Start Submission

Reading: Home to Roost: Some Problems for the Nested-Types Theory of Musical Works, Versions, and Aut...


A- A+
Alt. Display

Critical Note

Home to Roost: Some Problems for the Nested-Types Theory of Musical Works, Versions, and Authentic Performance


Andrew Kania

Trinity University, US
X close


A critical note on Nemesio G. C. Puy, ‘Interpretive Authenticity: Performances, Versions, and Ontology’, Estetika 59 (2022): 135–52.

How to Cite: Kania, Andrew. “Home to Roost: Some Problems for the Nested-types Theory of Musical Works, Versions, and Authentic Performance”. Estetika: The European Journal of Aesthetics 59, no. 2 (2022): 153–64. DOI:
  Published on 15 Sep 2022
 Accepted on 09 Aug 2022            Submitted on 09 Aug 2022

An authentic performance of a work of Western classical music is one that is faithful to the work.1 But what does such faithfulness amount to? Julian Dodd has recently argued that the fundamental faithfulness is interpretive: an authentic performance elicits a ‘subtle or profound’ understanding of the work in listeners.2 Such interpretive authenticity is fundamental in that it trumps not only instrumental, ‘garden-variety performance values [such as] interestingness, liveliness, [and] originality’ but also ‘score compliance authenticity’, something valued for its own sake in the tradition. Stephen Davies and I disagree, arguing (in different ways) that score compliance cannot be abandoned for the sake of insight into the work; performances that abandon score compliance are at best performances of versions or large parts of the work in question.3 Nemesio García-Carril Puy argues for a middle way between these two views.4 He claims that Davies and I are right that such performances are of versions of the work in question, but that Dodd is right that such performances are nonetheless of the work in question itself. How could both these claims be correct? According to Puy’s nested-types ontology, musical works are not as determinate as most philosophers and music scholars have assumed. A work can only be experienced in a version that specifies more determinate values of variables in the work, such as the details of a melodic line, the structure and details of a passage, or the work’s instrumentation. In turn, the version (what others call the work) can only be experienced in a performance that specifies determinate values of the version’s variables, such as the specific tempo, particular embellishments, and so on. The upshot is that a controversial performance such as Andreas Staier’s playful account of Mozart’s ‘Rondo alla Turca’ (henceforth ‘the Rondo’) is a performance of a version of Mozart’s work, but is thereby a performance that is both interpretively faithful to Mozart’s work – a work that is thinner than the version notated in the score – but also score-compliantly faithful to that version.

I. Hatching Problems for Score Compliance

Puy attempts to summarize Davies’s and my objection to Dodd’s view in his three-premise ‘objection from ontology’ (p. 142). I cannot speak for Davies, but I have reservations about the argument as it stands.5 As Puy notes (p. 142, note 22), my reservations are connected with his view (which he seems to think Davies and I should endorse) that an authentic performance of a work must be a properly formed instance of the work: ‘[d]isobeying W’s score results in improperly tokening W, which disqualifies those performances as authentic performances of W’ (p. 141). It seems to me that requiring an authentic performance of a work to be a properly formed instance of it misses the point of introducing the concepts of (im)properly formed instances, and thus norm types, in the first place. If a work can have improperly formed instances, then these are authentic – though not ideal – instances of the work.

As I read Davies, he would finesse this point by claiming that authenticity is a matter of degree, and thus that, to the extent that one departs from the score (in error or on purpose), one’s performance is less authentic, until the threshold of unrecognizability is reached, at which point the performance ceases to be of its target work.6 There are various ways in which Puy’s argument from ontology could be altered to reflect Davies’s view, but I leave that exercise to the reader.

My own view, as Puy notes, is that only intentional departures from the score preclude a performance’s being authentic; mistakes are another matter.7 Puy objects that this view is untenable: suppose Herbert von Karajan, on a visit to New Zealand, directs the Auckland Philharmonia to perform Beethoven’s Fifth Symphony as he has long led the Berlin Philharmonic in performing it – with horns playing the fanfare that is scored for bassoons in the recapitulation of the first movement. Forgetting Karajan’s instructions in the heat of the moment, however, the horns remain silent and the bassoons play the fanfare, as they have always done. Puy claims that this performance must count as inauthentic on my view, since Karajan intended to depart from the score. Yet the performance does not actually depart from the score, so it must be authentic. Hence my view must be modified.

There are a number of responses available. I could accept Puy’s proposed amendment that only successful intentional departures from the score count against authenticity. Alternatively, I could resist Puy’s claim that Karajan’s Auckland performance is an authentic performance of Beethoven’s Fifth. The basic, and widely accepted, idea would be that a performance’s sonic properties are never sufficient for that performance’s being of a particular work; also required is some sort of intentional connection between the sounds and the work. If a malicious demon sabotages the orchestra’s instruments so that, when they try to play Beethoven’s Fifth, the sounds of Mahler’s First emerge, it is not obvious that (on philosophical reflection) we should consider the result a performance of Mahler’s First. Less extravagantly, if we discovered a long-lost manuscript of Carl Maria von Weber’s that was (coincidentally) qualitatively identical to the score of Beethoven’s Fifth, we would not thereby discover that all performances of Beethoven’s Fifth were also performances of Weber’s work.8

The best response to Puy’s supposed counterexample, however, is to question its coherence. Puy claims that the Auckland performance counts as inauthentic on my view because ‘Karajan’s intention was not to comply with Beethoven’s score’ (p. 142, note 22). But this is misleading in two ways. The first can be illustrated by a simpler case: a pianist, inspired by Dodd’s work, intends to perform the ‘Rondo’ in a playful, Staier style. Alas! On stage, she chokes and ends up performing the work in her usual, perfectly score-compliant way. In what we might call the distal sense, this performer intends (or intended?) to depart from the score – that is (was?) her plan. But this is not the sense of ‘intention’ relevant to theories such as mine, which we might call the proximal sense – intentions as a kind of real-time, action-guiding mental state. In this case, the performer’s intentional actions are best described as complying with the work’s score, just as my intentional actions while driving a car include indicating my upcoming turn, even if I intended (in the distal sense) to live on the edge by not indicating my turns on this trip.9

Puy’s characterization of the Karajan case is misleading, second, in that he considers only Karajan’s intentions, which are not obviously the relevant ones. How best to understand collaborative intentional action is contested.10 But the intentional actions resulting in the notes emerging from the bassoons (and the silence of the horns) – at least in this case – are surely best understood as those of the players, not those of the conductor. Assume that Karajan cued the horns at the relevant point. Since the horns (and bassoons) ignored this cue, it was not an action that (directly, at least) determined the sonic properties of the performance, while the horns’ and bassoons’ actions did determine such properties. So the Auckland performance is properly described as one in which the performers intentionally complied with the score.11

Thus far, I have adopted Puy’s use of Dodd’s ‘score compliance’ terminology. As I have discussed elsewhere, however, this is potentially misleading, perhaps question-begging.12 For, according to Davies and me, one (properly) complies with the score only because scores are the method by which works are publicly specified. If something like that view is correct, ‘score compliance’ can only be a kind of shorthand for ‘work compliance’ or, even better, ‘work instantiation’ or ‘work performance’. Of course, to insist on that terminology might also give the appearance of question-begging. But all parties to the debate must keep clearly in mind that one disputed issue is whether ‘score compliance’ is valued fundamentally, as the only way of properly performing the work (as I would have it), or less than fundamentally, as Dodd and Puy would have it.13

It is also worth remembering that the determination of the work by the score (on views such as Davies’s and mine) is a complex matter. Sometimes, Puy seems to suggest that score compliance is a matter of translating all and only those properties recorded in a score into sound (pp. 136, 140–41). But, as Davies has discussed at length, the situation is much more complicated.14 A score is a contextual artefact, whose understanding requires familiarity with many conventions concerning, for instance, transposing instruments, the presence of a conductor, stylistic matters concerning articulation, rhythm, ornamentation, and so on.15

Keeping these issues clearly in mind neutralizes one supposed problem Puy raises for my (and presumably Davies’s) view. He points out (i) that we often have no, or only a partial, autograph score (that is, one handwritten by the composer) for many pre-twentieth-century works, (ii) that autograph scores often do not capture the composer’s ‘final instructions’, and (iii) that editors, who may make changes to what the composer intended, are inevitably involved in the production of published scores. This is particularly relevant for the central example of the ‘Rondo’, according to Puy, because four of the nine pages constituting Mozart’s autograph score of the sonata have been lost, and only one of the extant pages contains any part (the ending) of the ‘Rondo’.16 Thus ‘there is no […] score’ for the work (p. 143).

This conclusion strikes me as absurd. Mozart oversaw the publication of the first edition of this sonata, and the discovery, in 2014, of four of the five (known) extant pages of the autograph score confirmed that the published score is almost identical to that autograph in all relevant respects. The discovery did allow, as is often the case, for a couple of minor clarifications, but even here editors and musicians were aware that small errors must have been introduced into the first edition, and were unsure only of how to properly correct them. Moreover, the problems Puy sees with both the role of editors in the publication of scores and the fact that composers revisit their scores in a range of ways betray a simplistic intentionalism about the nature of artworks. As with a published novel, the content of a published musical work is basically determined by what ends up in print in an ‘authorized’ edition. Publication, though, is a messy process, involving the participation and influence of many people, even in the case of what we consider single-authored works. To insist that the work is determined by the artist’s ‘final intentions’, independent of the published text, score, and so on, suggests an outmoded Romantic-genius theory of artistic creation. I do not mean to deny that we are in a less-than-ideal epistemic position with respect to the details of many artworks, including musical works with authorized published scores. But these issues seem to me quite similar in lack of importance to the main questions in this debate as issues of practical considerations, such as whether one has the relevant instrument on hand. We can safely bracket them to discuss the question of authentic work-performance.17

Suppose, despite what I have just said, that Puy’s criticisms were sound. One advantage of his view would then be that, though the lack of reliable scores for many works might be an epistemic problem on his view, it would not be a (more serious) metaphysical problem, as it is for Davies and myself. Since the primary route to figuring out the content of the work (that is, the nesting type on Puy’s view) is clearly through the published score, our access to the content of musical works is unreliable to the extent that published scores are unreliable. But if the content of works is determined (in large part) through published scores, their unreliability renders works vague at best.18 However, this raises two central, related problems for Puy’s view that will occupy the rest of my discussion. First, there is the question of what determines the content of a nesting work (as opposed to its nested versions). Second, there are reasons for thinking that such nested types cannot be musical works in the relevant sense.19

II. Building the Nest

According to the nested-types view, musical works are thinner than they are typically taken to be. But how much thinner, exactly? Think of the wide variety of ways in which a version can differ from the original work (or version, as Puy would have it). In some cases, entire movements are added, removed, or replaced. (I consider below the example of Beethoven’s String Quartet No. 13, op. 130.) In other cases, instrumentation is radically altered. (We might consider Webern’s orchestration of the ‘Ricercar’ from Bach’s Musical Offering.) Often it is the way in which musical material is developed that is changed. (Puy’s example of Bruckner’s Symphony No. 8 is [partly] of this kind.) As David Davies points out, if all these possibilities are to be explained via the nested-types theory, musical works are so thin that there is little of interest we can say about them.20 Bear in mind that the theory says that all works are higher-order types, not just those that actually have more than one version, just as standard theories of musical works say that they can be instanced in a range of different performances, independently of how often, or even whether, they have actually been performed.21

It is instructive to consider the ‘Rondo’ – one of Puy’s (and Dodd’s) central examples – in more detail. First, note that no one thinks this is, strictly speaking, a work; it is one movement of a three-movement piano sonata (K. 331). It thus seems that there could have been versions of the work, that is, the entire sonata, that replaced this movement with an entirely different one, just as there are two versions of Beethoven’s String Quartet No. 13, op. 130 (one with the Grosse Fuge as the last movement and another with a more conventional finale). Indeed, if Puy’s view is correct, there could yet be such versions of the work, scored or performed by a suitably insightful and skilled musician. It follows that Puy’s characterization of the nesting work is grossly overdetailed: this movement has no particular key, time signature, thematic material, and so on. It seems to follow that the other two movements are in the same boat. In fact, is even the number of movements of the work determinate? Perhaps there are some essential features to the work: it is in a classical style, say. But now it looks like many works are going to be identical; all the works in a classical style that could have had different keys, time signatures, thematic material, and so on.22

For the sake of argument, however, suppose that this problem can be solved. Let us proceed on the assumption that the ‘Rondo’ is a single-movement work. It still seems that Puy’s characterization of this (nesting) work is too detailed. Puy once characterized the work’s individuating properties in musical notation (see Figure 1).23

Musical notation of several bars from Mozart’s ‘Rondo alla Turca’
Figure 1 

Puy’s earlier representation of the nesting work (Mozart’s ‘Rondo alla Turca’).

In the final version of the paper, Puy characterizes the ‘Rondo’ only verbally, saying that:

Among the properties that individuate [the work] we can count those of ‘being tonal music’, ‘having a rondo’s form’, ‘being an ethnic showpiece’, ‘being in 2/4’, ‘starting with a texture of accompanied melody’, and ‘starting with a melody that unfolds the A minor chord in an ascendant arpeggio with some embellishments’. (p. 149)

Both these characterizations of the work face similar problems. Let’s first consider some problems that they share.24 (i) Both imply that the ‘Rondo’ is necessarily a piece of tonal music. But some artists claim that they can create versions of a work in different art forms or media.25 (ii) The claim that the ‘Rondo’ is essentially in 2/4 time raises difficult questions about both the nature of meter and relations between works, scores, and performances. For instance, if we rewrote a score of the ‘Rondo’ in 2/8, 2/2, or 4/4 time, would we have a different version, or merely a different notation of that version? Whatever our answer to that question, a version in 3/4 or 6/8 time would surely not be the same ‘version’ as Mozart’s. But it is surely acceptable to talk (as I just did) of such a piece of music as a version of the same work.

(iii) Both characterizations imply that the ‘Rondo’ must begin with an accompanied melody structured around an embellished ascending A minor arpeggio. But if Staier’s departures from the published score do not disqualify his performance as of a version of Mozart’s work, then why couldn’t there be versions of the ‘Rondo’ that start with the melody unaccompanied or, alternatively, with just the accompaniment, soon joined by the melody? Similarly, why couldn’t there be versions in keys other than A minor, or versions without embellishments of the basic arpeggio structure of the melody?

Now consider some problems raised only by the music-notational characterization of the work. (iv) It makes the ‘Rondo’ explicitly for piano, though presumably there could be a version with different instrumentation. (This might be one reason for a change of key.) (v) The parentheses around particular notes in the right hand indicate that these notes may (or must?) be embellished. But is it really the case that a version that also (or instead) embellished the other crotchets in this line would not count as a version of the work? (vi) The first two quavers in the left hand are slurred in the first two bars, but not the third, though presumably such matters of articulation are not necessary to the nesting type. (In fact, Staier departs wildly from the left hand as represented here in some instances of this material.) And so on.26

Finally, consider some problems raised only by the verbal characterization of the work. (vii) Could there really be no version of the work that is not in rondo form? I can imagine someone arguing that a version in the form of theme-and-variations captures its spirit more accurately. (viii) Mozart’s version of the piece (as Puy would have it) is plausibly ‘an ethnic showpiece’, but I can imagine someone arguing that this is an unfortunate artistic shortcoming, and perhaps politically offensive aspect, of this score and producing a ‘less Turkish’ performance or score – claiming, of course, that it is a version of Mozart’s work.

The resolution of these problems clearly requires some account of what determines the content of the nesting work, according to Puy’s theory. With this account in hand, we would be able to draw a principled distinction (at least in theory) between, as Puy puts it, versions of a single target work and distinct works inspired by the target work. But I see no such account of work-determination in Puy’s discussion of the theory here or elsewhere. I presume that, ultimately, such an account would be part of something like a rational reconstruction of the relevant musical practice. There is disagreement about how exactly such reconstruction should proceed,27 but more problematic here is the fact that musical practice is much less regimented when it comes to what counts as a version, transcription, arrangement, inspired work, and so on than it is with respect to what counts as a performance of the musical entity represented in a score. This suggests that any account, like Puy’s, that attempts to precisify the notion of a version will be justified primarily in terms of more general theoretical benefits rather than in terms of verisimilitude to musical practice.28 Yet Puy seems to think that the primary justification for his view is to be found in closer attention to musical practice – in particular, the practice of creating, performing, and referring to versions. I concur with David Davies in finding these justifications inadequate.29

III. Eggs and Baskets: Nesting Types, Nested Types, and Works

Let me now put aside this first problem of what determines the content of a work on Puy’s view and move on to the second. Suppose we accept Puy’s ontology for the sake of argument. It seems to me that now Dodd, Stephen Davies, and I will want to reframe our disagreement in terms of versions. That is, Davies and I will maintain that a fully authentic performance of a version must comply with the score of that version, while Dodd will insist that a performance may sacrifice such score compliance for the sake of eliciting a subtle or profound understanding of the version in listeners. But Puy presumably thinks that this debate rests on a mistake – a mistake that is thrown into relief once we have adopted his ontology and its terminology. For our primary interest in performances, all parties to this debate agree, is in their being performances of works. The mistake that Dodd, Davies, and I have made, according to Puy, is to misidentify where works reside in the hierarchy of thicker to thinner musical structures. We can see this by comparing visual representations of the various ontologies on offer here. Consider the representation of Puy’s view in Figure 2. (The arrows indicate something like derivation, so that ‘XY’ means something like ‘Y is derived from X’.)30

Nested-types view exemplified on the musical notation of Mozart’s ‘Rondo’
Figure 2 

The nested-types view.

Puy says that Davies and I are right that Staier’s performances (3 and 4, let’s say) are of versions of Mozart’s work, but, of course, Davies and I don’t think that W is Mozart’s work; we think that Mozart’s work is WM. Similarly, Puy says that Dodd is right that Staier’s performances are of Mozart’s work – they are performances of W by virtue of being performances of WS. But Dodd agrees with Davies and me that Mozart’s work is WM, not W.

This can sound like a merely verbal dispute because Davies and I don’t seem obliged to reject the existence of anything in Puy’s picture; we would simply label things differently (see Figure 3).

Davies’s and Kania’s view exemplified on the musical notation of Mozart’s ‘Rondo’
Figure 3 

Davies’s and my view.

If there’s more than a verbal dispute here, it seems to me that it resides in the role of the work concept. Like many others, I take ‘work’ to be a quasi-technical term referring to the kind of thing that is a primary target of appreciation in an artistic practice. And it seems obvious to me that when people talk about Mozart’s ‘Rondo alla Turca’, they are talking about a musical object with a specificity very close to that of Mozart’s score. And that is, in part, because of how scores function within the practice. Mozart could, of course, have written a score that continues in the style of Puy’s harmonic reduction – though it is doubtful that he would have called it ‘alla Turca’! Perhaps, given the conventions of the time, such a score would properly be embellished in performance in various ways, so that authentic performances of it could sound exactly like performances 1–2 or 1–4 in my diagrams. The fact that Mozart did not do this is strong evidence that he intended to specify a more detailed musical structure in this work.31

(For the record, I suppose Dodd’s view would be represented as in Figure 4, since he disagrees with Davies and me about what constitutes a performance’s being of a work, as opposed to being of a version of a work. Note, however, that the notation at the lowest level here is not of individual performances; it is intended to indicate that both ‘score-compliant’ performances (for example, 1 and 2) and ‘interpretively authentic’ but non-score-compliant performances (for example, 3 and 4) are of a single work, on Dodd’s view.)

Dodd’s view exemplified on the musical notation of Mozart’s ‘Rondo’
Figure 4 

Dodd’s view.

IV. Concluding Remarks

Nothing I have said here constitutes a knockdown argument against Puy’s ontology. Choosing between ontologies requires careful consideration of the full range of costs and benefits of each theory. This is not the place to do that, but I hope to have shown that the nested-types view cannot be used to resolve the debate between Dodd, on the one hand, and Stephen Davies and me, on the other, without further work to defend the plausibility of the view.


1Henceforth I take the ‘Western classical’ – and often ‘music’ – qualifiers as read. 

2Julian Dodd, Being True to Works of Music (Oxford: Oxford University Press, 2020), 163. 

3Stephen Davies, ‘Performing Musical Works Authentically’, British Journal of Aesthetics 53 (2013): 71–75; Andrew Kania, Philosophy of Western Music: A Contemporary Introduction (London: Routledge, 2020), 188–97; ‘The Heart of Classical Work-Performance’, British Journal of Aesthetics 62 (2022): 125–41. 

4Nemesio G. C. Puy, ‘Interpretive Authenticity: Performances, Versions, and Ontology’, Estetika 59 (2022): 135–52. DOI: All references to Puy’s work are to this article, unless otherwise stated. 

5In particular, as we shall see, Puy is wrong that ‘premise 3 captures Kania’s way of seeing Karajan’s and Staier’s performances’ (p. 142). For what it’s worth, I would endorse the argument if its terms were interpreted charitably according to my view as expressed in my ‘Heart of Classical Work-Performance’, and premise 3 were revised as follows: 

3. But a performance that intentionally departs from W’s score thereby fails to be a performance of the (complete) work.

But I prefer to approach the issues the argument raises from a different angle.

6Stephen Davies, Musical Works and Performances: A Philosophical Exploration (Oxford: Oxford University Press, 2001), 160–61. 

7In fact, as I suggest elsewhere (‘Heart of Classical Work-Performance’, 133), my view could be considered a different interpretation of various things Davies says from that given in the previous paragraph. 

8Comp. Davies, Musical Works and Performances, 161–81, and the references therein. 

9For more on artistic intentions, see Paisley Livingston, Art and Intention: A Philosophical Study (Oxford: Oxford University Press, 2005). 

10For general discussion, see Abraham Sesshu Roth, ‘Shared Agency’,in Stanford Encyclopedia of PhilosophyEdward N. Zalta, summer 2017 ed. (Stanford University, 1997–),; for consideration of artistic collaboration in particular, see Sondra Bacharach and Deborah Tollefsen, ‘We Did It: From Mere Contributors to Coauthors’, Journal of Aesthetics and Art Criticism 68 (2010): 23–32, and the ensuing discussion in the Journal of Aesthetics and Art Criticism. 

11I ignore the contributions made to the performance by the visual aspects of conducting and the effect of the orchestra’s failure to follow Karajan’s directions on his consequent actions. For some related discussion, see Stephanie Ross and Jennifer Judkins, ‘Conducting and Musical Interpretation’, British Journal of Aesthetics 36 (1996): 16–29; Vincent Bergeron and Dominic McIver Lopes, ‘Hearing and Seeing Musical Expression’, Philosophy and Phenomenological Research 78 (2009): 1–16. 

13It is not entirely clear to me what Puy takes the ‘normative profile’ of score compliance to be. According to Dodd, score compliance is valued ‘for its own sake, yet only under the condition of its instrumentality’ (Being True, 144). That is, it has final value, but only because of its tendency ‘to facilitate [works’] insightful interpretation’ (ibid, 163). Puy says that interpretive authenticity has the ‘same normative profile’ ‘in the practice of composing musical versions’ as it does ‘in the practice of performing musical works’, including being ‘a final value’ (p. 144). But since Puy thinks that scores directly notate versions, rather than works themselves, it would seem that version compliance would take the place (in Puy’s theory of the normative profile of the practice) that Dodd grants to score compliance. I would think that (version) score compliance would, in turn, be less fundamental than version compliance, according to Puy. Or do they share the second tier? 

15Pace Dodd, Being True, 47–87. For discussion, see Kania, ‘Heart of Classical Work-Performance’, 136–39. 

16A useful discussion of the work’s autograph score and its relation to the history of the work’s publication can be found on the website of the National Széchényi Library, where four of the extant pages were discovered in 2014: 

17This is all rather quick, because of space constraints. For further discussion, see Davies, Musical Works and Performances, 66–71, 247–49; Kania, Philosophy of Western Music, 188, 191. 

18Indeed, perhaps the view would be incoherent if Puy’s criticisms were correct. For how could work-determining scores be widely unreliable? 

19I should note up front that I am indebted, in what follows, to an excellent recent discussion of the nested-types view in David Davies, ‘Puy on “Nested Types”’, Journal of Aesthetics and Art Criticism 79 (2021): 251–55. 

20Ibid., 253. 

21A further unintuitive corollary of Puy’s view: every work-performance is, in addition, a performance of a particular version of a work. 

22Thus, Puy is wrong that considering the ‘Rondo’ as a work in itself ‘does not affect’ his arguments (p. 148, note 44). 

23In presenting an earlier version of his ‘Interpretive Authenticity’ at the 2021 Annual Meeting of the American Society for Aesthetics. 

24Since both the verbal and music-notational characterization of the work include only a sample of its individuating properties, it is difficult to tell whether the inclusion or omission of a property (for example, the instrumentation) represents a change of mind or simply a different sample of properties. 

25Imagine my delight when, while writing this commentary, I came cross the actual example of a supposed version of Bruckner’s Symphony No. 8 (another of Puy’s central examples) rendered in paint on canvas. The artist claims that ‘the original structure of the piece is retained along with the meaning. […] I do not paint music; I paint a composition [that is, musical work] which was written in the language of music. […] My paintings are basically re-orchestrated versions of the original composition – re-orchestrated, that is, for the eyes.’ Jack Ox, ‘The Systematic Translation of Anton Bruckner’s Eighth Symphony into a Series of Thirteen Paintings’, in Bruckner-Symposion Linz 1990: Bericht, ed. Othmar Wessely et al. (Linz: Bruckner Institut, 1993), 83 (italics added). For discussion of this and other ‘paintings of music’, see Michelle Liu, ‘Paintings of Music’, Journal of Aesthetics and Art Criticism 80 (2022): 151–63. 

26One might reasonably object that the thinness of nesting works precludes their clear and accurate representation in standard musical notation. (Perhaps this is why Puy refrained from supplying this representation in the published version of his paper.) After all, standard musical notation developed as a way to characterize the content of work-versions (in Puy’s terminology), not works themselves. To my mind, however, this raises issues about the concept of a musical work to which I turn in the next section. 

27See, for example, David Davies, Art as Performance (Malden, MA: Blackwell, 2004), 16–24; ‘The Primacy of Practice in the Ontology of Art’, Journal of Aesthetics and Art Criticism 67 (2009): 159–71; ‘Descriptivism and Its Discontents’, Journal of Aesthetics and Art Criticism 75 (2017): 117–29; and Julian Dodd, ‘Adventures in the Metaontology of Art: Local Descriptivism, Artefacts and Dreamcatchers’, Philosophical Studies 165 (2013): 1047–68. 

28For one such attempt that, in my view, strikes a better balance between theory and practice, see Stephen Davies, ‘Versions of Musical Works and Literary Translations’, in Philosophers on Music: Experience, Meaning, and Expression, ed. Kathleen Stock (Oxford: Oxford University Press, 2007), 79–92. 

29Davies, ‘Puy on “Nested Types”’. To give just one example of Puy’s view contradicting practice, he claims that Staier’s ‘version’ of ‘Mozart’s work’ is like Mozart’s version of Handel’s Messiah, and that Staier neglected to notate his version only because he didn’t need to ‘to communicate [it] to other performers’ (p. 146). But surely a pianist who ‘performed Staier’s version’ of the piece would be considered by most audiences to be giving a performance of Mozart’s work that is highly derivative of another performance of it, rather than a performance of a version of that work. 

30To return to an earlier issue: the inclusion of musical notation should not be misconstrued as implying that musical works or versions are scores on any of these views. The notation is just a convenient way of representing distinct musical entities, whatever their fundamental ontology. 

31For the relevance of this point to authentic performance, see Kania, ‘Heart of Classical Work-Performance’. 


For helpful discussion of these issues, I thank Nemesio García-Carril Puy, Jerrold Levinson, Julie Post, and the audience at the session of the 2021 Annual Meeting of the American Society for Aesthetics, at which Puy and I presented earlier versions of his article and my commentary.

Competing Interests

The author has no competing interests to declare.


  1. Bacharach, Sondra, and Deborah Tollefsen. ‘We Did It: From Mere Contributors to Coauthors.’ Journal of Aesthetics and Art Criticism 68 (2010): 23–32. DOI: 

  2. Bergeron, Vincent, and Dominic McIver Lopes. ‘Hearing and Seeing Musical Expression.’ Philosophy and Phenomenological Research 78 (2009): 1–16. DOI: 

  3. Davies, David. Art as Performance. Malden, MA: Blackwell, 2004. DOI: 

  4. Davies, David. ‘Descriptivism and Its Discontents.’ Journal of Aesthetics and Art Criticism 75 (2017): 117–29. DOI: 

  5. Davies, David. ‘The Primacy of Practice in the Ontology of Art.’ Journal of Aesthetics and Art Criticism 67 (2009): 159–71. DOI: 

  6. Davies, David. ‘Puy on “Nested Types”.’ Journal of Aesthetics and Art Criticism 79 (2021): 251–55. DOI: 

  7. Davies, Stephen. Musical Works and Performances: A Philosophical Exploration. Oxford: Oxford University Press, 2001. DOI: 

  8. Davies, Stephen. ‘Performing Musical Works Authentically.’ British Journal of Aesthetics 53 (2013): 71–75. DOI: 

  9. Davies, Stephen. ‘Versions of Musical Works and Literary Translations.’ In Philosophers on Music: Experience, Meaning, and Expression, edited by Kathleen Stock, 79–92. Oxford: Oxford University Press, 2007. 

  10. Dodd, Julian. ‘Adventures in the Metaontology of Art: Local Descriptivism, Artefacts and Dreamcatchers.’ Philosophical Studies 165 (2013): 1047–68. DOI: 

  11. Dodd, Julian. Being True to Works of Music. Oxford: Oxford University Press, 2020. 

  12. Kania, Andrew. ‘The Heart of Classical Work Performance.’ British Journal of Aesthetics 62 (2022): 125–41. DOI: 

  13. Kania, Andrew. Philosophy of Western Music: A Contemporary Introduction. London: Routledge, 2020. 

  14. Liu, Michelle. ‘Paintings of Music.’ Journal of Aesthetics and Art Criticism 80 (2022): 151–63. DOI: 

  15. Livingston, Paisley. Art and Intention: A Philosophical Study. Oxford: Oxford University Press, 2005. DOI: 

  16. Ox, Jack. ‘The Systematic Translation of Anton Bruckner’s Eighth Symphony into a Series of Thirteen Paintings.’ In Bruckner-Symposion Linz 1990: Bericht, edited by Othmar Wessely, Andrea Harrandt, Elisabeth Maier, and Uwe Harten, 83–101. Linz: Bruckner Institut, 1993. 

  17. Puy, Nemesio G. C. ‘Interpretive Authenticity: Performances, Versions, and Ontology.’ Estetika 59 (2022): 135–52. DOI: 

  18. Ross, Stephanie, and Jennifer Judkins. ‘Conducting and Musical Interpretation.’ British Journal of Aesthetics 36 (1996): 16–29. DOI: 

  19. Roth, Abraham Sesshu. ‘Shared Agency.’ In Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta, summer 2017 ed. Stanford University, 1997–. 

comments powered by Disqus