly been made – to hunt for identity conditions that allow enough ‘wiggle room’ to accommodate brilliant performances – strikes me as very problematic; and this is the final consequence that follows from the argument that I have attempted to offer here.

The problem, in a nutshell, is this. Any set of identity conditions is going to have to rule some putative instances of works out, otherwise we'll find ourselves allowing ‘Three Blind Mice’ to count as a legitimate instance of, say, Beethoven's Fifth. It is, moreover, going to have to rule out the relevant non-instances a priori, so that we can disallow this or that deviation from whatever norms the conditions specify in advance of any particular deviating instance. A lot of the time this won't be a problem. Many of the non-instances that are ruled out will be miserable or wretched performances, and no one should much mind or care about the exclusion of these – or indeed have any particular view about them either way. If the odd one has to go to the wall for the sake of a certain theoretical tidiness, then so be it. But every now and again the performance to be disallowed will be a ‘brilliant’ one; and these, as we have seen, function as trump cards. They show, just as they showed in the Goodman case, that conditions that had been taken to be indefeasibly necessary can, in fact, be no more than sufficient at most. They demand, that is, some extra ‘wiggle-room’ to be built in. And this means that any set of identity conditions, no matter how splendidly accommodating it might be of brilliant performances that we already know about, can only ever be provisional, since it simply cannot have the clout to rule out, a priori, the possibility of the next deviant, but brilliant, performance that nobody could have seen coming. We can say, then, that, in this context, the role of brilliant performances is to remind us that our aesthetic expectations, however reasonable, and however ingeniously formulated, are always, in principle, open to disruption. And this, it seems to me, poses a more or less insurmountable obstacle to the identity-project in which Goodman and his followers have been engaged.Footnote ¹⁶