THE SPOKEN WORK
This is a preprint of an article accepted for publication in Journal of Aesthetics and Art Criticism © 2004 American Society for Aesthetics
Jerrold Levinson defends the curious view that musical works are what he calls “indicated sound/ performance-means structures,” hybrid entities consisting of two sorts of abstract structural types as well as individual composers and times.[1] Levinson’s theory can be viewed as an attempt reconcile the fact a single musical work can be performed on a number of distinct occasions with the (putative) fact that musical works are created, and hence do not pre-exist compositional activity.[2] Taking musical works to consist in part of structural types enables Levinson to model the work/ performance relation on the type/ token relation. And taking works to consist in part of composers and times ensures that they come into existence as a result of (and only after) the compositional process.
The central difficulty with Levinson’s view is that taking musical works to be in part constituted by composers and times does not suffice for their being created in a suitably robust sense. The source of this difficulty is the type/token model of the work/ performance relationship – a model which renders the problem of composer creation intractable. David Kaplan, however, has offered an account of the nature of words that might be of help here.[3] According to Kaplan, the word/ occurrence relation is better modeled on the continuant/ stage relation that the type/ token relation. And what I mean to argue is that a similar theory of musical works can reconcile the competing demands facing any adequate account of works.
I: Creation on the Cheap
Before subjecting Levinson’s account of musical works to critical scrutiny, it is worth pausing for a moment to consider exactly what an indicated sound/ performance-means structure is. A sound structure consists of individual sound types (abstractly) arranged in a certain way.[4] A simple arrangement might be a temporal sequence of sound types. A performance-means structure consists of types of sound producing activities similarly arranged. A sound/ performance-means structure (or “S/PM structure”) is an amalgam of two such structures. It consists of an arrangement of pairs of sound types and sound producing activity types. An instance of an S/PM structure occurs when (i) a token of each sound type in the structure is produced by a token of the sound producing activity with which it is paired and (ii) the sound tokens occur in (or form) a pattern corresponding to the arrangement of sound types in the S/PM structure.
Indicated S/PM structures are entities that are brought into existence by means of compositional acts. A compositional act is one in which a composer indicates an S/PM structure by means of “…producing an exemplar of the structure involved, or a blueprint of it.”[5] And the art object brought into existence by this activity, the musical work, is a hybrid entity consisting of the S/PM structure, the composer, and the time – an “S/PM structure-as-indicated-by-X-at-t”.[6] An instance of a musical work W is a sound event which is an instance of its constituent S/PM structure that stands in a more or less direct intentional or causal connection to the composer’s creative activity.[7] And a performance of W is a sound event which is intended to be an instance of W and which “succeeds to a reasonable degree.”[8]
One central reason Levinson offers for preferring his view to one according to which musical works are pure sound (or S/PM) structures is that musical works are composer creations, brought into existence by compositional activity, and as such cannot pre-exist such activity. After all, indicated S/PM structures, unlike pure sound structures, come into existence only when a composer indicates a (pure) S/PM structure. Now, although I will not argue for it here, I think the thesis that musical works are composer creations is eminently plausible. As a result, for present purpose I am just going to take it for granted. But what I want to point out is that the fact that something, X, is brought into existence by someone, A’s, activity does not suffice to make X a creation of A’s. Moreover, I want to argue that indicated S/PM structures are just not the kinds of entities that can count as creations in any non-trivial sense.
Suppose, for
example, that I finish reading A Pair of
Blue Eyes this evening. As a result of such activity, something new will
have come into existence: A Pair of Blue
Eyes-as-read-by-Alward-on
It might be objected, however, that the reason indicated S/PM structures count as creations while these other facts do not is because, unlike the others, they are the products of creative activity. But an artist, through her creative activity, will bring into existence all sorts of new facts – coming to stand in new relations to all sorts of objects, old and new. And while there is little temptation to take her to have produced such a plethora of distinct art objects, there is also little ground for choosing one such fact as the unique product of her creative activity. Moreover, suppose a sculptor were to produce, for example, a bronze sphere. By producing an exemplar of Sphericality, she would have come to stand in the indicating relation to Sphericality and thereby brought into existence the corresponding fact – an “indicated shape” if you like. But this fact would not be the art object – the sphere would. By parity of reasoning, we ought to be suspicious of the claim that the product of composer creative activity is an indicated S/PM structure, as opposed to the sound event or score by whose production, as Levinson would have it, the composer indicates some pure S/PM structure of other.
The culprit here – what gets Levinson into trouble – is the type/ token model of the work/ performance relationship. This model requires taking musical works to be constituted, at least in part, by sound structures types. And such types pre-exist compositional activity. Unless a different model of the work/ performance relation is forthcoming, a theoretical treatment of musical works which vindicates their status as composer creations will be hard to come by.
II: Common Currency Words
David Kaplan has defended what he calls the “common currency conception” of words. In the same way that a single musical work can have a multiplicity of performances, a single word can have many different occurrences – it can be spoken or written on numerous occasions. The common currency conception offers a continuant/ stage model of the word/ occurrence relation, in contrast to the type/token model proffered by the competing “orthographic conception.” Words are conceived of as concrete, enduring constituents of the natural world. They are, in effect, complex networks of causally related events. And occurrences of words are the individual events that make up these networks.
Let me elaborate. Underlying the common currency conception of words is Kripke’s causal historical theory of referring.[9] Recall: on Kripke’s view, when a speaker uses a name her action stands at the end of a chain of appropriately causally linked events that began when some individual acquired a name. And the referent of the speaker’s use of the name is the individual who was named in the event which initiated the causal chain in question. On Kaplan’s view, however, a Kripkean naming ceremony is also a word-originating event – it brings into existence a new common currency name.[10] The written or spoken token produced in this event is (typically) perceived by various people and stored in memory. These people will in turn produce tokens of the remembered perceptual input which will be perceived by various other people. And so on. The word – as opposed to any one of its occurrences – is the whole big causal network itself.
In order for a speaker’s linguistic output to count as an occurrence of the same word as some prior perceived input, the input and the output have to stand in an appropriate causal relation to one another. On Kaplan’s view, all that is required here is that the output be produced with the intention to repeat – imitate or standardize (by the speaker’s own standards) – the perceived input.[11] As result, the input and the output need not stand in any interesting similarity relation to one another, as long as the differences between them are due to misperception (mishearing, e.g.) or misspeaking.[12] This feature of Kaplan’s theory does result in some thorny difficulties. For example, it runs afoul of the role words play in linguistic communication.[13] But it is important to note that this feature is inessential to the common currency conception. One could endorse the basic picture but put more stringent (or at least different) requirements on the appropriate sort of causal relation. What is important to note for our purposes, however, is how the general picture reconciles the fact that individual words have multiple occurrences with their (putative) creation in Kripkean naming ceremonies.
III: Common Currency Works
In order to provide a common currency theory of musical works – according to which works are networks of causally linked events – what is needed is (i) an account of work-initiating events and (ii) an account of the requisite causal link between various occurrences of a single work. It is important to note, however, a disanalogy between words and works. Words have both written and spoken occurrences which can be perceived and stored in memory.[14] Musical works can be performed and transcribed, but only performances count as genuine occurrences of the work. Transcription is better understood as a written analogue of memory than as a type of occurrence of a work.
The work-initiating events, in my view, are more or less Levinson’s indicating acts – acts in which a composer produces an exemplar or a blueprint of a given S/PM structure. That is to say, they are either sound events or events in which a score is produced. Of course, not any old sound or score-producing event will do. First, presumably the product of the event has to be intended: the accidental production of a sequence of sounds or (less likely) of inscriptions which constitute a score does not result in a musical work. Second, in order to distinguish between a finished work and a draft, the composer must have an appropriate psychological attitude towards the sound-event or score she has produced. She must think of it as completed or finished; that is, she must be in a state of mind towards it that she would express by saying things such as “That’s it!” or “That’s as good as I can make it.”[15] Third, we might, more contentiously, also require that, in order to be work-initiating, an event be the culmination of a creative process of a specific sort. The point of this restriction would be to distinguish between work-initiating events on the one hand and performances and transcriptions on the over. In any event, the sound sequence or score produced in the work-initiating event is, in the most fundamental sense, what the composer has created; it is the aspect (or “stage”) of the musical work that is under the greatest degree of composer control.
There are a number of different causal routes from a work-initiating event to putative performances of that work. If the work-initiating event was a score-producing event, a performer might (i) attempt to produce a sound-event that conforms to or is based on the score or (ii) attempt to produce a sound-event that conforms to or is based on a score derived (in some suitable way) from the original. And if the work-initiating event was a sound-event, (iii) someone present during the original event or someone might attempt to produce a sound-event that conforms to or is based on her memory of the original sound event, (iv) someone who has heard a recording of the original event might attempt to produce a sound-event that conforms to or is based on the recording, or (v) someone might try to produce a sound-event that conforms to or is based on a transcription of the original sound event. Before discussing these various causal pathways, a few comments are in order. First, it is worth emphasizing that these processes can iterate in various combinations. One might, for example, produce a sound event that is based on a recording of a sound event that that conforms to a transcription of a work-initiating sound event. Second, my talk of a sound event being based on a score or prior sound event rather than conforming to it is intended to capture cases in which the performer is attempting to offer a different interpretation or a new arrangement of the work she is performing. And third, in many (if not most) cases, the causal pathway from perceptual uptake of a sound-event or recording or score to a performance involves the more or less complex process of learning how to perform the work in question.
Let P1 and P2 denote two distinct musical performances. And let us suppose that the following sequence of events occurs: P1 occurs; information regarding the sound/ performance-means structure instantiated by P1 is recorded; this information is preserved over time; the preserved information is retrieved and used in the production of P2. Our concern here is with the nature of the relation that must obtain between P1 and P2 in order for them to be performances of the same musical work. I will proceed here by first considering an ideal case in which it is uncontroversial that P1 and P2 are performances of the same work, and then considering a number of dimensions of deviation from ideality. In the ideal case, (i) there is no breakdown in the recording, preservation, or retrieval of the information about the S/PM structure of P1, (ii) the performance-intention underlying P2 is an S/PM structure that exactly conforms to the retrieved information, and (iii) the performance-intention underlying P2 is achieved.
One deviation from ideality we can tolerate is a failure to fully meet condition (iii). As noted above, even Levinson will count a sound event which fails to be an instance of the requisite S/PM structure as a performance of a musical work as long as it is intended to be an instance of that structure and “succeeds to a reasonable degree.” If the odd blown note was sufficient to undermine the claim to be performance of a given work, most works would be performed far less frequently than we think they are. In addition, I think we can tolerate minor failures with respect to (i). The fact that the original transcription of P1 was off by a note or two, or a recording of it has undergone minor decay over time, does not prevent P2 from being a performance of the same work.[16]
The more interesting deviations from ideality, however, are the intended ones (in contrast to the accidents heretofore mentioned). What I have in mind are failures to meet condition (ii); that is, cases in which the performance-intention is to produce a sound event which does not exactly conform to (the retrieved information about) the S/PM structure of P1. I want to suggest that despite the differences between P1 and P2 which such performance-intentions will typically (and are designed to) yield, on some occasions they will remain the same work nonetheless. What is needed, of course, is a theory of distinction between new arrangements/ interpretations of old works on the one hand and new but derivative works on the other. I am not going to offer such a theory here, but I will make a few suggestions. First, a necessary condition for being performances of the same work is that P1 and P2 be similar in important respects. Minimally this requires that they share central melody lines. A second more contentious condition might be the requirement of a certain psychological state on the part of the performers of P2, perhaps along the lines of an arranging (or interpreting) intention as opposed to a creating-intention. A lot more needs to be said, but this will have to do for present purposes.
There are a couple of upshots of the picture of musical works sketched here that need to be brought to the fore. First, the cumulative effect of permissible differences between “adjacent” performances along a given branch in the causal network that constitutes a work can yield performances of it that are radically dissimilar from the original performance or score. This is, of course, compatible with there being at the same time performances nearly identical to the original. And second, on the picture musical works can grow and evolve in a way that is largely out of the composer’s control. Although the composer creates the work, in a sense it subsequently comes to have a life of its own. One might be tempted to take only the work-originating event to be the work, but this would sever the link between a work and its performances.
IV: Applications
An important virtue of the common currency conception of musical works is its ability to reconcile the fact of composer creation with the fact that works can have multiple performances. What remains to be shown is that it can adequately handle a broad range of problems that arise for musical works. Consider, for example, Levinson’s Principle of Fine Individuation:
Musical works must be such that composers composing in different musico-historical contexts who determine identical sound structure invariably compose distinct musical works.[17]
Levinson argues that the aesthetic and artistic properties of a musical work are in part determined by the composer’s musico-historical context. For example, Levinson claims that a Stamitz symphony containing a “Manheim rocket” – a novel device in Stamitz’s day – is an exciting musical work, but a work written today with the same sound structure would be funny rather than exciting.[18] Since the Stamitz symphony and the (hypothetical) contemporary sound-identical work differ in properties, it follows from the indiscernibility of identicals that they are distinct musical works.
As should be obvious, the common currency conception satisfies the principle of fine individuation. Works produced by distinct composers always have distinct initiating events and, as a result, are distinct works, whether or not the sound sequences are the same. Moreover, the common currency conception has the advantage over Levinson’s alternative of allowing musical works to undergo (and survive) changes in qualities over time. Kivy, for example, has argued that if the judgement that a musical work is exciting is understood as the judgement that it is exciting to its audience, Stamitz symphonies were once but are no longer exciting.[19] Levinson has attempted to avoid this result by suggesting that excitement is correctly understood as excitingness not to any old audience but only to an audience that hears it correctly, “…in a way that reflects its provenance and musico-historical position.”[20] The trouble with Levinson’s manoeuvre here is that even a contemporary audience that hears a Stamitz symphony correctly is unlikely to be literally excited by it. They will, however, be well-placed to judge that it was likely exciting to the original audiences that heard it correctly. For the common currency conception, such changes in the properties of musical works are no more problematic than the changes a person undergoes between the various stages of her life. Just as a person can be energetic as a child but lazy as an adult[21], a common currency work can be exciting in its infancy and boring in the later stages of its existence.
Similar considerations can be brought to bear with respect to Levinson’s Principle of Inclusion of Performance Means:
Musical works must be such that specific means of performance or sound production are integral to them.[22]
A problem that arises for this principle is that it seems to imply the counter-intuitive result that a re-arrangement of a musical work on different instruments yields a distinct albeit derivative musical work.[23] Levinson himself simply bit the bullet on this particular point.[24] The common currency conception, however, can satisfy this principle while at the same time allowing that a re-arrangement of a work remains the same work. After all, being integral to the work-initiating event suffices for being integral to a common currency work. And any worries about the effect of re-arrangement on the aesthetic/ artistic qualities of a work can be alleviated as above by pointing out that a work can have different properties during different stages in its existence.
Finally, the common currency conception handles problems posed by improvisation better than its Platonistic competitors. Wolterstorff[25] argues that musical works are “norm-kinds” in that they can have properly and improperly formed examples. As a result, composing a work requires specifying a set of “requirements for correctness.” But in improvisation a performer does not select which of her performance is essential and which merely accidental. As a result, Wolterstorff concludes that “to improvise is not to compose.” As Kivy points out, this implies that if another person were to write down what the performer had improvised, this latter person would be the composer of the resultant work.[26] The common currency conception, however, easily evades this difficulty. An improvisation is a work-initiating event. Any subsequent event that stands in the right sort of causal relation to it is a distinct stage of the same work. And standing in the right kind of relation to the improvisation does not require that criteria of correctness have been selected in any sense.
V: The Spoken Work
It is perhaps surprising that the literature on the nature of words would prove so fruitful a source of insight into the nature of musical works. Kaplan’s motivation for developing his account of words was in part to avoid a Platonistic alternative. I have argued elsewhere that this is something of a red herring. The real issue is whether the type/ token model is the best model of the word/ occurrence relationship. A positive answer to this question need not commit one to Platonism regarding types. The same thing is true of musical works. Despite Kivy and Levinson’s adherence to it, Platonism is beside the point. The issue is what the best model of the work/ performance relationship is. And, as I have argued here, the continuant/ stage model clearly wins the day.
Peter Alward
Department of Philosophy
peter.alward@uleth.ca
[1] Jerrold Levinson, “What a Musical Work Is,” The Journal of Philosophy 77, 1980. Reprinted in Jerrold Levinson, Music, Art and Metaphysics, Ithaca: Cornell University Press, 1990, pp. 63-88. All page references are to the latter.
[2] But see Peter Kivy, “Platonism in Music: Another Kind of Defense,” in The Fine Art of Repetition, Cambridge University Press, 1993, pp. 35-74.
[3] David Kaplan, “Words”, Proceedings of the Aristotelian Society: Supplement, v. 64, 1990, pp. 93-119.
[4] By an arrangement I mean an abstract mathematical structure, like a set or sequence.
[5] Levinson (1990), p. 81.
[6] Levinson (1990), p. 79.
[7] “[There] must also be some connection, more or less direct, between the sound event produced and A’s creative activity. Whether this is primarily an intentional or causal connection is a difficult question…” Levinson (1990), p.85.
[8] Levinson (1990), p. 86.
[9] Saul Kripke, Naming and Necessity, Harvard University Press and Basil Blackwell, 1972, 1980.
[10] Kaplan distinguishes between common currency names and generic names. In addition to my common currency name “Peter” and Peter Lorre’s name “Peter”, there is a generic name “Peter”, which is a distinct word from the former two. When one uses a pre-existing generic name in a Kripkean naming ceremony, one creates a new common currency name. Kaplan (1990), p. 111.
[11] Kaplan (1990), p. 104.
[12] Kaplan (1990), p. 105.
[13] See, e.g., Herman Cappelen, “Intentions in Words”, Nous, 33:1, 1999, 92-102 and Peter Alward, “Between the Lines of Age: Reflections on the Metaphysics of Words,” <http://home.uleth.ca/~peter.alward/papers/words.htm>.
[14] Of course, one can in addition keep a written record of both written and spoken occurrences of words. In fact, this is exactly what dictionaries are used for.
[15] Of course, a composer might decide only sometime after the fact that her composition was finished.
[16] We might, however, require that when such deviations occur the causal chain terminates; that is, we might stipulate any performance P3 based on a deviant P2 is not a performance of the same work as P1. One reason to balk at this is because if P1 is a performance of the same work as P2 and P2 is a performance of the same work as P3, then transitivity might drive us to conclude that P1 is automatically a performance of the same work as P3. Two possible responses would be to argue that P3 is the performance of no work at all or that P2 is at the same time the performance of two distinct works.
[17] Levinson (1990), p. 73.
[18] Levinson (1990), p. 71.
[19] Kivy (1993), p. 64.
[20] Levinson (1990), p. 226.
[21] A better example might be a man whose earring twenty years (or maybe thirty) ago made him daring and original, but now leaves him ordinary and relatively unhip.
[22] Levinson (1990), p. 78.
[23] Peter Kivy, “Orchestrating Platonism,” in The Fine Art of Repetition, Cambridge University Press, 1993, pp. 75-94.
[24] Levinson (1990), p. 234.
[25] Nicholas Wolterstorff, Works and Worlds of Art, (Oxford: The Clarendon Press, 1980) pp. 55-64.
[26] Kivy (1993), pp. 53-4. It is worth noting that Kivy offers a Platonistic solution to this difficulty, although one that I do not kind entirely persuasive.