Music Theory and 'Good Comparison': A Viennese Perspective

Nicholas Cook

Journal of Music Theory, Vol. 33, No. 1. (Spring, 1989), pp. 117-141.

Stable URL:
http://links.jstor.org/sici?sici=0022-2909%28198921%2933%3A1%3C117%3AMTA%27CA%3E2.0.CO%3B2-6

Journal of Music Theory is currently published by Yale University Department of Music.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/yudm.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For
more information regarding JSTOR, please contact [email protected].

http://www.jstor.org
Sat Apr 7 22:29:27 2007
 MUSIC THEORY AND

'GOOD COMPARISON':

A VIENNESE PERSPECTIVE

Nicholas Cook

Most contemporary theorists feel uncomfortable about ascribing sig-

nificance to inaudible relationships in music; we tend to assume that there
should be some meaningful relationship between analysis and auditory
experience. There is no obligatory reason why this should be so. The idea
that what is significant in music should coincide with what is perceptible in
it is not a universal one; according to Dahlhausl this is no more than a
dogma of the last two centuries or so, while even today there are some
types of non-Western music whose structural organization lies principally
in the physical actions involved in performance, rather than in the sound?
And David Lewin has recently argued that as theorists we place too much
emphasis on perception as distinct from the broad range of activities
through which people express their responses to music? But the fact
remains that, in the West today, it is the perception of musical sound that is
generally considered to be paramount in defining the meaning of a piece of
music. Hence one of the most crucial questions we can ask about any the-
ory of music-one which bears directly upon the validity which we can
ascribe to it-is how it relates to the perceptual experience of the listener.
In this article I try to show the problems inherent in any simple answer to
the question, and to outline a rationale for the practice of analysis that over-
comes these problems.
 The simplest possible relationship between analysis and auditory expe-
rience would be if the analysis consisted of a list of "things to hear''. Essen-
tially this is what the much-abused formal analyses produced around the
turn of the present century consisted of: first the first subject, then a bridge
passage, then the second subject and so forth. Now there is no doubt that
people can be quite easily taught to hear these things. Some fifteen years
ago, Alan Smith4 carried out a series of experiments in which seventh-
grade students were given instruction in classical forms. They were then
played curtailed recordings of classical works and had to identify the struc-
tural point at which the music stopped. The seventh-graders performed
quite well. Perhaps more revealing, however, was a further series of tests
which Smith carried out, in which music majors were tested in the same
way. In the first test, in which they did not know what they would be
required to do, these more advanced students performed poorly. But in sub-
sequent tests-when they knew what was expected-they did well. What
this shows is that people who know how to track the musical form as they
listen-how to identify the first subject, the transition and so on-do not in
fact do so in the ordinary way. And this is perhaps not surprising, given
what many theorists, following Schenker, see as the superficial and even
meaningless nature of these traditional categorizations of musical form.
Schenker's attack on the categories of traditional formal analysis rather
resembles a phenomenological reduction. He "brackets" them in the same
way that phenomenologists strip off the sedimented accretions of historical
or theoretical knowledge, so that the underlying phenomenon will reveal
itself to consciousness. But Schenker's theory of music is not in fact a phe-
nomenology in any normal sense. He reduces music, not to a phenomenon,
but to a theory of structural levels. Schenker believed that only the genius
is directly aware of music's large-scale structure-of the Urlinie and its
transformations, that is to say. People who lack genius cannot become
aware of this directly; they can only do so through the mediation of the con-
ceptual representation of the music provided by the analysL5 So it would
not be correct to think of Schenker's analyses as attempts to represent the
manner in which listeners ordinarily perceive music; as Ruth Solie says,
"Schenker predicates his notion of totality not upon perceptual mecha-
nisms in the observer, but upon the work of art itself :6 I shall return to this
at a later point.
Our contemporary interest in Schenkerian analysis, however, derives
largely from the fact that it does somehow seem to reflect the auditory expe-
rience of music. And in the adaptation and extension of Schenkerian theory
offered by Lerdahl and Jackendoff in their Generative Zheory of Tonal
Music this perceptual basis is made explicit, for the authors describe their
work as a model of "the organization that the listener attributes to the music
he hears"? Though this listener-the "experienced listener", as Lerdahl and
Jackendoff sometimes call him-is an abstraction (because individuals may
vary in the manner in which they hear a piece), the theory is put forward
as "an empirically verifiabe or falsifiable description of some aspects of
musical organization, potentially to be tested against all available evidence
from contrived examples, from the existing literature of tonal music, or
from laboratory experiments". So far the only such tests to have appeared
in print are those by Irkne Delikge; and they do provide some confirmation
of Lerdahl and Jackendoffs model of tonal perception. But two qualifica-
tions need to be made. In the first place, Delikge's subjects were given
sheets on which each note of the music was shown as a dot, and asked to
indicate the segments into which the music fell. So they were not simply
listening; they were making decisions on the basis of a visual representa-
tion of the music. And this raises a problem to which Burton Rosner called
attention in his review of Lerdahl and Jackendoffs book, in which he said
that what the authors really seem to be talking about aie
the intuitions of a sophisticated listener who has the score of the piece! The
present theory (and others that claim perceptual validity) generally rest on
the tacitly accepted but little noticed use of scores. . . . Lerdahl and Jack-
endoff are not alone in assuming that complex hierarchical trees in visual
form are isomorphic to the auditory results of listening, which necessarily
occurs across time and involves memory. This assumption seems very dub-
ious psychologically, when carried to larger levels?
Rosner's mention of larger levels raises the second point I mentioned.
Delikge's tests were concerned only with the perception of small-scale
structure, on the level of phrases rather than extended formal sections.
Now Lerdahl and Jackendoff treat large-scale structure on the model of
small-scale structure; like Schenker, they model the unity of a movement
on the unity of a phrase. But it is by no means obvious that listeners hear
movements as integrated structure's in the same way as they do phrases. Ros-
ner elaborates this point in an article co-authored with Leonard B. Meyer,
and draws attention to an important distinction between hierarchical theor-
ies of music and the linguistic theories with which they are often compared:
The top node of a grammatical tree is an immediately observed datum: a sen-
tence or an utterance. . . . The lowest nodes in music-theoretic tree struc-
tures, however, represent a datum: an actual stretch of music. Quite often,
only fragments of it are held in memory. . . . We cannot therefore believe
that the increasingly higher nodes, which represent ever more rarified selec-
tions, form the core of music perception.1°
And the general point that Rosner and Meyer are making extends beyond
strictly hierarchical analysis. The concept of large-scale tonal closure-the
idea that a movement or even a multi-movement work should begin and
end in the same key-is fundamental to our understanding of classical
form, but experiments which I have published elsewherell suggest that
listeners only have a direct perception of tonal closure when the time-scale
involved is in the order of a minute or less. In other words it seems that
when we talk about the tonal coherence of movements lasting several
minutes, we are not talking about what people actually hear at all (or at
least most people). Yet it is this kind of coherence which forms the primary
focus of the majority of analytical studies of music. In this way the percep-
tual reality of what the music theorist is saying seems to become parlous
just at the point that it becomes of the greatest aesthetic interest.
Eugene Narmour and Alan Keiler,'2 among others, have pointed out the
implausibility of Schenkerian theory as a theory of perception (something
which, as I said, it was never originally intended to be). The basis of their
criticisms is that Schenker's middleground and foreground are only defined
by the background, so that the perception of the parts presupposes the per-
ception of the whole-a model of perception which it is difficult, if not
impossible, to maintain. Both Narmour and Keiler have accordingly been
trying to develop models of musical structure which are genuinely reduc-
tive, in the sense of showing how the listener can derive a structural inter-
pretation of the music from the observed phenomena of the musical
s~rface!B ~ ut the best-known attempt to model musical structure in this
manner is the one that Benjamin Boretz published in six issues of Perspec-
tives of New Music under the general title "Meta- variation^"?^
The basis of Boretz' system is what he called the "qualia" of pitch and
time-ordering, which constitute the building-blacks of musical perception.
These are structural categories: they represent the smallest difference in
pitch or time that is of significance within a given musical culture. The
smallest significant interval in serial music, for instance, is the tempered
semitone; a flat Db is still a Db, and as such categoricaly distinct from a C,
even a sharp one. (In another culture, however, the flat Db might represent
a different structural category from the in-tune one.)I5 Now it is obviously
important that people should be able to discriminate between different
structural categories-it would never do to have music in which there were
a hundred different structural values to the semitone. But this has nothing
to do with the musical structure as such. It is simply a question of the gen-
eral psycho-acoustical constraints within which music must operate, and as
Babbitt (whose influence is felt throughout Meta-Variations) puts it, "the
discovery and formulation of these constraints fall in the province of the
psycho-acoustician"!6 In other words a rigid methodological division is
made between the psycho-acoustician's investigation of auditory percep-
tion, and the music theorist's investigation of musical structure.
In line with this, the rest of Meta-kriations consists of formalized
models of some of the ways in which the qualia of pitch and time-ordering
can be organized in order to yield different types of musical structure.
There is hardly any direct consideration of the manner in which listeners
perceive these structures; the theoretical assumption seems to be that as
long as the building-blocks of musical structure-the qualia-are percep-
tible, any structure built out of them ought to be perceptible too!' And Bab-
bitt makes such an assumption explicit when, speaking of serial music, he
says that the transformations of a series S "require for the perception of
their relation to S merely the ability to identify interval classes".18 Now this
remark is in flat contradiction with the results of a considerable number of
experimental studies which have shown that listeners are not, in fact, gen-
erally able to perceive such relationship^!^ And this is not surprising,
because Babbitt's statement involves an illegitimate generalization between
two quite different contexts of listening. It is worth going into this in some
detail.
In his writings, it was Babbitt's avowed aim to develop a scientifically
adequate approach to music theory, that is to say one whose terms could be
defined in a rigorous, abstract manner.20 But people's subjective experi-
ences of music cannot be analyzed in rigorous, abstract terms. Scores, on
the other hand, can. To this extent, Babbitt's analyses of music are, in actu-
ality, analyses of musical scores. And scores consist, among other things,
of specifications of pitch and time-point. However, when people listen to
music in the ordinary way, they don't hear pitches and time-points. To be
sure, they hear tunes and harmonies, which are broken up on the page into
distinct notes, but they do not hear the notes as separate entities and indeed
they sometimes do not hear them at all, at least in a manner that directly
corresponds to what is visible in the score. Of course there are certain cir-
cumstances, for instance in the ear-training class or the psychologist's labor-
atory, where people make a special effort to hear music in terms of distinct
notes and rhythmic values. But even under such conditions, there are
severe limits in the extent to which this can be done; there is a drastic
asymmetry between the music that listeners can cope with as an ear-
training exercise, and what they can appreciate in the concert hall.
If the notes in the score do not represent things that people ordinarily
hear, then, do they have some kind of objective reality? A pianist plays
notes, in the sense that he presses keys. But singers do not sing notes in this
sense; indeed they sometimes glide from one note to another in such a way
that it becomes hard, or even arbitrary, to say where one note stops and the
next one starts. In such cases, or in the case of jazz saxophone playing, or
a rapid scale played on the violin-or any number of avant-garde pieces-
the notes cannot be said to be in the sound at all in any objective sense. As
Mary Louise Serafine puts it, they "arise only as a result of reflection upon
music and notation of it"? In other words the score does not directly cor-
respond to any psychological or acoustical reality. It is not even an approx-
imation to reality. Instead it is a model, or a based on a
comparison between the experience of musical sounds in an actual musical
context and the judgements of pitch or interval that would be made if the
same sounds were heard individually. What Babbitt has done is to confuse
the primary and the secondary subjects of the metaphor- that is, to treat
the model of ear-training on which the score is based as if it were the musi-
cal reality that it representsJ3
If in a significant sense the formalized music theory of Babbitt and
Boretz is a theory of musical scores, then the same is also true of both
Schenker and Schoenberg. Boretz' notorious statement that "sounds . . .
are not part of music, however essential they are to its transmi~sion"~~
echoes something that Schoenberg told Dika Newlin in 1940:
Music need not be performed any more than books need to be read aloud,
for its logic is perfectly represented on the printed page; and the performer,
for all his intolerable arrogance, is totally unecessary except as his interpre-
tations make the music understandable to an audience unfortunate enough
not to be able to read it in print?5
What this implies-and there are other statements by Schoenberg which
tend to the same conclusion-is that the work exists as an ideal entity essen-
tially distinct from its acoustic realization. And Schenker spelled this out
more specifically in an article from Das Meistenverk called "Let's do away
with the phrasing slur"F6
Schenker's main purpose in this article is to distinguish between the edi-
torial phrasing slur-which is simply an instruction for immediate
execution-and the legato slurs found in the autographs of the master com-
posers, which (according to Schenker) constitute an essential part of the
musical structure. Indeed the basic proposition Schenker is putting for-
ward is that "the masters' manner of notation represents the most complete
unity of inner and outer form, of content and symbols" (p. 55). For this rea-
son, he continues, the composer's "struggle over notation always goes hand
in hand with a struggle over the content; but once the content is worked
out, then the only possible notation is also immediately present" (p. 74).
For Schenker, then, the score is not a representation of the music; it is the
music. In a musical sense, therefore, a score is realized not through being
performed, but through being understood as an organic whole, an expres-
sion of the Urlinie. This explains Schenker's otherwise strange comment
about the analytical graphs in Der Freie Satz that "the graphic representa-
tion is part of the actual composition, not merely an educational means"?'
And it is in this sense-a decidedly Hegelian one-that, as Ruth Solie said,
Schenker's theory is based not on the listener's perceptions, but on the
work of art itself.
We have arrived at a paradox. Babbitt and Boretz' attempts to develop a
rigorous theory for music-one which will meet the demands of scientific
discourse-turn out to be based on a conception of the score that makes
more sense in terms of nineteenth-century metaphysics than twentieth-
century psychology. Nor is this the only way in which supposedly scientific
explanations of music turn out to be based on aesthetic presuppositions of
limited historical or geographical scope. John Rahn justifies what he calls
"the presumptuous assumption that our interest is primarily focused on par-
ticular pieces of music' on the grounds that 'after all, we never musically lis-
ten to anything else"?s But, unless we are willing to gloss "musically" as
"analytically" (in which case the statement becomes more or less circular),
this is not as obvious as Rahn implies. For, as Patricia Carpenter has
shown?9 the concept of the musical work as a reified entity is one that only
developed around 1800; and, in Dahlhaus' words, though the concept of the
composition seemed "self-evident in the nineteenth century according to
the letter of the aesthetic law, it was of restricted validity and always in peril
from the context of actual musical behavi~ur".~~
Certainly the tests I mentioned earlier, in which listeners failed to
respond to large-scale tonal closure, would support this last remark. Were
the subjects of these tests musically listening to particular pieces of music,
to borrow Rahn's words, in such a way that they experienced the particular-
ity of a work as a consequence of its structural organization? Apparently
not; and this means that, if these listeners experienced what they heard as
compositions, this must have had as much to do with their aesthetic
preconceptions-the interpretative attitude with which they approached
what they heard-as with the music's structure. Or perhaps they experi-
enced the music in the same way as background music, and not as consti-
tuting particular pieces at all-not "musically", to use Rahn's term. Either
interpretation is uncomfortable from the theorist's point of view. In the one
case the unity of the work lies in the province of historian or even the soci-
ologist, since it is the result of changing aesthetic attitudes; in the other
case there is no work, and hence no focus for music theory at all. A better
account of the relationship between perception and music theory is needed
if the discipline is not to collapse into the history and sociology of aesthe-
tics on the one hand, and the psychology and pedagogy of note-to-note
structure on the other.

Of course there is an easy way to solve these problems, though it means

dispensing with the image of a scientific music theory. This is to maintain
that the purpose of analysis is not to reflect how people listen to music, but
to explain how they ought to. Such was Schenker's position. Like Hanslick,
Schenker was reacting against what he saw as a decline in Western musical
culture, a decline that stemmed from a failure of hearing. Schenker says in
Das Meistenverk that "theorists as well as performers . . . plod along from
one passage to the next with the laziest of ears and without the slightest
musical imagination. All they hear is the constant change between tonic
and dominant, cadence after cadence, melodies, themes, repetitions, pedal
point"3l It is this that Schenker's work was intended to combat. And for
Schenker, of course, there was no question of his imposing his own per-
sonal tastes or values upon other musicians: the Urlinie and its transforma-
tions were not something he had dreamt up to express or rationalize his
own aesthetic viewpoint, but part of the objective structure of the musical
masterwork.32 But if we cannot today accept Schenker's Hegelian views,
then the idea that analysis says how music should be heard becomes proble-
matic. A prescriptive stance may easily appear to be purely authoritarian
(this is how you should hear the music), with the only alternative being an
unbridled relativism (this is how I hear the music; you can hear it as you
like).
What we need is a rationale for adopting Schenker's analytical methods
while rejecting his epistemology. And a clue to how such a rationale might
be formulated can be found in the concept of Darstellung as developed by
other writers in Vienna during the early decades of the present century3L
just the time that Schenker was bringing his own theories to fruition.
Schenker seems to have been unaffected by these broader intellectual devel-
opments. Schoenberg, however, was not. In the first chapter of his Harmo-
nielehre he attacks the work of previous theorists who put forward what
they claimed to be the natural laws governing music. Schoenberg does not
deny that such natural laws exist. But nobody, he says, has yet discovered
what they are, and he adds, "I believe they will not be discovered very
soon".34 However that does not mean that there is no place for music the-
ory. Schoenberg continues:
Efforts to discover laws of art can . . . , at best, produce results something
like those of a good comparison: that is, they can influence the way in which
the sense organ of the subject, the observer, orients itself to the attributes of
the object observed. In making a comparison we bring closer what is too dis-
tant, thereby enlarging details, and remove to some distance what is too
close, thereby gaining perspective. No greater worth than something of this
sort can, at present, be ascribed to laws of art. Yet that is already quite a lot.
And he goes on to outline the pedagogical value of this kind of theory:
What we do achieve can be enough, if it is given as a method of teaching, as
a system of presentation (Darstellungl-a system whose organization may
aim, sensibly and practically, towards the goals of instruction; a system
whose clarity is simply clarity of presentation, a system that does not pre-
tend to clarify the ultimate nature of the things presented
-as, of course, Schenker's theories
If Schoenberg put forward his concept of Darstellung by way of a crit-
icism of other writers in the field, the philosopher Ludwig Wittgenstein-
another Viennese citizen-turned the same arguments against his own ear-
lier work. Wittgenstein's Tractatus Logico-philosophicus, which was first
published in 1919, was an attempt to reduce language to its formal essen-
tials; and like Boretz' Meta-briations, which belongs to the same intellec-
tual t r a d i t i ~ n ?it~ makes use of propositional calculus for this purpose.
When he first completed the Tractatus, Wittgenstein believed that it pro-
vided answers to all the principal problems of philosophy; he said as much
in his preface to the work. But in later years he came to think that the elab-
orate system of the Tractatus had, in fact, amounted to no more than a sys-
tem of presentation. Janik and Toulmin put it like this (p. 190):
The propositional calculus had attracted Wittgenstein, in the first place, as
the intellectual instrument required for a fully rigorous "critique" of lan-
guage in general. By the time he had finished, it turned out to have given
him only the scaffolding for an elaborate metaphor. Unless one saw the pos-
sibility of modeling "facts" by "propositions" having the same "real logical
form'', no independent demonstration was possible to prove that the propo-
sitional calculus can be used to describe real "states of affairs".
And Wittgenstein also put forward a critique of Freudian psychotherapy
which ran along similar lines. According to Freud, the function of the ther-
apeutic process was to uncover events which had actually occurred in the
patient's childhood and which, through being repressed, had caused a
neurosis. Once the events had been made conscious, the neurosis would
disappear-rather in the manner of a physical symptom that disappears
when the underlying medical condition has been treated. Now Wittgen-
stein did not doubt that Freudian psychoanalysis might work. But he
rejected Freud's concept of causality. He argued that the very concept of
mental illness was no more than a metaphor-a "good comparison", to use
Schoenberg's phrase, between physical symptoms and behavioral abnormal-
ity. He argued that the childhood events supposedly discovered during the
therapeutic process had probably never happened and were certainly not
the cause of the patient's problems. But the psychoanalyst's reconstruction
of them was valid, all the same, to the extent that it was accepted by the
patient and so helped him to come to terms with his predicament.
Like Freudian psychoanalysis, Schenkerian analysis is based on a meta-
phor. In Schenker's case the comparison is between the note-to-note struc-
ture of Fuxian counterpoint and the freely elaborated surface of real
music?' Technically speaking, one might say that Schenkerian theory con-
sists of a number of transformations which may be invoked in order to
account for the discrepancies between a particular piece of music and the
rigid note-to-note specifications of Fuxian counterpoint. And the effect of
this is, as in the case of Freud's theory, to render intelligible something that
is problematic (large-scale musical structure in Schenker's case, abnormal
behaviour in Freud's) through formulating it in terms of something familiar
(Fuxian counterpoint in the one case, illness in the other). To say this is not
to say that the same laws govern large-scale musical structure and Fuxian
counterpoint (or abnormal behaviour and illness); it is just to make a com-
parison. We are dealing with what Schoenberg called a system of presenta-
tion, something that does not-or at least should not-pretend to clarify the
ultimate nature of the things presented.
It is easy enough to see what this means in principle. But what does it
mean in practice? Consider the status of consecutive fifths in Schenkerian
analysis. These are forbidden under all circumstances in Fuxian counter-
point, which deals only with the musical surface. Following Brahms'
Schenker regarded foreground consecutives as legitimate when they
result from the interaction of different structural levels: in such cases.
Schenker argued, the consecutives are apparent rather than real. But he
accepted the traditional prohibition of real consecutives. Now Schenker
a
did not see his theory as metaphor; he believed that there are natural laws
which operate equally at the level of large-scale and that of note-to-note
s t ~ c t u r e . ~So
9 one might expect to find an equally strict prohibition of con-
secutive fifths when these are generated at a single structural level in the
middleground. In the event Schenker is more pragmatic, saying that forbid-
den intervallic successions may be found in the middleground, but that "it
is then the task of the foreground to eliminate them"POIn practice, however,
both Schenker and present-day Schenkerians tend to avoid middleground
consecutives.
To take a specific example, how would you analyze the second half of
Schubert's song "Das Wandern"? (Figure la shows a prkcis of the score.)
The fall of the fundamental line from 5 to 2 is supported by a sequential
bass line FCC-EbF4 giving an intervallic progression 6-3-6-3 between the
outer lines [b]. But, as the slurs and the Roman letters in [b] indicate, the
F# and Eb are merely approach tones to the G and F; they have a subordi-
nate role. Therefore at a more remote level the structure is as shown in [c];
the melody arpeggiates the harmonies VI and V, which in turn support the
descent of the fundamental line. But now we have glaring consecutive fifths
between the outer parts. The analysis looks uncouth. Would you not avoid
graphing the piece this way, instead either including more foreground
detail so that the fifths are disguised-as in [b]-or interpreting the struc-
tural motion differently, for instance as in [dl or [el? But in what sense do
these alternative analyses make better musical sense of Schubert's song
than the one shown in [c]? (Why is the V harmony in [dl given precedence
over the VI, when each receives precisely the same support at foreground
level? Do you really hear a structural dominant in mrn. 17-20, as [el
implies, with their rocking alternation of F and Bb?) I would maintain that
[c] is not in itself a less accurate formulation of the tonal structure of "Das
Wandern" than [dl or [el, but that, because of the consecutive fifths, it is
less satisfactory as an expression of that structure in terms of the metaphor
Figure 1
of Fuxian counterpoint. It makes the music look ungrammatical and, there-
fore, incoherent. But this is not because the middleground consecutives
contravene any natural law of musical organization. It is because they run
counter to the representational means adopted in Schenkerian analysis.
They spoil the comparison between Schubert's song and Fuxian counter-
point.
If a Freudian explanation acquires its validity through being accepted by
the patient, a Schenkerian explanation is validated when its reader accepts
it as a satisfying account of the music in question. (To put it this way per-
haps makes it sound like just a matter of taste. But a Schenkerian analysis
does not simply present an interpretation; it provides reasons for the inter-
pretation, implicitly if not explicitly.) Now the idea that an analysis is val-
idated through the reader's acceptance of it may seem weak and subjective
by contrast with the criterion of verifiability, as discussed by Brown and
Dempster. As a matter of fact there is a good deal of scientific work that
does not adhere to the verifiability criterion; the social sciences, for
instance, generally adopt the less rigorous criterion of replicabilityP1 and
in a sense the reader's validation of Schenkerian analysis could be consid-
ered as falling under this heading. But what I would prefer to emphasize is
the larger trend in musical studies away from explanatory criteria borrowed
from the sciences. Kenneth Gourlay expresses this in a statement that
might have been designed specifically to counter what Babbitt said about
scientific language and method: "the assumption that there is only one sci-
entific method", he writes, "results in analogous attempts to apply it in
humanistic fields by eliminating the personal and subjective, as in anthro-
pology or sociology, without fully considering whether it is applicable"P2
What Gourlay is essentially saying is that the ethnomusicologist needs
to recognize the centrality of his own role in cultural interpretation. Dahl-
haus says much the same thing from the historian's standpoint: "whatever
precision it is within our powers to attain will never be reached by leaving
ourselves, the observers, out of the picture, but solely by making the
observer's position an integral part of the agreement reached on the cogni-
tive process chosen and the results ~btained"?~ And this is also the basic
premise of Richard Taruskin's critique of the performance practice
movementP4 In each of these cases, the study of music is being viewed as
a creative interaction between the musicologist and the music in question,
in which the musicologist seeks to define or redefine the music and, in so
doing, constructs a cultural identify for himselfP5
Wittgenstein stressed the necessity of "seeing" how the propositional cal-
culus of the Tractatus could be applied to real-life situations, and any meta-
phor has to be "seen" in the same sense as a joke: that is to say, not
conceptualized in analytical terms, but grasped through what could be
called a kind of inner performance. Schenkerian analysis lends itself to this
kind of inner performance because it builds on the familiar foundations of
species counterpoint. This makes it relatively easy for a trained musician
to read a Schenker graph, not in the way one reads an abstract argument,
but in the way one reads a score - in terms of the fusion of aural perception
and imagination that we refer to as "hearing"P6 A Schenkerian analysis, in
other words, yields up its meaning when it is experienced, or rather when
the music is experienced in terms of it. In saying this, I want to suggest that
the significance of an analysis lies not so much in the product-that is to
say, the published graph or table-as in the actual process of writing or
reading it. (David Lewin makes the same point more figuratively when he
refers to the products of analysis as "ski tracks tracing the poetic deeds that
were the perceptions themselves':47) And if we accept this view-if we
regard an analysis not as an objective representation of musical structure
but as a suggestion for how the music can be experienced-then we may
find that a number of the problems of contemporary music theory simply
evaporate.
One of these is the distinction that Narmour and Keiler emphasize, and
which I mentioned earlier, between background-to-foreground and
foreground-to-background derivation. Keiler-who is not hostile to Schen-
kerian analysis, as Narmour is-uses tree structures as a way of formaliz-
ing the transformations that are shown more impressionistically in a
conventional Schenkerian graph. Now this can obviously be useful as a
way of clarifying just what is involved when we talk, say, about interrup-
tion technique; if we are going to use theoretical models then we want to
have a clear conception of what the models actually are (though it might
possibly be argued that some of the formal distinctions that tree-structure
representation forces one to make do not correspond to musically signifi-
cant distinctions). But what is questionable is whether there is any virtue
beyond this in distinguishing, as Keiler insists that we should, between "the
structural analysis of a piece"-that is to say, the formalized interpretation
shown in the tree structure-and "the informal reductive strategy of the
analyst"P8 In formal terms, it may be true that Schenker's system is genera-
tive and not reductiveP9But the analyst does not begin with the fundarnen-
tal structure (how would he know which form of it to choose?) and then
elaborate it step by step. He is more likely to begin by reducing the score
to a series of middleground formations, only considering the background
at a later stage. He may begin by taking 3 as the primary tone in order to
see how such an interpretation will work out, but to do so is only to set up
a hypothesis which is liable to be set aside at any stage. In this way, doing
a Schenkerian analysis involves a constant alternation between background-
to-foreground and foreground-to-background derivation. Can we not main-
tain that the substance of the analysis lies precisely in this process of test-
ing alternative interpretations and seeing how they illuminate or contradict
the details of the surface-a process which is recapitulated when a reader
works through an analytical graph, rather than simply accepting it as some
kind of statement of fact? If so, Keiler's claim that "Schenker's graphic nota-
tion . . . constitutes a theoretical language that still requires considerable
decipherment" (p. 227) is open to the retort that we realize its meaning
every time we do Schenkerian analysis. And, after all, it was not Schenker-
ian analysis that prompted Wilson Coker to ask "but . . . will we be able
to interpret what our theories mean?"50 it was the use of formalized repre-
sentations of musical structure-representations that do not make use of
conventional notational symbols and so are harder to "hear" in musical
terms.
A more important problem, perhaps, for contemporary music theory is
the issue of universalism versus particularism, which Brown and Dempster
discuss at length in their article. Their basic contention is that there is an
"essential contradiction between forming general laws of music"- that is to
say, universalist or scientific explanation-"and explaining the uniqueness
of individual composition^"?^ Now there are a number of comments that
might be made about this. For one thing, not all scientific enquiry is con-
cerned with laws that are universal in their application, particularly in the
social sciences; in this respect Brown and Dempster's characterization of
scientific method seems a little old-fashioned. But a more important objec-
tion relates to Brown and Dempster's characterization of particularism.
They cite the following statement from Meta-Variations as an example of
particularism (p. 82):
The criterion for ''completion" of a "unit of syntactic structure" is and must
be external with respect to any individual utterance in language; in music,
such units of "completion" may again be contextually determined on the
basis of single instances from "internally", or "implicitly" defined criteria.
This is a particularist point of view, they say, because it implies that "an
individual art work both wholly determines its analysis or interpretation
and does so because it also determines its own best method of analysis or
interpretation" (p. 82). But Boretz is not saying that the general principles
governing the analysis-for example, that it should be concerned with iden-
tifying points of completion at different structural levels, and rationalizing
the relationships betweeen them-have to be created anew in each instance;
he is merely saying that the particular way in which the principles apply
will vary from one case to another. (The basic point of Meta-Variations
was, after all, to outline a theory of music that was broad enough to allow
for such varied applications.) We can make an analogy with psycholinguis-
tics here. The phoneme is defined in functional terms as a unit of linguistic
structure, but it also has an acoustic aspect. In acoustic terms any given
phoneme is defined by a certain combination of what Roman Jakobson
called 'distinctive features' (for instance gravelacute or nasalizedlnon-
nasalized). Now the general principles of phonemic structure are applica-
ble to all languages. But the particular combinations of distinctive features
that are significant vary from language to language; the principles have to
be applied differently in different cases, and as a result the psycholinguist
needs to determine the relevant features contextually, that is to say in terms
of each individual language. This does not stop psycholinguistic theory
from being scientific.
We can focus the issue of universalism versus particularism by consider-
ing how it applies to Schenkerian analysis. Making a Schenkerian reduction
involves generalization: the deeper the structural level, the more of the
composition's individuality is discarded, until the background is reached, at
which point the composition has been generalized out of existence. In this
sense Schenkerian theory is not at all particularist, in Brown and Dempster's
terms. Furthermore, Schenker based his theory on laws which he considered
universal in their application, though most people today would regard them
as of limited chronological and geographical scope; it is presumably
because of this universalism that Brown and Dempster cite Joseph Kerman's
criticisms of Schenker as an example of the kind of attacks that can be
directed against scientific music theory-so implying that Schenkerian the-
ory is not only universalist but also scientific. On the other hand, Schenker
himself repeatedly said that his theories should not be thought of as sci-
entific, instead emphasizing what he called the analyst's "obligation toward
parti~ularity"?~ And many contemporary theorists would agree with Wil-
liam Benjamin's view that "the great strength of Schenkerian theory lies in its
ability to characterize an individual tonal work in terms which highlight its
uniqueness and, especially, its uniqueness at higher level^''.^^ So there seem
to be reasons for regarding Schenkerian theory as both universalist and par-
ticularist. We need to go into this in a little more detail.
Kerman's main criticism of Schenkerian analysis is that it "repeatedly
slights salient features in the music"?4 For instance, he complains that
Schenker's analysis of 'Aus meinen Thranen spriessen", the second song
from Dichterliebe, completely ignores one of the song's most telling fea-
tures: the cadences that come at the end of each couplet. (Even more tell-
ing, perhaps, is the fact that the vocal line never resolves to the tonic; only
the piano does.) Now at one level what Kerman says is of course perfectly
true, just as it is true that Schenkerian theory more or less ignores rhyth-
mic structure. But at another level neither statement is true at all. As I have
argued el~ewhere?~ a good Schenkerian analysis presents the results of a
careful consideration of rhythmic factors, though it does so implicitly; and
Schenkerian theory also provides a framework within which a more
explicit rhythmic theory can be developed (which is why so many of the
most important contributions to rhythmic theory in recent years have come
from committed Schenkerians like Carl Schachter, Maury Yeston and Wil-
liam Rothstein). In the same way, Schenker's graph of 'Aus meinen Thranen
spriessen" may not show any specific consideration of the repeated
cadences, but the fact that they disappear in the reduction should not be
taken to mean that they have no significance; it means that, despite their
prominence at surface level, they do not have a structural role in tonal-
functional terms. And that, surely, is one of the reasons for their telling
effe~t.5~ Similarly, while the failure of the vocal line to resolve may not be
explicitly mentioned, it is thrown into sharp relief against the norms of
voice-leading represented by Schenker's graph, and so emerges from the
analysis as a striking discrepancy - as something which runs counter to nor-
mal expectations, which is of course exactly what it is. A great deal of the
value of Schenkerian analysis, it seems to me, lies precisely in the discrep-
ancies that arise between the analytical representation and the familiar sur-
face of the music in question.
A convenient illustration of this is Schenker's middleground sketch of
the opening theme from Beethoven's Sonata, Op. 90 (see Figures 2 and
3).57 A surface reading of this theme would revolve around the half cadence
at m. 16, which is marked by a pause sign, and followed by a I6 chord mov-
ing through I17 and V toward the close. But both the dominant harmony
and the caesura disappear without trace in Schenker's sketch. Instead, he
reads a tonic prolongation extending right up to the I17 chord. In the same
way, he suppresses the registral and dynamic contrasts that are particularly
characteristic of this theme; what are much the highest notes of the passage
(the Es at rnm. 16-17 and 20-1) appear in Schenker's sketch as an inner
voice. One can easily imagine what Kerman might have to say about this!
But we do not need an analytical method to point out the registral con-
trasts; they are among the most immediately obvious features of the music.
What we want an analysis for is to explain the powerful sense of cohesive-
ness and direction that pervades the discontinuities of the musical surface;
and this is precisely what Schenker's sketch does. In the same way, we do
not need Schenkerian analysis to tell us that there is a break at m. 16; we
need it in order to understand why this break seems so curiously evanes-
cent, with the musical motion continuing after it as if nothing had hap-
pened (Beethoven marks the continuation "in tempo"). Again the sketch
does just this. Schenker's analysis, in other words, reveals the divergence
between surface design and underlying structure; it shows how the music is
animated by the tension between foreground and background, whole and
~arts.5~
Each level of a Schenkerian analysis represents a stylistic norm, or a
more or less systematically derived model, against which the elaborations
of the next level stand out in all their particularity. The function of the anal-
ysis, then, is not to reduplicate the composition in question; it is to focus
attention on its individual qualities. And this means that it is wrong to
judge an analysis according to how directly it mi,rrors the surface of the
music, with its tunes and silences and abrupt changes of texture. What mat-
ters is the extent to which it illuminates the surface. Schoenberg's wise-
crack (if he ever actually made it59)about his favourite bits of the "Eroica"
Mit Lebhaftakeit und durchaus mit Empfindung und Ausdruck

Figure 2

Figure 3
Symphony being in the tiny notes of Schenker's graph is all very well, but
it misses one of the main points of the analysis, which is to show how
highly salient features of the music arise as purely surface elaborations
(something, incidentally, that seems to parallel Nottebohm's remark, based
on his studies of the sketches, that Beethoven's creative faculty "often rose
to its greatest heights only at the last moment"60). Schenker's 'Eroica' anal-
ysis does not reduplicate the salient features of the listener's experience; it
interprets them in terms of the metaphor of large-scale contrapuntal struc-
ture. In fact it precisely fulfils the criteria that Schoenberg himself set out
for a "good comparison", in which "we bring closer what is too distant,
thereby enlarging details, and remove to some distance what is too close,
thereby gaining perspective".
Stephen Davies remarks that, in general, analyses "are best to be seen
as illustrations of the theory [that informs them], rather than as evidence
for it"!' This seems to me to go to the heart of the supposed dichotomy
between universalism and particularism. Davies is suggesting that the main
purpose of general principles-theories-is to articulate the analyst's
insights into individual works. This is rather like historical writing, in
which general principles are on the whole invoked as a starting point for
the consideration of the individual instance, and not the other way roundPz
That is to say, the historian's general principles are to be understood not as
universal laws, but as systems of presentation. And when, exceptionally,
such principles are elevated into laws, so that proving their validity
becomes the historian's primary concern, the result is the now discredited
kind of historiography represented by Spengler's or Toynbee's writings.
Schoenberg of course lumped Spengler and Schenker together in his caus-
tic essay "Those who complain about the decline",63 and it is certainly pos-
sible to view Schenker's analyses as being designed to prove the validity of
his theory of musical structure (a theory which, in Schenker's view, ren-
dered all other theories obsolete). In their article, Brown and Dempster
talk a good deal about the ways in which Schenkerian theory can or cannot
be proved to be correct. My complaint is not that this is illegitimate, nor
that it is unfair to Schenker, but simply that it is not the most profitable way
for us to approach what Schenker had to say.
To regard a Schenkerian analysis-or indeed any other type of analysis-
as a demonstration that a given piece is unified is less illuminating than
regarding it as a demonstration of how the piece is unified, or for that mat-
ter of how it is not unified but rather characterized by irreducible structural
contrastsP4 And when we use the word "demonstration" in this sense we
mean something quite different from a mathematical demonstration, which
demands to be accepted (or rejected) on the grounds of its strictly logical
coherence. One cannot reasonably reject a mathematical proof just because
one does not like it. But the same does not apply to a Schenkerian analysis,
or to any other analysis that can be "heard" in musical terms. Of course we
are influenced by the arguments an analyst puts forward, by the corrobora-
tion provided through surface detail, register and so forth. But this is not
enough in itself. Davies writes that
The analyst who sees in the score relationships which cannot be heard by
anyone will not convince us that he has exposed the source of a work's unity.
But the analyst whose analysis allows us to hear relationships of which pre-
viously we were unaware may well convince usPS
However this does not go far enough, either. We do not accept an analytical
interpretation just because we can "hear" it. (After all, it is easy to "hear"
the most preposterous structural relationships if one wants to.) We accept
an analytical interpretation because we find it persuasive; that is, because it
satisfies the particular demands that led us to the analysis in the first place-
demands such as, in Benjamin's words, "the need to memorize a work, the
desire to relate it to other works, or the simple impulse to understand it
better'?

"Choosing our words with great care'', writes Kerman, "we might say
theory deals with those aspects of music that might be thought analogous to
vocabulary, grammar, syntax, and rhetoric in the field of language"P7 The-
orists like Keiler or Lerdahl and Jackendoff think of music theory more or
less on the model of linguistics. But the rationale for analytical practice
which I have outlined would imply an analogy not so much with linguistics
but with literary criticism, the thrust of which (as Judith and A.L. Becker
put it) is "to understand, not to explain"P8 Other ethnomusicologists, such
as Kenneth Gourlaypg have also endorsed literary criticism as a research
model, while Anthony Newcomb has recently applied techniques explicitly
derived from literary criticism to the study of Schumann's music.70 And in
the field of musical historiography, Leo Treitler expresses a similar concep-
tion of the musicologist's work when he writes that "verifiability as the mea-
sure of lawfulness yields ground to intelligibility, coherence, potential
explanatory power"?l Writers like Newcomb and Treitler-and one might
cite Kerman and Gary Tomlinson as well - aim primarily at the reconstruc-
tion of musical meaning through a full consideration of the context in
which the music arose, and they tend to deplore the way in which theorists
ignore the historical context of the music they analyze. Treitler, for
instance, complains that "prevailing modes of structural analysis are anti-
historical, in two respects: they decontextualize their objects in their ration-
alistic treatment of them; and they are taught and practiced without notice
taken of their own historicality or, in general, of the role that particular
models play in the organization of under~tanding"?~
 Nobody can claim that this charge is wholly unwarranted; after all, music
is never heard outside some interpretative context, whether that in which it
originated or another, and to this extent its meaning is obviously determined
contextually. But the issue is not as simple as it might seem. Dahlhaus writes
that "it is quite defensible in methodological terms for us to isolate an object
so long as we do not question the reality of the connections from which it has
been extracted"73 Of course the composition suffers a loss of meaning in this
process. But the loss of meaning may not be so great as to override the
methodological gains, particularly when one is dealing with a culture as
institutionalized and reified as that of Western art m u ~ i c . 7Indeed
~ I would
argue that such a methodology is indispensable not only from the theorist's
point of view-after all, it defines what he does-but also from the historical
musicologist's. For, as Dahlhaus and Treitler have made abundantly clear,
historical facts do not just exist; they are created through historical interpre-
tation. And this applies with particular force to the musical work, which is
not something given but rather (as Dahlhaus puts it) "dissolves into a source,
an authentic text, a composer's intention and a historian's notion as to the
musical significance of the acoustical substrate sketched out by the text or
realized according to the guidelines laid down within In other words,
the musicologist constitutes the musical work through his own interpretation
of it, through conceiving of it as an aesthetic object. And to do this involves
taking up some kind of theoretical position vis-i-vis the work, whether the
theory involved is explicit or merely implied, because it is only by virtue of
some kind of theory that a piece of music can be conceived as a discrete
entity. In this way any study of the music of the past-to the extent that it
maintains the concept of the work at all, rather than breaking up into recep-
tion history-is by definition theoretical as well as historical. To say this is
not to claim primacy for theory over history, as Schenker did when he out-
lined the way in which the history of music would have to be rewritten in the
light of his the0ries.7~(One shudders to think what such a history would have
looked like; maybe Schoenberg's remarks about Spengler were not so far off
the mark.) It is merely to endorse Treitler's own statement that "as investig-
ative procedures, neither analytical nor historical methods can be absolutely
prior to the other. They inform one another in a continuous circle".77
The central proposition I have put forward in this article is that music the-
ory acquires validity not, like scientific knowledge, from being verifiable,
but from serving some useful purpose - in enabling the analyst to arrive at an
interpretation, communicate an insight, or resolve a problem. And I have
cited music history in order to exemplify the kind of way that theory can be
useful and even indispensable within the broader context of musical studies.
But of course there are many other examples that could have been cited too.
Theoretical thinking about music is important just to the extent that it leads
to better musicology, ethnomusicology, composition, or performance. The-
ory, in short, can best be justified by practice.
 1. Carl Dahlhaus, Analysis and Value Judgment (New York: Pendragon Press, 1983 [An-
alyse und Werturteil, Mainz, 1970, trans. S. Levarie]), p. 54.
2. See for instance Bell Yung, "Choreographic and kinesthetic elements in performance
on the Chinese seven-string zither", Ethnomusicology 28 (1984): 505-17, and John
Baily, "Musical structure and human movement", in Peter Howell, Ian Cross and
Robert West (4s.). Musical Structure and Cognition (London: Academic Press,
1985). pp. 237-58.
3. David Lewin, "Music theory, phenomenology, and modes of perception", Music Per-
ception 3 (1986), especially pp. 374-91.
4. Alan Smith, "Feasibility of tracking musical form as a cognitive listening objective",
Journal of Research in Music Education 21 (1973): 200-13.
5. Sylvan Kalib, "Thirteen essays from the three yearbooks Das Meistenverk in der
Musik by Heinrich Schenker: an annotated translation" (Ph.D. diss., Northwestern
University, 1973). ii, p. 293.
6. Ruth Solie, "The living work: organicism and musical analysis'', 19th Century Music
4 (1980): 151.
7. Fred Lerdahl and Ray Jackendoff, A Generative Theory of Tonal Music (Cambridge,
Mass.: MIT Press, 1983), p. xii.
8. Irttne Delitge, "Grouping conditions in listening to music: an approach to Lerdahl
and Jackendoffs grouping preference rules", Music Perception 4 (1987): 325-59.
9. Burton Rosner, '2 Generative Theory of Tonal Music by Fred Lerdahl and Ray Jack-
endoff' (review), Music Perception 2 (1984): 289-90. See also Eric Clarke, "Theory,
analysis and the psychology of music: a critical evaluation of Lerdahl, F. and Jack-
endoff, R., A Generative Theory of Tonal Music:' Psychology of Music 14 (1988):
15-16.
10. Burton Rosner and Leonard B. Meyer, "The perceptual roles of melodic process, con-
tour, and form", Music Perception 4 (1986): 37.
11. Nicholas Cook, "The perception of large-scale tonal closure'', Music Perception 5
(1987): 197-205. In these tests, listeners heard pieces of music in two versions, one of
which was tonally closed while the other was not; they were required to say which ver-
sion was more coherent, created a stronger sense of completion, etc.
12. See Eugene Narmour, Beyond Schenkerism: the Need for Alternatives in Music Anal-
ysis (Chicago: Chicago University Press, 1977) and Alan Keiler, "On some proper-
ties of Schenker's pitch derivations", Music Perception 1 (1983): 200-28.
13. See in particular Narmour's "Some major theoretical problems concerning the con-
cept of hierarchy in the analysis of tonal music", Music Perception 1 (1983): 129-199.
Lerdahl and Jackendoffs theory is also reductive in this sense, because their prolon-
gational reduction-whose background-to-foreground formation corresponds to
Schenker-is complemented by time-span reduction, which works the other way
round.
14. For a full bibliographical citation see note 3 of Brown and Dempster's article.
15. See Boretz' "Nelson Goodman's Languages of A n from a musical point of view'', in
Benjamin Boretz and Edward T. Cone (eds.), Perspectives on Contemporary Music
Theory (New York: Norton, 1972), p. 35. (Orig. publ. in Journal of Philosophy 67
[I9701: 540-52 .)
16. Milton Babbitt, "Past and present concepts of the nature and limits of music", in
Boretz and Cone, op. cit., p. 9.
17. I do not mean to say that Boretz doesn't care about perceptual viability: "I believe
there's a real world out there, because not all of my fantasies work" ("Two replies",
Perspectives of New Music 1512 [1977]: 242). But the decision as to whether an ana-
lytical interpretation "works" is a purely intuitive one; the theory itself is not con-
cerned with perception at all.
18. Milton Babbitt, "Twelve-tone invariants as compositional determinants", in Paul
Henry Lang (ed.), Problems of Modem Music (New York: Norton, 1962). p. 120.
(Orig. publ. in Musical Quarrerly 46 [1960]: 246-59.)
19. A summary of such experiments will be found, with references, in Jana K. Millar,
"The Aural Perception of pitch-class set relations: a computer-assisted investigation"
(Ph.D. diss., North Texas State University, 1984). Millar's list should be supple-
mented by Paul Pedersen, "The perception of octave equivalence in twelve-tone
rows'', Psychology of Music 312 (1975): 3-8, and Carol L. Kmmhansl, Gregory J. San-
dell and Desmond C. Sergeant, "The perception of tone hierarchies and mirror forms
in twelve-tone serial music", Music Perception 5 (1987): 31-78.
20. See Babbitt's famous statement about scientific language and scientific method,
quoted at the beginning of Part I of Brown and Dempster's article.
21. Mary Louise Serafine, Music as Cognition: the Development of Thought in Sound
(New York: Columbia University Press, 1988). p. 60. A parallel argument to the one
I have put forward, leading to a similar conclusion, would be to contrast the Eucli-
dean representation of time in the score with the multiplicity of temporal perceptions
revealed through the phenomenological analysis of music; David Lewin discusses
this in "Music theory, phenomenology, and modes of perception". See also Boretz'
"What lingers on (, when the song is ended)", Perspectives of New Music 1611 (1977),
especially pp. 107-8.
22. On the score as a model see Henry Martin, "Modes of explanation in analytical dis-
course", Perspectives of New Music 1512 (1977): 174-191. On models and metaphors
see Christopher Lewis, "Mirrors and metaphors: reflections on Schoenberg and
nineteenth-century tonality", 19th Century Music 11 (1987): 26-42.
23. This argument is presented in more detail in my book Music. Imagination, and Cul-
ture (forthcoming from Oxford University Press).
24. Benjamin Boretz, "Nelson Goodman's Languages of Art': p. 34.
25. Dika Newlin, Schoenberg Remembered: Diaries and Recollections (1938-76) (New
York: Pendragon Press, 1980), p. 164.
26. Translated in Kalib, op. cit., ii, pp. 52-83.
27. Free Composition (New York: Longman, 1979 [Der Freie Satz. Vienna, 1935, trans.
E. Oster]), p. xxiii.
28. John Rahn, "Logic, set theory, music theory", College Music Symposium 1911 (1979):
114.
29. "The musical object", Current Musicology 5 (1967): 56-87.
30. Carl Dahlhaus, Schoenberg and the New Music: Essays by Carl Dahlhaus, trans. D.
Puffett and A. Clayton (Cambridge: Cambridge University Press, 1987). p. 221.
31. Kalib, op. cit., ii, p. 170.
32. "I was given a vision of the urlinie, I did not invent it!" (Kalib, op. cit., ii, p. 218).
I have discussed this issue in more detail in "Schenker's theory of music as ethics",
Journal of Musicology (forthcoming).
33. See Allan Janik and Stephen Toulmin, Wittgensteink Vienna (New York: Simon and
Schuster, 1973), especially pp. 183-4. Darstellung is of course a perfectly standard
German word meaning "model" or "representation" and does not necessarily carry
the particular interpretation I am discussing here; for instance, Schenker called his
study of Beethoven's Fifth Symphony a Darstellung.
34. Arnold Schoenberg, Theory of Harmony (London: Faber, 1978 [Harmonielehre,
Vienna, 1911, trans. R. E. Carter]), p. 10.
35. Schoenberg's account of "good comparison" seems to echo Goethe: "Thinking by
means of analogy . . . has the advantage that it concludes nothing and aims at nothing
final. Induction, on the other hand, is easily tainted; it holds a predetermined goal
before its eyes and sweeps true and false on before it as it works toward that goal"
(from Marimen und Rejlexionen; translated by William Pastille in "Ursatz: the musi-
cal philosophy of Heinrich Schenker" [Ph. D. diss., Cornell, 19851, pp. 87-8).
36. It is extraordinary how much twentieth century analytical thinking has its origins in
Vienna. Boretz, who was particularly influenced by Carnap, was working in the tra-
dition of Viennese positivism. Schenker came from what is now Poland but worked in
Vienna, and his first-generation disciples were Viennese. So were Schoenberg and the
main analysts who worked under his influence (Keller was born in Vienna; R6ti was
born in Serbia but studied and worked in Vienna). Perhaps this reflects the enduring
influence of Eduard Hanslick, who was the first professor of music at the University
of Vienna, and whose writings in effect demonstrated the necessity for structural anal-
ysis, without clarifying in what specific ways it might be done; all the analysts I have
mentioned might be seen as trying, in different ways, to carry out Hanslick's program.
As for style analysis, the program for this was spelt out by Hanslick's successor at the
University, Guido Adler.
37. Or more precisely, the freely elaborated surface visible in the score of the piece in
question. To the extent that the score itself is (as I said above) a metaphor or model,
we are dealing with a double metaphor, a model of a model (cf. Martin, "Modes of
explanation in analytical discourse", p. 183).
38. In his Octaven u. Quinten u. A . , which Schenker edited and annotated (English ver-
sion "Brahms's Study, Octaven u. Quinten a. A., with Schenker's Commentary Trans-
lated by Paul Mast" in The Music Forum 5 [1980]: 1-196).
39. "The principles of voice-leading, organically anchored, remain the same in back-
ground, middleground and foreground, even when they undergo transformations"
(Free Composition. pp. 5-6).
40. Free Composition, p. 56. For a discussion of Schenker's contradictory statements
about middleground consecutives, see William Benjamin, "Schenker's theory and the
future of music", Journal of Music Theory 25 (1981): 163-4.
41. See Robert Dentan's "Response to Field and Roseman" in Erhnomusicology 28
(1984): 464.
42. Kenneth Gourlay, "Towards a reassessment of the ethnomusicologist's role in
research", Ethnomusicology 22 (1978): 10.
43. Carl Dahlhaus, Foundations of Music History (Cambridge: Cambridge University
Press, 1983 [Grundlagen der Musikgeschichte, Cologne, 1967, trans. by J. B. Robin-
son]), p. 88.
44. See for instance Taruskin's "On letting the music speak for itself: some reflections on
musicology and performance", Journal of Musicology 1 (1982): 338-349.
45. The phrase is taken from Arthur Nestrovski, "Music theory, Saussure, Theoria:' In
Theory Only 1016 (1988): 10.
46. Maybe, as Benjamin seems to imply in "Models of underlying tonal structure: how can
they be abstract, and how should they be abstract?'(Music n e o r y Spectrum 4 [1982]:
48), the difficulty of "hearing" music in terms of such abstract analytical models as set
theory is simply a consequence of their not being built on such familiar foundations.
At the same time, I have difficulty in imagining quite what it might mean to "hear" the
difference between, say, a K and a Kh relationship. But it is easy to be too glib about
this. There is a continuum from "hearing" to abstract conception, and perhaps all one
can sensibly propose is that no analytical model should be more abstract than it has to
be in order to communicate the analyst's interpretation of a work.
47. "Music theory, phenomenology, and modes of perception", p. 382. This, it seems to
me, is really Boretz' position too. Despite the scientific appearance of Boretz' earlier
writings (the later ones, to put it mildly, look different), Brown and Dempster's char-
acterization of him as a would-be scientific theorist obscures what is surely the pri-
mary aim of his work: to contribute to the experienced "vividness (in the senses of
individuality, lucidity, and depth) of musical works" ("Two replies'', p. 241). Perhaps
we would all have understood Meta-kriations better if it had been called "Poetics of
Music Theory in the Form of Six Lessons".
48. "On some properties of Schenker's pitch derivations", p. 224.
49. After all, Schenker introduced his earliest full-scale reductive graph (in the Erlaute-
rungsausgabe of Op. 101) with the words: "Here are shown the lines that Beethoven's
imagination followed". And his method of presentation in Der Freie Satz follows a
generative, not a reductive, sequence.
50. Wilson Coker, "Richmond Browne (ed.), Music Theory: Special Topics" (review),
Music Theory Spectrum 4 (1982): 130.
51. Brown and Dempster, p. 84. This is reminiscent of Lewin's argument, in "Behind the
beyond: a response to Edward T. Cone" (Perspectives of New Music 712 [1969]: 62).
that "analysis is really antithetic to theory"; analysis, that is to say, focusses on the
individual case whereas theory is concerned with generalization. But, as Lewin
implied and as Cone made explicit in his reply (pp. 70-2). the antithesis is a matter
of emphasis rather than of underlying principles. It is not, in other words, a question
of contradiction.
52. Kalib, op. cit., ii, p. 82. Elsewhere in Das Meistenverk Schenker writes that "The
general . . . easily seduces man into becoming complacent, into taking no further
pains with the particular. Continued misperception of the particular unsouls, so to
speak, the understanding of the general; it no longer ripens into truth; it solidifies into
a schema" (translated in Pastille, "Ursatz", p. 159).
53. "Models of underlying tonal structure", p. 36.
54. Joseph Kerman, Musicology (London: Fontana, 1985; American edition, Contemplat-
ing Music, Cambridge [Mass.]: Harvard University Press, 1985), p. 82. Kerman's
remarks about "Aus meinen Thdnen spriessen" will be found in "How we got into
analysis, and how to get out'', Critical Inquiry 7 (1980): 311-31; for Schenker's anal-
ysis see Allen Forte, "Schenker's conception of musical structure", Journal of Music
Theory 3 (1959): 1-30, reprinted in Maury Yeston (ed.), Readings in Schenker Anal-
ysis and Other Approaches (New Haven, 1977): 3-37.
55. Nicholas Cook, A Guide to Musical Analysis (London: Dent and New York: Brazil-
ler, 1987). pp. 61, 77.
56. What Kerman might have reasonably complained about is the fact that Forte did not
draw attention to this in his commentary on Schenker's analysis. The trouble with
Kerman's attack on Schenkerism, as he calls it, is that he tends not to distinguish
between the analytical methodology on the one hand, and the manner in which it is
used-which may be sensitive or insensitive, critical or uncritical-on the other.
57. Schenker's sketch, shown in Fig. 3, is taken from Der Freie Satz, E x . 109a, (1).
58. See Pastille, "Ursatz", pp. 67ff., 134. Present-day Schenkerian analysis, it seems to
me, tends to put too much emphasis on unity per se, rather than on the dynamic rela-
tionships between different levels.
59. See Milton Babbitt (ed. S Dembski and J. N. Straus), Ubrds about Music (Madison:
Wisconsin University Press, 1987), pp. 139-40.
60. Gustav Nottebohm, Two Beethoven Sketchbooks: a Description with Musical Extracts
(London: Gollancz, 1979 [Ein Skiuenbuch von Beethoven, Leipzig, 1865, and Ein
Skiuenbuch von Beethoven aus den Jahre 1803, Leipzig, 1880, trans. J. Katz]), p. 98.
61. Stephen Davies, "Attributing significance to unobvious musical relationships", Jour-
nal of Music lheory 27 (1983): 204.
62. Patrick Gardiner, The Nature of Historical Explanation (London: Oxford University
Press, 1961), p. 90.
63. Translated in Style and Idea: Selected Writings of Arnold Schoenberg, ed. L Stein,
with translations by L. Black (London: Faber, 1975): 203-4.
64. See David Neumeyer's "Reply to Larson", In lheory Only 1014 (1987): 34.
65. "Attributing significance to unobvious musical relationships", p. 212.
66. "Models of underlying structure", p. 28.
67. Musicology, p. 13.
68. Judith and A.L. Becker, "Response to Feld and Roseman", Ethnomusicology 28
(1984): 455. See also Boretz' remarks in "What lingers on'', p. 106.
69. "Ethnomusicologist's role", pp. 25-6.
70. Anthony Newcomb, "Schumann and late eighteenth-century narrative strategies",
19th Century Music 11 (1987): 164-74.
71. Leo Treitler, "On historical criticism", Musical Quarterly 53 (1967): 191.
72. Leo Treitler, "'To worship that celestial sound': motives for analysis'', Journal ofMusi-
cology 1 (1982): 159.
73. Foundations of Music History, p. 27.
74. See Harold Powers, "Language models and musical analysis'', Ethnomusicology 24
(1980), p. 8.
75. Foundations of Music History, p. 35.
76. Free Composition, p. 27.
77. "'To worship that celestial sound"', p. 155.