by Jos Kunst (1987)
This is an electronic version of an article published in Interface, Vol. 16 (1987) pp. 1-11. Journal of New Music Research, the continuation of Interface, is available online at http://journalsonline.tandf.co.uk/openurl.asp?genre=journal&issn=0929-8215.
Abstract. A number of different approaches to analysis are reviewed and compared as to their status with respect to the scientificity some of them lay claim to. Moreover, a plea is made for a branch of musicology that would be not only empirical in the weak sense of "taking account of" reality, but also in the stronger sense of incurring risks, of making theoretical predictions, being falsified by the "resistance" of reality, refining itself, and, in this way, giving rise to a new kind of growth in musicological knowledge.
Let us begin by squarely admitting that anything even remotely verifiable and purported to be true of some given instance of music must count as analysis. Cruxes are therefore (a) in what counts as a verifiable statement about music and (b) in what counts as a given instance of music. We will argue these two issues to be related.
The apparent boldness of this statement is meant to achieve generality: those readers who would prefer a narrower definition of the analytical genre should still take seriously the reflections on it we are going to offer, or else they should be in need of arguments showing that our definition of the field, for all its broadness, is still too narrow in some respect.
I take it that most analysis is conducted orally, viz., in the classrooms where music is taught. As such, it is one of the ways the culture has of transmitting itself; as we all know, a lot of authoritarian behaviour is unavoidably bound up with activities of that kind; many (seemingly) truth-claiming statements are made not so much in order to impart truths as, rather, to mold people – in principle, we have no quarrel with that. According to theories we have been advocating in the past, the production of shared concepts, and, more precisely, the production of them in real people's minds, is one of the culture's foremost concerns; not least of that part of it commonly called the arts (in the general sense of the term, including also literature, music and the like).
One of the more pressing problems confronting us here is to find rational ways of deciding on the status of the cultural activity called analysis. Our main contention will be that analysis, as currently practiced, is an art, and that, for once, this may well be no reason for self-congratulation.
Let us review our arguments. We will bring them, for now, under three headings: first, how is conviction carried? second: why are theories (eventually) rejected? and third: what part of reality do analyses suggest or say they are describing?
How is conviction carried? Ostensibly, by constant reference to scores. Scores are cast in the role of "hard facts". This, we submit, is essentially a rhetorical trick (obviously, we do not mean that no reference should be made to scores; we mean that reference to scores should not carry conviction the way it does, at least not in a context taken to be "scientific". On the other hand, such tricks do belong to the rightful stock-in-trade of art). What we are saying is that the more one's statements become interesting, the less directly they will be grounded in "hard facts": only the most trivial assertions, such as: there are exactly 57 b-flats in this score, could possibly count as hard in a musicological context. All interesting statements are based on interpretation. Now it is well-known that the case of history itself is exactly parallel: it grows more interesting as the historical certainty of its statements decreases. Nevertheless, Bach's birth-year being 1685 is part of a network of statements, many of them interesting (and therefore interpretation-based), that would be simply wrecked should his birth-year suddenly turn out to be 1585 instead. So "hard facts" can be assigned the popperian status of negative touch-stones: even if they are hardly useful in underpinning important statements, they certainly can be used in order to attack them. Therefore, the only realistic kind of positive underpinning by (relatively) hard facts will consist in a relation of noncontradiction.
So far, so good. Analysts invoking "the score" are, it seems, somewhat easy on themselves: they propose a theory, while, in making it up, they had all the time at their disposal all and any possible material that might eventually contradict it. The essential difference with predictive theories seems to be that in the latter case a theory is and remains at risk from its conception onwards, also in its published (and therefore socialized) form; whereas in the former case it has been at risk (and may have been revised more than once as a consequence of that fact) only from its conception until its publication – unless, of course, both author and editors have been blundering. Its final completion constitutes an end point; growth of knowledge on that particular subject stops. But then, as we all know, theoretical style will change, and, we will argue in the following paragraphs, the subjects themselves are changed in the process. A 1986 article on Beethoven's opus 133 will not treat the same subject as a 1936 article on that work.
It is well-known that, given a closed set of data, such as constituted by a score, an indefinite number of theories, all mutually exclusive, can be formed in such a way that they are, all of them, consistent with these data. The real problem for the analyst is therefore not so much to come up with a theory that is not contradicted by the facts of his score, but rather to come up with a theory that is somehow right, or even: the right one. Now if we translate this as: a (or the) convincing one, our problem becomes one of rhetoric (in order to avoid over-rash conclusions, let us stress in passing that this concept obviously can in no way be reduced to mechanical or quasi-mechanical terms: any functioning rhetoric constitutes a serious and subtle problern of literary theorizing). On the other hand, there may be other kinds of rightness for theories, such as some kind of, say, correspondence to the facts; only what facts could be meant? We will come back to this later (1.3).
For the moment it is clear that the only substantial risk a published analytical theory is exposed to is that of being ignored – and it shares that lot with, again, works of art. In essence, this boils down to the idea that no analytical theory that is not contradicted by a score can ever be corrected on the basis of the "hard facts" contained in that score; a simple idea if ever there was one. This state of noncontradiction is, in itself, extremely easy to achieve; much more easy than other, more substantial, kinds of convincingness. However, some analysts, troubled by what they felt to be the arbitrariness of analytical results generally (we would say: some of the more conscientious ones) have explored the alternative of trying to increase the weight of the evidence, and have produced various ways of "blowing up", as one does with details of a photograph, the score's hard facts into their interpretation of the music. We will try to clarify what we mean by invoking the following two well-known examples.
The first one is not so much an individual concrete example, but rather something like a trend, practiced especially in relation to XXth century music, and forcefully represented by both classroom practice and books like Cogan & Escot (1976). Also Karkoschka's work (see his 1976, and also his present contribution to Interface) may, at least in part, count as representative. What we are referring to is the idea of subjecting the score to mechanical transcriptions and/or transformations, with a view of making more clear, and therefore more convincing, what, I take it, the analyst had first discovered in it by himself: because so many transformations are possible and so few are thought to be adequate for each given score, they cannot be very useful as heuristical devices, but serve well as pedagogical and expositional tools. Moreover, and this must concern us here, they bring with them, because of the exact quantifications involved, something like the disciplinary behaviour and, thereby, the flavour of science.
This could be another rhetorical trick.
But is it not more than that? Is not the arbitrary, or ideological, character of the analyst's heuristics really quite similar to the equally anarchistic and ideology-determined ways scientists have in getting their theoretical inspirations? Indeed, coming by ideas remains essentially a matter of unregimentable intuitions in all walks of life. The difference comes later, and resides, at least according to one respectable conception of science, in what can be done to test the theory. Data that have gone into finding it can no longer be used in order to test it. Expressed intuitively and perhaps somewhat naïvely, the guaranteed existence of such "fresh" data ensures that the theory in question is not about some theorist's construct modifiable at will, but about part of the outside world, about something substantial and resistant to what theorists and their audience may come to want of it. As a direct consequence, theories may be checked against new facts, be shown up to be inadequate as yet, and, by undergoing refinements, be made adequate also to the new facts; and this leads to growth of knowledge. Necessary conditions for this mechanism to get started are generalizability of theories, and repeatability of tests.
Our contention is that, for analysis as it stands, non-trivial repeatability of tests fails due to lack of generalizability (see also below, 1.2).
On the other hand, if analysis, even when using exact methods, fails to achieve scientific status, at least according to some widely accepted standards, this does not mean that the exact methods involved should be deemed useless. If the business of analysis is to help ensuring the creation and transmission of culture, exact methods may contribute a convincingness of their own. Their status, we will want to say, is then indeed rhetorical, but this does not disqualify them at all in this case; on the contrary, rhetorical tools are central ones, perhaps even the most central ones, in the culture's creation and transmission.
We must add to this picture the fact that important and interesting parts of XXth century composition are to varying degrees inspired and even informed by methods borrowed from the exact and/or natural sciences, and that on any conception of the analyst's work that sees him in a kind of intermediary role between composer and listener he will have to take on to explain these methods; analysis itself then will involve an account of the exact methods at issue, and that, in turn, can only be given in exact terms. We also feel, however, that this conception of the analyst as just a provider of what might be termed social lubricant does not wholly satisfy the historical role analysts have, in practice, come to fulfil. Their function seems indeed different from, say, the journalist's or music critic's; but what could be the essential ingredient left out in such a description of his role will have to remain, for now, an open problem – we will be offering a tentative solution below (2.1).
A related, but in some ways fundamentally different case is offered by the Nattiez analytical method as exposed in his 1975. The similarity is in the quasi-mechanical transformation applied to score material; the difference is in its ideal applicability to all and any scores. Here also we find a need to conform to scientific standards; only here the science invoked as a model is linguistics. There seems to have been some evolution in the Nattiez approach over the years, notably since the publication of Ruwet (1975), in which precisely the scientific standards of the enterprise were subjected to critical discussion.
With respect to the standards we invoked above, the Nattiezian "mise en série" is indeed repeatable, but it constitutes no test, because no acceptable conditions of failure, not even improbable ones, can be specified. In fairness, however, the Nattiez theory shares this lot with all analytical theories reviewed so far. We will come back to this in 1.2.
In his 1985, Nattiez seems, in some ways, a sadder and a wiser man – convincingness, in his opinion, is carried by means of trust in the erudition of the analyst on the one hand, and by means of effective use of rhetorics on the other. (One is tempted to add that the first element may be largely, if not totally, determined by the second.) In passing he seems to suggest (see his p. 44; undoubtedly, there is a rhetorical name for the gimmick) that the case of the sciences themselves is not very different. If I reconstruct him rightly, he takes Kuhnian revolutions as models for all changes in scientific theories, including those in music analysis. If that were right, music analysis should be in a state of never relenting turmoil, which, to my feeling, it clearly is not. Also, and more importantly, if we look to science for some rational way of deciding between theories, paradigm clashes is the place where we should not be looking in the first place. Besides, even in looking for ways of reducing irrationality it will not do to seek to annihilate all irrational and irreducibly intuitive behaviour – better assign it its proper place in the scheme of things, viz., in the context of discovery rather than in the context of justification.
Anyway, traditional analysis functions historically like Kuhnian normal science, and equally historically fails to live up to the standards of normal science, be it physics or linguistics. And, we submit, the Nattiez approach is in the same case. No testing can prove any Nattiezian wrong. Nevertheless, the approach has a convincingness of its own; the "mise en série" embodies, we feel, respectable theoretical assumptions of a very general perception-theoretic nature that, precisely because of their generality, defy, for the moment, attempts at making them testable (at least where the field of music is concerned). The reader interested in these theories is referred to Neisser (1976).
The last decade has seen a new kind of analytical heuristics, again inspired by linguistics, this time by its generative grammar variant. Lerdahl & Jackendoff, whose enterprise is at least in competition with the traditions of analysis, do indeed offer a strategy that claims generalizability to the whole of the field of classical tonal music. Of course, this is no small undertaking, and it is a good sign that spirited discussions are going on around it. From our viewpoint, however, the most serious objection will have to center on the testability claim they attach to their theory. Unhappily, they entirely fail to specify how this claim can be implemented in practice, and as a consequence, it must count, at least as I see it and until further notice, as just another rhetorical device. Baroni & Jacoboni, on the other hand, do offer testability chances akin to those available in linguistics. But the analysis embodied in their programs describes only what might be called a "muzak" part of our musical competences. This does not make it uninteresting, but keeps it rather apart from the analytical tradition.
Why, then, are the central analytical theories (eventually) rejected? Not because anyone has shown them to be false, it is now clear. The problem precisely is that they are lacking in propositional content. Of the two characteristic~functions traditionally ascribed to scientific theories, viz., the "searchlight" function and the "truth-content" function, they embody the former only. What is propagated through them is not some set of truth-claiming statements, but rather a way of looking at music (linked, we should hope, to a way of listening to it as well). The philosopher and literary theorist J.J.A. Mooij has tackled in his 1979 the difficult problem of how, in the absence of falsification, rejection of this kind of theories takes place.
What Mooij has shown may be summarized as follows. Science in fact uses many respectable but entirely unfalsifiable theories, one might perhaps say: theories as tautologous as mathematics itself, because they could be described as forms of applied mathematics. Examples are information theory, game theory, cybernetics. Like mathematics, they serve as the source-fields of concepts, the intrinsic logic of which they take on to investigate. In the three-level schematization due to Bunge (1973) they belong, with mathematics itself, to the highest level, on which are found theories scoring highest in versatility and lowest in testability (cf also Kunst & vd Bergh, 1984). As one goes down the levels, a trade-off is seen in which versatility is lost and testability gained. One goes down the levels through a process of specification and, possibly, combination of the more general theories, at the end of which one gets specific, non-versatile and testable micro-theories. Through the testing of these micro-theories, and therefore only indirectly and probably over longer periods of time, the more general theories are being put to the test.
Now where is an analysis to be located in Bunge's three-level scheme? Paradoxically, it seems to combine a total loss of versatility with no gain at all in testability. Given the fact that the genre is functioning and, in its own way, often flourishing, this constitutes a problem. The facts of the matter must be somewhat different.
The parts of this puzzie fall into place if we view published analyses as demonstration material for (often inexplicit) middle-level theories about music. In that case their interest would lie in the propagation of relatively general strategies they try to "explain", to bring home, by making them work – a well-documented pedagogical skill.
Again, analysis, if we are right, turns out to be a pedagogical undertaking. And according to the theory on art that we have been advocating an artist is an informal (or differently formal) teacher of concepts. So we will not be astonished if it is shown that analytical theories are rejected for the same reasons literary fashions are: according to Mooij analysts'theories may come to fail only through persistent lack of self-enriching fruitfulness – and is not that precisely what decides the eventual survival chances of literary (and artistic) fashions and styles generally?
What part of reality do analyses suggest or say they are describing? An answer one encounters not unfrequently is that they describe the pieces themselves. Very often this formulation counts as satisfactory, which makes it, in a sense, unassailable, but also hopelessly obscure. Nattiez has gallantly tried to escape this quandary by positing his niveau neutre concept (originally due to J. Molino). This purports to exclude both the composer's (niveau poïétique) and the listener's (niveau esthésique) activities. Because of the fact that anyone related to music is a listener anyway, the obvious countermove is to say: but is not anyone capable of reading music and getting his hands on a score, in particular the analyst, related to it as a listener is? Nattiez tries to get the sting out of this by heavily restricting what the analyst may do to the score: he has to limit himself to the mechanical execution of a certain kind of transformation, or rewriting, of it. Rhetorically, this has the effect of seeming to eliminate his subjectivity, which, again rhetorically, serves as the listener's hallmark. At least this is how we tentatively reconstruct some of his intuitions. The whole thing, however, remains hardly satisfactory. Any choice for a transformation of the score, be it mechanically executable or not (and we do not go here into the problem whether Nattiez' is) represents a theoretical choice, and therefore embodies a theory. In this case, a theory about what?...
In musical classroom practice, it is very often understood or claimed that the analytical theory is about the composer's intentions. This makes them historical hypotheses, and therefore subject to the checks and balances any historical claim is subjected to. In this case, these mostly do not amount to very much – the subject being restricted to, e.g., one work, and this work permitting, in principle, a plethora of exegeses, the usual cross-referencing by historians is of no great help. On the other hand, historians commonly avail themselves of consensus reached – therefore, if certain conventional ways exist of talking of, e.g., a late Beethoven string quartet, analysts are, in an important sense, justified in conforming to these conventions.
However, it is interesting to note that this knife cuts both ways. Epistemological problems concerning the past are decided on the basis of a social reality existing in the present, and therefore open to non-historical research (i.e. its reality and its content, whatever justifies it through its functioning, can be investigated non-historically; its genesis and/or historical justification remain a historical problem).
But there is something even more fundamental. If we view, as I would urge us to do, communication through music, the functioning of music, as among the essential interests of a musicologist, we might even say that the composer is nothing more (nor less, of course) than the prime mover of musical communication processes in which he, after finishing and publishing his work, becomes a participant among others. (Of course, it may be that nobody forgets that he has been the prime mover.) Once dead, he is not even a participant any more. If there is a sense in which he "lives on", it must be as the principal character in the "story" the users of his music coordinate themselves among each other to make the current one – according to the Lewis definition of the concept, a conventional story. And obviously, it is to this character that intentions actually are ascribed. In this view the composer himself is to be found just where pieces and musical styles are to be found: in the present, in the coordinated behaviour of the participants of the culture he belongs to.
The last alternative could be, I take it, to locate what analysis is about in the listener. Here again, we find three, more or less competing, alternatives: it could be (a) in what listeners could do, (b) in what listeners should do, and (c) in what listeners actually do. We take these cases in turn.
This explains why analysts so often attribute their intentions to him. The analyst, too, is coming up with hypotheses about the culture, i.e. about what is there in the real world. If this is so, and if he has a precise hypothesis to offer, then he and I may go out into that real world and look whether he is right or not. Nattiez (cf the last pages of his 1985) still despairs of this. For him, the matter remains purely one of coordination between analysts, and the best that can be done is, therefore, to establish conventions. There is not one and unique TRUTH about pieces, he says. Indeed, who would want to predict to what future use any piece can be put? As long as we keep wanting to tackle the "pieces in themselves", we are stuck with what has been called the "plasticity of the art-work" – not only do meaning and use of pieces remain in a state of constant change, but any successful analytical theory produces, to a varying extent, its own truth: by being successful, it makes itself true. Analysis, thus conceived, remains a literary genre, a form of art.
Let us at least admit of the possibility in principle, and the interest, of a descriptive and non-rhetorical kind of analysis. It may be said to be about pieces; but it locates these largely outside the analysts' circle, viz., in the memory spaces of their normal users. These memory spaces are not limited to the inside of the human skull, but include external, public and social spaces such as libraries, the accessible (i.e. socially effective) competences of musicians, etc.. This kind of reality belongs to the world of the products of the human mind, Popper's World 3, a world in its own way as "objective" as the physical world itself. In Kunst & van den Bergh (1984) we have shown how to use it in testing analytical hypotheses about a given musical event.
In this conception it is easy to see that the plasticity of the artwork really is the plasticity of its (future) users. Humans are molded by the culture they grow up in and keep adjusting to; and the culture includes not only musical compositions but also analytical essays about them, be they published in learned journals or only accessible through conventions governing "informal talk" about music. Let us admit of the possibility and interest of a musicology that takes on as its subject this plasticity itself, its extent, its limitations with respect to a given state of the culture, etc., and by constructing hypotheses about it, falsifying them, refining them, etc., etc., try to contribute to a new kind of growth of knowledge in musicology. 
It is hoped that the present issue of Interface will contribute to a fruitful discussion of analytical problems, both those arising from a critical appraisal of the genre itself, such as provided by the Karkoschka contribution and, more implicitly, by the essay of Camilleri and co-workers, and those concerning the interpretation of specific compositions, such as provided by the other contributors, who all, to a greater or lesser extent, propose and try out more or less generalizable analytical strategies. So the DeLio contribution gives a text-oriented interpretation of a case that, by many of its features, is unique, but through others (e.g., the idea of spatial symmetry in the use of twelve tone rows) is quite general indeed. The Perone contribution shows how pcs analytical vocabulary can be used to identify a veiled tonal center in a reputedly atonal piece; the Hamman essay explores a specific user-and-site-determined kind of existence certain contemporary music practices confer to the "pieces" they produce, and demonstrates how far will take us certain common conceptions of analysis, when extended to hitherto unfamiliar fields. Special use is made of the composer's own verbal comments to his composition – and it is, in fact, a noteworthy (and, in the light of some of the foregoing, understandable) phenomenon in XXth century music, that composers quite often take the initiative, so to speak, in the analytical work to be done on their pieces. So even the Landy piece on Shinohara owes its interest, in part, to the composer's having been accessible to the analyst: analysts can, and do, make fruiful use of composers'analysis-oriented attitude. Finally, this issue offers a composer on himself: in his contribution on Heterostase Vriend explicitly tries to formulate the links between what he meant the listener to perceive in his piece on the one hand, and the chosen means of production of it on the other.
It would seem to me that the analysts represented in this issue make amply clear, not only in their vocabulary but also in the substance of what they have to say, that they see themselves as theorists, and surely not as just providers of social lubricant. Now if, possibly by mistake, they have ended up practicing an art form instead of something like science (and it must in all fairness be conceded that they share this lot with the whole of the analytical mainstream) this might well be, as we said in the beginning of this essay, no reason for self-congratulation. For the quality of art is grounded exclusively upon the critical feedback of its audience. If the audience is small, it must become all the more fanatical in order to compensate for its smallness. I am not confident that the audience for written analysis actually satisfies this condition. If, as I am tempted to see it, it is essentially dependent on, and a by-product of, the oral classroom practice (I feel in sympathy with Karkoschka's remarks on the matter; however, he may not agree with mine) it lacks, in fact, a sufficient feedback system of its own, such as scientific theories find in the tests they are subjected to. Analyses are tentative theories, akin to science in a truncated form.
Baroni, M. & Jacoboni, C. (1978). Proposal for a Grammar of Melody: the Bach Chorales, Montréal (PUM).
Bunge, M. (1973). Method, Model and Matter, Dordrecht/Boston (Reidel).
Cogan, R., & Escot, P. (1976). Sonic Design, Englewood Cliffs (Prentice Hall).
Karkoschka, E. (1976). Analyse, Herrenberg (Döring).
Kunst, J. (1978). Making Sense in Music: an Enquiry into the Formal Pragmatics of Art, Ghent (Communication & Cognition).
Kunst, J. & vd Bergh, H. (1984). 'The Analysis of Musical Meaning: A Theory and an Experiment' in: Interface, 13, pp. 75-106.
Leman, M. (1985). 'Dynamical-Hierarchical Networks as Perceptual Memory Representations of Music' in: Interface 14, pp. 125-164.
Lerdahl, F. & Jackendoff, R. (1983). A Generative Theory of Tonal Music, Cambridge, Mass. (MIT).
Mooij, J. J. A. (1979). 'The Nature and Function of Literary Theories', in: Poetics Today, 1, pp. 111-135.
Nattiez, J.-J. (1975). Fondements d'une sémiologie de la musique, Paris (10-18).
Nattiez, J.-J. (1985). 'Les concepts de mise en série et d'intrigue dans l'analyse musicale', in: Lönn, A. & Kjellberg, E. Analytica, Studies in the Description and Analysis of Music, Uppsala, pp. 35-46.
Neisser, U. (1976). Cognition and Reality, San Francisco (Freeman & Co).
Ruwet, N. (1975). 'Théorie et méthodes dans les études musicales' in: Musique en jeu, 17, pp. 11-36.
 It is here, we take it, that we part ways with Marc Leman (1985) who, as we understand him, assumes something like total plasticity for his theoretical construct of the listener. His listener seems to be totally determined by what he is hearing here and now – by his short-term memory contents only. We still think that it is impossible to give anything like an algorithmic procedure for generating hypotheses on listening on the basis of the score alone.