Aesthetics for Birds

Aesthetics and Philosophy of Art for Everyone

Allographic Artworks & Their Instances


What follows is a guest post by Lee Walters. Lee is a Lecturer in Philosophy at the University of Southampton where he teaches courses on metaphysics, philosophy of language, and fiction and fictionalism. His research focuses on the metaphysics of art; fictional characters and their names; and the logic and semantics of conditionals.

It is widely, though not universally, accepted that there can be distinct allographic artworks (e.g. pieces of classical music, and novels) that are indistinguishable with respect to their intrinsic, structural or notational properties. What consequences does this claim about works have for an allographic work’s instances? In particular, can there be intrinsically indiscernible instances which fail to be instances of the same work? More generally, when is x an instance of some particular work, w?

Jerrold Levinson thinks that it follows from the fact that we can have distinct structurally indiscernible works (multiple works with the same word or sound/performance means structure) that we can have intrinsically indiscernible instances which fail to be instances of the same work:

Given a sonnet or sonata is a structure-as-indicated-by-X-at-t (i.e., an indicated structure) the instances of such a type are more than (though they must also be) instances of the associated pure word or S/PM structure. In order to be an instance of the indicated structure, an instance of the pure structure must in addition be related in a certain way to the dated act of indication (determination, fixation) of the pure structure by composer or poet which brings the sonata or sonnet into existence (Levinson 1980: 98).

And this same, natural inference is seemingly made by a number of other authors. Here is Greg Currie explicitly making a similar inference:

in reading the text [i.e. the word sequence] of a work one is not necessarily reading an instance of that work … And this is merely a consequence of my claim that a work is not exhausted by its text (Currie 1989: 123).

It is consistent with what Greg says above that the text is both an instance of w and of a distinct work w*, but that one is reading only one of these works. It is clear, however, from his discussion that he thinks this text can fail to be an instance of one of two structurally indiscernible works because he thinks that x can only be an instance of w if it stands in some causal relation to w’s creation.

My colleague Chris Janaway also makes this inference I think:

Goodman thinks exact compliance within a given notation is necessary and sufficient for a piece of music or poem to count as the work itself, or for a performance to count as a performance of the work itself. To say that such exactness is necessary is, I think, unrealistically strict … But what about the idea that exact compliance in some definite notation is sufficient for identity of the work [i.e. sufficient for x to be an instance of w]? What this entails is that history of production is irrelevant to the identity of a musical or literary work (Janaway 1997: 7-8).

Finally, Amie Thomasson, accepting draws the same conclusion:

books of the same text type may fail to be copies of the same composition. If, for example, by mere coincidence two authors compose identically worded tales, we may have books of the same text type, but two different compositions (Thomasson 1999: 64).

Now although the above inference is a fairly natural one, it is invalid, as Peter Lamarque (2010: 82-83) notes. To see this, consider the kind of Goodmanian individuation of allographic works that the authors above reject, which says that any structurally/notationally/intrinsically indiscernible allographic works are identical:

Now what Levinson and the others authors conclude on the basis of rejecting Goodman’s individuation of allographic works is that an analogous principle regarding the instances of allographic works is also false. In particular, they reject that structurally/notationally/intrinsically indiscernible instances of allographic works are instances of the same work(s):

It is quite easy to see, however, that the principle regarding instances can be maintained even if we reject Goodman’s individuation of allographic works. Just to hammer the point home, here is a model which demonstrates this:

My purpose here is not to criticize the above philosophers for making this invalid inference, since it is promoted by the correct thought that, at least in some cases, “instances of compositions are propagated through copying texts” (Thomasson 1999: 64). Moreover, I have been guilty of something like the above inference myself (Walters 2013: 463-464). The point, rather, is to ask, given that there is no smooth transition from the individuation of allographic artworks to the negative claim that conformity with w’s notation is not sufficient for being an instance of w, let alone for the positive claims found in Levinson, Currie, Thomasson, and Walters that some kind of (causal) connection is required for being an instance of w, what does determine the instances of an allographic work?

Having noted the invalidity of inference above, Lamarque considers when x is an instance of w. We can, I think, find the materials for three different answers in Lamarque’s discussion. Before delineating the three accounts, we should note that his discussion, drawing on Janaway’s (1997), starts from the claim that “Any given token of the Don-Quixote-text-type … can count as a type-instantiating token of either Cervantes’ or Menard’s work” (Lamarque 2010: 91-92). That is, as long as a text has the correct word structure it can count as a copy of Don Quixote. Similar things can be said about scores and musical performances.

The first account of the instances of allographic works found in Lamarque the Goodmanian answer considered above, namely that any instance that conforms with a work’s notation is an instance of that work: “What then is a copy of a poem? It is a token of a text-structure-type. Identical text-structure-types [such as Cervantes’ and Menard’s distinct, but indiscernible works] share identical tokens” (Lamarque 2010: 93).

Note that this view does not follow from Lamarque’s starting point that any token of a text-type can count as an instance of a work with that text type: that something could be an instance does not show that it isan instance. Moreover, this view of a work’s instances generates two related oddities. First, this view is committed to saying that x can be a token of a work the work has been created. Similarly, assuming that a repeatable artwork is destroyed when all traces of it are destroyed, the current view implies that there could be instances of a work which no longer exists, since there could be instances of the relevant text-type even after all traces of a work have been destroyed. For these reasons, then, I think we should grant the negative claim that conformity with a work’s notation is not sufficient for being an instance of that work, even though this claim does not follow from the fine-grained individuation of allographs we are assuming.

The obvious fix to the above type of account is to ask when does an instance of a text-type count as an instance of a work-type? There seem to be two non-equivalent answers here. Janaway (1997: 10) says that an instance of w’s text-type is an instance of w when it is accompanied by an instruction to interpret it as an instance of w. Perhaps this is what Lamarque has in mind when he writes that “A copy of a poem can be thought of as a route to the work. By attending to the copy, supplemented with an instruction, implicit or explicit, to read it as a copy of a particular work a reader comes to appreciate the work.” (Lamarque 2010: 93).

An alternative given by Lamarque (and perhaps this is what he meant in any case in the quote above) is that an instance of w’s text-type is an instance of w when it is interpreted as an instance of w: “What determines the matter [of whether a token of a text of a work] is a simple, unconstrained, decision made by the person reading the text … The underlying truth remains that it is always a matter of decision (by the reader) how a text is read. ” (Lamarque 2010: 86).

Now obviously, these two accounts can come apart, since one can be issued with an instruction to interpret x as an instance of w, but fail to do so, and in the other direction, one can interpret x as an instance of w and yet not be issued an instruction to do so. But let us not pursue how exactly we think about these subtleties. Rather, let’s ask what can be said in favour of these accounts from Janaway and Lamarque.

The key idea, I take it, is that in the relevant circumstances, any x which conforms with w’s notation gives us access to w, and so we are not missing anything by attending to x as opposed to some instance which is causally connected to w’s creation:

If a devotee of Black is keen to engage with Black’s poem but has no copy of it to hand, only a copy of White’s volume, then he might well – in the full knowledge that the works are distinct if textually identical –  read the text in White’s volume as the text of Black’s poem. A conscious decision of this kind could not be counted as a mistake or a misreading. Nor would it in any way detract from a full appreciation of the work. Of course that is not to say that a mistake could not be made. A reader who intended to read the text as a poem by White and believed that this was White’s poem but who identified in it aesthetic and artistic properties associated with Black’s poem has misread the poem.” (Lamarque 2010: 86)

As tempting as the thought is, it is a commonplace that we can access the properties of one thing by examining (an instance of) another. Indeed, Lamarque admits that non-instances can provide access to the aesthetic properties of artworks:

You haven’t see the Mona Lisa if you haven’t been to the Louvre. … (Incidentally, you might reasonably claim to be knowledgeable about the painting and even have an appreciation of its beauty and power through studying reproductions alone.) (Lamarque 2010: 59)

As well as reproductions, photos of objects give us access to them. And depending on one’s views, one might think that we can learn about humans/persons by studying distinct things, i.e. dead bodies, and that we can learn about an alien mammalian species, the schmiger, by studying tigers. Similarly, I could learn about the footprints of dinosaurs by studying recreations of them which, because not caused by dinosaur feet, are not themselves dinosaur footprints. If this is correct, then why could one not have access to a work’s properties by having an instance of the work’s text-type and knowing facts about the generation of the work, without the instance of the text-type itself being an instance of that work?  Nevertheless, there is, I think, something very intuitive about Janaway and Lamarque’s claims that in the appropriate circumstances any instance of w’s text-type can serve perfectly well as an instance of w. (We should note that even if Janaway and Lamarque are correct, they only provide sufficient conditions for being an instance of w. And presumably being an instance of w’s text-type is not required for being an instance of w, for there can be misprints and imperfect performances).

So although the kind of causal picture advocated by Levinson, Currie, Thomasson, and Walters seems very intuitive once we embrace the fact that there can be distinct yet structurally indiscernible works, there is no inference from the latter to the former. Moreover, once we consider treating a copy of w* as a copy of w, there is something plausible about saying the copy of w* is thereby a copy of w. But nonetheless, the type of argument given by Janaway and Lamarque falls short of establishing this.

By this stage, you might be thinking who cares? The discussion has moved on quite some way from common sense thoughts about repeatable and allographic artworks and the kinds of copying which is usually involved in propagating them. The issue, one might think, is purely of academic interest and nothing turns on it, we could choose either option. Now whilst there is something to be said for this way of thinking, it is, I think, an overreaction.

One may think, as Levinson, Janaway, and Lamarque seem to, that resolving what counts as an instance of a work helps us to get clear on the autographic/allographic distinction. This is a red herring, but I shall have to argue for that on another occasion. Rather than issues to do with forgery, my interest in works and their instances is generated by the metaphysics of the works: what kind of entities are they? To be sure, this question might only appeal to a certain kind of philosopher, but there are metaphysical issues here which can only be resolved by taking a stand on the instances of allographic artworks.

To see this, note that it is plausible that a work survives as long as there is some instance of it. So the existence condition of a work presupposes what counts as an instance of that work. Relatedly, Thomasson (1999: 65) provides the following plausible two-level criterion of identity for fictions, where f(x) is read as ‘the fiction of which x is a token’:

Now although I have some quibbles with this precise formulation which are orthogonal to the discussion above, we can see that Thomasson’s whole approach is incorrect if Janaway and Lamarque are correct about what counts as an instance of a fiction since there will be instances of a fiction which stand in no copying relation to other instances of that fiction. Moreover, if Janaway and Lamarque are correct, it is hard to see how we could give a two-level criterion of identity for works which is ‘quasi-reductive’ in Hawthorne and Lepore’s (2011: 479) sense of utilising a relation on the right-hand side of the criterion which is capable of being grasped independently of the domain for which one is providing the second-level criterion.

I am still attracted to the causal account of instances which I advocated in my 2013 (although the thrust of the paper does not rely on that claim), but I agree with Lamarque that “This is a problem not sufficiently addressed … by those who hold that distinct musical and literary works can be constituted by identical sound- or word-structures” (2010: 79).


Currie, G. 1989: An Ontology of Art. London: Macmillan.
Hawthorne, J. and Lepore, E. 2011: On Words. Journal of Philosophy CVIII: 447-485.
Lamarque, P. 2010: Work and Object: Explorations in the Metaphysics of Art. Oxford: Oxford University Press.
Levinson, J. 1980a: Autographic and Allographic Art Revisited. Philosophical Studies 38: 367-383. Reprinted in Music, Metaphysics & Art: Essays in Philosophical Aesthetics. Oxford: Oxford University Press.
Janaway, C. 1997: Two Kinds of Artistic Duplication. British Journal of Aesthetics 37: 1-14.
Thomasson, A. 1999: Metaphysics and Fiction. Cambridge: Cambridge University Press.
Walters, L. 2013: Repeatable Artworks as Created Types. British Journal of Aesthetics 53: 461-477.


  1. The following comment is from Donald Brook:

    It would be easier to negotiate one’s way through the intricacies of Lee Walters’ paper and much of its cognate literature if the criteria of identity and individuation applicable to those particulars (or instances) that we take to be autographic and those that we take to be allographic were clearly and consistently divided into two sorts.

    We are interested in the way particulars or instances of all sorts should be identified and individuated as particulars or instances, and we are also interested in the way in which the restricted set of particulars or instances that we distinguish as works of art should be identified and individuated as works of art.

    There are of course particulars or instances of a general sort that do not qualify as works of art, to which the autographic/allographic distinction is relevant. Such things as Newton's Laws of motion, Microsoft’s trademark and the argument that I am now constructing, are obvious examples. It has occasionally been suggested that just one criterion of identity (although not of individuation) for works of art is that they are all in principle allographic entities. While this might seem to be a comfortable enough hypothesis in its application to the recognition of such things as poems and songs (whether or not as works of art) it raises familiar problems about the identification of marble carvings and oil paintings as works of art.

    I do not aim either to dissolve these problems or to argue that their ineradicability refutes all allographic hypotheses about the way in which particular work of art are identifiable, as such. The point of my intervention is only to exhort those who theorize in this domain to display, for the benefit of rest of us, a clear and consistently maintained sense of the difference between the relevance of the autographic/allographic distinction to particular or instantiated entities in general and its application to those particular or instantiated entities that we deem to qualify additionally (on some other ground) as works of art.

  2. Thanks for your comment Donald. I’m not sure what you main point is though. It is clear that I am only talking about artworks in my post. Of course, I agree that one might want a metaphysics of types that covered more than artworks and that much of what can be said about artworks can be said about other created types, such as makes of car, and other kinds of type, such as species. And yes I agree that the autographic/allographic distinction applies more widely that simply to artworks. Indeed, see the Janaway paper cited above for a brief discussion of the distinction in relation to types of car.


    It might be interesting to pursue the issue further, Lee, but it would probably take the discussion too far away from the specifics of your paper to be of general interest. Broadly (as to my “main point”) I do not take works of art to be tokens of a type as you evidently do. I take them to be items of a kind, on the understanding that the successive items of a kind, unlike the tokens of a type, are generically related one to another in an evolutionary way. I take the generic relationship to be memetically propelled in the case of the items of cultural kinds, in strict analogy with the way in which the items of biological kinds are genetically propelled.

    In case you are interested, the briefest version of my account of these matters to which I can easily refer you may be in a little paper called “Experimental Art,” available at:

    I think, though, that a pursuit of these question about art and about works of art in the direction in which I would like to see it go may not count as a helpfully constructive response to most of the particularities of your own essay.

  4. Thanks Donald! I think there is more agreement here than you realise. Although I employ the type/token distinction, I identify the works themselves with the types, not the tokens. The tokens are the instances of the work, for example, the individual copies of a novel or the performances of a work. More importantly, although I think works are types, I do not think such types are individuated in purely qualitative terms like Wolterstorff and Dodd, but are individuated in terms of their origins (I think the fact that you think of works as kinds and I think of them as types, might just be a verbal point). It is then a natural thought to say that such types are propagated through chains of reproduction in a way which I think is similar to your ideas of evolutionary or memetic propagation. I am sympathetic to this idea and have defended it myself in the above cited paper. But as I try to spell out in the post above, this natural thought about propagation does not follow from the individuation of types I favour. So how instances of works are propagated is still an open question that needs more attention.

    So in summary, I too think that artworks and other artifacts are analogous to species, but at least in my own case, this position has not been argued for sufficiently. But I’ll be sure to check out your own work – thanks for the reference.

  5. [LEE: For some reason, Donald Brook has been unable to post any comments from his computer to the blog, and so I've been relaying his comments to here instead. If anyone else has experienced vanishing comments or similar troubles, let me know. I'm on the case and hope to have the issue resolved shortly. Until then, here's Donald!]


    So maybe I'm not forcing your train of thought off its rails after all.

    A way of grasping what (if anything) is in dispute is opened up by your declaration that you identify works of art with types, not with tokens. Contrastingly, although perhaps not necessarily contradictorily, I identify works of art with items of a cultural kind and not at all with the cultural kinds of which they are the items.

    It seems to me that the relevance of evolutionary processes to the elucidation of the histories of both cultural kinds and of biological kinds is not explicit in your story. The suggestion that it is nevertheless implicit is difficult to reconcile with your identification of works of art with types and not with tokens. I take it that a particular lion is an item of the lion biological kind, just as the colour-inflected canvas called “Blue Poles” (of which the insurance value turns on its recognition as an autographic work of art) is an item of the American-abstract-expressionist-painting cultural kind.

    In both cases the kinds evolve, the items do not. They have largely accidental vicissitudes. American-abstract-expressionist-painting is a cultural kind. It is not a work of art and it does have a hisory. “Blue Poles” is an item of this cultural kind. Unlike the kind of which it is an item it is a work of art; and as such it does not have a history.

  6. I only identify repeatable artworks with types: the novel Perfume is not identical to any copy of that book, rather it is a type of which the individual copies are tokens. In the case of non-repeatable artworks, such as paintings, I'm happy to identify them with material objects (which may fall under non-work types of the kind you have mentioned).

    But I'll need to make a study of your work before I can discern on what we agree and disagree. And I look forward to doing so.

  7. Thanks Lee and Donald,
    Very interesting discussion. I think I see where you part ways with one another though. The issue arises because we can't coherently say that prints and photographs are tokens of a type whilst also holding that paintings are a type. We have to look higher up the chain to determine the point where the many disparate entities we call artworks share something in common. What we don't find there though, despite years of diligent enquiry, is some essential feature that inheres in all artworks that we could finally call art. Some (perhaps including Christy) would like to say that intentions or reasons lie at the heart of the story and in a sense that must be true. But what we get when we make intention or reasons the crucial criterion is something less than art — something more like craft. If art were always intentional then there could never be unexpected discoveries of genuinely new ways of doing things. But we can't make the discovery of genuinely new ways of doing things the defining criterion of artworks because revelatory discovery is not the exclusive province of the artworld. Furthermore, artworks are often not the product of genuinely new ways of doing things. So, the best way of defining artworks (but not art) is that provided by the institutional theory.

Leave a Reply

Required fields are marked *.