You are on page 1of 377

Conceptualizing Music:

Cognitive Structure,
Theory, and Analysis

Lawrence M. Zbikowski


AMS Studies in Music
Lawrence F. Bernstein, General Editor

Editorial Board
Joseph Auner
Scott Bur nham
Richard Crawford
Walter Fr isch, Chair
Sarah Fuller
Robert Judd
Janet Levy
Jessie Ann Owens
Kerala Snyder
Judith Tick
Gary Tomlinson
Cognitive Structure, Theory, and Analysis

Lawrence M. Zbikowski

Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
So Paulo Shanghai Singapore Taipei Tokyo Toronto

Copyright 2002 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.
198 Madison Avenue, New York, New York 10016

Oxford is a registered trademark of Oxford University Press

All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the pr ior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data
Zbikowski, Lawrence Michael.
Conceptualizing music : cognitive structure, theory, and analysis / Lawrence M. Zbikowski.
p. cm. (AMS studies in music)
Includes bibliographical references (p. ) and index.
ISBN 0-19-514023-0
1. Musical perception. 2. Musical analysis. 3. Cognition. I. Title. II. Series.
ML3838 .Z25 2002
781'.11 dc21 2001058756

1 3 5 7 9 8 6 4 2
Printed in the United States of America
on acid-free paper
dem andenken meiner Mutter:
Anneliese Margerite Zbikowski, 1926 1999
This page intentionally left blank

O n picking up a book with the title Conceptualizing Music: Cognitive Structure,

Theory, and Analysis, one might reasonably assume that it deals with music cog-
nition and how our knowledge of that discipline can be applied to music theory and
analysis. This book does not do that, or at least not in a simple way. To begin, it does
not have much to say about the fairly large body of research usually placed under the
rubric music cognition. This work, having been developed out of music psychol-
ogy and infor med by recent research in the brain sciences and mind sciences, pro-
ceeds by carefully crafted exper iments, which are subjected to closely argued statis-
tical and logical analysis. As practiced by such eminent and able researchers as Carol
Krumhansl, John Sloboda, and David Huron, the study of music cognition has told
us much about how humans process sonic and musical infor mation.
But this book proceeds in a somewhat different way. Drawing on the same body
of research from the brain sciences and mind sciences that shaped studies in music
cognition, it explores how basic cognitive capacities are specified for understand-
ing music. The project takes inspiration from recent work in linguistics and rhetor ic
by researchers like Ronald Langacker, Gilles Fauconnier, Mark Turner, and George
Lakoff, and it is based on the assumption that musical understanding relies not on
specialized capacities unique to the processing of patterned sound but on the spe-
cialized use of general capacities that humans use to structure their understanding
of the everyday world. The methodology, in consequence, relies not on exper i-
mental design and data analysis but on using a broad and quite extensive body of
research to inter pret recurrent tropes of musical understanding. These tropes
involve such things as the importance to musical understanding of relatively small
and compact musical phenomena like motives, themes, and chords; the use of
terms grounded in nonmusical domains terms like space and depth to char-
acter ize musical events; and the reliance on patter ns of logical inference to reason
about music.
The result of this investigation is a theoretical perspective on musical organiza-
tion, but one rather different from what usually counts as music theory. To make
sense of this claim requires a bit of explanation about the intellectual context of
music theory, for music theory is, within the rolling seas of humanistic studies, a
viii pre fac e

rather strange fish. Put bluntly, it is clear that much of what music theory does, as a
discipline, does not count as any sort of theory in moder n scholarship. This is exem-
plified by each of the two distinct but related and intertwined strands that make up
contemporary music theory. One strand is occupied with pedagogy, the other with
speculative and highly systematic approaches to musical organization.
Music theor y, as it is presented in the classroom, is most often engaged with a
careful and often relentless explication of what, for want of a better ter m, we can call
musical grammar. Consider the following, from Edward Aldwell and Carl Schachters
Harmony and Voice Leading:
Like VII,V has 2 as its bass.V , in fact, resembles VII6 so closely that they are almost
interchangeable chords. The bass of V is a more neutral tone than that of V (or, as
you will see,V ) and can move convincingly either to 1 or to 3. Consequently,V , like
VII, forms a natural connection between I and I and appears very frequently as a
passing chord within an extended tonic.1

The prose and ter minology are impressively dense. But one should not be misled
into thinking that the authors are concer ned only with abstruse compositional
techniques, for immediately after this excer pt Aldwell and Schachter refer to a pas-
sage from an impromptu by Franz Schubert that illustrates the niceties of voice
leading with which they are concer ned. Their assumption is that the reader is famil-
iar with the music and counts it as typical, and it is this familiar ity that provides a
phenomenological anchor for what might appear to be rather thick jargon. If you
know Schuberts impromptu, or (better yet) can summon it in your sonic imagina-
tion when reading the example in score, Aldwell and Schachters point about the
harmonization of the second scale-step in the bass is not just clear but even obvious.
At the heart of pedagogical music theor y are familiar or typical examples of
music, the mysteries of which are revealed by a music theor ist (or theor ists) eager to
share the secrets and wonder of this music with others. As elegant and persuasive as
this approach might seem, it is, within our cur rent cultural climate, more than a lit-
tle unrealistic: music by Schubert and his contemporar ies is often unfamiliar to the
students who read Aldwell and Schachters text (or any of a number of similar texts)
and is not typical of the music that resonates through these students digitized and
hypercommercialized environments. That this should be so is often regarded as
symptomatic of an illness of the late twentieth centur y, an illness that leads to an
insufficient engagement with the g reat traditions of Western culture. For Classical
music (as it is so styled), the antidote is music theory. Music theory, with its careful
explication of the musical grammar of Mozart and Beethoven and Schubert, thus
becomes the last redoubt against the dissolution of Western culture represented by
a dwindling interest in the music of eighteenth- and nineteenth-centur y Europe.
If, for a moment, we step back from Aldwell and Schachters text and generalize
its intent beyond the specific repertoire relative to which it is framed, we might be
able to avoid this rather sanctimonious stance.We could argue that music is a highly
complex and idiosyncratic mode of human communication and that having a knowl-

1. Edward Aldwell and Carl Schachter, Harmony and Voice Leading, 2nd ed. (San Diego: Harcourt
Brace Jovanovich, 1989), 112.
p re fac e ix

edge of the grammar of this mode of communication is essential to its deeper

appreciation, no matter what for m music might take. The argument is a familiar
one to me, not the least because I often find myself making it. And yet, something
rings hollow. The grammar that music theor y teaches is unavoidably tied to the
repertoire to which it refers, and just how this is generalized to apply to other reper-
toires is not immediately apparent: I know of no theory text that explains how the
grammar of Schuberts musical discourse is manifested in the music of, say, John
Coltrane or Pr ince. Given its isolation from contemporary culture, the music the-
ory of the classroom appears to be little more than a ghost that haunts the echoing
halls of a crumbling cultural empire.
The second strand of music theory partakes of the systematic quality inherent
in grammars but generalizes it away from natural language and toward a free-stand-
ing intellectual construct. As an example of this sort of theory, consider the follow-
ing brief passage from David Lewins analysis of a section of the opening of Claude
Debussys piano prelude Reflets dans leau. In this excer pt, X, Y, and Z represent
specific collections of musical notes, RI refers to the compound operations retro-
grade and inversion, RICH is a function that effects ser ial transfor mations, and T
refers to transposition:
In measure 10 the music of measure 9 is repeated and extended. The crescendo recurs.
In the melody the repetition gives rise to a rotated for m of Z1, marked rot Z1 on
figure 10.10. Rot Z1 is Bb-Ab-F-Eb; it embeds ser ially the or iginal for m of Y, Ab-F-
Eb, and precedes this Y by its overlapping inverse-RI-chained for m Bb-Ab-F. (Bb-Ab-
F is RICH1(Ab-F-Eb).) This relationship is more or less inherent in the der ivations
of X,Y, T(X), their repetitions, and Z1.2

It is, of course, something of a challenge to evaluate this passage in isolation. Not

only is it just one part of a larger analysis, but also it comes late in a book occupied
with various and sundry applications of formal algebra and mathematical mapping
theory to music. Nonetheless, what should be clear is that more than familiar ity
with Debussys prelude is required to make sense of Lewins inter pretation of the
passage. The reader must also be familiar with a style of abstract thought that is
bound to appear cabalistic to the uninitiated, one in which the transfor mation of
musical entities is at least as important as the entities themselves. For some, the inac-
cessibility of this mode of thought is one of its char ms. For others, it is proof of the
hermeticism and ir relevance of music theory.
Before continuing, I should make clear that I have deep respect for the theor ists
whose work I have cited. I use this work in my teaching and continue to be intr igued
and stimulated by it. I also want to emphasize that these excer pts by no means
reflect all that there is to music theory. I take them as representative of two strands
of thought that are replicated and woven together in all sorts of different ways to
create the texture of contemporar y theoretical practice.What is important for my
purposes here is that the practice of contemporary music theory is not like that of
contemporary cultural or social theory. Instead of probing the cultural or histor ical

2. David Lewin, Generalized Musical Intervals and Transformations (New Haven, Conn.:Yale Univer-
sity Press, 1987), 234.
x pre fac e

context for musical utterances, or the complex networks of social interaction that
give rise to musical behavior, music theory continues to focus on details of musi-
cal discourse with an obsessiveness that is both maddening and quixotic to cultural
and social theor ists.
Given the impressive traditions of music theory and their influence on my own
thought, I cannot guarantee that what I offer here is a great improvement on this
situation. One of the things I want to do in the chapters that follow, however, is to
develop a somewhat different view of music theory one that sees music theory as
a response to a problem. The problem is that of musical understanding: how it is
that we can make sense of sequences of nonlinguistic patter ned sound, that we can
do so with amazing rapidity, and that (often as not) we can return to these or sim-
ilar sound sequences and find continued reward. I will argue that our understand-
ing of music relies on a play of concepts and conceptual structures that emerges
from training basic cognitive capacities on musical phenomena and that music the-
ory and music analysis der ive from this play.
This mode of inquiry is not one common in the discipline of music theor y,
despite its similar ities to work done by Leonard B. Meyer, Eugene Nar mour, and
Robert Gjerdingen. It does, however, share features with the approach to music evi-
dent in Susan McClarys recent Conventional Wisdom. That this should be so reveals
a debt on my part, for it was Professor McClar y who first suggested that I read
Mark Johnsons The Body in the Mind back in the late 1980s, and this had a profound
effect on my research. In her book, McClary explores the cultural and histor ical
forces that have shaped genres like opera and the blues and compositional practices
like tonality. My focus is on a somewhat different set of forces: those that shape the
way humans think. It seems inevitable that these forces are in some sort of grand,
if ill-defined, dialogue with cultural and histor ical agencies: it is, after all, human
cognitive processes and human culture and histor y about which we are talking.
That cultural and social theor y have turned a deaf ear to this dialogue is not sur-
prising: cognitive theory has had little room for and less patience with culture, and
the detail of its investigative method is no match for the epic sweep of high theo-
retical practice at its best. And yet it seems we must, at some point, come to ter ms
with cognitive structure, for if we do not develop an understanding of how cogni-
tive processes shape the basic mater ials of thought, we risk accepting these mater i-
als as things g iven by nature, just as culture and histor y and music, for that mat-
ter were once assumed to be given by nature.
A glimpse of the problem can be seen in McClarys compelling analysis of Robert
Johnsons 1936 Cross Road Blues. McClary contrasts Johnsons blues with those of
Bessie Smith, noting that the influence of Johnsons music on white British blues play-
ers of the 1960s was due in part to a misconception: the Br itish musicians believed
that the idiosyncrasies of Johnsons blues style represented authentic blues practice.
Describing the unique sound of Johnsons Cross Road Blues, McClary writes,
An affect of dread and entrapment pervades this tune partly the result of his stran-
gulated, falsetto vocals and his uncanny replication of that timbre on the guitar.
Moreover, Johnsons percussive guitar pulse, which locks in at the eighth-note level,
allows almost no sensual movement: even though Johnsons singing constantly strains
p re fac e xi

against that beat, the listeners body is regulated by those short, aggressively articulated
units. The guitar thus seems to represent simultaneously both oppressive outside forces
and a desperate subjectivity fighting vainly for escape.3

I have no quar rel with McClar ys analysis indeed, in this short passage, she has
captured a number of the essential features of Johnsons performance style. The
difficulty comes with the ultimate justification for the affect produced by the song.
Why do Johnsons vocals and guitar work yield dread and entrapment rather than
joyful anticipation and a feeling of liberation? Clearly, the rhythmic framework is
important, but why is it that the r igidity of Johnsons beat constrains us rather than
providing a secure foundation from which we can coolly regard his plaint? These
are not easy questions, all the more so because of the relatively unique character of
Johnsons recording when compared with other blues recordings of the per iod, and
because hear ing Cross Road Blues as something other than a moving, haunting
song is to misunderstand it rather thoroughly. Cultural, social, and histor ical context
cannot, by themselves, explain the or igin of our affective response to the song, for
our broad agreement on the effect of Johnsons music transcends these implements
of high theory (even if they have a profound influence on what we do with Cross
Road Blues once we have heard it). I propose that explor ing the way cognitive
structure infor ms our understanding of music gives us a way to account for the
source of our broad agreement on the affect that pervades Johnsons blues and can
help us understand better the ways culture, society, and history reshape musical
Again, the way I want to accomplish this is by reconceptualizing what it means
to theor ize about music. This can be done by approaching music theor y from the
perspective provided by recent work in cognitive science. My intellectual guide-
posts include not only the wealth of work done in the mind sciences and the brain
sciences but also contemporar y and histor ical ways of theorizing about music.
These theor ies of music now with theory understood in somewhat more tra-
ditional ter ms capture important aspects of how it is we structure our under-
standing of music. At their best, they represent a technical and systematic articula-
tion of an accord on what matters in music that is similar in kind, if not ter minology,
to our accord on how Robert Johnsons blues move us: we agree on how music is
put together and we agree on what music means because both are structured by
basic cognitive processes through which we organize our understanding of the
world. And so music theor y this rather strange fish in the seas of humanistic
scholarship may yet tell us some quite interesting things about the cultural and
social construction of music. Understanding the way music theory instantiates cog-
nitive processes will also help explain its continuing value. It will offer such help
whether that theor y be prosaic as within the classroom or within cr itical dis-
course (since even the most radical of cultural or sociological theor ists inevitably
makes recourse to basic music-theoretical constructs) or poetic, as with the elu-
sive and allusive constructs of abstract theory.

3. Susan McClary, Conventional Wisdom: The Content of Musical Form (Berkeley: University of Cali-
fornia Press, 2000), 51.
xii pre fac e

As should be apparent from the preceding, my intent here is to address not only
music theor ists but also musicologists and ethnomusicologists who find the chal-
lenges of theorizing about music intr iguing.What follows will also be of interest to
those with either a professional or avocational interest in cognitive science, for
music presents a number of interesting problems for cognitive processing, not the
least of which are its embeddedness in culture and the demands it places on real-
time processing. In entertaining the thought of such an audience, however, I should
briefly clar ify a distinction I draw between sound and music and the cognitive abil-
ities related to each, which is based on three suppositions. First, not all sound is
music. Second, an account of how humans process sound is not the same thing as
how they understand music. Third, phenomena relevant to musical understanding
exist at a conceptual level that is, at a level of cognitive activity at least potentially
accessible to conscious thought. I should emphasize that I regard the conceptual
level as occupying only a small part of our total cognitive activity, and I am not at all
opposed to efforts by music psychologists and others who try to explain how struc-
tures at the preconceptual level connect with and motivate structures on the con-
ceptual level. For me, however, it is at the conceptual level that I find the most pro-
foundly interesting questions, for concepts are the tools that allow us to construct
the complex notions essential to musical understanding. From this perspective, con-
ceptualizing music is fundamental to inquir ies about music, whether those be from
the perspective of music cognition, or ethnomusicology, or musicology, or theory.

I n the course of writing this book I enjoyed the assistance of many people, all of
whom improved the product immeasurably. Among the many colleagues who
read drafts of material or responded to presentations I have given, are the following:
Kofi Agawu, Holly Aksnes, Jeanne Bamberger, Larry Barsalou, Philip Bohlman,
Candace Brower, Scott Bur nham, Clif Callender, Thomas Chr istensen, Martin
Clayton, Rick Cohn, Nick Cook, Arnie Cox, Bob Gjerdingen, Robert Hatten, Bob
Holzer, Brian Hyer, Rich Janda, Carol Krumhansl, the late Jim McCawley, Marc
Perlman, Anne Robertson, John Rothgeb, Janna Saslaw, Martin Stokes, and Mark
Turner. Their comments, suggestions, and quer ies helped me strengthen and refor-
mulate my arguments. Ben Brinner, Gilles Fauconnier, Douglas Hofstadter, Travis
Jackson, Bobby Short, Sumarsam, and Susan Youens were generous in their response
to questions I posed to them and provided me with infor mation and insight I sim-
ply could not have gained elsewhere. Students in my classes and seminars at the
University of Chicago listened to many of the ideas presented in this book, to
which they responded with questions, challenges, and comments. Although it
would be impractical to acknowledge all of them here, they contributed immensely
to the shape those ideas have taken in this book.
Deborah Gillaspie of the Chicago Jazz Archive and Regenstein Library patiently
endured my requests for obscure recordings and was indispensable in helping me
find mater ials that I needed. The musical examples were expertly prepared by Jr-
gen Selk of Music Graphics Inter national. Philipp Goedicke and Heinr ich Jaeger
helped me with the translation of difficult passages in Ger man, and my brother
Gene Zbikowski assisted in decipher ing some of the more recalcitrant clauses in
Rameaus and Prousts French. Jos Antnio Martins and Aron Topielski served as
my research assistants, ably tracking things down, organizing my sprawling files, and
getting to know the photocopier, perhaps better than they wanted to. Maribeth
Payne, when she served as Executive Editor for Music at Oxford University Press,
listened to my ideas for the book (at least once at the r isk of missing her train) and
then provided valuable encouragement and assistance in the process of turning a
proposal into a final manuscript.
Institutional support came through a yearlong fellowship at the Franke Institute
xiv ac k nowle dg m e nt s

for the Humanities (for merly the Chicago Humanities Institute) at the University
of Chicago and from the Department of Music of the University of Chicago. The
latter provided not only a home for my research but financial support as well. Some
of the recordings discussed in chapter 5 were made available by the Archives for
Traditional Music in Bloomington, Indiana, and by the Bowling Green Sound
Recordings Archives, Bowling Green, Ohio.
Finally, production of this book was facilitated by subventions from two schol-
arly societies. The Society for Music Theor y provided a subvention to defray the
costs of producing the musical examples, and the result of this assistance can be seen
in the number and appearance of the examples I was able to include. The Publica-
tions Committee of the American Musicological Society was equally enthusiastic
about the book and ag reed to include it in its ser ies, AMS Studies in Music. This
decision not only made the book more affordable but also gave me an excellent and
endlessly patient editor in Lawrence Ber nstein.
My family Vicky Long, Anna Katia, and Andrei Nicolai had to put up with
sharing me with this project, and I am afraid they got the short end of the deal more
than once.Without their support, understanding, and uncr itical companionship, this
book would have been much the poorer.

My musical examples generally adhere to the readings of the following editions: the
new Beethoven Werke; the Johannes Brahms Smtliche Werke; the Neue Mozart Aus-
gabe; Casimir is edition of the works of Palestr ina; the Neue Schubert Ausgabe; Clara
Schumanns edition of the Lieder of Robert Schumann; the new Richard Wagner:
Smtliche Werke (in conjunction with the C. F. Peters score of Tristan); and the Col-
lected Works of Giaches Wert for Corpus mensurabilis musicae. Some indications of
editor ial intiative in those editions generally with respect to dynamic markings
added on the basis of parallel readings have been realized tacitly in my examples
for the sake of clarity of appearance. The setting of Bernhard Kleins Trockne Blu-
men was prepared from the 1822 edition by E. H. G. Christiani in Berlin.
When dealing with histor ical sources wr itten in languages other than English, I
have generally followed one of two practices. Where there is no translation of the
work, or where I think existing translations are not wholly satisfactory, I provide my
own translation, with the or iginal in a footnote.Where satisfactory translations exist,
I use these and do not include the or iginal in the footnote. In a few cases, I have
thought readers might like to make reference to both the or iginal and a translation
and consequently have included citations for both in the footnotes.
Some of the mater ial that follows has been previously published elsewhere and
is used here by kind per mission of the journals in which it or iginally appeared. This
material includes the following: Conceptual Models and Cross-Domain Mapping:
New Perspectives on Theor ies of Music and Hierarchy, Journal of Music Theory
41/2 ( 1997, Journal of Music Theory); Musical Coherence, Motive, and Catego-
rization, Music Perception 17/1 ( 1999 by The Regents of the University of Cali-
fornia); and The Blossoms of Trockne Blumen: Music and Text in the Early
Nineteenth Century, Music Analysis 18/3 ( 1999, Blackwell Publishers, Ltd.).

Introduction: Conceptualizing Music 3

part i. aspects of cognitive structure

1. Categorization 23
2. Cross-Domain Mapping 63
3. Conceptual Models and Theor ies 96

part ii. analysis and theory

4. Categor ization, Compositional Strategy, and Musical Syntax 137
5. Cultural Knowledge and Musical Ontology 201
6.Words, Music, and Song: The Nineteenth-Century Lied 243
7. Competing Models of Music: Theories of Musical Form and Hierarchy 287

Conclusion: Cognitive Structure, Theory, and Analysis 325

Bibliography 335
Index 353
This page intentionally left blank
This page intentionally left blank
conceptualizing music

E arly in the opening volume of A la recherche du temps perdu, Marcel Proust pre-
sents the first of a number of memorable accounts of listening to music. He
describes Charles Swanns initial encounter with the andante of Vinteuils sonata for
violin and piano:
Doubtless the notes which we hear at such moments tend, according to their pitch
and volume, to spread out before our eyes over surfaces of varying dimensions, to trace
arabesques, to give us the sensation of breadth or tenuity, stability or capr ice. But the
notes themselves have vanished before these sensations have developed sufficiently to
escape submersion under those which the succeeding or even simultaneous notes have
already begun to awaken in us. And this impression would continue to envelop in its
liquidity, its ceaseless overlapping, the motifs which from time to time emerge, barely
discernible, to plunge again and disappear and drown, recognized only by the partic-
ular kind of pleasure which they instill, impossible to descr ibe, to recollect, to name,
ineffable did not our memor y, like a laborer who toils at the laying down of firm
foundations beneath the tumult of the waves, by fashioning for us facsimiles of those
fugitive phrases, enable us to compare and to contrast them with those that follow.
And so, scarcely had the exquisite sensation which Swann had exper ienced died away,
before his memory had furnished him with an immediate transcr ipt, sketchy, it is true,
and provisional, which he had been able to glance at while the piece continued, so
that, when the same impression suddenly returned, it was no longer impossible to

What Proust summons in this lyr ical, enchanted vignette is the awakening and ini-
tial consolidation of musical understanding. Swanns first impressions of Vinteuils
sonata are vague and unfor med, his mind simultaneously struggling with and savor-
ing the ineffability of the music. But then, with the aid of memory, patterns emerge.
Although these are incomplete and subject to revision, they offer him a way to
make sense of the music, even as it continues to play.
Conceptualizing Music provides an exploration of the process of musical under-

1. Marcel Proust, Swanns Way (vol. 1 of Remembrance of Things Past), trans. C. K. Scott Moncr ieff
and Terence Kilmartin (New York: Vintage Books, 1981), 227.

4 introduct ion

standing that is, the process through which those liquid impressions spoken of by
Proust are transfor med into structures that make it possible to grasp music. In what
follows, I argue that Swanns or anyones understanding of music draws on the
same cognitive processes that humans use to organize their understanding of the
world as a whole. Confronted with musical sound, these processes create musical
concepts, the things that enabled Swann to gain a g rasp of the music. The act of
conceptualizing music is the beg inning of a whole chain of cognitive events that
allow us to theor ize about music and to analyze the things that populate our aural
past, present, and future.
The notion that Swanns musings might give rise to musical concepts demands
some further consideration. Concepts are often thought of as highly stable cognitive
structures of considerable complexity, a view hardly commensurate with the ephemera
attended to or produced by Swann. Recent work in the brain sciences and the mind
sciences, however, has changed how we view concepts. There are now persuasive
arguments that concepts are quite fluid, that they are not ir revocably wedded to
words or to concrete representations, and that they are not even unique to our
species.2 In consequence, the provisional replicas of musical phrases that make it
possible for Swann to secure a foothold in the unfamiliar ter rain of Vinteuils sonata
need not automatically be excluded from the conceptual domain. In fact, they are
very much like the concepts we use to structure our understanding of the every-
day world.
This same work in the brain sciences and the mind sciences suggests a way to
account for the apparent simplicity and immediacy of musical understanding, which
seems incommensurate with the complexity of musical structure. For instance, the
sonata by the fictional Vinteuil is intended to be a relatively complex contemporary
work that has captured the fancy of the musical elite of the Paris salons. Nonethe-
less, Swann, whose connoisseurship does not extend to music, is able to gain a grasp
of the work almost immediately. That he is able to do so is not simply novelistic
license but is, in fact, thoroughly plausible: almost everyone has had the exper ience
of listening to an unusual composition or exotic repertoire and being able to make
something of it. This possibility has suggested to some a latent musicality in humans
comparable to the sort of competence for language proposed by Noam Chomsky.3
Competencies of this sort raise as many questions as they answer, however, partic-
ularly where cultural entities such as music are concer ned. It seems more promising
to follow the path of researchers who have rejected linguistic competence as a given

2. See, for instance, Douglas R. Hofstadter and the Fluid Analogies Research Group, Fluid Concepts
and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought (New York: Basic Books,
1995), chaps. 5 6, 8 10; Gerald M. Edelman, The Remembered Present: A Biological Theory of Consciousness
(New York: Basic Books, 1989), chap. 8; and Donald R. Griffin, Animal Minds (Chicago: University of
Chicago Press, 1992), chap. 6.
3. The notion of musical competence has generated a range of commentar y and scholarship. See
John Blacking, Music, Culture, and Experience, in Music, Culture, and Experience: Selected Papers of John
Blacking, ed. Reginald Byron, with a foreword by Bruno Nettl (Chicago: University of Chicago Press,
1995), 228 31, on musical competence and culture; Steven Pinker, How the Mind Works (New York:
W.W. Norton, 1997), 528 38, on musical competence and its relation to other competencies; and Allan
Keiler, The Origins of Schenkers Thought: How Man Is Musical, Journal of Music Theory 33 (1989):
273 98, on a nineteenth-century conception of musical faculties akin to musical competence.
conc e p tual i zi ng mu s i c 5

and who have set about explor ing the cognitive foundations of language.4 Their
task has been to discover what processes are basic to human cognition and then to
determine how they are specified for language. For my part, I would like to explore
how some of these same general cognitive processes are specified for music. By this
means I hope to account for the apparent ease and real rapidity with which we can
conceptualize a highly complex, completely unfamiliar music on our first encounter,
without having to postulate the faculty of musical competence.
At a bit more of a remove, but no less important for a complete account of cog-
nitive processing, is the way concepts come to be organized into the more extended
cognitive structures with which our thought is usually occupied. This is where the-
ories come into play, for theor ies are the cognitive tools that guide the way we rea-
son about the things we exper ience. At first, this might seem to be a rather special-
ized use of the notion of theory, for the theory with which much current literature
is occupied within and without music scholarship is hardly the stuff of every-
day exper ience. Recent research has countered this view of theory by demonstrat-
ing that the elegant and abstract theor ies of science have much in common with the
tools for reasoning used by very young children.5 Theories are the basic means by
which we make our exper ience coherent and guide further action. The rough-and-
ready transcr ipt that guides Swanns listening thus has something in common with
the more fully articulated and systematic structures we usually associate with the
idea of music theory.
That music theory might have alliances with everyday thought processes is a
provocative claim. On the one hand, music theory often manifests itself as a relent-
lessly practical discipline: a codification of the scales, chords, and grammatical rules
proper to a highly circumscr ibed portion of musical discourse, assembled with the
intent of rendering music comprehensible to those who would become musically
literate. On the other hand, music theory can reach into the far cor ners of abstrac-
tion to embrace complicated mathematical concepts or the arcane symbologies of
voice-leading graphs, as any reader of the Journal of Music Theory, Music Theory Spec-
trum, or Music Analysis will quickly discover. Nonetheless, I want to argue that music
theory, in all its diverse forms, reflects the same basic processes that guide our under-
standing of the everyday world. Theorizing about music is an activity specialized
only in its domain, not in the cognitive processes it involves.
What might these cognitive processes be, and how would they manifest them-
selves? To answer these questions, let us begin at a beginning, with two theor ies of

4. For general introductions to some of the working assumptions of cognitive linguists, see George
Lakoff, The Invariance Hypothesis: Is Abstract Reason Based on Image-Schemas? Cognitive Linguistics
1 (1990): 39 74; and Michael Tomasello, Introduction: A Cognitive-Functional Perspective on Lan-
guage Structure, in The New Psychology of Language: Cognitive and Functional Approaches to Language Struc-
ture, ed. Michael Tomasello (Mahwah, N.J.: Erlbaum Associates, 1998), viixxiii. In the field of cognitive
linguistics, Ron Langackers work is particularly notable for its thoroughness and its systematic approach.
See Langacker, Theoretical Prerequisites, and Descriptive Application, vols. 1 and 2, respectively, of Foundations
of Cognitive Grammar (Stanford, Calif.: Stanford University Press, 1987, 1992); idem, Grammar and Con-
ceptualization (Berlin: Mouton de Gruyter, 2000).
5. See Alison Gopnik and Andrew N. Meltzoff, Words, Thoughts, and Theories (Cambridge, Mass.:
MIT Press, 1997); and Alison Gopnik, Andrew N. Meltzoff, and Patricia K. Kuhl, The Scientist in the Crib:
Minds, Brains, and How Children Learn (New York:William Morrow, 1999).
6 introduct ion

music from Greek antiquity. These theor ies, and the music to which they refer, are so
unfamiliar that even many who make music theory the focus of their research have
only passing knowledge of what they involve. But there is an advantage in this unfa-
miliar ity, for the disor ienting effect it can have also serves to loosen our notions about
what a theory of music, or a theoretical construct, should be. Despite connections
with music theory as it was practiced in Europe and (to a lesser extent) in the Arabic
world, Greek music theory of antiquity is the theory of an alien society. Nonetheless,
this theory was a beginning one of the starting points for accounts of musical orga-
nization in the Western tradition. We can discern within it, therefore, the theor ists of
antiquity grappling with basic constructs equivalent to, if still different from, the sort
of constructs taught in beginning music theory classes today. Both aspects will allow
us to see the role basic cognitive structures play in our understanding of music.

ancient music theory and

modern cognitive science
Those who wrote on Greek musical practice in antiquity concer ned themselves
with a wide variety of topics, including the place of music in society, musical aes-
thetics, the construction and nature of musical instruments, and the organization of
pitch mater ials. Specific discussions of pitch mater ials the usual topic of disquisi-
tions more directly or iented to music theory centered around the set of pitch
relationships that has come to be called the Greater Perfect System.6 As shown in
figure I.1, this consisted of a set of four tetrachords (hypaton, meson, diezeugmenon,
and hyperbolaion), which, together with one additional note (called Proslambanom-
enos), provided the framework for a two-octave system of pitches basic to Greek
music. The end points of the tetrachords (the notes Hypate hypaton, Hypate meson,
Mese, and so on), together with Proslambanomenos, were regarded as stable, unmov-
able pitches. There was, however, no fixed standard for tuning the note-names
given in figure I.1 are simply for the pur poses of illustration.Within the boundar ies
marked by each tetrachord were two other pitches, whose placement varied accord-
ing to which of three different genera was understood to be in play. (In the tetra-
chord hypaton, for instance, the variable notes above Hypate hypaton were Parhypate
hypaton and Lichanos hypaton.) The diatonic genus located the movable pitches in a
manner analogous to moder n diatonic scales (for instance, given the reference
pitches on fig. I.1, Parhypate hypaton in the diatonic genus would be equivalent to
F3, and Lichanos hypaton would be equivalent to G3), but the chromatic and enhar-
monic genera situated the pitches in ways that have no comfortable analogue in
modern scale construction. The result was a system in which the placement of
movable pitches could vary widely and in which intervals between successive
pitches could be smaller than a half step and larger than a whole step.

6. For a more detailed discussion of the Greater Perfect System and its place in Greek theor y of
antiquity, see Andrew Barker, ed. and trans., introduction to Harmonic and Acoustic Theory (vol. 2 of Greek
Musical Writings), Cambridge Readings in the Literature of Music (Cambr idge: Cambridge University
Press, 1989), 12 13. For a more general overview of Greek music theor y and a thorough discussion of
the sources, see Thomas J. Mathiesen, Apollos Lyre: Greek Music and Music Theory in Antiquity and the
Middle Ages (Lincoln: University of Nebraska Press, 2000), part 4.
conc e p tual i zi ng mu s i c 7

!"# !"#"$%&'"()*+,-*.$
$%# !"#"$/-"0"123".*.$

&%# 4,(,3"5"

!%# 6"5"
$'# 7&',#"$3"5*.$

&'# 7&',#"$%&',#*.$
!'# 4(*5+,3),.*3".*5$

figure I.1 Diagram of the Greater Perfect System of Greek music theory

There were two main theoretical approaches to presenting the elements of the
Greater Perfect System. The Pythagorean approach der ives from a metaphysics asso-
ciated with Pythagoras of Samos, who lived during the sixth centur y b.c. Its hall-
marks are a persistent interest in number and the deployment of numer ical con-
ceptions in cosmological contexts. In contrast to this, the Aristoxenian approach,
associated with the Peripatetic school of the fourth century b.c., places little reliance
on number, trusting instead in observation and reason as the means to knowledge
about music.

Pythagoras and the Blacksmiths

Nicomachus of Gerasa, a mathematician and har monist wr iting around the begin-
ning of the second century a.d., described Pythagorass discovery of the basic pr in-
ciples of music theory as follows:
He [Pythagoras] was plunged one day in thought and intense reasoning, to see if he
could devise some instrumental aid for the hear ing which would be consistent and
not prone to er ror, in the way that sight is assisted by the compasses, the measur ing rod
and the dioptra, and touch by the balance and by the devising of measures; and hap-
pening by some heaven-sent chance to walk by a blacksmiths workshop, he heard the
hammers beating iron on the anvil and giving out sounds fully concordant in combi-
nation with one another, with the exception of one pairing; and he recognized among
them the consonance of the octave and those of the fifth and the fourth. He noticed
that what lay between the fourth and the fifth was itself discordant, but was essential
in filling out the greater of these intervals. Overjoyed at the way his project had come,
with gods help, to fulfillment, he ran into the smithy, and through a great variety of
experiments he discovered that what stood in direct relation to the difference in the
8 introduct ion

sound was the weight of the hammers, not the force of the str ikers or the shapes of the
hammer-heads or the alteration of the iron which was being beaten. He weighed
them accurately, and took away for his own use pieces of metal exactly equal in weight
to the hammers.7

Nicomachus continues the story by descr ibing how Pythagoras used the weights to
conduct further exper iments. After suspending the weights from identical str ings,
Pythagoras plucked pairs of strings and discovered the same concords as he had
heard produced by the blacksmiths. He further discovered that the interval of an
octave was produced by weights in a 2:1 ratio, that of the fifth by weights in a 3:2
ratio, and that of the fourth by weights in a 4:3 ratio, as shown in figure I.2a. The
one discordant interval that of a second sounded by the middle two weights
was the product of a 9:8 ratio. Additional exper imentation showed that the small-
est weight sounded a fourth, with the next to smallest weight (8:6 ! 4:3) and a fifth
with the next to largest (9:6 ! 3:2; see fig. I.2b). The octave could thus be viewed
as the product of either a fourth plus a fifth (12:9:6; fig. I.2c) or of a fifth plus a
fourth (12:8:6; fig. I.2d). According to Nicomachus, Pythagoras also discovered that
these ratios held constant throughout the musical domain. It made no difference
whether the constituent notes of the intervals were produced through string ten-
sion, string division, beating on pots, or blowing on tubes the relationships
between these notes always reduced to the self-same ratios.8
The Pythagorean view of music outlined by Nicomachus assumes that music has
its origins in the natural world and that the natural world has a basic (if often
unseen) order that can be expressed through number. It is thus important to Nico-
machuss story that the r inging of the hammers is accidental and not contr ived: their
harmony has everything to do with the inherent order of the world and almost
nothing to do with the blacksmiths. It is also significant that the concordant inter-
vals are immediately apparent to Pythagoras and that he can discer n them even
amid the discordant clang of the major second. Not only is the basis of musical
order natural, but also it is manifest to all who have ears to hear. The association of
these intervals with the pounding hammers provides the computational tool for

7. Nicomachus, Enchiridion, in Barker, Greek Musical Writings, 2: 256 57. The dioptra was a rod that
was used for the indirect measurement of the height of tall objects.
8. Nicomachus, Enchiridion, 258. It should be noted that Nicomachuss story is a complete fiction.
There is no evidence whatsoever that Pythagoras ever conducted any empirical research on the acoustic
origins of harmonic relationships, with or without blacksmiths. Perhaps more important, the ratios
described by Nicomachus simply do not work. To sound the intervals descr ibed in the story, the values
of the weights must be squared that is, the weights must be in the ratio 4:1 to produce the octave, 9:4
to produce the fifth, and 16:9 to produce the fourth. The ratios given by Nicomachus only work when
used to segment a str ing into different sounding lengths. One-half the length of a str ing will sound an
octave with the entire length of the string; two-thirds the length of the str ing will sound a fifth with the
entire length of the str ing; and three-fourths the length of a str ing will sound a fourth with the entire
length of the str ing. The importance of Nicomachuss story lies in its influence: in the for m Boethius
gave it in the sixth century, it became the standard account of basic Pythagorean principles for the Mid-
dle Ages and Renaissance; see Ancius Manlius Severinus Boethius, Fundamentals of Music, ed. Claude V.
Palisca, trans. Calvin Bower, Music Theory Translation Ser ies (New Haven, Conn.:Yale University Press,
1989), 17 19. A somewhat different perspective on Nichomachuss story can be found in Mathiesen,
Apollos Lyre, 399.
conc e p tual i zi ng mu s i c 9

%3* +,* -* .* /*


43* +,* -* .* /*

&3* +,* -* /*
012* 21,

)3* +,* .* /
21,* 012

figure I.2 Ratios of the octave, fifth, and fourth from Pythagorean legend

which Pythagoras had searched, for it allowed him to translate the constituent notes
of the har monic intervals into magnitudes that is, into number.9 These numbers
then provide further proof of the order of nature, for the ratios of the concordant
intervals (the octave, fifth, and fourth) are all simple (1:2, 2:3, and 3:4), while the
ratio of the discordant second is relatively complex (8:9). Finally, the numbers
involved in the ratios of the concords1, 2, 3, and 4 were also those of the tetrak-
tys of the decad, which Pythagoreans regarded as the fount and root of ever-
flowing nature.10
The account of musical organization presented by the story of Pythagoras and
the blacksmiths is a model of concision. At its core are but three intervals: the
octave, fourth, and fifth. The notes that make up these intervals are assigned mag-
nitudes, the cor respondence of which yields numerical ratios. These ratios come
to stand for the intervals the ratio 1:2 is the octave and can also be used to
describe relationships between them, leading to a precise account of the composi-

9. The notion of a computational tool I employ here derives from the work of Edwin Hutchins,
especially that presented in the second chapter of his Cognition in the Wild (Cambridge, Mass.: MIT Press,
1995). Hutchins proposes that various of the navigational tools used by seafarers are in fact computational
tools in that they facilitate computation by transfor ming analog information into digital infor mation.
This is exactly what the hammers did for Pythagoras: they transfor med the analog information of sound
into the digital infor mation of hammer weights (with an inter mediary stage occupied by the computa-
tional tool of a scale for measur ing the weights).
10. Barker, Greek Musical Writings, 2: 30. A tetraktys is any coordinated group of four items; those of
the tetraktys of the decad sum to 10, which is the basis of the base-10 number ser ies used by the Greeks.
10 introduct ion

tion of intervals. The interval between the Mese and Proslambanomenos of figure I.1
is an octave, a 1:2 ratio. This octave is made up of a fourth (from Mese to Hypate
meson, a 3:4 ratio) and a fifth (from Hypate meson to Proslambanomenos, a 2:3 ratio).
The fifth is in tur n made up of a fourth (from Hypate meson to Hypate hypaton,
another 3:4 ratio) and a tone (from Hypate hypaton to Proslambanomenos, an 8:9
ratio). With these components in place, a Pythagorean theor ist could descr ibe,
through number, relationships between any of the fixed notes of the Greater Perfect
System and eventually character ize the relationships that obtained among the
movable notes of the various genera. The order of the cosmos, which was for
Pythagoreans the order of number, thus found sounding expression in the domain
of music.

Aristoxenus and Aristotelianism

Although the Pythagorean perspective on musical order was an influential one
for instance, it infor med Platos and Aristotles writings on music it was not the
only one available to antiquity. The alter native offered by Aristoxenus in his Ele-
menta harmonica (most likely written toward the end of the fourth century b.c.) starts
with a definition of the science of harmonics:
It is to be understood as the science which deals with all melody, and inquires how the
voice naturally places intervals as it is tensed and relaxed. For we assert that the voice
has a natural way of moving, and does not place intervals haphazardly.We try to give
these matters demonstrations which confor m to the appearances, not in the manner
of our predecessors, some of whom used arguments quite extraneous to the subject,
dismissing perception as inaccurate and inventing theoretical explanations, and saying
that it is in ratios of numbers and relative speeds that the high and the low come
about. Their accounts are altogether extraneous, and totally in conflict with the ap-
pearances. Others delivered oracular utterances on individual topics, without giving
explanations of demonstrations, and without even properly enumerating the percep-
tual data. We, on the other hand, try to adopt initial pr inciples which are all evident
to anyone exper ienced in music, and to demonstrate what follows from them.11

This account of harmonics reveals Aristoxenus to be in conflict not only with the
Pythagoreans (the unnamed antagonists who dismiss perception and explain pitch
relations through ratios) but also with earlier har monic theor ists whose empir ical
work he found deficient because they did not explain their methods of proof or
properly descr ibe their observations.
Aristoxenuss alter native was to apply Aristotles intellectual method to music
more rigorously than did Aristotle himself .12 This entailed restricting the account
of music to ter ms and concepts that could properly be said to belong to the domain
of music. It excluded descr iptions that made recourse to ratios (which are in the
domain of number) or to theor ies about the propagation of physical sound (which

11. Aristoxenus, Elementa harmonica, in Harmonic and Acoustic Theory (vol. 2 of Greek Musical Writings),
ed. and trans. Andrew Barker, Cambridge Readings in the Literature of Music (Cambr idge: Cambridge
University Press, 1989), 149 50.
12. For commentar y on Aristoxenuss approach, see Barker, Greek Musical Writings, 2: 66 69, 119.
conc e p tual i zi ng mu s i c 11

are in the domain of physics). Once the definition of these basic musical concepts
was accomplished, an explanation of the entire domain of music could then follow.
Aristoxenuss demonstration proceeds in three steps. First, he identifies two
forms of vocal motion, the continuous and the intervallic. In the continuous for m,
which is associated with speech, the voice appears to traverse space without stop-
ping, until the point of silence. In the intervallic for m, which is associated with
singing, the voice appears to stand still at specific points, and then pass over some
interval of space before coming to rest at another point.13 The various pitches upon
which the voice pauses when singing constitute musical notes; the spaces between
these notes are musical intervals.
Aristoxenuss second step is to draw distinctions among the various musical
intervals. The first distinction is made with regard to magnitude, which reflects the
amount of space between the two notes of the interval. That such a space exists is
inferred from the difference between the two pitches that adjoin the interval; the
size of the space can be reckoned in ter ms of how many other notes could be put
inside it.14 The second distinction is made with regard to concord and discord. Aris-
toxenus identifies the concordant intervals as the fourth, fifth, and octave (and their
octave duplications). These are the only concordant intervals he accepts as de-
termined by the intr insic nature of melody all other intervals are by definition
The final step toward assembling the basic definitions and pr inciples of har-
monics is the der ivation and division of the tone. Aristoxenus defines the tone as the
difference between the first two concords (the fourth and the fifth) and explains
that it can be divided in half (yielding the semitone), in thirds (yielding the least
chromatic diesis), or in fourths (yielding the least enhar monic diesis, which is the
smallest interval recognized as melodic).16 These distinctions allow him to locate
various notes within the tetrachords of the different genera and thereby to specify
the scalar structure of each.
With these steps, Aristoxenus lays out the basic mater ials for his account of musi-
cal organization. Singing involves a specific way of using the voice that creates
musical notes and musical intervals. Intervals can be distinguished according to size
and whether they are concordant or discordant. Concordant intervals the fourth,
fifth, and octave are accepted as axiomatic to melody and thus representative of
the basic mater ials from which the various intervals of the Greater Perfect System
can be developed.

13. Aristoxenus, Elementa harmonica, 133.

14. Aristoxenus, Elementa harmonica, 136.
15. Aristoxenus, Elementa harmonica, 139. Aristoxenus further distinguishes between composite and
incomposite intervals (p. 137); however, this distinction is not necessary for a basic understanding of his
16. Aristoxenus, Elementa harmonica, 140. The different dieses apply to the three genera mentioned
above: the half-tone is used in the diatonic genus; the third-tone is used in the chromatic genus; and the
fourth-tone is used in the enhar monic genus. For further discussion of Aristoxenuss Elementa harmon-
ica, see Mathiesen, Apollos Lyre, 319 34.
12 introduct ion

Greek Theory and Cognitive Structure

As mentioned, the music theor ies of Pythagoras and Aristoxenus belong to a world
remote from our own. Not only did these theor ists have to grapple with the most
basic of principles, but also the music they would descr ibe is a microtonal one that
is primarily concer ned with the successive notes of melody rather than the simul-
taneous notes of harmony. Despite this or perhaps because of it Pythagorean
and Aristoxenian accounts of musical organization give us a glimpse into how the-
ories are formed and, more important, the cognitive processes that are basic to these
theories. In particular, three cognitive processes can be seen at work: categorization,
cross-domain mapping, and the use of conceptual models.

categorization Our ability to categor ize things is a cognitive process so

basic and so pervasive that it can easily escape our notice.Were you to lift your eyes
from this book and survey your sur roundings, you might well see chairs, lamps,
tables, and other books; were you outside, you might see trees, birds, clouds, cars, and
bicycles. If you considered the other things that populate your day, you might think
of friends and family members, facial expressions and gestures, actions and activities.
Your recognition of these things reflects the categor ies through which we structure
our thought: to recognize a book is to identify it as a member of the category book;
to recognize a tree is to identify it as a member of the category tree. Categorization
occurs in all sensor y modalities and throughout the range of mental activities: we
categor ize smells and sounds, thoughts and emotions, skin sensations and physical
Categor ies are not just basic to thought; they also give insight into our thought
processes. At one time it was thought that categor ies reflected the structure of the
real world, but recent research has shown that the categor ies humans use are shaped
by their interactions with their environments. Our reasons for developing and
employing a given category are part and parcel of the category itself: categor ies are
not only not given by nature, but also they are subject to change and modification
as our thought unfolds.
Two categor ies basic to Pythagorean and Aristoxenian music theory are those for
consonant or dissonant intervals. Consonant intervals (such as the octave, fifth, and
fourth) are fundamental to the conceptualization of Greek music: they mark the sta-
ble pitches of the Greater Perfect System and are the source of derivation for all fur-
ther intervals, both consonant and dissonant. The process of categor ization is also
exhaustive: any interval that can be conceived belongs to one of these two cate-
This is not to say, however, that consonant and dissonant intervals are given by
nature in any simple way, Pythagoras and the blacksmiths notwithstanding.17 As one

17. I should note here that psychoacousticians distinguish between musical consonance, which is a
cultural construct framed relative to a particular set of musical practices, and sensory consonance, which
is a consequence of how sound waves are processed by the hear ing mechanism (which involves the
cochlea and the auditory cortex). Sensory consonance is thus a fairly straightforward product of nature.
Although musical consonance has its basis in sensory consonance, there is some freedom in how the sen-
sory data are inter preted.
conc e p tual i zi ng mu s i c 13

example, consider the way Aristoxenians and Pythagoreans classified the interval of
an octave plus a fourth. Aristoxenians considered the interval a consonance, since it
was simply the combination of two smaller consonances. Pythagoreans, in contrast,
classified intervals according to the numer ical ratio for med by their constituent
pitches. As explained by the anonymous (and thoroughly Pythagorean) author of
the Sectio canonis (fourth century b.c.), consonant intervals are those whose ratios are
either multiple (of the for m [mn]:n) or epimor ic (of the for m [n + 1]:n). Dissonant
intervals are those whose ratios are epimer ic (of the for m [n + m]:n, where m is
greater than 1 and neither equal to nor a multiple of n).18 Because the octave plus
a fourth had the epimer ic ratio 8:3, it was regarded as a dissonance.
Another example of how categor ies shape our understanding of phenomena is
provided by Greek theor ists treatment of thirds and sixths. Although thirds and
sixths sound fairly consonant, they were nonetheless categor ized as discords. Two
factors bear on this classification. First, forming thirds and sixths requires using the
movable pitches of the Greater Perfect System at best, a third or a sixth will
involve only one of the stable pitches bounding the constituent tetrachords of the
system. Thirds and sixths were intervals that necessar ily varied in size, and so they
were placed among the dissonances. Second, in the classification of intervals Greek
theory followed a tradition of dichotomous categor ies: there was concord, discord,
and nothing else. By contrast, neither of these factors played a part in the music the-
ory of early India. Indian music theor ists were consequently free to focus on the
qualitative aspect of intervals rather than on their cor respondence with the fixed
notes of a tuning system and to construe intervallic relationships as concordant, dis-
cordant, or neutral.19
These two examples show that while the categor ies for consonant and dissonant
intervals may be basic to Pythagorean and Aristoxenian theor y, just how they are
defined reflects the context and goals of categorization: consonance and dissonance
are not naturally occur ring properties, but ways of constructing an understanding of
musical organization.
Of course, there are numerous other categor ies important for Pythagorean and
Aristoxenian music theor y, including those for pitches, intervals, and numer ical
ratios. These categor ies and others are basic to the sort of systematic account of
musical phenomena provided by these theor ies indeed, it is simply not possible to
have a theory of music, or of anything else, without first having categor ies.

cross-domain mapping Cross-domain mapping is a process through

which we structure our understanding of one domain (which is typically unfamil-
iar or abstract) in ter ms of another (which is most often familiar and concrete). For
example, one way to think about the elusive concepts of electr ical conductance is
in terms of a hydraulic model: flipping the light switch tur ns on the juice, and elec-

18. Sectio canonis, in Harmonic and Acoustic Theory (vol. 2 of Greek Musical Writings), ed. Andrew
Barker, Cambridge Readings in the Literature of Music (Cambr idge: Cambridge University Press, 1989),
19. Lewis Rowell, Music and Musical Thought in Early India, Chicago Studies in Ethnomusicology
(Chicago: University of Chicago Press, 1992), 157 60.
14 introduct ion

trical cur rent flows to the light bulb to light the room. By this means we take what
we know about a fairly concrete and familiar source domain the flow of water
and other liquids and map it onto a rather abstract and unfamiliar target domain:
that of electr icity. As a wealth of research on analogy and metaphor has shown, the
process of mapping structure from one domain to another is basic to human under-
One place cross-domain mapping is evident is in the Pythagorean and Aristox-
enian construal of interval. Because musical pitches are ephemeral and virtually
intangible, relationships between pitches musical intervals represent something
of a challenge to understanding. One way to meet this challenge is to map structure
from the physical world onto music, a process evident in Nicomachuss story of
Pythagoras and the blacksmiths. Pythagoras hears harmonious sounds, traces their
origins to the blacksmiths hammers, and then proceeds to conduct various exper-
iments using weights equivalent to those of the hammers. These exper iments lead,
among other things, to a highly pragmatic objectification of musical pitch, as pitches
are translated into physical objects that can be weighed, studied, and preserved. By
performing a mapping from the concrete physical domain proper to the black-
smiths hammers onto the domain of musical sound, Nicomachuss story allows us
to structure the latter domain in ter ms of the for mer. Of course, musical notes are
not physical objects that can be weighed, studied, and preserved they remain
ephemeral and virtually intangible. Nonetheless, we are so accustomed to the map-
ping between concrete physical objects and musical sound that we sometimes have
to be reminded that notes are not endur ing physical objects.
Aristoxenuss construal of musical interval involves a slightly different mapping.
As we have seen, according to Aristoxenus, when the voice moves intervallically, it
appears to stand still at a given place (a musical pitch) and then pass over an inter-
val of space (a musical interval) before coming to rest at another place (another
musical pitch). Underlying this account is a mapping from the familiar domain of
two-dimensional space onto that of music. This mapping allows us to apply the
methodology of measur ing space to music. The difference between two linear mea-
sures yields a third measure; similarly, the difference between the intervals of a fifth
and a fourth yields the interval of a tone. Since linear measures can be easily divided
into equal halves or thirds or fourths, the musical tone can be similarly divided,
something impossible from the Pythagorean perspective.
On closer inspection, the Pythagorean and Aristoxenian construals of interval are
indeed incommensurate. From the Pythagorean perspective, pitches are physical
objects, and an interval descr ibes the relationship between these objects. From the
Aristoxenian perspective, pitches are breadthless points that simply mark out an
expanse of two-dimensional space, and an interval is the expanse itself. Each map-
ping gives an account of interval, but each leads to a different conceptualization of
musical structure. This point can be generalized for music theory as a whole: map-
ping structure from a nonmusical domain onto music is a way of creating musical
structure, and different mappings will lead to different accounts of musical structure.

conce ptual mode ls Both categor ization and cross-domain mapping

provide the basis for fundamental ontological assertions about musical mater ials: this
conc e p tual i zi ng mu s i c 15

interval is a consonance; the pitches of an octave are physical objects. They can also
lead to conditional statements: if the interval is an octave, then it is a consonance; if
a pitch is an object, then its properties are measurable. Propositions like this are basic
to conceptual models, which act as guides to reasoning and inference. In their sim-
plest for m, conceptual models consist of concepts in specified relationships, which
pertain to a specific domain of knowledge.
For an example of a conceptual model, let us return to the classification of con-
sonant and dissonant intervals presented in the Sectio canonis, according to which all
consonant intervals have either multiple or epimor ic ratios, and all dissonant inter-
vals have epimer ic ratios. This classificator y system relies on a conceptual model
that organizes concepts related to interval, concord, discord, and the three classes of
ratios. The simple patter n of inference that follows from this model is that if an
interval has a multiple or an epimor ic ratio, it is a concord; if it has an epimer ic
ratio, it is a discord.
Integral to this model are the products of categor ization and cross-domain map-
ping. Two types of categories are involved in the model: those pertaining to music
(the categor ies of concord and discord) and those pertaining to number (the mul-
tiple, epimoric, and epimer ic ratios). Cross-domain mapping cor relates the two types
of categor ies by construing musical interval as a relationship between two objects
(namely, musical pitches) to which magnitudes (in the for m of numbers) can be
assigned. Specific classes of ratios can then be used to distinguish between the musi-
cal categor ies.
The robustness of this particular conceptual model is reflected in the debate over
the status of the octave plus a fourth that continued into the Middle Ages. In the
second centur y a.d., Ptolemy showed the speciousness of the cor relation of con-
cord with multiple or epimor ic ratios and argued for a classification of intervals
based on empir ical evaluation and the postulate that a concord added to a concord
produces a concord.20 Although Ptolemy still used ratios to descr ibe various inter-
vals, they were no longer part of the conceptual model through which intervals
were classified into concords and discords. In the sixth century, Boethius presented
both the Pythagorean and Ptolemaic models but took no position on which he pre-
ferred.21 After Boethius, when an author wished to invoke the author ity of the
Pythagorean approach, the Pythagorean model of intervallic classification was cited;
when an author wished for a more empirically satisfying classification, the Ptole-
maic model was used.22
Conceptual models provide the first level of organization for concepts. They are
too limited and localized, however, to provide the comprehensiveness we expect
from theor ies of music. Theories achieve this comprehensiveness by integrating

20. Ptolemy, Harmonics, in Harmonic and Acoustic Theory (vol. 2 of Greek Musical Writings), ed. Andrew
Barker, Cambridge Readings in the Literature of Music (Cambr idge: Cambridge University Press, 1989),
286 90.
21. Boethius, Fundamentals of Music, 81 82, 169.
22. For a discussion of these modes of reasoning, see C. Andr Barbera, The Consonant Eleventh
and the Expansion of the Musical Tetractys: A Study of Ancient Pythagoreanism, Journal of Music The-
ory 28 (1984): 191 223. Barberas treatment of Aurelians Musica disciplina on p. 210 is especially illumi-
16 introduct ion

clusters of conceptual models. And as we shall see in the following chapters, con-
ceptual models also play a role in categor ization and cross-domain mapping. There,
as in theor ies, they provide guides for reasoning and inference about specific and
circumscr ibed domains of knowledge.

cognitive processes and music theory Much has been left out
of this discussion of theories of music from Greek antiquity, with respect to both the
theories themselves and the cognitive processes behind them. To be sure, these the-
ories are of a different order than Swanns musings on Vinteuils sonata. Nonethe-
less, the cognitive processes we have seen at work in Pythagorean and Aristoxenian
theory are the same processes through which we organize our understanding of the
world as a whole. Just how this occurs how categor ization, cross-domain map-
ping, and the use of conceptual models shape our theor ies of music and guide our
analyses of musical works is the subject of the remainder of this book.

I have divided the chapters that follow into two parts. In the first, I present a
detailed overview of research on the three cognitive processes highlighted in this
introduction. This overview is itself framed around specific musical topics, such as
motivic transfor mation, text painting, and the ways in which we structure our
understanding of a specific musical domain. The research that has been done in
cognitive science over the past three decades has been extensive and far ranging, and
one of the jobs of this portion of the book is to br ing this work to bear on basic
issues of musical understanding. Another objective is to show in some detail how
these processes relate to one another and how they form the bedrock for our
thought about music. The second part of the book moves from this foundation to
analytical studies of specific musical issues. These issues include relationships
between categor ization and musical syntax, the problem of musical ontology, text-
music relations, and conceptions of musical for m and musical hierarchy.
Chapter 1 begins the overview of research in cognitive science with a close look
at processes of categorization. For centur ies, writers in the West regarded categor ies
as fixed and immutable, and any variation in categor ization was taken as evidence of
the failure of the human intellect to deal with the structure of the real world. It took
the pioneer ing work of Eleanor Rosch and others in the 1970s to show that cate-
gory structure was not as simple as first believed. In particular, some levels of cate-
gorization are preferred over others, and some members of a category are regarded
as better representing the categor y than others (a phenomenon known as graded
The key to how this research can be applied to music is provided by the musi-
cal motive (or, as Proust would have it, the motif ). Motives are generally reckoned to
be one of the basic building blocks of musical works, but they are also a bit slippery:
the same motive typically assumes a number of diverse shapes over the course of
a work. Thinking of a motive as a cognitive categor y makes it possible to account
for its identity, as well as its diversity, and reveals how aspects of categorization are
embodied by musical mater ials. These preliminary applications of categorization to
conc e p tual i zi ng mu s i c 17

music also show ways musical mater ials can be organized over the course of a work
and offer an explanation of how it is possible to have musical concepts that are inde-
pendent of language.
If categor ization can be said to be the source of musical concepts, cross-domain
mapping is the means by which these concepts are placed in cor relation with oth-
ers. Chapter 2 examines the process of cross-domain mapping in some detail, begin-
ning with the work of cognitive linguists who, in the 1980s, proposed that metaphor
was a basic structure of understanding. This proposal gained added weight when it
was shown that metaphor ical projection (which is one way to accomplish cross-
domain mappings) was a general process not restricted to linguistic expressions but
grounded in embodied exper ience.
One example of cross-domain mapping that involves music in a rather immedi-
ate way is the technique of text painting, a compositional device that aims to rep-
resent in music specific images summoned by the text of a vocal work. Text paint-
ing provides a point of departure for the exploration of how cross-domain mapping
is manifested in our understanding of music, as it leads to an extension of cross-
domain mapping called conceptual blending. In a conceptual blend, elements from
two cor related domains are projected into a third, giving rise to a r ich set of possi-
bilities for the imag ination. As I show in the latter part of chapter 2, text painting
can lead to such blends, as can program music.
Chapter 3, which focuses on conceptual models and theor ies, gets to the heart of
the perspective on cognition developed in part I. My point of departure is research
by Jeanne Bamberger on childrens representations of musical structure. In my
analysis of Bambergers study of one specific eight-year-old boy, I show the part
played by categor ization and cross-domain mapping in the conceptual models used
by this boy to come to ter ms with a musical environment. I also show how these
models are combined to for m a theor y of music and how this theor y changes in
response to changes in the task at hand. This close-up glimpse of the structure and
role of conceptual models and theor ies leads, in the middle of the chapter, to a more
generalized character ization of these knowledge structures, which I connect with
work on similar structures in artificial intelligence, cognitive anthropology, ethno-
musicology, and developmental psychology. In the latter part of the chapter, I return
to music theory and explore the role of conceptual models and theor ies (that is,
theories framed relative to a cognitive perspective) in analyses by Jean-Philippe
Rameau and Heinr ich Schenker, two of the best-known music theor ists of the last
three hundred years.
Although the features of cognitive structure discussed in part I might seem to
be relatively detailed, in truth all are associated with relatively high-level cognitive
processes. My reason for focusing on this level is quite simple: it allows me to engage
in issues of immediate and occasionally central importance to music scholarship and
to do so in a way that connects with extensive research in cognitive psychology and
cognitive linguistics. Part II explores this possibility in g reater depth by consider-
ing various problems of musical understanding from the perspective on cognitive
structure and music theory developed in part I.
In chapter 4, I turn to the matter of how musical mater ials are organized within
a work more properly, the problem of musical syntax and, by extension, musical
18 introduct ion

semiotics. Although semioticians are usually quick to grant that music has a syntax,
they are more doubtful about whether its semantic level has any depth. By taking
a close look at how composers make use of categor ies of musical events in this
case, the way Mozart and Beethoven use motives in the opening movements of
three str ing quartets I am able to provide insight into how musical mater ials are
organized in the service of musical discourse, as well as how features of this orga-
nization contr ibute to meaning construction as a whole.
In chapter 5, I confront a somewhat larger problem one that may seem ir re-
deemably abstract: the problem of what counts as a work of music. I view this prob-
lem, usually called the problem of musical ontology, as one of cultural knowledge
and try to show that, as opposed to being hopelessly recondite, the problem is of
immediate importance for understanding music. By approaching the entire work of
music as a categor y a category that includes all the scores, performances, repre-
sentations, and such that are said to be of the piece I develop a model for the
cultural knowledge upon which judgments about musical ontology are made.
Determinations of what counts as an instance of a particular musical work are thus
one of the ways members of a musical community construct and negotiate their
identity. My examples for this chapter are two songs taken from the traditions of
popular music and jazz: I Got Rhythm and Bye Bye Blackbird. The latter offers
an intr iguing case of how the cultural knowledge relative to which deter minations
of musical ontology are made can become complicated when implicated in the lay-
ered discourse structures Mikhail Bakhtin called double-voiced discourse, and
which were extended to African American culture through Henry Louis Gates Jr.s
notion of Signifyin(g).
Chapter 6 returns to the analysis of individual musical works by pursuing one
of the entailments of text painting noted in chapter 2: under certain circumstances,
combinations of words and music, through the process of conceptual blending, cre-
ate worlds for the imagination well beyond those that spr ing from words or music
alone.Where only a few fairly circumscr ibed instances of text-music relations were
considered in chapter 2, here research on conceptual blending is applied to the
whole of five Lieder from the nineteenth centur y. These analyses offer a way to
flesh out the theory of conceptual blending as it applies to music and provide a fur-
ther perspective on musical syntax. In these songs, we see words and music com-
bining to create r ich domains in which the imagination can play, as well as discover
how musical syntax shapes our understandings of the words themselves.
The final analytical chapter (chap. 7) turns to music theory itself, specifically to
the theor ies of musical for m and hierarchy that go back to the eighteenth and early
nineteenth centur ies. Accounts of the for m a musical work takes, or of how its ele-
ments relate to one another, are basic to theor izing about music indeed, we can
see these emerging in the course of M. Swanns ruminations on Vinteuils sonata
but at times it seems that theorists are talking about quite different things. For instance,
there are two common ways to talk about musical for m: the first approaches for m
as der iving from the assembly of relatively static building blocks that are combined
to create the finished work; the second approaches for m as an emergent property of
the work, which becomes manifest only as the music unfolds over time. The first
approach yields a view of musical for m that is quite static, the second a view rather
conc e p tual i zi ng mu s i c 19

more dynamic. Using the analytical framework provided by cross-domain mapping,

I discuss the source of these two models for musical for m, as well as two models for
musical hierarchy, and explore some of the ways these models interacted over the
course of the history of music theory.
In the conclusion I return to M. Swann and to his final encounter with Vinteuils
sonata after a year in which it became thoroughly intertwined with his love affair
with Odette, the courtesan with whom he had become acquainted around the same
time he first encountered the andante. This will provide a frame for a review of the
points made in the preceding chapters and an instrumentality for drawing conclu-
sions from the whole.

cognitive structure,
theory, and analysis
A central claim of this book is that through developing an appreciation of how
aspects of cognitive structure shape our understanding of music we can better appre-
ciate the active role of theories of music in that understanding. A further claim is
that our analyses of musical phenomena from the most mundane and localized of
accounts to the most abstract and comprehensive similarly reflect cognitive struc-
ture, in that every analysis is based on some sort of theory of music. Musical analy-
ses are in truth dialogues, and not just dialogues between the analyst and an imag-
ined audience: musical analyses are also dialogues between the analyst and some
body of theoretical knowledge. Analysis rarely, if ever, simply cor roborates a theory:
analysis pulls theor y and pushes it, extending and chang ing theory just as it also
extends and changes our understanding of musical phenomena.
The analyses I present throughout this book are no different, except that they
engage cognitive theory as well as music theor y. The intent of the analyses is to
show how our understanding of particular musical phenomena can be character ized
in terms of specific cognitive processes and structures and thereby connect that
understanding to research in cognitive science as a whole. The analyses are not intended
as definitive statements about how we can account for such understanding; they are
intended to be the initiation of a dialogue with cognitive theory, a dialogue whose
purpose is to expand our knowledge of both music and cognition. Analysis is thus
a central concer n of what follows, but, unlike cognitive structure or theor y, I have
not treated it as a central topic for investigation. Instead, analysis will be a funda-
mental tool to explore both and to provide new insight into the conceptual worlds
wrought by musical sound.
This page intentionally left blank
part i
aspects of cognitive
This page intentionally left blank
chapter one

I n the opening eleven measures of the prelude to Richard Wagners Tristan und
Isolde there are three successive statements of a melodic motive that foreshadows,
and then later accompanies, the tragic arc of the opera. Three times this motive
emerges from silence, swells through the cellos and English hor n, and then sinks
back into the stillness from which it came. After the third statement, the motive dis-
appears, then reappears a few long minutes later near the end of the prelude. Here
it is stated first in the cellos and violas (and completed by the oboes and bassoons),
then in the English hor n, and finally in the bass clar inet. After the last of these, the
prelude dissolves into the first scene of the opera. All of these statements are shown
in example 1.1.
Although our first inclination is to treat these various statements as somehow the
same that is, all can be regarded as statements of the opening motive from Wag-
ners Tristan und Isolde there are actually notable differences among them. There
is, of course, the change in the instrumentation used for the statements at the end of
the prelude, which contrasts with the distinctive timbral stamp put on the motive in
the opening measures. There is a similar diversity in the pitches through which suc-
cessive statements are given voice, since versions of the motive begin on either A3
(or Ab3), B3 (or B2), or D4. Although there are significant similar ities in the rhyth-
mic figuration used for the various statements, there is no patter n of duration and
accentuation common to all. And there are also somewhat smaller variations: the
first statement begins with a minor sixth instead of the major sixth used by the oth-
ers, and the fourth is the only one shaped by both a diminuendo and a rallentando.
Despite their differences, these seven melodic fragments sound similar enough to
one another to be regarded as functionally equivalent. Such equivalence was impor-
tant to Wagners compositional style, for it provided the mater ial foundation for
works uninter rupted by conventional for ms. By recalling specific motives at cru-
cial points within his musical and dramatic discourse Wagner was able to create a
sense of shape and unity that seemed inevitable without being predictable. By the
time of Tristan, however, such motives were no longer simple building blocks to
be assembled and per muted but had become infinitely malleable constructs. Carl
Dahlhaus, recognizing both the structural importance and mutability of these

24 aspects of cogn it ive st ruc ture

example 1.1 Statements of the opening motive from Richard Wagners Tristan und
Isolde, Prelude
Vnc. 1 Eng. Hn.

Vnc. 5 Eng. Hn.

cresc. dim.

8 Vnc. Eng. Hn.


85 Vnc., Vla. Ob., Bn.



88 Vnc., Vla. Ob.


Eng. Hn.

Bs. Cl. 104


motives within the opera, likened them to woven threads that surface only to divide
and then disappear from sight.1
To understand this music to make sense of the sonic texture Wagner weaves
requires being able to assimilate these various musical phrases into a single cognitive
construct and then recall that construct, often after an hour or more of Wagnerian
effusion. Understanding Wagner or most music, for that matter requires being
able to think in ter ms of categor ies of musical events. In this chapter, I explore what
recent research into categor ization can tell us about this process. First, however, I
develop a slightly clearer picture of two basic aspects of understanding music, which
will provide the g roundwork for applying research on categor ization to situations
of the sort presented by the opening of Tristan. I do this by reviewing a theory of
musical motive developed by one of the many students of Wagner. Not that Wag-

1. Carl Dahlhaus, Richard Wagners Music Dramas, trans. Mary Whitall (Cambr idge: Cambridge Uni-
versity Press, 1979), 63.
cate gorization 25

ner ever actually taught composition. But the influence of his music was enor mous,
and many lessons were taken away from his operas. Among these were lessons on
compositional technique absorbed, refined, and retransmitted in both treatise and
score by Arnold Schoenberg.

musical coherence and motive

Schoenberg thought of himself first and foremost as a Brahmsian, but, under the
influence of Alexander von Zemlinsky, he also found a passion for Wagner. Schoen-
berg noted that his early works showed the influence of both masters: In my Ver-
klrte Nacht the thematic construction is based on Wagner ian model and sequence
above a roving harmony on the one hand, and on Brahmss technique of develop-
ing variation as I call it on the other.2 Wagner was also known for his har-
monic innovation, although Schoenberg took pains in one of his late essays to show
that Brahmss harmonies were every bit as progressive as Wagners.3 Nonetheless,
it was Wagners harmonic language that was perceived as posing the most for mida-
ble challenge to composers of the late nineteenth century, for the free treatment of
dissonance integral to this language loosened prevailing conceptions of tonal orga-
nization and called into question the laws of tonality that had for merly held sway.
From Schoenbergs perspective,Wagners music presented few dangers for tonal-
ity. As Schoenberg had argued in his Harmonielehre of 1911, tonality had its or igins
in acoustic phenomena.4 But shaping music in confor mance with acoustic phe-
nomena was not all there was to musical composition: composition also required
insight into how the human mind worked. In an essay that attempted to sort out
some of the problems raised by the emancipation of the dissonance Schoenberg
Tonalitys origin is found and rightly so in the laws of sound. But there are other
laws that music obeys, apart from these and the laws that result from the combination
of time and sound: namely, those governing the working of our minds. This latter
forces us to find a particular kind of layout for those elements that make for cohe-
sion and to make them come to the fore, often enough and with enough plastic-
ity so that in the small amount of time granted us by the flow of the events, we can
recognize the [musical] figures, grasp the way they hang together, and comprehend
their meaning.5

In Schoenbergs view, the laws governing the workings of our minds require the
composer to wr ite in such a way that listeners can quickly recognize musical figures
and the way they cohere. The listener, upon grasping this coherence, will then be
able to comprehend the work.
Key to musical coherence and comprehension were musical motives. According

2. Arnold Schoenberg, My Evolution (1949), in Style and Idea: Selected Writings of Arnold Schoen-
berg, ed. Leonard Stein, trans. Leo Black (New York: St. Martins Press, 1975), 80.
3. Arnold Schoenberg, Brahms the Progressive (1947), in Style and Idea, 398 441.
4. Arnold Schoenberg, Harmonielehre (Leipzig:Verlagseigentum der Universal-Edition, 1911); trans.
as Theory of Harmony by Roy E. Carter (Berkeley: University of California Press, 1978), 23 25.
5. Arnold Schoenberg, Opinion or Insight? (1926), in Style and Idea, 259.
26 aspects of cogn it ive st ruc ture

to Schoenberg, comprehension starts with recognition, and recognition starts with

basic musical figures that is, motives: Motive is at any one time the smallest part
of a piece or section of a piece that, despite change and variation, is recognizable as
present throughout.6 More specifically, motives consist of intervals and rhythmic
patterns combined to produce a shape or contour that, once recognized, can be eas-
ily remembered.7 A particularly str iking illustration of Schoenbergs concept of
motive is provided by his analysis of a subordinate theme from the first movement
of Brahmss Sextet in Bb major, Op. 18 (1860), shown in example 1.2.8
As Patricia Car penter and Severine Neff have noted, rhythm is central to
Schoenbergs concept of motive.9 This centrality is clearly evident in his analysis of
the Brahms sextet. Each family of motives is distinguished by a specific rhythmic
patter n: all of the a motives use a qde pattern; and all of the b motives use a
qdeq pattern. (Both patter ns, of course, contribute to the marvelous waltz-like feel
of the theme.) Contour also serves to distinguish the two families, although here
some variation among the different motive forms can be noted. Motives a1, a2, and
a4 all preserve the basic contour of motive a: the second note of the motive is lower
than the first note, and the last note of the motive is higher than the first note.
Motive a3 preserves only a portion of the contour of a: although the second note of
the motive is lower than the first, so is the last note. In a similar fashion, motives b1
and b3 preserve the basic contour of motive b (the second note is lower than the
first note, and the third note is lower than the second), while b2 preserves only a
portion of the contour of b (the second note is lower than the first note, but the
third note is higher than the second).
As this example shows, Schoenbergs concept of motive is broad and dynamic:
a family of motives can be distinguished simply by rhythmic figure or contour;
motive forms can be various and need have only some of their features in common.
This conception is rather different from, for example, Heinrich Schenkers or Rudolph

example 1.2 Schoenbergs analysis of the Brahms Sextet, Op. 18, first movement,
mm. 84 93
a b a1 b1

89 a2 b2 b3 a3 a4

6. Arnold Schoenberg, The Musical Idea and the Logic, Technique, and Art of Its Presentation, ed. and
trans. Patricia Car penter and Severine Neff (New York: Columbia University Press, 1995), 169.
7. Arnold Schoenberg, Fundamentals of Musical Composition, ed. Gerald Strang and Leonard Stein
(London: Faber and Faber, 1967), 8.
8. Schoenberg, Brahms the Progressive, 417.
9. See the commentary in their edition of Schoenbergs The Musical Idea and the Logic, Technique, and
Art of Its Presentation, 27 29.
cate gorization 27

Rtis concept of motive. Both Schenker and Rti regard specific intervallic rela-
tionships as constituent of motive; rhythm and contour are regarded as secondar y
aspects of motivic organization.10 The difference between Schenkers or Rtis con-
ception of motive and that of Schoenberg lies with Schoenbergs notion of coher-
ence and the role it plays in making music comprehensible. Coherence comes about
when the various parts that make up a musical entity are connected in such a way
that those parts similar to other entities become prominent. The work is most com-
prehensible to the listener when the ar rangement of these parts is such that their rela-
tionship to each other and to the whole is manifest.11 Thus there is no need to
restrict the distinguishing features of a family of motives to specific intervallic rela-
tionships, nor is there a need to require that all motive forms share exactly the same
features. All that is required of motive is recognizability and a potential for connec-
tion to other motive forms. This potential for connection contr ibutes to coherence,
which is, in turn, the basis for comprehensibility.
According to Schoenberg, then, the process of comprehension starts with rec-
ognizable bits (motives) that are easily remembered. Motives hang together not sim-
ply because their constituent parts are connected to one another but because these
connections emphasize similar ities to other motives. Coherence thus reflects prop-
erties shared by collections of motives; it is not, properly speaking, a property of any
one individual motive. Motive forms are of necessity variable, for differences between
forms reveal most clearly what is typical of the collection of motives as a whole.
And, although attention to coherence is important to the composer who wishes to
craft a convincing work, the apprehension of coherence is essential to the listener
who would make sense of that work.
Many composers, working in the long shadow cast by Ludwig van Beethoven
across the nineteenth century, relied on musical motives to make their works coher-
ent. As Edward Cone has noted, the use of thematic transfor mation as a unifying
device became increasingly important for composers of the later nineteenth cen-
tury, composers whose lengthy movements often embraced stylistic extremes of tempo,
meter, texture, and mood.12 However, the prevalence of the device also poses prob-
lems for the listener, for it is often difficult to sort out relationships among various
motivic entities. Consider the relationship of derivation, in which later motive
forms are derived from earlier motive forms; thus motive b3 in example 1.2 is said
to be der ived from motive b. As straightforward as this might appear, tracing der iva-
tional relationships is rarely a simple task: motive b can itself be understood to be
derived from one of the motives prevalent in the preliminary subordinate theme
heard starting in m. 61 of Brahmss sextet (shown in ex. 1.3), which has the same
rhythmic patter n but the reverse contour. Could b3 then be regarded as der ived
directly from this earlier motive form, or is it still to be regarded as der ived from b?
Schuberts Wanderer Fantasie, Op. 15 of 1822 presents an even more complicated

10. Heinrich Schenker, Der freie Satz, Neue musikalische Theor ien und Phantasien, 3 (Vienna: Uni-
versal-Edition A.G., 1935); Rudolph Rti, The Thematic Process in Music (New York: Macmillan, 1951). I
discuss histor ical construals of motive below in the first part of chap. 7.
11. Arnold Schoenberg, Coherence, Counterpoint, Instrumentation, Instruction in Form, ed. Severine
Neff, trans. Severine Neff and Charlotte M. Cross (Lincoln: University of Nebraska Press, 1994), 20 23.
12. Edward T. Cone, On Der ivation: Syntax and Rhetor ic, Music Analysis 6 (1987): 238.
28 aspects of cogn it ive st ruc ture

example 1.3 Johannes Brahms, Sextet, Op. 18, first movement, mm. 61 64

61 in tempo





situation. The work is based on an ar ietta from Schuberts 1816 cantata Der Wan-
derer, Op. 4/1, D. 489, which is quoted in the works second movement Adagio;
the relevant passage from the cantata appears in example 1.4a. If we have the ar i-
etta in mind, we can hear the motivic mater ial of the first movement as der ived
from the ar ietta and then presented instrumentally in the subsequent Adagio (the
motive from the first movement is labeled c in ex. 1.4b; its use in the Adagio is
demonstrated in ex. 1.4c).13 The temporal order of the process of derivation is thus
reversed: c is heard as der ived from c1. If we are coming to the work fresh, we will
most likely hear the Adagio as der ived from the first movement: c1 is thus der ived
from c. Finally, if we know the ar ietta but have temporar ily forgotten it, its reprise
in the Adagio, anticipated by the derived motives of the first movement, can be a
moment of great rhetor ical impact: the pianistic musings of the first movement have
summoned forth the ar ietta that now stands before us.
It would seem that, as Cone argued, hearing relationships among motive forms
is an ineluctably subjective affair. Following his argument still further, how we hear
these relationships is important to how we respond to a work, for, as Schuberts fan-
tasy suggests, it infor ms our understanding of musical rhetor ic. The most dramatic
case occurs in a work in which there are multiple der ivations of motivic mater ial.
In the course of such a work we can lose our sense of how motive forms are con-
nected, until the composer often unexpectedly reveals how the mater ials relate
to one another by bringing them into rapprochement. The rhetor ical aim of this
sort of compositional strategy is what Cone calls epiphany, and, when successful,

13. While the annotations in ex. 1.4 focus on the distinctive rhythmic figure common to each of the
excerpts, it is worth noting that each passage also features a move to the dominant.
cate gorization 29

example 1.4 Franz Schubert, (a) Der Wanderer, Op. 4 No. 1, D. 489, mm. 23 26; (b)
Wandererfantasie, D. 760, mm. 1 3; (c) Wandererfantasie, mm. 189 90
23 3 3

Die Son ne dnkt mich hier so kalt, die Bl te welk, das Le ben alt,

1 Allegro con fuoco ma non troppo

189 Adagio


it can compel the listener to realize a previously unsuspected or at most uncon-

firmed relationship among diverse motives.14 Put another way, the strategy that
leads to epiphany compels the listener to realize aspects of musical coherence that
become evident only over the course of time.
Motives and motivic relationships are important for both the coherence and the
rhetor ic of music, but understanding the part they play in musical understanding
poses some unique challenges. If the recognition of motives is central to the process
of comprehending music, it must occur very rapidly. This rapidity has led some to
argue for cognitive modules dedicated to the processing of such basic musical mate-
rials.15 The theory of cognitive modular ity, an adaptation of faculty psychology first
proposed by the philosopher Jerry Fodor, bears some resemblance to the notion of
competence discussed br iefly in my introduction.16 The coherence of musical

14. Cone, On Derivation, 246.

15. See, for example, David Temperley, Motivic Perception and Modular ity, Music Perception 13
(1995): 141 69; and Eugene Nar mour, The Analysis and Cognition of Basic Melodic Structures: The Impli-
cation-Realization Model (Chicago: University of Chicago Press, 1990).
16. Jerry A. Fodor, The Modularity of Mind: An Essay on Faculty Psychology (Cambridge, Mass.: MIT
Press, 1983). According to the theor y Fodor proposed in The Modularity of Mind, cognitive structure is
organized into modules dedicated to specific processing tasks. There are thus modules for such things
30 aspects of cogn it ive st ruc ture

materials, however, is an emergent property that is less well explained from a mod-
ular perspective since it involves the active compar ison and evaluation of a number
of motive forms.
One model that can r ise to these challenges is offered by recent research into
categor ization, which has demonstrated that categor ization is both extremely rapid
and evaluative. From the perspective provided by this research, the coherence of
music reflects our ability to group musical events into categor ies: the a and b motives
of example 1.2 each constitutes a cognitive categor y, with five members in one
category (a, a1, a2, a3, and a4) and four in the other (b, b1, b2, and b3). To descr ibe
the laws of the mind that Schoenberg noted laws that force the composer to
arrange musical elements so that they cohere is to descr ibe the process of cate-
gorization. In what follows, I shall review research on processes of categor ization
and show how it can be used to account for the role of motive as a starting place
for higher-level cognitive processes and for relationships among diverse motive

processes of categorization
For the most part, the categor ies through which we structure our understanding of
the world are not cause for a great deal of reflection. To all appearances, these cat-
egories are simply givens. A book is a book, a tree is a tree, and any equivocal cases
are just a matter of insufficient knowledge. This commonsense view of categoriza-
tion was the predominant view in philosophy and psychology at least since the time
of Plato: categories, because they were constitutive of knowledge, were regarded as
stable and universal. It was only in the 1950s that the adequacy of this view began
to be called into question, most notably by Ludwig Wittgenstein in his Philosophi-
cal Investigations. There, in the midst of a discussion of language games, Wittgenstein
paused to consider the category of game itself and found that category structure
was far more complicated than it first appeared.17 Research carried out in the 1960s
and 1970s showed that numerous aspects of the way humans categor ize could not
be reconciled with the traditional view of categorization, and by the 1980s a new
view of categorization had begun to develop. The categor ies through which
humans organize their understanding of the world came to be seen as ways of hav-
ing knowledge, rather than as reflecting what could be known.
The research crucial to developing this new view of categor ization focused on
two topics. The first concer ns the relationship between the process of categoriza-
tion and the taxonomic structures to which it gives r ise. The second concer ns the

as vision, hearing, language processing, and, presumably, the identification of musical motives. Fodor has
since made arguments against modular ity; see his recent The Mind Doesnt Work That Way (Cambridge,
Mass.: MIT Press, 2000).
17. Ludwig Wittgenstein, Philosophical Investigations, trans. G. E. M. Anscombe (New York: Macmil-
lan, 1953), 31 32. For summar ies of the recent history of research into categor ization, see Eleanor Rosch,
Categor ization, in Encyclopedia of Human Behavior, ed.V. S. Ramachandran (San Diego: Academic Press,
1994), 1: 513 23; and George Lakoff, Women, Fire, and Dangerous Things: What Categories Reveal about the
Mind (Chicago: University of Chicago Press, 1987), chap. 2.
cate gorization 31

internal structure of categor ies that is, how members of a category relate to one
Most of the categor ies we use in daily life can be organized into hierarchically
arranged taxonomies. The cat prowling around my feet is a calico cat, a housecat, a
feline, a mammal, and a living organism. The levels in taxonomies of this sort are related
by inclusion: feline includes all of the different sorts of housecats, a category that
includes all of the different sorts of calico cats, a category that includes my particular
cat. In categor izing things, it seems logical to work from the top of the hierarchy
down, moving from the most abstract level to the most concrete. I note first that the
thing before me is a living organism, and as I increase my level of discrimination
determine exactly what sort of organism it is. In actually categor izing things in the
world around us, however, we start neither at the most specific level (the cat
prowling around my feet) nor at the most general (living organism) but at a level in
between; this level has come to be called the basic level. Research on the basic level
tells us where, for practical pur poses, the process of categorization begins. The first
section that follows reviews research on the basic level and shows how it supports
Schoenbergs contention that musical comprehension begins at the level of the
It is similarly a commonplace to regard each level of a taxonomy as having clear,
fixed boundar ies: a creature either is or is not a housecat. Categories are thus un-
changing anchors of understanding, their structure immutable and constant. The
categories we actually use in daily life, however, cannot be so neatly character ized.
These categor ies have a somewhat more complicated structure: membership is by
degrees; category boundar ies are relatively fluid and often hazy; and category struc-
ture can change over time. In these cases, the process of categorization is a highly
dynamic one, requiring that we evaluate potential categor y members relative to
existing categor y members which, together, define a standard of typicality. Cate-
gories with a variable membership are thus said to show typicality effects. The sec-
ond section that follows reviews research on categor ies that illustrate typicality
effects and shows how it can be used to model the coherence and variability of
motive forms basic to Schoenbergs theory of motive.

Basic-Level Categor ization

the basic leve l The psychologist Roger Brown was perhaps the first to
note that the categor ies we most frequently employ are not at the lowest level of a
taxonomy (and concer ned with individuals), nor are they at the highest level (and
concer ned with broad classes); instead, they are in the middle of a taxonomy, at a
level of maximum utility.18 For instance, we typically call the thing in our purse or
pocket a dime, even though we might also call it money or a metal object (names at the
superordinate level of the taxonomy), or my favorite 1956 dime (which would be at
the subordinate level of the taxonomy).

18. Roger Brown, How Shall a Thing Be Called? Psychological Review 65 (1958): 14 21; idem,
Social Psychology (New York: Free Press, 1965), 319 21.
32 aspects of cogn it ive st ruc ture

table 1.1. Sample Taxonomies Showing the Basic Level

Superordinate Basic Level Subordinates

Musical instrument Guitar Folk guitar

Classical guitar
Piano Grand piano
Upright piano
Fruit Apple Delicious apple
Mackintosh apple
Peach Freestone peach
Cling peach
Furniture Table Kitchen table
Dining room table
Lamp Floor lamp
Desk lamp

Adapted from Eleanor Rosch et al., Basic Objects in Natural Categor ies, Cognitive Psychology 8
(1976): 388.

Subsequent research into ethnobotanical systems of classification by Brent Berlin

and his associates supported the importance of mid-level classification and gave
empirical evidence that this level is psychologically basic: at this level, people name
things more readily, languages have simpler names, categories have greater cultural
significance, things are remembered more readily, and things are perceived holisti-
cally.19 Research directed specifically at mid-level classification was car ried out by
Eleanor Rosch and her associates in the early 1970s. In their report, this level was
named the basic level, a term which has been generally adopted in the literature on
categor ization.20 Examples of sample taxonomies showing the basic level are given
in Table 1.1.
Rosch and her associates suggested that two contrasting pr inciples influence the
taxonomic level at which people prefer to categor ize.21 The first is the efficiency prin-
ciple, according to which people prefer to minimize the number of categor ies they

19. Brent Berlin, Speculations on the Growth of Ethnobotanical Nomenclature, Language in Soci-
ety 1 (1972): 51 86; idem, Ethnobotanical Classification, in Cognition and Categorization, ed. Eleanor
Rosch and Barbara B. Lloyd (Hillsdale, N.J.: Erlbaum Associates, 1978), 9 26; Brent Berlin, Dennis E.
Breedlove, and Peter H. Raven, Principles of Tzeltal Plant Classification: An Introduction to the Botanical
Ethnography of a Mayan-Speaking People of Chiapas (New York: Academic Press, 1974).
20. Eleanor Rosch, Carolyn B. Mervis,Wayne D. Gray, David M. Johnson, and Penny Boyes-Braem,
Basic Objects in Natural Categor ies, Cognitive Psychology 8 (1976): 382 439.
21. This inter pretation of Roschs work is drawn from Lawrence W. Barsalou, Cognitive Psychology:
An Overview for Cognitive Scientists (Hillsdale, N.J.: Erlbaum Associates, 1992), 181 84. See also Rosch et
al., Basic Objects in Natural Categor ies; Eleanor Rosch, Principles of Categorization, in Cognition
and Categorization, ed. Eleanor Rosch and Barbara B. Lloyd (Hillsdale, N.J.: Erlbaum Associates, 1978),
27 48; Carolyn B. Mervis and Eleanor Rosch, Categor ization of Natural Objects, Annual Review of
Psychology 32 (1981): 89 115; and Mary E. Lassaline, Edward J.Wisniewski, and Douglas L. Medin, Basic
Levels in Artificial and Natural Categor ies: Are All Basic Levels Created Equal? in Percepts, Concepts, and
Categories: The Representation and Processing of Information, ed. Barbara Bur ns, Advances in Psychology, 93
(Amsterdam: North-Holland, 1992), 327 78.
cate gorization 33

must consider in making a categor ization. For instance, if you were concer ned with
things in my house, categor izing at the level of musical instrument would yield the most
efficient categor y, since the number of contrasting categor ies at this level (such as
furniture and fruit) is significantly less than at the next level (guitar, piano, apple, peach,
table, lamp). The second pr inciple is the informativeness principle, according to which
people tend to maximize the infor mativeness of their categor izations. Since the
most infor mation about any entity is found at the most specific level of a taxonomy,
you would use grand piano to categor ize the thing sitting in the living room. Rosch
and her associates proposed that the inter mediate level of a taxonomy (in this case,
piano) optimizes both efficiency and infor mativeness and is thus the preferred level
for basic categor izations. A number of empirical operations converge at the basic
level. The basic level is the highest level whose members have similar and recog-
nizable shapes; it is also the most abstract level for which a single mental image can
be formed for the category. The basic level is also the highest level at which a per-
son uses similar motor actions for interacting with categor y members. The basic
level is psychologically basic: it is the level at which subjects are fastest at identifying
category members, the level with the most commonly used labels for categor y
members, the first level named and understood by children, the first level to enter
the lexicon of a language, and the level with the shortest pr imary lexemes.22
In her initial work, Rosch assumed that the basic level depended on bundles of
perceived-world attr ibutes that for med natural discontinuities and that such attr i-
butes were inherent in the real world. She later came to doubt this assumption, and
subsequent research has shown that the basic level is variable. It depends on cultural
context and level of expertise and on whether one is dealing with categor ies that
occur naturally, as humans interact with their environments, or with categor ies cre-
ated for exper imental ends.23 Other research has indicated that the basic level is
strongly deter mined by the ways in which humans interact with their environment
and that it is highly dependent on the way humans perceive the spatial configura-
tion of physical parts.24 In all cases, however, Roger Browns fundamental insight has
held up: the preferred categor ization for most of the things we encounter occupies
a maximally useful level in the middle of a taxonomy.

22. Rosch et al., Basic Objects in Natural Categor ies; Eleanor Rosch, Human Categor ization,
in Studies in Cross-Cultural Psychology, vol. 1, ed. Neil Warren (London: Academic Press, 1977), 1 49; Bar-
bara Tversky and Kathleen Hemenway, Objects, Parts, and Categories, Journal of Experimental Psychol-
ogy: General 113 (1984): 169 93.
23. On Roschs reservations about the objectivity of the basic level, see Rosch, Principles of Cat-
egorization; on the relationship between expertise and the basic level, see James W. Tanaka and Marjor ie
Taylor, Object Categor ies and Expertise: Is the Basic Level in the Eye of the Beholder? Cognitive Psy-
chology 23 (1991): 457 82; on the contrast between natural and artificial categor ies in this context, see
Lassaline et al., Basic Levels in Artificial and Natural Categor ies.
24. On relationships between the basic level and the environment, see Barbara Tversky and Kath-
leen Hemenway, Categor ies of Environmental Scenes, Cognitive Psychology 15 (1983): 121 49; on the
basic level and humans interactions with physical objects, see Tversky and Hemenway, Objects, Parts,
and Categor ies; Irving Bieder man, Recognition-by-Components: A Theor y of Human Image
Understanding, Psychological Review 94 (1987): 115 47; and Lawrence W. Barsalou, Deriving Cate-
gories to Achieve Goals, Psychology of Learning and Motivation: Advances in Research and Theory 27
(1991): 1 64.
34 aspects of cogn it ive st ruc ture

the basic level and musical motive s In a close analogue to

what are, perhaps, the most distinctive features of basic-level categor ies, Schoen-
bergs theory of musical coherence begins not with individual musical events or
with four- or eight-measure phrases, but at a level somewhere in between. Al-
though the motive is the smallest recognizable part of a musical work, it is in fact
made up of still smaller parts namely, its constituent pitches, intervals, and durations.
The cognitive salience of the motive thus mir rors that of the basic level: in both
cases, the focus is on the whole rather than on the parts. Each of Schoenbergs fam-
ilies of motives has a distinctive shape or contour that allows us to distinguish it
from other motives. Similarly, the basic level is the highest taxonomic level at which
category members have similarly perceived overall shapes and the highest level at
which a single mental image can reflect the entire categor y. Finally, motive recog-
nition is not something given directly by perception and can change with expertise:
what counts as a motive, just as what counts as the basic level, is not immutable.
To illustrate cor respondences between a musical motive and a basic-level cate-
gory, let us consider the opening of Beethovens Fifth Symphony, given in example
1.5. The passage is a familiar one, and it illustrates important matters of musical
organization that we will later explore in Wagners Tristan.25 To the extent that we
hear a motive in these opening gestures, we hear a figure that compr ises four
notes. The motive has a distinctive shape or contour that allows us to distinguish
it from other motives (although such are in rather short supply in this particular
example). Lastly, there is a certain minimal level of expertise involved in picking out
mm. 1 2 and its ilk as the motive. One can easily imag ine an alter nate hear ing
that would take mm. 1 2 and mm. 3 4 as the main thematic mater ial for this sec-
tion; the ensuing music would then represent a sort of fragmentation of this the-
matic mater ial, rather than successive statements of various for ms of the motive.26

summary Research on basic-level categor ies provides support for Schoen-

bergs proposal that musical comprehension beg ins at the motivic level, for such a
level maximizes both efficiency and infor mativeness. In addition, there appear to be
a number of correlations between musical motives and basic-level categor ies. Both
are concer ned with wholes, rather than the parts of these wholes; shape (whether as
visual property or auditory analogue) is important to both; and both are subject to
the influence of expertise. Finally, while entities such as motives may be basic to
the process of comprehending music, they are actually relatively high-level cogni-
tive constructs (as are basic-level categor ies). Although such constructs are funda-
mental to reason (and may even represent an important starting point for conscious

25. That a passage from Beethoven might be used to shed light on Wagners compositional strategies
is not accidental, for Beethovens compositional style had a profound influence on Wagner. For discus-
sion, see Klaus Kropfinger, Wagner and Beethoven: Richard Wagners Reception of Beethoven, trans. Peter
Palmer (Cambr idge: Cambridge University Press, 1991); and Thomas S. Grey, Wagners Musical Prose:
Texts and Contexts (Cambridge: Cambridge University Press, 1995).
26. It is worth noting, however, the role of the fermatas in suggesting that mm. 1 2 are one entity
and mm. 3 4 another, since these add time for the listener to process the musical infor mation presented
by Beethoven. This additional time may also play a role in what is considered the most typical form
of the motive, a topic explored in greater depth later in this chapter.
example 1.5 Ludwig van Beethoven, Symphony No. 5 in C Minor, Op. 67, first
movement, mm. 1 38

1 Allegro con brio ( = 108)



Clarinets in B


Horns in E

Trumpets in C

Timpani in C, G

Allegro con brio ( = 108)

Violin I

Violin II




36 aspects of cogn it ive st ruc ture

example 1.5 (continued)



reasoning processes), they are the product of myriad other still more-basic processes
that operate below the level of consciousness.

Typicality Effects

graded structure and a typology of categories As noted,

the variability of motive forms was important to Schoenbergs notion of musical
coherence, for this variability tended to emphasize how motive forms were similar to
cate gorization 37

example 1.5 (continued)















one another. Research on categor ization has demonstrated an analogous structure for
numerous categor ies. Membership in these categor ies is not fixed but is instead
graded through a dynamic process in which the attr ibutes of potential category
members are compared with the attr ibutes most typically found within the category.
As an example of such a graded structure, consider the category bird. Experimental
rankings show that subjects view robins and sparrows as the best examples of birds, with
38 aspects of cogn it ive st ruc ture

example 1.5 (continued)


owls and eagles lower down in the rankings, and ostriches, emus, and penguins among
the worst examples.27 All are considered members of the category bird, but some bet-
ter represent the category than others. Consequently, category structure is graded
according to typicality: category members range from the most typical to the least typ-

27. Eleanor Rosch, On the Inter nal Structure of Perceptual and Semantic Categor ies, in Cogni-
tive Development and the Acquisition of Language, ed. Timothy E. Moore (New York: Academic Press, 1973),
111 44; idem, Cognitive Representations of Semantic Categor ies, Journal of Experimental Psychology:
General 104 (1975): 192 233.
cate gorization 39

example 1.5 (continued)


ical, with the former securely inside the bounds of the category (robins and sparrows)
and the latter in danger of being excluded from the category (emus and penguins).
Graded membership has been shown to be pervasive among what are conven-
tionally called natural categories, so named based on their emergence from the
interaction of humans with their natural environments.28 For the sake of clarity, and
28. On the pervasiveness of natural categor ies, see Lawrence W. Barsalou, The Instability of Graded
Structure: Implications for the Nature of Concepts, in Concepts and Conceptual Development: Ecological
and Intellectual Factors in Categorization, ed. Ulric Neisser (Cambr idge: Cambridge University Press, 1987),
102. On the ter m natural applied to certain categor ies, see Barsalou, Cognitive Psychology, chap. 2.
40 aspects of cogn it ive st ruc ture

example 1.5 (continued)









to avoid certain of the associations summoned by the ter m natural, I call this sort
of category a Type 1 category.
Type 1 categor ies stand in contrast to categor ies in which membership is based
on individually necessar y and jointly sufficient conditions, such that the members
of category x must have features y and z, and if something has features y and z it is
a member of category x. Because the conception of categories of this latter sort is
often traced backed to the work of Aristotle, such categor ies have been called clas-
cate gorization 41

sical categories; because the structure of these categor ies is not typically an analog
representation of the natural world, such categor ies have also been called artificial.29
I call this sort of category a Type 2 category. Most evidence indicates that Type 2
categories simply represent a specialized for m of Type 1 categor ies take a Type 1
category, specify limits for the categor y through the imposition of necessar y and
sufficient conditions for category membership, and you create a Type 2 category.30

prototype effects and frame structure Eleanor Rosch and

Carolyn Mervis, in their pioneer ing work on Type 1 categor ies, suggested that such
categories were organized around a stable cognitive construct called a prototype,
which encapsulated the statistically most-prevalent features of members of the cat-
egory and against which potential category members were compared.31 Type 1 cat-
egories were thus said to exhibit prototype effects. It is clear, however, that the
prototype offers more structure than simply a record of the attr ibutes cor related
with a particular category: although wings and beak are both highly cor related with
bird, we know that wings are used for locomotion and beaks are used for eating. Rela-
tionships between attr ibutes, as well as the values assigned these attr ibutes, are thus
important to the function of the prototype.
As a way of captur ing the relational structure inherent in most categor ies,
Lawrence Barsalou proposed using the notion of a frame, as developed in artificial
intelligence research, to represent category structure.32 As Barsalou has shown, the

29. On classical or artificial categor ies, see Edward E. Smith and Douglas L. Medin, Categories and
Concepts (Cambridge, Mass.: Harvard University Press, 1981), chap. 1; Sharon Lee Armstrong, Lila R.
Gleitman, and Henry Gleitman, What Some Concepts Might Not Be, Cognition 13 (1983): 263 308;
Lila R. Gleitman, Sharon Lee Armstrong, and Henry Gleitman, On Doubting the Concept Concept,
in New Trends in Conceptual Representation: Challenges to Piagets Theory?, ed. Ellin Kofsky Scholnick
(Hillsdale, N.J.: Erlbaum Associates, 1983), 87 110; Lakoff, Women, Fire, and Dangerous Things, chap. 2;
Barsalou, Cognitive Psychology, chap. 2.
30. Although the typology of categories I introduce here places a burden on those familiar with the
traditional labels, it has two advantages. First, it avoids the complications caused if natural is inter preted
as given by nature; while Type 1 categor ies result from humans interactions with their environment,
they are not given by that environment in any simple way. Second, the typology classifies categor ies on the
basis of their internal structure. This is especially useful when consider ing research on categor ization, since
artificial categor ies (such as things to pack in a small suitcase, as discussed by Barsalou in his Deriving Cat-
egories to Achieve Goals) often have the same basic structure as do natural categories (such as bird).
31. Eleanor Rosch and Carolyn B. Mervis, Family Resemblances, Cognitive Psychology 7 (1975):
573 605. See also James Hampton, Prototype Models of Concept Representation, in Categories and
Concepts: Theoretical Views and Inductive Analysis, ed. Iven van Mechelen, James Hampton, Ryszard S.
Michalski, and Peter Theuns, Cognitive Science Ser ies (London: Academic Press, 1993), 67 95.
32. Barsalou, Cognitive Psychology, chap. 7; idem, Frames, Concepts, and Conceptual Fields, in
Frames, Fields, and Contrasts: New Essays in Semantic and Lexical Organization, ed. Adrienne Lehrer and
Eva Feder Kittay (Hillsdale, N.J.: Erlbaum Associates, 1992), 21 74; idem, Flexibility, Structure, and Lin-
guistic Vagary in Concepts: Manifestations of a Compositional System of Perceptual Symbols, in Theo-
ries of Memory, ed. A. F. Collins, S. E. Gathercole, M. A. Conway, and P. E. Morris (Hillsdale, N.J.: Erlbaum
Associates, 1993), 29 101.
The exploration and implementation of frame structures in work on artificial intelligence is demon-
strated in Marvin Minsky, A Framework for Representing Knowledge, in The Psychology of Computer
Vision, ed. Patrick Henry Winston (New York: McGraw-Hill, 1975), 211 77; Marvin Minsky, The Society
of Mind (New York: Simon and Schuster, 1985); and Douglas B. Lenat and R. V. Guha, Building Large
42 aspects of cogn it ive st ruc ture

%##$546#"7* 8%96"7* 5:)585)6%97*

<")56<* =$":*

7<%99* (#

4$(=:* (#

&(9($* ='5#"*

8-(/$ >$"":*

&'5$?7* (#
7(6:)* &96&@7*




figure 1.1 A partial frame for the category bird (asterisks indicate the values of the
prototype that are stored in memory)

basic structure of a Type 1 category can be conceived of as a relational network

made up of relatively abstract attr ibutes, to which concrete values are assigned by
each individual. For the categor y bird, for example, attributes might include size,
color, sound, and locomotion, as shown in figure 1.1. For the attr ibute size, values might
include large, medium, and small. Each member of the categor y exemplifies specific
values for each of the attr ibutes: a wren is small, brown, chirps, and flies; a chicken is
large, white, clucks, and (for the most part) runs.
Of course, individuals with none of the attr ibutes of the category would be
excluded from it. Prototype effects ar ise from the compar ison of the values repre-
sented by individuals with the values of the prototype for the categor y that are
stored in memory (indicated with aster isks in fig. 1.1). If the values of the prototype
for bird were small, brown, chirps, and flies, then a wren would be most typical of the
category, a male cardinal (small, red, sings, flies) would be somewhat less typical, and
a chicken (large, white, clucks, runs) would be least typical.

applications to music Returning to the opening of the Fifth Sym-

phony, we can see that, as suggested by Schoenbergs account of motive, there are a

Knowledge-Based Systems: Representation and Inference in the Cyc Project (Reading, Mass.: Addison Wesley,
cate gorization 43

number of different for ms that Beethovens motive takes in the first thirty-seven
measures of the movement, and these vary with respect to contour, dynamics, and
orchestration. Approached from the perspective of categorization, each motive form
can be viewed as a member of the categor y motive forms from the opening of Bee-
thovens Fifth Symphony.33 Attributes of this category might include things such as
orchestration, dynamic, and melodic profile. In the following discussion, I show how
these three attr ibutes could play a part in deter mining membership in the category
motive forms from the opening of Beethovens Fifth Symphony; however, I should empha-
size that, for present pur poses, what is of concer n is less the specific attr ibutes and
more the model of categor ization they illustrate.
To character ize melodic profile, it is useful to account for intervallic relationships
among motive forms in ter ms of diatonic intervals. That is, within the diatonic con-
text of the movement, the descending third from G4 to Eb4 can be reckoned to be
equivalent to the descending third from F4 to D4, even though the for mer is major
and the latter is minor. Here I represent a descending third with 3, an ascending
third with +3, and a unison with a u.With this in mind, the attr ibute-values for the
first two statements of the motive would be tutti, fortissimo, and u, u, 3.
Although rhythmic profile is an equally important attr ibute of the category, it is
not as useful for distinguishing among the various motive forms found in the open-
ing measures shown in example 1.5. All the motive forms found there would assign
the same value to this attr ibute: short, short, short, long (where short is equivalent to
an eighth note and long varies between a quarter note and three tied half notes). In
consequence, the attr ibute rhythmic profile is relatively transparent in the opening of
the symphony, as is the attr ibute played by instruments; both will be omitted from the
following account of category structure.
Figure 1.2 presents a frame diag ram for the categor y motive forms from the open-
ing of Beethovens Fifth Symphony; it shows the values that thirteen instances of the
motive assign to the attr ibutes orchestration, dynamic, and diatonic melodic profile. The
result is a rather dense diag ram (owing in part to the number of individuals cate-
gorized), which gives an intr iguing and rather counter intuitive picture of what is
most typical of the various motive forms gathered there. Based on the evidence pro-
vided by these individuals, the most typical for m of the motive would assign the
values solo, piano, and u, u, 3 to the relevant attr ibutes. Although the last value is
what we might expect, the first two are not: it is difficult to think of Beethovens
theme as anything other than tutti and fortissimo.
Were we to take into account a further fifteen instances of the motive, the profile

33. What follows is by no means the first attempt to use research on categor ization to model vari-
ation among musical elements; notable are the studies of Robert L.Welker,Abstraction of Themes from
Melodic Variations, Journal of Experimental Psychology: Human Perception and Performance 8 (1982): 435 47;
and Lucy Pollard-Gott, Emergence of Thematic Concepts in Repeated Listening to Music, Cognitive
Psychology 15 (1983): 66 94. The following extends this work by taking advantage of research done since
the early 1980s and by endeavoring to give a more thorough account of the process of categor izing
musical events. Mention should also be made of Robert O. Gjerdingens discussion of typicality in A
Classic Turn of Phrase: Music and the Psychology of Convention, Studies in the Cr iticism and Theor y of
Music (Philadelphia: University of Pennsylvania Press, 1988), 93 94, which anticipates many of the
points made here.
44 aspects of cogn it ive st ruc ture

!""#$%&"'() *!+&'() $,-$*$-&!+()

5#67'("#!"$5,) (5+5)




-<,!.$6) $%&'$(#




-$!"5,$6) ../)04108

.'+5-$6) &>)&>)13)
@#5A$+') ../)08109)



figure 1.2 Diagram of the category structure for the category motive forms from the
opening of Beethovens Fifth Symphony, mm. 1 21

of typicality would be skewed even further away from what we might expect (this
is shown in fig. 1.3, where the frame diagram has been rearranged somewhat in the
interests of legibility). The difference between the evidence provided by figures 1.2
and 1.3 and our usual recollection of how Beethovens theme sounds is due in part
to a limitation of frame diagrams: they are essentially statistical representations of the
most common values assigned attr ibutes by individuals. They may not represent the
most important attribute values for a given category, however.

beyond the prototype: conceptual models Recent empir i-

cal work on categor ization has shown that category structure is even more variable
than had been assumed by the researchers who first proposed the prototype theory
of category structure.34 In most Type 1 categor ies, some attr ibutes are more impor-
tant to categor y structure than others. As a result, some of the statistically most
prevalent features of members of a category may be disregarded in for mulating the

34. Armstrong et al., What Some Concepts Might Not Be; Gleitman et al., On Doubting the
Concept Concept ; Barsalou, The Instability of Graded Structure; Lawrence W. Barsalou, Intracon-
cept Similar ity and Its Implications for Interconcept Similar ity, in Similarity and Analogical Reasoning, ed.
S.Vosniadou and A. Ortony (Cambr idge: Cambridge University Press, 1989), 76 121.
cate gorization 45

228#?67# C5CC*# 228#7"679#

1-4-# 228#7967:#

228#96:# 228#7:67B#

228#:6B# ./#..# 228#7B67A#

228#B6A# 228#7A6'@#

228#?@6??# 228#'@6'?#

228#??6?7# 228#'?6'7#

228#?76?'# 228#'76''#
228#?%6?"# ;<4,8#=>#

228#?"6?9# 5/#5/#5/# ;<4,8#==/#<4+8>#

228#?96?:# 228#'%6'"#

228#?:6?B# ;<4,8#=>#
228#?B6?A# ;<4,8#==/#<4+8>#

228#7767%# D%/#5/#5# 228#'96':#

figure 1.3 Diagram of the category structure for the category motive forms from the
opening of Beethovens Fifth Symphony, mm. 1 37

prototype. For example, attributes such as natural environment and means of locomo-
tion have a strong influence on how we categor ize whales, leading us to disregard
such taxonomic attr ibutes as mode of respiration and means of reproduction. In conse-
quence, a whale seems closer to a prototypical fish than to a prototypical mammal,
despite its proper Linnean categor ization.35 Research has also shown that judgments
of typicality change over time, even when no new individuals are categor ized. It
appears that, rather than constituting a stable cognitive entity, the prototype for
a category is actually fluid and variable and it changes with the circumstances of
An alter native view of category structure suggests that categor ies are organized
around conceptual models. In brief, conceptual models consist of concepts in specified
relationships; each model pertains to a specific domain of knowledge. Such models
are central to reasoning and inference, and they will be discussed in g reater detail
in chapter 3; in what follows, the emphasis is on the role conceptual models play in
categorization, rather than on the structure of such models. With regard to catego-
35. For a discussion of some of the factors involved in this particular categor ization, see John Dupr,
Are Whales Fish? in Folkbiology, ed. Douglas L. Medin and Scott Atran (Cambr idge, Mass.: MIT Press,
1999), 461 76.
46 aspects of cogn it ive st ruc ture

rization, the primary function of the conceptual model is to supply a guide for rea-
soning about accepted and potential members of the category. This is accomplished
through a simplified representation of category structure that incor porates knowl-
edge about what values are most typical for a select group of attributes for the given
category. These attr ibutes are selected according to the goals of categor ization,
which are themselves infor med by more global conceptual models applicable to a
broad range of categorization tasks.36 The conceptual model for a g iven category
thus reflects infor mation of the sort summar ized in frame diagrams but conditioned
by knowledge about the overall goals of categor ization.
As a means of developing a preliminary understanding of the concepts and rela-
tionships constituent of a conceptual model for a category, let us return once more
to the opening of Beethovens symphony and consider a conceptual model for the
category motive forms from the opening of Beethovens Fifth Symphony. The model, dia-
grammed in figure 1.4, consists of four correlated conceptual elements, character ized
as the things necessary for a melodic fragment to count as an instance of the motive:
the rhythmic patter n h
a statement by the full orchestra
a fortissimo dynamic
an intervallic patter n consisting of three repeated notes followed by a
descending diatonic third37
These elements g ive priority to select attr ibutes of the motive (rhythmic patter n,
orchestration, dynamics, and intervallic profile) and assign default values to these
attributes that is, they define what counts as a typical member of the category.
This network of attributes and values then provides a guide for categor izing the
various versions of the motive that occur in the opening measures. Accordingly, ver-
sions of the motive such as those of mm. 1 2 and mm. 22 24 (which instantiate all
aspects of the conceptual model) would be judged most typical of the category;
motive forms such as those of mm. 6 7 and mm. 29 30 (which instantiate only
two aspects of the conceptual model) would be less typical; and motive forms such
as those of mm. 7 8 and mm. 35 36 (which instantiate only the rhythmic aspect of
the conceptual model) would be least typical of the category.
What, then, of the evidence provided by figure 1.3? That is, what can one make
of the lack of correlation between the conceptual model and the individual cate-
gory members, since most of these would be reckoned to be less typical members
according to the conceptual model for the categor y? The answer to this question
requires delving a bit deeper into the way conceptual models both reflect and shape
our understanding of music.

36. Barsalou, Deriving Categor ies to Achieve Goals; Lawrence W. Barsalou, Wenchi Yeh, Barbara
J. Luka, Karen L. Olseth, Kelly S. Mix, and Ling-Ling Wu, Concepts and Meaning, in Chicago Lin-
guistics Society 29: Papers from the Parasession on the Correspondence of Conceptual, Semantic, and Grammatical
Representations, ed. Kathar ine Beals, Gina Cooke, David Kathman, Sotaro Kita, Karl-Er ik McCullough,
and David Testen (Chicago: University of Chicago, Chicago Linguistics Society, 1993), 23 61.
37. I should emphasize that the conceptual models I descr ibe here are by no means the simplest ele-
ments of conceptual structure. Each conceptual cluster of the present model, for instance, could be rep-
resented by a still smaller conceptual model cor relating the concepts that make up the cluster.
cate gorization 47

!""!""!" #"



figure 1.4 Conceptual model for the category motive forms from the opening of
Beethovens Fifth Symphony

The conceptual model shown in figure 1.4 gives undeniable priority to the
statements of the motive that open the movement. The preference given the first
statements of musical mater ials we encounter is not absolute there are, of course,
works in which the pr incipal mater ials emerge only gradually but it does reflect
the fact that we must process musical mater ials in time and that it is often the case
that significance accrues to the first events in any psychological process. Note that
Beethoven departs from his opening mater ials only gradually; in particular, diver-
gences from the intervallic patter n of the motive (the aspect of the typicality shown
by fig. 1.3 that best matches the proposed conceptual model) occur somewhat later
in the opening measures and cor relate with the compositional strateg ies of devel-
opment and transition.38 The initial statements of the motive thus represent anchor
points for the process of elaboration that Beethoven undertakes: the variations evi-
dent in figure 1.3 are variations on the musical mater ials summar ized in the con-
ceptual model of figure 1.4.
The conceptual model of figure 1.4 is also infor med by our ideas about musical
themes: that is, such themes should be strongly and clearly stated at pr ivileged points
within a movement (such as the beg inning of the movement or after a change of
key or tempo), after which they may be modified or embellished in confor mance
with compositional strategies. Our local model for the motive of Beethovens sym-
phony is thus shaped by a global model of what constitutes an appropriate musical
theme. Such global models are developed through abstraction from any number of
local models, and, perhaps more important, they are taken from the broader base of
knowledge that constitutes culture.39 For instance, in the case of Beethovens sym-
phony, the relevant global model for a musical theme reflects the influence of nine-
teenth-century German and Austrian musical culture and, in particular, the efforts

38. Relationships between processes of categor ization and compositional strategy are developed
more fully in chap. 4.
39. This perspective on culture is explored more fully in chaps. 3 and 5.
48 aspects of cogn it ive st ruc ture

of the nineteenth-centur y critic and theor ist A. B. Marx and others to come to
terms with Beethovens compositional style.40
Our global model of what counts as a theme (which applies to any number of
works by any number of composers) is thus in rapport with the local models through
which we organize our understanding of specific musical works. Without a global
model, it can be difficult to decide what aspects of a phenomenon are relevant for
categorization. Without local models, there is no way to anchor global models to
specific phenomena.
It should be emphasized that the conceptual model sketched here and dia-
grammed in figure 1.4 represents limited features of a network of interrelated
propositions that might be used to guide reasoning about how motive forms in the
opening of Beethovens symphony relate to one another. The model does not rep-
resent, in a simple way, what someone has in mind when confronting this music.
The simplicity of the model (which reflects in part the global models relative to
which it is framed) contr ibutes to its efficiency. It is specific enough to allow for
clear distinctions between the motive of the opening measures and other thematic
mater ials in the symphony (such as the new theme that enters in m. 63). It is gen-
eral enough, however, to accommodate the wide range of different for ms the open-
ing motive takes. The model also reflects our intuition that there is just one main
form of the opening motive, despite evidence to the contrary. Intuitions such as this
contribute to the plausibility of Type 2 categor ies, with their cr isp boundar ies and
clear definitions. Features of the conceptual model for a g iven categor y will often
be transfor med into the necessar y and sufficient conditions character istic of Type
2 categor ies: take a Type 1 category and specify limits for it by regarding the con-
stituent elements of the conceptual model relative to which it is structured as nec-
essary and sufficient conditions for category membership, and you create a Type 2
category. In descr iptions of music, this process yields textbook definitions of the
sort that often lead to considerable debate among music theor ists. Although such
debates appear to be about music, they are in fact about how to define the cate-
gories through which we organize our understanding of music.

summary: typicality effects and musical motive Cate-

gories that show typicality effects offer a way to model both the coherence and the
variability of motive forms basic to Schoenbergs theory of motive. The property of
coherence is analogous to the shared attr ibutes of category members that is, our

40. Marxs is perhaps the most important nineteenth-century voice within the tradition of analyz-
ing the thematic aspects of Beethovens music. Although Marx did not restrict himself to a definition of
theme as such, preferring a broader and more organic account of musical organization, he did offer the
notion of the Satz. Formed from motives, the Satz is a complete self-sufficient musical idea and repre-
sents the first coherent unit of musical organization. See Adolph Ber nhard Marx, Musical Form in the Age
of Beethoven: Selected Writings on Theory and Method, ed. and trans. Scott G. Burnham, Cambridge Stud-
ies in Music Theor y and Analysis, 12 (Cambridge: Cambridge University Press, 1997), 68. The Satz is
roughly comparable to Schoenbergs notion of a Grundgestalt, as discussed in Schoenberg, The Musical
Idea, 169. It is no accident, of course, that Marxs formal theor ies reflect his efforts to make explicit
Beethovens compositional strategies. That is, to the extent that our notion of what constitutes a theme
resonates with Marxs definition, that notion is infor med by works such as the Fifth Symphony. Further
discussion of Marxs notion of a motive is taken up in chap. 7.
cate gorization 49

sense that a collection of things coheres as a categor y reflects the attr ibutes shared
by those things. The variability Schoenberg noted is analogous to the variation
among members of a Type 1 category. Just as certain motive forms are taken as the
source for the der ivation of further motive forms, so certain category members will
be regarded as more typical of the category than others. While typicality effects
reflect, to a certain extent, statistical invariances among the attr ibute-values demon-
strated by category members, they are also a function of the goals of categor ization.
These goals shape the categorys conceptual model, which represents what is regarded
as most typical of the category and which also provides a guide for reasoning and
inference about potential category members.
An important difference between collections of musical motives and cognitive
categories is that the for mer must be processed under the temporal constraints of
musical perfor mance. Under most circumstances, these constraints will infor m the
conceptual model for a categor y of motive forms, with priority given to those
forms that occur early in a work. As the example of Schuberts Wanderer Fantasie
suggests, however, other factors may cause our concept of what constitutes the cen-
tral motive of a work to change as the piece unfolds.

categorization and
wagnerian leitmotiven
Wagners Leidensmotiv
Approaching the seven statements of the opening motive from Tristan und Isolde
given in example 1.1 as a cognitive categor y offers a way to explain why it is that
we consider them the same even when they are manifestly different, and it gives
some insight into their role as a point of entry into Wagners sonic fantasy. Knowing
something about the structure of cognitive categor ies also gives us a way to glimpse
features of Wagners compositional strategy, which both exploits and challenges our
ability to categor ize.
Before we explore the motive as a category, let us first situate the motive in its
original context. As shown in example 1.6, the motive is accompanied by harmonies
(provided by the clar inets and bassoons) and a r ising chromatic line in the oboes.
Hans von Wolzogen, in his classification of the motives from Tristan und Isolde (which
he referred to as Leitmotiven), called the main motive the Leidensmotiv (the suffer ing
motive) and the r ising chromatic line the Sehnsuchtsmotiv, or the yearning motive.41
The har mony on the downbeat of m. 2 is traditionally known as the Tristan chord.
As Carl Dahlhaus has noted, all of these elements are intertwined to such an extent
that taking any one of them out of context distorts the whole.42 Nonetheless, it is the

41. Hans von Wolzogen, Thematischer Leitfaden durch die Musik zu Richard Wagners Tristan und
Isolde: Nebste einem Vorworte ber den Sagenstoff des Wagnerschen Dramas (Leipzig: Feodor Reinboth,
1890), 15. This work was also published in an English edition as Guide through the Musical Motives of
Richard Wagners Tristan and Isolde: With a Preface on the Legend and the Poem of Wagners Drama (Leipzig:
Feodor Reinboth, 1889).
42. Dahlhaus, Richard Wagners Music Dramas, 63.
50 aspects of cogn it ive st ruc ture

example 1.6 Opening motivic group from Richard Wagners Tristan und Isolde

changes to the Leidensmotiv that are the most noticeable (not the least because it is
the first sounding element of the ensemble), and so in the following I will focus on
it, with the understanding that it is but one part if a rather prominent one of a
larger whole.43
With these thoughts in mind, let me propose that the categor y statements of the
Leidensmotiv is organized around a conceptual model consisting in two cor related
elements, character ized as the things necessar y for a melodic fragment to count as
an instance of the motive:
an ascending sixth, followed by three descending half-steps
, ,
the rhythmic patter n Z eqd q eqd e44
Again, this model is greatly simplified, for in addition to ignor ing the mater ials that
invariably accompany the Leidensmotiv, nothing is said of dynamics or orchestra-
tion, and the notion of a motive or theme is simply assumed (as part of the
global model that frames this local model). The model is sufficient, however, to dis-
tinguish the Leidensmotiv from other motivic mater ials in Wagners music drama
and to evaluate the typicality of different versions that appear over the course of the
Turning to the versions of the motive given in example 1.1 we can note the fol-
lowing range of typicality: statements 1, 2, 6, and 7 are in complete confor mance
with the model (they are typical of the category), but statements 3, 4, and 5 confor m
rather less well (they are less typical of the category).Viewed from the perspective of
compositional strategy,Wagner begins with two thoroughly typical members of the
category and then follows them with a markedly less-typical member that adds a
note and changes the rhythmic profile. After a considerable span of music, he re-
turns to the motive, but again in a less than typical for m indeed, the two tempo-
rally compressed versions of mm. 85 90 enact the sense of reminiscence they are
meant to evoke by presenting to our ears a distillation of the most salient aspects of

43. Robert P. Morgan has recently offered an analysis of the Tristan Prelude which focuses more on
harmonic structure than on the appearance of motive forms and which offers an intr iguing alter native
to what I present in the following. See Morgan, Circular Form in the Tristan Prelude, Journal of the
American Musicological Society 53 (2000): 69 103. A selection of analyses of the Prelude spanning the
twentieth century can be found in Robert Bailey, ed., Prelude and Transfiguration from Tristan and Isolde
(New York:W.W. Norton, 1985).
44. The actual duration of the concluding note of the motive is variable; in all cases, however, its
duration is of moderate length relative to that of the other notes that make up the motive.
cate gorization 51

the motive. Ten measures further (beginning in m. 100) there are two final appear-
ances of the motive before the curtain goes up, each a thoroughly typical for m.
Wagners compositional strategy thus falls into three stages. In the first stage (asso-
ciated with statements 1 and 2), he establishes the motive. In the second stage (asso-
ciated with statements 3, 4, and 5), he presents less typical for ms of the motive; these
are cor related with a larger set of developmental strategies, with which the central
portion of the Prelude is concer ned. Finally, in the third stage (associated with state-
ments 6 and 7), he returns to more typical for ms of the motive, preparatory to con-
cluding the Prelude and beginning the music drama proper.45
Two things emerge from this initial sketch. First, in the different versions of the
Leidensmotiv that appear in the Prelude, we have a category that develops over
time. Because categor ization is an active response to the environment, it always has
a temporal dimension: a comparative categor ization of birds, for instance, demands
that we evaluate each in turn.With music the temporal aspect is slightly more promi-
nent, since musical entities are thoroughly ephemeral: not only must we evaluate
each musical entity in tur n, but also this evaluation is practically the only evidence
that the entities ever existed at all. Wagner is clearly conscious of the way the Lei-
densmotiv functions through time and uses it to g ive shape to the Prelude as a
whole. Second, the view of Wagners compositional strategy der ived from compar-
ing the versions of the motive in the Prelude, fascinating as it might be, is also mis-
leading: the sort of circularity that it suggests is fundamentally at odds with his goal
of perpetually progressive music. By looking at further instances of the Leidensmo-
tiv over the course of Tristan und Isolde we can develop our understanding of both
of these things and refine our inter pretation of Wagners treatment of the motive in
the Prelude.

The Leidensmotiv and Musical Syntax

The Leidensmotiv is prominent at five moments in Tristan und Isolde: in act 1, scene
2, mm. 26 33 (p. 36; see ex. 1.7a) and scene 5, mm. 439 41, 456 58, and 468 72
(pp. 175 78; ex. 1.7b); in act 2, scene 3, mm. 260 78 (p. 448; ex. 1.7c); and in act 3,
scene 2, mm. 93 107 (pp. 594 96; ex. 1.7d) and scene 3 (near the conclusion of the
opera), mm. 153 57 (p. 630; ex. 1.7e).46 In each case, there are three entr ies of the
motive, beginning (as did those in the opening of the Prelude) on A (or, in the case
of the first entry in act 3, scene 2, Ab), B, and D.
Each group departs from the conceptual model proposed above in various ways.
The versions of act 1, scene 2 have one less pitch (and, of course, one less interval)
than the model and a different rhythmic profile. Those of act 1, scene 5 have a dif-
ferent rhythmic profile. In act 2, scene 3, the versions of the motive are quite close

45. This syntactic patter n is analogous to the Beg inningMiddleEnd paradigm discussed by Kofi
Agawu, but here it is expressed purely in ter ms of categor ies, rather than in ter ms of tonal organization.
See Victor Kofi Agawu, Playing with Signs: A Semiotic Interpretation of Classical Music (Princeton, N.J.:
Princeton University Press, 1991), chap. 2.
46. Measure numbers are given by scene; page numbers refer to the Dover reprint of the C. F. Peters
1911 edition of the score: Richard Wagner, Tristan und Isolde (New York: Dover Publications, 1973).
52 aspects of cogn it ive st ruc ture

to the model, but the first two have extended endings, while those of act 3, scene
2 have an altered rhythmic profile, so that their overall duration is greatly expanded.
Finally, the versions of the motive in act 3, scene 3 are greatly compressed (as are
those of act 1, scene 2). Within each group, however, the versions of the motive are
relatively consistent (with the exception of those of act 1, scene 5, about which
more in a moment), although the last statement (beginning on D) is always a bit dif-
ferent from the preceding two. This consistency suggests that each g roup of three
statements of the Leidensmotiv forms a syntactic unit, which represents its own
local standard of typicality. These units are themselves members of a superordinate
category (to which we could give some sort of cumbersome name like successions
of the Leidensmotiv beginning on A/Ab, B, and D) whose members span almost the
entirety of Wagners music drama.47
For the most par t, the statements that make up these syntactic units are con-
junct, and, although they may stretch across a number of measures, the connection
between the successive statements is quite obvious. The one exception comes
with the three statements of act 1, scene 5, which occur at a climactic moment in
the drama. Tristan has just dr unk what he believes is a death potion. Isolde wrests
the cup from him, sings Verrther, ich tr ink sie dir! (Betrayer! I dr ink to
you!), and then empties what remains in the cup. Unbeknownst to either, the
draught they have consumed will lead not to death but will instead cause them to
fall even more deeply in love. The melody for Isoldes unintentionally ironical
Ich tr ink sie dir! is the version of the Leidensmotiv that begins example 1.7b.
For the fifteen measures of music that follow, the stage directions read: She
drinks, then throws the cup away. Both [Tristan and Isolde], seized with awe, in
the greatest excitement but motionless, look fixedly into each others eyes, in
which the expression of defiance of death soon gives way to the glow of love.
The cellos then enter with the second statement of the Leidensmotiv, beginning
on B. Further stage directions follow, to accompany another ten measures of
music: Trembling seizes them: they clutch convulsively at their breasts and pass
their hands over their foreheads. Unison cellos and violas then br ing in the last
statement of the Leidensmotiv, beginning on D. In this instance, what draws the
three versions of the motive together into a single categor y is not so much their
similar ity to one another (although this is still quite audible) or their temporal
proximity, but the way Wagner pr ivileges them within his musical discourse. Not
only is there nothing else to compete with them at the moment of their appear-
ance, but also their par ticipation in this pivotal moment suggests that they are
autonomous agents within the drama itself .
We should, if but briefly, consider the notion of syntax developed in the pre-
ceding paragraphs. First, the basic idea of syntax that I employ here is what Websters
calls a connected or orderly system. I flesh out this idea further in chapter 4; for
now, it is enough to say that this appears to be how at least some music theor ists

47. Robert P. Morgan has also pointed out that each successive phrase within these units takes up the
same notes outlined in the top voice of the preceding one, making it seem that each phrase emerges from
the one just before; see Morgan, Circular Form in the Tristan Prelude, 95, 102.
example 1.7 Statements of the Leidensmotiv from Richard Wagners Tristan und Isolde:
(a) act 1, scene 2, mm. 26 33 (Richard Wagner: Smtliche Werke [WSM], vol. 8, part 1,
mm. 313 20); (b) act 1, scene 5, mm. 439 41, 456 58, 468 72 (WSM; vol. 8, part 1,
mm. 1754 56); (c) act 2, scene 3, mm. 260 78 (WSM, vol. 8, part 2, mm. 1890 1908);
(d) act 3, scene 2, mm. 93 107 (WSM, vol. 8, part 3, mm. 1301 15; (e) act 3, scene 3,
mm. 153 57 (WSM, vol. 8, part 3, mm. 1581 85)

26 Eng. Hn.


439 456 Vnc.

Isolde: Ich trink sie dir!

468 Eng. Hn.

Vnc., Vla.

sehr ausdrucksvoll

260 Eng. Hn.




93 Hns., Bns. Hns., Bs. Cl.

ausdrucksvoll dim.
Eng. Hn., Bs. Cl., Hns.

ausdrucksvoll dim.

153 Vnc. Vla. Vn. II 3

poco cresc.
54 aspects of cogn it ive st ruc ture

construe the notion.48 Second, the structural integrity of the syntax associated with
the Leidensmotiv emerges only over time. Although all of its elements are present
from the opening moments of Wagners music drama, confir mation of the syntax
must wait until well into the opera. Third, this syntax is specific to Tristan und Isolde.
Not only would we expect a different motive to be used for another work, but this
motive would almost certainly be deployed in different ways. Musical syntax, at least
with regard to this example, is thus created over time and is idiosyncratic to one
particular composition. Finally, semantic value may become associated with syntac-
tic structure, something evident in Wagners manipulation of the three statements of
the motive in the climactic moment that occurs in act 1, scene 5. By disrupting or
modifying his musical syntax, Wagner can create a type of meaning that is thor-
oughly and exclusively musical.

Syntactic Structure and Compositional Strategy

As a general type, the category statements of the Leidensmotiv has a temporal span
coextensive with that of Tristan und Isolde. More specifically,Wagner carefully artic-
ulates the category over the course of the drama, with the motive appear ing in syn-
tactic units of three successive statements so that, with each entry of the motive on
A (or Ab), we expect to hear two further entr ies, replicating the patter n set in the
opening measures. Wagner relies on this expectation in act 1, scene 5 to underline
both the tragedy and passion of Tristans and Isoldes love. From this perspective, the
further appearances of the Leidensmotiv in the Prelude after the initial three seem
somewhat anomalous, for here they are only in pairs: statements on B and D in
mm. 85 90, and on Ab and B in mm. 100 106. But Wagner is playing a deep, even
subliminal, game: buried beneath the overheated music in mm. 80 83 are three
rhythmically displaced versions of the Leidensmotiv, all beginning on Ab (as shown
in ex. 1.8). Although little more than wraiths appear ing in the midst of a fury of
motives, these statements have their continuation indeed, their consequence in
the entrances on B and D in the measures that follow. This hidden syntactic unit
implies that the next round of statements, which start with the entrances on Ab and
B in mm. 100 106, should be followed by one additional entrance of the Leidens-
motiv on D.Wagner withholds the third statement, however, and instead calls for the
curtain to go up: in response to our anticipation of a third statement of the motive,
Wagner gives us an entire opera.49
The shadowy appearances of the Leidensmotiv in mm. 80 83 of the Prelude
and the somewhat less-typical versions that follow in mm. 85 90 suggest that Wag-

48. Cone, in On Der ivation, pairs the organizational system of syntax with the expressive tools
of rhetor ic. As another example of this sort of approach, Robert P. Morgan, in a recent essay, describes
tonality as compr ising not only the abstract collections of diatonic pitches (plus their chromatic alter-
ations), which have significant symmetr ical properties of their own, but various asymmetr ical syntactical
functions, harmonic and linear, that transfor m these collections into a full-fledged compositional lan-
guage (Morgan, Symmetr ical Form and Common-Practice Tonality, Music Theory Spectrum 20 [1998]:
2, emphasis added).
49. Morgan points out that the har monic syntax for these final statements is also incomplete; see
Circular Form in the Tristan Prelude, 83.
example 1.8 Statements of the Leidensmotiv from Richard Wagners Tristan und Isolde,
Prelude, mm. 80 83
zu 3.

Fl. 3



espress. pi
espress. pi

1. 2.

Fg. 3.




1. 2.




3 3
pi 3 3
3 3

pi 3 3






example 1.8 (continued)
zu 8.

Fl. 3









1. 2.



1. 1. 2. zus.

1. 2.

Ps. pi




3 3
3 3
3 3
3 3

3 3



cate gorization 57

ner is both playing with and relying on our capacity to remember the motive (a
capacity directly related to our ability to categor ize it). But what Wagner wants is
not the sort of recognition that calls attention to itself , distracting us from the flow
of his drama as we become aware of its artifice, but something much more subtle.
In consequence, the motive undergoes continual change, as the different versions
collected in example 1.7 show, the only constant being the opening sixth and
descending half steps. (Contrary to Schoenbergs concept of a motive admittedly
developed with reference to Brahms rhythmic figuration is only weakly linked
to the identity of the Leidensmotiv within the opera.) At the same time, the
repetitions of the motive proper to each syntactic unit which set up local, if
ephemeral, standards for typicality give each unit the illusion of stability and thus
of properly representing the Leidensmotiv.What results is a syntax that is in con-
stant transfor mation.
Further proof of Wagners attention to musical memory can be seen in the rel-
ative atypicality of two of the rare isolated statements of the Leidensmotiv within
Tristan und Isolde. One such statement occurs in the third scene of act 1. Isolde,
recounting to Brangne how her wrath against Tristan was tur ned to pity in her
first encounter with him, sings Er sah mir in die Augen. Seines Elendes jammerte
mich! Das Schwert ich lie es fallen! (He looked into my eyes. His anguish
touched my heart! The sword I let it fall!). As shown in example 1.9a, the mid-
dle phrase is sung to a modified version of the Leidensmotiv. This statement begins,
unlike any other, on G4, and the chromatic passage that leads away from the E5
extends to four descending half-steps. And, relatively early in the long dialogue with
Tristan that makes up scene 2 of act 2, Isolde sings Wie schmerzte tief die Wunde!
(How deeply the wound smarted), again using a less-typical version of the Lei-
densmotiv (as shown in ex. 1.9b). The statement is doubled and or namented by the
strings, and its chromatic descent continues unhindered through six half steps. Both
of these atypical appearances of the Leidensmotiv atypical in their for m relative
to the conceptual model of the motive, and atypical in their isolation function as
remembrances of the motive rather than statements of it.
Carl Dahlhaus, in compar ing the treatment of themes in the first movement of
Anton Br uckners Sixth Symphony to Brahmss method of developing variation,
noted Bruckners use of a similar strategy.When Bruckner returns to a theme, the

example 1.9 Isolated statements of the Leidensmotiv from Richard Wagners Tristan und
Isolde: (a) act 1, scene 3, mm. 135 38 (Richard Wagner: Smtliche Werke [WSM], vol. 8, part
1, mm. 672 75); (b) act 2, scene 2, mm. 341 43 (WSM, vol. 8, part 2, mm. 886 88)


Isolde: Sei nes E len des jam mer te mich


Isolde: wie schmerz te tief die Wun de!

58 aspects of cogn it ive st ruc ture

correspondence with the or iginal version is often only approximate. Dahlhaus

That the one version is able to substitute for the other means, aesthetically, that instead
of developing variation, where each variant represents . . . a consequence of the pre-
ceding one and a prerequisite for the next one, Bruckner makes use of an analytically
elusive but clearly perceivable similar ity by association, which makes the later version
seem like a wr itten-out memor y of the earlier one. The logic of discourse, as con-
ceived by Brahms, gives way to a system of approximate cor respondences.50

The preceding analysis of transfor mations to the Leidensmotiv suggests that a more
apt compar ison would have juxtaposed Bruckner to Wagner rather than to Brahms,
for it is written-out memor ies such as those shown in example 1.9 that con-
tribute much to the power of Wagners music dramas: we respond to such memo-
ries not with our intellect but in a way that seems far more intuitive and thus far
more deeply felt.
Wagner, of course, could have had no explicit knowledge of how humans struc-
ture their understanding through categor ization, since empir ical evidence for this
cognitive process did not begin to accumulate until a hundred years after the 1865
premiere of Tristan und Isolde. And yet Wagner almost assuredly had an implicit
understanding of what Schoenberg called the laws that govern the working of our
minds. Wagners music shows that he knew how to lay out his musical mater ials
to make them come to the fore, often enough and with enough plasticity so that
his listeners could comprehend their meaning. His music also shows that he knew
how to challenge this ability, to modify the structure of categor ies of musical mate-
rials so that these same mater ials seemed new, or to have an entirely different onto-
logical status (changing from a theme into a remembrance of a theme). An account
of the way Wagner transfor med the Leidensmotiv cannot exhaust his inexhaustible
music, but it can give a glimpse into the processes through which we come to under-
stand his music, as well as the way Wagners compositional strateg ies both exploit
and challenge that understanding.

categorization and
conceptualizing music
Schoenbergs theory of motive, shaped by his understanding of the constraints on
musical composition imposed by human cognitive processing (his laws of the
mind), offers insight into how we organize our understanding of music in ter ms of
categories of musical events. Motives were, for Schoenberg, where musical recogni-
tion started, and the coherence of musical motives was what made musical com-
prehension possible. As we have seen, these two functions are replicated in cognitive
categor ies. Our preferred level of categor ization is at a level in the middle of a tax-
onomy the basic level which, like the motive, strikes a compromise between
efficiency and infor mativeness. The categor ies we use most often in everyday life

50. Carl Dahlhaus, Nineteenth-Century Music, trans. J. Bradford Robinson (Berkeley: University of
California Press, 1989), 273.
cate gorization 59

are not defined in ter ms of individually necessary and jointly sufficient conditions.
Instead, they have variable membership: some members are more typical of the cat-
egory than others, and membership in the category is graded rather than an all-or-
nothing affair. Musical motives show a similar amount of variation, and it is through
developing a notion of what they have in common a conceptual model for what
is typical of the category that we begin to comprehend the musical organization
of works like those that fascinated Schoenberg.
As the analysis of Wagners use of the Leidensmotiv shows, Schoenbergs con-
ception of a motive is not adequate for all circumstances (especially in its reliance on
rhythmic figuration), nor are composers always content to establish a single model
for a motive and stick with it. Indeed, they may vary it constantly (as did Wagner in
the case of the Leidensmotiv), or they may combine the models for different motive
forms in any number of ways (as Cones work on the problem of motivic der ivation
shows). Although focusing on motives offers a place to start, it is on the more gen-
eral notion of a musical category free as it is from associations with Beethovens
compositional process, fascinating as that is that I want to focus. While a collec-
tion of assorted motive forms is a good example of a musical categor y, categor ies
can be much more various and structured around whatever set of musical relation-
ships seems best to account for what is salient about a particular repertoire. The rel-
evant units may be harmonic (as in a chaconne), involve a repeated bass patter n (as
in a passacaglia), or combine the two (as in the linear cadential patter ns that Susan
McClary has shown provide a structural framework for music of the sixteenth cen-
tury);51 music of the twentieth centur y suggests an even greater number of possi-
bilities. And music (construed as patter ned sound) may be just one of a number of
contributing factors: one could also envision musical categor ies infor med by the
text being set (of particular importance for liturgical music) or the affect that is to
be summoned (a concer n for some types of dramatic music). In all cases, the choice
and structure of the relevant musical categor ies will be guided by the global con-
ceptual models relative to which the local models for each category are framed.
My proposal, then, is that categories are where our conceptualization of music starts.
If to think is to think in terms of categories, then to think of music is to think in terms
of musical categories.Viewed this way, categories constitute an analogue or perhaps
replacement for the notion of a group as it has been developed by Fred Lerdahl and
Ray Jackendoff. For Lerdahl and Jackendoff, musical understanding begins with the
process of segmenting a musical work into contiguous sequences of pitch-events, drum
beats, or the like.52 Each such segment constitutes a group, and groups are arranged into
a strict hierarchy that ultimately yields the entire work of music.53 Musical categories

51. Susan McClary, Conventional Wisdom: The Content of Musical Form (Berkeley: University of Cal-
ifornia Press, 2000), 14 16.
52. Fred Lerdahl and Ray Jackendoff, A Generative Theory of Tonal Music (Cambridge, Mass.: MIT
Press, 1983); the process of grouping is descr ibed on p. 13; the definition of a group is given on p. 37.
53. By strict hierarchy, I mean a hierarchy in which the boundar y points at each superordinate
level confor m to boundar y points at the level immediately subordinate. Lerdahl and Jackendoff define
this sort of organization as hierarchical; see Lerdahl and Jackendoff, A Generative Theory of Tonal Music,
13, 37. Conceptions of hierarchy as they are applied to music are discussed in greater detail in the latter
part of chap. 7.
60 aspects of cogn it ive st ruc ture

can absorb many of the functions of groups specifically, the role of the group as the
first level of organization for musical phenomena but do not require absolute con-
tiguity or str ict hierarchical organization. As a result, a theory of musical categor ies
can easily account for the multiple strands of polyphonic music in a way that group-
ing theory cannot. Each of the overlapping entr ies of a work in canon, for instance,
could be regarded as a member of a category; this sort of analysis is simply impossi-
ble with grouping theory, since overlaps between groups are not per mitted.
If categories are where our conceptualization of music starts, they must be
closely linked to concepts themselves. At the r isk of some confusion, I would like to
go further and suggest that musical categor ies are concepts. Now, to attempt any
definition of a concept is to (at least potentially) embroil myself in a long-standing
debate within philosophy and cognitive science.54 To skirt this debate, I concentrate
on a definition of concept that reflects recent work in the mind sciences and the
brain sciences, and that can be applied to music. Although I think this definition can
inform the larger debate on concept, I shall leave the substance of the debate for
another time.
The definition of concept I develop here is influenced by the work of Gerald
Edelman, who has been interested in developing a biological approach to con-
sciousness. Edelmans definition of a concept is developed as a descr iption of the
capacities necessar y for the control of complex interactions between an organism
and its environment:
An animal capable of concepts is able to identify a particular thing or action and con-
trol its future behavior on the basis of that identification in a more or less general way.
It must act as if it could make judgments based on recognition of category member-
ship or integrate particulars into universals. This recognition rests not just on per-
ceptual categor ization (although a concept may have a highly sensory content) but, to
some degree, must also be relational. It can connect one perceptual categor ization to an-
other even in the absence of the stimuli that tr iggered these categor izations.55

Thus to have concepts involves not only the process of categor ization but also rec-
ognizing relationships between categor ies.56 What is also important is that having
concepts is a capacity that is not limited to humans, a point also made by Donald
Griffin.57 Concepts are thus not necessar ily tied to language.

54. For a recent and stimulating sally in this debate, see Jerry A. Fodor, Concepts: Where Cognitive Sci-
ence Went Wrong (Oxford: Clarendon Press, 1998).
55. Gerald M. Edelman, The Remembered Present: A Biological Theory of Consciousness (New York:
Basic Books, 1989), 141.
56. Other works that either make a strong connection between categor ies and concepts or assert
their equivalence include James Hampton and Danile Dubois, Psychological Models of Concepts, in
Categories and Concepts: Theoretical Views and Inductive Analysis, ed. Iven van Mechelen, James Hampton,
Ryszard S. Michalski, and Peter Theuns, Cognitive Science Ser ies (London: Academic Press, 1993),
11 33; Barsalou, Flexibility, Structure, and Linguistic Vagary in Concepts; Barsalou et al., Concepts
and Meaning; Smith and Medin, Categories and Concepts; and Gregory L. Murphy and Douglas L.
Medin, The Role of Theories in Conceptual Coherence, Psychological Review 92 (1985): 289 316.
57. Donald R. Griffin, Animal Minds (Chicago: University of Chicago Press, 1992), chap. 6. A simi-
lar approach can be seen in the work of Douglas Hofstadter and his associates, which has resulted in a
number of computer programs that deal with relatively compact bundles of information they call con-
cepts; see Hofstadter and the Fluid Analogies Research Group, Fluid Concepts and Creative Analogies:
Computer Models of the Fundamental Mechanisms of Thought (New York: Basic Books, 1995).
cate gorization 61

With this perspective in mind, let me propose that a musical concept has three
character istics. First, it is a product of the process of categor ization. A musical cat-
egory, then, is quite literally where our conceptualization of music begins. Second,
a musical concept is an essential part of the means through which we guide present
and future actions. These actions thus constitute a sort of indirect evidence for a
cognitive structure almost as ephemeral as music itself. Third, a musical concept can
be related to other concepts, including concepts associated with bodily states (both
physical and emotional), perceptual categor ies (including sound, which, after all, is
not necessar ily music), and linguistic constructs.
From the perspective provided by this definition, language in the sense of nat-
ural language is not required in order to have musical concepts. I have, of course,
used language to character ize musical categor ies, as well as the conceptual models
around which musical categor ies are organized. But we could imagine a listener
perhaps a particularly astute listener, but perhaps one no more skilled than Prousts
M. Swann who develops a musical category without recourse to language.While
language is still nearly indispensable for communicating features of musical concepts
and for developing such concepts in r icher contexts, it is not required.58 The notion
of concepts independent of language is more than a little provocative, and with con-
sequences that extend beyond music and the general debate on concept noted
above. Where some authors have excluded music from the conceptual realm
(because, by their definition, concepts require language),59 music can join dance, the
visual arts, and any of the other nonlinguistic modes of human expression as a prop-
erly conceptual activity.
This perspective on musical understanding also offers a new way to think about
musical syntax. Music theory has traditionally (if not explicitly) character ized syn-
tax in ter ms of relationships between categor ies of musical objects or events. For
instance, a perfect authentic cadence, as it is usually defined, consists (in part) of a
dominant chord followed by a tonic chord, with bass motion from 5 to 1, and with
the soprano concluding an octave above the bass. The definition involves categor ies
like scale step (5 and 1), harmonic function (dominant and tonic), voice (bass and
soprano), and interval (the octave). Members of these categor ies, when placed in
specific relationships (succession and above), produce a component of musical
syntax (the perfect authentic cadence).60 Musical syntax can also be conceived of

58. The idea that language was not required for musical analysis was one of the ideas behind Hans
Kellers wordless analyses. See Keller, Wordless Functional Analysis no. 1: Mozart, K. 421, The Score and
I.M.A. Magazine 22 (1958): 56 64; idem, The Musical Analysis of Music, in Essays on Music, ed.
Christopher Wintle (Cambr idge: Cambridge University Press, 1994), 126 28; idem, Functional Analy-
sis no. 9a: Mozarts Piano Sonata in A Minor, K. 310, in Essays on Music, ed. Christopher Wintle (Cam-
bridge: Cambridge University Press, 1994), 129 38.
59. Ray Jackendoff, Consciousness and the Computational Mind (Cambridge, Mass.: MIT Press, 1987);
Mark DeBellis, Music and Conceptualization (Cambridge: Cambridge University Press, 1995).
60. It is worth noting that Type 2 categor ies are chiefly involved in this definition: within a given
context, something either is or is not an example of 5; something either is or is not a dominant chord;
and so on. Such categor ies are typical of the sorts of taxonomies established by formal music theor y.
However, note that at least one categor y (for voice) is less explicitly defined and functions more like a
Type 1 category. For instance, the soprano is, by convention, the highest sounding voice within a mul-
tivoice texture, but in instrumental works it may actually sound well below the range of human sopra-
nos (and thus be a somewhat atypical example of a soprano voice).
62 aspects of cogn it ive st ruc ture

as an emergent property of sequences of musical events. Wagners treatment of the

Leidensmotiv offers one example of this, but the same can be said of most of the
musical works that interest us: it is through the development of local levels of syn-
tax that musical works become distinctive, and it is the multivalence of this syntax
that makes repeated listening to the same work a rewarding exper ience. Of course,
syntactic structure that is shared between musical works is also important, and in
most music there is a dialogue between the two analogous to the dialogue between
the local and global models relative to which musical categor ies are framed.61 Elab-
orating this approach to musical syntax is the main goal of chapter 4. There we will
see how composers rely on syntactic conventions to provide a framework for the
development of more local patter ns of syntax specific to individual works.
Before we can more fully investigate the role musical concepts play in under-
standing and the way musical syntax comes about, we need to consider two impor-
tant issues. First, there is the issue of how relationships between concepts are estab-
lished. This will be taken up in chapter 2, which explores how conceptual domains
such as music and language are cor related through the process of cross-domain
mapping. This process allows structure from one conceptual domain to be brought
to bear on another and for relationships among concepts to be established. Second,
there is the issue of where conceptual models come from and how they are used to
guide reasoning. Chapter 3 shows how concepts and conceptual relationships are
combined into conceptual models and how conceptual models are connected
together to create theor ies of music.

61. My perspective on musical syntax owes a debt to Eugene Nar mours construal of musical style.
Narmour distinguishes between intraopus style (which is specific to a g iven work) and extraopus style
(which can be seen across a number of works). See Narmour, The Analysis and Cognition of Basic Melodic
Structures; and idem, The Analysis and Cognition of Melodic Complexity: The Implication-Realization Model
(Chicago: University of Chicago Press, 1992).
chapter two
cross-domain mapping

A bout a quarter of the way through the Credo of his Pope Marcellus Mass
(printed 1567), Giovanni Pierluigi da Palestrina indulges in a marvelous bit of
text painting with telling effect. The text Palestr ina sets here is Qui propter nos
homines et propter nostram salutem descendit de clis (Who for us men, and for
our salvation, came down from heaven). As shown in example 2.1, with the first
statement of the word descendit, each voice begins a scalar descent. Christs
descent from heaven is thus represented with a cascading fall through musical space,
a ser ies of overlapping movements down the musical scale.
Palestrinas text painting is a str iking embodiment of the conventional construal
of pitch as high and low. This way of thinking about pitch relations has a ven-
erable tradition. Its basic elements can be seen in Aristoxenuss descr iption of pitch
relations in ter ms of two-dimensional space, which was discussed in the introduction
to this volume. Pitches are understood as points in space, and musical intervals are
reckoned in ter ms of distances between these points. However, the spatial or ientation
crucial to Palestr inas text painting the cor relation of the up and down of
physical space with specific pitch relations, such that a musical scale can descend
was seldom used in Greek music theory. As Andrew Barker has noted, the standard
Greek for what would now be called high-pitched is oxys, which meant sharp,
pointed, or keen-edged; its musical opposite was barys, which meant heavy
(but not, in opposition to oxys, blunt).1 Just when up and down consistently
came to be cor related with musical pitch is unclear, but the linkage was in place at
least by the beginning of the tenth century, a good six hundred years before Pale-
strina wrote his Mass. From around the tenth century, thus, musicians in the West
began writing about and depicting pitch in ter ms of high and low, mapping
structural relations from the domain of vertically or iented, two-dimensional space
onto the domain of music.
Perhaps more remarkable than the long tradition of construing pitch relations in
terms of up and down are the ready reminders of how arbitrary a construal it is.

1. Andrew Barker, ed., Harmonic and Acoustic Theory (vol. 2 of Greek Musical Writings), Cambridge
Readings in the Literature of Music (Cambr idge: Cambridge University Press, 1989), 69, n. 2.

64 aspects of cogn it ive st ruc ture

example 2.1 Giovanni Pierluigi da Palestrina, Credo of the Pope Marcellus Mass,
mm. 53 58


tem de scn dit de c lis.


de scn dit de c lis, de scn dit de c lis.

Tenor I

tem de scn dit de c lis.

Tenor II

de scn dit de c lis. de scn dit de c lis.

Bassus I

de scn dit de c lis.

Bassus II

tem de scn dit de c lis.

Consider, for instance, up and down on the piano: how can D4 be above C4
on the piano when they are both on the same hor izontal plane? Think of playing the
two notes on the cello to play the higher D4, we have to move our left hand
down, so that it is closer to the ground. In fact, there are countless reminders of the
artificiality of describing pitch in ter ms of high and low. And yet Palestr inas text
painting seems anything but arbitrary there seems to be an aptness to his portrayal
of the descent from heaven that goes beyond mere traditions of depiction.
In this chapter, I explore the function of cross-domain mappings of the sort that
underlie this small bit of text painting in the Credo of the Pope Marcellus Mass.
Cross-domain mapping plays two important roles in musical understanding: first, it
provides a way to connect musical concepts with concepts from other domains,
including those associated with language; second, it provides a way to ground our
descriptions of elusive musical phenomena in concepts der ived from everyday
experience. Both of these contr ibute to the establishment of relationships between
concepts, relationships that are fundamental to the prospect of theorizing about
music. In the first section that follows, I provide an introduction to the theor y of
cross-domain mapping as it has been developed in recent work by cognitive lin-
guists and others. This introduction provides a framework for explaining what
makes mappings possible and why some mappings are more effective than others.
Cross-domain mapping also makes it possible to cor relate the musical domain
with others, such as the domain of physical space or of gesture. Under certain cir-
cumstances, such cor relations provide the basis for r ich worlds of the imagination.
c ro s s - doma i n map p i ng 65

Correlating musical pitches with vertically or iented, two-dimensional space, for

instance, leads quite naturally to an imag inary world in which pitches become
things that move through space: the successive notes of a scale g radually descend
and ascend; in other passages, some notes leap, while still others fall. Within this
imaginary world, each traversal of space has a specific and unmistakable sound
that is, descent sounds one way, ascent another. And this is not something limited to
text painting of the sort demonstrated by Palestrina, as any number of cartoon
soundtracks confir m. Never mind that actual traversals of space sound nothing like
those of the hyperkinetic Roadrunner or the hapless Wile E. Coyote; if the cor re-
lation between the domains is properly established, elements from each will blend
together to create novel relationships and elements. In the second section that fol-
lows, I descr ibe the process that leads to this sort of blending and show the role con-
ceptual blending plays in text painting and program music.

an introduction to
cross-domain mapping
Cross-Domain Mapping and Metaphor
The theory of cross-domain mapping is a product of a generalized approach to lin-
guistic metaphor first taken by George Lakoff and Mark Johnson in 1980. Perhaps
the most common conception of metaphor is of a literary device, a manifestation of
the figural use of language to create colorful if imprecise images. Lakoff and John-
son accumulated a substantial body of evidence demonstrating that metaphor was
not simply a manifestation of literary creativity but was, in fact, pervasive in every-
day discourse.2 As an example, consider the way up and down are used to character-
ize emotions, consciousness, and health:

Im feeling up. My spir its rose. Im feeling down. I fell into a depression. My spir-
its sank.
Get up. Im up already. He rises early in the mor ning. He fell asleep.
Hes at the peak of health. Shes in top shape. He came down with the flu.

Each character ization suggests not a literal representation of the spatial domain
implied by the or ientation updown but, instead, uses our knowledge of physical
space to structure our understanding of emotions, consciousness, or health.
Based on evidence provided by a large number of similar examples of the

2. George Lakoff and Mark Johnson, Metaphors We Live By (Chicago: University of Chicago Press,
1980). Their work has recently been extended in Lakoff and Johnson, Philosophy in the Flesh: The Embod-
ied Mind and Its Challenge to Western Thought (New York: Basic Books, 1999).
66 aspects of cogn it ive st ruc ture

appearance of metaphor ical constructions in everyday discourse, Lakoff and John-

son proposed that metaphor was a basic structure of understanding through which
we conceptualize one domain (the target domain, which is typically unfamiliar or
abstract) in ter ms of another (the source domain, which is most often familiar and
concrete). Further study has provided a wealth of empirical evidence for this pro-
posal and contr ibuted to the development of the field of cognitive linguistics.3
Fundamental to the theor y of metaphor that sprang from Lakoff and Johnsons
work is a distinction between conceptual metaphors and linguistic metaphors. A
conceptual metaphor is a cognitive mapping between two different domains; a lin-
guistic metaphor is an expression of such a mapping through language. For instance,
the conceptual metaphor state of being is orientation in vertical space maps
relationships in physical space onto mental and physical states.4 The cross-domain
mapping wrought by this conceptual metaphor then gives rise to innumerable lin-
guistic expressions. Some of these expressions are commonplace, such as Maxwell
seems a bit down today. Others summon a r ich imagistic world, such as that of
John Keatss Ode to a Nightingale:
My heart aches, and a drowsy numbness pains
My sense, as though of hemlock I had drunk,
Or emptied some dull opiate to the drains
One minute past, and Lethe-wards had sunk.

Here the descent to the mythical r iver gives a physical cor relate to the narcotic state
of the nar rator: the act of sinking is mapped onto a melancholy emotional state.
Thus the same conceptual metaphor (state of being is orientation in vertical
space) is behind both linguistic metaphors, one commonplace (Maxwell seems a
bit down today), the other poetic.
With respect to music, the high and low used to descr ibe pitches reflect the
conceptual metaphor pitch relationships are relationships in vertical space.
This metaphor maps spatial or ientations such as updown onto the pitch contin-
uum. The mapping yields a system of metaphors replete with possibilities for
describing musical pitch. We can speak in ter ms of pitch contour (meaning succes-
sions of pitches, which are located at different places in pitch-space), gesture (mean-

3. For a review of the empir ical evidence supporting metaphor as a basic cognitive process, see Ray-
mond W. Gibbs, Jr., The Poetics of Mind: Figurative Thought, Language, and Understanding (Cambridge:
Cambridge University Press, 1994). For discussion of the link between the study of metaphor as a cog-
nitive process and the central concer ns of cognitive linguistics, see George Lakoff, The Invariance
Hypothesis: Is Abstract Reason Based on Image-Schemas? Cognitive Linguistics 1 (1990): 39 51.
4. By convention, conceptual metaphors are represented in capital letters. Thus love is a journey
designates a conceptual metaphor of general application, and Love is a journey a specific linguistic
expression based on that metaphor.
The conceptual metaphor state of being is orientation in vertical space is a variant of the states
are locations conceptual metaphor discussed by George Lakoff and Mark Turner in More Than Cool
Reason: A Field Guide to Poetic Metaphor (Chicago: University of Chicago Press, 1989). On cross-domain
mapping as a general phenomenon, see Lakoff and Turner, More Than Cool Reason, 4; George Lakoff,
The Contemporar y Theory of Metaphor, in Metaphor and Thought, 2nd ed., ed. Andrew Ortony
(Cambridge: Cambridge University Press, 1993), 202 51; Gibbs, The Poetics of Mind; and Gilles Faucon-
nier, Mappings in Thought and Language (Cambridge: Cambridge University Press, 1997).
c ro s s - doma i n map p i ng 67

ing successions of pitches with a specific directionality and contour), and musical
space (meaning a three- or four-dimensional extension of the basic two-dimen-
sional mapping).5 This system is g iven graphic representation in traditional musi-
cal notation: notes that are the result of more rapid vibrations of the sounding
medium are placed higher on the page than notes that result from less rapid vibra-
tions (with the exception of sharps and flats). The two-dimensional space of the
musical page thus cor relates with the spatial or ientation ascr ibed to pitch.6 The
systematic quality that results from mapping spatial or ientations onto the pitch
continuum thus leads to an entire vocabulary for descr ibing relationships among
pitches that provides a r ich set of possibilities for furthering our conceptualization
of music.
As common as conceiving of pitches as high or low seems, not all cultures
describe pitch relationships in purely spatial ter ms. As I noted here, Greek theor ists
of antiquity used oxys (sharp or pointed) and barys (heavy) to character ize
pitches. And traversing histor ical distance is not the only way to discover alter native
conceptualizations of pitch relations. Consider three examples in which it is culture,
rather than time, that creates distance:
1. Steven Felds research has shown that the Kaluli of Papua New Guinea
describe melodic intervals with the same ter ms they use to character ize
features of waterfalls. For instance, in Kaluli sa means waterfall, and a
mogan is a still or lightly swirling waterpool; sa-mogan is the flow of a water-
fall into a level waterpool beneath it. Sa-mogan is also used to descr ibe a
melodic line that descends to a repeated note, the contour of which repli-
cates that of a waterfall flowing into a pool. In contrast, there are no
specific names for ascending intervals, which nonetheless do occur in Kaluli
song.7 Behind this account of musical intervals is the conceptual metaphor
pitch relationships are waterfall characteristics, which provides the
basis for a r ich set of descriptive terms that capture some aspects of melody
but not others.
2. In Bali and Java pitches are conceived not as high and low but as
small and large.8 Here the conceptual metaphor is pitch relation-
ships are relationships of physical size, a mapping that accurately

5. For further discussion of mappings between the spatial and pitch domains, see Arnie Walter Cox,
The Metaphor ic Logic of Musical Motion and Space (Ph.D. diss., University of Oregon, 1999).
6. It is worth noting that the spatial domain of the musical page is itself metaphor ical, since the way
the page is or iented in physical space that is, whether it is propped on a music stand or lying flat on a
table, turned right side up or upside down does not change what we regard as the top and bottom of
the page. Two factors, each independent of the actual or ientation of the page in physical space, help
establish the spatial domain of the page: the conventional or ientation of the symbols on the page (either
rightside up or upside down) and the relative distance of the symbols on the page from the reader.
Rightside up symbols that are farthest from the reader are at the top of the page; rightside up sym-
bols that are nearest the reader are at the bottom of the page.
7. Steven Feld, Flow Like a Waterfall: The Metaphors of Kaluli Musical Theory, Yearbook for Tra-
ditional Music 13 (1981): 30 31.
8. Personal communication from Benjamin Br inner, 8 July 1997. See also Wim van Zanten, The
Tone Mater ial of the Kacapi in Tembang Sunda in West Java, Ethnomusicology 30 (1986): 85.
68 aspects of cogn it ive st ruc ture

reflects the nor ms of acoustic production: small things typically vibrate

more rapidly than large things. This acoustic fact is represented through-
out the numerous parts of the gamelan, the collection of instruments cen-
tral to the musical practice of Bali and Java.
3. The Suy of the Amazon basin do not have an extensive vocabulary for
describing pitch relationships. When they are descr ibed, however, it is in
terms of age: pitches are conceived not as high and low but as young
and old. The conceptual metaphor that guides this mapping is pitch
relationships are age relationships, which accurately reflects the Suys
beliefs that the pitch of the voice becomes deeper with age.9
With each of these conceptions of pitch relationships it is easy to focus on what
they lack rather than what they offer. That is, it is natural (if not quite excusable) to
reckon difference in somewhat chauvinistic terms: that the Kaluli do not have terms to
describe ascending intervals, where we do; that the way pitch relationships are charac-
terized in Bali and Java does not transfer into graphic representations with the same
ease as do high and low. It has to be remembered, however, that mapping high
and low onto music has its own limitations: high and low cannot reflect the sub-
tle play of flowing water, nor do they provide much of an explanation for how acoustic
features correlate with pitch relationships. Such differences show that each mapping
between domains makes some conceptualizations possible, while it disables others.

Image Schema Theory

The variety of conceptual metaphors used to character ize pitch relations leads to
the question of the ultimate g rounding of the process of cross-domain mapping.
Even if we grant that we understand a target domain (such as pitch relationships)
in ter ms of a source domain (such as or ientation in vertical space), how is it that we
understand the source domain in the first place? Mark Johnson answered this ques-
tion by proposing that meaning was grounded in repeated patter ns of bodily expe-
rience.10 These patter ns give rise to what Johnson called image schemata, which pro-
vide the basis for the concepts and relationships essential to metaphor. An image
schema is a dynamic cognitive construct that functions somewhat like the abstract
structure of an image and thereby connects together a vast range of different expe-
riences that manifest this same recurring structure.11 Image schemata are by no
means exclusively visual the idea of an image is simply a way of captur ing the
organization infer red from patter ns in behavior and concept for mation.
As one example of an image schema, consider the verticality schema, which
might be summarized by a diagram of the sort given in figure 2.1.We grasp this struc-
ture repeatedly in thousands of perceptions and activities that we exper ience every
day. Typical of these are the exper iences of perceiving a tree, our felt sense of stand-
ing upr ight, the activity of climbing stairs, forming a mental image of a flagpole, and

9. Anthony Seeger, Why Suy Sing: A Musical Anthropology of an Amazonian People (Cambridge:
Cambridge University Press, 1987), 100 02.
10. Mark Johnson, The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason
(Chicago: University of Chicago Press, 1987).
11. Johnson, The Body in the Mind, 2.
c ro s s - doma i n map p i ng 69

figure 2.1 Diagram of the

verticality schema

watching the level of water rise in the bathtub. The verticality schema is the abstract
structure of the verticality experiences, images, and perceptions. Our concept of
verticality is based on this schema, and this concept is in turn invoked by the various
conceptual metaphors that use vertical space as a source domain through which to
structure such target domains as emotions, consciousness, health, and musical pitch.
By definition, image schemata are preconceptual: they are not concepts, but they
provide the fundamental structure upon which concepts are based. In consequence,
it is important to emphasize that any diagram used to illustrate an image schema is
intended to represent the key structural features and inter nal relations of the
schema; it is not meant to summon a r ich image or mental picture that we some-
how have in mind and use actively to structure our thought. More directly, what-
ever actually occupies our thoughts is not, by definition, an image schema. We can
conceive of image schemata, just as we can conceive of any of a number of non-
conceptual or preconceptual cognitive processes. We can also note general patter ns
in the way concepts are structured, which can be attr ibuted to image schemata.
However, there are, by definition, no image-schema concepts.
The relationship between the verticality schema and our character ization of
musical pitch with reference to the spatial or ientation updown is fairly immediate:
when we make low sounds, our chest resonates; when we make high sounds, our
chest no longer resonates in the same way, and the source of the sound seems
located nearer our head. The up and down of musical pitch thus cor relate with
the spatial up and down the vertical or ientation of our bodies. The verti-
cality schema offers a straightforward way to explain why we character ize musi-
cal pitch in ter ms of high and low even when the actual spatial or ientation of the
means through which we produce pitches either does not reinforce the character-
ization or runs directly counter to it.
At present, the image schema remains largely a theoretical construct.Work across
a variety of fields, however, has made strong arguments for the importance of such
a construct, including that by Leonard Talmy in linguistics, Gerald Edelman in neu-
roscience, David McNeill in psychology, and Raymond W. Gibbs Jr. and Herbert L.
Colston in psycholinguistics.12 Recently, Lawrence Barsalou and his associates have

12. Leonard Talmy, Concept Structuring Systems (vol. 1 of Toward a Cognitive Semantics) (Cambr idge,
Mass.: MIT Press, 2000), 409 70; Gerald M. Edelman, The Remembered Present: A Biological Theory of Con-
sciousness (New York: Basic Books, 1989), chap. 8; David McNeill, Hand and Mind: What Gestures Reveal
about Thought (Chicago: University of Chicago Press, 1992); Raymond W. Gibbs, Jr., and Herbert L. Col-
ston, The Cognitive Psychological Reality of Image Schemas and Their Transformations, Cognitive
Linguistics 6 (1995): 347 78.
70 aspects of cogn it ive st ruc ture

developed a comprehensive theory of mental representation based on perceptual

symbols (which are analogical structures similar to image schemata) and have begun
to provide empir ical work to support the theory.13 Antonio Damasio, working from
a neurophysiological perspective, has made compelling arguments for the impor-
tance of the body to consciousness and thought.14 Although research on image
schemata and similar structures is still preliminary, it is also highly promising and
offers some of the best prospects for solving some of the problems of the relation-
ship between mind and body that have dogged cognitive research throughout this

The Invariance Pr inciple

The theory of image schemata provides a way to explain how conceptual meta-
phors are grounded. It does not, by itself, explain why some conceptual metaphors
seem intuitively better than others. For instance, the conceptual metaphor pitches
are fruit could provide the grounding for such expressions as You must play the
first note more like an apple, the second more like a banana. Although such map-
pings are possible, they are certainly not common. Pitches and fruits just do not
seem to be a good match.
To account for why some metaphor ical mappings are more effective than oth-
ers, George Lakoff and Mark Turner proposed that such mappings are not about the
imposition of the structure of the source domain on the target domain, but are
instead about the establishment of correspondences between the two domains.
These cor respondences are not haphazard, but instead preserve the image-schematic
structure latent in each domain. Lakoff and Turner for malized this perspective with
the Invariance Pr inciple, which Turner states as follows: In metaphor ic mapping,
for those components of the source and target domains deter mined to be involved
in the mapping, preserve the image-schematic structure of the target, and import
as much image-schematic structure from the source as is consistent with that preser-
vation.15 According to the Invariance Pr inciple, mapping the spatial or ientation

13. Lawrence W. Barsalou,Wenchi Yeh, Barbara J. Luka, Karen L. Olseth, Kelly S. Mix, and Ling-Ling
Wu, Concepts and Meaning, in Chicago Linguistics Society 29: Papers from the Parasession on the Corre-
spondence of Conceptual, Semantic, and Grammatical Representations, ed. Kathar ine Beals, Gina Cooke, David
Kathman, Sotaro Kita, Karl-Er ik McCullough, and David Testen (Chicago: University of Chicago,
Chicago Linguistics Society, 1993), 23 61; Lawrence W. Barsalou, Karen Olseth Solomon, and Ling-Ling
Wu,Perceptual Simulation in Conceptual Tasks, in Cultural, Psychological, and Typological Issues in Cog-
nitive Linguistics: Selected Papers of the Bi-Annual ICLA Meeting in Albuquerque, July 1995, ed. Masako K.
Hiraga, Chris Sinha, and Sher man Wilcox (Amsterdam: John Benjamins, 1999), 209 28; Lawrence W.
Barsalou, Perceptual Symbol Systems, Behavioral and Brain Sciences 22 (1999): 577 609.
14. Antonio R. Damasio, The Feeling of What Happens: Body and Emotion in the Making of Conscious-
ness (New York: Harcourt Brace, 1999), chaps. 5 and 6.
15. Mark Turner, Aspects of the Invariance Hypothesis, Cognitive Linguistics 1 (1990): 254. For
additional wr itings on the Invariance Pr inciple (which at first was called the Invariance Hypothesis), see
Lakoff, The Invariance Hypothesis; Mark Turner, An Image-Schematic Constraint on Metaphor, in
Conceptualizations and Mental Processing in Language, ed. Richard A. Geiger and Brygida Rudzka-Ostyn,
Cognitive Linguistics Research, 3 (Berlin: Mouton de Gruyter, 1993), 291 306; and Mark Turner, The
Literary Mind (New York: Oxford University Press, 1996), chap. 3.
A preliminary discussion of a similar sort of topographical invariance, with applications to music, can
c ro s s - doma i n map p i ng 71

up down onto pitch works because of correspondences between the image-schematic

structure of components of the spatial and acoustical domains. Both space and the
frequency spectrum are continua that can be divided into discontinuous elements.
In the spatial domain, division of the continuum results in points; in the acoustic
domain, it results in pitches. Mapping updown onto pitch allows us to import the
concrete relationships through which we understand physical space into the domain
of music and thereby provide a coherent account of relationships between musical
pitches. Mapping various fruits onto musical pitches works less well because fruit
does not (in any ordinary way) constitute a continuum. To employ this mapping is
to highlight instead both the discontinuity among musical pitches and how they are
unlike one another (an emphasis on difference suggested by the idiomatic phrase
like apples and oranges).

Cross-Domain Mapping and Conceptual Models

According to cur rent theory, cross-domain mappings are grounded in repeated pat-
terns of embodied exper ience called image schemata. These schemata provide the
basic structure employed in the mappings: the verticality schema is thus funda-
mental to our understanding of two- or three-dimensional spaces as having direc-
tionality and of musical pitch as high and low. Image schemata also constrain
the possibilities for mapping between two domains, a constraint reflected in the
Invariance Pr inciple. Because the verticality schema can be applied to both the
spatial and the musical domains, we can use our understanding of the for mer to
structure our understanding of the latter. It remains to be explained why one map-
ping would be preferred over another why, for instance, we tend to descr ibe pitch
relations in ter ms of high and low rather than small and large.
On the f ace of it, both of these mappings are equally viable. Both draw on
aspects of our embodied exper ience: on the one hand, countless exper iences with
the seeming or igin of our own voices; on the other hand, countless exper iences
with the sounds g iven out by physical objects in the world around us. Both map-
pings allow us to descr ibe musical pitches as elements within a continuum. And
each mapping can be easily understood from the perspective provided by the
other. For instance, in Camille Saint-Sanss Le Carnaval des animaux, the music for
the elephant is played by the contrabass (large is low), that for the birds by the flute
(small is high).16 Although musicians educated in the West sense the novelty of the
mapping, they can nonetheless understand it perfectly well. In a similar f ashion,
musicians from Bali and Java, when confronted with Western conventions for nota-
tion, have few if any problems translating small and large into high and

be found in Peter Grdenfors, Semantics, Conceptual Spaces, and the Dimensions of Music, in Essays
on the Philosophy of Music, ed.Veikko Rantala, Lewis Rowell, and Eero Tarasti, Acta Philosophica Fennica,
43 (Helsinki: Philosophical Society of Finland, 1988), 9 27.
16. Camille Saint-Sans, Le Carnaval des animaux: Grande fantasie zoologique, ed. Felix Aprahamian
(Zurich: Eulenberg, 1974): Llphant (no. 5), p. 11, Volire (no. 10), pp. 23 27.
17. Personal communication from Sumarsam, 10 April 1998.
72 aspects of cogn it ive st ruc ture

The reason we prefer one mapping over another has to do with the global con-
ceptual models we absorb from culture and that supply crucial support for the pre-
ferred mapping.18 In the West, the descr iption of pitch relations in ter ms of up
and down arose around the same time musicians began to develop ways of notat-
ing polyphonic compositions. These notational systems often relied, either directly
or indirectly, on the physical placement of symbols on the page.19 The attr ibution
of high and low to musical pitches is thus cor related with a system of notation
that permitted both the visualization and the preservation of musical works. In turn,
this notation relied on a global model that made three important assumptions: first,
pitches could be regarded as objects that were independent of the sound source that
produced them; second, graphical symbols could be used to represent these pitch-
objects; and third, the surface on which these symbols appeared was analogous to
physical space. In Bali and Java, the perfor mance of music was associated with the
rich palette of instruments through which the music was effected. Thus the attr i-
bution of small and large to pitches cor related with character istics of the musi-
cal instruments intr insic to musical perfor mance. The conception of musical pitches
as physical objects relies on a global model that does not, at some fundamental level,
disassociate a pitch from the sound source that produces it.20
As noted earlier, the character ization of pitch relations can be infor med by map-
pings other than high and low and small and large. In each case, the basic
mapping relies on embodied knowledge and on the cor relation of the musical
domain with a more concrete domain. The specific mapping chosen within a tra-
dition of discourse about music reflects not so much absolute musical structure as
it does the broader cultural practice within which music and its understanding are
embedded: mappings reflect the conceptual models that are important to culture.
The cross-domain mappings employed by any theory of music are thus more than
simple cur iosities they are actually key to understanding music as a r ich cultural
product that both constructs and is constructed by cultural exper ience.

18. These models and the relationship between global and local models will be discussed in greater
detail in the next chapter.
19. With regard to an indirect reliance on physical placement, it was often a part of the pedagogy
of daseian notation (developed in the ninth century, apparently with an eye toward accommodating the
needs of organum at the fourth) to place high notes higher on the page, even though no such discr im-
ination was required by the notation (since it relied str ictly on a limited repertoire of letter-like sym-
bols to indicate different pitches). See Musica et Scolica enchiriadis una cum aliquibus tratatulis adiunctis, ed.
Hans Schmid, Verffentlichungen der Musikhistor ischen Kommission, Bayerische Akademie der Wis-
senschaften, 3 (Munich: Verlag der Bayerische Akademie der Wissenschaften, 1981); and the Musica
enchiriadis and Scolica enchiriadis, ed. Claude V. Palisca, trans. Raymond Er ickson, Music Theory Transla-
tion Series (New Haven, Conn.: Yale University Press, 1995). These treatises are cur rently thought to
have been wr itten no later than the tenth centur y and possibly as early as the middle of the ninth. For
further discussion of the histor y of this notational convention, see Mar ie-Elisabeth Duchez, La
Reprsentation spatio-verticale du caractre musical grave-aigu et llaboration de la notion de hauteur
de son dans la conscience musicale occidentale, Acta musicologica 51 (1979): 54 73.
20. It should be observed that a disassociation of pitch from sound does not obviate an image-
schematic basis for understanding pitch relations. It only bears witness to a reification of pitch after its
initial conceptualization.
c ro s s - doma i n map p i ng 73

Cross-Domain Mapping, Music, and Music Theory

pale strinas text painting The theory of cross-domain mapping

outlined in the preceding discussion provides a relatively simple way to account for
the effectiveness of Palestr inas text painting. According to this theor y, we use the
verticality schema, the product of countless bodily exper iences, to give physical
space an up and a down analogous to the up and down of our bodies. This
orientation is then mapped on to a metaphor ical musical space made up of pitches
and the relations (or distances) between them. Pitches that result from more rapid
vibrations of the sounding medium are regarded as higher than pitches that result
from less rapid vibrations of the sounding medium. As each of the voices in Pale-
strinas six-part texture takes up the word descendit, it begins to descend
through musical space. The notion of descent summoned by the text is thus given
sounding image by a specific ser ies of musical events.
Of course, we could also hear (through the ears of a Javanese musician) the
pitches growing larger, or hear (through the ears of the Suy) the pitches g rowing
older, or hear (through the ears of a Greek theor ist) the pitches g rowing heavier.
Distinct from these ways of hearing, Palestrinas text painting relies on a conceptual
model that character izes pitches as objects in two-dimensional, vertically or iented
space, in which up and down describe specific pitch relations. This model is
strongly associated with the conventions of musical notation as developed in the
West: as pitches get lower in sound, they are also wr itten lower on the musical
There is an additional reason why Palestrinas text painting is as convincing as it
is, however. The image of descent created by Palestrina relies on more than just the
general notion of descent, which could have been evoked in a variety of ways: by
a downward leap of a single interval; by a short sequence that alter nated descending
thirds with ascending seconds; or through having each voice enter in tur n, begin-
ning with the highest voice and ending with the lowest. The scalar descent chosen
by Palestr ina, however, provides a str iking analogue for the descent of our bodies
through physical space (when this descent is unaided by artificial means). Such a
descent involves a lessening of potential energy and a continuous action in one
direction, articulated by the regular transfer of weight from one leg to another.21

21. Most of us find descent especially scalar descent well represented by the thought of walking
down a staircase, but I think it could be argued that walking down a hillside works as well. The reason
is that it is not so much the neat, two-dimensional image of stairs that is operative but the regular trans-
fer of weight from one leg to another. Indirect evidence is provided by a striking anecdote related by
John Hockenberry, from a time when he was a reporter for National Public Radio. In order to get to a
group of Kurdish refugees on a remote edge of Iraqi Kurdistan dur ing the after math of the 1991 Gulf
War, Hockenberry, a paraplegic since 1976, had to abandon his wheelchair temporar ily and r ide on a
donkey. He comments on the effect of becoming reacquainted with a non-wheeled mode of trans-
portation through the rhythm of the donkeys gait: It was walking, that feeling of groping and climb-
ing and floating on stilts that I had not felt for fifteen years. I had long ago grown to love my own wheels
and their special physical grace, and so this clumsy leg walk was not something I missed until the sensa-
tion came rushing back through my body from the shoulders of a donkey. Hockenberry, Moving Vio-
lations: War Zones, Wheelchairs, and Declarations of Independence (New York: Hyperion, 1995), 2 3.
74 aspects of cogn it ive st ruc ture

The act of singing a descending scale cor relates well with the basic structure of this
event: the relaxation many singers feel as they sing a descending scale matches the
lessening of potential energy; the temporary pauses on each note of the scale match
the regular transfer of weight, which articulates a physical descent. The text paint-
ing of descendit is thus supported by our embodied knowledge of descent, as well
as by the conventions of ascribing high and low to musical pitches.
In its exploitation and enr ichment of a mapping between the domains of phys-
ical space and music, Palestr inas text painting also gives us a glimpse into a process
of meaning construction. Conventionally, text painting is understood to operate
through a crude sort of mimesis: physical descents are represented by musical descents.
By contrast, I have argued that hear ing a succession of musical events as a descent
is an act that is thoroughly mediated.22 What mimesis there is is highly conditioned
by the choice of cross-domain mappings through which discourse about music is
structured; in turn, these mappings reflect the global models of a given cultural per-
spective and histor ical moment. Palestr inas text painting is not just woven into this
web of meaning construction; it also spins its own threads. Some result from specific
features of physical descents summoned by the passage, some from the point within
the larger musical and dramatic discourse at which this str iking moment occurs, and
some from the sonic attr ibutes that get mapped back onto the notion of physical
descent. The meaning constructed is not, in the final analysis, simple or direct but
multivalent and contingent, and it reflects the r ich set of correspondences activated
by mapping between the two domains.

cross-domain mapping and music theory Because it provides

a way to br ing an integrated system of terms and structural relations to bear on
problems of musical understanding, cross-domain mapping plays an important part
in theor ies of music. Indeed, every theory of musical organization employs cross-
domain mappings of one sort or another. Often, the appeal of such mappings is
strong, and the mappings seem intuitively correct (much as high and low seem
intuitively cor rect for the character ization of pitch relations). Further investigation,
however, reveals that these mappings are every bit as mediated as those that are less
systematically developed.
Rudolf Louis and Ludwig Thuilles character ization of tonality in their Har-
monielehre provides a case in point. The relevant passage, which appears at the begin-
ning of the first chapter, runs as follows:
The unity within the diversity of all har mony is ensured through the law of tonal-
ity. This asserts that any succession of harmonies can become musically understand-
able only when each (independently appear ing) chord is perceived to be in a specific
relationship of dependence with respect to the principal chord that underlies the entire
harmonic context. Both melodic and har monic succession require a center, a point of
repose, around which everything twists and tur ns. For melodic relationships, this sta-
tionary middle point is the tonic note, the Tone, whose central place in the scheme of
the scale is made manifest by the melodic motions that we direct away from tonic and

22. This same argument, mutatis mutandis, applies to McClarys analysis of Robert Johnsons blues
mentioned above in the preface.
c ro s s - doma i n map p i ng 75

then back toward it again. The same role of a fixed middle point and resting place that,
within melodic unity, falls to the tonic note, is played, with respect to har mony, by the
tonic chord: the consonant triad that has the tonic note as its root.23

The view set forth by this character ization is that musical organization is analogous
to that of the physical world. Music consequently obeys laws that are independent
of human beings tonality is, as it were, a natural force.24 Musical understand-
ing relies on apprehending the har monic dependencies that reflect these immutable
laws. The focal point of these dependencies is the center of the system either the
tonic note or the tonic chord which acts as both a center of gravity and an axis of
symmetry.25 To support this perspective, Louis and Thuille appeal to the empir ical
evidence provided by actual melodic and har monic progressions, which they be-
lieve show that every properly constructed melody tur ns around a single tonic note,
and that every properly constructed har monic progression tur ns around a single
tonic chord.
Louis and Thuilles character ization of tonality does a marvelous job of captur-
ing the compelling coherence wrought by well-composed tonal music. Given the
evidence of ethnographic and histor ical studies of musical cultures, however, the
idea that Western European tonality is a naturally occur ring force seems doubtful,
as would be the notion of a science aimed at discovering the laws behind this force.
Also implicit in Louis and Thuilles theory is the concept that tonality exists apart
from musical syntax: successions of musical events do not g ive rise to the law of
tonality but serve only to provide evidence of its existence. According to this per-
spective, successions of musical events that do not provide evidence of tonal rela-
tions cannot be understood as music. Of course, such a view places rather profound
restrictions on what counts as music.
Theorizing about music requires that we bring order, even if of a tenuous sort,
to an ephemeral and often intang ible domain. Cross-domain mapping aids this
process by bringing systems of relationships to bear on the musical domain and by

23. Die Einheit in der Mannigfaltigkeit aller Har monie wird gewhrleistet durch das Gesetz der
Tonalitt. Dieses sagt aus, da irgendwelche Folgen von Har monien nur dann musikalisch verstndlich
wirken knnen, wenn jeder (selbstndig auftretende) Accord in einem bestimmten Verhltnis der Ab-
hngigkeit von einem dem ganzen har monischen Zusammenhang zug runde liegenden Hauptaccord
aufgefat wird.Wie die melodische Folge, so bedarf auch die Accordfolge eines Centrums, eines ruhen-
den Pols, um den sich alles dreht und wendet. Fr die melodischen Beziehungen ist dieser feststehende
Mittlepunct die Tonica, der Ton, dessen centrale Stellung im Schema der Tonleiter dadurch zum Aus-
druck gebracht wird, da wir deren melodische Bewegung von ihm ausgehen und zu ihm wieder
zurckkehren lassen. Dieselbe Rolle eines fixen Mittle- und Ruhepuncts, die innerhalb der melodischen
Einheit der Tonica zufllt, spielt in harmonischer Hinsicht der Tonica-Accord: der consonierende Dreiklang, der
Tonica zum Grundton hat. Rudolf Louis and Ludwig Thuille, Harmonielehre, 7th ed. (Stuttgart: Carl
Grninger Nachf. Ernst Klett, 1920), 7.
24. I should point out that Louis and Thuille are not concer ned with tonality as it might be broadly
defined, but with tonality as it is represented in Wester n European music of the eighteenth and nine-
teenth centur ies. This is made all the more clear by their character ization of tonality in har monic ter ms,
something that has limited applicability to a fairly wide range of music cur rently recognized as tonal.
25. In the pages that follow this excer pt, Louis and Thuille explore in greater detail the idea that
tonic is at the center of a symmetr ical scheme of pitches and har monies. See Louis and Thuille, Har-
monielehre, 8 10.
76 aspects of cogn it ive st ruc ture

correlating r ich, if at times complicated, images with essential musical concepts.

Such mappings are a way to both structure our understanding and extend it
indeed, much work in music theory is devoted to explor ing the ramifications of
these mappings. Nonetheless, each mapping reflects the cultural values and imper-
atives relative to which it is framed, not the verities of absolute musical structure.
Tonality is not, in a simple way, a centr ic and symmetr ical system, any more than it
is a reflection of logic (as Hugo Riemann would have it), psychological energetics
(as Ernst Kurth would have it), or gravity (as Arnold Schoenberg would have it).26
Instead, tonality is a way of understanding musical organization. The cross-domain
mappings through which notions such as tonality are articulated are not simply
essential to theor ies of music; in fact, they constitute what counts as music in the
first place.

Cross-domain mapping is a general cognitive process through which we structure
an unfamiliar or abstract domain in ter ms of one more familiar or concrete. Cross-
domain mapping plays two important roles in musical understanding. First, it pro-
vides a way to connect musical concepts with concepts from other domains. As we
saw here, pitch relations within the domain of music have been connected with
concepts associated with vertical space, waterfalls, physical size, and human aging.
Each such mapping made possible systematic accounts of the ways pitches related to
one another. Second, cross-domain mapping allows us to g round our descr iptions
of elusive musical phenomena in concepts der ived from everyday exper ience, since
the structural relations basic to cross-domain mapping have their source in repeated
patterns of bodily exper ience that is, in image schemata.
As we have seen, the mappings we use to structure our discourse about music are
not accidental but reflect two constraints. One constraint is the Invariance Pr inci-
ple, which proposes that the best cross-domain mappings are those that preserve as
much of the image-schematic structure of both target and source domains as pos-
sible. The other constraint is provided by the global conceptual models relative to
which cross-domain mappings are framed. Taken together, these constraints suggest
that cross-domain mappings not only provide a way to structure our understand-
ing of music but also shape our ideas about what we include under the rubric
music. For an element to count as musical, it must be able to serve as a target for
the cross-domain mappings that guide our discourse about music.
Because cross-domain mapping offers a way to connect what are often elusive
musical concepts with concepts from more concrete domains, and because these
connections give rise to integrated systems of terms and relations, cross-domain
mapping is essential to our theor izing about music. Indeed, as we shall see in the

26. Hugo Riemann, Musikalische Logik: Hauptzge der physiologischen und psychologischen Begrndung
unseres Musiksystems (Leipzig: C. F. Kahnt, 1874); Ernst Kurth, Die Voraussetzungen der theoretischen Har-
monik und der tonalen Darstellungssysteme (Bern: Akademische Buchhandlungen von Max Dreschel, 1913);
Arnold Schoenberg, Harmonielehre (Leipzig:Verlagseigentum der Universal-Edition, 1911); trans. as The-
ory of Harmony, by Roy E. Carter (Berkeley: University of Califor nia Press, 1978).
c ro s s - doma i n map p i ng 77

next chapter, the concepts der ived from processes of categor ization and the relations
created by cross-domain mapping provide the basic mater ials for music theory.

conceptual blending
Palestrinas text painting relies on our understanding of pitches as high and low
and on our exper iences with physical descents down staircases, slopes, and hillsides
to create a vivid aural representation of the text. However, the imaginary world
summoned by Palestr ina also extends beyond the immediate bounds of text and
music. Using the basic cor relation between text and music as a point of departure,
we can enter an imag inary domain in which the pitches to which descendit is
sung become objects descending through musical space.Within this domain, every
physical descent is accompanied by sounds that result from a smooth transition from
very rapid to less rapid vibrations of the sounding medium. This extension of
Palestr inas imaginarium cannot be predicted simply from linking the domain of
text with the domain of music. It results instead from blending elements and events
from these two domains to create a new one with its own structures and relations,
a domain populated by such things as pitch-objects and the sound of descent.
It must be admitted that an imaginary domain populated with pitch-objects mov-
ing through musical space is a somewhat rarefied one. Even so, the process of con-
ceptual blending through which it comes about is itself exceedingly common. For
instance, concepts about humans and concepts about animals are often brought
together in childrens stor ies to produce talking animals. Such creatures are powerful
devices in storytelling, for they offer nar rative possibilities beyond those offered by
characters with only human or animal attr ibutes. In a like fashion, the combination of
musical concepts with those from other domains creates possibilities for meaning con-
struction that reach far beyond those of music alone. As an introduction to a method-
ology for explor ing such conceptual blends in greater detail, let us turn to one of the
talking animals from recent literature and discover what he can tell us about the
process behind the imaginary world summoned by Palestr inas text painting.

Talking Animals and

Conceptual Integration Networks
The Old Grey Donkey, Eeyore, stood by himself in a thistly cor ner of the forest, his
front feet well apart, his head on one side, and thought about things. Sometimes he
thought sadly to himself , Why? and sometimes he thought, Wherefore? and
sometimes he thought, Inasmuch as which? and sometimes he didnt quite know
what he was thinking about. So when Winnie-the-Pooh came stumping along, Eeyore
was very glad to be able to stop thinking for a while, in order to say How do you
do? in a gloomy manner to him.27

Thus A. A. Milne introduces a new character into the stor ies he crafted for his son,
in this case a character built around a stuffed toy donkey Christopher Robin had

27. A. A. Milne, The World of Pooh: The Complete Winnie-the-Pooh and the House at Pooh Corner, illus-
trated by E. H. Shepard (New York: E. P. Dutton, 1957), 45 46.
78 aspects of cogn it ive st ruc ture

received as a Chr istmas present. Eeyore is, unquestionably, a donkey, fond of thistles
and content to stay outside in all sorts of weather. But Eeyore is also endowed with
human character istics: he is able to talk, to be perversely gloomy, to work his
own convoluted chains of logic, and, on occasion, to be capable of exquisite irony.
In fact, Eeyore is a blend of some of the concepts associated with donkeys and with
In order to study conceptual blends such as that represented by Eeyore, the
rhetor ician Mark Turner and the linguist Gilles Fauconnier developed the notion
of conceptual integration networks (CINs).28 Each CIN consists of at least four cir-
cumscr ibed and transitor y domains called mental spaces. Mental spaces temporar ily
recruit structure from more-gener ic conceptual domains in response to immediate
circumstances and are constantly modified as our thought unfolds.29 For instance,
Milnes sketch of Eeyore sets up two closely related mental spaces. The first is that
of the Old Grey Donkey solitary, graceless, and phlegmatic. The second is that of
a somewhat morose and plodding intellect, tangled in its own thoughts and happy
enough to leave them behind at the first opportunity for social interaction. Aspects
of these two spaces are combined in a third space, producing the character of
Eeyore. Turner and Fauconnier use CINs to for malize the relationships between
the mental spaces involved in a conceptual blend, to specify what aspects of the
input spaces are imported into the blend, and to descr ibe the emergent structure
that results from the process of conceptual blending.
The CIN for the conceptual blend used by Milne is diagrammed in figure 2.2.
Its network involves four interconnected mental spaces, which are shown as circles.
Central to the network are two cor related input spaces, the donkey space and the
human space. The solid double-headed ar row linking these two spaces indicates

28. Research on conceptual blending and conceptual integ ration networks is discussed in Gilles
Fauconnier and Mark Turner,Conceptual Projection and Middle Spaces, UCSD Department of Cog-
nitive Science Technical Report 9401 (San Diego: University of Califor nia, San Diego, 1994); idem,
Blending as a Central Process of Grammar, in Conceptual Structure, Discourse, and Language [based on
papers presented at the First Conceptual Structure, Discourse and Language Conference, San Diego,
1995], ed. Adele E. Goldberg (Stanford, Calif.: Center for the Study of Language and Infor mation, 1996),
113 30; idem, Principles of Conceptual Integration, in Discourse and Cognition: Bridging the Gap: [Sec-
ond] Conceptual Structure, Discourse, and Language Conference, Buffalo, 1996, ed. Jean-Pier re Koenig (Stan-
ford, Calif.: Center for the Study of Language and Infor mation, 1998), 269 83; idem, Conceptual
Integration Networks, Cognitive Science 22 (1998): 133 87; Mark Turner and Gilles Fauconnier, Con-
ceptual Integration and Formal Expression, Metaphor and Symbolic Activity 10 (1995): 183 204; idem,
Conceptual Integration in Counterfactuals, in Conceptual Structure, Discourse, and Language, II, ed. Jean-
Pierre Koenig (Stanford: Center for the Study of Language and Infor mation, 1998), 285 96; Turner, The
Literary Mind; Fauconnier, Mappings in Thought and Language; Seana Coulson, Semantic Leaps: Frame-Shift-
ing and Conceptual Blending in Meaning Construction (Cambridge: Cambridge University Press, 2001),
115 202. The most comprehensive study of conceptual blending as of this writing is Gilles Fauconnier
and Mark Turner, The Way We Think: Conceptual Blending and the Minds Hidden Complexities (New
York: Basic Books, 2002).
29. The theor y of mental spaces is developed in Gilles Fauconnier, Mental Spaces: Aspects of Mean-
ing Construction in Natural Language, 2nd ed., with a foreword by George Lakoff and Eve Sweetser (Cam-
bridge: Cambridge University Press, 1994); Fauconnier, Mappings in Thought and Language; Gilles Fau-
connier and Eve Sweetser, eds., Spaces, Worlds, and Grammar (Chicago: University of Chicago Press,
1996); see also Turner, The Literary Mind.
c ro s s - doma i n map p i ng 79

F 3,''-.2$)"#>"".$

G:?6#*7?%&" G:?6#*7?%&"*

I52+,#1)+E0 &-,G0H#1)+E0
F* '"(5*.$ F*/*.@"&
F* 2+**3&$/-5'*5-#-*. F*5+*>A3*9-.2$
F "3)(,B"5$/-5B*3:*(#$ F*",#5$#%-5#+"5?$5#,&5$
,5$.,#1(,+$5#,#"$ *1#5-/"$-.$,++$5*(#5$
F* B,.$#%-.@$,./$#,+@ *:$>",#%"($

F* #,+@-.2$/*.@"&$ H9":)")*7?%&"*
F 2+**3&?$'+*//-.2$
F* B+135&?$-++A,#A",5"$

figure 2.2 Conceptual integration network (CIN) for the anthropomorphic blend used
for A. A. Milnes Eeyore

that elements within them serve as structural cor relates: person is correlated with
donkey, and gloomy disposition is correlated with slow-moving. Guiding the process of
mapping between these domains is the generic space, which defines the core cross-
space mapping and basic topography for the CIN.30 Throughout this network,
beings are mapped onto beings, and character traits onto character traits. Guided by
the conceptual framework provided by the gener ic space, structure from each of the
input spaces is projected into the fourth space, called the blend, which results in the
anthropomorphic character of Eeyore. The mapping is only partial, however,
reflecting the limitations imposed by the gener ic space. Since the gener ic space
does not map between physical character istics, we do not expect Eeyore to be of
human appearance or to be human-sized.31
The dashed ar rows linking the gener ic space to the input spaces, and the input
spaces to the blended space, indicate the directions in which structure is projected:

30. The topography of a domain consists of the elements that populate the domain and the relations
that hold between them. By definition, all of the domains in a CIN must have a unifor m topography,
which is represented in its abstract for m by the gener ic space. The requirement of uniform topography
within the cor relational structure of a CIN can be seen as another manifestation of the Invariance Pr in-
ciple discussed earlier in this chapter.
31. Inasmuch as a number of the characters in Milnes stor ies or iginated as stuffed toys, they are easy
to imagine as rather diminutive in size; in fact, this is how they are rendered in E. H. Shepards illustra-
tions. In his text, Milne is relatively silent on this matter, although various contextual clues suggest that
the animals with which he deals are of the stuffed variety.
80 aspects of cogn it ive st ruc ture

from the gener ic space to the input spaces, and from the input spaces to the blended
space. The arrows are double-headed because, under certain circumstances, struc-
ture may also be projected from the blended space back into the input spaces, and
from the input spaces back into the gener ic space. The idea of anthropomorphic
animals that emerges in the blended space may thus influence the way we think
about regular animals, leading us to talk to them and act toward them as though
they had human character istics. Similarly, our exper iences with actual animals and
people give life to the abstractions of the gener ic space. The double-headed ar rows
also serve as a reminder of the limitation of all of the diagrams of CINs I shall use:
mental spaces are dynamic structures, as are the CINs that are built from them. Thus
figure 2.2 represents a sort of analytical snapshot of this particular network, framed
with the intent of capturing its essential features, but making no claim to exhaust-
ing the possibilities for descr iption. Hints about how the CIN and its spaces may
develop can be gleaned from the diagram, but a full account would require a ser ies
of such snapshots.
Although the diag ram given in figure 2.2 is standard in the literature on men-
tal spaces, it can lead to two misunderstandings. First, since the blend is at the bot-
tom of the diagram, it gives the impression that concepts precipitate down into
the blend. Second, the function of the gener ic space can be a bit confusing, since
it does not seem to be directly involved in the blend. In the interests of clarify-
ing these points I offer figure 2.3, which represents the essential components of
a four-space CIN in a slightly different for mat. Here the gener ic space is properly
represented as both the backg round and the foundation for the entire network.
The two input spaces are concrete representations of the abstract str ucture rep-
resented by the gener ic space, and the conceptual blend is a fur ther projection
from these.
An important aspect of the topography of CINs is the basic logic established by
the gener ic space. For the CIN of figure 2.2, the assumption is that beings visibly
manifest their character traits. Old donkeys are slow-moving and balky; morose
humans are given to gloomy pronouncements on the state of the world; and Eeyore
talks and acts like the clumsy, gloomy character he is. The topography of the net-
work also guides three operations composition, completion, and elaboration
that produce emergent structure unique to the blend.
Composition puts together elements from the input spaces to create new entities
in the blended space, yielding the character of Eeyore: although donkeys cannot
actually think and talk, and humans do not have four legs or hooves, Eeyore has all
these traits. Completion extends the image suggested by the initial mapping from the
input spaces, drawing on our backg round knowledge of the circumstances sum-
moned by the CIN. For instance, we know that gloomy characters expect the worst
of situations. Eeyore can thus be relied on for solemn pronouncements of impend-
ing disaster on even the sunniest of days, to greet each calamity as confir mation of
his estimation of the world, and to eat his thistles with little sign of relish. Elabora-
tion is a more extensive operation than completion is; it develops the structure of the
blended space by building on the pr inciples and logic evinced by the blend. In
effect, the input spaces decrease in importance and the focus is directed toward the
rich imaginary possibilities of the blended space. At this point, we can start to wr ite
c ro s s - doma i n map p i ng 81




figure 2.3 Alternative representation of a four-space CIN

our own stor ies about the Old Grey Donkey.Were we to place Eeyore aloft in an
airplane, we could be sure he would profess no enjoyment of the view but would
instead focus on the various aeronautical catastrophes that, from his perspective,
were almost certainly imminent. Given Eeyores clumsiness and the awkwardness of
his hooves, we would be much less likely to imagine him flying the plane, leaving
this to the more dexterous animals or to Chr istopher Robin.
Although the conceptual blend that yields anthropomorphic animals is impor-
tant to Milnes story, it is not the only blend in evidence. Because Eeyore originated
as a stuffed animal, he retains some of the character istics of the species. Accordingly,
in the story in which he is introduced, he loses his tail only to have it recovered by
Winnie-the-Pooh and nailed back on by Christopher Robin, a sequence of events
unlikely were he based solely on real donkeys. This points to yet another concep-
tual blend, in which one of the input spaces is structured around stuffed, rather than
real, animals (and one that I shall not pursue in the present discussion).
A. A. Milnes Eeyore and most of the other characters from the Winnie-the-Pooh
stories provide clear examples of conceptual blending. As indicated, however, the
process of blending is not restricted to childrens stor ies. On the one hand, concep-
tual blending is also common in everyday discourse. Witness statements such as
The car is being stubbor n today I just cant get her to start, which blends con-
cepts related to the physical properties of inanimate objects and those related to the
behavior of humans. On the other hand, conceptual blending also has its place in
literature. Consider Prousts descr iption of one feature of the spr ingtime walks along
the Msglise way in Combray:
We would leave town by the road which ran along the white fence of M. Swanns
park. Before reaching it we would be met on our way by the scent of his lilac-trees,
come out to welcome strangers. From amid the fresh little green hearts of their foliage
they raised inquisitively over the fence of the park their plumes of white or mauve
82 aspects of cogn it ive st ruc ture

blossoms, which glowed, even in the shade, with the sunlight in which they had

Here Proust blends features of the lilac trees with actions of which only animals or
humans are capable, thus creating a memorable image that perfectly captures the
intense sensual engagement prompted by spring.
Conceptual blending is a pervasive and often transparent cognitive process. A
given situation or story may involve any number of conceptual blends, as the char-
acter of Eeyore constructed from the attr ibutes of both stuffed and real animals
shows. Each such blend can be descr ibed through a conceptual integ ration net-
work, which per mits a systematic descr iption of specific features of the blend. Finally,
the CIN associated with the character of Eeyore can be used to explain why the
Old Grey Donkey has some, but not all, of the features of humans, and why we can
easily imagine him taking part in some activities but not in others.

Conceptual Blends and Text Painting

In the conceptual blends just discussed, the input spaces for each CIN were sum-
moned by language. As Fauconnier has noted, however, mental spaces are very gen-
eral and are constructed for many cognitive purposes; language is but one way
to prompt the constr uction of a mental space.33 Under cer tain circumstances,
music can also prompt space construction. In the case of Palestrinas text painting,
the mental space is relatively circumscr ibed and focuses on an orderly progression
of pitches that lead resolutely toward a cadence. This space is in cor respondence
with that set up by the text, which focuses on Chr ists movement from the heavenly
to the mundane and the lessening of potential energy associated with physical
Correlations between the musical and textual spaces involved in Palestr inas text
painting set up the CIN shown in figure 2.4. The gener ic space for the CIN is struc-
tured around the notion of elements that are in directed relations conceived with
respect to a teleological framework.Within the network, motions through physical
and musical space are directed motions. As with the conceptual blend associated with
the character of Eeyore, the emergent structure of the blended space can be descr ibed
in terms of the operations of composition, completion, and elaboration.
Composition puts musical pitches together with the act of descent to yield
pitch-objects that descend. Composition also associates descent with a specific
sound, which actual descents may or may not produce. Completing this image, we
might infer that the lower the pitch, the closer it is to the ground and to the mun-
dane.We could also infer that, since descent can be given a sound, ascent can also be
given a sound. Palestr ina in fact confir ms the latter inference in mm. 92 93 of the

32. Marcel Proust, Swanns Way (vol. 1 of Remembrance of Things Past), trans. C. K. Scott Moncr ieff
and Terence Kilmartin (New York: Vintage Books, 1981), 147 48. Further blends in Proust are discussed
in Turner, The Literary Mind, 91 92, 120 23; see also my concluding chapter.
33. Personal communication from Gilles Fauconnier, 8 September 1998. This perspective is also rep-
resented, somewhat less explicitly, in Fauconnier and Turner, Conceptual Integration Networks; and
Fauconnier, Mappings in Thought and Language.
c ro s s - doma i n map p i ng 83

figure 2.4 CIN for Palestrinas text painting

Credo, when, as shown in example 2.2, he sets ascendit with steadily r ising
pitches. Elaborating the blend, we might imagine pitch-objects doing all sorts of
things, not just ascending and descending. Something like this is behind Louis and
Thuilles character ization of tonality as a system of forces operating on pitches dis-
tributed throughout the field of musical space.
The extension of the conceptual blend suggested by mm. 92 93 of the Credo

example 2.2 Giovanni Pierluigi da Palestrina, Credo of the Pope Marcellus Mass,
mm. 92 94

Et a scn dit in c lum

Et a scn dit in c lum

Et a scn dit in c lum, se

Et a scn dit in c lum

84 aspects of cogn it ive st ruc ture

notwithstanding, Palestr inas text painting is a relatively isolated instance of mean-

ing construction involving music and language. Indeed, text painting is usually
regarded as a clearly circumscr ibed compositional technique, most often restricted
to a single word or image. Under certain circumstances, however, music and lan-
guage can combine to create rather more extended possibilities for meaning con-
struction, as is shown by an instance of text painting in Giaches de Werts seven-part
madrigal Tirsi morir volea, first printed in 1581.
The text for Werts madrigal is from a poem by Giovanni Battista Guar ini that
is probably the most popular madr igal text of the late sixteenth century. It recounts
a sexual encounter between a shepherd and a nymph.34 The passage relevant to dis-
cussion here appears in example 2.3; the text for the entire poem, with the verses
that appear in example 2.3 underlined, is as follows:

tirsi morir volea thyrsis wished to die

Tirsi mor ir volea, Thyrsis wished to die,
Gli occhi mirando di colei ch adora, gazing at the eyes of his beloved,
Quandella, che di lui non men ardea, when she, whose ardor equaled his,
Li disse: Ahim, ben mio, said: Alas, my love,
Deh non mor ir ancora do not die yet,
Che teco bramo di mor ir anch io. since I long to die with thee.

Fren Tirsi il desio, Thyrsis curbed the desire,

Chebbe di pur sua vita allor finire, which by then had almost ended his life:
E sentea morte, e non potea mor ire; he felt death near, yet could not die;
E mentre il guardo suo fisso tenea and while he kept his gaze
Ne begli occhi divini, fixed upon those eyes divine,
E l nettar amoroso indi bevea, and drank from thence the nectar of love,
La bella Ninfa sua, che gi vicini his pretty Nymph, who felt

Sentea i messi dAmore, Loves heralds near,

Disse con occhi languid e tremanti: said with languishing and trembling looks:
Mori, cor mio, chio moro. Die my heart, for I die.
Cui r ispose il Pastore: At which the Shepherd replied:
Ed io, mia vita, moro. And I, my life, die.
Cos mor iro i fortunati amanti, Thus the happy lovers died
Di morte s soave e s gradita, a death so sweet and pleasant,
Che per anco mor ir tornaro in vita. that in order to die again, they returned to life.

The text painting beg ins in mm. 31 32 when the word tremanti is set to a
written-out or nament (a tr ill or gruppo) in all parts. Tremanti (trembling) sets
up a mental space focused on an intense and partially involuntary physical reaction
to stress that produces repeated oscillating motions. The written-out or nament sets
up a mental space focused on the rapid alter nation of a minor chord with Eb3 in
the lowest voice, and a major chord with FS4 in the highest (which creates an

34. The context for Guar inis poem and Werts setting of it have been discussed by Laura Macy in
Speaking of Sex: Metaphor and Performance in the Italian Madrigal, Journal of Musicology 14 (1996): 1 34.
example 2.3 Giaches de Wert, Tirsi morir volea, mm. 28 39


ci ni Sen tea i mes si da mo re, Dis se con oc chi

ci ni Sen tea i mes si da mo re, Dis se con oc chi

ci ni Sen tea i mes si da mo re, Dis se con oc chi

ci ni Sen tea i mes si da mo re, Dis se con oc chi


Mo ri, cor mio, chio mo ro,

Mo ri, cor mio, chio mo ro,

Mo ri, cor mio, chio mo ro,

lan guid e tre man ti: Cui

lan guid e tre man ti: Cui

lan guid e tre man ti: Cui

lan guid e tre man ti: Cui

example 2.3 (continued)


Mo ri, cor mio, chio mo ro,

Mo ri, cor mio, chio mo ro,

Mo ri, cor mio, chio mo ro,

ris pos il Pa sto re: Ed io, mia vi ta, mo ro, Ed io, mia

ris pos il Pa sto re: Ed io, mia vi ta, mo ro, Ed io, mia

ris pos il Pa sto re: Ed io, mia vi ta, mo ro, Ed io, mia

ris pos il Pa sto re: Ed io, mia vi ta, mo ro, Ed io, mia


Mo ri, cor mio, chio mo ro. Co s mo rir no i

Mo ri, cor mio, chio mo ro. Co s mo rir no i

Mo ri, cor mio, chio mo ro. Co s mo rir no i

vi ta mo ro, Ed io, mia vi ta, mo ro. Co s mo rir no i

vi ta mo ro, Ed io, mia vi ta, mo ro. Co s mo rir no i

vi ta mo ro, Ed io, mia vi ta, mo ro. Co s mo rir no i

vi ta mo ro, Ed io, mia vi ta, mo ro. Co s mo rir no i

c ro s s - doma i n map p i ng 87

F* B%,.2"5$-.$

K0LC#1)+E0# M51*E#1)+E0#
F* J#$"<%:#5K*LJ#$"<495:>KM* F* >(-##".$*1#$#(-++$-.$
-.#".5"$,./$',(#-,++&$ ,++$',(#5$CB(*55$
-.9*+1.#,(&$'%&5-B,+$ ("+,#-*.?$,+#"(.,#-*.$
(",B#-*.$#*$5#("55$ *:$3,D*($,./$3-.*($

F* '-#B%"5$5133*.$
F* 5*1./A*:A#("3)+-.2$

figure 2.5 CIN for mm. 31 32 of Werts Tirsi morir volea

augmented-second cross-relation).35 Similar ities between these spaces suggest cor-

relations between their elements: each chord in the musical space cor relates with an
end-point of the physical motion summoned by the textual space, and the inten-
sity of the physical reaction cor relates with the rapid alter nation of major and minor
chords and the cross-relation between the outer voices.
It is the cor relations between these spaces that set up the CIN shown in figure
2.5. The gener ic space for the CIN is structured around the idea that changes of
disposition represent motion.36 Physical trembling of the sort summoned by tre-
manti is one instantiation of this notion, and the rapid alter nation of musical mate-
rials is another. Once again, the emergent structure of the blend can be descr ibed in
terms of the operations of composition, completion, and elaboration. Composition
combines musical pitches with specific features of trembling, so that the pitches
summon a sense of physical motion even though the sound-source producing them
remains relatively fixed: the singers of Werts madrigal need not actually tremble
when they sing tremanti. Composition also suggests that trembling has a sound,

35. At the time Werts madrigal appeared, terms such as minor chord and major chord were not
conventionally used to descr ibe simultaneous sonor ities. Writers on music dur ing the per iod preferred
to descr ibe sonor ities in ter ms of their intervallic makeup. In the present context, this more accurate ter-
minology is just a bit unwieldy, and I have consequently used somewhat more familiar and concise mod-
ern ter ms. I trust their anachronistic application here will not prove unduly confusing.
36. This is analogous to the phi phenomenon noted by Gestalt psychologists. See Kurt Koffka, Prin-
ciples of Gestalt Psychology (London: Routledge and Kegan Paul, 1935), 179.

88 aspects of cogn it ive st ruc ture

even though most actual trembling is done in relative silence. If we complete the
pattern suggested by the basic mapping, we can infer that once the physical action
stops the sound will stop. Finally, through elaboration we can imagine that succes-
sions of restricted groups of pitch mater ials could depict intense physical actions
more complex than trembling, perhaps even extending to the interaction of physi-
cal bodies.
Although the elaboration of a blended space is often left to our imagination, at
times it may be explicitly developed. This is indeed what Wert does in the measures
immediately following the setting of tremanti. Here are some of the salient fea-
tures of the music of mm.33 38:
Starting with the pickup to m.33, the womens and mens voices engage
in a rapid alter nation of entrances.
The beginnings and endings of the entrances are elided, so that the
singing is seamless.
The spacing between entrances is gradually compressed, culminating in
the joining of mens and womens voices in mm.37 38; in fact, this is the
first time in the madr igal that the two vocal groups sing together.
The setting is almost entirely homophonic within each group of voices.
The har monic mater ial is highly restricted, consisting almost entirely of
triadic sonor ities built on G, D, and F.
In m. 38, the music breaks off, and there is a half measure of silence.
The alter nation of restricted groups of pitch mater ial, the charged interaction of
the vocal groups, the sudden breaking off of the voices in m.38, and the context
provided by the blend produced by the text painting of tremanti all summon an
image of intense physical activity followed by a sudden suspension of that activity. It
is an image that gives sounding representation to the sort of death sexual orgasm
about which the poem obsesses. This image also offers a way to ground the dou-
ble entendre exploited by Guarini, built on the paradoxical connection between life
at its most intense and death. Both sexual climax and death are marked by sharp,
nearly immediate contrasts: orgasm followed by quiescence; the clamor of life fol-
lowed by the silence of death. Werts music, building from the common g round
shared by sexual climax and death, breathes life into a play of words that, by this
point in the sixteenth century, had become a rather common commonplace.37
It is important to emphasize that not all combinations of text and music will
yield rewarding conceptual blends. For a blend to occur, there not only needs to be
some cor relation between two domains, but the domains must also have a unifor m
topography. In some cases certain strophic songs, for instance the cor relation
between text and music is simply too general to generate a compelling blend. In

37. My inter pretation of connections between sexual climax and death is somewhat different from
that offered by Laura Macy. Noting that early physiological accounts saw both sex and death as an emis-
sion of spirit, Macy writes that this link is surely at the bottom of the metaphor ic use of the word death
to mean sexual climax (Speaking of Sex, 5).While I would not want to deny the importance of this
connection, my argument is that commonalities between the larger physiological processes of both
orgasm and death could also serve to ground the metaphor and that Werts music provides a musical ana-
logue for these.
c ro s s - doma i n map p i ng 89

other cases formulaic songs composed for opera or reflecting the more commer-
cial side of popular music the cor relation between text and music may be so ten-
uous as to be virtually nonexistent. Also important are the conceptual models that
inform the inter pretation of the cor relation between the input spaces for a blend. A
careful listener who is nonetheless unaware of the context of Palestr inas music, for
example, will almost certainly realize that something is going on in the passage cited
in example 2.1 but may interpret the music as sad or losing energy. If concep-
tual models associated with the act of descent, or with this particular portion of the
Mass, are not activated, the listener may get no farther than this. Similarly,Werts cli-
mactic moment, for a listener who does not understand the text (either for lack of
knowledge of Italian or out of innocence of the double entendre), may simply be
heard as exciting. In both cases, the conceptual blends will remain latent: potential
but not actualized opportunities for the construction of meaning.
As Werts madrigal suggests, conceptual blends involving text and music may be
relatively extended. Chapter 6 will consider such blends in greater depth, showing
how they develop over the course of entire songs, how the same text set to new
music can give rise to two very different songs, and how music alone can elaborate
a blend first set up by text and music.

Conceptual Blending and Program Music

In theory, conceptual blends may involve mental spaces associated with any domain
of thought. Musical domains could thus be cor related not only with domains asso-
ciated with language but also with physical gesture or color.38 In the following, I
would like to consider the conceptual blends that occur when instrumental music
is associated with an extra-musical program. Although such programs are almost
always invoked through language, the connection between the musical and linguis-
tic domains is much looser, and the concepts that emerge from the blend are some-
what more variable. The generality of such blends also allows them to cover greater
expanses of music in the example with which I am concer ned here, an entire
Ludwig van Beethovens Symphony No. 6 in F major, Op. 68 (Pastorale) is
among the most well-known of instrumental works with which extra-musical pro-
grams have been associated. In the case of the Sixth Symphony, this association was
aided by the descr iptive titles Beethoven gave individual movements (such as

38. The correlations I refer to here are between the conceptual domains associated with music and
with colors or gestures; this fits with my notion of concepts as not necessar ily linguistic. However, I am
not thinking of synaesthesia, where stimuli applied to one of the five senses produces responses in one or
more of the other senses as well; I view this phenomenon as functionally automatic and thus not involv-
ing concepts. Synaesthesia between music and color is common enough to have warranted relatively
extensive research, and no end of commentar y. See, for instance, Lawrence E. Marks, The Unity of the
Senses: Interrelations among the Modalities (New York: Academic Press, 1978); Wilson Lyle, Colour and
Music: An Introduction, Music Review 43 (1982): 261 64; Michael Hurte, Musik, Bild, Bewegung: The-
orie und Praxis auditiv-visueller Konvergenzen (Bonn:Verlag fr systematische Musikwissenschaft, 1982); and
Jonathan W. Bernard, Messiaens Synaesthesia: The Cor respondence between Color and Sound Struc-
ture in His Music, Music Perception 4 (1986): 41 68.
90 aspects of cogn it ive st ruc ture

Scene by the Brook). As Richard Will has pointed out, the programmatic aspects
of the Sixth Symphony created difficulties for those who sought to give an account
of the work, for the simplistic representation of natural sounds seemed in conflict
with the received wisdom that Beethovens music referred to nothing outside
Symptomatic of this uneasiness with Beethovens program was Donald Francis
Toveys outr ight rejection of any link between such natural sounds and the music of
the Sixth Symphony. Tovey writes, In the whole symphony there is not a note of
which the musical value would be altered if cuckoos and nightingales, and country
folk, and thunder and lightning, and the howling and whistling of the wind, were
things that had never been named by any man, either in connection with music or
with anything else.40 With the cor relation between the domain of nature and the
domain of music sundered, Tovey paves the way to approach the work as a perfect
classical symphony and to claim for it a place in the pantheon of Beethovens works.
Nevertheless, Toveys analysis of the symphony seems to speak of something
other than absolute music. Describing an important moment in the slow move-
ment, he writes,
The deep shadow of this remote key of G flat becomes still deeper as C flat, which,
changing enhar monically to B natural, swoops round to our or iginal key B flat. At the
outset of this wonderful passage the theme was that of the first subject with the mur-
mur of the brook becoming articulately melodious in the clar inet and the bassoon.
At the moment when the melody gathers itself up into a sustained phrase and makes
its enhar monic modulation, there comes a phenomenon full of deep meaning. From
this point nothing is left of the melody but sustained notes and bird-song tr ills; the
whole of the rest of the return to the main key is harmonic and rhythmic. In this as
everywhere else the movement remains true to type, a perfect expression of the hap-
piness in relaxation.41

Here we have key areas that swoop, clarinets and bassoons that are articulate, and
melodies capable of independent motion, all within a movement that expresses the
happiness there is to be found in relaxation.
At first glance, Toveys prose seems to retain the very correlation between the
domains of music and nature that he had earlier rejected, and to develop a concep-
tual blend based on this cor relation. Still present are the murmur of the brook, the
bird-song tr ills of melody, and the general atmosphere of the pastoral, all wedded
to Beethovens music. Closer consideration, however, shows that the cor relation of
music and nature cannot explain the structural attr ibutes of the imaginative domain
summoned by Tovey. Although nature includes living beings, music does not; at
best, music compr ises sounds that occur within a temporal framework. The inde-
pendent agency implied by Tovey, evident in the articulate and goal-directed enti-
ties he imagines, also seems somewhat out of step with the commonplace view of

39. Richard Will, Time, Morality, and Humanity in Beethovens Pastoral Symphony, Journal of the
American Musicological Society 50 (1997): 271 329.
40. Donald Francis Tovey, Symphonies (vol. 1 of Essays in Musical Analysis) (London: Oxford Uni-
versity Press, 1935), 45.
41. Tovey, Essays in Musical Analysis, 1: 51. The reference is to mm. 78 85 of the slow movement.
c ro s s - doma i n map p i ng 91

nature as a domain in which events are guided not so much by individual volition
as by larger forces working from without. In short, it is one thing to link elements
of nature and musical sounds to create a domain in which the cuckoo sings with the
timbre of the clar inet and the sound of a brook is summoned by undulating pat-
terns in the str ings. It is rather another to animate musical sounds and to endow
them with their own volition.
As it tur ns out, Toveys inter pretation does not rely on a simple cor relation
between music and nature. Instead, Tovey activates a much larger network of men-
tal spaces to produce his inter pretation. Central to this network is one of the most
common of conceptual blends, that of anthropomorphism (which played an impor-
tant role in the blend that produced A. A. Milnes Eeyore). The basic cor relation that
underlies this blend can be seen in statements like The car is being stubbor n today
I just cant get her to start. In this mapping, the domain of inanimate objects is
structured by the human domain; more specifically, human being is mapped onto car,
and a volitional state proper to humans (stubbornness) is mapped onto the mechani-
cal state of the car (not-starting).42 A blend that exploits this mapping might then
attribute additional volitional or emotive states to the vehicle (I think the car wants
to stay home, Shes mad because I havent changed her oil), or extend into the
nonvehicular domain (I dont think the toaster likes me, it keeps burning the
bread). In its most abstract for m, this blend involves two input spaces, as shown in
the CIN of figure 2.6. One recruits structure from the human domain, the other
from some non-human domain. The cor relation of these two spaces gives r ise to
the blended space of anthropomorphism, whose emergent structure reflects the
counter part correlations of the input spaces. Within this space, non-human entities
become endowed with the character istics and intellects of humans while retaining
many of the features that make them distinct from humans. An anthropomorphic
car may be stubbor n and sullen, but it will also be able to stay outside year round
(even though it may not like to do so).
The gener ic space for the CIN connects entities with entities and states with
states. According to the logic proper to the network, the events that occur within
each source domain have causes that follow from the rules that govern the behavior
of entities that populate the domain. Stubbornness in the human domain reflects per-
sonal idiosyncrasies and may be abetted by exter nal circumstances that render the
current situation unsatisfactory. According to conventional wisdom, people are stub-
born either because they were born that way or because they are unhappy with the
way things are and will not take any action (other than being stubbor n) until things
change. Not-starting in the domain of automobile mechanics is caused by electro-
mechanical conditions that are insufficient to initiate or produce sustained com-
bustion. The car will not start because there is something wrong with the engine. It
is important to note that, although the gener ic space may have a structure that is
strongly image-schematic, generic spaces are often r ich with detail. Generic spaces

42. Note that to ascr ibe a state associated with a living entity to a nonliving entity is not the same
thing as to endow the latter with life. The statement The car wont start I think it died does not
entail a belief in animistic vehicles, but only cor relates a particular state associated with living beings
(being dead, or being no-longer-alive) with the mechanical state of the car (never-to-start-again).
92 aspects of cogn it ive st ruc ture

F 3,''-.2$)"#>"".$

I52+,#O-2+*,# P-,6Q52+,#O-2+*,#
F* 713,.$".#-#-"5$ F !*.A%13,.$
,./$"9".#5$("+,#"/$ ".#-#-"5$,./$"9".#5$
#*$#%"3 ("+,#"/$#*$#%"3$

F !*.A%13,.$".#-#-"5$

figure 2.6 CIN for anthropomorphic blend

are not necessar ily primitive or schematic they only define the basic topography
common to all of the spaces in the blend.
Tovey activates the mental space of anthropomorphism early in his essay when
he writes, The Pastoral Symphony has the enor mous strength of some one who
knows how to relax.43 Within the blend that results from cor relating the human
and musical domains, musical events become actors endowed with their own char-
acteristics and capable of their own acts of volition. Thus the transition that follows
the first subject of the opening movement of the symphony leads in three indolent
strides to a second subject which slowly stretches itself out over tonic and dominant
as a sort of three-part round.44 Tovey develops an anthropomor phic perspective on
nature more gradually, bringing it to fruition only in his account of the end of the
slow movement. Suddenly for a moment all is silent; we have no ears even for
the untir ing brook, and through the silence comes the voice of the nightingale, the
quaint rhythmic pipe of the quail, and the syllabic yet impersonal signal of the
cuckoo.45 Here elements of nature (brooks and birds) become actors, each endowed
with individual character istics and capable of individual acts of volition. In both
cases, the cor relation of human and non-human domains gives rise to imaginative
conceptual blends: symphonies that know how to relax, musical passages that str ide,
brooks that do not tire, and cuckoos capable of impersonal signals.

43. Tovey, Essays in Musical Analysis, 1: 46.

44. Tovey, Essays in Musical Analysis, 1: 47.
45. Tovey, Essays in Musical Analysis, 1: 52.
c ro s s - doma i n map p i ng 93

F* ,.#%(*'*3*('%-B$

R,CQF-)2-F)Q*T0O# R,CQF-)-2-F)Q*T0O#
251*E# ,+C5F0#
F* 315-B,+$".#-#-"5$ F* ".#-#-"5$,./$"9".#5$
,./$"9".#5$>-#%$ :(*3$.,#1("$>-#%$
%13,.$B%,(,B#"(-5#-B5$ %13,.$B%,(,B#"(-5#-B5$

F* 5#*(35$>-#%$

figure 2.7 CIN for Toveys analysis of the Beethoven Pastoral Symphony

Construing both music and nature in anthropomorphic ter ms makes it possible

for Tovey to connect the two in a further blend, shown in figure 2.7. The gener ic
space for this network is the blended space of figure 2.6. One of the input spaces is
that of anthropomorphized nature, which includes natural agents who act within
and respond to their native environment (nature). The other input space is that of
anthropomorphized music, which includes musical agents who act within and
respond to their native environment (music). In the blend, aspects of the input
spaces are combined to create unique entities and relations. Here the murmur of a
brook becomes an articulate melody for clar inets and bassoons; here Beethovens
music can be conceived as natural without having been reduced to nature.
The three operations that produce emergent structure in the blended space
composition, completion, and elaboration are all evident in Toveys account of
the end of the fourth movement: The stor m moves in grand steps to its climax.
This is marked by the entry of the trombones. . . . Then the stor m dies away, until
with the last distant mutter ings of the thunder the oboes give a long slow fragment
of bright sustained melody on the dominant of F. This has been aptly compared
with a rainbow.46 Composition unites the lives and interests of natural agents and
musical agents. In the final twenty measures of the fourth movement, the thunder
mutters with the voices of the timpani and double basses, who also prepare the
tonality of the closing fifth movement precipitated by the retreat of the stor m.

46. Tovey, Essays in Musical Analysis, 1: 55.

94 aspects of cogn it ive st ruc ture

Completion supplies additional structure. Beethovens terse title for the fourth
movement is Gewitter, Sturm (Thunderstorm, Tempest), which says nothing of
the after math of a stor m. Nonetheless, we know that rainbows often follow stor ms,
and so the bright sustained melody of the oboes can summon the image of a rain-
bow, a completion suggested but not necessitated by the input spaces.47 Elaboration
develops the blend by pursuing its logic. The stor m, as a magnificent entity, moves
with grand steps. The climax of its ter ror and wrath (enacted through musical
agents) is marked by the entry of new musical agents (the trombones). As these pas-
sions abate, musical agents g radually disappear (a change only implicit in Toveys
prose) until only a bare few remain. Perhaps most str iking about this elaboration is
that a single natural agent (the stor m) appears to be brought fully to life only through
a multiplicity of musical agents. Such a genesis, while inexplicable in the natural
domain, is fully consonant with the idiosyncrasies of the musical domain. There,
individual musical entities (such as movements or entire symphonies) are often
understood to emerge from the actions of the subentities (such as themes or tonal
areas) that they compr ise.
Tovey thus mobilizes a number of mental spaces to provide an account of
Beethovens symphony. The most active spaces are those that contr ibute to a blend
of anthropomorphized natural and musical elements. This blend gives Tovey a way
to descr ibe music that seems naturalized but does not reduce to a simple cor relation
of natural and musical sounds. Less active, but still important, are spaces involved in
blends fundamental to this pr imary network. These include spaces built up from the
domain of human entities and events, along with the domains of nature and music.
Toveys descr iption makes only spar ing use of these subsidiar y blends, since they
cannot provide the naturalized account of music that is the goal of his essay. Far in
the background is the gener ic space common to anthropomorphic blends, which
defines the core cross-space mapping and logic that underlie the entire network.

Conceptual blending is a dynamic process of meaning construction that involves
small, interconnected conceptual packets called mental spaces, which temporar ily
recruit structure from conceptual domains in response to local conditions. When
blending occurs, a portion of the structure from two cor related input spaces is pro-
jected into a third, blended space. As part of this process, the operations of compo-
sition, completion, and elaboration produce structure within the blend that is not
found within either of the input spaces. This structure only becomes possible
through the concepts and relations produced by conceptual blending.
The str ucture common to the mental spaces within a conceptual integ ration
network is reflected in the gener ic space. This space defines the core cross-space
mapping and is organized according to a basic logic that remains consistent
throughout the network. Generic spaces are not necessar ily pr imitive or image-

47. This suggestion is also followed by the animators of Walt Disneys Fantasia, who not only sup-
ply a rainbow to accompany this moment in Beethovens symphony but integrate it within the multi-
leveled, quasi-mythical story they set to the music.
c ro s s - doma i n map p i ng 95

schematic they only define the basic topography common to all of the spaces in
the blend.
In a basic blend, four spaces will be active (although not necessar ily to the same
degree): the gener ic space, the input spaces, and the blend. It often will be the case,
however, that additional spaces will be activated, to a greater or lesser extent, as the
construction of meaning proceeds. In certain passages within Toveys account of
Beethovens Sixth Symphony, as many as six spaces may be active. In addition to the
blend, these include mental spaces built up from the domains of music and nature,
from the domains of anthropomor phized music and nature, and from the gener ic
domain of anthropomorphism. This multiplicity of spaces explains the complexity
of Toveys account, as well as its inter pretive richness. It also serves to explain the
seeming contradiction entailed by Toveys desire to deny the programmatic aspect
of Beethovens symphony while responding to the powerful impressions to which
it gives rise.
More generally, evidence from conceptual blending lends credence to the idea
that music is an independent conceptual domain. As I have shown in my analyses of
text painting and program music, musical concepts combine with concepts from
other domains to create blended concepts that suggest str iking possibilities for the
imagination. More to the point, musical syntax (from the perspective developed in
chapter 1) can be seen contr ibuting to the blended concepts associated with Werts
text painting and Beethovens program music. In the for mer,Werts deployment of
musical mater ials sets the stage for an enactment both by the voices of the singers
and by our imaginations of the sort of death with which Guar inis poem is con-
cerned. As but one example from the latter, Beethovens conclusion of the fourth
movement and preparation for the fifth (the distant mutter ings of the timpani
notwithstanding) provide a moment of contrast and articulation that clearly sup-
ports our imagining one process ending (the stor m) and another beg inning (the
shepherds song of joyful thanksgiving). And musical spaces can infor m our under-
standing of gener ic spaces: to hear Werts musical climax is to get new insight into
climax as a general phenomenon, be that climax sexual, cerebral, or even visual; to
hear Beethovens transition is to get a glimpse of how natural events like stor ms can
be thought of not simply as personages but also as processes.
Each of these aspects of conceptualizing music will be taken up in more detail in
chapter 6, within the context of the analysis of the nineteenth-century Lied. There
it will be possible to see, to a greater extent, how processes of cross-domain map-
ping and conceptual blending construct and contr ibute to our understanding of
both music and the world as a whole.
chapter three
conceptual models
and theories

A s part of her efforts to understand the everyday knowledge with which people
make sense of music, the music theor ist Jeanne Bamberger has often asked
young children to develop ways to represent familiar melodies or simple rhythmic
patter ns. One way Bamberger does this is to have a child ar range a specially con-
structed set of bells, called Montessor i bells, so that he or she can play a given tune.
Bamberger then asks the child to invent instructions so someone else could play the
tune on the bells as theyve been ar ranged.
For those who have even a passing acquaintance with music, an encounter with
Montessor i bells can be a truly disor ienting exper ience. Developed as part of the
mater ials for the Montessor i curriculum, the bells are designed to encourage chil-
drens exploration of sound and music.With this in mind, they are carefully con-
structed to be identical in appearance: the lowest-sounding bell looks just like the
highest-sounding one, as do all those in between. Only by striking each bell with
a small mallet can the different pitches of the various bells be discovered. Each bell
is attached to a wooden stem mounted on a wooden base so that the bells are phys-
ically independent of one another. There is, in consequence, no set ar rangement of
the bells, just as there are no visual cues as to their pitch: how the bells might be
arranged, whether in casual play or in the perfor mance of some tune, has to be
determined by each person who uses them. As a further challenge (or an additional
resource), a complete set of Montessor i bells contains duplicates: there are eight bells
with white bases, which make up a diatonic scale from C4 to C5, and thirteen bells
with brown bases, which make up a chromatic scale from C4 to CS5. Thus for each
white-based bell there is one brown-based bell with a matching pitch but again,
just which bells match must be discovered by the user.
In the course of her research, Bamberger discovered that children use one of two
general strategies when building a tune with the bells. Children following the first
strategy ar range the bells in the order in which the pitches occur in the tune. To
build the first phrase of Happy Birthday, for instance, the bells might be placed
one after another to create a bell-path of the sort shown in figure 3.1, with the
arrow indicating the order in which the bells are set down. (Note that two G-bells
are used, one for each time the pitch G occurs anew in the tune. That is, a separate

conc e p tual mode l s and th e ori e s 97

I N I O H*
figure 3.1 Bell-path no. 1 for the first phrase of Happy Birthday (open bases, white
bells; solid bases, brown bells; arrows show direction of play)

bell is not provided to accommodate repeated notes in immediate succession. In this

illustration, the second G-bell is taken from the set of brown-based bells.) Playing
this much of the tune is a relatively simple matter of retracing the building process
and str iking each bell in tur n (with the first bell getting two hits). The spatial
arrangement of the bells, then, is in direct cor respondence with the succession of
pitches in the first phrase of the tune.
Children following the second strategy start by putting the bells in order, typi-
cally from low to high. Once this is done, they figure out the sequence in which to
strike the bells in order to play the tune. Figure 3.2 illustrates the results of this strat-
egy for the first phrase of Happy Birthday. The bells are arranged from low to
high (left to r ight); numbers indicate the order in which the bells should be struck,
and arrows indicate the direction of motion. (Note that only one G-bell is required
by this strategy; however, the path through the bells is considerably more complex.)
As figure 3.2 shows, the placement of the bells no longer cor responds to the suc-
cession of pitches that make up the tune. The second strategy does not rely on a
spatial analogue for the succession of pitches in the tune, as does the first. Com-
pensation for this loss in immediacy is provided by a gain in flexibility: to play a dif-
ferent tune, the child needs simply to discover what new succession of hits will pro-
duce the cor rect results the bells themselves do not need to be reordered.
Bamberger calls the representation produced by the first sort of strategy figural
and that produced by the second formal.1 Figural representations highlight the
unique figures (or groupings of musical mater ials) that make the tune what it is.

1. Jeanne Bamberger, The Mind behind the Musical Ear: How Children Develop Musical Intelligence
(Cambridge, Mass.: Harvard University Press, 1991), 24 26; idem, Cognitive Issues in the Development
of Musically Gifted Children, in Conceptions of Giftedness, ed. Robert J. Sternberg and Janet E. David-
son (Cambr idge: Cambridge University Press, 1986), 398 99. In personal cor respondence (20 May
2000), Bamberger notes that for mal strategies, as far as she has seen them among both children and
adults, only occur with people who have had some training in music or who play an instrument.
Bamberger also observes that representation (the ter m I favor here) tends to imply more fully
worked-out cognitive processes than a ter m like description. Although I respect this distinction (espe-
cially where it applies to the activities of young children who may not really be concer ned with rep-
resenting anything), I shall retain representation in my discussion, but with the understanding that it
applies to a quite general set of cognitive structures. Representation, defined thus, embraces image-
schematic constructs, perceptual symbols, concepts, and actions whatever is necessar y to make real-
world phenomena available (to re-present them) to the higher-level cognitive processes associated with
theories. For further discussion of this approach to representation, see Mark Turner, Design for a The-
ory of Meaning, in The Nature and Ontogenesis of Meaning, ed.Willis F. Overton and David S. Palermo
(Hillsdale, N.J.: Erlbaum Associates, 1994), 91 108; and Antonio R. Damasio, The Feeling of What Hap-
pens: Body and Emotion in the Making of Consciousness (New York: Harcourt Brace, 1999), 320.
98 aspects of cogn it ive st ruc ture

I N* H O*
+ ,* P 0*
figure 3.2 Bell-path no. 2 for the first phrase of Happy Birthday

Formal representations, in contrast, tend to obscure the unique character istics of the
tune but situate musical elements within a more general context applicable to any
number of tunes. Bambergers research suggests that figural representations are
among the earliest for med by children. Formal representations develop somewhat
later as the childs sense of how to organize and represent mater ials matures.
Bambergers primary interest in the distinction between figural and for mal rep-
resentations was with the developmental changes that led to the transfor mation of
the one into the other. Nonetheless, the contrast between these two strategies also
points to different ways to construe relationships among musical concepts. That is,
figural strategies offer one way to character ize these relationships, and for mal strate-
gies another. Perhaps more important, both remain options even after the passing of
childhood, as is shown by the transfor mational networks employed by David Lewin
in his analyses of late-nineteenth- and early-twentieth-century music.While Lewins
networks certainly have their abstract side they owe much to theor ies of mapping
that are employed in for mal algebra they also offer a way to represent unique rela-
tionships among musical events: they are figural as well as for mal.2
These different ways to think about the essential relationships that obtain within
what is putatively the same sonic domain (here, the first phrase of Happy Birthday)
demonstrate the role of two basic cognitive structures conceptual models and the-
ories in our understanding of music. Conceptual models, which are the more basic
structures, are made up of relatively limited cor relations of concepts and are specific
to a given domain. (We caught glimpses of such models in the discussion of catego-
rization in chapter 1 and saw the role they played in cross-domain mapping in chap-
ter 2.) Theories, in contrast, are extended structures that coordinate a number of con-
ceptual models to provide a more comprehensive account of the world.3 Figural

2. See David Lewin, Generalized Musical Intervals and Transformations (New Haven, Conn.:Yale Uni-
versity Press, 1987); and idem, Musical Form and Transformation: Four Analytic Essays (New Haven, Conn.:
Yale University Press, 1993). Lewin makes explicit reference to Bambergers work in his analysis of Karl-
heinz Stockhausens Klavierstck III; see Lewin, Musical Form and Transformation, 45 47. I should note that
Bambergers work is by no means limited to early childhood development, and that she, too, takes the
position that figural and for mal strategies are not restricted to developmental stages in childhood. For
instance, in dialogues built into The Mind behind the Musical Ear, she attributes figural strateg ies to one
college-age interlocutor and for mal strategies to another.
3. As is evident from this initial character ization, theory, as I define it, embraces cognitive struc-
tures far more modest than those with which moder n scientific theor y is occupied. The connection
between such cognitive structures and scientific theory is explored in greater detail in the next section.
conc e p tual mode l s and th e ori e s 99

strategies, on the one hand, rely on one set of conceptual models and lead to one
theory of musical organization. Formal strategies, on the other hand, rely on another
set of conceptual models and lead to a somewhat different theory of musical orga-
Both conceptual models and theor ies are central to our understanding of the
world as a whole; however, both often remain implicit. It is only when they are
teased out in the course of work such as Bambergers, or as the result of deliberate
reflection on our thought processes, that either conceptual models or theor ies come
to the fore. In this chapter, I explore the structure and function of conceptual mod-
els and theor ies and show the part they play in both infor mal and for mal accounts
of musical organization. In the first section I examine the role conceptual models
and theor ies play in the musical representations of one of Bambergers subjects, an
eight-year-old boy named Jeff. Jeff s encounters with music engage him in a strug-
gle to come to ter ms with basic musical concepts, and they reveal music theory
that is, a theor y specific to music in the making. In the second section, I show
how the cognitive structures demonstrated by Jeff s nascent theor ies of music con-
nect with similar structures proposed by researchers in a wide variety of fields, from
artificial intelligence and cognitive science to anthropology and ethnomusicology.
This will serve to generalize the construal of conceptual models and theor ies beyond
Jeff s specific situation and lay out some of their additional properties. In the third
section, I turn to two of the more influential theor ies of music within the last three
centur ies of the Western tradition: the music theory of Jean-Philippe Rameau and
that of Heinrich Schenker. My objective is to show how they employ the same
basic cognitive structures that can be seen in childrens figural and for mal strategies
for tune building.

developing a theory of
musical organization
Learning to Play Twinkle on the Bells
Jeanne Bamberger worked with Jeff for about six months in the Childrens Learn-
ing Lab at MIT, starting in the fall of 1975.4 She set him the task of building the
tune Hot Cross Buns, first with the sonic resources provided by a simple com-
puter program and later with the Montessor i bells. After Jeff achieved some facil-
ity with these tasks, Bamberger had Jeff concentrate on the tune Twinkle, Twinkle,
Little Star, using only the Montessor i bells. In consequence, at the point we take up
Jeff s story he has already lear ned some things about tunes and the properties of the
bells, although Twinkle, with its comparatively more complex structure, presents
new challenges. In what follows, the focus is on the first phrase of Twinkle
(Twinkle, twinkle, little star, / How I wonder what you are; the musical notation
is given in ex. 3.1) and on two of the bell-paths Jeff constructed in order to play it.

4. Jeff is a pseudonym adopted by Bamberger to preserve the boys anonymity. An extensive

account of Bambergers work with Jeff is given in chaps. 6 12 of The Mind behind the Musical Ear.
100 aspects of cogn it ive st ruc ture

example 3.1 The first phrase of Twinkle, Twinkle, Little Star

Twin kle, twin kle, lit tle star, How I won der what you are.

In preparation for Jeff s work with Twinkle, Bamberger had selected eleven
bells from the complete Montessor i set and put them out on a table.5 Included were
all of the white-based bells (providing a complete C major scale) and three brown-
based bells: the C-, E-, and G-bells. Once Jeff had familiar ized himself with the
bells, Bamberger proposed that he tr y building Twinkle. After beginning with
the C-bell, Jeff searched among the available bells until he found one that matched
the next pitch of the tune. He placed this bell next to the C-bell, and then played
both to re-create the beginning of the tune (that is, the pitches that fit the words
Twinkle, twinkle). To find the bell for little, he once again searched among the
available bells, and when he found the r ight one he placed it next in line. He again
played what he had assembled of the tune (Twinkle, twinkle, little) before return-
ing to the remaining bells to find the one cor responding to the next pitch of the
Jeff s preference was to use the white-based bells first. He turned to the brown-
based bells only when there was no white-based bell left that would provide the
pitch for which he was looking; this first happens with the recurrence of the note
G on the word star. As a way of giving a visual representation of the division
between the two subphrases (or figures) of Twinkle, Jeff left a space between
the first four bells and the second four bells (marking the first four bells as one
group and the second four bells as another). This produced the ar rangement shown
in figure 3.3. The result is a figural representation of the tune, with each successive
bell cor responding to a successive pitch of Twinkle. Performing the tune is sim-
ply a matter of beginning at one end of the bell-path and proceeding to the other
(but with some bells getting two hits, and others only one).
After Jeff had completed his bell-path, Bamberger asked him to come up with
instructions so someone else could play Twinkle on the bells as he had set them
out. Jeff s initial solution was quite straightforward: he simply drew eight short ver-
tical lines, with each line cor responding to a bell. The cor responding notation
looked something like this:

" " " " " " " "

His intent was that the player should start at the beginning and play straight ahead
to the end (mimicking, incidentally, his process of setting up the bell-path), some-

5. Bambergers account of Jeff s first work with Twinkle can be found in The Mind behind the
Musical Ear, 183 92.
6. The strategy of playing through as much of the tune as he had built before searching for the next
pitch is one Jeff used throughout the tune-building exercises I discuss here.
conc e p tual mode l s and th e ori e s 101

O* I N I* Q R S O*

figure 3.3 Jeff s initial bell-path for the first phrase of Twinkle, Twinkle

thing made manifest when, at Bambergers prompting, he put in numbers to show

the sequence of actions:

" " " " " " " "

1 2 3 4 5 6 7 8

With a bit of additional prompting, Jeff concluded his notation by adding another
set of numbers to show how many times each bell was to be struck:

2 2 2 1 2 2 2 1
" " " " " " " "
1 2 3 4 5 6 7 8

With this, Jeff s first session with Twinkle was complete: he had constructed
a bell-path with which he could play the first phrase of the tune, and he had come
up with a notation that would allow another child to do the same (once the mean-
ing of the numbers and lines had been explained).7

Conceptual Models and Theor ies in Jeff s First

Version of Twinkle
Jeff s actions, as he constructs a bell-path for Twinkle and then invents a notation
for both the path and the actions to be perfor med on it, are based on a number of
The pitch-events of the tune are embodied by the bells that reproduce
them.8 In this way, each pitch-event (which is invisible and ephemeral) is
given visible manifestation by a bell.
Each of the pitch-events of the tune is unique. It follows that there will
be a different bell for each pitch-event; Jeff does not, at this point, recog-
nize the pitch equivalence of the two G-bells or of the two C-bells.
The succession of temporal events that make up the tune can be repre-
sented through the spatial disposition of the bells. Since this succession is

7. It is worth noting that Jeff s notation is tied to his bell-path: were the additional space between
the fourth and fifth bells eliminated, his notation would almost assuredly lead to a distortion of the tem-
poral spacing of the pitches of Twinkle.
8. By pitch-event I mean a pitched sound framed relative to some temporal context (so that it is
an event). As will be seen, because building the tune involves ar ranging the bells in space, the tempo-
ral aspect of pitch-events is almost always conflated with spatial location.
102 aspects of cogn it ive st ruc ture

conceived of as linear, the bell-path is likewise linear; similarly, the pause

between the two subphrases is represented by a greater amount of space
between the fourth and fifth bells.
Playing the tune is a matter of performing actions along a path specially
constructed for this tune. These actions are unidirectional the player
starts at one end of the bell-path and proceeds straight through to the
other and mimic the process of building the path. Playing the tune
constitutes one large, continuous motion, articulated by the smaller
motions of striking each individual bell.
These suppositions, the relationships between them, and the way they help guide
Jeff s actions as he sets up his bell-path can be character ized in ter ms of four sor ts
of cognitive constructs: concepts, conceptual models, conceptual domains, and
Concepts. Musical concepts, as noted in chapter 1, have the following attr i-
butes: they are a product of processes of categorization, they are part of the
means through which we guide present and future action, and they are not
necessar ily linguistic. The concepts basic to Jeff s perspective include (but
are not limited to) those associated with pitched sounds, with the individual
Montessor i bells, with the pitch-events that make up the first phrase of
Twinkle, and with the notion of performing the tune (as distinct from,
say, listening to the tune).
Conceptual models. Conceptual models consist of concepts in specified rela-
tionships. Three conceptual models play a prominent role in Jeff s figural
strategy. The first model cor relates pitch-events and objects. Accordingly,
the constituent pitch-events of Twinkle are construed as objects that
endure even after the acoustic signal has faded away. This perspective is
reinforced by the physical and visual evidence provided by the bells. In
turn, each bell can stand for a pitch-event, even when it has not been
struck.9 The second model translates temporal relationships into spatial rela-
tionships. Events separated by equal amounts of time are inter preted as
objects separated by equal amounts of space; events separated by unequal
amounts of time are inter preted as objects separated by unequal amounts
of space. This model infor ms Jeff s arrangement of the bells on the table, so
that there is a greater space between the two bells that mark the end and
the beginning of the two consecutive subphrases (the fourth and fifth bells)
than between the bells within each subphrase. The third model links physi-

9. While I believe that the conceptual models I descr ibe here accurately reflect the cognitive struc-
tures that guide Jeff s actions, they should be generalized only with great caution. Bamberger notes that
some children believe that a bell will sound differently if it is moved to another place. Indeed, there is
good evidence that Jeff is more concer ned with spatial location than with pitch. For discussion of a par-
ticularly revealing moment in his development, see Bamberger, The Mind behind the Musical Ear, 204 09.
To simplify matters I shall continue to focus on pitch-events (since these ultimately infor m the mean-
ing given to any particular location along the bell-path or action-path), with the understanding that the
full descr iption of an event requires specification of both its temporal and spatial coordinates.
conc e p tual mode l s and th e ori e s 103

cal actions with a sequence of sounds. As a result, Twinkle acquires a cor-

poreal dimension to go with its acoustic dimension. Each action produces
another of the pitch-events of Twinkle, and, just as the pitches of Twin-
kle form a coherent whole, so does the sequence of physical actions.10
Conceptual domains. Each of these conceptual models sets up a conceptual
domain. The first model sets up a domain of sound-objects, populated by
physical embodiments of ephemeral sonic events. The second model sets up
the domain of pitch-space, quite apparent in the ar rangement of bells on
the table. As Jeff has constructed it, pitch-space consists of temporal and
structural relationships between sound-objects. The third model sets up a
domain associated with a practiced sequence of actions, what Bamberger
calls a felt path.11 It should be noted that these conceptual domains are
not completely exclusive of one another: in general, conceptual domains
will overlap to the extent that the conceptual models on which they are
based share concepts and relationships.
Theories. Theories coordinate a number of conceptual models in order to
guide inference and reason. In this case, the three conceptual models
described above are coordinated by a theory Jeff s theory of Twinkle.
According to this theory, a representation of the tune can be constructed by
arranging a ser ies of unique sound-producing objects in space in order of
their appearance in the tune, producing a unidirectional path that, if tra-
versed with the appropriate actions, will yield a sounding version of the
tune. Jeff s notation for Twinkle is a symbolic manifestation of this the-
ory: each sound-producing object is represented by a separate line.12 The
uniqueness of these objects is indicated by the ser ies of numbers that is
placed beneath these lines; the actions that will yield the tune are indicated
by the numbers that are placed above.
There will be more to say about each of these constructs, but first let us see how
Jeff s theory of Twinkle changes as Bamberger alters one aspect of the task of
constructing a representation of the tune.

Jeff s Second Version of Twinkle

Jeff continued to work with Twinkle for a number of weeks and became quite
adept at building the tune according to his first theory. One day, as an outgrowth
of a pitch-matching task that she had asked Jeff to do, Bamberger changed the set of

10. In some cases, these physical actions are constitutive of the tune in a way that the ar rangement
of the bells is not. For instance, the bell-path gives no clues as to how many times each bell should be hit,
which, of course, is an integral part of the action-path.
11. Bamberger, The Mind behind the Musical Ear, 213.
12. My use of symbol with respect to Jeff s notation is not very heavily freighted the symbols
are at best ad hoc, and under different circumstances Jeff might use quite different symbols to accomplish
the same notational goals. However, I do want to say that Jeff s notation cor relates with his concepts
that is, the marks he puts down stand as symbols for some of the concepts in ter ms of which he under-
stands Twinkle.
104 aspects of cogn it ive st ruc ture

O I N Q* R S*
figure 3.4 Jeff s modified bell-path for the first phrase of Twinkle, Twinkle

bells Jeff had to work with by removing the brown-based bells.13 This left Jeff with
only one bell for each pitch-type. Jeff built the tune as usual until he came to the
place where, in accordance with the strategy based on his first theory, he would find
and position the brown-based G-bell. When he could not find the bell, he stopped
his work and said, I need another bell.
Bamberger did not press Jeff further at this point, but a few days later she
returned to the task. This time, Jeff decided to look for the needed bell among those
he had already arranged in the bell-path. After first trying the C-bell, he moved to
the G-bell (second in the bell-path) and discovered that it provided the pitch he was
looking for. Jeff eventually was able to build up the entire tune using only the
white-based bells, creating the bell-path and action-path shown in figure 3.4.While
this ar rangement of the bells is more parsimonious only six bells are needed
instead of eight the action-path through the bells is more complex, requiring
both switchbacks and leaps over intervening bells. As a result, the bell-path and
action-path are no longer in simple cor respondence with one another.
As before, Bamberger asked Jeff to come up with a notation that would enable
someone else to play Twinkle on the bells as he had ar ranged them. This time,
however, the complexities of the action-path required to play the tune on the new
arrangement of bells posed some challenges, and it was only after a number of
attempts that Jeff was able to develop the notation shown in figure 3.5. The nota-
tion shows two important innovations. First, by means of arrows, Jeff is able to show
just how one negotiates the switchbacks and leaps necessary to play Twinkle. Sec-
ond, the numbers now refer to only one thing: the number of hits each bell gets.
The ser ial notation Jeff used for his first version of Twinkle has been abandoned,
since there is no longer a one-to-one mapping of bells to pitches.

Conceptual Models and Theor ies in Jeff s

Second Version of Twinkle
While some of the suppositions behind Jeff s construction and notation of the sec-
ond version of Twinkle are the same as those for his first version, some are dif-
ferent. Unchanged are the notions that the pitch-events of the tune can be embod-

13. Bamberger did this in part to see if she could move Jeff from a figural representation of the tune
toward a for mal representation. For a fuller account of this aspect of Bambergers work with Jeff, see The
Mind behind the Musical Ear, chap. 10.
conc e p tual mode l s and th e ori e s 105

+* +

,* , , ,* ,* ,*

figure 3.5 Jeff s notation for his modified bell-path for the first phrase of Twinkle,

ied by the bells that reproduce them and that playing the tune is a matter of per-
forming actions along a path in a specially constructed pitch-space. Changed are
Jeff s ideas about pitches and about the relationship between temporal events and
spatial disposition. In a conceptual shift that ultimately had profound implications
for Jeff s understanding of musical organization (as Bambergers further work with
Jeff showed), he now recognizes that pitch-events are not entirely unique: some
pitches will be repeated over the course of a tune. A given bell can thus serve as a
sound-source for more than one pitch-event. This realization has the consequence
of loosening the relationship between when pitches happen in the tune and how
representatives of these pitches can be disposed in space. As a result, playing Twin-
kle is no longer a matter of following a unidirectional path through pitch-space.
Even though time might still be conceived of as moving unifor mly in one direc-
tion, the motions needed to play Twinkle are no longer unidirectional and ser ial:
the player moves forward and backward, sometimes str iking the bells ser ially, some-
times leaping over one or more bells to find the next pitch of the tune.
The conceptual shift associated with Jeff s changed ideas about pitches, pitch-
space, and temporal order is an example of what Douglas Hofstadter calls conceptual
slippage. As Hofstadter defines it, conceptual slippage is the context-induced dis-
lodging of one concept by a closely related one, within the mental representation of
some situation.14 Here, the context is provided by Bambergers removal of the
brown-based bells from those Jeff had to work with in constructing Twinkle.
What has slipped is Jeff s concept of a pitch-event specifically, how pitch-events
are unique. Jeff still clearly realizes that the G that comes after the first C of Twin-
kle is different from the G that comes after the first A after all, they occur at dif-
ferent points in the tune. He now knows, however, that both can be played on the
same bell. This in tur n means that the temporal uniqueness of pitch-events will not
necessar ily translate into spatial uniqueness: the two distinctly different Gs of the
tune can be found in the same place on the bell-path. Jeff s concept of pitch-event
as a specific pitch in Twinkle has been dislodged by a concept of pitch-event as a
general type of sonic event that is manifested at a particular point in the temporal progress of
Because all of the conceptual models through which Jeff structures his under-

14. Douglas R. Hofstadter and the Fluid Analogies Research Group, Fluid Concepts and Creative
Analogies: Computer Models of the Fundamental Mechanisms of Thought (New York: Basic Books, 1995), 198.
The idea of conceptual slippage has proven quite important in Hofstadters work in that it introduced a
level of flexibility and fluidity into the computer programs his group developed, features that were absent
from earlier artificial intelligence programs.
106 aspects of cogn it ive st ruc ture

standing of Twinkle incorporate the concept of pitch, the slippage of this concept
cascades through the system. The cor relation between sounds and objects provided
by the first conceptual model is now looser than before, and the domain of sound-
objects it sets up is populated with gener ic, rather than specific, entities. In conse-
quence, the functional distinction Jeff made between brown-based bells and white-
based bells (preferring to work with the latter first) is no longer necessar y now
there are only bells. The second model, which maps temporal relationships onto
spatial relationships, is more profoundly affected. While temporal and spatial events
are still connected ( Jeff s bell-path still has a beg inning and an end), the indepen-
dence of temporal succession and the spatial ar rangement of sound-sources is far
more evident. Reflecting this, the domain of pitch-space has been reshaped: once
pitches are conceived as gener ic entities, they start to have more to do with each
other than they do with the specific sounds that make up Twinkle. These alter-
ations in the notion of spatial disposition then influence the third conceptual
model. While the physical actions that produce Twinkle can still be seen as a
coherent whole, it is a whole made of a larger number of disparate parts (including
switchbacks and leaps). This creates a physical manifestation of the growing separa-
tion of the tune from the medium through which it sounds: one can now get lost
along the action-path in a way that would have been unlikely with the first bell-
path Jeff constructed.
As a result of these changes, Jeff s theory of Twinkle is changed as well.
According to this modified theor y, Twinkle can be built using various gener ic
sound-producing objects in space in the approximate order of their occur rence in
the tune. A relatively complex action-path is then imposed on this ar rangement,
which, if properly executed, will yield a sounding version of the tune. Again, Jeff s
notation for Twinkle is a symbolic manifestation of this theory: there are no
longer unique entities that cor respond to each of the constituent pitches of the
tune, and perfor ming the tune involves a ser ies of complex actions. In ter ms of the
two basic strategies Bamberger observed, Jeff has begun to move from a figural rep-
resentation to a for mal one.

The Function and Structure of Theories

At this point, it will be useful to pause and consider five features of theories revealed
by Jeff s two approaches to constructing Twinkle on the bells.
First, theories serve as a guide for actions, both mental and physical. Each of Jeff s
theories of Twinkle allows him to create an ar rangement of the bells that will
enable him to play the tune. Using these theor ies, Jeff would be able to find a way
to arrange similar objects glasses filled with varying amounts of water, marimba
bars, pots and pans so that he could play Twinkle on them. Should something
change in one of these ar rangements (for instance, if one of the bells was covertly
switched for a bell with a different pitch), Jeff would most likely be able to find the
problem and suggest ways of remedying it. In order to avoid confusion later, I want
to broaden the character ization of this feature and say that theor ies serve as a guide
to understanding and reasoning. Although this seems a bit g rand when applied to
Jeff s situation after all, Jeff is someone who is much more predisposed to action
conc e p tual mode l s and th e ori e s 107

than to talking about action it makes clear that the actions with which I am con-
cerned are not of an automatic or unconscious sort but are a manifestation of either
implicit or explicit patter ns of reasoning.
Second, theories provide answers to conceptual puzzles.When Jeff, using his first
theory of Twinkle, could not find a brown-based bell to complete the tune, his
theory told him that he could not continue building the tune, and he stopped. Of
course, this is not a very satisfying answer to the puzzle, not the least because it
points to the inadequacy of the theor y as a guide for further actions. Jeff s second
theory of Twinkle, in which the concept of a pitch-event has slipped toward a
more gener ic perspective, is able to provide a solution by reinterpreting the func-
tion of one of the bells already in the bell-path, thereby permitting him to continue
building the tune.
Third, theories simplify reality. Jeff does not attempt to make the proportional
spacing of the bells replicate the temporal intervals between the pitch-events of the
tune it is enough that there is a somewhat larger space between the bells that end
the first subphrase and those that begin the second. Jeff s notation (and, presumably,
his theory) also do not specify the tempo at which Twinkle should be played,
how hard the bells should be struck, whether the player should be seated or stand-
ing, and so on.
Fourth, theories are relatively extended cognitive structures. As analyzed here,
both of Jeff s theories involve at least three conceptual models, which descr ibe rela-
tionships among a number of concepts and which set up conceptual domains. Each
conceptual model can operate independently of the others: construing pitches as
objects does not necessar ily require arranging the objects in any particular way, nor
does it necessar ily involve physical actions perfor med on these objects (to which
mental rehearsal by musicians bears testament); temporal successions (particularly
those that do not involve music) can be represented spatially without involving
pitch-objects, and so on. Theories, however, coordinate such models into systems
for inference. The systematic quality of theories (in the sense that they involve
coordinated and thus inter related conceptual models) is evident in the way the
slippage of Jeff s concept of pitch-event cascades through his theory of Twinkle:
it affects each of the conceptual models involved.
Fifth, theories are dynamic.When Jeff s concept of pitch-event changes, his the-
ory of Twinkle changes as well. Theories thus respond to changes in circum-
stances and (as Bambergers further work with Jeff shows quite clearly) can g row
even more extensive as they are developed to deal with a broader range of circum-
Although Jeff s efforts to construct and notate bell-paths for playing Twinkle
give a glimpse into how we structure our understanding of music, the view is lim-
ited by the specificity of his situation, which is that of a young boy struggling with
basic musical concepts. In order to know what it means to have a theory, and to
connect Jeff s functional and implicit theor ies with the abstract and explicit theo-
ries of writers like Rameau and Schenker, we need to consider in greater depth the
nature and function of the cognitive structures conceptual models and theor ies
basic to our understanding of the world as a whole.
108 aspects of cogn it ive st ruc ture

conceptual models and theories

As chapter 1 demonstrated, the last quarter of the twentieth century bore witness to
a great deal of research on categor ization research that fundamentally changed
our understanding of the process of categor ization and its role in human cognition.
And as shown by chapter 2, at about the same time, work in metaphor theory and
cross-domain mapping blossomed, resulting in similar changes to our understand-
ing of the role played by mapping structure from one domain onto another. It does
not appear, however, that we will come to regard this time as one in which work on
conceptual models and theor ies came to fruition. While much interesting work has
been done work to which I shall refer in the following it has not converged on
broad agreement as to what counts as a conceptual model (or even what such struc-
tures should be called), nor has it produced any unified view of what constitutes a
Two factors have contr ibuted to this situation. First, the functions we call the
mind and the organ we call the brain are both stagger ingly complex indeed,
Gerald Edelman has called the brain the most complicated mater ial object in the
known universe.15 Further, mind and brain are not coextensive in any simple way:
mind functions extend at the least to the central nervous system, and much of what
the brain does is not directly involved with what we conventionally accept as mind
functions.16 Mind and brain thus admit of any number of structural descr iptions,
and in consequence there is great diversity in how wr iters have character ized con-
structs like conceptual models and theor ies. The second factor, which is related to
the first, is that researchers have tended to focus either on conceptual-model-like
structures or on theor y-like structures, but rarely on both. The advantage of this
strategy is that it limits the amount of cognitive structure that has to be explained.
The disadvantage is that it makes it difficult to explain how relatively low-level
structures like conceptual models coordinate with relatively high-level structures
like theor ies.
In what follows, I give an account of cognitive structure that begins with con-
cepts and extends through theor ies. In line with my analysis of the cognitive struc-
ture behind Jeff s strategies for building Twinkle, I distinguish among concepts,
conceptual models, conceptual domains, and theor ies. I should make clear that I
assume humans rarely, if ever, make explicit reference to structures of this sort in the
process of reasoning about a g iven situation. Instead, as is evident from Jeff s re-
sponses to the challenges presented by building tunes with the Montessor i bells, we
most typically are completely unaware of things like concepts and conceptual mod-
els, no matter how evident they are in our patter ns of reasoning. Concepts, con-
ceptual models, conceptual domains, and theor ies are, by and large, implicit cogni-
tive structures: they can be teased out through exper iment or analysis, but they need

15. Gerald M. Edelman, Bright Air, Brilliant Fire: On the Matter of Mind (New York: Basic Books,
1992), 17.
16. Things get even more complicated when emotions and feelings are factored into our account of
cognition, for there is strong evidence that these are both of the body and of the mind. Antonio
Damasio has offered compelling arguments on this point; see in particular Damasio, Descartes Error: Emo-
tion, Reason, and the Human Brain (New York: Avon, 1994); and idem, The Feeling of What Happens.
conc e p tual mode l s and th e ori e s 109

not be explicit to be useful to us in reasoning about the world. I also do not want to
suggest that these constructs have unitary representations, either cognitively or
physiologically. Indeed, recent work indicates that the lines between entities such
as concepts, models, domains, and theor ies are more often blurry than clear and that
there are no simple neurobiological explanations for any of them. My distinctions
are thus pragmatic ones, intended to reflect the different sorts of cognitive work
done at different levels of structural complexity and the compass of the conceptual
structures that result.
In the first subsection that follows, my focus is on conceptual models and their
relationships to concepts and conceptual domains. In the second subsection, I turn
to theor ies and explore what recent research in anthropology and developmental
psychology can tell us about what it means to have a theory.

Conceptual Models

basic features of conceptual models As I define it here, a con-

ceptual model consists of concepts in specified relationships. To get a better idea
what this means, let us return to one of the models used by Jeff in his initial attempts
to construct a bell-path for Twinkle, the model that translates temporal events
into spatial events. The concepts required for the model are event and object. Both
of these are very broad Type 1 categor ies: the typical event is something that hap-
pens in time; the typical object is something that persists over time. And, as always
with this sort of category, there are boundary cases that challenge what is typical: a
rainbow is in some respects an event (since it happens only as a result of certain
atmospher ic conditions), but, in other respects, an object (since it persists if
briefly over time). In the model, events and objects are related to one another by
a cross-domain mapping between pragmatic construals of time and space, which
interprets durations (that is, the amount of time between events) as spatial disposi-
tions (that is, the amount of distance between objects).17 Large units of time are rep-
resented by distantly spaced points, small units of time are represented by closely
spaced points, and equivalent temporal units get equivalent spacing. This way of
interpreting time can be seen in the distr ibution of marks for minutes on the face
of an analog clock: marks for five-minute intervals are relatively far apart, marks for
one-minute intervals are relatively close together, and equivalent units get equiva-
lent spacing.18
Conceptual models are drawn from two sources. The first, and perhaps most
important, is culture, especially as it is represented by the people around us.We learn
conceptual models by observing and imitating the actions of others and by abstract-
ing models from the lessons we are taught. Conceptual models are thus part of the
shared body of knowledge that constitutes culture. As Naomi Quinn and Dorothy

17. Modern physics, of course, does not retain the clear separation between the categor ies time and
space that is necessary for this mapping.
18. The spatial representation of temporal units on the face of an analog clock reflects in part the
mechanism that moves the hands at a steady rate across the clock face. Compare this to the face of a sun-
dial, on which the marks for equal units of time are not spaced equally.
110 aspects of cogn it ive st ruc ture

Holland put it, culture is not a peoples customs and artifacts and oral traditions, but
what they must know in order to act as they do, make the things they make, and
interpret their exper ience in the distinctive way they do.19 Conceptual models may
also be created through cross-domain mapping. The propositions and structure of
a model framed relative to one domain are applied to another domain; if the process
is successful (if it allows us to structure effectively our knowledge of some aspect of
the target domain), a new conceptual model is bor n.
Models, once lear ned or created, are stored in memory and called up to organize
our understanding of a given set of circumstances. Each of the models Jeff used in
his theory of Twinkle was lear ned or developed before he attempted to build the
tune (even if, as in the case of the model that cor relates pitch-events with objects,
it was lear ned only when he worked with Hot Cross Buns) and was recalled from
memory when needed for the new task.
Although there has been wide interest in constructs like conceptual models,
there has been an almost equally wide variety of construals of mental models of this
sort.20 There is general agreement on two basic features, however. First, conceptual
models are of relatively restricted extent that is, they structure a quite limited por-
tion of our knowledge of the world. Second, conceptual models are abstractions
from exper ience they do not replicate the outside world in all its detail but model

19. Naomi Quinn and Dorothy Holland, Culture and Cognition, in Cultural Models in Language
and Thought, ed. Dorothy Holland and Naomi Quinn, Cambridge Readings in the Literature of Music
(Cambridge: Cambridge University Press, 1987), 4.
20. Analogues for conceptual models, and the relevant literature, include the following.
mental models: Philip Nicholas Johnson-Laird, Mental Models: Towards a Cognitive Science of Lan-
guage, Inference, and Consciousness (Cambridge, Mass.: Harvard University Press, 1983); Lawrence W. Barsa-
lou, Cognitive Psychology: An Overview for Cognitive Scientists (Hillsdale, N.J.: Erlbaum Associates, 1992);
Dedre Gentner and Albert L. Stevens, eds., Mental Models (Hillsdale, N.J.: Erlbaum Associates, 1983); Jane
Oakhill and Alan Garnham, eds., Mental Models in Cognitive Science: Essays in Honour of Phil Johnson-Laird
(East Sussex: Psychology Press, 1996).
idealized cognitive models: George Lakoff, Women, Fire, and Dangerous Things: What Categories
Reveal about the Mind (Chicago: University of Chicago Press, 1987); Robert N. McCauley, The Role of
Theories in a Theory of Concepts, in Concepts and Conceptual Development: Ecological and Intellectual Fac-
tors in Categorization, ed. Ulric Neisser (Cambr idge: Cambridge University Press, 1987), 288 309.
cultural models: Quinn and Holland, Culture and Cognition; Roy G. DAndrade and Claudia
Strauss, eds., Human Motives and Cultural Models, Publications of the Society for Psychological Anthro-
pology (Cambr idge: Cambridge University Press, 1992); Roy G. DAndrade, The Development of Cogni-
tive Anthropology (Cambridge: Cambridge University Press, 1995); Bradd Shore, Culture in Mind: Cogni-
tion, Culture, and the Problem of Meaning (New York: Oxford University Press, 1996).
cognitive domains: Ronald W. Langacker, Theoretical Prerequisites and Descriptive Application, vols. 1
and 2 of Foundations of Cognitive Grammar (Stanford, Calif.: Stanford University Press, 1987, 1992).
frames: Marvin Minsky,A Framework for Representing Knowledge, in The Psychology of Computer
Vision, ed. Patrick Henry Winston (New York: McGraw-Hill, 1975), 211 77; idem, The Society of Mind
(New York: Simon and Schuster, 1985); Douglas B. Lenat and R.V. Guha, Building Large Knowledge-Based
Systems: Representation and Inference in the Cyc Project (Reading, Mass.: Addison Wesley, 1989);
knowledge structures: Robert P. Abelson and John B. Black, Introduction, in Knowledge Struc-
tures, ed. James A. Galambos, Robert P. Abelson, and John B. Black (Hillsdale, N.J.: Erlbaum Associates,
Needless to say, these constructs are by no means equivalent to each other. However, all involve the
basic features I outline for conceptual models.
conc e p tual mode l s and th e ori e s 111

aspects of it as a guide to understanding. Modifying a distinction drawn by Clifford

Geertz, these are models for understanding a particular set of circumstances rather
than models of the circumstances themselves.21
Conceptual models are thus a very basic and yet abstract construal of how the
world is organized they are examples of how concepts first become useful through
being related to other concepts.22 Because we cannot think without thinking in
terms of relationships of concepts, conceptual models are part of the warp and woof
of our knowledge of the world: they represent the way the world is. However, this
is not to say that conceptual models reflect, in a simple way, an objective and unchang-
ing reality, even if they are crucial to the way we construct our reality.
In earlier work, I proposed, following George Lakoff, that models of the sort I
am concer ned with here were idealized.23 Idealization is a way to account for
instances when there is a poor fit between a conceptual model and actual circum-
stances. For example, the conceptual model that sets up the notion of a sound-
object runs into something of an obstacle when it comes to the human voice, since
it is difficult to find the physical object that cor relates with vocal sound. Most of the
time this difficulty is ignored: confronted with the sounds of the human voice, we
continue to think in ter ms of sound-objects. A conceptual model concer ned with
sound-objects could thus be regarded as an ideal to which actual circumstances may
or may not confor m.
I now think it better to regard idealization as belonging to a later stage in con-
ceptualization. As they typically function, conceptual models are immediate and
transparent: they represent our working notion of reality. Accounting for why a
conceptual model does not fit a given situation belongs to a less immediate level of
cognitive structure occupied with reasoning and causal explanation. This is the level
of the theory. One job of theories, then, is to explain discrepancies between cir-
cumstances and conceptual models, even if this explanation is little more than the
recognition that the model works in most, but not all, cases. The distinction reflected
in this perspective is one I shall maintain throughout the following: conceptual
models are immediate and not consciously evaluative; theories are always evalua-
tive they provide an inter pretation of circumstances and possible courses of action
based on this inter pretation.
Conceptual models are a response to, not a simple reflection of , the outside
world. They deal with only a small portion of that world and represent its structure

21. Geertz draws his distinction between models without reference to cognitive processing, focus-
ing only (but quite importantly) on how models are used. A model for is used to guide physical rela-
tionships; a model of is a symbolic representation of a particular situation. See Clifford Geertz, Reli-
gion as a Cultural System, in Anthropological Approaches to the Study of Religion, ed. Michael Banton (New
York: Frederick A. Praeger, 1966), 7 8.While I assume that conceptual models have a symbolic repre-
sentation, I do not assume these symbols are necessar ily available for conscious manipulation (that is, they
may be what Lawrence Barsalou calls perceptual symbols, which were mentioned above in chap. 2).
22. In function and structure, conceptual models bear a resemblance to the very simple, image-
schematic stor ies descr ibed by Mark Turner in The Literary Mind (New York: Oxford University Press,
1996), chaps. 2 and 3. Such stor ies may involve little more than a record of the action of throwing a ball
but nonetheless are essential to our understanding of the world.
23. Lawrence Zbikowski, Large-Scale Rhythm and Systems of Grouping (Ph.D. diss.,Yale Uni-
versity, 1991), chap. 4.
112 aspects of cogn it ive st ruc ture

in ways that have more to do with cognitive efficiency than with accuracy. None-
theless, changes in the structure of the outside world will at times require changes
to a conceptual model. As we saw in the preceding discussion, the absence of the
brown-based bells caused Jeff s concept of a pitch-event to slip, resulting in changes
to the conceptual models basic to his theor y of Twinkle. On a more local level,
biological processes will have a similar, if more incremental, effect on the structure
of our conceptual models. For instance, every time we remember something, we
strengthen some synaptic connections and weaken others,re-wiring our brain on
a micro-level and subtly transfor ming the memor y we would seek to recall. Such
changes will have an influence, slight but cumulative, on our conceptual models.
Conceptual models are thus dynamic at both the global and the local level, although,
because of their immediacy, this fluidity is rarely evident.

representing conceptual models My thinking about conceptual

models has been shaped in important ways by research in artificial intelligence from
the 1970s and 1980s in particular, the work of Marvin Minsky and of Roger
Schank and Robert Abelson24 and I find it useful to character ize conceptual
models in ter ms of simple propositions linked together to create a very basic system
of inference. I thus represented the conceptual models in chapter 1 as a collection
of interlinked nodes and shall use similar representations in chapters 4 and 5. I
should note, however, that I propose these models, and this way of representing
them, much in the spir it of that early work in artificial intelligence: as a place to start
a descr iption of the processes basic to reason and inference. The models are gener-
alized, and they are presented in a way that connects with a wide body of research.
But they are not meant to represent a true, or even empirically verified, picture of
what humans have in their heads when they reason about various situations. This
is not to say that there is no empir ical evidence for conceptual models, only that
my purpose in this book is to show how conceptual models are used rather than
to make an argument for exactly what they compr ise or how they are structured.
As I discuss in the concluding chapter, I believe much work has to be done before
a detailed picture of the structure and function of conceptual models can be

conceptual models and conceptual domains In the con-

ceptual model for sound-objects, the concept of a sound is linked with the concept
of an object through a process of cross-domain mapping. Elements from the invis-
ible and ephemeral domain of sound are cor related with those from the visible and
enduring domain of objects, and structure is mapped between the domains, giving
rise to sound-objects. The domains involved in this mapping the domain of
sound and the domain of objects are broad and relatively gener ic. In compar ison,
the domain of sound-objects is rather more specific. I shall call the more specific
domains set up by conceptual models conceptual domains.
The specificity of conceptual domains is manifested in two ways. First, because

24. Minsky, The Society of Mind; Roger C. Schank and Robert P. Abelson, Scripts, Plans, Goals and
Understanding: An Inquiry into Human Knowledge Structures (Hillsdale, N.J.: Erlbaum Associates, 1977).
conc e p tual mode l s and th e ori e s 113

some of the relationships proper to a conceptual domain may not obtain in the
more gener ic domain (or domains) from which it descends, the scope of the con-
ceptual domain is more limited: for example, not everything that can be said of
sound-objects can be said either of sounds or of objects. Second, conceptual domains
are typically associated with specific phenomena or exper iences. The conceptual
domain of sound-objects set up by Jeff s model pertains to his exper ience with the
Montessor i bells and with the task of building a bell-path for Twinkle. This par-
ticular domain is instantiated by the various bells used to construct the bell-path for
Twinkle, but it can be expanded to include things whose relationship to sounds
or objects is more abstract, such as marks on a page or physical places on a musical
I regard conceptual domains as distinct from mental spaces; again, this is a depar-
ture from my earlier work. As descr ibed in chapter 2, mental spaces are complex and
transitory sites for meaning construction and for reasoning. Furnishing a mental
space requires drawing on a number of coordinated conceptual domains. One way
to achieve this coordination is through the recursive application of conceptual
models. Through recursion, the structur ing principles that give rise to a conceptual
model are subsequently applied to the elements of a conceptual domain or iginally
created by the same, or another, conceptual model. For example, consider applying
a conceptual model for anthropomor phism to the conceptual model for sound-
objects. (Anthropomor phism, it will be recalled from the discussion of anthropo-
morphic blends in the second part of chapter 2, maps the properties of living beings
onto objects.) This application maps the properties of living beings onto sound-
objects, yielding a mental space populated with animistic pitches. Admittedly, such
a space is fairly fantastic even those with a steadfast belief in the viability of an
organicist inter pretation of pitch structure usually rely on conceptual blends rather
than on the rather cur ious idea of animistic pitches but not entirely dissimilar
from mental spaces populated by pitches that obey the laws of classical physics.25
Another way to achieve the coordination of conceptual domains is by linking
conceptual models into a system for reasoning about the world that is, by creat-
ing a theory.

When I outlined the structure and function of theories in the preceding discussion,
I did so with reference to an eight-year-old boys attempts to ar range some rather
unusual bells so as to play the first phrase and only the first phrase of Twinkle,
Twinkle, Little Star. But is this really a theory in the way Darwins theor y of
natural selection is a theory, or in the way Einsteins special theory of relativity is a
Scientific theor ies, viewed from a broad perspective, have the features of theories

25. Organicist construals of pitch relationships are discussed in chap. 7. Early in his career, Jean-
Philippe Rameau construed pitch relations in ter ms of Newtonian physics; for a discussion, see Thomas
Christensen, Rameau and Musical Thought in the Enlightenment, Cambridge Studies in Music Theory and
Analysis, 4 (Cambr idge: Cambridge University Press, 1993), 7 11.
114 aspects of cogn it ive st ruc ture

I outlined above. They serve as a guide for understanding and reasoning, and they
provide answers to conceptual puzzles. Thus Darwins theory sought to explain
(among other things) why finches on various islands of the Galpagos g roup had
different-shaped beaks and why similar variations could be observed in other
species.26 Scientific theor ies also simplify reality. For instance, Darwins theory did not
take into account the effect of inheritance on variation: if natural selection were
the only process affecting variation, variation should become less with each succes-
sive generation until it finally disappears altogether. However, because there is a
recombination of parental traits in offspr ing, each successive generation has new
variations. Finally, scientific theor ies are extended cognitive structures that are sub-
ject to change: Darwinian evolutionary theory was eventually refined to take into
account the effect of inheritance, as well as natural selection, on the evolution of
Despite these similar ities, there are nonetheless significant differences between
scientific theor ies and theor ies of the sort that Jeff used in his encounters with
Twinkle. Jeff s theories are of quite limited extent, pertaining only to the way
bells or similar objects might be ar ranged to play a particular song, and they have
the limited function of guiding his actions within this specific context. Scientific
theories, by contrast, attempt to account for as broad a range of phenomena as pos-
sible, and they offer predictions that can be proved true or false by any of a num-
ber of researchers working independently. These features have contr ibuted to the
success of scientific theor ies as a type, a success demonstrated by the incredible
explosion of knowledge in the hard sciences over the past two centur ies. It is no
wonder that explorations of the structure and function of scientific theor ies gen-
erated an extensive literature in the twentieth centur y and that scientific theor ies
have become a standard to which all other theor ies are compared.
The distance, then, between Jeff s theories and scientific theor ies is considerable,
enough so that using theory for the for mer might seem ill advised. Nonetheless,
work by anthropologists and others over the past thirty years has provided evidence
that scientific theor ies are but one sort of general theory and that the process of the-
orizing is both various and omnipresent. More recently, research by developmental
psychologists has shown that even very young children make use of theory-like
structures to guide their actions. These two lines of research, which I review briefly
in the following paragraphs, do not diminish the distance between Jeff s theories
and those of modern science, but they do suggest that these two types of theory
represent points along a continuum rather than discrete entities.

cultural theorie s The notion of a cultural theory (as distinct from a

scientific theory) came to prominence in the 1960s and reflected a growing aware-

26. One can see the puzzle that challenged Darwin emerging in his earliest wr itings. After noting,
in The Voyage of the Beagle (first published in 1839, three years after the Beagle returned home), the gra-
dation in the sizes of the finchess beaks, Darwin makes this prescient comment: Seeing this gradation
and diversity of structure in one small, intimately related group of birds, one might really fancy that from
an or iginal paucity of birds in this archipelago, one species had been taken and modified for different
ends. Charles Darwin, The Voyage of the Beagle, with an introduction by Leonard Engel (Garden City,
N.Y: Doubleday, 1962), 381 (chap. 17).
conc e p tual mode l s and th e ori e s 115

ness among anthropologists that Wester n models for representing knowledge,

including those of anthropology, were not always adequate or appropriate for non-
Western cultures. Here are three examples of the kind of knowledge that caused
anthropologists and those working in related fields to rethink what was meant by a
1. Research by Brent Berlin and his associates on ethnobotanical classifi-
cation (mentioned in chapter 1 in connection with the way processes of
categor ization were reconceptualized dur ing the 1970s) demonstrated that
the taxonomic systems of indigenous cultures could be every bit as com-
plex as those of Wester n biologists. Berlins further work has shown that
ethnobotanical classification reflects, in great part, the order that is in-
herent in nature and that it also achieves a measure of comprehensiveness
by classifying unknown species through making reference to known
2. For more than a thousand years, long-distance noninstrumental navigation
has been practiced in Polynesia and Micronesia. Experienced navigators
routinely sail their outr igger canoes up to one hundred fifty miles be-
tween islands in an area where less than 0.2 percent of the surface is land.
Edwin Hutchins perhaps uniquely qualified as someone with expertise
in navigation, ethnography, and artificial intelligence has shown that this
method of navigation is based on a computational system distinct from
(yet comparable to) that of Western navigation.28
3. The cure for snakebite among the Rama Indians of Nicaragua is a com-
plicated affair involving ritual, diet, beliefs in the super natural, and herbal
medicines especially prepared by snakebite doctors. For instance, after
someone has been bitten, the victim must make sure the snake is killed; if
the snake is not killed, the Rama believe the person cannot be cured.
While under cure, the patient is placed on a special diet, with foods pre-
pared in a special way; the Rama believe that this is necessar y to avoid
endanger ing the lives of other members of his or her household. Although
it is difficult to sort out just which aspects of the snakebite cure have med-
ical efficacy and which do not, the cure appears to be reasonably effective,
saving the patient over three-fourths of the time.29
One way to account for the stability and efficacy of the knowledge represented
by each of these sets of cultural practices is to construe it as organized by a theory
(a theory of botanical classification, or navigation, or snakebite cure). In that this
knowledge is specific to a particular culture or ethnographic context, anthropolo-

27. Brent Berlin, Ethnobiological Classification: Principles of Categorization of Plants and Animals in Tra-
ditional Societies (Princeton, N.J.: Princeton University Press, 1992); idem, How a Folkbotanical System
Can Be Both Natural and Comprehensive: One Maya Indians View of the Plant World, in Folkbiology,
ed. Douglas L. Medin and Scott Atran (Cambr idge, Mass.: MIT Press, 1999), 71 89.
28. Edwin Hutchins, Cognition in the Wild (Cambridge, Mass.: MIT Press, 1995), 65 93.
29. Franklin O. Loveland, Snakebite Cure among the Rama Indians of Nicaragua, in Medical
Anthropology, ed. Francis X. Grollig and Harold B. Haley (The Hague: Mouton, 1976), 97.
116 aspects of cogn it ive st ruc ture

gists have designated the cor responding theor ies cultural theor ies, folk theor ies, or
ethnotheor ies.30
One thing that distinguishes cultural theor ies from scientific theor ies is that sci-
entific theor ies have an explicit tradition of verification: it has to be possible to
replicate exper imental results that support the theor y, or the theor y is cast into
question.While cultural theor ies clearly have efficacy the classification of a wealth
of plant life, navigation between distant islands, and cure from snakebite all actually
occur they are not subjected to the open and constant scrutiny that marks the
modern scientific method. Cultural theor ies, in contrast to scientific theor ies, orga-
nize a relatively stable body of knowledge that changes only gradually and most
often imperceptibly.31 Cultural theor ies, as a further manifestation of cultural
knowledge, thus both reflect and constitute culture.
As I construe them, cultural theor ies need not be explicit; indeed, theories are
often manifested as patter ns of behavior rather than as explicit statements descr ib-
ing how a specific situation should be understood or what actions are appropriate
for a given set of circumstances.32 This is particularly important for music, for the-
ories of music may be both implicit and nonverbal.33 A glimpse of an implicit and
at least partially nonverbal theor y of music was provided by Steven Felds research
on the Kaluli of Papua New Guinea, which was touched on in chapter 2. Feld
showed that Kaluli ter minology centers around waterfall imager y, which yields a
number of highly distinctive ways to descr ibe the contour of musical descents. By
contrast, there is only a single ter m in Kaluli to descr ibe musical ascents.34 This does
not mean that the Kaluli conception of musical ascent is undifferentiated were an

30. Of course, ethnotheor y, as a sort of Third World shadow of theory, is not a very comfort-
able concept: theory always remains the pr ivileged ter m. For discussion of the related topic of the rela-
tionship between ethnopsychology and psychology, see Cather ine Lutz, Ethnopsychology Compared
to What? Explaining Behavior and Consciousness among the Ifaluk, in Person, Self, and Experience:
Exploring Pacific Ethnopsychologies, ed. Geoffrey M.White and John Kirkpatr ick (Berkeley: University of
California Press, 1985), 35 79.
For an illuminating discussion that also bears on this matter, framed with respect to folk models
(which have the function I here ascr ibe to theor ies), see Ladislav Holy and Milan Stuchlik, The Struc-
ture of Folk Models, in The Structure of Folk Models, ed. Ladislav Holy and Milan Stuchlik (London: Aca-
demic Press, 1981), 1 34.
31. Although histor ical evidence suggests that changes in the theor ies (and the cor responding con-
ceptual models) that are constitutive of culture have often been so gradual as to escape notice, the incur-
sion of Wester n commercial and technological culture into traditional societies dur ing the past two
decades has accelerated the rate of change markedly all of the researchers I cite note the effect of this
incursion into the cultural practices they study. Further discussion of this point, framed relative to musi-
cal practice, is taken up in chap. 5.
32. My construal of cultural theor ies as at least potentially implicit differs from that of Roy DAn-
drade, who has proposed that cultural theor ies are always explicit. See DAndrade, The Development of
Cognitive Anthropology, 172 73.
33. My thinking here follows that of John Blacking, who proposed that implicit theor y embraces
both verbal and non-verbal discourse about music. See Blacking, Music, Culture, and Exper ience, in
Music, Culture, and Experience: Selected Papers of John Blacking, ed. Reginald Byron, with a foreword by
Bruno Nettl (Chicago: University of Chicago Press, 1995), 228 31.
34. Steven Feld, Flow Like a Waterfall: The Metaphors of Kaluli Musical Theory, Yearbook for Tra-
ditional Music 13 (1981): 31 32.
conc e p tual mode l s and th e ori e s 117

ascending melody played improperly, they would be quick to call attention to the
mistake but that, within the broader cultural context relative to which musical
concepts are framed, ascents are not pr ivileged and thus require only the barest of
descriptive terms.35
In sum, cultural theor ies have the same general features that Jeff s theories and
scientific theor ies have. They serve as a guide to reasoning and inference, provide
answers to conceptual puzzles, and simplify reality. They are extended cognitive
structures, and they are subject to change. But cultural theor ies suggest two exten-
sions to the conception of theory that I have outlined thus far. First, theories should
be seen as manifestations of cultural knowledge. Jeff s theor ies must thus be under-
stood as developed within the context of a larger set of cultural practices, and so
should those of modern science. Second, theories can be implicit and to some
extent nonverbal.Without a doubt, explicit theor ies are immeasurably important to
the advancement of knowledge, for their precepts are available for public debate,
discussion, and modification. Even so, implicit theor ies make up a great part of our
cultural and pr ivate knowledge, knowledge that becomes visible only when these
theories are subjected to challenge or analysis.

theory in very young children The fact that theor ies are every-
where in science, in culture, and in our personal construals of the world sug-
gests that they are basic to human understanding. This perspective has received sup-
port from recent research by Alison Gopnik and Andrew Meltzoff, who have shown
that children use patter ns of reasoning quite similar to those of scientists to organize
their understanding of the world. Focusing on exper imental evidence gathered
from studies with nine- and eighteen-month-old children, Gopnik and Meltzoff
have demonstrated that childrens understanding of the world replicates the struc-
ture and function of scientific theor ies.36 For instance, six-month-old infants do not
appear to realize that spatial contact is required to move an object: they assume that
if the object moved shortly after they perfor med some action (like kicking their
legs), it was this action that caused the object to move. At around eight to ten
months, however, infants begin to develop a theor y of action on objects that is,
they appreciate that their actions can affect objects only if they are spatially in con-
tact with those objects. As they grow older, children come to realize that this the-
ory does not work in all circumstances: if an object is out of reach and she is given
a toy rake, the one-year-old will use the rake to make contact with the object

35. For other work on implicit theor ies of music, see Simha Arom, African Polyphony and Polyrhythm:
Musical Structure and Methodology, trans. Martin Thom, Barbara Tuckett, and Raymond Bond (Cam-
bridge: Cambridge University Press, 1991), 139 40; idem, New Perspectives for the Descr iption of
Orally Transmitted Music, World of Music 23 (1981): 43; James R. Cowdery, The Melodic Tradition of Ire-
land (Kent, Oh.: Kent State University Press, 1990); and Ruth M. Stone, Let the Inside Be Sweet: The Inter-
pretation of Music Event among the Kpelle of Liberia (Bloomington: Indiana University Press, 1982).
36. Alison Gopnik and Andrew N. Meltzoff, Words, Thoughts, and Theories (Cambridge, Mass.: MIT
Press, 1997). Those who have worked with infants and young children will appreciate the cleverness with
which the exper iments which produced this data were designed and the r igor with which they were car-
ried out. For a report intended for a more general audience, see Alison Gopnik, Andrew N. Meltzoff, and
Patricia K. Kuhl, The Scientist in the Crib: Minds, Brains, and How Children Learn (New York: William
Morrow, 1999).
118 aspects of cogn it ive st ruc ture

(where the six-month-old might simply wave the rake at random), although this
contact alone will not make the object come any closer. By eighteen months of age,
children refine the theory of action on objects further still, and, now realizing that
just touching the object is not enough, an eighteen-month-old will adjust the rake
so that it captures the object cor rectly.37
According to Gopnik and Meltzoff s view, infants and young children develop
their knowledge much as scientists do. They start with an initial theor y for a par-
ticular set of circumstances (the relationship between actions and objects). They
conduct a variety of exper iments (reaching out to g rab objects, observing how
objects move as a result of their actions). Then they refine and adapt the theor y on
the basis of the exper imental results (when touching the object alone does not
produce the desired results, for example, infants and young children will modify
the theory to acknowledge that spatial contact must be of a specific sort). Indeed,
Gopnik and Meltzoff go even further to argue that it is not so much that there is
a scientist in the child as it is that there is a child in the scientist: the cognitive role
of theories in earliest childhood is the same as the cognitive role of theories at
The strong claim that Gopnik and Meltzoff make in Words, Thoughts, and The-
ories that theory (scientific or developmental) is a means of finding truth and that
theorizing is an evolutionary adaptation conducive to the survival of the species
gives occasion for pause, if only because each theor y frames truth a bit differ-
ently.38 Nonetheless, the weaker version of the claim that the theor ies of children
resemble those of scientists is somewhat more promising. Thomas Kuhn, as part
of the central conceptual frame for his essay on thought exper iments, noted that
both children and scientists sometimes hold paradoxical views of phenomena and
that both lear n when these paradoxes are made evident.39 According to Kuhns
argument, thought exper iments which introduce no new evidence and would
thus seem to offer no opportunity for lear ning are one way of making paradox-
ical views of phenomena evident and thus advancing scientific knowledge. Although
only implicit in his discussion, Kuhn appears to consider the thought processes of
children to be analogous to those of scientists inasmuch as both exper ience the
same sort of nonempir ical learning.
While we have much to lear n about the cognitive structures that guide the
behavior of very young children it is hardly a pun to say that developmental psy-
chology is still in its infancy evidence has begun to accumulate in support of the

37. Gopnik and Meltzoff, Words, Thoughts, and Theories, chap. 5.

38. Paul L. Harris offers a somewhat different argument against what Gopnik and Meltzoff call the
theory theory (that is, the theory that children have theor ies of the world), based on two points: first,
that the theor ies of children and lay adults do not count as true theor ies; second, that the pursuit of truth
is not what guides the scientists day-to-day activities. As is evident from the preceding, I differ from Har-
ris on the first point; however I am in broad agreement with his second point. See Harris, Thinking by
Children and Scientists: False Analogies and Neglected Similar ities, in Mapping the Mind: Domain
Specificity in Cognition and Culture, ed. Lawrence A. Hirschfield and Susan A. Gelman (Cambr idge: Cam-
bridge University Press, 1994), 294 315.
39. Thomas S. Kuhn, A Function for Thought Exper iments, in The Essential Tension: Selected Stud-
ies in Scientific Tradition and Change (Chicago: University of Chicago Press, 1977), 242 46.
conc e p tual mode l s and th e ori e s 119

notion that theor ies are a basic structure of human understanding.40 Whether the
theories of children are of exactly the same sort as those of modern science remains
to be seen; however, it does seem that these two sorts of theories have a similar
structure and that they are responses to the same impulse to understand the world.

cognitive theories and music theories Based on his work

building Twinkle with the Montessor i bells, I do not think that there is a strong
case to be made for character izing Jeff as a scientist. Jeff is not really involved in a
wide-ranging search for knowledge, aided by a protocol of experiment and analy-
sis. Jeff is instead doing his best to meet a fairly specific and, for him, relatively chal-
lenging task. He is similar to a scientist, however, in that theory is an important cog-
nitive tool, serving to guide his actions and make the elusive world of sound a bit
more comprehensible. On the basis of evidence from anthropology and develop-
mental psychology, Jeff is not alone in using theory in this way, for theor ies are part
of the very fabric of human thought. Given this context, the fact that Jeff should
have a theory of music, primitive as it may be, is not sur prising. Indeed, we should
only be sur prised if we could find no evidence for the sort of cognitive theor ies I
have described. Even so, if the distance between Jeff s theor ies and those of mod-
ern science is considerable, the same can be said of the distance between his theo-
ries and those of music theory as a discipline. In the next section, I shrink this dis-
tance just a bit by showing how the same cognitive structure evident in Jeff s
theories can be seen in the mature work of two authors who made music theory
their lifes work: Jean-Philippe Rameau and Heinr ich Schenker.

conceptual models, theories,

and music theory
During the eighteenth century, Jean-Philippe Rameau (1683 1764) was widely rec-
ognized as the foremost opera composer in France. Rameau had intellectual aspi-
rations, as well as musical ones, however, and these he first brought to fruition in
his Trait dharmonie of 1722. The Trait offered two ideas that changed thinking
about musical organization throughout Europe. First, it suggested that the r ich but
complex lexicon of chordal sonor ities used in early-eighteenth-centur y music
could be simplified and rationalized by granting certain ar rangements of notes pr i-
ority (the theory of chord inversion). Second, it proposed that musical organization
was governed by privileged relationships between specific har monic entities (the
theory of harmonic progression). In part because of the novelty of these ideas, in
part because of Rameaus famously treacherous prose, these theor ies were subject to
a variety of reinterpretations and misunderstandings by those who adopted them,
and Rameau himself continued to revise and refine his theor ies through no fewer

40. For further work on the role of theories in reasoning by children, see Frank C. Keil, Concepts,
Kinds, and Cognitive Development (Cambridge, Mass.: MIT Press, 1989); and Frank C. Keil, Daniel T.
Levin, Bethany A. Richman, and Grant Gutheil, Mechanism and Explanation in the Development of
Biological Thought: The Case of Disease, in Folkbiology, ed. Douglas L. Medin and Scott Atran (Cam-
bridge, Mass.: MIT Press, 1999), 285 319.
120 aspects of cogn it ive st ruc ture

than five additional treatises and a host of other wr itings until almost the last years
of his life.41
Heinrich Schenker (1867 1935) was an Austrian pianist, critic, and music the-
orist who supported himself primarily through giving pr ivate lessons. Schenkers
writings include the jour nal-like ser ies Tonwille and Meisterwerk, analyses of Bee-
thovens piano sonatas and Ninth Symphony, and three main treatises: Harmonielehre
(1906), Kontrapunkt (1910 1922), and Der freie Satz (1935). Through these publica-
tions, Schenker developed two ideas that had a profound influence on Anglo-
American music theor y during the latter half of the twentieth centur y. First, he
maintained that highly complex musical compositions were the result of elabora-
tions of more basic and abstract structures. Second (and related), he held that musi-
cal works consisted of a number of hierarchical levels, proceeding from the most
basic and abstract to the most complex and concrete, and that each successive level
involved a further elaboration of a preceding level.42
The theor ies of music developed by Rameau and Schenker are complex, highly
formalized systems for descr ibing and analyzing music. In what follows, I consider
each from the perspective on theor ies developed in this chapter, using short musi-
cal analyses by Rameau and Schenker to provide glimpses into their respective the-
ories. This will allow me to show some of the cognitive substructure of their the-
ories, and it will also reveal some important features of extended theor ies of music.

Rameau on a Recitative by Lully

As do most of the music treatises of the early eighteenth centur y, Rameaus con-
tain few if any analyses of actual works of music. An exception is found in his Nou-
veau systme of 1726, where he provides an analysis of an entire recitative from Jean-
Baptiste Lullys opera Armide of 1686, written on a libretto by Philippe Quinault.43
An excer pt from the analysis is given in example 3.2. The top two staves are as they
would have appeared in Lullys score: a vocal line with text, below which is the basse

example 3.2 Rameaus analysis (1726) of mm. 18 22 of Armides recitative from act 2,
scene 5 of Lullys Armide

A che vons . . . je fr mis! ven geons nous je sou pi re!

9 7 7

7 7 7 7

41. For a more complete and more elegant account of Rameaus music theor y, see Christensen,
Rameau and Musical Thought in the Enlightenment.
42. The notion of hierarchy employed by Schenker is somewhat complex and is discussed further in
chap. 7.
43. Jean-Philippe Rameau, Nouveau systme de musique thorique et pratique (Paris: Ballard, 1726), 80 90.
conc e p tual mode l s and th e ori e s 121

continue part, with Arabic-numeral figures to indicate how the har monization
should be completed. Added beneath these is a staff for Rameaus basse fondamentale,
which encapsulated his analysis of Lullys music.
Rameaus basse fondamentale had its or igins in the precept that each chord had
one fundamental sound (in moder n ter ms, its root), from which the remaining
notes were derived. In his analyses and descr iptions of various successions of chords,
Rameau would notate each of these fundamental notes on an additional staff, thus
giving an account of what he considered to be the basis of any given harmony. The
basse fondamentale was more than just a record of chord roots, however: it also
allowed Rameau to g ive an account of musical syntax. The way Rameau accom-
plished this is more than a little complicated, but the essential points are as follows.44
To explain why chord Y followed chord X rather than chord Z, Rameau exploited
the notion of cadence prevalent in eighteenth-centur y music theory: chord Y fol-
lowed chord X because the two were part of a nor mative cadence for mula. As part
of this for mula, chord X would invariably be a major chord to which was added a
minor seventh (in moder n ter minology, a dominant-seventh chord). Rameau came
to regard the dissonance introduced by this seventh as the mechanism that led to
cadence: the dissonant notes str ike together, and the collision leads to the appear-
ance of the chord that completes the cadence what Rameau called the tonique.
The chord that preceded the tonique, with its fundamental bass a fifth above that
of the tonique, was called the dominante-tonique.45 Of course, music of even the sim-
plest sort involves more than tonic chords preceded by their dominants, but the
basic patter n gave Rameau a place to begin his explanation of musical syntax. This
he did by using the basse fondamentale to show the cadential relationships he took
to be the motive force in music, even if this meant that the note indicated in the
basse fondamentale could be related through only the most circuitous means to the
chord that actually appeared in the music.46
Returning to the passage from Lullys recitative, there are two places where the
note of the basse fondamentale is different from that of the basse continue. In m. 19,
Rameau indicates that the fundamental of the chord B-D-G is G, an inter pretation
wholly in keeping with moder n thought about chordal inversion. In m. 20, how-
ever, Rameau takes E as the fundamental of the C-E-G chord and implies that a D
should be added to the chord; thus the 7 above the E of the basse fondamentale.
This inter pretation reflects his belief that the chord with A as its fundamental
should be preceded by its dominante-tonique, which has E as its fundamental and
which should be accompanied by the dissonant seventh; thus the E and D of m. 20.
As an additional complication, Rameau added the figure 9 to the basse continue
part of the version of Lullys score that appeared in the Nouveau systme. In this way
he was able to accept Lullys bass-note C and accommodate his own added D by

44. My discussion draws on that of Christensen in Rameau and Musical Thought in the Enlightenment,
106 23.
45. As mentioned in n. 25, Rameau employs a cross-domain mapping between the physical and
musical domains to character ize musical progression. Consequent to this mapping, individual pitches are
construed as physical objects that obey the laws of Newtonian physics.
46. For more on the function of the basse fondamentale as an extra-musical commentary on music,
see Allan Keiler, Music as Metalanguage: Rameaus Fundamental Bass, in Music Theory: Special Topics,
ed. Richmond Browne (New York: Academic Press, 1981), 83 100.
122 aspects of cogn it ive st ruc ture

suggesting that the latter should appear as a dissonant ninth above the bass. The fact
that the E of the basse fondamentale appears above the C of the basse continue is not
a problem for Rameau, since the C is explained by his notion of suppositionit
functions as an or namental note below the true bass note E.47 From the perspective
of modern music theory, Rameaus inter pretation of m. 20 is viewed as completely
wrongheaded, whereas his inter pretation of m. 19 is viewed as essentially r ight. It is
important to note, however, that both interpretations follow from Rameaus theory
of musical organization and its account of musical syntax.
Rameaus treatment of Lullys recitative in the Nouveau systme was not to be his
last. In 1753, Jean-Jacques Rousseau offered a competing analysis of Lullys recita-
tive, taking Rameaus discussion as a point of departure.48 Rousseaus intent was to
show that French musicians even the most celebrated had no idea of how to set
text (in contrast to Italian musicians who, in Rousseaus view, did). Singling out the
passage cited in example 3.2, Rousseau wr ites
Certainly here is the most violent moment in the whole scene. It is here that the
greatest struggle is taking place in Armides heart. Who would believe that the
Musician has left all this ag itation in the same key, without the slightest intellectual
transition, without the slightest har monic distinction, in a manner so insipid, with a
melody so little distinguished and so inconceivably clumsy, that instead of the last verse
spoken by the Poet,
Achevons; je frmis. Vengeons-nous; je soupire
[End it; I tremble. Avenge myself; I sigh]
the Musician says precisely this:
Achevons, achevons. Vengeons-nous, vengeons-nous.
[End it; end it. Avenge myself; avenge myself.]49

Rameau responded to Rousseaus polemic with the 1754 Observations sur notre
instinct pour la musique, et sur son principe, which set out to prove two points: that har-
mony is the most important aspect of music (prior to both the melody or the text
that is sung) and that French music excels in the way it uses har mony.
The idea that har mony was the most important aspect of music was not new for
Rameau, for he had argued the point in his 1722 Trait. In the Observations, how-
ever, the argument takes a slightly different for m, based on the premise that our
understanding of music begins at an instinctual level that reflects a natural pr inciple.
This pr inciple exists, Rameau argues, in the har mony that results from the reso-

47. For further discussion, see Chr istensen, Rameau and Musical Thought in the Enlightenment, 120 29;
and E. Cynthia Verba, The Development of Rameaus Thoughts on Modulation and Chromatics, Jour-
nal of the American Musicological Society 26 (1973): 69 91.
48. Jean-Jacques Rousseau, Letter on French Music, in Essay on the Origin of Languages and Writ-
ings Related to Music, trans. and ed. John T. Scott, Collected Writings of Rousseau, 7 (Hanover, N.H.: Uni-
versity Press of New England, 1998), 168 73.
49. Rousseau, Letter on French Music, 171 72. At this most violent moment, the princess and
sorceress, Armide, is about to take her revenge on the hero Renaud, who, enchanted, sleeps before her.
However, confusion freezes Armides vengeful hand as love casts a spell more powerful than that which
grips the sleeping hero. At the point singled out by Rousseau, Armide swings back and forth between a
steely deter mination to kill her foe and a desire to melt into his ar ms.
conc e p tual mode l s and th e ori e s 123

nance of every vibrating system, or corps sonore, including musical instruments and
the human voice.50 The essential elements of this har mony the principle it reveals
are the intervals of the octave, twelfth, double octave, and seventeenth, or, in a
more compact representation, root, third, and fifth.
According to Rameau, these intervals are generated not only above the root (as
overtones) but below the root as well.51 These undertones gave Rameau a way to
explain the or igins of the minor mode. In the Observations, he also used the inter-
vals below the root to explain the affectual qualities of the subdominant in contrast
with those of the dominant. The dominant is a product of the overtones; in conse-
quence, it arouses feelings of vigor and joy. The subdominant literally, the under-
dominant is a product of the undertones; thus, it gives rise to feelings of weak-
ness, gentleness, tenderness, and sadness.52
Having established this perspective on har mony and har monic function,
Rameau argued that Quinaults text was served perfectly by Lullys music, most par-
ticularly so in the case of the passage quoted in example 3.2. Rameaus argument,
illustrated by the example given in example 3.3, runs as follows: shortly before this
point, the music has been in E minor. With the D dominante-tonique of m. 18
(which sets the ang ry, ascending Achevons), the music establishes G major (Ton
de Sol in ex. 3.3). At je frmis, this G is converted into the dominante-tonique
of C. This represents a move to the subdominant of G (Ton dUt), which is accom-
panied by a melodic descent. With vengeons-nous, the dominante-tonique of D
is introduced. This represents a move to the dominant of G (Ton de Re), which is
accompanied by a melodic ascent. Finally, with je soupire, D is converted into the
dominante-tonique of G. Although this signals a return to G (Ton de Sol), it is also
a move to the subdominant of the immediately preceding key of D, and it is accom-
panied by a final melodic descent. According to Rameau, Lully relies on our

example 3.3 Rameaus analysis (1754) of mm. 18 22 of Armides recitative from act 2,
scene 5 of Lullys Armide
[18] [22]

A che vons . . . je fr mis! ven geons nous . . . je sou pi re!

7 6 5 7 7
Ton de Sol Ton dUt Ton de Re Ton de Sol

50. Jean-Philippe Rameau, Observations sur notre instinct pour la musique, et sur son principe (Paris: Prault
fils, 1754; facs. ed., Monuments of Music and Music Literature in Facsimile. Second Ser ies, Music Lit-
erature, 54 [New York: Broude Brothers, 1967]), 2.
51. The claim is first made in Rameaus Gnration harmonique ou trait de musique thorique et pratique
(Paris: Prault fils, 1737; facs. ed. Monuments of Music and Music Literature in Facsimile. Second Ser ies,
Music Literature, 6 [New York: Broude Brothers, 1966]), 4 5.
52. Rameau, Observations, xiixiii.
124 aspects of cogn it ive st ruc ture

instinctual responses to these oscillations between the dominant and the subdomi-
nant to summon Armides swings between revenge and love: each move toward the
dominant arouses in us feelings appropriate to her revenge; each move toward the
subdominant arouses feelings appropriate to the dissolution of her resolve. Rameaus
analysis of this passage serves as the capstone to his argument in the Observations: the
excellence of French music, the appreciation of which had eluded Rousseau, is
embodied in the way it exploits our instinct for music in the service of musical
Rameaus revised analysis necessitated some important changes in his basse fon-
damentale for this passage. In the Observations, he reads the second half of each of
mm. 18 21 as including a dissonant seventh (which is in inversion in m. 19), reflect-
ing the dominante-tonique chords that establish the keys of G, C, D, and finally G.
Perhaps more important, the chord in the first half of m. 20 no longer serves as the
dominante-tonique of the chord in the second half of the measure, for this would
undercut Rameaus assertion that C major is in effect at the beginning of the mea-
sure: the note in the basse fondamentale is now C instead of E, and there is no 7
above it. As Rameaus theory changed, so did his analysis.
Rameaus intent with this analysis, as with the theor y of music on which it is
based, is to explicate the pr inciples of musical organization and musical syntax. He
does this by focusing on har monies and har monic relationships, reading out the
larger implications of the basse continue that is, not only the notes that will com-
plete the musical texture but also the musical sense made by the bass line itself
and making them explicit through the basse fondamentale. The project is a large
and complex one, and Rameaus analysis of Lullys recitative lets us glimpse only a
portion of all that he would try to say about musical organization. Nonetheless, this
vignette is enough to show how Rameaus music theor y is a theor y in the more
general sense developed in this chapter, for all of the five features of theories are
clearly in evidence:

theorie s guide unde r standing and reasoning Rameaus

theory gives an account of harmonic progression that is, it provides an explana-
tion of why one chord follows another. Given a bit of context connected with a
particular chord, we could make a reasoned judgment about what the following
chord would most likely be. Through the notion of chordal inversion, his theory
also explains relationships between chords that compr ise the same pitch classes. In
the analysis shown in example 3.3, the har mony in the first half of m. 19 is the
same as that in m. 22, even though their bass notes are different. Finally, by locating
the origin of music in the naturally occur ring corps sonore, Rameaus theory offers
a way to account for the emotions music evokes in humans. One set of emotions
(broadly, the happy ones) is associated with notes generated above the cor ps sonore;
another set (the sad ones) is associated with notes generated below the cor ps sonore.

theories provide answers to conceptual puzzles One of

the profound puzzles that Rameau sought to solve in the Observations is why music
moves us. This he did by proposing that the cor ps sonore was not simply an acoustic
phenomenon but a natural one, too: we are moved by music in confor mance with
conc e p tual mode l s and th e ori e s 125

natural laws. A more parochial puzzle concer ns successions of bass notes that do not
confor m to his notions of harmonic progression (such as the C3 to A3 in m. 20 of
ex. 3.2, from Rameaus first analysis in the Nouveau systme). Rameau suggests that
the given bass note has, essentially, an or namental function and the real bass note
(indicated in the basse fondamentale) is the one that confor ms to the proper pro-
gression of harmonies. Of course, the puzzle here could be seen as simply a
manifestation of the limitations of Rameaus theory it is one that disappears when
the theory changes, as it has by the time of the Observations. Nonetheless, the issue
raised by this puzzle/non-puzzle points up something important about theor ies and
conceptual puzzles: theories, through the way they descr ibe situations, tend to cre-
ate puzzles, for which the theor y then offers solutions. When the theor y changes,
these puzzles often disappear, to be replaced by new puzzles specific to the new

theories simplify reality As a composer of many successful operas,

Rameau knew well that music consisted of more than just successions of dominant
and tonic chords. Focusing on this succession to the exclusion of others, however,
allowed Rameau to give an explanation of why one har mony followed another, and
thus to give an explanation of harmonic progression. And Rameaus account of the
way humans respond to music greatly simplifies the emotional response to music,
reducing it to an essential binar ism at a variance with the range of emotion evoked
by effective musical works.

theories involve a number of conceptual models Rameaus

complete theor y coordinates a large number of conceptual models, ranging from
models for pitch-events and intervallic relationships to models for or nament and
expressivity. Three models stand out in both of his analyses of Lullys recitative: the
conceptual model associated with the cor ps sonore, which uses basic acoustic prop-
erties to explain relationships among musical events; the conceptual model for
musical mechanics, which maps Newtonian physics onto musical events to explain
musical syntax; and the conceptual model for chordal inversion, which descr ibes
relationships between the notes that make up a har mony in terms of a set of depen-
dencies created by chord generation. Although the notion of the cor ps sonore was
ultimately absorbed into general thinking about the acoustic or igins of music, the
latter two models had a profound effect on the music theory of succeeding gener-
ations: the success of Rameaus theory was such that later theor ists either incor po-
rated these models into their own theor ies or found substitutes for them in alter-
native conceptions of chordal progression and inversion.

theories are dynamic As is evident from the two analyses of Lullys

recitative, Rameaus theor y of music changes over time: in the 1726 version, the

53. This situation is somewhat more common in the humanistic disciplines than in the hard sci-
ences. In the latter, the puzzles are typically independent of humans they obtain whether humans are
observing the phenomena or not whereas in the for mer the puzzles (and the disciplines, for that mat-
ter) would not exist without humans.
126 aspects of cogn it ive st ruc ture

chord at the beginning of m. 20 acts as a dominante-tonique for the chord that fol-
lows; in the 1754 version (expanded to account for the drama evoked by the pas-
sage), the same chord represents a temporar y, but affectually important, harmonic
goal. It should be said that when Rameau changed his theor ies, it was not always for
the better: even though these changes allowed him to account for a wider range of
phenomena, they often introduced complications that made his theor y more
Although Rameaus is a complex and highly elaborate theory, embedded in his-
tory and pertaining to a relatively limited musical repertoire, it nonetheless shares
significant features with the cognitive theor ies that are the basic tools of human
understanding. To fully comprehend Rameaus theor y requires our reclaiming its
historical context and br inging our own conceptual models into cor respondence
with those on which his theor y is based. Certain models, such as that for chordal
inversion, pose few problems. Others, such as the conceptual model for supposition,
have little in common with our own models and require much greater effort to
understand. Nonetheless, the distance between his way of understanding and ours,
or between either of these and Jeff s, is only a matter of the concepts and concep-
tual models employed. The basic cognitive structures that make this understanding
possible are the same.

Schenker on a Waltz by Schubert

Although Heinr ich Schenker adopted some of Rameaus theoretical precepts, his
attitude toward them changed over the course of his wr itings.54 In his Harmonielehre
of 1906, Schenker speaks approvingly of Rameaus work, but after World War I,
Schenker became highly cr itical of all things French, Rameau included. His late
essay Rameau or Beethoven? is a sustained diatr ibe in which he portrays Rameau
as at best ignorant, at worst a malevolent force who contr ibuted to the downfall of
music.55 Rameau was not the only theor ist Schenker cr iticized, however. Even
before the war, Schenker had become convinced that Ger man music was being
pulled into mediocr ity and that music theor ists had aided and abetted the process.56
A central and explicit goal of Schenkers writings on music was to retard or ar rest
this process by showing the basis for the super iority of a select repertoire of music
associated with the her itage of German-speaking countr ies. Schenker believed that
only when the work of composers from Johann Sebastian Bach to Johannes Brahms
was once again fully appreciated could humanity be saved from its own excesses.

54. For a fuller account of how Schenkers attitude toward Rameau changed over the course of his
writings, see Harald Krebs, Schenkers Changing View of Rameau: A Compar ison of Remarks in Har-
mony, Counterpoint, and Rameau or Beethoven?, Theoria 3 (1989): 59 72.
55. Heinrich Schenker,Rameau or Beethoven? Creeping Paralysis or Spir itual Potency in Music?,
in The Masterwork in Music: A Yearbook, trans. Ian Bent, ed.William Drabkin, Cambridge Studies in Music
Theory and Analysis 5 (Cambr idge: Cambridge University Press, 1994 97), 3: 1 9.
56. Heinrich Schenker, Counterpoint: Book I: Cantus Firmus and Two-Voice Counterpoint, ed. John
Rothgeb, trans. John Rothgeb and Jrgen Thym, New Musical Theor ies and Fantasies, 2 (New York:
Schirmer Books, 1987), xxviixxx.
conc e p tual mode l s and th e ori e s 127

Basic to Schenkers music theory were two ideas: first, that the forces that shape
music are independent of humans; second, that these forces deter mine fundamen-
tal aspects of musical syntax.57 These forces are made manifest in what Schenker
called the Naturklang, or chord of nature, consisting of the overtones generated from
a single fundamental, with priority granted to those closest to the fundamental.58
Couching his argument in strongly organicist ter ms, Schenker proposed that these
overtones were first witness to the desire of the tone to propagate itself .59 This
desire also gave rise to various Urstze, or fundamental structures, each of which
began with either the third, fifth, or octave of the fundamental. As shown in exam-
ple 3.4, an Ursatz consists of a simple two-part counter point. The notes with up-
ward stems make up the Urlinie (or fundamental line), which descends stepwise
from the Kopfton (or head note here, the third of the fundamental).60 Below this,
in one-to-one counter point, is the Brechung (or arpeggiation); in example 3.4, this
consists of an alter nation between the fundamental and fifth of the Naturklang. The
Ursatz represents an idealized succession of musical mater ials, all of which are in
clearly specified relationships with each other. It thus provides an account abstract,
minimal, but nonetheless complete of musical syntax: that is, how musical mate-
rials are organized to produce the elements of musical discourse.
As Schenker would have it, the skilled composer, in intuitive response to the
powerful urges of the Naturklang, elaborates an Ursatz further and further until,
finally, a complete work is bor n. The possible elaborations of Urstze are manifold

example 3.4 A Schenkerian Ursatz beginning on the third of the fundamental


57. These two aspects of Schenkers music theory are discussed in a bit more depth in chap. 7.
58. Although there are obvious affinities between Rameaus cor ps sonore and Schenkers Natur-
klang, the latter is more complex in both the overtones proper to the system and their application to
musical structure. In general, the idea that there was an acoustical basis to musical organization was a
commonplace of nineteenth-century musical thought, and there is little evidence that Schenkers think-
ing on this matter was directly linked to that of Rameau.
59. It is self-evident that the urge to produce unending generations of overtones belongs to every
tone in equal measure. One might also compare this urge to that of animals, for it appears in fact to be
in no way infer ior to the procreative urge of a living being. [Es ist selbstverstndlich, da den Trieb,
Generationen von Obertnen ins Unendliche zu zeugen, jeder Ton in gleichem Mae besitzt. Man darf,
wenn man will, auch diesen Trieb einem animalischen vergleichen, denn er scheint in der Tat dem
Fortpflanzungstr ieb eines Lebewesens durchaus nicht nachzustehen.] Heinr ich Schenker, Harmonielehre,
Neue musikalische Theor ien und Phantasien, 1 (Stuttgart: J. G. Cottasche, 1906), 13, p. 42.
60. Translating Schenkers terminology is a notor iously fraught issue. For a discussion, see Robert
Snarrenberg, Competing Myths: The American Abandonment of Schenkers Organicism, in Theory,
Analysis and Meaning in Music, ed. Anthony Pople (Cambr idge: Cambridge University Press, 1994),
29 56. In the following, I shall rely on Schenkers original ter ms (followed, at their introduction, by
common translations) as presented in Schenker, Der freie Satz, Neue musikalische Theor ien und Phan-
tasien, 3 (Vienna: Universal-Edition, 1935).
128 aspects of cogn it ive st ruc ture

and, for all intents and pur poses, are coextensive with the repertoire Schenker
sought to explain. Within these possibilities, two main types of elaboration can be
distinguished. The first is a manifestation of the desire of individual components of
the Ursatz to assert control over longer spans of music. This occurs through the
process of Verwandlung (transfor mation or prolongation), realized through a variety
of compositional stratagems.61 The second type of elaboration replicates whole sec-
tions of the Ursatz on subsidiar y levels of musical structure, thus presenting the
Ursatz in a sort of structural diminution. In both cases, the Ursatz is a first level of
musical syntax, whose elements are subsequently transfor med or reiterated to cre-
ate more extensive musical utterances. This notion of syntax is different from
Rameaus in two important respects. First, rather than descr ibing local, event-to-
event relationships, Schenkers theory attempts to account for relationships over the
course of an entire work. Second, where harmony was the starting point for
Rameau, for Schenker, musical syntax starts with the contrapuntal voice-leading of
the Ursatz.
Of course, composers rarely if ever committed the successive elaborations of the
Ursatz to paper. According to Schenker, creating a musical composition that both
confor med to and properly expanded on an Ursatz was the work of genius. The
best a nongenius could do was to come to an appreciation of the finished work by
retracing the successive elaborations of the Ursatz through a ser ies of hierarchical
levels, which proceed from the backg round (the most minimal elaboration of the
Ursatz) to the foreground (which often resembles a simplified version of the
finished work).62 The result is a str iking image of musical structure that shows
musical syntax operating on a number of levels and within a number of different
time frames within each composition.
To represent the elaborations of an Ursatz and the hierarchical levels that
resulted, Schenker developed a unique set of graphical symbols that he used in his
analyses. These allowed him to show relationships among structural elements and
also made manifest the replication of elements across the various levels of an elab-
orative hierarchy. As an example, let us tur n to Schenkers analysis of the first of
Franz Schuberts Valses nobles, Op. 77. The score for the work is given in example
3.5; Schenkers graphic analysis appears in example 3.6 (the Ursatz for the analysis is
the same as that g iven in ex. 3.4). Although not all of the features of Schenkers
graphic notation need concer n us here, a few will be useful for or ienting ourselves
and for understanding his analysis. Open note heads in the graph (with the excep-
tion of the F5 neighbor note after the double bars) represent components of the
Ursatz; generally speaking, slurs and beams mark off larger syntactic units. The
graph omits the rhythmic figuration, filled-out chords, and much of the registral
detail of Schuberts waltz, as well as various subsidiar y passing and neighbor notes

61. The transfor mation (or prolongation) of elements of the Ursatz is more than a little complicated,
and Schenker used a number of terms in order to capture the variety of ways it is accomplished. See
Schenker, Der freie Satz, 45, p. 49.
62. Actual musical analysis most often proceeds in the reverse direction, going from completed
musical work, to foreground, to background.
conc e p tual mode l s and th e ori e s 129

example 3.5 Franz Schubert, Valse noble, Op. 77 No. 1, D. 969


example 3.6 Schenkers analysis of Schuberts Op. 77 No. 1

and br ief arpeggiations. Nonetheless, it does not take too much imagination to see
(and hear) the outlines of the opening of the waltz in the first portion of Schenkers
After the double bars, the connection between Schuberts music and Schenkers
graph is somewhat less immediate, and the graph becomes somewhat more com-
plicated. Schenkers basic argument, however, is clear: mm. 9 11 serve as contrast-
ing mater ial that prolongs the Kopfton, and after m. 12, the music returns to a
modified version of the opening mater ial that culminates in the completion of the
Ursatz. The completion of the Ursatz is accomplished in spite of two complica-
tions. First, scale degree 2 is not actually present in the music Schenker shows this
by placing the 2 of the Ursatz in parentheses. Second, the Ursatz is brought to com-
130 aspects of cogn it ive st ruc ture

pletion in a higher octave than that established by the Kopfton in m. 8, which runs
counter to Schenkers principle of obligate Lage (obligatory register).63
Schenkers theory, then, offers a view of Schuberts waltz as a sort of abstract dis-
course structure. The first eight measures (the a1 section) serve as introduction, pro-
viding an Anstieg (initial ascent) from C5 through D5 to the Kopfton E5. The next
four measures (the b section) offer a contrast, prolonging the Kopfton through a
neighbor-note figure (indicated with Nbn in ex. 3.6). The final four measures
(the a2 section) return to the opening mater ial, modified in order to complete the
Urlinie and br ing about the conclusion of the piece. According to Schenker, to the
extent that we understand this discourse structure, we understand how the Ursatz is
elaborated to create this unique composition. The force of this understanding is
such that we supply the D6 of the Ursatz at the conclusion of the waltz, even
though it is not literally present in the music.
In the hundreds of analyses that accompanied the exposition of his theor y,
Schenker showed evidence of similar discourse structures in the works of all of the
principal composers from Germany and Austria dur ing the eighteenth and nine-
teenth centur ies, and a few others besides. The theory of music that emerges, not
unlike the works whose structure it seeks to explicate, is fascinating in its extent and
complexity. Nonetheless, it is a theory with all of the features of a cognitive theory
outlined in this chapter.

theorie s guide unde r standing and reasoning Schenkers

theory provides a broad account of musical syntax by explaining how musical mate-
rials relate to one another in confor mance with the rules of counter point and har-
monic progression. His intent is to descr ibe the entirety of musical works. Given
this objective, and the organicist imperative relative to which the theory is framed,
Schenkerian analysis provides an explanation for the function and meaning of even
the smallest part of a musical composition: one can proceed, in an orderly fashion,
from the graph of example 3.6 back to the music of example 3.5. From a larger per-
spective, Schenkers theory gives an account of form, understood as the various roles
that different sections of music play within an overall discourse structure. The indi-
cation of a return to the opening mater ial of Schuberts waltz in m. 12 (indicated by
Schenkers a2 below the staff ) marks not only a reprise of the opening figure but
also a return to the syntactic process initiated by the opening of the waltz. Finally,
Schenker also offers an aesthetic theory: he believed that the works most worthy of
his theoretical accounts were also the most valuable aesthetically.

theories provide answers to conceptual puzzles One of

the modest puzzles presented by Schuberts waltz concer ns the attention we should
pay to it. It is one of dozens if not hundreds of waltzes Schubert tossed off to gen-

63. The principle of obligate Lage is a reflection of the urge of the notes of the Ursatz to return to
their source, which in all cases should involve a descent from the Kopfton to the octave above the fun-
damental (as in ex. 3.4). However, Schenker himself did not seem to regard obligate Lage as obligatory in
all cases he notes the Schubert waltz discussed here as an instance in which the pr inciple is violated.
See Schenker, Der freie Satz, 268, p. 175.
conc e p tual mode l s and th e ori e s 131

erate income. Does it bear the mark of his compositional genius, or is it simply
another bit of doggerel chur ned out to fit the needs of the moment? Schenkers
analysis decidedly affir ms the for mer: whether it was wr itten in a few minutes or
over the course of several days, the waltz confor ms to and properly expands on one
of the Urstze basic to music. As with Rameaus account of Lullys recitative, we also
see the theory providing answers to puzzles specific to the theory. In the analysis of
the Schubert waltz, these are the complications presented by the apparent depar-
tures from the Urlinie in the concluding measures.

theories simplify reality One of the superficial cr iticisms made of

Schenkers theory is that his graphic analyses leave out all of the surface detail that
makes a musical work distinctive and rewarding, and that register, dynamics, and
rhythmic figuration are often obscured or ignored. Indeed, the graph given in
example 3.6 bears this out. This is not a deficiency, however, but simply something
that marks Schenkers theory as a theor y: one that simplifies reality as do all theo-
ries.64 The process of theorizing, then, is one of selecting data from a given situation
and using them to build a more comprehensive understanding of that situation.

theories involve a number of conceptual models Schenkers

propensity for neologisms, his attempts to vivify the realm of tones through com-
parisons to human and animal domains, and his adaptations of the theoretical work
of earlier theor ists give rise to an astounding number of conceptual models. Two
in particular give his analysis of the Schubert waltz a distinctive flavor: the concep-
tual model associated with the Naturklang, which supplies the organicist basis for
his account of musical syntax; and a conceptual model of hierarchy (also organicist,
and discussed in further detail in chap. 7) that cor relates numerous layers of musi-
cal syntax as a way of explaining how basic syntactic structures account for the syn-
tax of an entire work.

theories are dynamic In as br ief a summary as this, it is not possible to

show in any compelling way the dynamic aspect of Schenkers theory over time.
Indeed, about the best that can be done is to note Schenkers flexibility with regard
to obligate Lage, which suggests that his approach to musical structure is not as r igid
as he would sometimes have us believe. Nonetheless, there is a wealth of research that
shows that Schenkers theory was a dynamic one, changing in response to both the
repertoire and problems he engaged.
Many claims have been made for Schenkers theory of music, a theory that, in
its ambition and extent, has few competitors in the Western tradition. No one to my
knowledge, however, has claimed that it is cut from the same cloth as the music the-
ory of Rameau or that of an eight-year-old boy from the Boston area. Indeed, I
would not want to make such a claim, or at least not without considerable qualifi-

64. Ronald Langacker makes a similar point about the analyses produced by theor ies of language,
noting that there is certainly some truth in the view that analysis and descr iption inevitably distort sub-
ject matter since they cannot be their subject matter. Langacker, Foundations of Cognitive Grammar, 1: 14.
132 aspects of cogn it ive st ruc ture

cation. In the things that matter most to music theor ists, Schenkers theory is simply
of another order. Nonetheless, in its general features in the cognitive work that
it does, and in the way that it does it it shares much not only with Rameaus the-
ory of music and with Jeff s theor ies of Twinkle but also with cognitive theor ies
as a whole.

Features of Extended Theor ies

The theor ies of Rameau and Schenker, unlike those of Jeff, are sophisticated intel-
lectual constructs that can account for a wide range of complex musical phenom-
ena. Such theor ies, which I shall call extended theories (in order to differentiate them
from more ad hoc theor ies like Jeff s), have three prominent features. First, extended
theories usually engage a problem or problems of long standing, like those of musi-
cal expression or musical coherence. Ancillary to this, extended theor ies often reveal
or create problems through the constructs they apply to music.
Second, extended theor ies, as the product of traditions of discourse that often
stretch across generations, are utterly embedded in culture. This was quite apparent
in the nationalism that infor med indeed, fueled both Rameaus and Schenkers
work. Although it might be easy to dismiss such cultural trappings as extraneous to
the business of music theory, judging from the wr itings of Rameau and Schenker,
it is hardly so: they see music as thoroughly implicated in the business of history and
Third, theories like those of Rameau and Schenker are invested in developing
systematic accounts of musical organization. This systematic quality can take the
form of developing an exhaustive account of musical structure. In some measure,
this is what Schenker aspired to, as did Allen Forte in his efforts to provide a the-
ory of pitch organization for atonal music.65 The systematic aspect of a theory, how-
ever, may be somewhat more circumscr ibed, especially when its focus is musical
syntax. Here the goal is not so much to explain everything about musical organi-
zation but, rather, what is s in musical organization. This is closer to the goals of
Rameaus theory, and it tempered both the repertoire Schenker chose to deal with
and the problems he engaged.
The topic of musical syntax, especially as a manifestation of the systematic qual-
ity of extended theor ies, brings us back to the paradox that theor ies simultaneously
treat and create problems. The very notion of musical syntax the idea that musi-
cal discourse is in some way organized is in part a creation of theories that seek
to explain musical organization. Depending on the models music theory brings to
bear on the problem, musical syntax can be seen to be quite different things: a mat-
ter of how one har mony leads to another; an account of how basic contrapuntal and
harmonic frameworks are elaborated into full-fledged musical works; or a different
explanation altogether. As shall be seen in the next chapter, my own approach to
musical syntax is different yet, involved as it is with the way musical mater ials are
arranged to take advantage of and to exploit our ability to categor ize musical
65. Allen Forte, The Structure of Atonal Music (New Haven, Conn.:Yale University Press, 1973).
conc e p tual mode l s and th e ori e s 133

Theories, as general cognitive constructs manifested in a wide range of human
activity, accomplish their work by correlating a number of conceptual models, each
of which sets up a conceptual domain. In some cases, these conceptual domains will
be quite similar to one another, but in others, the domains may have very little in
common. This means that the topography of any theory, reckoned in ter ms of the
conceptual domains it compr ises, will not necessar ily be unifor m. A theor y may
thus appear to make contradictor y claims when in fact it has simply shifted focus
from one conceptual domain to a second, and incommensurate, conceptual domain.
Thinking about a theory in terms of its constituent conceptual models can thus be
a tool for coming to understand it better, especially where these models are novel or
lead the theory in surprising directions.
Theories are also part of culture. The culture specific to music theory, it must be
admitted, is a quite rarefied one. Nonetheless, the broader influence of culture can
be seen here, too, as was demonstrated by the nationalistic claims made by both
Rameau and Schenker. These claims are not extraneous to Rameaus and Schenkers
theories but are in fact integral to them manifestations of the broader culture of
which these theor ies are a part and relative to which they are framed. This is not
to deny the uneasiness we may feel when reading the str ident claims of the music
theory of another era but only to say that they are not so very different from the
notions about pitch relationships and chordal inversion common within the same
cultural milieu.
Finally, theories of music can tell us something about cognitive structure as a
whole. Because the focal domain for music theory is a nonlinguistic one, theories of
music demonstrate how we have and cor relate nonlinguistic concepts. The abstrac-
tion of Rameaus and Schenkers theories notwithstanding, theories of music are
often action or iented. These actions can be of a quite visible kind, as demonstrated
by Jeff s attempts to set up different bell-paths for Twinkle, but they can also
involve actions less apparent to the eye: theories often suggest ways to attend to or
respond to music. Given their involvement with nonlinguistic entities and their
concer n with action, theories of music can be seen to be deeply involved with
embodied understanding in ways that few other highly developed cognitive theo-
ries are.
While the preceding chapters have sketched how work in cognitive science can
be applied to problems of musical understanding or, put another way, how our
accounts of musical understanding can be brought into line with what is cur rently
known about the mind and brain much work remains to be done to fill out this
picture. The chapters that follow make a start by showing how the framework
developed thus far can be used to give accounts of musical syntax, cultural knowl-
edge, musical ontology, text-music relations, and the structure of theories. As I
noted in the introduction to this volume, I regard every analysis as based on some
theory or collection of theories. The theor ies I highlight in part II, however, are
theories about how basic cognitive processes are manifested in and exploited by
musical understanding. In consequence, the analyses that follow are of a somewhat
different character than musical analysis as it is usually practiced, for my intent is
134 aspects of cogn it ive st ruc ture

not only to shed light on specific issues of concer n to music scholars but also to
refine a theor y of music cognition. The examples around which I build my argu-
ment are not simply case studies, but are also tools that will help us probe further
into the issues discussed thus far.66
As should be clear, the view of cognitive structure, theory, and analysis that I have
presented thus far is of necessity preliminary, and it deals with only a small part of
the potential issues raised even when the domain of inquiry is restricted to the con-
ceptual level of cognition. Nonetheless, as the following chapters will make clear,
even this partial view offers str iking new perspectives on musical understanding and
on what it means to theor ize about music.

66. This approach to the role examples play in explorations of the structure of thought is similar to
that noted by Mark Turner in his discussion of cognition and rhetor ic. See in particular Turner,Figure,
in Figurative Language and Thought, ed. Albert N. Katz, Cristina Cacciar i, Raymond W. Gibbs, Jr., and
Mark Turner (New York: Oxford University Press, 1998), 80.
part ii
analysis and theory
This page intentionally left blank
chapter four
categorization, compositional
strategy, and musical syntax

E arly in his Theory of Semiotics, Umberto Eco surveyed the communicative

processes and cultural systems that might be included in the field of semiotics,
moving through a list worthy of Jorge Luis Borges: zoosemiotics, olfactory signs,
tactile communication, codes of taste, paralinguistics, medical semiotics, kinesics and
proxemics, formalized languages, written languages, unknown alphabets, secret
codes, natural languages, visual communication, systems of objects, plot structure,
text theory, cultural codes, aesthetic texts, mass communication, and rhetor ic. Mid-
way through this list he considered what he called musical codes. He began by
proposing that the whole of musical science since the Pythagoreans has been an
effort to descr ibe musical communication as a r igorously structured system.1 How-
ever, Eco also observed that, in spite of this tradition of structuralism from which
musical semiotics can draw, music poses ser ious difficulties for semiotic theor y:
Music presents, on the one hand, the problem of a semiotic system without a
semantic level (or a content plane): on the other hand, however, there are musical
signs (or syntagms) with an explicit denotative value (trumpet signals in the ar my)
and there are syntagms or entire texts possessing pre-culturalized connotative value
(pastoral or thr illing music, etc.).2
Eco developed a somewhat fuller account of what he meant when he charac-
terized music as a semiotic system without a semantic level a bit further on, just
prior to setting out the combinator ial rules for semiotic codes. The problem was
not that musical syntax cannot be inter preted indeed, Eco argued that any syn-
tactic system can be inter preted on some level but that the inter pretation of
music is resolutely indeter minate. There is thus no depth to the semantic levels pro-
duced by musical syntax.3

1. Umberto Eco, A Theory of Semiotics, Advances in Semiotics (Bloomington: Indiana University

Press, 1976), 10. The idea that musical communication is a r igorously structured system is, of course,
another way to theor ize about music. Interest in such an approach has waxed and waned along with
musicians interest in numerology, mathematics, and science as possible models for musical organization.
2. Eco, A Theory of Semiotics, 11.
3. Eco, A Theory of Semiotics, 88 90. In advancing this point, Eco engages in a dialogue with Louis
Hjelmslevs distinction between semiotic and non-semiotic systems; see Louis Hjelmslev, Prolegomena to

138 analysis and theory

I find Ecos comments interesting, not for the view of musical semiotics they
provide (since the passages to which I have referred are pretty much all he has to say
about music in A Theory of Semiotics), but for their clear focus on the importance of
musical syntax for musical understanding. Musical syntax is where any semiotics of
music must begin. In what follows, I build on this idea, using the perspective on
musical syntax introduced in chapter 1. There, I proposed thinking of syntax as a
connected or orderly system. In the case of music, organized sequences of musical
events properly recognized as syntactic are typically created through a set of musi-
cal practices shared among a number of musicians. Within most traditions of music
making, however, there is also some latitude in how musical mater ials can be
arranged, and the syntax specific to a particular musical work may emerge only over
time. This was the case with the Leidensmotiv from Wagners Tristan und Isolde, for
the importance of three successive statements of the motive was something that
became evident only once we were well into the opera. In situations such as this,
certain aspects of the systematic quality associated with musical syntax are deter-
mined by the expressive goals of the individual ultimately responsible for ar rang-
ing the musical mater ials. My term for the way such expressive goals are realized is
compositional strategy.4
Compositional strategy, conceived in these ter ms, assumes an alliance between
syntax and processes of meaning construction: composers arrange musical mater i-
als with relatively specific expressive or communicative goals in mind. This approach
takes inspiration from recent work in cognitive linguistics, which does away with a
hard-and-fast distinction between syntax and semantics and instead construes lan-
guage as made up of symbolic structures that consist of form-meaning pairs.5 A
similar perspective can be seen in Mark Turners work on the iconicity of rhetor i-
cal figures. Turner argues that, in some cases, the for m of a rhetor ical figure is
matched to the meaning the speaker wishes to convey, connecting the image-
schematic structure of the for m with the image-schematic structure of the mean-
ing. Thus a rhetor ical figure based on repetition, such as anaphora (which involves
the repetition of the same word or group of words at the beginning of successive
clauses, sentences, or lines), can be used to summon the image of repeated blows,
as in an example attr ibuted to Longinus: By his manner, his looks, his voice, when

a Theory of Language, rev. ed., trans. Francis J.Whitfield (Madison: University of Wisconsin Press, 1963),
111 13. Hjelmslev, for his part, leaves open the question of whether music is a semiotic system in the
strict sense he develops.
4. It should be clear from the general way I have formulated the notion of compositional strategy
that these strateg ies are not the exclusive property of those traditionally recognized as composers. In
certain situations, performers and even listeners may use such strategies to organize musical mater ials.
Throughout this chapter, I shall simplify things by assuming that composers are the architects of com-
positional strategies, with the understanding that (mutatis mutandis) these strategies can also be imple-
mented by those not conventionally recognized as composers.
5. Ronald W. Langacker, Conceptualization, Symbolization, and Grammar, in The New Psychology
of Language: Cognitive and Functional Approaches to Language Structure, ed. Michael Tomasello (Mahwah,
N.J.: Erlbaum Associates, 1998), 2. Langackers notion of a symbol is spelled out in g reater detail in his
Theoretical Prerequisites (vol. 1 of Foundations of Cognitive Grammar)(Stanford, Calif.: Stanford University
Press, 1987), 11 12, 56 58.
cate g ori zat i on, st rate g y, and sy ntax 139

he strikes you with insult, when he str ikes you like an enemy, when he str ikes you
with his knuckles, when he str ikes you like a slave. The efficacy of such a connec-
tion is straightforward enough as Turner notes, Involving members of the audi-
ence in the image schema of the iconic for m automatically involves them in the
basic structure of the meaning, thus moving them part way toward accepting the
whole.6 In what follows, I show how compositional strateg ies in music make use
of similar connections between expressive goals and the for ms expressions take.
Music being music, the meaning that is conveyed is not as direct as that which con-
cerned Longinus, nor as quotidian as that which concer ns cognitive linguists.
Nonetheless, I argue for a direct connection between compositional strategy and
musical syntax: the organization of musical elements into systematic structures is
part and parcel of meaning construction in music.7
The particular focus of what follows is on the part processes of categorization
play in compositional strateg ies. As discussed in chapter 1, categor ies of musical
events have immediate salience for listeners and are thus important guides to any
understanding of how musical mater ials are organized. Building on the perspective
developed in chapter 3, we can say that they are basic to any theory of music. Inas-
much as composers try to shape their music to confor m to the capacities of listen-
ers (trying, in Schoenbergs ter ms, to create musical coherence so that the work can
be comprehensible), processes of categorization will also play an important part in
compositional strategies. Musical categor ies are thus a place where the concer ns of
listeners and the concer ns of composers meet.
In the first section of this chapter, I explore the role categor ies of musical events
play in the compositional strategies basic to a musical game attr ibuted to Wolfgang
Amadeus Mozart. The game in question is one of a genre, which seems to represent
the very antithesis of compositional strategy, since such games produce finished or
nearly finished musical works simply through tosses of the dice. Embedded within
their structure, however, is a set of rather traditional strategies for wr iting two-part

6. Mark Turner, Figure, in Figurative Language and Thought, ed. Albert N. Katz, Cristina Cacciar i,
Raymond W. Gibbs, Jr., and Mark Turner (New York: Oxford University Press, 1998), 50 51. Turners
discussion of the iconicity of rhetor ical for m appears on pp. 49 51. The quotation attr ibuted to Long-
inus is from Aristotle, On the Sublime, trans. W. Hamilton Fyfe, Loeb Classical Library, 199 (Cambr idge,
Mass.: Harvard University Press, 1995), 20, p. 190. Both a poetic and musical use of anaphora can be
found in Bob Dylans Knockin on Heavens Door, which first appeared in his 1973 album Pat Garrett
and Billy the Kid. The chorus of the song consisting of reiterations of the line Knock, knock, knockin
on heavens door makes use of a repeated note in the melody, as well as a repeated word. Both repe-
titions cor relate with the meaning of the text. Note as well that the meaning is at least potentially figura-
tive, rather than literal, since nowhere in the song is it assumed that heaven actually has a physical door
on which one can knock.
7. As I have character ized it here, compositional strategy bears a strong resemblance to the tactical
logistics of rhetoric, and it might thus be seen as another instance of the use of ideas drawn from rhetor ic
to conceptualize musical design (something discussed in greater detail in the first part of chap. 7). As will
become apparent, I construe the idea of compositional strategy as rather more general than its rhetor i-
cal counter part, and more involved with issues of immediate importance to musical expression (such as
how the temporal succession of musical mater ials should be ar ranged). This construal of compositional
strategy seems largely consonant with the approach Edward Cone took in his article On Der ivation:
Syntax and Rhetor ic, Music Analysis 6 (1987): 237 55, which was discussed in chap. 1.
140 analysis and theory

musical for ms, realized through the manipulation of categor ies of musical events.
During the eighteenth century, these same strategies were applied, on a much larger
scale, to sonata for m, guiding the way harmonic and thematic mater ials were laid out.
They could also be applied to melodic mater ial alone, however. In the second section
of this chapter, I show how this was accomplished through an analysis of the syntax
of the principal motive from the first movement of Mozarts Dissonance Quartet
(K. 465). Taking the different versions of the motive as members of a Type 1 category,
the patter n of typicality descr ibed by these members as they appear over the course
of the movement shows strong similar ities to the basic strategies for sonata for m. The
syntax for these motive forms nonetheless occupies a different layer than the syntax
that guides the for mal organization of the movement as a whole.
As with the grouping of statements of Wagners Leidensmotiv, the syntax of the
principal motive from the first movement of the Dissonance Quartet emerges
over the course of the movement. Both the emergence of musical syntax and the
independence of syntactic levels is further demonstrated by the first movement of
Ludwig van Beethovens String Quartet, Op. 18 No. 6, which is taken up in the
third section of the chapter. Although there is little variation of the principal
motivic mater ial for the movement, Beethoven is nonetheless able to imbue two
versions of the motive with distinctly different syntactic functions. The categor ies
of motive forms that compr ise these versions then become elements at a higher
level of syntax. The nesting of syntactic functions can also be seen, in a slightly more
complicated way, in the first movement of Beethovens String Quartet, Op. 18 No.
1, which is discussed in the fourth section of the chapter. There, different versions
of the main motive are distinguished by melodic structure and by compositional
strategy, and the categor ies for each become syntactic elements within the move-
ment as a whole. In a concluding section I review the perspective on musical syn-
tax developed over the course of the chapter and g ive a bit further consideration
to how syntax contr ibutes to the construction of meaning in music.

musical games and

compositional strategy
A Musical Game Attributed to Mozart
Europe in the latter half of the eighteenth centur y saw the blossoming of a num-
ber of musical games that allowed any person, without the least knowledge of
Musick, to compose minuets, marches, waltzes, and contredanses simply by toss-
ing dice or spinning a top.8 As if by magic, a ser ies of throws or spins produced a
workable piece of music that was different every time the game was played.
The game descr ibed here the Musikalisches Wrfelspiel allowed its players to

8. One example is Pier re Hoegis A Tabular System Whereby the Art of Composing Minuets Is Made So
Easy That Any Person, without the Least Knowledge of Musick, May Compose Ten Thousand, All Different, and
in the Most Pleasing and Correct Manner (London: Welcker, 1770). For further infor mation about such
games and comprehensive lists of them, see Leonard G. Ratner, Ars combinatoria: Chance and Choice in
Eighteenth-Century Music, in Studies in Eighteenth-Century Music: A Tribute to Karl Geringer on His Sev-
entieth Birthday, ed. H. C. Robbins Landon (New York: Da Capo Press, 1979), 343 63; Stephen A.
Hedges, Dice Music in the Eighteenth Century, Music and Letters 59 (1978): 180 87; and Sebastian Klotz,
cate g ori zat i on, st rate g y, and sy ntax 141

figure 4.1 Number table for mm. 1 8 of the Musikalisches Wrfelspiel

write a sixteen-measure waltz. It is one of several such games attr ibuted to Mozart
and is catalogued as K. Anh. C 30.01, but the ascr iption is hardly secure.9 Mozarts
authorship, in any event, is not my central concer n. Instead, I want to focus on the
compositional strategies that can be teased out of this particular game.
The game requires the use of a set of dice, and of two tables a number table
(the one used for the first eight-measure per iod of the waltz is shown in fig. 4.1)

Ars combinatoria oder Musik ohne Kopfzerbrechen: Kalkle des Musikalischen von Kircher bis Kir n-
berger, Musiktheorie 14 (1999): 231 45.
9. See Wolfgang Amadeus Mozart, Musikalisches Wrfelspiel: Eine Anleitung Walzer oder Schleifer mit
zwei Wrfeln zu componieren ohne Musikalisch zu seyn, noch von der Composition etwas zu verstehen, ed. Karl
Heinz Taubert (Mainz: B. Schotts Shne, 1956). On the matter of the authenticity of the attr ibution of
this particular game to Mozart, see Paul Lwenstein, Mozart-Kuriosa, Zeitschrift fr Musikwissenschaft 12
(1930): 342 46; Otto Er ich Deutsch, Mit Wrfeln komponieren, Zeitschrift fr Musikwissenschaft 12
(1930): 595; Herbert Gerigk, Wrfelmusik, Zeitschrift fr Musikwissenschaft 16 (1934): 359 63; Ludwig
Ritter von Kchel, Chronologisch-thematisches Verzeichnis smtlicher Tonwerke Wolfgang Amad Mozarts, 8th
ed., ed. Franz Giegling, Alexander Weinmann, and Gerd Sievers (Wiesbaden: Breitkopf and Hrtel, 1983),
581, 910; and Hedges, Dice Music in the Eighteenth Century, 183.
It is known from sketch studies (unrelated to the particular game under consideration) that Mozart
had an interest in musical diversions of this sort; see Kchel, Chronologisch-Thematisches Verzeichnis, 581.
And a musical dice game survives that is perhaps more reliably attributed to Mozart; it appears, in his hand,
on the verso of a manuscr ipt that contains a fragment also in Mozarts hand of the G minor Str ing
Quintet K. 516. See Hideo Noguchi, Mozart Musical Game in C Major, K.516f, Mitteilungen der Inter-
nationalen Stiftung Mozarteum 38 (1990): 89 101. K. 516f is significantly different from the game to be dis-
cussed below, however, and the dispar ities that distinguish the two could be seen as arguing against the
authenticity of the latter one. The game from K. Anh C 30.01 that produces the waltz, and that I will con-
sider, was published under Mozarts name soon after the composers death by Johann Julius Hummel
(Amsterdam and Berlin, 1793). This is hardly a guarantee of Mozarts authorship, since his name could
readily have been used simply to boost the sales of a work by another (and unknown) author.
142 analysis and theory

figure 4.2 Portion of the notation table for the Musikalisches Wrfelspiel

and a notation table (a portion of which is shown in fig. 4.2). For the first measure
of music, the player throws the dice and uses the result to find a number in the first
column of the table. Thus, the throw 11 gives the result 3 in the first column. The
player then tur ns to the notation table, finds the music designated with a 3 there,
and uses it for the first measure of the waltz. For the second measure of music, the
player throws the dice again and uses the result to find a number in the second col-
umn of the number table. The throw 10 gives the result 142 in the second column;
the music that cor responds to 142 on the notation table is then used for the second
measure of the waltz. A complete waltz produced by the game is shown in exam-
ple 4.1. Although the dance is hardly distinctive, it is nonetheless completely ser-
viceable, equal to if not exceeding the quality of the Tafelmusik produced by knowl-
edgeable amateurs of the period.
Of course, the point of such games was not the production of works of music,
but diversion. There were a number of ways the game could be played: a group of
players might sit in a circle and take turns throwing the dice, so that each succes-
sive measure was deter mined by a different player; players might each create their
own waltzes, which could then be compared with one another; the entire piece
might be wr itten down and then perfor med; or players might take turns perfor m-
ing the piece as it progressed. In any event, the game offered an amusing and pleas-
ant way to spend a portion of an evening, with the added benefit that music pro-
duced by what appeared to be total chance was orderly and coherent, no matter
what the skill of the player.
In truth, chance played little part in the success of the music produced by such
games. Instead, what was required of the compilers of games such as this one were
cate g ori zat i on, st rate g y, and sy ntax 143

example 4.1 Waltz produced by the Musikalisches Wrfelspiel

Dice throw 11 10 9 5 8 5
Result in
notation table 3 142 114 85 99 2

3 7 3 5
134 91 117 176
7 1 2

7 6 9 8 8 5
11 150 125 51 168 89 170

a little knowledge about how to put the game together and an understanding of the
formal design of waltzes, contredanses, and the like. One first needs to construct a
number table.10 A table for a sixteen-measure waltz using two dice, for instance,
starts with an 11 by 16 grid (for the eleven possible throws of the dice and sixteen
measures of the piece; the table shown in fig. 4.1 provides only one-half of this
grid). The numbers 1 through 176 are then randomly distr ibuted on the 176 squares
of the grid to complete the table. To construct a notation table, it is necessar y to
write eleven variations on a sixteen-measure waltz, preferably arranging these on a
single sheet with the first variation at the top and the eleventh at the bottom. The
various numbers of the number table are then mapped onto the individual measures
of the eleven variations. For instance, the number table used in our Musikalisches
Wrfelspiel would map the number 96 (found at the intersection of the first row and
first column) onto the first measure of the first variation, the number 22 onto the
second measure of the first variation, and so on for all 176 measures of the eleven
variations. An example of how this might look, derived by working backward from
the number and notation tables, can be seen in example 4.2.11 Once each measure

10. Another approach would be for a number table to be adopted from a previous publication. In
fact, the number table for our waltz game is exactly the same as that for the the one published in the
Weimar Journal des Luxus und der Moden of February 1787. See Karl Heinz Taubert (ed.) Das Menuett:
Musikalisches Wurfelspiel, Paris 1786 (Zurich: Musikhaus Pan, 1988).
11. There is no way to know whether the eleven variations shown in ex. 4.2 are actually the source
variations for the game attr ibuted to Mozart: the number found in the first square of the number table
144 analysis and theory

example 4.2 Eleven variations with numbers derived from the number table for the
Musikalisches Wrfelspiel

m. 1 2 3 4 5 6 7 8
96 22 141 41 25 122 14 30

1. **2.


32 6 128 63 146 46 34 31

2. **2.


69 95 158 13 153 55 110 24

3. **2.


40 17 43 85 161 2 159 100

4. **2.


48 4 163 45 80 97 36 107

5. **2.

* The notes with stems down are used for the first ending.
** The notes with stems up are used for the second ending.

could, in theory, be mapped onto any of the variants of m. 1, and the game would still function prop-
erly. However, my focus in the following is not on establishing the or iginal for m of each of the eleven
variations but on the way the mater ial of each measure is varied. The principal function of ex. 4.2, which
puts each measure into its proper column, is to aid discussion of this aspect of the musical game.
Readers who would like to try their luck at composing a piece with the Musikalisches Wrfelspiel can
use ex. 4.2 (supplying their own dice). The number cor responding to each variation can be read as the
result of the throw of the dice; the columns denote the measures of the dance, as in the or iginal num-
ber table. Of course, playing the game with the number table and the notation table collapsed together
(as in ex. 4.2) takes most of the mystery and pretty much all of the fun out of the game.What remains
is the somewhat thin pleasure of finding out what combinations of measures chance will dictate.
cate g ori zat i on, st rate g y, and sy ntax 145

example 4.2 (continued)

9 10 11 12 13 14 15 16
70 21 26 9 112 44 109 14

117 39 126 56 174 18 116 83

66 139 15 132 73 58 145 79

90 176 7 34 67 160 52 170

25 143 64 125 76 136 1 93


has been assigned a number, all of the measures are rewritten, in ascending numer-
ical order, to create the notation table.
The permutational combination required by the game each of the variants of
m. 1 had to make sense with each of the variants of m. 2; each of the variants of m. 2
had to make sense with each of the variants of m. 3; and so on placed significant
constraints on what could be varied and what sort of music was produced by the
game. In general, all of the variants of a given measure had to have the same har-
example 4.2 (continued)

m. 1 2 3 4 5 6 7 8
104 157 27 167 154 68 118 91

6. **2.


152 60 171 53 99 133 21 127

7. **2.


119 84 114 50 140 86 169 94

8. **2.


98 142 43 156 75 29 62 123

9. **2.


3 87 165 61 135 47 147 13

10. **2.


54 130 10 103 28 37 106 5

11. **2.

* The notes with stems down are used for the first ending.
** The notes with stems up are used for the second ending.

example 4.2 (continued)

9 10 11 12 13 14 15 16
138 71 150 29 101 162 23 151

16 155 57 175 43 168 99 172

120 88 48 166 51 15 72 111

65 77 19 82 137 38 149 8

102 4 31 164 144 59 173 78

35 20 108 92 12 124 44 131

148 analysis and theory

mony, and they typically had the same bass note. The melody for each measure was
restricted to gener ic figures arpeggios, an occasional scalar run, or sometimes a
single note that could easily be concatenated with other such figures. In truth, the
compiler of the game wrote not so much variations on a waltz as variations on the
harmonic plan for a waltz, using melodic fragments that served the minimal con-
trapuntal plan of the whole.12 Rhythmic figuration offered the chief means of
deriving the variants of a given measure, although changes to the beg inning
melodic note, overall melodic contour, or number of nonhar monic tones were also
used to render each variant unique. The result of a game constructed according to
this plan was a set of variations that was, for all practical pur poses, inexhaustible:
there are theoretically 11 that is, 45,949,729,863,572,161 different waltzes that
could be wr itten from the mater ials provided in example 4.2. The waltzes would
not be great music, but their for mulaic melodies and predictable har monic plan
were perfectly in keeping with dance for ms of the late eighteenth century.
The success of the compositions produced by such games is often attr ibuted to
the rationalization of musical mater ials effected dur ing the eighteenth centur y,
which allowed specific compositional functions to be attr ibuted to musical mate-
rials.13 Although such rationalization is quite evident in the work of music theor ists
like Joseph Riepel and Heinr ich Chr istoph Koch,14 and while it is supported by the
interchangeability of the components of example 4.2, it is not the only thing that
makes the waltz of example 4.1 believable as music.What is also at work is a com-
positional strategy that shapes the whole of the waltz. The strategy is evident in the
overall har monic plan for the dance, an outline of which is given in figure 4.3. The
waltz is divided into four-measure sections; the first section (mm. 1 4) establishes
the tonic, the second (mm. 5 8) modulates to the dominant, and then the two sec-
tions repeat. The third section (mm. 9 12) remains in the dominant, the fourth sec-
tion (mm. 13 16) returns to the tonic, and these, too, are repeated. Every one of the
waltzes generated by the game will have this basic har monic syntax. However, a suc-
cession of harmonies alone does not create convincing music Rameaus har-
monic theor y notwithstanding, every succession of dominant to tonic does not
make for a cadence. What is crucially important for the compositional strategy
embodied by the Musikalisches Wrfelspiel is the way the mater ial for each measure
is varied. For instance, note that only one version of m. 8 appears in example 4.2

12. Mostly, the counter point above the bass is relatively free (within the str ictures of style). In a few
cases, however mm. 9 10, for example, where the dominant in position leads to a tonic , and at the
cadences that conclude each per iod the melodic possibilities are constrained by the overall counter-
point in the outer voices.
13. Ratner, Ars combinatoria, 344 45.
14. See Joseph Riepels Anfangsgrnde zur musicalischen Setzkunst (1752 68), included in his Smtliche
Schriften zur Musiktheorie,Wiener musikwissenschaftliche Beitrge (Vienna: Bhlau, 1996); and Heinr ich
Christoph Koch, Versuch einer Anleitung zur Composition (Leipzig: Adam Fr iedrich Bhme; Rudolstadt:
gedruckt mit Schr iften der Lwischen Erben, und Schirach, 1782 [vol. 1]; Leipzig: Adam Fr iedrich
Bhme, 1787 93 [vols. 2 3]; repr. Hildesheim: Georg Olms, 1969), esp. vol. 2. For a translation of por-
tions of the latter, see Heinr ich Chr istoph Koch, Introductory Essay on Composition: The Mechanical Rules
of Melody, Sections 3 and 4, trans. Nancy Kovaleff Baker, Music Theory Translation Ser ies (New Haven,
Conn.:Yale University Press, 1983). The work of both Riepel and Koch is discussed more extensively
in the first half of chap. 7 in this volume.
cate g ori zat i on, st rate g y, and sy ntax 149

T"&#5(:*+* T"&#5(:*,*
+* ,* 2* 0* P* /* U* .*

O1*G* G* V* G* I1*V* G* GV* V* /*W*P*

GG* V* /*W*P*

T"&#5(:*2* T"&#5(:*0*
-* +Y* ++* +,* +2* +0* +P* +/*

I1*V* G* GV* G* O1*G* G* GV* V* G*

GG* V*
figure 4.3 Basic har monic plan for the Musikalisches Wrfelspiel

and, save for the ending of the tenth variation, only one version of m. 16.15 By con-
trast, none of the versions of m. 1 are identical. Shaping each of the trillions of
waltzes that might be generated by the game is a template that regulates the degree
and kind of variation that can occur within the sixteen-measure framework of the

A Template for Compositional Strategy

We can discover the plan of the template if we approach each column of measures
in example 4.2 as a Type 1 category of the sort discussed in chapter 1. Each measure-
category can be defined in ter ms of attributes (such as har mony, rhythmic figura-
tion, beginning melodic note, and so on), to which each of the members of the cat-
egory (that is, the different variants included in the category) supplies values. Each
measure-category also has a distinctive profile of typicality, determined by the range
of values its members assign to the various attr ibutes and the weight given any par-
ticular set of values. For example, the members of measure-category 16 assign two
different values for the attr ibute rhythmic figure, but assign one of these ten out of
eleven times (as shown in fig. 4.4). With respect to the attr ibute rhythmic figure, we
can say that the profile of typicality for measure-category 16 is clearly defined. By
comparison, the members of measure-category 15 assign seven different values for
the attr ibute rhythmic figure; one value is assigned five times, the other six are
assigned once each. The profile of typicality for rhythmic figuration is thus some-
what less clearly defined for measure-category 15.
Two patterns can be glimpsed from figure 4.4. The first concerns which attribute-
values change from measure-category to measure-category and which stay the
same for instance, there is no rhythmic figure in common between measure-
categor ies 15 and 16. The second patter n concer ns changes in the profile of typi-

15. The lack of variation in mm. 8 and 16 means that only (1114) # 2 merely 759,499,667,166,482
different waltzes can be wr itten using this particular game.
150 analysis and theory

<"%76$"D <"%76$"D
&%#">($E*+P* &%#">($E*+/*

$'E#'<5&* :6<4"$*(B* $'E#'<5&* :6<4"$*(B*

B5>6$"* #5<"7*67")* B5>6$"* #5<"7*67")*

!"!"!"!"!"! !" &

$ "%" P* $ $ $ " +Y *

! ! ! +* ! ! &" +
$ $ $ " $ "%"

!"!"!"!"!"!" +*

!"!" !" !" +*

$"$" $" $

!" !"!"!"! +*
$ "%"

!!" !" !! +*
$ "%"

!"!" !" ! +*
$ "%"

figure 4.4 Values assigned the attr ibute rhythmic figure in measure-categor ies 15 and 16

cality between measure-categor ies as can be seen, there is a relatively sharp con-
trast between the profiles of measure-categor ies 15 and 16. These patter ns, taken
together, create the template that shapes each of the waltzes generated by the game.
A somewhat more complete perspective is provided by looking at the values
assigned to just two attr ibutes harmony and rhythmic figure in the first section of
the waltz, as illustrated by figure 4.5. As can be seen, there is little change in the val-
ues that members of measure-category 1 and measure-category 2 assign to these
attributes: one new rhythmic figure is introduced and one is dropped, but otherwise
the categor ies are (at least with respect to these two attr ibutes) identical. While
measure-category 3 retains most of the rhythmic figures of measure-category 2, the
values assigned the har mony are completely different and markedly more various.
Finally, measure-categor y 4 has significant changes in the values assigned both
attributes: two new rhythmic figures come to prominence, and the attr ibute-values
for har mony change once more.
cate g ori zat i on, st rate g y, and sy ntax 151

/* ! ! ! !"!"!"!"!"! ! ! ! !" !" ! !" !""!""!""! ! ! ! !"!"!"!"!"! !"!"!"!"! !" !"!" !

/* /* U* $ "&" '" $ "&" $ $ $ " $"$"$"$"$"$" $ "&" $ "%" $ "%" $ "&" $ "%"
G* G V* V V VP*

<3*+* ++* 0* 0* +* +* +*

<3*,* ++* 0* 0* +* +* +*

<3*2* P* ,* 2* +* 2* 0* +* +* ,*

<3*0* 0* U* ,* +* 0* 0*

figure 4.5 Attribute-values for harmony and rhythmic figure in measure-categor ies 1
through 4 (left side: the values assigned to the attr ibute harmony by the members of
measure-categor ies 1 through 4 and the number of members within each measure-
category that assign that value; right side: the values assigned to the attr ibute rhythmic figure
and the number of members within each measure-category that assign that value)

There are also changes to the profile of typicality for the measure-categor ies in
this section. While the profile for measure-categor ies 1 and 2 is identical a typi-
cal variant will have a root-position tonic har mony and use either the first or sec-
ond rhythmic figure it is less clearly defined for measure-categor y 3. Although
the number and range of rhythmic figures do not change dramatically, there are
now four possible values for the attr ibute har mony, and only a bit of additional
weight is put on one of these attr ibute-values (the root-position dominant is
assigned by five of the eleven members of the category). Things come back into
focus with measure-category 4 the most typical variant is a first-inversion har-
mony using either of the last two rhythmic figures but the profile of typicality is
still not quite as clearly defined as it was in the first two measure-categor ies.
The syntax of this section (reckoned in ter ms of harmony and rhythmic figure)
is one in which new mater ials are gradually introduced and displace previous mate-
rial. At the same time, measure-categor ies with a fairly clear profile of typicality give
way to ones that have a less clearly defined profile (a trend reversed somewhat by
measure-category 4). Similar syntactic patter ns structure the remaining three sec-
tions of the waltz. The most important features of the syntactic template for the
waltzes produced by the game, taken as a whole, are as follows:16
New attr ibute-values especially for rhythmic figure are introduced
throughout the first three sections. At the beginning of section 4, this process
is temporar ily halted and, coincident with the return C major in m. 13, rhyth-
mic figures are for the first time brought back.
Cadences (such as those that occur in measure-categor ies 7 8 and 15 16)
are marked by a distinctive treatment of attribute-values: values that occur in

16. In developing this character ization of the changes in attr ibute-values and typicality that can be
found in the measure-categor ies collected in ex. 4.2, I considered the attr ibutes of contour, number of
non-chord tones, bass motion, and starting pitch of the melody, as well as har mony and rhythmic figure.
I found, however, that focusing on the latter two attr ibutes provides the clearest picture of typicality
within and across the measure-categor ies. In consequence, my account of the syntactic template
tends to emphasize these attr ibutes.
152 analysis and theory

no other measure-category occur here, and the number of different attr ibute-
values within each measure-categor y is shar ply attenuated. In particular, the
measure-categor ies that cor respond with har monic ar rivals (that is, 8 and 16)
have the most clearly defined profile of typicality of any of the sixteen.
The syntax of sections 1 and 2 involves complementary processes: section
1 begins with measure-categor ies that have clearly defined profiles of typi-
cality; the profiles of the measure-categor ies at the end of the section are less
clearly defined. Section 2 begins with measure-categor ies that have less clearly
defined profiles of typicality; at the end of this section, the profiles of the
measure-categor ies are clearly defined.
The syntax of sections 3 and 4 involves contrasting processes: section 3 has
the broadest range of attribute-values of any of the sections (there are, for
instance, over two times as many different rhythmic figures in section 3 as
there are in section 1) and the most poorly defined profiles of typicality. By
compar ison, the range of attribute-values is much more limited in section 4,
and the clar ity of typicality is perhaps second only to section 1.
The compositional strategy embodied in this syntactic template is one that
would be familiar to aspir ing composers of the eighteenth century, for it is in com-
plete accord with the compositional strateg ies for sixteen-measure dance for ms
proposed in treatises of the per iod. Theorists like Riepel and Koch taught that the
tonic should be plainly stated in the first section (the measure-categor ies with tonic
harmony in this section have the most clearly defined profile of typicality); there
should be clear cadences at the end of the second and fourth sections (in the musi-
cal game, this is accomplished by the distinctive treatment accorded the end of sec-
tions 2 and 4); and the third section should be planned so that it does not become
independent from the whole in Kochs words, it must be for med so that it loses
its completeness and is bound more closely with the following fourth melodic sec-
tion17 something achieved in the template by the large amount of variation
among the members of measure-categor ies 9 through 12.
At a more abstract level, the compositional strategy incor porated in the template
is, of course, similar to late-eighteenth-century conceptions of what is now conven-
tionally called sonata for m.18 These conceptions divided sonata for m into two main

17. Koch, Introductory Essay on Composition, 100. Kochs discussion of sixteen-measure dance for ms is
on pp. 78 117.
18. The classic study of eighteenth-century conceptions of sonata for m is Leonard Ratners Har-
monic Aspects of Classic Form, Journal of the American Musicological Society 2 (1949): 159 68. The via-
bility of eighteenth-century conceptions of sonata for m, and their relevance for recent literature on the
topic of form, have been given a thorough discussion recently in Mark Evan Bonds, Wordless Rhetoric:
Musical Form and the Metaphor of Oration, Studies in the History of Music, 4 (Cambridge, Mass.: Harvard
University Press, 1991). A thoughtful cr itique of Bondss work can be found in Peter Hoyt, Review of
Mark Evan Bonds, Wordless Rhetoric, Journal of Music Theory 38 (1994): 123 43.
It should be noted that recent scholarship has argued for a substantially more complicated view of
sonata for m than that presented by either Ratner or Bonds. For an insightful discussion, see Scott Bur n-
ham, The Second Nature of Sonata Form, in Music Theory and Natural Order from the Renaissance to the
Early Twentieth Century, ed. Suzannah Clark and Alexander Rehding (Cambr idge: Cambridge Univer-
sity Press, 2001), 111 41.
cate g ori zat i on, st rate g y, and sy ntax 153

parts, whose functions were determined by harmonic syntax: the Exposition (which
set out the tonic and, in major keys, closed on the dominant); and the Development
and Recapitulation (which closed on the tonic).19 The Exposition presented the pr in-
cipal mater ials of the movement, much as sections 1 and 2 of the musical game pre-
sent tonic and dominant; the Development departed from these mater ials, as does
section 3 of the game; and the Recapitulation returned to the mater ial of the Expo-
sition, just as there is a reprise of previously used mater ial in section 4 of the game.
Given these similar ities, we can see that the conception of compositional strat-
egy behind the template provides a link between modest sixteen-measure dance
forms and sonata for m: both rely on the same syntax for organizing musical mate-
rials.20 This points out something rather important about the teaching of compo-
sition in the eighteenth centur y: the highly detailed accounts of dance for ms by
Koch and others did indeed set out the means by which modest utilitar ian works
could be generated, but they also schooled students in compositional strategies that
could be applied to larger works. These larger works, then, were not simply gener-
ated by repeated applications of the compositional techniques descr ibed for minuets
and the like, but by the reinterpretation of these techniques as strateg ies for com-
position that could be applied to categor ies of musical events. When these cate-
gories were realized over longer spans of time, extended works (such as those by
Mozart and Beethoven discussed in the following sections) were born.

Typicality Effects and Musical Syntax

What emerges from this analysis of the Musikalisches Wrfelspiel is the significance of
typicality effects for compositional strategy. Important structural points in the syn-
tactic template the establishment of tonic in the opening measures, the cadence on
the dominant at the end of the second section, and the cadence on the tonic at the
end of the fourth section were all marked by measure-categor ies with clearly
defined profiles of typicality. Indeed, a functional distinction between tonic and
dominant can be made simply on the basis of the amount of variation present
among the members of the measure-categor ies associated with each. Taking the
template as a whole, there is less variation in (and thus a clearer profile of typical-
ity for) measure-categor ies associated with the tonic (that is, measure-categor ies
1 2, 4, 13 14, and 16); there is comparatively more variation in (and consequently
a less clear profile of typicality for) measure-categor ies associated with the dominant
(that is, measure categor ies 3, 6, 8, 10, 12, and 15).21

19. In the following, I capitalize Exposition, Development, and Recapitulation when they
refer to components of sonata for m, reserving the uncapitalized versions of the ter ms for the composi-
tional techniques through which musical mater ials are laid out, developed, and reprised. There is, of
course, a certain measure of anachronism in applying these ter ms to eighteenth-century works, especially
when sonata for m is conceived as a two-part structure; nonetheless, the ter ms do single out composi-
tional strategies recognized, if not so named, by eighteenth-century theorists.
20. The absence of an explicit link between the discussion of the appropriate compositional strategy
for dance for ms and that for sonata for m has been noted by Bonds; see Wordless Rhetoric, 26 27.
21. I interpret the C major har mony of measure-category 11 as the subdominant of G major; note
the FS3 in the bass of the sixth variation.
154 analysis and theory

What also emerged was the possibility of describing the syntactic structure for
a two-part form purely in ter ms of a general set of compositional strategies realized
through the manipulation of categories of musical events. There is no doubt that, in
the case of the waltz created by the game, harmonic syntax is of central importance;
nor is there any doubt that those who have descr ibed for m with an emphasis on
harmonic syntax do it with the assumption that this syntax also embodies compo-
sitional strategies applied to other aspects of music. Nonetheless, the account I have
provided shows how certain compositional strategies viewed as a generalized set
of processes on musical mater ials operate independently of the strategies associ-
ated with har monic syntax.
With these thoughts in mind, let us tur n to some actual musical compositions
as opposed to recreational games that make use of some of the compositional
strategies seen in the Musikalisches Wrfelspiel, as well as more novel strategies that
are appropriate for extended and adventuresome instrumental works.

mozart, string quartet in c major

(k. 465), first movement. alleg ro
The Allegro of Mozarts String Quartet in C major (K. 465) was one of Arnold
Schoenbergs favorites for illustrating motivic development of the sort discussed in
chapter 1 of this volume. He found the work appropriate for these pur poses in part
because of the prominence of the motive among the various musical mater ials and
in part because it is easy to find traces of the motive even after it has left the scene
(thus supporting the organic conception of musical composition of which Schoen-
berg was so fond).22 Because my concer n here is not with developing an account of
the way a musical whole is der ived from a single theme, but with the motive as an
element in an unfolding musical discourse, my account approaches the syntax of the
principal melodic motive as independent from syntactic levels occupied with har-
mony and stylistic for mulae, and independent from musical signs or syntagms of the
sort noted by Eco. I shall, however, consider how these are cor related with the syn-
tax of the principal motive toward the end of my analysis.
Mozart completed the quartet in C major dur ing January of 1785, and in Sep-
tember of that year it was published by Artaria in Vienna as one of a set of six quar-
tets dedicated to Joseph Haydn.23 The first movement of K. 465 begins with an
Adagio (mm. 1 22) remarkable for its har monic peregrinations and rather odd
pitch juxtapositions, which together ear ned the work its nickname as the Disso-
nance Quartet. The ensuing Allegro, by contrast, has hardly anything dissonant

22. For a discussion of Schoenbergs analyses of this movement, see the Commentar y by Patricia
Carpenter and Severine Neff in their translation of Schoenbergs The Musical Idea and the Logic, Tech-
nique, and Art of Its Presentation (New York: Columbia University Press, 1995), 53. Schoenberg first devel-
oped his idea of developing variation which he later, and famously, applied to the music of
Brahms in an analysis of this movement made in 1917; see Arnold Schoenberg, Coherence, Counterpoint,
Instrumentation, Instruction in Form, ed. Severine Neff, trans. Severine Neff and Charlotte M. Cross (Lin-
coln: University of Nebraska Press, 1994), 41.
23. For a fuller account of the quartet and its companions, see John Irving, Mozart: The Haydn
Quartets, Cambridge Music Handbooks (Cambr idge: Cambridge University Press, 1998).
cate g ori zat i on, st rate g y, and sy ntax 155

about it. The Allegro is cast in sonata for m, and the pr incipal melodic motive appears
prominently in five places: the opening and end of the Exposition (mm. 23 51 and
91 106, respectively); the Development section (mm. 107 54); and the opening
and end of the Recapitulation, extending into the Coda (mm. 155 70 and 211 46).
My discussion concentrates on the first three of these, with br iefer comments on
the beginning of the Recapitulation and the Coda.

Measures 23 51 of the Allegro

Example 4.3 g ives the first twenty-nine measures of the Allegro. The principal
motive of the movement is presented in the first violin in mm. 23 24, and then
again in mm. 25 26 (symbolized with $ in ex. 4.3). Using an analytical perspective
that will infor m each of the remaining analyses, I shall approach the various for ms
of this motive as a cognitive category, structured around a conceptual model. I shall
leave open the question of the status of these conceptual models that is, whether
they are empir ically verifiable constructs or simply convenient analytical tools
and proceed on the general understanding developed in chapter 1 and br iefly
reviewed at the beginning of this chapter: categor ies are basic structures in ter ms
of which we understand the world, they are organized around conceptual models,
and they represent a common g round where the concer ns of listeners and com-
posers meet.
The model for the categor y consists of three cor related conceptual elements,
character ized as the things necessary for a melodic fragment to count as an instance
of the motive:
a rhythmic patter n that spans two measures of common time:
ch, h q
a rather specific diatonic contour, consisting of four elements (each is
shown in ex. 4.4); in order of succession, they are
(i) a sustained first note
(ii) an ascending stepwise motion through a third
(iii) an upward skip of a third
(iv) a downward step
an implied har monic change from the first to the second measure of the
The contour of the motive gives rise to two distinctive intervals: the fifth spanned
by ii and iii; and the fourth between i and the last note of iv. Although the implied
change in har mony is masked in mm. 23 24 (owing to the violas pedal point), in
most other cases it is quite apparent.
After mm. 23 26, there are three further instances of the motive that cor respond
to this model: two in mm. 31 34 (with the motive accompanied by the second vio-
lin), and one in the cello in mm. 44 45 (with the cellos concluding half note tak-
ing the place of the quarter-note and quarter-rest of the original). Somewhat less-
typical for ms appear in the first violin in mm. 27 28 and in the ascending cascade
of mm. 45 48 (% in ex. 4.3). In each case, the very end of the motive has been
example 4.3 W. A. Mozart, String Quartet K. 465, first movement, mm. 23 51




example 4.3 (continued)












158 analysis and theory

example 4.4 Schoenbergs analysis of the main motive from W. A. Mozart, String Quar-
tet K. 465, first movement

i iii

ii iv

modified: in mm. 45 48, the most significant change is to the end of the rhythmic
figure; in mm. 27 28, both the end of the rhythmic figure and contour patter n have
been changed. One additional version of the motive, heard in the first violin in
mm. 35 36 (& in ex. 4.3), is still less typical, retaining most of the rhythmic patter n
of the original, but only element ii of the contour.
One other bit of musical mater ial (shown with ' in ex. 4.3) seems to demand
affiliation with the motivic category. As but a subcomponent of the motive the
ascending gesture that ends on a strong beat it is hardly typical of the category.
Indeed, hearing the fragment as a member of the category requires that it appear in
close proximity to more typical members of the categor y.When a similar gesture
occurs in mm. 62 63 and 64 65 in the midst of the first subsidiar y theme (as
shown in ex. 4.5), it is decidedly subservient to the alter nation of ascending and

example 4.5 W. A. Mozart, String Quartet K. 465, first movement, mm. 61 66






cate g ori zat i on, st rate g y, and sy ntax 159

descending skips that dominates the musical scene, having lost its connection to the
principal motive. The change in the way we hear this fragment in mm. 48 51,
as belong ing to the categor y that includes the pr incipal motive; in mm. 62 65, as
a rather gener ic bit of filler is an example of the phenomenon of conceptual slip-
page noted in chapter 3, but one applicable to purely musical concepts. The context
that induces this slippage is the absence of the principal motive and the presence of
new mater ial associated with the first subsidiary theme.
The melodic syntax for these opening measures, then, involves setting out the
most typical members of the category first indeed, these provide the basis for the
conceptual model for the categor y.24 Less typical members of the categor y follow
(in mm. 27 28 and 35 36), and, in mm. 37 43, the motive temporar ily disappears
altogether (save for the ' fragments). It then reappears in its most typical for m in
m. 44, immediately followed by three slightly less typical versions. After this, there
are further ' fragments that lead to the gener ic cadential mater ial that prepares for
the entrance of the first subsidiary theme in m. 56.
Taken as a whole, the motive forms presented in mm. 23 51 are a relatively nor-
mative Type 1 category.What is str iking, however, is the dynamic shape of the cat-
egory as it is presented over the course of time. Although typical members domi-
nate the musical texture in the first half of the section, they momentar ily disappear
from the texture in the second half , to be brought back at an important moment
preceding the transition to the first subsidiary theme. The importance of the prin-
cipal motive for the melodic syntax of mm. 23 51, then, is less a function of its con-
stant presence and more a function of its memorability and cognitive utility. These
make possible a musical syntax of both exposition and development: exposition as
the motive is first established in its most typical for m; and development as less-
typical variants are then introduced.

Measures 90 106 of the Allegro

After its gradual departure in mm. 44 51, the motive disappears completely for
forty measures, during which time two subsidiary themes are presented in the dom-
inant. When the motive does return, it is at a moment of opportunity, following a
final emphatic cadence of the sort that usually signals the conclusion of a section
or that prepares the way for additional thematic mater ial. By bringing back the
main motive at this point, Mozart ties together the beginning and end of the Expo-
sition. The bonds are relatively loose ones, however, for it is a variant of the main
motive rather than the or iginal that is introduced, and the compositional strategy
applied to the statements that follow is closer to development than it is to exposi-
As shown in example 4.6, the motive enters in the second violin in m. 91, just
as the first violin achieves its cadential G5.25 The first violin responds by taking up

24. As noted in chap. 1, it is also possible for the conceptual model for a category of motive forms
to emerge over time, although the longer the process is delayed, the more the conceptual model comes
into conflict with the global model for a musical theme.
25. The sustained note at the beginning of the motive allows Mozart to introduce it under the cover
of cadence, something he exploits at various developmental moments throughout the movement.
160 analysis and theory

example 4.6 W. A. Mozart, String Quartet K. 465, first movement, mm. 90 106










cate g ori zat i on, st rate g y, and sy ntax 161

the motive as well, beginning a short game of motivic tag in which some for m of
the motive enters in each of the succeeding measures. Here, the initial motive forms
are of the % variety, highly similar to the or iginal for m but with a slightly modified
ending. Although har monic change is still evident, the constant overlap of motives
limits the har monic possibilities to the tonic and dominant of G major. The sense
of harmonic progress that accompanied the or iginal statement of the motive is
markedly attenuated: contr ibuting factors are the constant alter nation between
tonic and dominant, the cellos pedal point in mm. 91 95, and the halving of har-
monic rhythm that follows in m. 96.
This reprise of the motive bears witness to new, and somewhat more exten-
sively varied, members of the motivic category. In mm. 93 94 and 94 95, the sec-
ond and first violins present embellished versions of the motive forms they have
just stated (indicated with a & in ex. 4.6); were it not for the temporal proximity
of these to the preceding motive forms, it might be difficult to make a case for
their der ivation from the or iginal. More obvious is the inverted for m of the motive
that appears in the first violin in mm. 96 99, and which is then used in the simul-
taneous presentation of motive forms in all four instr uments in mm. 103 104.
Finally, the ' fragment that provided a measure of unity at the opening of the
exposition reappears in mm. 100 103 to serve as a link between the motive forms
that precede and follow it.
Given the absence of the or iginal for m of the motive in these measures, a case
could be made for modifying the conceptual model for the category so that variants
of the sort shown with a % in example 4.6 would be regarded as typical (indeed,
such a move would be the first step along a slippery slope of motivic development
of the sort that Wagner leads us down with his transfor mations of the Leidensmo-
tiv). Nonetheless, retaining the sense that we have moved away from what is typi-
cal of the category is important for what happens next. If the repeat sign of m. 106
is observed and the Exposition reprised, the typicality of the original version of the
motive will be reaffirmed: we will have moved from the edges of the category (as
represented by mm. 91 106) back to the center. If we instead proceed from m. 106
into the Development, we will move yet farther toward the fr inges of the category.

Measures 107 35 of the Allegro

As shown in example 4.7, Mozart begins the Development with a slightly modified
version of the or iginal motive, but with changes now made to the middle of the
motive rather than to the end: he starts the ascending-third figure a step higher than
the sustained note, instead of repeating the sustained note at the beg inning of the
figure. As a result, the motive, as a whole, now spans a fifth and implies no change of
harmony. The chromatic alteration in the first violin at m. 111 moves the motive a
bit further away from its typical for m, but it is the version in the viola at mm. 112 13
that accelerates the process of transfor mation. In this version (indicated with a & in
ex. 4.7), the ascending-third figure is replaced with an ar peggio, which becomes a
feature of the following motive forms. At m. 117, the arpeggio figure takes over, and
the subsequent dissolution of the motive leads to a momentar y cadence on the
dominant of A minor in m. 121.
example 4.7 W. A. Mozart, String Quartet K. 465, first movement, mm. 107 35








example 4.7 (continued)





164 analysis and theory

At m. 121, the cello returns to the for m of the motive used at the beginning of
the Development (but now within the context of A minor). After two statements,
the rising arpeggio figure once again replaces the ascending third within the
motive, after which point it displaces the main motive or its variants altogether (as
indicated by the preponderance of ' symbols after m. 125 in ex. 4.7). In m. 147
(which is not g iven in ex. 4.7), the for m of the motive used at mm. 107 and 121
makes a final appearance (beg inning with the cellos G2, but pointing toward
C minor). Given the events that followed the previous two appearances of this
form of the motive, this third appearance seems to suggest yet another episode of
development. Nonetheless, in m. 151, the motive yields to an ascending cascade of
arpeggio-figures (all outlining a G dominant-seventh), which br ings the Develop-
ment to a close.
As was the case in the Exposition, in the Development more typical for ms of the
motive are followed by less typical for ms.What is distinctive is the repetition of this
pattern in close order (which occurs three times within the forty-odd measures of
the Development), and the emergence of the ascending ' figure as the means for
Mozart to advance his musical argument. While the slightly varied for m of the
motive that marks the beginning of each patter n has prominence as a point of ref-
erence for this musical argument (and thus represents what is typical of thematic
mater ial), it lacks the sense of balanced har monic progression summoned by the
original for m of the motive. Given the variation of the motive that takes place over
the course of mm. 107 35, however, just what is typical of the or iginal motive is
something of an open question by the time we arrive at the dominant-seventh of
mm. 151 55.

Measures 155 60 and 211 46 of the Allegro

The question of motivic identity posed by the Development is answered by the
understated but nonetheless confident return of the original motive in mm. 155 58,
shown in example 4.8. Although the accompaniment is somewhat different than
that used in mm. 23 26, the motive itself is unchanged from its or iginal for m: this
is the motive at its most typical. Mozart is not content to leave it at that, however.
The accompaniment for the second statement of the paired motives, in mm. 163 66,
is not only fuller, but it br ings back the overlapped motive forms used at the end of
the Exposition and at the beg inning of the Development. These slightly varied
forms of the original (indicated with % in ex. 4.8) recall the developmental cast of
the earlier passages, as do the ascending ' figures in the cello. Mozart nonetheless
quickly tur ns away from any further dissolution of the main motive and instead
moves with dispatch toward the first subsidiary theme, which commences at m. 176.
The end of the Recapitulation restates the end of the Exposition almost exactly:
mm. 91 106 are transposed into C major for mm. 211 26. From this point, the
music proceeds either back to the beg inning of the Development (if the repeat is
taken) or forward to the Coda (which occupies mm. 227 46). As shown in exam-
ple 4.9a, the latter grows out of the opening of the Development but uses yet
another modification of the original motive. This version owes much to that used
in mm. 211 13 (the transposed quotation of mm. 91 93), but applies a chromatic
example 4.8 W. A. Mozart, String Quartet K. 465, first movement, mm. 155 70





166 analysis and theory

inflection to the note at the beginning of the ascending-third figure. A further vari-
ant then concludes the movement, which is shown in example 4.9b: the end of the
motive, shortened from the or iginal half-notequarter-note figure to the successive
quarters of mm. 227 29, is shortened yet again to two successive eighth-notes. Al-
though a similar figure occurs in mm. 229 30, that variant refers back to mm.
27 28, for the successive eighths lead immediately to additional mater ial (in the
case of mm. 229 30, a repetition of the ascending-third figure). In mm. 242 45, the
eighths simply represent a compression of the original motive, and it is with this
abbreviated for m of the motive that the movement ends.
Although the concluding section recalls the opening of the Development,
Mozart allows the prospect of further divergence from the main motive to linger
for only a moment before banishing it with a vigorous and tr iumphant cadence
in C major in mm. 234 35. Nonetheless, it cannot be said that the movement
ends with an unequivocally typical for m of the motive: if the or iginal for m of the

example 4.9 W. A. Mozart, String Quartet K. 465, first movement: (a) mm. 227 30;
(b) mm. 242 46





cate g ori zat i on, st rate g y, and sy ntax 167

motive represents the stable center of the cognitive categor y of motive forms, the
conclusion does not return us to that center. Two factors infor m the ending; both
are related to global models, which provide a context for the conceptual model
that defines what is typical of this motivic categor y. First, the ending is har mon-
ically quite stable: the final version of the motive introduced in mm. 242 45 re-
inforces this stability by outlining scale deg rees 5 and 1 and by placing the ar rival
on 1 as close as practicable to the fir st beat of the metr ic cycle while retaining
the appogiatura figure character istic of the motive. Second, the most typical
version of the motive, in accordance with the global model for music themes rel-
ative to which it is framed, represents musical mater ial to be acted upon it is a
theme in the sense of a topic for discourse. At this point in the movement, how-
ever, discourse must be concluded: what is required is not a highly typical version
of the main motive but a reminiscence of the motive suited to the pur poses of

Compositional Strategy, Categor ization,

and the Allegro
Although I chose not to emphasize it in the preceding discussion, the overall com-
positional strategy that shapes the Allegro (as opposed to its pr incipal motive alone)
has, not sur prisingly, a high deg ree of conformance both with models for sonata
form proposed by music theor ists dur ing the late eighteenth century and with the
syntactic template for the musical game discussed above. The first main section of
the work (mm. 23 106, which I have called the Exposition) moves from tonic to
dominant, with themes or thematic groups for each key area. The second main sec-
tion (mm. 107 204) begins by moving through a number of different key areas (the
Development) and then tur ns to music that emphatically confir ms the tonic (the
Recapitulation). The syntax of this for m is articulated by important cadential
arrivals (such as the ar rival on the tonic in m. 44 or the ar rival on the dominant in
m. 91), and although melodic themes are important for projecting a key and ren-
dering a particular key area distinctive, they are not the dr iving force behind this
portion of the musical discourse.
Running parallel to this for mal syntax, but independent from it, are the compo-
sitional strateg ies that shape the categor y compr ising the various versions of the
principal melodic motive of the Allegro. While the most typical members of this
category are associated with tonic (and, in their typicality, serve to emphasize the
importance of tonic within the for mal syntax), members with a high degree of con-
formance to the conceptual model for the category are also used to initiate presen-
tations of the motive at points not associated with the tonic (such as the end of the
Exposition or beg inning of the Development). Each of these presentations of the
motive (whether on or off the tonic) leads to an increasing number of less typical
members of the category, a process that ends either with the introduction of a more
typical for m (as in m. 44 or m. 155), or with a move toward new thematic mater ial
(as happens following m. 49). Through the accumulating evidence of these processes
we come to understand the motive as a basis for Mozarts musical discourse
168 analysis and theory

something to be introduced, varied, and ultimately reprised. A similar understand-

ing does not develop in connection with the subsidiary themes first introduced in
the dominant dur ing the Exposition; instead, these are melodic mater ials with the
highly localized syntactic function typically attr ibuted to such themes by eighteenth-
century theor ists.
Understanding the pr incipal motive as the basis for Mozarts musical discourse,
however, is not the same thing as regarding it as the generator of discourse through-
out the movement. As I pointed out in connection with example 4.5, snippets of
material reminiscent of some other aspect of the motive can be found elsewhere in
the movement.26 Nonetheless, without the context provided by more typical mem-
bers of the motivic category, we tend not to hear such fragments as related to the
principal motive and thus do not hear them contr ibuting to the discourse of which
the motive is a part.
It is important to emphasize that the syntax associated with the pr incipal motive
of the Allegro is not entirely unique to this movement. As Robert Gjerdingen has
shown, composers of the late eighteenth century relied on a variety of stylistic for-
mulae for both the substance and shape of their melodic and har monic mater ial,
and these for mulae infor m the compositional strateg ies applied to the motive.27
However, the way this particular motivic category unfolds how many typical
members of the category are presented before variation ensues, the relative typical-
ity of subsequent members, and the amount of time before more typical members
return is something specific to the Allegro.
The principal motive of the Allegro should also be understood as an instance of
the sort of musical topics to which eighteenth-centur y composers regularly made
recourse (and to which Eco referred in his acknowledgement of musical signs).
Mozarts motive is a good example of the singing style as character ized by Leonard
Ratner, and the lyr ical presuppositions of this topic do much to shape just what
components of the motive are called on for the pur poses of variation.28 Nonethe-
less, I think Mozarts motive is also more than simply a musical topic, for in its sta-
tus as a remarkable musical categor y remarkable in the variety of members the
category compr ises, and remarkable in the way the manipulation of these members
contributes to Mozarts overall musical argument it transcends topicality to be-
come an essential element within the musical discourse peculiar to the Allegro of
the C major quartet.

26. A thorough analysis that makes an argument for the omnipresence of the principal motive can
be found in Bonds, Wordless Rhetoric, 102 10.
27. See in particular Robert O. Gjerdingen, A Classic Turn of Phrase: Music and the Psychology of Con-
vention, Studies in the Cr iticism and Theor y of Music (Philadelphia: University of Pennsylvania Press,
1988); idem, Revisiting Meyers Grammatical Simplicity and Relational Richness, in Cognitive Bases
of Musical Communication, ed. Mari Riess Jones and Susan Holleran (Washington, D.C.: American Psy-
chological Association, 1992), 225 43; and idem, Courtly Behaviors, Music Perception 13 (1996):
365 82.
28. For examples and discussion of musical topics, see Leonard G. Ratner, Classic Music: Expression,
Form, and Style (New York: Schirmer Books, 1980), 9 29. For a discussion of musical topics of this sort
within the specific context of musical semiotics, see Victor Kofi Agawu, Playing with Signs: A Semiotic
Interpretation of Classical Music (Princeton, N.J.: Princeton University Press, 1991), 17 34.
cate g ori zat i on, st rate g y, and sy ntax 169

beethoven, string quartet in bb major,

op. 18 no. 6. first movement
Joseph Kerman devoted the first quarter of his classic study of the Beethoven str ing
quartets to the Op. 18 set (which date from 1798 to 1800), rightly believing that
these early works held important keys to understanding how Beethoven trans-
formed what had been a relatively modest genre into the ne plus ultra of chamber
music.29 But Kerman had little patience with the first movement of the sixth quar-
tet, for he found scant evidence of the sort of motivic development that would
become the hallmark of Beethovens mature style.30 In truth, the opening motive of
the work undergoes hardly any transfor mation over the course of the movement: its
syntax is not easily explained in ter ms of relative degrees of typicality among cate-
gory members. What is involved instead is a categor y that includes two distinctly
different for ms of the motive, forms distinguished not only by the mater ials from
which they are fashioned but also by the way they are used. The different uses of
these two forms suggests a higher level of syntax that includes both, and this in tur n
gives a glimpse into the process by means of which musical topics, of the sort dis-
cussed by Ratner and alluded to by Eco, are formed.
Ratner, for his part, character ized this movement as pure buffa, a dancing car-
toon.31 As can be seen in example 4.10, the two forms of the principal motive con-
tribute much to the effect, providing a musical Tweedledum and Tweedledee to the
listeners Alice.32 The first of these ($ in ex. 4.10) sketches a rapid ascent through a
broken ar peggio on the tonic and, aided in the first case by a slightly overemphatic
fortepiano, sets out the first beats of mm. 3 and 5 as points of arrival. The second for m
of the motive (% in ex. 4.10), suave and somewhat understated, introduces the dom-
inant and is metr ically more open. In mm. 7 and 9, it ends on the second main beat
of the measure; in its third appearance, it ends on the first beat of the measure, but
only after its ending is extended by mm. 12 13. The overall har monic contour of
the % form is also relatively flat: in mm. 5 7, for instance, it begins and ends on Bb2,
while m. 6 is preoccupied with a simple alter nation of the fifth and root of the
dominant chord.
29. With this character ization I mean to take nothing away from the accomplishments in this genre
by Haydn or Mozart, but only to suggest that it is with Beethoven that the quartet eventually becomes
a site for the active contestation of musical values and that this represents a new possibility for chamber
music. See, for instance, Daniel K. L. Chua, The Galitzin Quartets of Beethoven: Opp. 127, 132, 130
(Princeton, N.J.: Princeton University Press, 1995). The issue for musicologists, then, is the extent to
which the early quartets prepare the way for this transfor mation of the genre.
30. Joseph Kerman, The Beethoven Quartets (New York:W.W. Norton, 1966), 71 74.
31. Leonard G. Ratner, The Beethoven String Quartets: Compositional Strategies and Rhetoric (Stanford,
Calif.: Stanford Bookstore, 1995), 89.
32. In Through the Looking Glass, Alice quotes a nursery rhyme about Tweedledum and Tweedledee,
and, after hear ing this, the two brothers reply:
I know what youre thinking about, said Tweedledum; but it isnt so, nohow.
Contrar iwise, continued Tweedledee, if it was so, it might be; and if it were so, it would be; but
as it isnt, it aint. Thats logic.
Lewis Carroll, The Annotated Alice: Alices Adventures in Wonderland and Through the Looking Glass,
introduction and notes by Martin Gardner, illustrated by John Tenniel (New York:W.W. Norton, 2000),
181. In a note to this passage, Gardner observes that the nursery rhyme may descend from a bit of dog-
gerel by John Byrom (an eighteenth-century hymn writer and teacher of shorthand), which recounted
the rivalry between George Fr ideric Handel and Giovanni Battista Bononcini.
example 4.10 Ludwig van Beethoven, String Quartet, Op. 18 No. 6, first movement,
mm. 1 30

Allegro con brio

Violino I

Violino II







cate g ori zat i on, st rate g y, and sy ntax 171

example 4.10 (continued)







Commenting on the deliberate and somewhat stagy entrances of $ and % is a

fragmentar y form of the motive (& in ex. 4.10). This consists of only a tur n and a
leap (first ascending, then descending); its entrances reverse the expansion of the
distance between strong metr ic accents occasioned by the extension of mm. 12 13,
shrinking it first to one-measure and then (in mm. 17 18) half-measure intervals.
This process of acceleration culminates in an ar rival on the dominant in m. 18,
embellished with a run in the cello that returns the music to the tonic and to the
$ form of the motive.
To better understand the distinctive features of and relationships between these
three forms of the motive, as well as their progress through the movement, let us
consider conceptual models for each; diagrams for the models are given in figure
4.6. The conceptual model for the categor y of $ forms consists of five cor related
conceptual elements, character ized as the things necessar y for a melodic fragment
to count as an instance of this version of the motive:
the rhythmic patter n C qqqq hd
the motive begins with a tur n figure (here, around 5)
the continuation of the motive involves a broken ar peggio with ascending
contour (here beginning on 3 and ascending through a sixth to 1)
172 analysis and theory
%3* 43 &3

(" !"!"!"!" !""!""!""!" #" !" (" !"!"!"!" !"

(" !"!"!"!" !""!""!""!" #)"
$'E#'<5&*?%##"$:* $'E#'<5&*?%##"$:*

4">5:7*=5#'*#6$:* 4">5:7*=5#'*#6$:* 4">5:7*=5#'*#6$:*

B5>6$"* B5>6$"* B5>6$"*

" =5#'*5:#"$:%9*7@5?* # ":)7*=5#'*%7&":)5:>*
&(:#(6$* (B*%*B5B#'*

:(*&'%:>"*(B* '%$<(:E*%9#"$:%#"7* :(*&'%:>"*(B*

'%$<(:E* GDVDG* '%$<(:E*

<"9()E*=5#'* 7#%#")*5:*)5%9(>6"* 7#%#")*5:*)5%9(>6"*


figure 4.6 Conceptual models for the three forms of the principal motive from
Beethoven, String Quartet, Op. 18 No. 6, first movement, mm. 1 30: (a) the $ form;
(b) the % form; (c) the & form

there is no change of harmony implied by the motive

the motive is presented as a melody supported by accompaniment in the
remaining str ings33
Each of these elements could, of course, be broken down into smaller parts. This is
particularly true of the rhythmic patter n, which is extended by the % form of the
motive and truncated by the & form. As indicated by figure 4.6b, the conceptual
model for the % form retains most of the first two components but alters the last
three. The & form continues this process, retaining only the tur n figure that is the
motives most character istic feature.
Slightly varied restatements of the $ and % forms of the motive in mm. 19 29
the texture thickened somewhat, and with the two violins taking part in the dia-
logue proper to the % form of the motive lead to a long pedal on the dominant
of F major preparatory to the statement of the second theme, which commences
in m. 45.What is remarkable, from the perspective on musical syntax I have devel-
oped thus far, is the nearly complete lack of typicality effects among these mater i-
als.While it would be quite possible to include them all in a single category for the
motive, framed relative to a conceptual model that combines the essential elements
of those included in figure 4.6, there does not seem to be a compelling argument
for regarding either the $ and % form as more typical of the category. A better case
could be made for conceiving of the & form as a less typical member of the cate-
gory (if only on account of its fragmentar y nature), but it is a bit hard to say just

33. There are, of course, other ways the motive could be character ized. For instance, the rhythmic
figure I identify does not include the very first note of the piece. My chief reason for seizing upon this
figure is for compar ison with other for ms of the motive. Should one wish to include the first note, the
remaining for ms of the motive would represent even more radical departures from the standard estab-
lished by the $ form.
cate g ori zat i on, st rate g y, and sy ntax 173

what it is less typical of . It could be der ived from either the $ or the % form, and
without further variants it remains somewhat difficult to settle the question.
Subsequent to a cadence on F major in m. 80 that wraps up the exposition of the
second theme, the % form of the motive reappears, as shown in example 4.11.
Although it has been transposed to F, and its third statement in the cello is accom-
panied by parallel tenths in the viola and by a reprise of the passagework from the
bridge in the first violin, it is otherwise exactly the same as when it appeared in the
context of the first theme g roup. This restatement has two consequences. First, it
confir ms the character of the % form of the motive that was established in the
opening measures that is, it reinforces the conceptual model outlined in figure
4.6b. Second, it draws out a function of % implicit in the initial exposition of motive
forms, which was to close off musical discourse.Where the $ form of the motive
sets out the conditions for musical discourse establishing a two-measure hyper-
meter and opening the melodic range to cover more than two octaves the %
motive, with its pedestr ian alter nation of tonic-dominant-tonic and restricted
range, provided closure. Following m. 80, the motive has a similar function, but now
it serves to close off the discourse of the entire Exposition.
The syntactic functions of the $ and % forms of the motive are due in part to

example 4.11 Ludwig van Beethoven, String Quartet, Op. 18 No. 6, first movement,
mm. 79 87






174 analysis and theory

their resemblance to similar figures in other compositions: $ has all the features of
a prototypical opening gesture within eighteenth-century style, and % has much in
common with prototypical closing gestures. It should be emphasized, however, that
how such motives are used is also important. As Janet Levy has pointed out, Joseph
Haydn, at least on occasion, used closing gestures as his opening mater ial.34 The use
of a closing gesture in this way is in part comic it creates a piece that seems to end
before it has begun but it also makes possible some marvelously ambiguous
seams: one is not sure whether a particular passage is a beginning or an ending. The
possibility of this ambiguity testifies to the importance of the way musical mater i-
als are used within a particular composition, as distinct from how similar mater ials
are used in other compositions. For a gesture to signify opening, it has to be used
in openings; for a gesture to signify closing, it has to be used for closings.
The $ form of the motives function as an opening gesture is confir med when
it initiates the Development, which begins with a tutti statement of the motive,
immediately echoed by a solo statement in the first violin (as shown in ex. 4.12).
These slightly less than typical versions of the motive (which has heretofore always
appeared as melody with accompaniment) are followed immediately by the & form,
which tosses a repetition of the head of the $ form just heard back and forth
between the upper str ings. In mm. 103 110, two further variants of the $ form
appear, which extend the motive both metr ically (to four measures) and intervalli-
cally (to a compass of an eleventh, and then a twelfth). These expansions lead to G
minor, which initiates a rapid cycle of keys explored through the medium of the
thematic mater ial that first served as a br idge between Bb and F back in mm.
33 43.
In m. 139, there is a weak cadence on F, followed immediately by a reappearance
of the % form of the motive (shown in ex. 4.13). Stated, as it is, in F major, it recalls
the end of the Exposition; however, this particular version does not confor m com-
pletely to the conceptual model diag rammed in figure 4.6b. The rhythmic profile
has been changed and is now the same as that of the $ form; the dialogue presen-
tation typical of % forms is absent, replaced here by one unaccompanied and one
accompanied statement. Nonetheless, although close to thirty measures of the
Development remain, this entry of the % form of the motive signals the end of the
harmonic exploration initiated by the variant $ forms in mm. 103 10: F major hav-
ing been regained, the remainder of the Development simply serves to postpone the
inevitable return to the opening mater ial. Just as it has each time before, the % form
of the motive confir ms a closing of harmonic structure.
With the return to Bb in m. 175, there is an exact reprise of the first eighteen
measures of the movement, after which the & form of the motive is presented in a
series of close-order repetitions that initiate the br idge to the second theme. At the
conclusion of the movement, the % form returns once more, as it did in
mm. 80 86, only this time confir ming a final close on Bb.

34. Janet Levy, Gesture, Form, and Syntax in Haydns Music, in Haydn Studies: Proceedings of the
International Haydn Conference, Washington, D.C., 1975, ed. Jens Peter Larsen, Howard Serwer, and James
Webster (New York:W.W. Norton, 1981), 355 62.
example 4.12 Ludwig van Beethoven, String Quartet, Op. 18 No. 6, first movement,
mm. 91 110

91 2















176 analysis and theory

example 4.13 Ludwig van Beethoven, String Quartet, Op. 18 No. 6, first movement,
mm. 138 43


As Kerman noted, there is little development of the principal motive in the first
movement of Beethovens Quartet, Op. 18 No. 6. This is not to say that the motive is
uninvolved in the syntax of the movement, however, only that the musical discourse
is not, in any significant way, about this motive. Thus the $ form of the motive sig-
nals an opening for musical discourse, but it is not discourse about $. Even when the
$ form makes an attempt to grab center stage at the beginning of the Development,
the effort ends up being little more than comedic bluster: knowing only how to
begin, it cannot quite figure out how to continue, and after two attempts (in
mm. 102 10), it gives up. Similarly, the % form of the motive closes off discourse, but
it is not always a discourse in which % has been involved (as in its appearances at the
end of the Exposition and Recapitulation, and in the Development). Thus the two
motive forms, by virtue of the way they are used, contr ibute to a larger syntax con-
cerned not with them but with the progress of the movement as a whole.
Despite their similar ities, I believe these two motive forms are distinct enough in
their features to be regarded as separate, though related, categories (something illus-
trated by the conceptual models of fig. 4.6). The syntactic function that accrues to
each over the course of the movement with the $ form used only for openings,
and the % form used only for closings gives further support to this distinction.
Nonetheless, I regard distinguishing the motive forms on the basis of syntactic func-
tion to belong to a later and more theoretical stage of understanding. Accordingly,
categor izing the motives is relatively immediate, while understanding the way they
function within the work emerges only over time.
As I have noted, these two forms of the motive bear resemblance to stylized
musical gestures that signal openings and closings. However, I would argue that, at
least on the abstract level concer ning processes that pertain to musical categor ies,
their function within the movement does not der ive from these gestures. To be
blunt, I do not believe knowledge of such stylized figures is necessary to understand
Beethovens musical argument. That it would enr ich this understanding I have no
doubt, but this is not the same as asserting such knowledge as a precondition for
making sense of the discourse of the movement. Instead, the function of the $ and
cate g ori zat i on, st rate g y, and sy ntax 177

% forms of the motive is an emergent property consequent to the way they are con-
stituted (as contrasting yet similar motivic structures) and the way they are used
within this movement.
The notion that Beethovens use of these motives bestows particular functions
on them further suggests how musical topics come to signify the things they do.
Granted, the musical mater ial for any topic needs to meet certain prerequisites of
harmony and counter point and confor m to general expectations about affect. But
beyond this, one must acknowledge the significance of the way composers have
used particular musical gestures that is, their prevalence within a certain histor i-
cal and cultural milieu in establishing the things these gestures come to signify.
Put another way, signification is not an inherent property of musical mater ials but
derives from how they are deployed in particular works of music that is, from the
syntactic function with which they are invested.

beethoven, string quartet in f major,

op. 18 no. 1, first movement
Where the discourse of the first movement of Beethovens Bb major quartet tends
not to be about its opening motive, the discourse of the first movement of the F
major quartet has difficulty being about anything else: as A. B. Marx famously
observed, the opening motive occurs no fewer than 131 times within the 427 mea-
sures of the movement.35 Within this multiplicity, two categor ies of motive forms
can be distinguished, differentiated both by their constituent features and by the
compositional strategy that is applied to them. The distinction between these two
categories becomes especially plain in the Development, when a new and highly
dynamic variant of the motive appears that combines features of both. The impor-
tance of this variant for developing Beethovens musical argument is affir med in the
Coda, which revisits all three forms of the motive.
The compositional strategy associated with these various motive forms is rela-
tively complicated, and it was not completely successful in Beethovens first attempt
at the movement, a version descr ibed in great detail in a thorough study by Janet
Levy.36 In what follows, I discuss how this strategy is manifested in the Exposition,
Development, and Coda of the second (and final) version of the F major quartet, and
then turn to Beethovens first version to explore how it was brought to its final for m.

The Exposition
In a manner similar to the first movement of the Bb major quartet, the F major
quartet opens with a pr incipal motive that takes two distinctly different for ms. The
first ($ in ex. 4.14) is assertive, direct, and ends on the first beat of the measure. The
second (% in ex. 4.14) is voiced in a lighter manner, is har monically fluid, and ends

35. Adolph Ber nhard Marx, Ludwig van Beethoven: Leben und Schaffen, 6th ed., ed. Gustav Behncke
(Berlin: Otto Janke, 1908), 203.
36. Janet M. Levy, Beethovens Compositional Choices: The Two Versions of Opus 18, no. 1, First Move-
ment, Studies in the Cr iticism and Theory of Music (Philadelphia: University of Pennsylvania Press, 1982).
example 4.14 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
mm. 129

Allegro con brio

Violino I

Violino II









cate g ori zat i on, st rate g y, and sy ntax 179

on the third beat of the measure. Joining these is a fragmentary form (& in ex. 4.14)
whose character is less distinctive, and which plays a much more subsidiary role in
the Exposition. This fragment consists of little more than the stylized tur n figure
common to all for ms of the motive (and which was almost certainly the basis for
Marxs impressive tally). It is this feature of commonality that, as Ratner has noted,
allows the motive to fit seamlessly into a variety of different contexts.37
As with the preceding analyses, we can character ize the distinctive features of
and relationships between various motive forms by consider ing conceptual models
for each; diagrams for the models are given in figure 4.7. The conceptual model for
the category of $ forms consists of five cor related conceptual elements, character-
ized as the things necessar y for a melodic fragment to count as an instance of this
version of the motive:
the rhythmic patter n T qd q !!
the first part of the motive outlines a stylized tur n figure
the motive ends with a small downward skip
the motive is stated tutti
the motive has a relatively static har monic context
Save for the change in dynamics from the first to the second pair of statements
piano replaced by forte all four of the statements of this for m of the motive in the
opening measures confor m to this model. Put another way, the members of the cat-
egory presented in mm. 1 4 and 9 12 assign a relatively restricted range of values
to the attr ibutes picked out by the conceptual model.
Where the members of the $ motivic category are all quite similar to one
another, the members of the % category are rather more diverse. In mm. 13 18, the

%3 43 &3*

* * *
+" !)" !"!"!" !" !" , , " +" !)" !"!"!" !" #" !" +" !)" !"!"!" !"
$'E#'<5&*?%##"$:* $'E#'<5&*?%##"$:* $'E#'<5&*?%##"$:*
B5$7#*?%$#*(B* B5$7#*?%$#*(B*
<(#58"*(6#95:"7*7#E95;")* <(#58"*(6#95:"7*7#E95;")* B5$7#*?%$#*(B*
#6$:*B5>6$"* #6$:*B5>6$"* <(#58"*(6#95:"7*7#E95;")*

! <(#58"*":)7*=5#'*
" 6?=%$)*7@5?C*B(99(=")*

7#%#")*#6##5* =5#'*%&&(<?%:5<":#

$"9%#58"9E*7#%#5&* $"9%#58"9E*B965)*
'%$<(:5&*&(:#"Z#* '%$<(:5&*&(:#"Z*

figure 4.7 Conceptual models for the three forms of the principal motive from
Beethoven, String Quartet, Op. 18 No. 1, first movement, mm. 1 29: (a) the $ form; (b)
the % form; (c) the & form
37. Ratner, The Beethoven String Quartets, 12 13.
180 analysis and theory

first measure of each statement of the motive is unaccompanied, an appogiatura

over a fully-diminished seventh chord is introduced (in mm. 14 and 16), and the end
of the last statement (in mm. 17 18) has a slightly modified rhythmic profile.38 Ver-
sions of the motive that appear in mm. 21 26 are similar to those of mm. 13 18,
but include echoes provided by the & fragment as part of their accompaniment. The
category of % motives is, as a consequence, much more diverse: category members
represent a relatively wide range of attribute-values.
In the remainder of the Exposition, not only are these two categor ies of motives
maintained but also Beethoven continues to treat them as he did in laying out the
first key area. The $ motives (for the most part) continue to stay close to the con-
ceptual model around which their motivic category is structured, and the % motives
(for the most part) continue to diverge in various ways from the conceptual model
around which their motivic category is structured. The Exposition thus reinforces
both the structure of the two categor ies and the compositional strateg ies through
which they were first presented.
After the opening thematic group, the $ form of the motive is prominent in four
places, putting in appearances twice in the transition to C major: once in the mid-
dle of the second thematic group, and preparatory to the cadences that conclude the
exposition. In each case, one of two relatively minor variants is used.
The first variant, which initiates the transition that begins in m. 30 (as shown
in ex. 4.15a), differs from the model in that it is stated solo against accompa-
niment and in alter nation with a competing figure in the first violin and then
concludes with a skip down an octave. From a har monic perspective, this
variant is more static than the or iginal $ form in that it prolongs but a single
pitch-class (here, C) through the combination of the tur n figure and an
octave skip; Ratner has noted that, in this for m, the motive evokes the sound
of a musette.39
The second variant of the $ form of the motive, introduced by the viola
in mm. 41 42, is similar to the first, but it concludes with a repeated note
instead of an octave skip (as shown in ex. 4.15b): rather than a single pitch-
class, now only a single pitch (Eb3) is prolonged. Appearing against the $
form is a % form (in the first violin), whose features are discussed below.
This same variant of $ is used by the cello in mm. 72 77 in the midst of
the second theme group (see ex. 4.15c).40 An interesting complication here is
that the figure with which this appearance of the variant competes (stated in
the first violin) can easily be heard as a further variant of the $ form of the
motive: admittedly, only the rhythmic profile and the beg inning of the tur n
figure are retained (and, unlike other $ variants, this one is sequenced up-
ward), but placed against the second $ variant the similar ities are striking.

38. In the first version of the quartet, the initial statement of the % form of the motive (in mm. 5 6)
was also unaccompanied. The difference in texture, however, is less str iking than the clar ification of har-
monic context provided by the accompaniment Beethoven added, which helped emphasize the contrast
between the $ and % forms.
39. Ratner, The Beethoven String Quartets, 14.
40. Note that the cello appears here an octave lower than in most published scores. This reading of
the passage reflects that of the new Beethoven Werke.
example 4.15 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
variants of the principal motive: (a) mm. 30 35; (b) mm. 41 48; (c) mm. 72 77;
(d) mm. 101 8







182 analysis and theory

example 4.15 (continued)







Heard as a variant of the $ form of the motive, it offers the first real challenge
to the standard of typicality for the motivic category.
The first variant of $ returns in the cello in mm. 101 104 (shown in ex.
4.15d) against a figure in the first violin that can be related to the % form of
the motive; this passage is discussed in a bit more detail below.
cate g ori zat i on, st rate g y, and sy ntax 183

In contrast to the ubiquity of the $ form of the motive, the % form makes only
two appearances after m. 30. The first, already mentioned above, is in m. 42 (see ex.
4.15b). Here the motive appears with an altered rhythmic profile (lacking its first
note and using the variant ending first introduced in m. 18), and, in its role as coun-
terstatement to the modified $ form in the viola, with less independence than
before. The third entrance of the variant (in m. 46) departs further still, with its end-
ing a durational augmentation of that used in the preceding measures. A similar
augmentation of the tail of the % form is heard in mm. 89 91 (as shown in ex. 4.16)
after a general pause that inter rupts the mater ial from the second thematic g roup.
Save for this augmentation, this variant of the % form of the motive is quite close to
the very first version of %, heard in mm. 5 6.
The variant heard in mm. 89 91 is the last explicit reference to the % form of
the motive before the conclusion of the Exposition. Nonetheless, the appogiatura
figure that came to be a feature of % beginning in m. 14 figures quite prominently
in the counterstatement to the final appearance of the $ form in mm. 101 104
(shown in ex. 4.15d). When this counterstatement is embellished with a repeated
turn figure in m. 106 (answering the embellishment of the $ variant in m. 105), the
connection becomes somewhat more plausible: the counterstatement could be
heard as a further variant of the % form of the motive. This inter pretation is sup-
ported by two things: first, there is the close relationship between the $ and % forms
of the motive throughout the Exposition (exemplified by their juxtaposition in
mm. 41 48). Second, the tur n figure typical of all for ms of the motive is replicated,
in compressed for m, by the combination of the trill and ter mination on the third
beat of m. 102 and the appogiatura of m. 103 (and further varied in mm. 104 105).
The possibility of extending the category of % motives to include the counter-
statement of mm. 102 105, together with the connection between the category of
$ motives and various counterstatements (in mm. 41 48, 72 77, and 101 108),
suggests a way our concept of the % form of the motive might slip further still. If we
return to the transition that begins in m. 30 (ex. 4.15a), we note a similar ity between

example 4.16 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
mm. 88 91





184 analysis and theory

the accompanimental figure in the first violin and the % variant from mm. 17 18:
the rhythmic patter n and contour of mm. 31, 33, and 35 is the same as that of m. 18.
Our return to these measures becomes literal rather than imaginary if the Exposi-
tion is repeated: hearing mm. 30 35 after the several pair ings of the $ form of the
motive with counterstatements and the extensions of the % form that occur dur-
ing the first pass through the Exposition could provide enough pressure to cause
significant slippage within our concept of the % form. This, in turn, might lead us
to embrace the accompanimental figure of mm. 30 35 within that for m when we
encounter it for the second time.
Perhaps more remarkable than the conceptual fluidity of the % form of the
motive is the resistance of the $ motive to change: despite numerous juxtapositions
with rival mater ials, it remains constant. Indeed, these two categor ies of motive forms
can be further distinguished by the compositional strategies associated with each.
Over the course of the Exposition, $ forms are subjected to relatively little variation,
and most members of the category are close to the conceptual model outlined in
figure 4.7a. In contrast, % forms are varied almost from the outset. While the con-
ceptual model for the category of % forms still makes sense it captures the salient
features of members of the category as presented in the opening measures and serves
to distinguish them from members of the category of $ motives the range of val-
ues that members assign to the attr ibutes picked out by the conceptual model is
much broader, and a given variant is rarely repeated. The difference between the
compositional strategies applied to each category of motive forms lends a distinctive
character to them: $ as stable in the face of various challenges; % as an agent of
change and transition. Although qualities such as these are part of the elemental
materials of sonata for m, as they are manifested in the first movement of the F major
quartet they also pose a kind of problem. The sort of fluidity with which the % form
of the motive is associated operates pr incipally in contrast to the stability of the $
form. By itself, it is not quite sufficient to create the dynamic energy the sparks of
real transfor mation that the Development section will require. For this sort of
drama, a new form of the motive is required, but one that draws on the musical
materials and syntax developed over the course of the Exposition.

The Development
After a br ief A major flour ish, the Development revisits $ and %, recalling the open-
ing of the Exposition (as shown in ex. 4.17; the example does not include the A major
flourish).While each motive departs slightly from its respective conceptual model
the $ motive includes light accompaniment in the upper str ings, and the % motive
acquires an echo for the most part the conceptual model for each category is
reaffirmed. Beethoven then introduces a new form of the motive that reconfigures the
properties of $ and % and transfor ms the possibilities for musical discourse.
The new motive form (shown with a ' in ex. 4.17) is launched in m. 129
through a ser ies of imitative entrances. Although motive forms have appeared at
one-measure intervals before, here they differ only in the octave in which they are
stated. This manner of presentation makes reference to a topic to which Beethoven
often made recourse in the Development sections of his instrumental works: the
example 4.17 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
mm. 119 46




186 analysis and theory

example 4.17 (continued)



learned style that includes fugue and canon.41 This new motive form resembles a
member of the category of $ motives. It is new in its isolation, harmonic stasis, and
clear assertion of the downbeat of each measure as a point of initiation (rather than
a point of articulation within a larger gesture, as it is in the second measure of a typ-
ical member of the % category of motives). However, the dissonant interval for med
by the downward leap consequent to the tur n figure (a diminished seventh Bb3
to CS3 in mm. 129 30) demands continuation beyond the downbeat, a fluidity
reminiscent of the % form of the motive.42 This amalgam of features and the way
they dominate the measures that follow suggest that ' constitutes a new category of
41. For a discussion of the learned style, see Ratner, Classic Music, 23 24.
42. Given both the emphasis of the large downward leap and the fact that the leap is dissonant, the
' form of the motive is more ambiguous at the hypermetric level than either the $ or % forms: where
the previous for ms of the motive can be heard in ter ms of pairs of strong and weak measures, such a
grouping in the case of the ' motive is much more arbitrary. This is not to say that deter mining which
measure of the $ and % forms of the motive is strong or weak at any given point is a simple matter, how-
ever. Janet Levy and Roger Sessions, for instance, give interpretations of the accentuation of mm. 1 2
that are the exact opposite of one another. See Levy, Beethovens Compositional Choices, p. 10, n. 28; and
Roger Sessions, The Musical Experience of Composer, Performer, Listener (Princeton, N.J.: Princeton Uni-
versity Press, 1950), 13 14.
cate g ori zat i on, st rate g y, and sy ntax 187

figure 4.8 Conceptual model for the ' form of the principal motive from Beethoven,
String Quartet, Op. 18 No. 1, first movement, mm. 129 44

motive forms. The conceptual model for this categor y (diagrammed in fig. 4.8)
consists of five cor related conceptual elements, character ized as the things necessary
for a melodic fragment to count as an instance of this version of the motive.While
this model has much in common with those diag rammed in figure 4.7, it also
demonstrates enough differences to warrant its being used to distinguish the ' form
from any of the other for ms of the motive important in the movement.
Beethoven follows mm. 129 34 with two similar sections, the first beginning
in m. 135, the second in m. 141. Each section begins with a slightly different patter n
of entrances among the instruments, and each section prolongs a single fully di-
minished seventh chord: CSo (mm. 129 34), FSo (mm. 135 40), and Bo (mm.
141 46).43 Each section also includes a variant of the & fragment, based on the tur n
common to all for ms of the motive but with a sforzando on the second beat: it first
appears in the second violin in mm. 133 34, and subsequently in the viola and cello
(mm. 139 40) and first violin (mm. 145 46). In truth, the variant of & is difficult to
hear, for within these imitative sections it must compete with the qd figure that
comes to dominate the first half of each measure in which the fragment appears.
Nonetheless, as can be seen in example 4.18, the fragment is gradually transfor med,

43. This is in fact the same sequence of fully diminished seventh chords initiated in mm. 14 16;
however, that sequence led not to a Bo chord but to a Bb pedal preparatory to the first full cadence on
F in mm. 19 20.
In his analysis of mm. 129ff., Ratner inter prets each fully diminished seventh chord as a leading-tone
harmony with a nor mal resolution; see Ratner, The Beethoven String Quartets, 15. Indeed, on the very last
eighth-note of mm. 128 and 129 there is a D minor chord, as there is a G minor chord in mm. 139 and
140, a C minor chord in mm. 145 and 146, and F and Bb minor chords in mm. 147 50. However, I hear
these har monies as simply embellishing the fully diminished seventh chords that dominate these mea-
sures. This hear ing is not without its theoretical problems it suggests the kind of dissonant prolonga-
tion that has vexed a number of music theor ists over the past two decades but it has the advantage of
capturing the intensity and uncompromising nature of this passage.
188 analysis and theory

example 4.18 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
mm. 145 54



over the course of mm. 147 51, into the & fragment as it or iginally appeared. So
transfor med, it leads in a cascading sequence to the ar rival on the dominant of F
major in m. 167. In one sense, however, this is not the & fragment as it was first
heard in mm. 22 27, for that fragment was one-half of a two-measure motive. By
contrast, the fragment of m. 151 and following is heard as the extension, in one
voice, of the imitative entrances initiated in m. 129.
With its abrupt introduction of the lear ned style, combination of features from
both the $ and % forms of the motive, and prolongation of dissonant har monies, the
' form of the principal motive provides elements of contrast and continuity with
the preceding music. And compared with the various changes visited upon $ and %
in the course of the Exposition, the repetitions of mm. 129 46 both on the level
of the individual motives that make up the imitative passages and on the level of the
separate sections that prolong three different fully diminished seventh chords
make for an island of stability within the changing fortunes of the principal motive.
Nonetheless, this moment of stability provides its own disruptions, and not only
because it is based around a sequence of fully diminished seventh chords. The bal-
ance that had existed between the $ and % forms of the motive one associated
cate g ori zat i on, st rate g y, and sy ntax 189

with a syntax of stability, the other with a syntax of fluidity, and each necessary for
the other has now disappeared: what is left after ' has done its work is only the
chatter ing of the & form of the motive. This, and passagework scales over a domi-
nant pedal, is all that remains until the $ form of the motive reasserts itself at the
beginning of the Recapitulation.

Recapitulation and Coda

The appearance of motive forms in the Recapitulation, which begins with a literal
return to the opening of the movement in m. 179, follows the same general outline
as that of the Exposition. Missing, however, are almost all of the members of the %
category of motives. After the initial response to the $ motives (mm. 183 86, an
exact repetition of mm. 5 8), % disappears, along with other vestiges of the open-
ing twenty-odd measures, including the sequence of fully diminished seventh
chords from mm. 14 16 that had their extension in mm. 129 46. Beethoven
replaces mm. 9 29 with a br ief episode that veers off toward Gb major, which in
its latter half makes reference to the transition that began in m. 30, and which also
absorbs the function of the statement and counterstatement of motive forms in
mm. 41 48.44 The consequence of these changes is that the $ form of the motive
goes virtually unchallenged in the Recapitulation, for both the % form and the syn-
tax of fluidity with which it was associated are absent. What we are left with is a
clear impression of the stability even obstinateness of the $ form of the prin-
cipal motive.45
Beethoven does offer a final challenge to this stability in the Coda, which begins
in m. 274. After a pair of short ascending passages in staccato quarter notes
(mm. 274 81) that recall the quarter notes that followed ' in the Development and
a brief descending figure built from the paired sixteenth-note portion of the styl-
ized tur n figure (mm. 282 83), the main motive reenters in the first violin in m. 284
(see ex. 4.19). With the violas entrance at the octave below in m. 285, and contin-
uation with the same sort of staccato quarter-notes that began the Coda, the refer-
ence is clear: Beethoven means to summon the ' form of the motive, although
without its fortissimo entrances or dissonant har monies. On the one hand, the sub-
sequent entr ies (beginning in m. 288) hint at the possibility of further development
as the motive is tossed back and forth between different voices; on the other hand,
they all serve to prolong the dominant. The possibility of development becomes a
bit more real with the move to the submediant, in m. 294, and the reprise of a vari-
ant of the % form of the motive.While the rhythmic figure is different from a typ-
ical member of the % category, the syntactic references are clear: the motive is
sequenced upward (as it was in mm. 13 18, 21 26, and 123 27), and its har monic

44. Levy notes that these changes tend to draw together various features of the original transition;
see Levy, Beethovens Compositional Choices, 79 83.
45. The impression of the solidity of the $ form of the motive is given further support by
Beethovens decision to eliminate the repeat of the second section of the movement (that is, the Devel-
opment and Recapitulation), one of the significant changes he made between the first and second ver-
sions of the quartet.
example 4.19 Ludwig van Beethoven, String Quartet, Op. 18 No. 1, first movement,
mm. 284 302








cate g ori zat i on, st rate g y, and sy ntax 191

context is thoroughly fluid, including four fully diminished seventh chords (in
mm. 295, 296, 298, and 299) and a cadential dominant (in mm. 300 301). Against
the resolution of the over this dominant, the $ form of the motive finally enters
(in the first violin in m. 301), and for the first and only time in the movement, its
concluding downward leap outlines a falling 5-1, signaling an end to the peregri-
nations suggested in the preceding measures.
After this closing gesture, there is one more bit of dialogue between $ and % (not
shown in ex. 4.19). The cello, beginning and ending on F2, repeats the variant of $
that ends with a repeated note (in mm. 302 303 and 304 305, a variant first heard
in mm. 41 42). It is answered by a variant of % reminiscent of that just heard in
mm. 294 299 (played in thirds in the first and second violins in mm. 306 307 and
308 309). True to for m, these % variants pull away from F major by introducing its
b6 (Db), but they are answered by a cadential gesture that returns the movement, for
once and for all, to F major.
With regard to the pr incipal motive of Beethovens F major quartet, the Coda
has a summational function that is twofold. First, the Coda revisits each of the three
categories of motive forms that appeared over the course of the movement, doing
so in reverse order of their initial appearance. Second, the Coda revisits the syntax
with which each of these categor ies was associated. The Coda begins (much as did
that of the first movement of Mozarts Dissonance Quartet) with a reference to
the Development; the subsequent allusions to ' suggest a challenge to the integrity
of the principal motive (through fragmentation) but also take place within a con-
text of overall stability (with the bulk of the entrances linked to the dominant of F).
The % motives that follow these ' forms exemplify the syntax with which % is most
often associated: although the broad range of features exhibited by members of the
category of % motives is not in evidence, the har monic fluidity typical of the cate-
gory is well represented by the fully diminished seventh chords prominent in the
passage. And the final $ motive, with its summoning of cadence, is the very embod-
iment of general, processive stability, yielded by the compositional strateg ies most
often used for the category of which it is a member.

Changes between the First and Second Versions

of the F Major Quartet
Janet Levy observed that one of the principal differences between the first and sec-
ond versions of the F major quartet is the greater concision of the second version.46
Beethoven accomplished this in a number of ways, which together resulted in the
elimination of a significant number of statements of the $ form of the principal
motive. Basic elements of the plan Beethoven followed in his revisions are evident
in the changes made to the episode involving the $ motive initiated in the midst
of the second key area. The passage as or iginally wr itten is given in example 4.20.
In the fourteen measures spanned by the episode, there are seven clear statements of
the $ motive. In the second version of the movement, this episode was cut back to

46. See Levy, Beethovens Compositional Choices, 24.

example 4.20 Ludwig van Beethoven, String Quartet, Op. 18 No. 1 (first version), first
movement, mm. 72 85












cate g ori zat i on, st rate g y, and sy ntax 193

six measures (which are given in ex. 4.15c), with only three statements of the $
form of the motive. But perhaps more important than simply reducing the appear-
ances of the $ motive, Beethoven appears to have rethought how it fits into the
overall musical fabric, as a closer inspection of his first attempt at this episode
The or iginal version of the passage divides into two parts, the first in C minor
(mm. 72 79), the second in C major (mm. 80 86). In both parts, there are three
main statements of the $ motive: in the C minor section these are in the second
violin, and each ends with a large ascending leap; in the C major section the state-
ments are given in the cello and end with a repeated note. (The latter for m of the $
motive is heard in a pr ior episode, given in ex. 4.15b, which is substantially the same
in both versions of the movement.) Each of the second violins statements is
answered by a counterstatement in the first violin. The counterstatements in
mm. 73 and 75 beg in with a tur n figure and share their rhythmic profile with the
$ form of the motive. On the other hand, the counterstatement of m. 77 answers
with the same variant of the $ motive that will be heard momentar ily in the cello.
The cellos three statements of this $ variant are not given similar counterstatements
but are answered instead by passagework in the first violin, which ultimately leads
out of the episode.
When revising this passage, Beethoven cut the C minor section, salvaging the
first violins counterstatements of mm. 73 and 75. These he placed an octave lower,
but still in the first violin (see ex. 4.15c). He omitted the counterstatement of m. 77,
changing the first violins music so that it starts with what appears to be another
counterstatement, but which then dissolves into passagework that leads out of the
episode. The effect of these alterations is not just concision but a view of the $
motive that is changed in three ways. First, by dropping the three statements of the
motive made by the second violin in the C minor section, Beethoven eliminated a
further variant namely, one that ends with a large leap from the category of $
forms. Second, the counterstatements made by the first violin, now closer in regis-
ter to the cellos statements of the motive, make a direct if limited challenge to the
structure of the $ motivic category. Third, changing the last counterstatement so
that it leads out of the episode (rather than echoing the cellos third statement of the
$ motive, as the or iginal version of m. 77 would) lets the challenge stand, albeit only
for a moment or two. Thus, in the revised version of the passage Beethoven has not
only achieved great concision, but he has also accomplished two strategic goals: he
has tightened the boundar ies of the $ motivic category, and he has intimated how
these boundar ies might be challenged.
A similar elimination of outright statements of the $ motive and modification of
the structure of motivic categor ies occurs with the final statement of the motive at
the end of the Exposition. In the first version of the passage, shown in example
4.21, the musette variant of the $ motive (in mm. 109 and 111) is answered by a
counterstatement with a tr ill and appogiatura (in mm. 110 and 112). When the
phrase is repeated (in mm. 113 16), the counterstatement is embellished by
repeated tur n figures, but the $ motive is not. In Beethovens revisions, both the
counterstatement and the musette variant are embellished when the phrase is
repeated (mm. 105 108 of ex. 4.15d). Although this removes two explicit state-
194 analysis and theory

example 4.21 Ludwig van Beethoven, String Quartet, Op. 18 No. 1 (first version), first
movement, mm. 109 16


ments of the $ motive (and, consequently, could be seen to weaken the structure
of that category), the change makes it possible to hear the first violins counter-
statements as further variants of the % motivic categor y (a possibility raised in the
earlier discussion of ex. 4.15d).
A further result of Beethovens decision to reduce the number of appearances
of $ in the revised version of the movement was a clearer distinction and more
even balance between the $ and % forms of the motive. In the Development of the
first version of the movement, for instance, the two forms of the principal motive
blend into one another. As shown in example 4.22, the entrances of the $ form
subsequent to the A major flour ish that opens the Development are first answered
by an $ form of the motive and then by a variant of % (mm. 127 32). Appearing
above the extended tail of this answering % is another % (a quite typical member
of the categor y), which leads, after a pause, to yet one more entrance of % (itself
answered by a & fragment, as it was in m. 21 and following). Thus an episode
beginning with the $ form of the motive ends with the % form of the motive,
blurring the distinction between the two. In Beethovens revised version (ex. 4.17),
example 4.22 Ludwig van Beethoven, String Quartet, Op. 18 No. 1 (first version), first
movement, mm. 127 43




196 analysis and theory

the two for ms are successive but do not overlap: $ and % stand out more clearly
as separate entities.
Beethovens revisions to the Development also resulted in a shar per distinction
between the ' form of the motive and the $ and % forms, especially where the syn-
tax of the motivic categor ies is concer ned. In the or iginal version, the answering
statements by the second violin (first $ in m. 128 of ex. 4.22, then % in m. 130) both
enter at the octave above the viola. They thus anticipate the imitative entrances of
' in mm. 137 and following.When these answers are removed in the second version
(cf. ex. 4.17, m. 119 and following), the entrances of ' seem fresher the introduc-
tion of the lear ned style seems more of an innovation. The entr y of ' in the first
version of the movement, moreover, leads to a consonant har mony (the dominant
of G minor), and the downward leap at this point is a minor sixth (ex. 4.22,
mm. 137 38). In the revised version, ' leads to a dissonant har mony (ex. 4.17,
mm. 129 30), and the downward leap is a diminished seventh. In sum, in the or ig-
inal version of the Development ' seems to grow out of a generalized idea of the
principal motive ($ and % blurring together); after Beethovens revisions, the intro-
duction of ' is more striking, and it contrasts shar ply with the statements of $ and
% that precede it.
Revisions to the Coda offer a final bit of evidence that the changes Beethoven
made were with the intent of more clearly differentiating between different versions
of his principal motive. The original version of the Coda, starting at the point that
references to the pr incipal motive become explicit, is given in example 4.23. In the
first half of this excer pt, Beethoven relies heavily on the $ form of the motive, pro-
viding statements of it that beg in variously on G and C in the first violin, and Bb
and A in the second violin. Because the latter have an ascending skip after the tur n
figure, they could also be heard as a variant of the % form of the motive, but the
quality of the $ form of the motive still pervades. In m. 312, with the shift away
from the dominant to the submediant, the % form of the motive enters in ear nest,
with a clear reference to the successive entr ies between % and & over the fully
diminished seventh chords found in the Exposition. This ser ies of % motives leads
ultimately to a conventional, if deliberately under played, cadence.
In the second version of the Coda (ex. 4.19), the $ motive is withheld, and the
first references are to the ' form. As noted in the discussion of example 4.19, the
syntax of the passage owes much to the fluidity associated with the process of devel-
oping musical mater ials and extends to the % motive that enters in m. 294. When
the $ form of the motive finally returns in m. 301, it is coincident with a perfect
authentic cadence and signals a definitive end to the possibility of further develop-
ment raised by the entrances of the ' motive.
The changes Beethoven made to the first movement of the F major quartet,
then, had two important results. First, they created a more balanced relationship
between the $ and % forms of the motive in the Exposition. Second, the distinc-
tions between the three main categor ies of motive forms are more sharply drawn.
These changes make the syntax associated with each categor y somewhat more
apparent that is, we can more easily hear the role of each motive form in the dis-
course of the movement as a whole.
example 4.23 Ludwig van Beethoven, String Quartet, Op. 18 No. 1 (first version), first
movement, mm. 302 19



? ?





198 analysis and theory

Compositional Strategy and Multiple Levels

of Musical Syntax
As A. B. Marx noted, the principal motive of the first movement of Beethovens F
major str ing quartet is a nearly constant presence, never far from the listeners ear.
The function of this dense network of motives, however, is not so much to lend the
movement coherence after all, Beethoven actually reduced the number of times
the $ form of the motive appeared when he revised the movement but to set up
a musical discourse separate from but coordinated with the har monic and thematic
syntax of sonata for m. This discourse involves not simply a syntax specific to the
principal for m of the motive (as in the case of the first movement of the Disso-
nance Quartet), but higher levels of syntax of the sort seen in the sixth quartet
from the Op. 18 set.
Beethovens compositional strategy depends on clear distinctions between dif-
ferent for ms of the motive. These distinctions are made not only on the basis of
audible features of the motive (of the sort captured in the conceptual models of fig.
4.7) but also on the compositional strategies that lend a distinctive character to each
category of motive forms: $ as stable in the face of various challenges and % as an
agent of change and transition. By the end of the Exposition, a delicate balance has
been established between the two, a balance not in evidence in the first version of
the quartet, where the category of $ motive-forms dominated.
This balance is maintained through the first part of the Development, but it is
disrupted by the introduction of the ' form of the motive and the topic of the
learned style (ex. 4.17, m. 129). It is only with the Recapitulation that the focus
comes back to the $ form of the motive, where it stays (with little or no competi-
tion from %) until the Coda. The disruption created by ' and the recuperation facil-
itated by $ suggest a function for the topic of the lear ned style within Beethovens
compositional practice: the fugato on the ' form of the motive creates the impres-
sion of development by temporar ily suspending the nor mative syntax for the pro-
jection of tonal areas associated with $ and %, a syntax restored only with the
reprise of $ at the beginning of the Recapitulation. This function of $ is in evi-
dence again at the end of the movement when, after hints at further development
and har monic fluidity in the Coda, $ returns as the melodic voice of the final
cadence in F major.
The distinctions among these three forms of the principal motive distinctions
that make possible a musical discourse that is focused on the changing fortunes of
the motive over the course of the movement, and that were made more apparent
in Beethovens revisions to the movement suggest that motivic analysis could be
refined by attending to how motives are used, rather than to how successfully they
can be traced back to some Ur-type. Such a refinement makes possible an appreci-
ation of multiple levels of syntax levels that interact with the overall har monic
and thematic syntax of the musical work, and which collectively engage more and
less immediate levels of cognitive processing. Ultimately, an appreciation of this
refinement leads to a better understanding of how compositional strateg ies shape
our conceptions of musical organization.
cate g ori zat i on, st rate g y, and sy ntax 199

For Umberto Eco, one of the mysteries of semiotics was that music could have a
fairly well developed syntax and yet no apparent semantic depth. Music meant
something, but outside of conventionalized signs like those associated with military
or pastoral music, it was quite difficult to say what this was. In this chapter, I have
not tr ied to address this problem head on; indeed, describing the or igins and scope
of musical meaning is an endeavor far more ambitious than what I undertake here.
I have, however, tried to argue that musical syntax and semantics are better viewed
as points along a continuum than as discrete domains. More directly, composers (as
well as others who br ing music to life) make use of strategies that disrupt and
redefine musical syntax as part of a process of meaning construction.
My path into these compositional strategies has been through categor ization, on
the assumption that categor ies of musical events represent a common ground where
the concer ns of composers and listeners meet. The viability of this assumption was
demonstrated by the structure of the waltzes produced by the musical dice game.
Where formal structure needed to be clearly defined (for instance, at the end of
each half of the waltz), the range of category members within each measure-
category was quite nar row, and what was typical of the relevant measure-category
was quite apparent.Where formal structure could be more open (for example, at the
beginning of the second half of the waltz), the range of category members was quite
wide and typicality harder to deter mine. Typicality could thus be used to define for-
mal syntax, a syntax that not incidentally coordinated with and thus confir med
harmonic syntax.
A somewhat different example of a compositional strategy organized around
typicality was seen in the category for the pr incipal motive of the first movement of
Mozarts Dissonance Quartet. At important moments in the movement at the
beginning and close of the Exposition and Recapitulation, and in the Coda either
thoroughly or highly typical for ms of the motive made their appearance. These
were followed by less typical members of the motivic categor y, leading either to a
reprise of a typical for m or to new musical mater ial. This basic process a syntax of
typicality yielding to atypicality was repeated three times in the course of the
Development, ending only with the appearance of a thoroughly typical for m that
signaled the start of the Recapitulation. While the compositional strateg ies that
shaped the different versions of the principal motive ran parallel to the for mal syn-
tax of the movement, they were nonetheless independent. The motive thus emerged
as a topic for musical discourse, whose progress could be followed along with the
other strands of musical discourse compr ised by the movement.
By contrast, the principal motive of the first movement of the Bb major quartet
from Beethovens Op. 18 never functioned as a topic for musical discourse. Al-
though the motive forms fell into two distinct categor ies, individual members never
strayed far from their respective conceptual models. In consequence, the immedi-
ate level of syntax for the categor ies of motive forms was not particularly interest-
ing, since members were almost always quite typical of their categor y. However,
over the course of the Exposition each category acquired distinct functions: one ($)
200 analysis and theory

signaled the opening of musical discourse; the other (%) signaled the close of dis-
course. This emergent syntax was then exploited in the Development and more
generally served to show how musical topics of the sort descr ibed by Ratner could
come into being.
It was in Beethovens F major quartet from Op. 18 that we saw both of these lev-
els of syntax come together. As with the Bb major quartet, Beethoven introduced
two categor ies of motive forms ($ and %) in the Exposition, with a third (&) play-
ing a decidedly subsidiary role. A compositional strategy focused on typicality was
applied to both categor ies, but where the range of variation for $ was relatively nar-
row, for % it was quite wide. Contextual pressures on % even suggested that the con-
cept of what counted as a member of the category could slip to include a number
of rather remote variants, a kind of slippage that never seemed to threaten $. By the
end of the Exposition, and even into the beg inning of the Development, $ and %
appeared to maintain a delicate balance, the for mer stable in the face of various
challenges, the latter an agent of change and transition. Beethoven then introduced
a new form of the motive (') to disrupt this balance, creating a species of develop-
ment of high drama: $ and % were swept from the stage, vanquished by a volatile
combination of properties drawn from each. Balance within the movement as a
whole (but not between $ and %) was restored with the return of $ at the start of
the Recapitulation. Even the developmental urges of the Coda could not threaten
this equilibr ium. The compositional strategies applied to the pr incipal motive, then,
operate on both a local level (within a g iven category) and global level (among
groups of categor ies). Again, these syntactic levels are independent of but coordi-
nated with those for the overall har monic and thematic structure of the move-
ment the syntactic strands associated with the pr incipal motive are multiple, but
they do not exhaust the syntax of the whole.
In sum, these analyses suggest that the construction of meaning in music can be
achieved through the way composers choose to deploy the elements of musical syn-
tax. Of course, compositional strategy is not the only source of meaning construc-
tion in music I have also pointed out the contr ibutions of musical topics and
cross-domain mapping to musical semantics but it is one to which our knowl-
edge of categorization can be profitably applied. As the meeting place for the con-
cerns of composers and listeners, categor ies of musical events are important to both
compositional strategy and musical syntax, for they represent a means through
which uniquely musical meaning can be created.
chapter five
cultural knowledge and
musical ontology

A s part of a search for an inter pretive theory of culture and consequent to a cr i-

tique of early work in cognitive anthropology, Clifford Geertz momentar ily
and somewhat famously tur ned his thoughts to music as an example of culture.
Taking a Beethoven str ing quartet as his case in point, he suggested that no one
would identify the quartet with its score, with the skills and knowledge needed to
play it, with the understanding of it possessed by its perfor mers or audience, with a
particular perfor mance of it, or with some mysterious entity transcending mater ial
existence. There was one perspective, however, that Geertz thought would meet
with general agreement: That a Beethoven quartet is a temporally developed tonal
structure, a coherent sequence of modeled sound in a word, music and not
anybodys knowledge of or belief about anything, including how to play it, is a
proposition to which most people are, upon reflection, likely to assent.1
By this example, Geertz intended to put to rest, for once and for all, the notion
that culture exists as some sort of program inside peoples heads. Nonetheless, if
music is not knowledge or belief, then what is it? Geertzs definition of music as a
coherent sequence of modeled sound is simply too broad, for it would have to
include str ings of Morse code, speech in an unknown tongue, and the cr ies of var-
ious animals. Indeed, from the perspective of this definition, there is, in principle,
nothing to distinguish music from noise or noise from music.2 Music is a category
constructed by humans, not a substance or set of relations floating free in the world.

1. Clifford Geertz, Thick Descr iption: Toward an Inter pretive Theory of Culture, in The Interpre-
tation of Cultures: Selected Essays (New York: Basic Books, 1973), 11 12.
2. Much can and has been said regarding the relationship between music and noise. Three sources
from the last ninety years represent the range of issues engaged by the topic: Luigi Russolo, The Art of
Noises (1916), trans. Barclay Brown (New York: Pendragon Press, 1986); Jacques Attali, Noise: The Politi-
cal Economy of Music, trans. Brian Massumi, with a foreword by Fredric Jameson and an afterword by
Susan McClar y, Theory and Histor y of Literature, 16 (Minneapolis: University of Minnesota Press,
1985); and the improvised performances by Derek Bailey, Pat Metheny, Gregg Bendian, and Paul Wertico
on The Sign of 4 (Knitting Factory Works, [CD] KFW 197, 1997).
I indicate the media of the recordings I cite with the following abbreviations: CD = compact disk;
LP = 33 RPM record; 78 = 78 RPM record.

202 analysis and theory

As such, the notion of music is dependent upon and constrained by human knowl-
edge, which presumably resides with individual humans. Having said this, it must
also be said that the continued existence of a Beethoven str ing quartet does not
depend on the personal knowledge of any one individual. Music is indeed a cultural
fact, and culture is largely public and inter personal.
Cultural knowledge, as represented by Geertzs example, is thus a bit of a para-
dox. On the one hand, it is not something that is deter mined by any one individual:
culture is, as Geertz maintained, public. On the other hand, individuals need such
knowledge if they are going to participate in culture, and it is through the partici-
pation of individual humans that culture and cultural knowledge is maintained.
The problem, as framed by Geertzs example of a Beethoven quartet, is more
specifically one of musical ontology that is, the ontological status of a work of
music. In the West, it is common to regard pieces of music as having an objective
status, speaking of a work of music as one might of a painting or sculpture. This
habit of thought persists in spite of the transience of musical phenomena: unlike a
painting or sculpture, a musical work has no endur ing mater ial existence. The
enigma of works that are of profound cultural importance and yet are totally
ephemeral has led to a number of attempts by philosophers and others to account
for the objective status of the work of music and, by extension, music itself; thus the
problem of musical ontology.3 A contrasting perspective has developed among ethno-
musicologists who, working chiefly with non-Western music, have noted the fluid-
ity of musical practice and a concomitant absence of any ascription of permanence
to the products of musical activity. These observations render doubtful any univer-
sal solution to the problem of the ontology of the musical work, or even music,
since many cultures do not show linguistic or practical evidence of equivalent con-
cepts.4 And yet, musical practice within these cultures, while fluid, is certainly not

3. Thoughtful discussions of the problem of musical ontology can be found in Lydia Goehr, The
Imaginary Museum of Musical Works: An Essay in the Philosophy of Music (Oxford: Clarendon Press, 1992);
Philip V. Bohlman, Ontologies of Music, in Rethinking Music, ed. Nicholas Cook and Mark Everist
(Oxford: Oxford University Press, 1999), 17 34; and Nicholas Cook, At the Borders of Musical Iden-
tity: Schenker, Corelli and the Graces, Music Analysis 18 (1999): 179 233. See also the essays collected
in Michael Talbot, ed., The Musical Work: Reality or Invention? (Liverpool: University of Liverpool, 2000).
4. Examples of cultures the languages of which lack a word equivalent to music or concepts
equivalent to that of a musical work include the Hausa of Nigeria, the Macuma Shuar of Ecuador, and
the Mapuche of Argentina; Charles Keil reports such a lack in over twenty languages of the African con-
tinent and gives detailed discussion of the methodological problems involved in studying such cultures.
See David W. Ames and Anthony V. King, Glossary of Hausa Music and Its Social Contexts (Evanston, Ill.:
Northwestern University Press, 1971), ix;William Belzner, Music, Modernization, and Wester nization
among the Macuma Shuar, in Cultural Transformations and Ethnicity in Modern Ecuador, ed. Norman E.
Whitten, Jr. (Urbana: University of Illinois Press, 1981), 735 36; Carol E. Robertson-DeCarbo, Tayil
as Category and Communication among the Argentine Mapuche: A Methodological Suggestion, Year-
book of the International Folk Music Council 8 (1976): 39; and Charles Keil, Tiv Song (Chicago: University
of Chicago Press, 1979), 27 52.
I should make clear that what I find dubious is a universal solution to the problem of the ontology
of a musical work. I do not think it is impossible, however, to speak about music in the case of a cul-
ture that does not show linguistic or practical evidence of such a concept, as long as one does not attempt
to colonize that culture by insisting that its members, when they try to make sense of their own expe-
riences and cultural practices, make use of the term.
cultural knowle dge and musical ontology 203

unregulated. In general, the habit of objectifying aspects of musical practice is

omnipresent; in consequence, the problem of musical ontology, while most properly
one that ar ises only among those who believe in an objective status for the work
of music, also has applications to situations where no such object is posited.
In this chapter, I propose that solutions to the problems of cultural knowledge in
general, and of musical ontology in particular, can be found in what we now know
about processes of categorization. In brief, knowing that Beethovens quartet is
music or that anything else is music, for that matter means knowing how to
categorize sequences of sound-events in accordance with conceptual models shared
with other members of a culture. More specifically, each work of music or
whatever other unit of cultural cur rency we choose to focus on constitutes a
Type 1 category of the sort discussed in chapter 2.5 On the one hand, the trans-
parency and immediacy of the conceptual models that guide the process of catego-
rization explain why we describe works of music in the same language we use for
objects: we speak of the Beethoven quartet just as we speak of the red cup, even
though the for mer (as sounding music) is a phenomenon completely unlike the lat-
ter.6 On the other hand, because membership in Type 1 categor ies is by degree, a
certain amount of fluidity is possible, varying in degree with the cultural context
relative to which the conceptual model for the categor y is framed. We might
include a wide variety of performances in the categor y for the Beethoven quar-
tet some more, some less competent, even some in an ar rangement for four-hand
piano but we could well draw the line at a version for pitched percussion and
automobile hor ns.
In what follows, I explore the idea that musical works can be thought of as Type
1 categor ies and that such categor ies are a manifestation of cultural knowledge. I
approach this issue by taking a close look at two popular songs from the early twen-
tieth century: I Got Rhythm and Bye Bye Blackbird. Songs like this challenge
our ideas about what constitutes a work of music: while they are almost always
associated with a score (as is a Beethoven quartet), they are also part of a perfor-
mance tradition that involves some level of improvisation (introducing the fluidity
typical of many non-Western perfor mance traditions).
I Got Rhythm provides a good example of this situation, for there were mul-
tiple competing versions of the song almost from the moment it appeared on
Broadway late in 1930. In the first section below, I consider a number of recordings

5. Of course, a work of music could be regarded as a Type 2 category were necessary and sufficient
conditions for categor y membership specified. This is how I would inter pret Nelson Goodmans pro-
posal that the score is the absolute deter minant of musical identity. Goodman wr ites, If we allow the
least deviation [from the score], all assurance of work-preservation and score-preservation is lost; for by
a ser ies of one-note er rors of omission, addition, and modification, we can go all the way from
Beethovens Fifth Symphony to Three Blind Mice. Thus while a score may leave unspecified many fea-
tures of a performance, and allow for considerable variation in others within prescr ibed limits, full com-
pliance with the specifications g iven is categor ically required. Nelson Goodman, Languages of Art: An
Approach to a Theory of Symbols, 2nd ed. (Indianapolis: Hackett Publishing, 1976), 186 87.
6. There are, of course, aesthetic perspectives that support this habit (such as that associated with the
work-concept, as descr ibed by Lydia Goehr in The Imaginary Museum of Musical Works, 89 119) and
technological innovations that give it credence (like music notation and sound recording), but these are
not preconditions for the notion of a musical object.
204 analysis and theory

of the song made dur ing the 1930s and 1940s, explore how the model for the song
changed and how it stayed the same, and show that both the fluidity and the stabil-
ity of this model are straightforward consequences of our use of categor ization to
organize our understanding of music.
Bye Bye Blackbird presents a somewhat different problem, which takes us
deeper into the relationship between categor ization and cultural knowledge.Writ-
ten as a popular song dur ing the 1920s and quite successful when it first appeared,
it never really entered the standard repertoire until a recording by Miles Davis in
1956 transfor med the song into a jazz tune.7 Because there was no tradition of grad-
ual change, it was possible for radically different models for the song to coexist and
come into conflict. In the second section below, I consider this sort of conflict in
light of what Mikhail Bakhtin called double-voiced discourse and what Henr y
Louis Gates Jr. calls Signifyin(g). I attempt to show that a fuller account of how we
categorize musical works requires that we go outside of the relatively parochial con-
cerns of musical perfor mance and consider the context in which musical perfor-
mance takes place.
In the third section, I offer some concluding thoughts on cultural knowledge,
musical ontology, and the cognitive processes through which we structure our
understanding of the world. I end the chapter with some reflections about how the
perspective on musical ontology I offer here can change the way we view music

categorization, musical ontology, and

cultural knowledge
A Multiplicity of Rhythms, 1930 1946
I Got Rhythm was wr itten by George and Ira Gershwin for their musical Girl
Crazy, which opened on Broadway on 14 October 1930. The song, intended to be
a show stopper, appeared near the end of the first act, and its refrain was repeated a
number of times: to accompany a tap dance chorus after the final vocal refrain; as an
encore for Ethel Mer man, whose success with the song became legendar y; as part
of the mostly instrumental entracte before the second act; and, sung by the whole
company, as the shows concluding number.
The song itself, the score for which is given in example 5.1, comprises an intro-
duction, verse, and refrain. The melody of the verse begins with an eight-measure
phrase in G minor (featur ing a prominent b5 at its midpoint), which is given a var-

7. I borrow the distinction between a song and a tune from Richard Crawford, who wr ites, in the
jazz tradition, we usually speak of tunes, not songs. A jazz tune is defined first and foremost by its struc-
ture: by the patter n of repetition and contrast in its melodic phrases and the har monic framework under-
lying them. Second, it is defined by its ethos: by the mood it projects and the tempo at which it is played.
Only third does its melody come into play, for in the jazz tradition the melody is often little more than
an entre into the perfor mance; after being heard, it is usually discarded for free melodic invention by the
performers. Richard Crawford, George Gershwins I Got Rhythm (1930), in The American Musical
Landscape (Berkeley: University of Califor nia Press, 1993), 221, 225.
cultural knowle dge and musical ontology 205

example 5.1 Version of George and Ira Gershwins I Got Rhythm used in act 1 of
Girl Crazy. I Got Rhythm, words and music by George Gershwin and Ira Gershwin,
1930 WB Music Cor p. (Renewed). All rights reserved. Used by permission. Warner Bros.
Publications U.S. Inc., Miami, FL 33014.


ied repetition. This is followed by a four-measure phrase, which is also repeated

(again, with variations), after which a two-measure transition leads to the refrain.
The refrain is in the AABA form typical of popular songs of this period, but it
includes an ascent to F5 near the end of the last A section. This climactic r ise
requires a two-measure tag to br ing the melody back to Bb4 and yields a 34-
measure AABA' form.
I Got Rhythm was only rarely recorded as it appeared in the or iginal show.
Although it contr ibuted significantly to her professional career, Ethel Mer man did
not record it until 1947. Many amateurs and semi-professional musicians must have
performed the song from the sheet music, which did reproduce the music as it
206 analysis and theory

example 5.1 (continued)

appeared in Girl Crazy. This was not the case, however, with professional musicians.
Even in the earliest recordings (made within a few weeks of the shows opening),
the verse loses its introductory (and nar rative) function and is used instead as an
interlude between repetitions of the refrain. For example, the recording made on 23
October 1930 by Red Nichols and His Five Pennies (a group drawn from the pit
orchestra for Girl Crazy) begins with an instrumental version of the or iginal refrain,
example 5.1 (continued)

208 analysis and theory

example 5.1 (continued)

followed by the refrain with vocal.8 Only after this does the music for the verse
enter, but without vocals. The recording concludes with two more statements of
the refrain, also without vocals. A very similar ar rangement was used by Fred Rich
and a studio band in their recording of 29 October 1930. The overall for m of their
rendition consists of three refrains (the second with vocal), followed by an instru-
mental verse, followed by one complete and one partial statement of the refrain (the
latter using only the B and A' sections to conclude the perfor mance).9
In the two years following the opening of the show, almost all of the recordings
that appeared made use of vocals, although none of which I am aware used the
words for the verse.10 With the instrumental version recorded by Don Redman and
8. Red Nichols and His Five Pennies, I Got Rhythm, in Rarest Brunswick Masters, 1926 1931: Red
Nichols and His Five Pennies (MCA Records, [LP] MCA 1518, 1982).
9. Fred Rich and His Orchestra, I Got Rhythm, in 1930 Youre Driving Me Crazy: Portrait of a
Year in Music (Phontastic, [CD] PHONT CD 7618, 1992).
10. A fairly comprehensive list of recordings of I Got Rhythm from 1930 to 1942, based on Br ian
Rusts The American Dance Band Discography 19171942 (New Rochelle, N.Y.: Arlington House, 1975), is
included in Crawfords George Gershwins I Got Rhythm, 222 25.
cultural knowle dge and musical ontology 209

His Orchestra on 30 June 1932, however, we encounter a new approach to the

song.11 In this ar rangement, the verse was eliminated completely, as was the role of
the vocalist. The result is something very much like the version of I Got Rhythm
that can be found in jazz fake books, such as shown in example 5.2.12 After 1932,
the number of recordings that include vocals drops off precipitously, and I Got
Rhythm became a tune used by dance bands for instrumental improvisation a
jazz standard as may be observed in recordings by Red Norvo and His Swing
Sextette in 1936 or Chick Webb and His Little Chicks in 1937.13 Seen from the
perspective of these recordings, George Gershwins own perfor mance of I Got
Rhythm as a set of variations for piano and orchestra is somewhat anomalous, for
there is little if any sense of improvisation, nor is this music intended for dancing.14
From 1932 to 1942, recorded perfor mances are remarkably consistent in their
approach to I Got Rhythm: all use the same thirty-four-measure form, all make

example 5.2 Jazz fake book version of George and Ira Gershwins I Got Rhythm.
I Got Rhythm, words and music by George Gershwin and Ira Gershwin, 1930 WB
Music Cor p. (Renewed). All rights reserved. Used by permission. Warner Bros. Publications
U.S. Inc., Miami, FL 33014.

11. Don Redman and His Orchestra, I Got Rhythm, in Don Redman and His Orchestra, 19311933
(Classics, [CD] 543, 1990).
12. Jazz fake books typically contain hundreds of songs. Generally, as is true of the piece in ex. 5.2,
each is given as a melody, along with chord symbols that outline a basic har monization and some min-
imal instructions for perfor mance; the words for the songs may or may not be included. These mater i-
als allow players familiar with jazz perfor mance practice to give an acceptable rendering of a song even
if they have never played it before that is, they can fake their way through a perfor mance.
13. Red Norvo and His Swing Sextette, I Got Rhythm, in Red Norvo (Time-Life Records, [LP]
STL-J14, 1980), originally recorded 16 March 1936; Chick Webb and His Little Chicks, I Got Rhythm,
in Spinnin the Webb: Chick Webb and His Orchestra (GRP Records, [CD] 513678L, 1994), originally
recorded 21 September 1937.
14. George Gershwin, I Got Rhythm, in Gershwin Performs Gershwin: Rare Recordings 19311935
(Musical Her itage Society, [CD] 512923A, 1991), originally recorded 19 February 1934.
210 analysis and theory

some reference to the or iginal melody, and all preserve the affect and tempo of the
original. However, an ar rangement recorded by the Jimmy Dorsey Orchestra on
28 November 1942 signaled yet another approach to I Got Rhythm, for after
employing the customar y thirty-four-measure format for the head, the band
switched to a thirty-two-measure format for the solos.15 In truth, this was not the
first time the musical mater ials of I Got Rhythm had been associated with a
thirty-two-measure form. Early on, the song had been used as the basis for a num-
ber of other compositions.16 These der ived tunes typically retain the har monic
framework, overall for m, and affect of the or iginal song but substitute a new
melody and g ive the whole a new title. In the case of I Got Rhythm, derived
tunes allowed jazz players to keep what was useful from the song its fast tempo,
lively affect, and straightforward harmonic structure yet escape the limitations of
its simple, predominantly pentatonic melody. The earliest recorded tune der ived
from I Got Rhythm, Sidney Bechet and the New Orleans FeetwarmersShag
(recorded 15 September 1932), did retain the thirty-four-measure form of I Got
Rhythm, with its two-measure tag at the end of the final A section.17 However,
Fletcher Hendersons Yeah Man, which appeared the following year, did away
with the two-measure tag, yielding a thirty-two-measure form in which all of the
A sections were of equal length.18 There were two advantages to this change. First,
the excision of the two-measure tag made the connection between Yeah Man
and I Got Rhythm more obscure: Yeah Man could sound a bit more like a
brand new tune. Second, jazz practice of this per iod was beginning to embrace
longer solos. The 34-measure for m of I Got Rhythm, it will be recalled, ends
with a two-measure tag added to its final A section. A musician who wanted to
take two choruses in succession would consequently find the regular flow of
eight-measure phrases in the backg round har monic structure inter rupted in mid-
stream with the appearance of the ten-measure A' section. When a musician took
two choruses on the 32-measure form, however, the succession of eight-measure
phrases was constant: the improvisation could stretch seamlessly across the two
refrains, allowing the player to build longer melodic lines and to shape the dynamic
structure of the solo over larger spans of music. Although after Yeah Man the
tunes der ived from I Got Rhythm were invariably based on the thir ty-two-

15. Jimmy Dorsey and His Orchestra, I Got Rhythm, in Jimmy Dorsey and His Orchestra,Vol. 2:
194244 (Hindsight Records, [LP] HSR-153, 1980). For perfor mances with an emphasis on improvisa-
tion that are guided only by a relatively informal ar rangement, the practice among jazz perfor mers is to
state the tune at the beg inning and end of the perfor mance. In these cases, the tune functions as the
head. The head is typically followed by solos over the har monic framework of the tune, which are, in
turn, followed by a concluding statement of the tune.
16. Richard Crawford calls such der ived tunes contrafacts, borrowing the ter m from one of sev-
eral medieval practices to which it is linked. The specific practice invoked is one in which new texts are
grafted on to preexistent music, which was itself freely adapted. See Crawford, George Gershwins I
Got Rhythm, 225, 228 29.
17. Sidney Bechet and the New Orleans Feetwarmers,Shag, in Sidney Bechet (Time-Life Records,
[LP] STL-J09, 1980).
18. Fletcher Henderson and His Orchestra, Yeah Man, in Fletcher Henderson, Developing an Amer-
ican Orchestra, 19231937 (Smithsonian Collection, [LP] R 006, 1977), originally recorded 18 August
cultural knowle dge and musical ontology 211

measure form, it was only around 1942 that this form was brought to I Got Rhythm
At first, the thirty-two-measure form was used only for solos the head
retained its thirty-four-measure form, complete with the two-measure tag (as in the
Jimmy Dorsey arrangement). By the mid-1940s, however, even the tag for the head
had fallen away.When Charlie Parker recorded I Got Rhythm with a group of
all-star perfor mers on 22 April 1946, the thirty-two-measure form was used for the
head, as well as for the twenty-three choruses that followed.19 The transfor mation
of I Got Rhythm into a platfor m for improvisation is even more evident in a
recording made two years earlier by Lester Young and the Kansas City Six. In this
performance, not only is the thirty-two-measure harmonic patter n used through-
out, but also the melody of the song is never stated.20 In this for m that is, as a
thirty-two-measure, AABA harmonic structureI Got Rhythm came to be one
of the basic elements of the jazz players musical vocabulary: the Rhythm changes
shown in example 5.3.21

example 5.3 Scheme for Rhythm changes (based on the har monic structure of the
thirty-two-measure for m of I Got Rhythm). I Got Rhythm, words and music by
George Gershwin and Ira Gershwin, 1930 WB Music Cor p. (Renewed). All rights
reserved. Used by permission. Warner Bros. Publications U.S. Inc., Miami, FL 33014.
-<%[U U* -<%[U U* -U -<%[U* -
N* H
<U* U*
I O Q* H
<U* U*
I O Q* Q*<U* H R R\U* H <%[U IU* O<U* Q*U

-<%[U U* -<%[U U* -U -<%[U* - -<%[U
N* H
<U* U*
I O Q* H
<U* U*
I O Q* Q*<U* H R R\U* H*<%[U* H*

U* U* U* U* U U U* U*
H* S S I I O O Q Q

N* -<%[U U* -<%[U U* -U -<%[U* - -
<U* U*
I O Q* H
<U* U*
I O Q* Q*<U* H R R\U* H <%[U* QU* H*<%[U

19. Charlie Parker and others, I Got Rhythm, in Bird: The Complete Charlie Parker on Verve
(Verve/Polygram Classics, [CD] 837 142 2, 1988).
20. Lester Young with the Kansas City Six, I Got Rhythm, in The Tenor Sax: Lester Young, Chu
Berry and Ben Webster: The Commodore Years (Atlantic, [LP] SD 2-307, 1973), originally recorded 28
March 1944. That the perfor mance by Youngs group resulted from an ar rangement employed for suc-
cessive performances rather than happenstance is suggested by an alter nate take of the same tune
recorded the same day (included on this album), which uses precisely the same ar rangement.
21. The ter m changes refers to a conceptual structure that consists of a succession of harmonies
correlated with a metr ic and hypermetric framework. Knowing this structure knowing where and
when the har monies change is essential for creating convincing improvisations.
On the importance of Rhythm changes as a basic har monic prototype, see Paul F. Berliner, Think-
ing in Jazz: The Infinite Art of Improvisation, Chicago Studies in Ethnomusicology (Chicago: University of
Chicago Press, 1994), 76 78.
212 analysis and theory

From the evidence provided by recordings, then, what counted as I Got Rhythm
changed from 1930 to 1946 as the song was transfor med from a Broadway or pop-
ular song, to a jazz standard, to a platfor m for improvisation. This prcis is not quite
accurate, however, for even as new conceptions of the song emerged, older ones
remained in circulation. Benny Goodman, for instance, continued to use the thirty-
four-measure form of the refrain even after others had adopted the thirty-two-
measure form;22 productions of the musical (including the 1943 movie Girl Crazy,
starring Mickey Rooney and Judy Garland) continued to use the entire song after
most other perfor mances had omitted the verse.23 From almost the moment of its
introduction, there were multiple Rhythms in circulation, and what belonged
under the rubric I Got Rhythm depended on where you stood in the stream of
American popular culture.

Categorizing I Got Rhythm

Collected in table 5.1 are all of the versions of I Got Rhythm discussed thus far.
How might someone go about categorizing these various versions were they encoun-
tered in the course of daily life rather than as a pre-selected list? That is, assuming one
were not so fortunate as to have each perfor mance identified as a perfor mance [or
recording, or rendering] of I Got Rhythm, what would be the basis for including
any or all of these in the category performances of I Got Rhythm? As shown in chap-
ter 1, categories like this are organized around a conceptual model that sets up an ide-
alized representation of what counts as a typical member of the category.24
As a place to begin, let us consider a conceptual model focused on the attr ibutes
of the song as it first appeared in Girl Crazy (leaving aside the variants created as the
song was reprised over the course of the musical). The model (which I shall call the
pop-music model, for reasons that will become clear below) is diagrammed in
figure 5.1 and consists of five cor related conceptual elements, character ized as the
things necessar y for an adequate perfor mance of the song. Two aspects of this
model should be clar ified. First, although the verses of popular songs were often
omitted in perfor mance, I take the fact that composers still bothered to wr ite verses
and that the verse was included, in some for m or another, in early recordings as evi-
dence of its importance to the identity of the song. Second, my distinction between
variation and improvisation is based on the extent rather than the presence of musi-
cal mater ials not specified by the score: if these mater ials are relatively limited, I
would tend to regard them as a product of variation; if they are extensive, I would
tend to regard them as a product of improvisation. Where to draw the line involves
another set of conceptual models for perfor mance practice; nonetheless, the dis-
tinction is relatively clear in early recordings.

22. Benny Goodman Sextet, I Got Rhythm, in Benny Goodman (CBS Records, [LP] P5 15536 CP,
1981), originally recorded 18 September 1945.
23. Although the plot line for the movie is completely different from that of the musical, the musi-
cal numbers are the same.
24. By categor ies like this I mean not categor ies involving musical works specifically, but the cat-
egories we use in daily life generally.
cultural knowle dge and musical ontology 213

table 5.1. Select Performances of I Got Rhythm

Performers Date of Recording

Red Nichols and His Five Pennies 23 October 1930

Fred Rich and His Orchestra 29 October 1930
Don Redman and His Orchestra 30 June 1932
George Gershwin 19 February 1934
Red Norvo and His Swing Sextette 16 March 1936
Chick Webb and His Little Chicks 21 September 1937
Jimmy Dorsey and His Orchestra 28 November 1942
Lester Young with the Kansas City Six 28 March 1944
The Benny Goodman Sextet 8 September 1945
Charlie Parker and others 22 April 1946

Applying this model to the perfor mances listed in table 5.1 yields the typicality
effects character istic of Type 1 categor ies. Most typical would be the perfor mances
by Red Nicholss and Fred Richs groups; performances from the or iginal produc-
tion of the musical (which are not included in the list of recordings) would of
course also be reckoned as highly typical.25 By contrast, the perfor mance led by
Gershwin, which drops the verse and lyr ics, would not confor m as well with the
model. Even less typical would be the perfor mances led by Redman, Norvo,Webb,
and Goodman, for all these add a substantial emphasis on improvisation. Finally,
on the fr inges of the category would be the perfor mances led by Dorsey, Young,
and Parker, for these not only omit the verse and lyr ics but also use a thirty-two-
measure form of the refrain as a platfor m for extended improvisations.
As we see in this example, conceiving of various renditions of a song as a Type
1 category organized around a conceptual model offers a way to explain the pecu-
liar ontological status of musical works. Our sense that there is only one thing that
should be called I Got Rhythm is supported by the idealization basic to the
model and to the models role as the conceptual anchor for the categor y. At the
same time, our ability to accept as members of the category things that do not fully
correspond to this model shows how the model functions as a guide to under-
standing the structure of the category as a whole. Although Charlie Parkers I Got
Rhythm is not like Ethel Mer mans I Got Rhythm, it shares enough attr ibutes
with the latter that we will not want to exclude it from the category outr ight. Fur-
ther, the model can provide a basis for descr ibing why we would grant Parkers per-
formance a liminal rather than a central status within the category we might say
something along the lines of He uses too much of that crazy improvisation, so he
really isnt playing I Got Rhythm.

25. There was no recording of the original production of Girl Crazy. However, there was a painstak-
ing recreation of the 1930 production undertaken by Elektra Nonesuch in 1990, and Lorna Lufts per-
formance there (in the part of Kate Fothergill or iginated by Merman) agrees in most respects with the
version of I Got Rhythm recorded by Merman in 1947. See George Gershwin and Ira Gershwin, Girl
Crazy (Elektra Nonesuch, [CD] 79250-2, 1990), recorded 26 28 February 1990.
214 analysis and theory

figure 5.1 Pop-music model for the category performances of I Got Rhythm

Useful as this approach to musical ontology is, it demands at least a bit of fur-
ther elaboration: we need to find a way to descr ibe how the conceptual model for
a musical work changes over time. After all, while Parkers 1946 perfor mance might
have been regarded as unusual by some, it is clear from the musical practice of his
contemporar ies that it was far from exceptional. Here Richard Crawfords account
of three different approaches taken to I Got Rhythm during the years 1930 1942
provides a helpful guide.26 The first approach, in which I Got Rhythm was
treated as a popular song, is similar to that assumed by the pop-music model dia-
grammed in figure 5.1. The second approach descr ibed by Crawford is one in
which I Got Rhythm is treated as a jazz standard. A conceptual model that fits
this approach is diagrammed in figure 5.2 and consists of five things necessary for an
adequate perfor mance.27 Adopting this model would, of course, change what
counted as a typical version of the tune. Now the perfor mances led by Redman,
Norvo, Webb, and Goodman would be central to the categor y, and the perfor-
mances led by Dorsey, Young, and Parker only a bit less typical. Far more atypical
would be the perfor mance led by Gershwin, with its relative lack of emphasis on
improvisation, and the perfor mances by Red Nicholss and Fred Richs groups, with
their inclusion of the verse.
The third approach to I Got Rhythm noted by Crawford was as a musical
structure that served as the basis for various der ived tunes. I would like to general-
ize this somewhat to accommodate the slightly expanded time frame represented by
the recordings collected in table 5.1. After 1942, I Got Rhythm served not only
as the basis for der ived tunes but also as a framework for improvised perfor mance

26. Crawford, George Gershwins I Got Rhythm, 218.

27. Explicit mention of the melody in this conceptual model is intended as a hedge against confus-
ing I Got Rhythm with any of the tunes der ived from it.
cultural knowle dge and musical ontology 215

figure 5.2 Jazz-standard model for the category performances of I Got Rhythm

a framework embodied in the Rhythm changes of example 5.3. This leads us to a

third model for I Got Rhythm, which is diagrammed in figure 5.3. This improv
model, conceived as a skeletal vehicle for improvisation and composition, consists of
four things necessary for an adequate perfor mance. Once again, what would count
as a typical version of the tune changes. From the perspective provided by the
improv model, the perfor mances led by Young and Parker would be most typical,
with that led by Dorsey following close behind. The perfor mances led by Redman,
Norvo,Webb, and Goodman, with their allegiance to the or iginal for m and melody
of Gershwins song, would now be regarded as less typical, and the perfor mances
led by Gershwin, Nichols, and Rich would have to be considered as thoroughly
Let us now return to the matter of how the conceptual model for a musical
work changes over time. If we take these three models as an histor ical sequence
the pop-music model followed by the jazz-standard model, which is followed by
the improv model we can see that each successive conceptual structure retains
some of the elements of the previous structure and either modifies or drops others.
The jazz-standard model retains from the pop-music model the elements
concer ning affect and the mater ials of the refrain. It changes those relating
to function and lyr ics, and it adds an element specifying the presence of the
melody. (The latter is implicit in the materials of the refrain inherited
from the pop-music model, but it is made explicit here).
The improv model retains from its jazz-standard counter part the element
concer ning affect. It changes the components of the model associated with
the refrain, melody, and function, and it drops any requirement for the lyr ics.
As is evident from table 5.1, these are not the only models possible, nor should my
descriptions of the model be taken as comprehensive or final. One could easily
216 analysis and theory

figure 5.3 Improv model for the category performances of I Got Rhythm

imagine idiosyncratic or transitional models that blend features of those outlined

here. Thus a model based on the perfor mance led by George Gershwin, for
instance, would bor row some features from the pop-music model and some from
the jazz-standard model; a model based on the ar rangement played by the Jimmy
Dorsey band would incor porate some aspects from the jazz-standard model and
some from the improv model. And were such models taken as the basis for our
determinations as to what belonged in the category performances of I Got Rhythm,
there would of course be changes in what counted as a typical or atypical member
of the category.
When people share the conceptual model for a song (realizing that all such shar-
ing is approximate), they will tend to make similar judgments about what counts as
a typical or an atypical rendering of the song (or whether a succession of sounds
should even be counted as an instance of the song). As I proposed in chapter 3, shar-
ing knowledge in this way is one of the bases of culture. This becomes especially
clear when we consider other roles played by the model for a musical work: as a set
of cognitive resources for perfor mance, as a basis for negotiations on how musical
practice should proceed, and as a focus for how musical practice can be regulated.

Conceptual Models and Performance

Up to this point, we have considered music that is part of a tradition of performance
at least partially regulated by the score. But what of performance traditions in which
scores play no part that is, music that is part of an aural tradition? Where scores (or
similar artifacts) are absent, performers must rely on a store of knowledge about
music and music making, knowledge often gathered through a long and painstak-
ing process that starts in earliest childhood and continues for many years.
Constantin Brailoiu was perhaps the first to suggest that musical perfor mance
within aural traditions relies on stored cognitive constructs. As a way of explaining
the source of the divergent perfor mances folk musicians gave of the same song,
Brailoiu proposed that such perfor mances were ephemeral incar nations of an ideal
cultural knowle dge and musical ontology 217

archetype stored in the mind of the musician.28 In his work on African music,
Simha Arom adopted Brailoius idea, but he called the archetype a model.29 James
Cowdery, whose research has focused on Ir ish music, proposed a similar cognitive
basis for perfor mance practice, but one der ived from the tradition of inquiry cen-
tered on the tune family.30 Cowdery called the construct a tune model and con-
ceived of it as a generating patter n in the mind of the perfor mer or musical com-
munity.31 Finally, James Porter has developed detailed descr iptions of a succession of
conceptual perfor mance models for the chang ing versions of the song My Son
David sung by the Scottish singer Jeannie Robertson between 1953 and 1960.32
For Porter, the conceptual perfor mance model represented a holistic, self-contained
concept of a particular song in the mind of the singer.33 This model enabled the
singer to g ive successive but nonidentical perfor mances of the same song on a
number of occasions, but the model was nonetheless subject to change over time
that is, the same song gradually became a different song as the conceptual perfor-
mance model changed.
The idea behind all of these approaches is that perfor mers who work within
aural traditions base their perfor mance of a given tune on a cognitive construct that
is stored in memory and that represents essential features of that tune. In their basic
features, such constructs are essentially the same as the conceptual models I
described in chapter 3. I have used these same models in this chapter to explain how
individuals deter mine whether a particular perfor mance is of a specific tune and
how typical that perfor mance is. The difference that distinguishes my models from
those of Brailoiu, Arom, Cowdery, and Porter is that the sort of constructs they pro-

28. The relevant texts in translation and the or iginal are as follows: Lacking an unchallengeable
[musical] text, we must admit that we never collect more than variants and that latent in the singers
minds lives an ideal archetype of which they offer us ephemeral incar nations. [A dfaut de tout texte
irrcusable, force nous est dadmettre que nous ne recueillons jamais que des variantes et que, dans
lesprit des chanteurs, vit, dune vie latente, un archtype idal, dont ils nous offrent des incar nations
phmres.] Constantin Brailoiu, Le Folklore musicale, in Musica aeterna: La Vie et la production musi-
cales de tous les temps et de tous les peuples, en tenant compte particulirement de la Suisse, de la Belgique, de la
France et de la musique de nos jours, ed. Gottfr ied Schmid (Zur ich: M. S. Metz, 1949), 2: 319.
29. Simha Arom, African Polyphony and Polyrhythm: Musical Structure and Methodology, trans. Martin
Thom, Barbara Tuckett, and Raymond Bond (Cambr idge: Cambridge University Press, 1991), 148.
30. The idea of a tune family was first proposed by Samuel P. Bayard in his classic essay Prolegom-
ena to a Study of the Pr incipal Melodic Families of British-Amer ican Folk Song, Journal of American
Folklore 63 (1950): 1 44.
31. James R. Cowdery, The Melodic Tradition of Ireland (Kent, Oh.: Kent State University Press, 1990), 33.
32. James Porter, Jeannie Robertsons My Son David: A Conceptual Performance Model, Journal
of American Folklore 89 (1976): 7 26; and idem, Context, Epistemics, and Value: A Conceptual Perfor-
mance Model Reconsidered, in Selected Reports in Ethnomusicology: Issues in the Conceptualization of Music,
7, ed. James Porter and Ali Jihad Racy (Los Angeles: University of California, Department of Ethno-
musicology, 1988), 69 97. Porters thoughtful appraisal of what might be involved in a conceptual per-
formance model goes well beyond the pr imitive models I sketch here and should be taken into account
in any further application of theories of categor ization to ethnographic research.
It should be noted that while Arom, Cowdery, and Porter all use the ter m model for their hypo-
thetical cognitive constructs, they do so without reference to any of the cognitive science research on
models or related constructs.
33. Porter, Context, Epistemics, and Value, 71.
218 analysis and theory

pose are used not for categor izing musical perfor mances, as I attempt to do, but for
producing them. In truth, we have already seen conceptual models used in this way,
for Jeff s conceptual models for Twinkle not only helped him organize the
Montessor i bells in particular patter ns but also guided his actions when he per-
formed the song. In the present context, however, the network of knowledge with
which we are concer ned is far more extensive than that with which Jeff was
engaged. The skills required for an adequate perfor mance of a tune like I Got
Rhythm (relative to the perfor mance tradition descr ibed above) are extensive: a
trumpet player, for instance, must know where to find the notes for the tune on her
instrument, how to produce these notes in the r ight order, how to shape this pro-
duction to create a convincing version of the tune, how to coordinate her playing
with that of other members of an ensemble, and so on.34 For perfor mers, then, there
will be a number of conceptual models related to the various challenges of creat-
ing musical sound, as well as the conceptual model for a particular piece.
The complexity of this network of knowledge emphasizes once again the
importance of the global models relative to which local models are framed. With
regard to the conceptual model for a musical work, the global model speeds the
process of learning the conceptual model for a new tune and also provides a frame-
work for discourse among musicians. For instance, among perfor mers the jazz-
standard model for I Got Rhythm (which was diagrammed in fig. 5.2) is framed
relative to a global model for jazz practice, which includes the five elements dia-
grammed in figure 5.4. A comparison of figure 5.4 with figure 5.2 shows that each
generic element in the for mer has its more specific cor respondent in the latter.
Confronted for the first time with a tune like I Got Rhythm, a musician who has
internalized the global model will already have a start on assembling a local con-
ceptual model to use as a guide in perfor ming the tune. The global model also
serves as a frame for discourse (both verbal and musical) about local models. That is,
approaching I Got Rhythm as a song that belongs to the perfor mance practice
of jazz means that features such as its har monic structure, melody, and character are
what are relevant, rather than whether there is a definitive score on which to base
Conceptual models, then, are used not only to structure our understanding of

34. Of necessity, this network of knowledge will include embodied knowledge: what it feels like to
make musical sounds and to coordinate this activity with other physical activities. Richard Powers
touches on the importance of such knowledge in his novel Galatea 2.2, which centers on a hypotheti-
cal exper iment to replicate aspects of human intelligence on a vast and spreading neural network, imple-
mented via parallel processing on a massive supercomputer. At one point, the implementation, named
Helen, asks How do you sing? The narrator responds by singing a fragment of a song he lear ned as a
child: Bounce me high, bounce me low, bounce me up to Jericho. To his amazement, the implemen-
tation not only absorbs this song but can later be heard singing it, over and over, like a child trying out
her voice.Yet sing is not quite the r ight word, for Powerss nar rator realizes that there is something
missing: Helen did not sing the way real little girls sang. Technically, she almost passed. Her synthesized
voice skittered off speechs earth into tentative, tonal Kitty Hawk. Her tune sounded remarkably lim-
ber, given the scope of that mechanical tour de force. But she did not sing for the r ight reasons. Little
girls sang to keep time for kickball or jump ropes. . . . Helen didnt have a clue what keeping time meant,
never having twirled a jump rope, let alone seen one. Richard Powers, Galatea 2.2 (New York: Farrar
Straus Giroux, 1995), 205.
cultural knowle dge and musical ontology 219



J#'"*[%;;*#6:"K* 46#*#'"7"*%$"*:(#*"77":#5%9*



figure 5.4 Partial representation of the global model of jazz practice

what we hear but also to create the things we hear. In both cases, the model is part
of a network of knowledge that, through being shared by the members of a musi-
cal community and realized in sound, constitutes musical culture.

Negotiating Conceptual Models

Musical practice is not a steady state, as may be seen in the different versions of I
Got Rhythm collected in table 5.1. Over time, conceptual models will be
modified in response to changes in cultural pr iorities. Rarely are such modifications
the result of creative fiat by one individual (although one such case is discussed
below, in connection with Bye Bye Blackbird). More commonly, they are the
product of a gradual ser ies of negotiations within a musical community.
The basic outlines of the process of negotiating the structure of conceptual
models can be illustrated by consider ing two highly simplified situations. Each
involves a musician who initiates the process of negotiation by suggesting changes
to the constituent elements of a given conceptual model. In the first case, the nego-
tiations occur within a g roup of musicians occupied with perfor ming the same
work; in the second case, the negotiations occur between the musician and an audi-
ence of non-perfor mers.
The first situation can be conceived of as taking place in a rehearsal or a perfor-
mance of the version of I Got Rhythm associated with the jazz-standard model.
Instead of playing the usual har monic progression for the refrain, the pianist, with-
out comment, introduces a completely new ser ies of harmonies for the A section
and truncates the overall for m to thirty-two measures. The other musicians may
accept this change (by adjusting their perfor mance to the new harmonies), reject it
220 analysis and theory

(by resolutely hewing to the structure of the or iginal model), or suggest further
modifications (by offering their own inter pretation of the har monic progression).
This process could also play out with the aid of language (the changed har monic
progression could be announced in advance) or symbols (a chord chart for the
changed refrain could be supplied to the other musicians), or through various com-
binations of musical perfor mance, language, and symbols by any or all participants.
The second situation can be conceived of as taking place dur ing a perfor mance
of I Got Rhythm by a solo pianist, playing for an audience of nonperfor mers.
Again, instead of the usual har monic progression for the refrain, the pianist intro-
duces a new ser ies of harmonies and truncates the overall for m to thirty-two mea-
sures. The audience may respond by accepting the change (through applause or a
similar activity), rejecting it (by shouting catcalls or throwing bottles), or suggest-
ing further modifications (by shouting More! or Too much!). (Of course, it is
quite probable that an audience would not even notice changes of this sort, but for
the sake of the argument, I shall assume that it did.)
In each of these imaginary situations, some elements of the conceptual model for
the tune are changed and some are retained. The conceptual models thus for m a
framework for negotiation that is, negotiation is conducted in ter ms of the con-
ceptual model for the tune. Series of such negotiations then lead to the more sub-
stantive changes in conceptual models of the sort suggested by my account of how
such models might be linked in a histor ical succession. Of course, the imaginary sit-
uations I have described are highly simplified: as the works of Ruth Finnegan, Paul
Berliner, and Ingrid Monson have demonstrated, actual negotiations among musi-
cians are much more complex than those outlined here.35 Musical discourse is r ich
and creative, and it includes comedy and irony, as well as meaningful exchanges
about the business at hand. Nonetheless, what makes this communication possible
is the knowledge shared among musicians, knowledge that is organized by concep-
tual models.

Limits on Negotiation
The mutability of conceptual models under the social and cultural pressures of
negotiation has the potential to destabilize cultural knowledge by transfor ming it
until it becomes alien and unfamiliar. Under certain circumstances, the intercon-
nectedness and dispersion character istic of cultural knowledge can counteract desta-
bilization, a process noted by Mantle Hood.36 Under other circumstances, however,
the challenge to stability may be significant. In the case of musical practice, such
challenges might come from a number of sources: from the innovations of influen-

35. Ruth Finnegan, The Relation between Composition and Performance: Three Alternative
Modes, in The Oral and the Literate in Music, ed.Yoshihiko Tokumaru and Osamu Yamaguti (Tokyo:
Academia Music, 1986), 73 87; Berliner, Thinking in Jazz; Ingrid Monson, Saying Something: Jazz Impro-
visation and Interaction, Chicago Studies in Ethnomusicology (Chicago: University of Chicago Press,
36. Mantle Hood, The Reliability of Oral Tradition, Journal of the American Musicological Society 12
(1959): 201 09.
cultural knowle dge and musical ontology 221

tial individuals, from contact with other cultures, from a sudden increase in the
complexity of musical practice, or from changes in technology. Faced with this sort
of challenge, a musical community may develop strategies to eliminate or control
the instability of the conceptual models upon which the practice of that commu-
nity is based.
One common strategy is to represent key structural elements of a local concep-
tual model through artifacts such as scores or recordings. Although the score illus-
trations used for I Got Rhythm carry the aura of author ity, they actually follow
from the conceptual models for the song: a score is an artifactual manifestation of
the elements of the conceptual model deemed most relevant to the musical practice
of which the model is a part, created as a means of stabilizing the model. The score
for the or iginal version of I Got Rhythm (ex. 5.1) can thus be seen as a reflection
of the conceptual model for pop songs (fig. 5.1). Such songs invariably had a verse,
even if it was only rarely perfor med. Since the songs were often perfor med by musi-
cians with only limited score-reading abilities, chord symbols (which often do not
match the chords indicated by the music notation) are almost always included, and
they serve to indicate the minimal mater ials necessary to accompany the voice. The
jazz fake book rendering of I Got Rhythm (ex. 5.2) does away with the verse and
any specifics about how the accompaniment is to be realized; it reflects the essen-
tial features of the jazz-standard model (fig. 5.2). And the Rhythm changes of
example 5.3 preserve only the bare essentials, cryptic to anyone not schooled in jazz
but more than enough to serve musicians as a starting point for improvisation, pro-
vided they are familiar with the concepts and relations embodied in the improv
model (fig. 5.3). Thus, both in its production and inter pretation, musical notation
reflects (rather than generates) the local and global conceptual models that consti-
tute musical practice.37
The idea that musical notation could serve as a way to control (rather than sim-
ply enable) musical practice is not a new one, having been proposed by Charles
Seeger over fifty years ago.38 With respect to Western practice, the limits on nego-
tiation that follow from this reliance on musical notation operate on two levels.
First, any given instance of musical notation, as an embodiment of the work of
music, limits negotiation to those aspects of music deemed nonessential to the
work (such as the specifics of dynamics, timbre, and tempo): these are the things left
relatively ambiguous by the score.39 Second, the terms for negotiation become arti-
factual: any aspect of musical practice that would vie for deter mining the identity of
a musical work must be such that it can be expressed artifactually. Arguments over

37. Leo Treitler has made similar points about the role of notation in medieval music. See Treitler,
Transmission and the Study of Music History, in Report of the Twelfth Congress, Berkeley 1977, ed. Daniel
Heartz and Bonnie Wade (Kassel: Brenreiter-Verlag; Philadelphia: The American Musicological Soci-
ety, 1981), 202 11; and idem, History and the Ontology of the Musical Work, Journal of Aesthetics and
Art Criticism 51 (1993): 483 97.
38. Charles Seeger, Oral Tradition in Music, in Funk and Wagnalls Standard Dictionary of Folklore,
Mythology, and Legend, ed. Maria Leach and Jerome Fried (New York: Funk and Wagnalls, 1950), 2: 825 29.
39. Nelson Goodman observes that where the score attempts to capture what I have called the
nonessential aspects of music it ceases to be, in his ter ms, notational. See Goodman, Languages of Art,
183 86.
222 analysis and theory

musical identity (such as those that play out in discussions about the Urtext for a
score) are arguments about the status and inter pretation of artifacts, and any com-
peting account of musical identity must be expressed in artifactual ter ms.
It bears mention that the author ity of notation is not monolithic: most musicians
accept notation as something to be inter preted and consider the way it is to be used
to create a musical perfor mance a matter for negotiation. And even where notation
has been introduced with the explicit intent of eliminating variation within musi-
cal practice, it may be circumvented. As Ruth Davis has shown, the modified West-
ern notation applied to Tunisian art music beginning in the 1930s for the pur pose
of standardizing its melodic tradition has been tur ned into a creative tool for rein-
terpreting and redefining that tradition.40 Notation has thus slowed, but not elimi-
nated, the fluidity of Tunisian musical practice.

Categorization, Musical Ontology, and

Cultural Knowledge
By the late 1940s, as we have seen, there were multiple versions of George and Ira
Gershwins I Got Rhythm in circulation. There was the show stopper and pop-
ular tune from the Broadway musical, the hot jazz tune played by any number of
dance bands, and the relatively abstract platfor m for improvisation used as the basis
for any number of derived tunes by jazz composers and for extended improvisations
by smaller ensembles. Confronted with this multiplicity of I Got Rhythms, how
might a listener discr iminate between them? Based on what we now know about
processes of categor ization, we can say that our listener would almost certainly rec-
ognize that, while all these versions belonged in the category performances of I Got
Rhythm, not all would represent the category equally well. Depending on the con-
ceptual model around which the category was organized, some versions would be
regarded as quite typical of the category, others as less typical, and still others in dan-
ger of being excluded from the category altogether. Following Richard Crawfords
description of three different approaches to I Got Rhythm, I proposed three
models the pop-music model, the jazz-standard model, and the improv model
which would yield different inter pretations of typicality when applied to a range
of performances of the tune. The knowledge represented by such models, when
combined with knowledge about musical technique and framed by global concep-
tual models applicable to an entire repertoire, can also serve as the basis for musical
performance within aural traditions. Because they are retained by individual mem-
bers of a musical community, who may wittingly or unwittingly change the model
as they use it, conceptual models are subject to change over time. When such
changes are brought to the attention of other members of the musical community,
a ser ies of negotiations about what constitutes the model may ensue. At times,
changes to conceptual models may be perceived as a threat to the integ rity of a
musical community, and the community may then seek either to halt or to retard
change by placing limits on what can be negotiated.

40. Ruth Davis, The Effects of Notation on Performance Practice in Tunisian Art Music, The
World of Music 34 (1992): 111.
cultural knowle dge and musical ontology 223

Given the various conceptual models we have considered and the ways these
models are inter preted and cor related with other types of knowledge, it is clear that
the network of knowledge within which any musical work (or similar construct) is
embedded is a complex one. And yet we have thus far restricted ourselves to an
almost purely musical world. Matters only stand to become more complicated
when we contemplate how this world might connect with the cultural context
within which musical practice is itself embedded. In the next section, we encounter
some of these complications and gain a sense of the richer understanding of musi-
cal practice to which they lead, as we explore some of the different ways Bye Bye
Blackbird was conceived during the middle of the twentieth century.

bye bye blackbird, signifyin(g), and

cultural knowledge
On 5 June 1956, Miles Davis, John Coltrane, Red Garland, Paul Chambers, and
Philly Joe Jones went into Columbias 30th Street Studio in New York and recorded
three tunes: Tadds Delight, Dear Old Stockholm, and Bye Bye Blackbird.41
Although the recording session took place at a time when Davis was actively shap-
ing the sound and repertoire of his quintet, the choice of Bye Bye Blackbird was
nonetheless somewhat odd. Unlike the other two tunes, Bye Bye Blackbird had
never been a part of the jazz repertoire, but was simply an upbeat popular song usu-
ally played as a foxtrot. Recorded by Davis as a ballad, it acquired a melancholy and
an irresistible groove that were thoroughly exceptional.
The choice was made odder still by the racial overtones associated with the tune.
First published in 1926, Bye Bye Blackbird played a role in the racially charged
election campaign for the mayoralty of Chicago in 1927. Big Bill Thompson, a
Republican who had successfully cultivated the African American vote to become
mayor of Chicago in 1915, returned in 1927 to challenge Democratic mayor William
E. Dever. As part of their strategy to fight off this challenge, the Democrats appealed
to racial prejudice, voiced specifically through Bye Bye Blackbird. The ploy ulti-
mately backfired, for it galvanized African American support for Thompson:
When some of the Dever supporters sent calliopes through the streets piping the
strains of Bye Bye Blackbird, and spreading a circular which displayed a trainload of
Negroes headed from Georgia with Thompson as pilot of the train, and the caption:
This train will start for Chicago, April 6, if Thompson is elected, the obvious answer
of the colored leaders was: Elect Big Bill or its going to be bye-bye blackbirds in

41. These three tunes, together with Round Midnight, Ah-leu-cha, and All of You, were
released on Miles Davis, Round about Midnight (Columbia, [CD] CK 40610, 1987), originally recorded
27 October 1955 and 5 June and 10 October 1956. There are extensive transcr iptions and a discussion of
this recording of Bye Bye Blackbird in Berliner, Thinking in Jazz; see esp. pp. 678 88.
42. Harold F. Gosnell, Negro Politicians: The Rise of Negro Politics in Chicago, with an introduction by
James Q.Wilson (Chicago: University of Chicago Press, 1966; originally publ. in 1935, with an intro-
duction by Robert Park, Social Science Studies, 35), 54 55. There is a discussion of a similar cartoon
from around the same time that invokes the phrase Bye Bye Blackbird in Lloyd Wendt and Her man
Kogan, Big Bill of Chicago (Indianapolis: Bobbs-Mer rill, 1953), 256.
224 analysis and theory

In a similar vein, Jack Chambers reports that dur ing the 1930s, the few African
Americans who participated in activities dominated by whites, like the semi-
professional National Football League, had the song sung to them by the all-white
What seems to be reflected in Daviss adoption of the tune is the sort of double-
voiced discourse first descr ibed by Mikhail Bakhtin: on one level of discourse,
Daviss Bye Bye Blackbird is an aging pop song, with racial overtones that tur n
against its perfor mers; on another level, it is a deeply felt rewriting felt both in
terms of melancholy and groove that makes a mockery of the simplistic messages
of the song, not least by rising above them. This process of repetition (in the sense
of returning to an earlier text) and revision is associated with a kind of inter-
textuality Henr y Louis Gates Jr. called Signifyin(g), which he argued is funda-
mental and specific to African American rhetor ical practice.44 While Gatess pri-
mary focus was on literature, he extended his argument to other domains, music
among them, and his arguments have proved compelling to musicologists. Adopting
one of Gatess examples, Ingrid Monson was able to show in some detail how John
Coltranes 1960 recording of Richard Rodgers and Oscar Hammersteins My
Favorite Things (written in 1959) accomplishes this sort of Signifyin(g). Coltranes
interpretation completely revises the song: the focus is changed from the verse to
the interludes, Rodgers and Hammersteins waltz is transfor med into a polyrhyth-
mic extravaganza, and what was an upbeat tune becomes a brooding one.45
There is, however, a problem with claiming, in any simple way, that Davis is Sig-
nifyin(g) on Bye Bye Blackbird, and that problem is profiled vividly by a record-
ing of the tune in 1955 by Peggy Lee. In her perfor mance, Lee adopts an excep-
tionally slow tempo slower than Daviss and completely vanquishes the upbeat
sentiments of the or iginal song.46 Lee, too, seems to have engaged in double-voiced
discourse, but would we want to say that she is Signifyin(g)? Indeed, this sort of per-
formance practice is so common in jazz within and without the African Ameri-
can community that Signifyin(g), present everywhere, would seem to signify
Two sets of issues, then, are raised by Daviss recording of Bye Bye Blackbird.
First, what is the nature of the double-voiced discourse of Signifyin(g)? Is it in any
way distinct from double-voiced discourse as a whole? Second, what part does the
reinterpretation of musical works or, from the perspective developed in the pre-

43. Jack Chambers, Milestones I: The Music and Times of Miles Davis to 1960 (Toronto: University of
Toronto Press, 1983), 237.
44. Henry Louis Gates, Jr., The Signifying Monkey: A Theory of African-American Literary Criticism
(New York: Oxford University Press, 1988), chap. 2. The parenthetical g in Signifyin(g) is meant to sig-
nal African American pronunciation, as well as to distinguish it from the more general process of
signification; for discussion, see The Signifying Monkey, 46.
45. Monson, Saying Something, 106 21.
46. Peggy Lee, Bye Bye Blackbird, in Songs from Pete Kellys Blues (Decca, [LP] DL 8166, 1955),
recorded 6 May 1955.
47. The practice of revisiting the work of other composers and musicians was also common in the
music of the previous century: we thus have Beethoven Signifyin(g) on Diabelli in his Op. 120 variations
and Brahms Signifyin(g) on Paganini in his Op. 35 variations.
cultural knowle dge and musical ontology 225

ceding section, revisions to the conceptual model for a musical work play in Sig-
nifyin(g)? In what follows, I begin by developing an approach to double-voiced dis-
course in ter ms of a cultural theory of information exchange proposed by the lin-
guist Eve Sweetser. I proceed then to explore some of the conceptual models for
Bye Bye Blackbird in circulation dur ing the per iod 1926 1955, and, finally, I
return to Daviss inter pretation of the tune.

Double-Voiced Discourse and a Cultural Theory

of Information Exchange
Mikhail Bakhtin developed his notion of double-voiced discourse with reference to
the novels of Dostoevsky, going so far as to claim that this rhetor ical mode was essen-
tial to Dostoevskys style.48 Double-voiced discourse is actually the third of three
types of discourse distinguished by Bakhtin. The first type is direct discourse,
which is essentially the unmediated disquisition of the author. The second type
Bakhtin called objectified or represented discourse; its most common for m is the
direct speech of characters. In double-voiced discourse (the third type), the author
takes the direct discourse of someone else and infuses it with the authors own inten-
tions and consciousness while still retaining the or iginal speakers intentions. Thus
two discourses and two consciousnesses are present at the same moment.49
With this taxonomy in place, Bakhtin further distinguished three types of dou-
ble-voiced discourse. In the first, the discourse of the author and that of the other
person essentially cooperate the author inflects, but does not substantially change,
the discourse of the other person. The second type of double-voiced discourse is
associated with parody and irony (this is essentially the sort of discourse Monson
deals with in her analysis of Coltranes My Favorite Things). Here the discourse
of the author is opposed to that of the other person; as Bakhtin puts it, the second
[that is, authorial] voice, once having made its home in the others discourse, clashes
hostilely with its pr imordial host and forces him to serve directly opposing aims.
Discourse becomes an arena of battle between the two voices.50 In the third type
of double-voiced discourse, the other persons discourse remains outside the
purview of the authors, which nonetheless becomes influenced by that of the other
person. The result is often a hidden polemic: the authors discourse represents a
reaction to the discourse of the other person, but the latter can be heard only by
implication it is never directly represented.
Each of these three types of discourse and, most important, the three types of
double-voiced discourse can be analyzed with reference to a cultural theor y of

48. Mikhail Mikhailovich Bakhtin, Problems of Dostoevskys Poetics, ed. and trans. Caryl Emerson, with
an introduction by Wayne C. Booth, Theory and History of Literature, 8 (Minneapolis: University of
Minnesota Press, 1984), chap. 5.
49. The premise of double-voiced discourse is that the two domains of discourse remain separate
and identifiable; it is thus different from the process of conceptual blending discussed previously in chap.
2 (in that blending is concer ned with how elements from two distinct domains come together in a third
domain), but it may be seen as a related phenomenon.
50. Bakhtin, Problems of Dostoevskys Poetics, 193. I shall follow Bakhtins practice and speak of the first
voice in double-voiced discourse as that of the other and the second voice as that of the author.
226 analysis and theory

information exchange developed by Eve Sweetser. Sweetser used the theor y to

explain the results of an exper iment by Linda Coleman and Paul Kay that was
designed to show that the English word lie (in the sense of to prevaricate) func-
tioned like a Type 1 categor y. That is, rather than following a clear-cut definition,
some statements were regarded as typical lies, others as atypical. Coleman and Kay
hypothesized that a statement is judged to be a more-or-less good example of a lie
on the basis of three conditions. First, the actual statement must be false (a condi-
tion that is commonly taken as necessar y and sufficient when lie is construed as
a Type 2 category). Second, the speaker must believe the statement to be false.
Third, in utter ing the statement, the speaker must intend to deceive the listener.51
To test this hypothesis, Coleman and Kay constructed eight stor ies that satisfied all,
only some, or none of these conditions. They then asked sixty-seven subjects to
determine whether, in the case of each story, a lie had been told. In support of their
hypothesis, the results indicated that lie functioned as a Type 1 categor y: judg-
ments of what constitutes a lie were not made comprehensively, but, rather, could
be made by degree.
Although they were able to prove their hypothesis, Coleman and Kay were not
able to explain satisfactor ily what it was about the word lie, or about the condi-
tions sur rounding its proper use, that made it function as a Type 1, rather than as a
Type 2, category. Sweetser, for her part, proposed that judgments about lies were
made relative to a cultural theor y of information exchange.52 The theory, as for-
mulated by Sweetser, consists of three relatively straightforward propositions, which
can be stated as rules:
1. Try to help (do not harm). This rule is combined with the belief that knowl-
edge is helpful and that misinfor mation is har mful (which leads to the
next rule).
2. Give knowledge (do not misinform). Sweetser argues that what counts as
knowledge, in everyday terms, is what we believe. All things being equal,
we assume that what we believe matches some state of affairs or possible
state of affairs: we take what we believe to be true, to count as knowledge.
Using this understanding of knowledge together with (1) and (2), we
arrive at the final rule.
3. Say what you believe (do not say what you do not believe). If what we believe
counts as knowledge, and if we wish to be helpful, we must give knowl-
This chain of entailments represents a cultural theory of information exchange that
allows us to reach the somewhat sur prising conclusion that if a speaker (S) utters a

51. Linda Coleman and Paul Kay,Prototype Semantics: The English Word Lie, Language 57 (1981):
26 27.
52. Eve E. Sweetser, The Definition of Lie: An Examination of the Folk Models Underlying a
Semantic Prototype, in Cultural Models in Language and Thought, ed. Dorothy Holland and Naomi
Quinn (Cambr idge: Cambridge University Press, 1987), 43 66. Sweetsers models have the same func-
tion as the theor ies descr ibed in chap. 3; I have simplified some aspects of her account in the interests
of concision.
cultural knowle dge and musical ontology 227

proposition (P), then that proposition is true. The causal chain can be represented
a. S said P. [as part of an infor mation exchange; from rules (1) and (2)]
b. S believes P. [(a) combined with rule (3)]
c. Therefore, P is true. [(b) combined with the understanding that we believe
what is true and do not believe what is false]
This is, without a doubt, a greatly simplified perspective.What is important is that
the model of information exchange set out by the theory works quite a bit of the
time, and when it does not work, the theory provides a way to inter pret where
things went wrong.
To illustrate, let us consider two situations. In the first, I claim that I saw a fr iend
of mine walking down the street. On closer inspection, it turns out to be someone
who just looked like my friend. According to the simplest definition, my claim is a
lie a statement that is not true. However, since my behavior as a whole fits with
the cultural theor y of information exchange (my intent was to provide helpful
information), my claim will most likely not be judged a lie. In the second situation,
I see someone walking down the street whom I mistake for my friend. Remem-
bering that I owe my friend money and being a cheap sort, I claim that the person
I see is not my friend. According to the simplest definition, this is not a lie, since my
claim is true: the person walking down the street is not, in fact, my friend. Nonethe-
less, since I deliberately said what I thought to be misinfor mation, I have violated
the theory of information exchange and my claim might well be judged a lie.
Of course, this straightforward theory of information exchange does not operate
in isolation but is inter preted through other cultural theor ies to explain types of dis-
course that seem to contravene the theory. Jokes, tall tales, and fiction all may have
limited truth-value, but their lies are permitted as part of a cultural theor y that
interprets certain types of exceptional behavior as entertainment. Social lies are
based on the recognition that not all for ms of knowledge are helpful: rarely when
someone remarks How are you? do they really want to know how we are. The
cultural theory of information exchange is thus filtered through a theory that says
maintaining social order through polite behavior is more important than saying
what you believe to be true.53
Applying Sweetsers example of a cultural theor y of information exchange to
Bakhtins first two types of discourse is relatively straightforward. All things being
equal, we assume that direct discourse the voice of the author involves the
truth; we similarly give the benefit of the doubt to the objectified discourse of the
authors characters.54 With double-voiced discourse, however, the application of the

53. For a related view of how information exchanges contr ibute to and constitute social order, see
Michael B. Bakans account of how various Balinese musicians reported the results of a gamelan com-
petition in Music of Death and New Creation: Experiences in the World of Balinese Gamelan Beleganjur,
Chicago Studies in Ethnomusicology (Chicago: University of Chicago Press, 1999), chap. 5.
54. The genre of fiction, which is the focus of Bakhtins analysis, requires an intr iguing nesting of
theories about infor mation exchange. On the level of everyday life, fiction is generally recognized as
nonfactual (except by the delusional).Within a given work of fiction, however, we usually apply the the-
ory of information exchange outlined by Sweetser: we assume that characters say what is true in the con-
228 analysis and theory

theory is a bit more complicated. With the first type of double-voiced discourse
where the discourse of the author and the other person essentially cooperate the
idea that both voices speak the truth is easily accommodated. What results is a sort
of heterophony of discourse: two voices saying the same thing in different ways.
With the second and third types parody and hidden polemic the contradictions
and tensions between the two voices suggest that one of the voices must be violat-
ing the pr inciples of the cultural theor y of information exchange. The traditional
resolution of the problem, implicit in Bakhtins analysis of discourse as a whole, is
that in double-voiced discourse, the voice of the author is to be pr ivileged: it is the
voice of author ity.55 This resolution requires modification of Sweetsers cultural
theory of information exchange by the addition of one further proposition:
4. The voice of authority is always helpful.
On the one hand, it is difficult to ascr ibe this inter pretation to Bakhtin without a
crushing sense of irony, given his personal history. But on the other hand, it is cer-
tainly the case that, in his analyses of double-voiced discourse, the first (non-
author ial) voice is the one he most often distrusts.
Given this analysis, equating Gatess notion of Signifyin(g) with Bakhtins dou-
ble-voiced discourse brings us to two conclusions. First, the voices of black rhetor ic
become voices of authority: they are the helpful voices that speak the truth. Second,
Signifyin(g) is by no means unique to the African American community indeed,
Dostoevsky and countless other authors are also Signifyin(g). But these conclu-
sions are hardly satisfying, for they trivialize the points Gates wishes to make about
African American literature and culture and miss the ways language can be used
against oppressors.
The solution to these problems lies in two important adjustments to the cultural
model of information exchange adumbrated here. First, a distinction must be made
between author and author ity. For the African American community, the first
voice in double-voiced discourse was almost always the voice of author ity. Since
experience had shown that the voice of authority was almost always unhelpful, it
was a voice to be distrusted. Rather than trusting one of the voices of a double-
voiced discourse, it was important actively to distrust one of those voices; by default,
the nonauthor itative voice of the author was to be regarded as the helpful one. Sec-
ond, it did not particularly matter if the statements of the author were true or not.
Indeed, as Gates points out, one of the common African American ter ms for Sig-
nifyin(g) is lies.56 What mattered more was whether the author, in some way,
adhered to the first principle of information exchange: to be helpful. And it is clear
that, to an oppressed people, speech that co-opts, mocks, or indirectly challenges the
voice of authority, whether the statements made in the course of that speech are

text of the story.When a character says something that is not true within the context of the story (even
though it may be true in the everyday world), we then have to fall back on this theory or its elaborations
to make sense of the utterance.
55. The pun is not sur prising, since authority and author share the same Latin root (auctor).
56. Gates, The Signifying Monkey, 56, 71.
cultural knowle dge and musical ontology 229

true or not, is singularly helpful.With these adjustments, distinctions between double-

voiced speech and Signifyin(g) become clear: Signifyin(g) relies on a modified the-
ory of information exchange that assumes that the first voice in double-voiced dis-
course is the voice of an extracultural author ity and that this author ity is not to be
believed. Under these circumstances, double-voiced discourse is, as Bakhtin charac-
terized it, an arena of battle between the two voices. However, it is an arena in
which the author is at a distinct disadvantage and must rely on whatever combina-
tion of truths, half-truths, and lies is necessary to win the day.
My intent with this analysis is to show how Bakhtins notion of double-voiced
discourse functions relative to a highly simplified, yet quite serviceable, rendering of
a cultural theory of information exchange and how the double-voiced discourse of
Signifyin(g) relies on a modified version of this theory. In applying this perspective
on discourse to music, I restrict my discussion to the level of the conceptual model
for a work. Musical discourse, understood in this way, is thus concer ned with the
realization and projection of the conceptual model for a work. In ter ms of the cul-
tural theor y of information exchange (extended to music), our assumption is that
the perfor mer intends to project this model and that it is valid for the piece. My
decision to impose this restriction is based on practical reasons. It will keep discus-
sion focused on the main topics of this chapter, it will greatly simplify the analyses,
and it requires no strong analogy between linguistic and musical discourse. There is,
however, no reason that, mutatis mutandis, this treatment could not be extended to
more finely grained levels of musical discourse.

Conceptualizing Bye Bye Blackbird, 1926 1955

Bye Bye Blackbird was wr itten in 1926 by the song-wr iting team of Mort Dixon
and Ray Henderson. The score, given in example 5.4, reveals a song typical of the
thousands wr itten dur ing the days of Tin Pan Alley. It consists of an eight-measure
introduction (taken from the concluding phrase of the refrain); a two-measure vamp
to allow the vocalist to get ready; a thirty-two-measure verse in AABA' form; and
a thirty-two-measure refrain in AA'BA" form (where A' presents the mater ial of the
A section a diatonic step higher, and A" changes the or iginal A section at midphrase
so that it ends on the tonic).57
The general topic of the song centers around a nar rator in alien or unhappy sur-
roundings, symbolized in the first verse by the trope of winter, the blackbirds dole-
ful pronouncements, and the use of minor mode. The nar rator rejects these condi-
tions and looks to where theres sunshine galore, a promise held out by the refrain:
a waiting loved one, a familiar bed, and a light burning in the window, all intro-

57. At least two other scores for Bye Bye Blackbird have been in circulation. Both of these are in
F major, and both shorten the verse to a sixteen-measure form by omitting the inter nal AB sections. One
score, with the copyright renewed and assigned to the Remick Music Cor poration, has a short, four-
measure introduction; the sixteen-measure verse; and the refrain. The other, copyrighted in 1948 by the
Remick Music Cor poration, was printed by Francis, Day and Hunter, London. It gives the refrain first,
followed by the introduction and verse. I have found no recordings of performances based on either of
these scores.
230 analysis and theory

example 5.4 Ray Henderson and Mort Dixons Bye Bye Blackbird, Fred Ahlert
Music Group

duced against the backdrop of a spr ightly G major. The second verse (which
replaces the blackbird with a bluebird) reflects the hope offered by the refrain, con-
ditioned nonetheless by the return to minor mode and references to longing, fad-
ing like a flower, and hours that are like one long tear.
Recordings of the period indicate typical treatments of the song. In their version
of 22 April 1926, for example, Bennie Kruegers Orchestra begins with a short
introduction (which bears no resemblance to the published one) and proceeds to an
cultural knowle dge and musical ontology 231

example 5.4 (continued)


instrumental version of the refrain, verse, and refrain.58 (As with the early recordings
of I Got Rhythm, the Krueger ar rangement uses the verse as an interlude
between the refrains.) This is followed by a sung refrain and one final instrumental
refrain. The tempo of the recording (h = 116 120) is a bit faster than the Mod-
erato indicated by the score, but it is perfectly suitable for a foxtrot.59 Gene Austins
58. Bennie Kruegers Orchestra, Bye Bye Blackbird (Brunswick, [78] no. 3186, 1926).
59. The song is in fact identified as a foxtrot on the label of the record, which also includes the
rubric for dancing.
232 analysis and theory

example 5.4 (continued)

recording, made a week later, follows the score rather more closely.60 Although he
adds his own introduction and piano accompaniment and varies the melody in the
third phrase of the verse, his perfor mance otherwise confor ms to the score given in
example 5.4, even adher ing to a somewhat slower tempo than did Kruegers group
(h = 96 100). By most standards, Bye Bye Blackbird was a hit: it was recorded at
least fourteen times between March and June of 1926, and Austins recording was
no. 1 on the Billboard charts of 4 September of that year.

60. Gene Austin, Bye Bye Blackbird (Victor, [78] no. 20044-B, 1926), originally recorded 29 April.
cultural knowle dge and musical ontology 233

example 5.4 (continued)

The various perfor mances of Bye Bye Blackbird recorded in 1926 could be
collected into a category organized around a conceptual model quite similar to that
used for early versions of I Got Rhythm. The model, which is diag rammed in
figure 5.5, consists of five cor related conceptual elements, character ized as the
things necessary for an adequate perfor mance of the song. However, where the pop-
music model for I Got Rhythm continued to have some cur rency in the years
just after the first appearance of the song (in part because of its association with the
musical), the model for Bye Bye Blackbird appears to have gone dor mant for a
234 analysis and theory

figure 5.5 Model for the category performances of Bye Bye Blackbird

time: there are few signs of the song having been recorded after 1926 and through-
out the 1930s.61
New recordings of Bye Bye Blackbird did begin to appear in the 1940s. There
are a number of indications, however, that by then the conceptual model for the
song had changed. The novelty recording by Freddie Schnickelfritz Fisher and
His Orchestra (made on 22 February 1941), with its rapid succession of unlikely
instruments (including solo whistling and bar itone saxophone) and its shave-and-
a-haircut ending, is just short of an all-out joke.62 In a 1941 Ger man propaganda
recording by Charlie and His Orchestra, the words of the song are altered to cele-
brate early Axis victor ies in the Second World War and to taunt the Br itish (Bye
Bye Blackbird becomes Bye Bye Empire).63 And in a tr ibute to the songs of
1926 recorded on 27 May 1942 by the piano duo of Marlene Fingerle and Arthur
Schutt, Bye Bye Blackbird is the final song in a medley that includes I Know
That You Know and Baby Face.64 In each of these recordings, only the refrain
of the song is used, and the tempo varies from the moderate swing of Charlie and
His Orchestra (h = 84) to the Fisher Orchestras foxtrot (h = 120 126).While there
are some signs the song has entered the repertoire of standards its verse has

61. Recording, of course, is not the only measure of popular interest in a song: sales of records, piano
rolls, and sheet music; programming of the song in live performances; and the playing of recordings on
radio broadcasts and in juke boxes would also provide an indication of continued interest in the song.
Unfortunately, no infor mation on these indicators is presently available for Bye Bye Blackbird.
62. Freddie Schnickelfritz Fisher and His Orchestra, Bye Bye Blackbird (Decca, [78] no.
25357/3788, 1941).
63. Charlie and His Orchestra, Bye Bye Blackbird, in Charlie & His Orchestra: German Propaganda
Swing 19411942 (Harlequin, [CD] HQ CD 03, 1990).
64. Marlene Fingerle and Arthur Schutt, Medley:I Know That You Know,Baby Face,Bye Bye
Blackbird, in Song Hits of 1926, Songs of Our Times (Decca, [LP] DL 5170, 1950).
cultural knowle dge and musical ontology 235



JHE"*HE"*H9%&@45$)K )%:&5:>]*?"$B($<"$7*67"*8%$5%#5(:*
L?(?*7(:>C*&%3*+-0+M 46#*95##9"*5<?$(857%#5(:*



figure 5.6 Model for the category performances of Bye Bye Blackbird, ca. 1941

fallen away, and it is readily available for parody it is also clearly marked as an
old song. The conceptual model for the categor y performances of Bye Bye Black-
bird, ca. 1941, then, would consist of five cor related conceptual elements (which are
diagrammed in fig. 5.6).65
As a consequence of a growing interest in the 1920s as an era, this model was, if
anything, reinforced dur ing the postWorld War II per iod. Following the phenom-
enal success of The Jolson Story in the fall of 1946,Warner Brothers films contracted
in 1947 for the filming of a similar story with (and about) Eddie Cantor. Although
Warner Brothers ultimately dropped the project, Cantor and Sidney Skolsky (pro-
ducer of The Jolson Story) eventually filmed The Eddie Cantor Story independently,
and it was released in 1953.66 As part of the soundtrack for the movie, Cantor
recorded Bye Bye Blackbird in something approximating the style of the 1920s,
although using only the refrain of the song.67 A similar approach can be heard in

65. I do not mean to suggest, with the idea that Bye Bye Blackbird was written for a previous
generation, that the fifteen years between 1926 and 1941 actually fill the span of a human generation.
However, I do think this per iod is long enough to mark a significant change in popular tastes, which is
what this portion of the conceptual model aims to capture.
66. Herbert G. Goldman, Banjo Eyes: Eddie Cantor and the Birth of Modern Stardom (New York:
Oxford University Press, 1997), 283 84.
67. Eddie Cantor, Bye Bye Blackbird, in Songs by Eddie Cantor from The Eddie Cantor Stor y
(Capitol, [LP] L-46, 1954). There is no evidence that Cantor recorded Bye Bye Blackbird during the
1920s or 1930s, although he might well have performed it in live shows; in fact, Goldman character izes
it as one of the standards of Cantors repertory from the 1920s. By most accounts, The Eddie Cantor Story
was a flop; its failure is generally attr ibuted to the inadequacies of Keefe Brasselle (who played the role of
Cantor) and to a scr ipt that was little more than a remake of The Jolson Story. Of the soundtrack, how-
ever, Goldman wr ites: The musical ar rangements were first rate, and the vocals better than might be
236 analysis and theory

Johnny Maddoxs recording of the song on piano (accompanied by a rhythm sec-

tion), released in 1955, which inter prets it in a quasi-ragtime style as part of a col-
lection of 1920s songs intended to accompany tap dancing.68
During this same per iod the general fascination with the character and music of
the 1920s was given a specific focus when there was a resurgence of interest in early
jazz, variously called New Orleans or Dixieland jazz.69 Partly, this came about
because of a str ike by musicians against the recording industry from 1942 to 1944,
which led record companies to reissue older recordings. Around the same time,
there emerged an audience who felt the swing era had passed but who were not
attracted to be-bop, the new jazz style that emerged dur ing the Second World War.
In 1950, the actor Jack Webb (now chiefly remembered for his role as Sergeant Fr i-
day on the Dragnet radio and television ser ies) developed a radio ser ies around a jazz
musician in 1920s Kansas City, reflecting contemporary interest in the roaring 20s
and Webbs own enthusiasm for early jazz.70 The radio ser ies was eventually devel-
oped into the movie Pete Kellys Blues, which Webb directed and in which he also
starred. In the movie, Pete Kelly is a bandleader threatened by a gangster who is
attempting to extract protection money from musicians. Kelly initially tr ies to resist
but eventually capitulates. As part of his agreement with the gangster, Kellys band
has to take on the gangsters girlfriend, Rose Hopkins (played by Peggy Lee), as a
singer. Rose is an alcoholic but can still render a moving ballad. A tragic character,
Rose is eventually reduced to the mentality of a five-year-old by alcoholism and by
a beating from the gangster.
As part of the promotion of Pete Kellys Blues as a jazz movie, Lee, together
with Ella Fitzgerald (who played a blues singer working in a roadhouse), recorded
an album of songs associated with the movie on 10 May 1955. Although Bye Bye
Blackbird appears in the movie, neither Lee nor Fitzgerald sings it. Instead, it is
rendered by a drunken chorus of band members returning from a late-night party.
On the album, Lee retains Roses persona from the movie and sings the song (con-
sisting only of the refrain) as a slow ballad (q = 72 76), with slight allusions to Bil-
lie Holidays distinctive style of inflection.71 The arrangement makes use of com-
plicated har monies typical of the treatment of ballads in 1940s and 1950s jazz and

expected, considering that Cantor had suffered a major heart attack less than three months earlier and
that his voice had lost its for mer brilliance after World War II. Goldman, Banjo Eyes, 284.
68. Johnny Maddox, Bye Bye Blackbird, in Tap Dance Rhythms (Dot, [LP] DLP 3008, 1955).
69. Chip Deffaa, Traditionalists and Revivalists in Jazz, Studies in Jazz, 16 (Metuchen: Scarecrow Press;
Institute of Jazz Studies, Rutgers, The State University of New Jersey, 1993), 132. In the following I refer
to this style of jazz as traditional jazz, retaining the quotation marks to highlight the function of the
term as a rubric rather than a judgment about the place of this music within any histor ical account of
jazz styles. See also Ber nard Gendron, Moldy Figs and Moder nists: Jazz at War (1942 1946), in Jazz
among the Discourses, ed. Krin Gabbard (Durham, N.C.: Duke University Press, 1995), 31 56.
70. Charles Emge, Meet Jazzman Jack Webb: Swing Era Helped Make Him Music-Conscious,
Downbeat 22 (10 August 1955): 8.
71. The character of Rose seems to have had a profound personal effect on Lee. In her autobiogra-
phy, she speaks of Rose in the third person, speculates on whether a visit from James Dean was intended
for her or for Rose, and wr ites of moving out of Roses trailer [on the set] once the film was com-
pleted. See Peggy Lee, Miss Peggy Lee: An Autobiography (New York: Donald I. Fine, 1989), 165 67.
cultural knowle dge and musical ontology 237

adds words for the second pass through the refrain that signal a new perspective on
the song:
A while ago when I was young, I heard a song and heard it sung
[male chorus sings as distant echo] Bye bye blackbird
I dont know why it makes me sad, a happy song should make me glad
[male chorus] Bye bye blackbird

Singing in her persona as a dissolute, tragic chanteuse, Lee evokes a melancholy

that, up to this point, had not been associated with Bye Bye Blackbird.
Pete Kellys Blues was, at best, a mixed success cr itically, and Lees treatment of
Bye Bye Blackbird might well have been forgotten (especially since she did not
sing the song in the movie). Fate, however, intervened: the singer and pianist Bobby
Short heard and liked Lees recording and decided to approach Bye Bye Blackbird
in a similar fashion, finding reflective sadness in the refrain rather than joyful antic-
ipation.72 In his recording of 15 September 1955, Short also adopts a slow ballad
tempo (q = 84) and fashions his accompaniment out of rich, progressive harmonies.
To provide an introduction, he starts with the B section of the refrain, the one sec-
tion that retains the sense of alienation generated by the or iginal verse: No one
here can love and understand me . . . The effect is to cast a shadow over the
remaining lyr ics of the refrain, emphasizing what is negative in them the nar ra-
tors care and woe, distant loved one, and need to travel late into the night rather
than the pleasant things that wait at home. Short also adds his own lyrics for the sec-
ond pass through the refrain (with words adapted from the song Im a Little Black-
bird Looking for a Bluebird):73
Ive been all over the east and the west, in search of clover to feather my nest,
Bye bye blackbird
I have a daydream the same as you do, why would a daydream have to be hoo-doo?
Bye bye blackbird
No one here can love and understand me,
Oh, what hard luck stor ies they all hand me. (etc.)74

With these two recordings by Lee and Short, Bye Bye Blackbird was trans-
formed from an upbeat popular song in a fast tempo, designed for dancing a
hop tune, as Short has called it75 into a melancholy jazz ballad in a slow tempo,
designed for listening. The motivation for this change was not, strictly speaking,
musical, for most of the music in Pete Kellys Blues kept to a traditional or Dix-
ieland style.76 Instead, the change was initially motivated (that is, in Peggy Lees

72. Personal communication from Bobby Short, 8 March 2000.

73. Im a Little Blackbird Looking for a Bluebird was wr itten by Clarke Grant, Arthur Johnston,
George W. Meyer, and Roy Turk. It was recorded by both Louis Armstrong and Sydney Bechet in the
74. Bobby Short, Bye Bye Blackbird, in Bobby Short (Atlantic, [LP] 1230, 1956), recorded 15 Sep-
tember 1955. As will be noted, Shorts revised lyr ics return to Dixons or iginal refrain at midpoint (with,
No one here can love and understand me).
75. Personal communication from Bobby Short, 8 March 2000.
76. The soundtrack for the movie (with pr iceless nar rations by Jack Webb) consisted exclusively of
238 analysis and theory

figure 5.7 Jazz ballad model for the category performances of Bye Bye Blackbird

recording) by a romanticized notion of the female jazz singer, itself derived from
any number of nineteenth-century fantasies about perfor mers.77
The new conceptual model for Bye Bye Blackbird suggested by Lees and
Shorts recordings, and diagrammed in figure 5.7, consists of four elements neces-
sary for an adequate perfor mance. Gone is the sense of history that had freighted
approaches to the song since the early 1940s; it is replaced by a sense of immediacy
generated by the necessar ily reflective nature of the perfor mances.
It may be useful at this point to consider differences in the way conceptions of
I Got Rhythm and Bye Bye Blackbird changed over roughly the same per iod
in time. Because I Got Rhythm was an endur ing part of the popular music and
jazz repertoire, its conceptual model changed only g radually, through a process of
slow but relatively constant negotiation. In contrast, no such process is in evidence
for Bye Bye Blackbird. After its first blossoming, the tune, and its conceptual
model, faded from popular consciousness. When it was recalled dur ing the early
1940s, it was as part of a larger process of revisiting, through music, an earlier era:
Bye Bye Blackbird was simply emblematic of a popular tune from the 1920s.
When Peggy Lee introduced her version, the change to the conceptual model was
sudden and dramatic, by fiat of her character in the movie.

traditional arrangements. See Pete Kelly and His Big Seven (Matty Matlock and others), Pete Kellys
Blues (BMG Special Products, [CD] DRC12081, 1998), originally recorded 7, 9, 11 June 1955.
77. For a recent, and stimulating, discussion of some of these fantasies, see Susan Ber nstein, Virtuos-
ity of the Nineteenth Century: Performing Music and Language in Heine, Liszt, and Baudelaire (Stanford, Calif.:
Stanford University Press, 1998), chap. 3.
My character ization of the persona developed by Lee in connection with Pete Kellys Blues should
not obscure the powerful presence of African American women who sang the blues and jazz, or the very
real tr ials to which these women were subjected. (For a discussion of the role of these women in shap-
ing the blues in particular, see Susan McClary, Conventional Wisdom: The Content of Musical Form [Berke-
ley: University of California Press, 2000], 42 49.) I do believe, however, that Lee, as a white entertainer
in the United States of the 1950s, was able to choose this persona (much as she chose to emulate Billie
Holidays voice in the course of her performance of Bye Bye Blackbird) and that part of the frame-
work for this choice was a romanticized notion of the suffer ing perfor mer.
cultural knowle dge and musical ontology 239

Thus by the end of 1955 there were two conceptual models for Bye Bye Black-
bird in circulation. The first conceptualized the tune as a somewhat dated but
nonetheless upbeat pop song; the second conceptualized the tune as a slow, melan-
choly ballad. There may have also been still circulating a third, and more sinister
model, derived from the first, that inter preted the song in racial ter ms a signal for
African Americans to be on guard, that they were not wanted here. It was into the
context provided by these models that Miles Davis introduced his version of Bye
Bye Blackbird.

Daviss Bye Bye Blackbird, Double-Voiced

Discourse in Music, and Signifyin(g)
The recording that Davis made with Garland, Chambers, Jones, and Coltrane was
strictly instrumental, at a tempo slower than the hop version of the tune (q =
120), but notably faster than the tempos adopted by Lee and Short. After a br ief
introduction by Garland, Chambers, and Jones that establishes a loose, almost lop-
ing, groove, Davis enters playing the refrain. From his first notes, Davis announces
the song as a jazz ballad, using a languid approach to the melody and muted trum-
pet to create an understated, slightly distant reflection on the song. After the head,
Davis takes two choruses, followed by two choruses from Coltrane that occasionally
hint at a double-time feel. Garland takes a chorus and a half , and, at the br idge of
Garlands second chorus, Davis reenters with the melody and guides the perfor-
mance to a conclusion.
Daviss principal contr ibutions are to infuse Bye Bye Blackbird with an ir re-
sistible groove and to establish it as a venue for instrumental commentar y and
expansion. Although these contr ibutions are by no means insignificant Daviss
recording virtually established Bye Bye Blackbird as a jazz standard, something it
had never been before one can also hear the influence of Lees and Shorts record-
ings in his version. There is, obviously, the slower tempo and hint of melancholy
adopted from the jazz-ballad model (although the g roove Davis sets up tends to
move the tune back toward the pop-music model). There is also, however, the same
sense of immediacy that marked Lees and Shorts perfor mances: Bye Bye Black-
bird is not, in Daviss hands, about nostalgia.78 Davis has modified the conceptual
model for the tune used by Lee and Short, but he has also retained enough of it that
we can still recognize its main elements.
The immediate relationship between the jazz-ballad model for Bye Bye Black-
bird and the model Davis set out in his recording suggests that double-voiced dis-
course can be accomplished in music when one conceptual model for a musical
work is understood to make reference to another conceptual model for the same

78. The element of nostalgia has continued to cling to Bye Bye Blackbird as a popular tune (as
distinct from a jazz standard) and has been exploited in a number of films. In Jonathan Demmes Melvin
and Howard (1980), Bye Bye Blackbird is the one song Howard Hughes is able to recall when Melvin
Dummar presses him on pain of being thrown out of Dummars truck to sing something, and it
becomes a deeply felt recollection of Hughess childhood. In Nora Ephrons Sleepless in Seattle (1993),
Bye Bye Blackbird is cast as the song the boys deceased mother used to sing to him when he couldnt
240 analysis and theory

work. If we know Peggy Lees recording of Bye Bye Blackbird, hear Miles Daviss,
and understand the latter as making reference to the for mer, musical double-
discourse takes place.79 Musical double-discourse thus requires, minimally, a listener
who knows at least two conceptual models for a given work. Connections between
the models may be drawn out by the perfor mer, but they may be more oblique and
constructed mostly by the listener. Each of the three types of double-voiced dis-
course descr ibed by Bakhtin can be created: an inflection of the first conceptual
model by the second (which is what Davis does); a parody of the first conceptual
model by the second (which is what Monson descr ibes Coltrane as doing); or tacit
reference to the first model by the second (which is one way to conceive how
derived tunes operate).
But what of Signifyin(g)? Signifyin(g) would require that the first conceptual
model be recognized as the authoritative one and that its author ity is not to be
believed: that is, it is not a valid model for the work. Validity is then g ranted by
default to the conceptual model used by the perfor mer. A consequence is that the
validity of this model depends on the rejection of the model associated with author-
ity; it is not granted independently. Under these circumstances, a case can be made
that, in his recording of My Favorite Things, Coltrane is Signifyin(g) on Rodgers
and Hammersteins song: the validity of the Broadway-show-tune model for this
song has been rejected, and it is Coltranes model of the tune that is accorded valid-
ity. It is much less clear, however, whether Davis is Signifyin(g) on Bye Bye Black-
bird. It is difficult to ascr ibe author ity of a sort that one should be skeptical of to
either Lees or Shorts recordings, and it is unclear whether Davis was aware of, or
cared about, the pop-music model for the tune (with or without its overlay of
racism). This does not mean that a listener might not hear the recording by Davis as
Signifyin(g) on Bye Bye Blackbird, only that the relationship of Daviss rendition
of the song to the concept of Signifyin(g) is much more equivocal.
Under what circumstances might we hear Davis as Signifyin(g) on Bye Bye
Blackbird? First, Davis could, in the course of a performance (recorded or other-
wise), make an obvious reference to one of the earlier models for the tune, perhaps
by playing a fragment of the verse or by evoking another nostalg ic tune from the
1920s. The other model for Bye Bye Blackbird would thus be brought to our
attention (even if only as something anomalous in Daviss performance) to be then
regarded as a voice of author ity under challenge. Second, we could br ing to Daviss
performance a knowledge or awareness of one of the earlier models, hear that he
reverses certain of the key elements of this model, and inter pret this as subverting
authority. Third, Davis could, through either word or deed, situate himself as in dia-
logue with what he regarded as an author itative version of the tune.
With this analysis of Daviss recording of Bye Bye Blackbird I do not mean to
suggest that Signifyin(g) is a rare occur rence in music, but only that it involves a
complex matr ix of cultural knowledge and that the listener must understand the
process of Signifyin(g) as taking place. When this matr ix of knowledge is not in
place, what may be left to hear is only double-voiced discourse, or unmediated

79. Although my focus here is on recordings, the same situation would obtain were our exper ience
with live, as opposed to recorded, performances.
cultural knowle dge and musical ontology 241

authorial discourse. (Indeed, all music could be heard as involved in double-voiced

discourse of a quite general kind, in that the conceptual model for each work is in
dialogue with the conceptual models for other works.) Double-voiced discourse,
as well as Signifyin(g), becomes interesting when tensions between the models
become obvious that is, when the tensions become as important as the models

Signifyin(g) and Cultural Knowledge

Given the cognitive complexity of the process of Signifyin(g), it is well worth ask-
ing why the notion has become a trope among those who work on jazz and pop-
ular music. Three reasons come to mind. First, Signifyin(g) offers a clear alter native
to the Saussurean model of meaning construction: in Signifyin(g), meaning is some-
thing that is not fixed but is created by the listener.80 Second, Signifyin(g) provides
a way of interpreting the cultural practice of jazz and popular music in ter ms devel-
oped to account for the unique profile of one of its most important influences,
African American culture. Third, applying Signifyin(g) to music allows music cr it-
ics to avail themselves of the power and prestige of literary criticism.
Of these three, it is perhaps only the second that has much relevance to music,
as opposed to the practice of writing about music. Musical Signifyin(g) is a complex
process, distinct from linguistic Signifyin(g) and thoroughly implicated in the dense
matrix of cultural knowledge. To explain how it works when it works requires
not only a theory of Signifyin(g) but also a way to account for musical ontology, an
explanation of the framework for discourse provided by cultural theor ies of infor-
mation exchange, and a theory of double-voiced discourse.With these in place we
can start to see and hear how Signifyin(g) shapes musical discourse.

cultural knowledge and

musical ontology
Clifford Geertz was skeptical about using descr iptions of individual cognitive
processes to build an account of culture, and rightly so. Cognitive structure is
incredibly complex, and the relationship between individual knowledge and cul-
tural knowledge is not a simple one. Nonetheless, one can discer n patter ns of con-
sistency in the features selected for defining and evaluating cultural products, pat-
terns that are indicative of cultural knowledge. The approach to cultural knowledge
I have developed in this chapter is meant to capture these consistencies and to facil-
itate a discussion of how cultural knowledge is organized, sustained, and changed. It
is not meant as a literal rendering of what someone has in mind in evaluating par-
ticular cultural products. The approach is thus an inter pretation of culture, but of a

80. This is essentially the approach taken by Robert Walser in Out of Notes: Signification, Inter-
pretation, and the Problem of Miles Davis, in Jazz among the Discourses, ed. Krin Gabbard (Durham,
N.C.: Duke University Press, 1995), 165 88.While my approach to Signifyin(g) is broadly consonant
with this treatment of meaning construction, my argument is made from the perspective of research in
cognitive science rather than structuralism.
242 analysis and theory

somewhat different sort than those called for by Geertz, for it assumes that culture
is what we know, rather than the specific things we do, make, or say.
As members of a culture, one of the things we know is how to categor ize
sequences of sound events as music (or whatever other ter m is operational in our
culture). Knowing that a Beethoven str ing quartet is music (to return to Geertzs
example) means that it confor ms to our conceptual model for music, no more and
no less. Similarly, John Cages 4'33" is music if it confor ms to our conceptual model
for music; if it does not, we may simply regard it as a rather strange r itual in which
a group of people sit in silence for a specified per iod of time.81 The conceptual
models around which we organize our categor ies of music are also the means by
which we negotiate or resist cultural change, guide perfor mance, and engage in
complex patter ns of musical discourse. They are, quite simply, the stuff of musical
Quite obviously, what I offer here is only a beginning. In particular, the account
of the conceptual models appropriate to cultural knowledge needs to be refined
through further study. Coleman and Kays study of the meaning of lie, together
with Sweetsers interpretation of that study, offers an interesting example of how
relevant empir ical research might be conducted. If conceptual models truly have
value as a tool for descr ibing cultural knowledge, then it should be possible to
develop exper imental protocols that test whether and how members of a culture
use such models. What will result is not a simple picture of what someone has in
mind when reasoning about a given situation (which is at best a dubious goal) but
a better understanding of the elements of knowledge that make it possible to func-
tion as a member of a culture.
Another obvious extension of the work presented in this chapter is to other
repertoires, both contemporar y and histor ical. In some cases (for instance, when
dealing with a remote histor ical per iod), the application may be so speculative as to
be useful chiefly in providing a fresh perspective, but in others it will yield sur pris-
ing results. There are processes every bit as complex and challenging as Signifyin(g)
that operate in other repertories for instance, the intr icate patter ns of influence
and innovation that abound in the music of Gustav Mahler and Arnold Schoen-
berg patterns that await only our interest and efforts to discover them.
Perhaps the most intr iguing consequence of the argument put forth in this chap-
ter is that, if musical works are cognitive categor ies, and if music is a cognitive cat-
egory, then music theory is about the study of categor ies or more typically the
conceptual models around which musical categor ies are organized. Understood this
way, music theory immediately escapes the g ravitational force field of the text
that has at times kept the literary criticism of the previous generation earth-bound
(even while remaining relentlessly recondite). Thus liberated, music theorys own
traditions of high abstraction and unabashed pragmatism may yet find a place in
contemporary intellectual discourse.

81. Although I have not tr ied (or wanted) to develop an overarching conceptual model for music
here, it is worth noting in this connection that Cages four-and-a-half minutes of silence has a score and
has been presented in concert numerous times. If these things are important elements of someones con-
ceptual model of music, they would most likely regard Cages piece as music.
chapter six
words, music, and song:
the nineteenth-century lied

O n 15 December 1822, the poet Wilhelm Mller noted the publication of a

group of songs that took his poems as their texts. In a letter to the composer
of the songs, Bernhard Klein, Mller observed, indeed, my songs lead . . . only half
a life, a paper-life, black upon white . . . until music breathes the breath of life into
them, or at least, when it slumbers within, calls it out and wakens it.1
That music could breathe life into poetr y is a sentiment also expressed in a
eulogy for a composer Mller chose not to address in the foregoing letter, but
whose settings have kept Mllers poems from disappear ing into the obscure
recesses of nineteenth-centur y German poetr y. Speaking of Franz Schuberts
Lieder, Josef von Spaun wrote:
That which moved the poets breast Schubert rendered true and transfigured in tone
in each of his songs as none had done before him. Every one of his songs is actually a
poem on the poem that he set to music and not only for sentiment, which is doubt-
less at home in songs, but also for all the magic of fantasy, for its super natural char m
and its noctur nal terrors, could he find the perfect moment yes, even for the lofty
seriousness of reflection he had a language.2

These two remarkably similar views of the relationship between poetr y and
music are, without a doubt, as much a testament to the intellectual and artistic life

1. Quoted in Carl Koch, Bernhard Klein (17831832): Sein Leben und seine Werke (Leipzig: Oscar
Brandstetter, 1902), 34, n. 8. The complete text runs as follows: In der That fhren meine Lieder, die
zu einem declamator ischen Vortrage, wenige ausgenommen, durchaus nicht geeignet sind, nur ein halbes
Leben, ein Papierleben, schwarz und wei(Ach, wie traur ig sieht in Letter n Schwarz auf Wei das Lied
mich an!) bis die Musik ihnen den Lebensodem einhaucht, oder ihn doch, wenn er dar in schlummert,
herausruft und weckt.
2. Was des Dichters Brust bewegte, hat Schubert in jedem seiner Lieder treu und verklrt in Tnen
wiedergegeben, wie keiner vor ihm. Jede seiner Lieder-Compositionen ist eigentlich ein Gedicht ber
das Gedicht, das er in Musik setzte; und nicht fr Empfindung nur, die im Liede wohl daheim ist, son-
dern auch fr jeden Zauber der Fantasie, ihren ber irdischen Reitz und ihre nchtlichsten Schrecken,
wute er den eigensten Augenblick zu finden, ja selbst fr den hohen Er nst des Gedankens hatte er eine
Sprache. Till Gerrit Waidelich, ed., Franz Schubert: Dokumente, 1817 1830,Verffentlichungen des Inter-
nationalen Franz Schubert Instituts, 10 (Tutzing, Germany: Hans Schneider, 1993), 516 17.

244 analysis and theory

and ideals of early nineteenth-century Germany and Austria as they are clear-eyed
proposals for a regard of song.3 And yet they capture something important about
the possibilities that ar ise when words and music come together: under certain cir-
cumstances the result is far more than simply the sum of its parts.
The topic of song has fascinated wr iters on music at least for the last two cen-
turies, and a study of the study of song could be its own book. In a 1992 essay, Kofi
Agawu offered a survey of some of this work (as well as his own contr ibution to it)
and identified four basic models for the analysis of song, of which three were fairly
well entrenched in the literature on song.4 The first model der ives from the work
of Suzanne Langer and operates on the premise that when words and music come
together in song, the music wholly absorbs the words. As Langer put it, Song is not
a compromise between poetry and music, though the text taken by itself be a great
poem; song is music.5 The second model proposes that there is an ir reducible rela-
tionship between words and music; as Lawrence Kramer has put it, A poem is
never really assimilated into a composition; it is incorporated, and it retains its own
life, its own body, within the body of music.6 The third model inter prets song as
a compound structure in which the words carry the primary semantic content and
the music colors or enhances what the words signify. Agawu observed that this
model is limited because it provides no independent account of the meaningfulness
of music.
In reviewing these three models, Agawu was rather too polite, resisting the
temptation to note that, within the tradition of song analysis, there have been rela-
tively few attempts to come to ter ms with how words and music combine to cre-
ate the phenomena called song. There are thorough and painstaking analyses of
song texts, with cursory mention of the music that sets the text; there are carefully
crafted analyses of musical structure, supported at tur ns by isolated instances of how
the words buttress or anticipate this structure; but there are few comprehensive
analyses that tr y to capture how the meaning constructed by the words and the
meaning constructed by the music come together to create the meaning associated
with a particular song.7 As an improvement, Agawu proposed a fourth model in
which song is construed as a confluence of three independent but overlapping sys-
tems: music and words and song. However, he almost immediately withheld full
endorsement of this model, noting that it provided no account of a concrete iden-

3. Both must also be situated within the more personal agendas of their authors. Mllers letter is
part of an attempt to get the composer to set more of his poems; Spauns eulogy is part of an attempt to
preserve what he viewed as the true legacy of his departed friend.
4. Victor Kofi Agawu, Theory and Practice in the Analysis of the Nineteenth-Century Lied, Music
Analysis 11 (1992): 3 36.
5. Susanne K. Langer, Feeling and Form: A Theory of Art Developed from Philosophy in a New Key (New
York: Charles Scr ibners Sons, 1953), 152.
6. Lawrence Kramer, Music and Poetry: The Nineteenth Century and After (Berkeley: University of Cal-
ifornia Press, 1984), 127.
7. One analysis that develops a comprehensive account of a song is David Lewins Auf dem Flusse:
Image and Background in a Schubert Song, in Schubert: Critical and Analytical Studies, ed.Walter Fr isch
(Lincoln: University of Nebraska Press, 1986), 126 52. However, Lewins main interest in this essay is in
developing an account of certain remarkable features of this particular song; his study, thus, did not set
out to delineate a broad methodological approach to the analysis of song.
word s, mu s i c, and s ong 245

tity for song but, instead, allowed song an identity only by default. This led him to
muse,Perhaps, then, what the model points to is song as process, not product.What
is interesting, in other words, is not what song is, but what it becomes in its per pet-
ual str iving for a concrete mode of existence.8
Nicholas Cook was able to steer a course between the Scylla of emphasizing the
text at the expense of the music and the Charybdis of emphasizing the music at the
expense of the text, as well as to give an account of a concrete identity for song, by
approaching song as an instance of multimedia: words were one medium, music
another.9 However, because Cook defined multimedia as the perceived interaction of
different media, he tended to regard songs as successful to the extent that the syntax
of their words and music remained separate indeed, in a state of contest. The
approach is effective for many songs, and the breadth offered by integrating song
analysis into the analysis of multimedia as a whole makes for a very comprehensive
perspective. Nonetheless, what is missing is a sense of how, despite their differences,
words and music work together in song how music can breathe the breath of life
into poetry or create a poem on a poem.
In this chapter, I explore further the nature of song and also the process of con-
ceptual blending. As noted in chapter 2, conceptual blending is a pervasive and
often transparent process in which elements from two cor related mental spaces
combine in a third.10 And, as suggested by the examples used in chapter 2, the men-
tal spaces basic to conceptual blends can be set up by music as well as by language.
In what follows, I expand on this idea, working on the assumption that music can
set up complete discourse structures, which can be cor related with the discourse
structure of a text to give rise to a unique domain for the imagination. What I am
concer ned with, then, are not isolated or very general cor relations between music
and text but a conceptual blend produced by an entire song.
In the first section that follows, I analyze two settings of Wilhelm Mllers poem
Trockne Blumen (Withered Flowers), taken from the cycle of poems he called
Die schne Mllerin (The Beautiful Miller Maid). The first is by the Berlin composer
Bernhard Klein, whom Mller addressed in the letter quoted above; the second is
by Franz Schubert. These analyses will allow me to show how different musical
environments interact with the same text to produce two quite different conceptual
blends, and thus two different songs. In the second section, I turn to Robert Schu-
manns Im Rhein (In the Rhein), from his song cycle Dichterliebe (Poets Love). In
Im Rhein, the music makes possible a conceptual blend whose emergent struc-
ture includes a nar rative considerably more extensive than that of the text alone. In
the third section, I consider two settings of a poem by Joseph von Eichendorff, In
der Fremde (In Foreign Lands), the first by Schumann, the second by Johannes

8. Agawu, Theory and Practice, 7 8.

9. Nicholas Cook, Analysing Musical Multimedia (Oxford: Clarendon Press, 1998); see esp. pp. 98 106
and Cooks analysis of Robert Schumanns Wenn ich in deine Augen seh from Dichterliebe on pp.
136 43.
10. A focus on the possibilities for meaning construction that ar ise in this blended space is one of
the fundamental differences between the idea of conceptual blending and both Cooks conception of
multimedia and Bakhtins notion of double-voiced discourse.
246 analysis and theory

Brahms. Again, two different songs result, but, in this case, the reasons lie with the
gener ic spaces that structure each conceptual blend.

trockne blumen
Wilhelm Mllers Trockne Blumen
Mllers Die schne Mllerin had its genesis in the diversions of a group of young
artists and intellectuals who gathered at the house of the privy councilor Fr iedrich
August von Stgemann dur ing the autumn and winter of 1816 17.11 There Mller
(1794 1827) together with the Stgemanns daughter Hedwig (1799 1891); the
artist Wilhelm Hensel (1794 1861); his sister, the poetess Luise Hensel (1798
1876); the publisher Fr iedrich Frster (1791 1868); and others put together an
informal drama structured around the story of a pretty millers maid and her suitors.
Each member of the group took a part, for which they wrote a verse scr ipt. In the
enactment of the resulting Liederspiel, the participants would recite their verses,
using whatever gestures or changes in vocal inflection the situation required in
order to tell their part of the story.
The evidence of the poems and anecdotes that survive from the Stgemann cir-
cle suggests that its Liederspiel had a thoroughly tragic ending. In what can be
reconstructed of the story they developed, the millers daughter, named Rose
(played by Hedwig von Stgemann), was courted not only by the young miller
(played by Mller) but also by a country squire (played by Frster), a hunter (played
by Wilhelm Hensel), and a gardener (played by Luise Hensel). The miller, after pre-
vailing br iefly over the other suitors, falls into despair when the hunter wins the
miller maids affections, and commits suicide. The maid, filled with remorse, then
joins him in death.
During the three years after the meetings of the Stgemann circle, Mller, appar-
ently intr igued by the possibilities of the story of the miller maid and her suitors,
went on to create his own version of the stor y, consisting of twenty-five poems
under the title Die schne Mllerin (Im Winter zum lesen), which was first published
in 1821.12 The cycle consists of a verse prologue (in the voice of the poet), twenty-

11. My account of Die schne Mllerin follows Susan Youenss excellent work on the or igins and his-
torical context of the cycle, which is reported in three of her publications:Youens, Behind the Scenes:
Die schne Mllerin before Schubert, 19th Century Music 15 (1991): 3 22; idem, Schubert: Die schne Ml-
lerin, Cambridge Music Handbooks (Cambr idge: Cambridge University Press, 1992); and idem, Schubert,
Mller, and Die schne Mllerin (Cambridge: Cambridge University Press, 1997). Another important
source is Graham Johnsons copious notes for his recording of the cycle. See Johnson, Die schne Ml-
lerin, The Poems and Their Backg round, in Die schne Mllerin, D795 (Hyperion, [CD] CDJ 33025,
Although the works I discuss here, both literary and musical, use trockne in their titles, the proper
word is trockene, which may be contracted to trockne. For the sake of consistency, I shall use
trockne throughout.
12. The cycle initially appeared as part of the fancifully titled Siebenundsiebzig Gedichte aus den hin-
terlassen Papieren eines reisenden Waldhornisten. The edition of Die schne Mllerin upon which I have relied
and that appears below is from Wilhelm Mller, Vermischte Schriften, ed. Gustav Schwab (Leipzig: F. A.
Brockhaus, 1830), 1: 1 56.
word s, mu s i c, and s ong 247

three poems proper to the drama, and a verse epilogue (again in the voice of the
poet), together intended to take about an hour to perfor m.13 In poems 1 5, we are
introduced to the miller, the millstream, the miller maid, her father, and the group
of young workers in his charge. In poems 6 11, we become aware of the millers
growing infatuation with the maiden and the depths of his unrequited love. In
poems 12 14, the miller becomes convinced that the miller maid is, as he says in
poem 12, Mine! Despite this belief, there is absent from the poetic nar rative any
explanation for what changed his perspective from one of dreamy mooning to a
self-assured sense of certain love; the poems provide no evidence that the miller
maid has actually returned his love. In poems 15 17, the hunter suddenly enters the
scene and, in the confused understanding of the miller, steals the miller maid away.
Just as suddenly, the bliss of the preceding poems is replaced by anger, jealousy, and
vindictiveness. In poems 18 21, the miller, torturing himself with visions of the
hunter and miller maid in fond embrace, gradually begins to dwell more and more
on death. It is in poem 21, Trockne Blumen, that the miller decides to commit
suicide, bringing to culmination the thoughts of death and the grave that came to
dominate the preceding poems. Poem 22 is a sad dialogue between the miller and
the stream, and poem 23 is given to the stream alone, as it gently and eer ily caresses
the drowned miller. The cycle concludes with an epilogue, which is cur iously
detached from the passion of the preceding poems, and which attempts to build a
bridge from Mllers intense world of fantasy back to the world of the everyday.
The basic stor y told in Trockne Blumen, the text and a translation of which
are given here, can be seen to proceed in three stages.14 In the first three stanzas, the
miller addresses a collection of withered, moribund flowers that are unaccountably
wet. In the second three stanzas, he contemplates the ineffectuality of his tears (the
tacit cause of the wetness of the flowers) and the insignificance of his existence in
the face of love that has died. This group ends with a stanza that echoes the first
stanza of the poem, both in its return to the image of the faded flowers and in the
recurrence, in reverse order, of the rhyming pair gab and Grab. Together, these
suggest a circular return that closes off the first six stanzas. The final stage of the
story commences with the last two stanzas of the poem. These speak of a future in
which the miller lies dead in his g rave, in which the miller maid recognizes the
truth of his love, and in which flowers bloom once more. The poem thus involves
sharp contrasts: dying flowers set against blossoming flowers, winter set against
spring, and dead love set against true love.

trockne blumen withered flowers

Ihr Blmlein alle, All you flowers
Die sie mir gab, That she gave to me,

13. Mller wr ites in the prologue, As its wintertime I expect you wont regret a br ief hour here
in the countryside (as it is depicted in the cycle of poems). [Erhoffe, weil es grad ist Winterzeit, / Thut
euch ein Stndlein hier im Grn nicht Leid.] Mller, Vermischte Schriften, 1: 1.
14. There are other ways to parse the poem, as we shall see later in this discussion. The analysis I
offer here is simply for the pur poses of drawing out some of the aspects of the poem important for a
number of its musical settings.
248 analysis and theory

Euch soll man legen You shall be laid

Mit mir ins Grab. With me in the grave.

Wie seht ihr alle Why do you all

Mich an so weh, Look at me so sadly,
Als ob ihr wtet, As if you knew
Wie mir gescheh? What happened to me?

Ihr Blmlein alle, All of you flowers,

Wie welk, wie bla? Why so faded, so pale?
Ihr Blmlein alle, All of you flowers,
Wovon so na? Why are you so moist?

Ach, Trnen machen Ah, tears do not make

Nicht maiengrn, The greenness of May,
Machen tote Liebe They do not make dead love
Nicht wieder blhn. Bloom anew.

Und Lenz wird kommen And spring will come

Und Winter wird gehn, And winter will go,
Und Blmlein werden And flowers will
Im Grase stehn. Grow in the grass.

Und Blmlein liegen And flowers lie

In meinem Grab, In my grave,
Die Blmlein alle, All the flowers
Die sie mir gab. That she gave to me.

Und wenn sie wandelt And when she walks

Am Hgel vorbei, Past the mound,
Und denkt im Herzen: And ponders in her heart:
Der meint es treu! His love was true!

Dann Blmlein alle, Then all you flowers,

Heraus, heraus! Come out, come out!
Der Mai ist kommen, May is here,
Der Winter ist aus. Winter is over.

One thing immediately evident is that this poem relies, as do most, on concep-
tual blends.When the miller addresses the flowers, for instance, it is not as plants but
as anthropomorphized beings, creatures that can not only look upon the miller but
do so with pathos: Wie seht ihr alle / Mich an so weh, / Als ob ihr wtet, / Wie
mir gescheh? This, of course, is another example of the anthropomorphic blends
discussed in chapter 2. Another blend is set up in the fourth stanza, where Mller
combines concepts related to the seasons and to romantic love. Thus maiengrn
becomes both a harbinger of spring and a sign of loves first bloom.Yet a third blend
crops up in the seventh and eighth stanzas, when concepts related to the seasons
combine with those of life and death: the consummatory Der Mai ist kommen /
Der Winter ist aus represents both the future tur n of seasons the miller envisions
word s, mu s i c, and s ong 249

and his rebirth as the eter nal lover of the miller maid. The poem is thus a r ich dis-
course structure in its own r ight, one with intr iguing possibilities for pair ing with
the discourse structures of music.

Bernhard Kleins Trockne Blumen

The Berlin composer Ber nhard Klein (1793 1832) apparently discovered Mllers
poetic cycle shortly after it was published and set Trockne Blumen as the last of
his five Lieder und Gesnge mit Begleitung des Piano Forte. This collection was brought
out by E. H. G. Christiani in 1822,15 and the inclusion of Trockne Blumen is all
the more remarkable in that the texts for the first four songs involve rather con-
ventional stor ies of wanderlust and relatively innocent romantic encounters.
Trockne Blumen, as Klein clearly recognized, is a text of quite another order.
Kleins setting, the score for which is given in example 6.1, is guided by his divi-
sion of the text into four strophes, each compr ising two stanzas of the poem.
Although the music for each strophe is different, so much mater ial is reused that the
song seems more strophic than through-composed, as the for mal outline g iven in
table 6.1 indicates.
The song begins with a br ief piano introduction that sets out the mater ial used
to begin each of the first three strophes. This mater ial, labeled A1 in table 6.1, con-
sists of three iterations of a motion from dominant to tonic, culminating in a passage
in parallel sixths that returns to the dominant. The whole is couched in a spare two-
and three-voice texture, with a repetitive melody built up from a compact succes-
sion of ascending and descending gestures that circulate around C5 and A4.
Although A as tonic is scarcely in doubt, it falls short of attaining a full measure of
stability, given the absence of root-position tonic tr iads in strong rhythmic positions
and the emphasis on the dominant har mony at both the beginning and end of each
A1 section. Sections B1 and B2 consist entirely in prolongations of the dominant,
either by temporar ily transfor ming it into a leading-tone har mony (as in mm.
11 12 and mm. 20 24) or through a pathos-drenched passage that introduces an
unprepared VII # of IV that leads through IV back to VII, and then to V (as in
mm. 13 16 and mm. 25 28). In each of the B sections, the melody struggles
upward to D5 three times, only to fall back either to B4 or CS5. Of these two sec-
tions, B2 is slightly more intense. In it, a repeated Fn4 replaces the repeated E4 that
signaled the opening of the B1 section, giving rise to a VIIo that lasts from m. 20 to
m. 24.
In the third strophe, Klein lightens the mood of the music by shifting to the par-

15. Mller is the author of all of the poems of Kleins Lieder und Gesnge of 1822; Trockne Blumen
is the only poem Klein selected from Die schne Mllerin. It is unclear what motivated Kleins choice of
poet. In his letter of 15 December quoted above, Mller praised Kleins settings and asked if he would be
interested in setting more of his poetry. Given the tone of the letter, it would seem unlikely that Mller
was an active participant in the selection of texts for the 1822 collection of songs. I should note, as well,
that another Berlin composer, Ludwig Berger, published a setting of Trockne Blumen in 1818; analy-
sis of this work, as well as the context for its publication, can by found in Lawrence Zbikowski, The
Blossoms of Trockne Blumen: Music and Text in the Early Nineteenth Centur y, Music Analysis 18
(1999): 307 45.
250 analysis and theory

example 6.1 Bernhard Klein, Trockne Blumen

Andantino A1

Ihr Blm lein


5 B1

al le die sie mir gab, euch soll man le gen mit mir ins Grab. Wie seht ihr


al le mich an so weh als wenn ihr w tet wie mir ge

15 A1

scheh? Ihr Blm lein al le wie welk, wie bla? Ihr Blm lein

allel major. The repeated notes that lent an air of intensity or obsession to the dom-
inant prolongations of the preceding B sections now reappear to keep the music
from stalling on the first and second beats of each bar of section A2. Section B3
begins with another prolongation of the dominant, but instead of being trans-
formed into a leading-tone har mony in m. 36, as it had at the analogous juncture in
sections B1 and B2, it now remains a dominant, leading initially to a I in m. 38, and
then to the first cadence on A thus far. With the ar rival on the I of m. 38, the
melody finally reaches E5, preparatory to its ar rival on A4 in m. 40.
example 6.1 (continued)

19 B2

al le wo von so na? Ach, Tr nen ma chen nicht


mai en grn, ma chen to te Lie be nicht wie der blhn!

A2 Poco pi lento

Der Lenz wird kom men, und Win ter wird gehn und Blm lein

32 B3

wer den im Gra se stehn, und Blm lein lie gen in mei nen

37 C

Grab die Blm lein al le die sie mir gab. Und wenn sie

example 6.1 (continued)


wan delt am H gel vor bei, und denkt im Her zen

47 A3

der meint es treu! dann Blm lein al le her aus her aus! der


Mai ist kom men der Win ter is aus der


Mai ist kom men der Win ter ist aus.

word s, mu s i c, and s ong 253

Table 6.1. Formal Outline of Bernard Kleins Trockne Blumen

Section and Key Text Stanza Musical Material and Melody Measures

Introduction A minor A1 , a 04
Strophe 1 Stanza 1 A1, a 48
Stanza 2 B1, b 816
Strophe 2 Stanza 3 A1, a 1620
Stanza 4 B2, b 20 28
Strophe 3 A major Stanza 5 A1, a 2933
Stanza 6 B3, b' 3340
Strophe 4 Stanza 7 C, c 40 48
Stanza 8 A2, a' 4858
Postlude A2, a' (fragment) 58 64

The C section beg ins another prolongation of the dominant, one whose
repeated notes are strongly reminiscent of sections B1 and B2. Absent, however, is
the relatively active bass line that opened each of the latter sections; it is replaced
here by a pedal on E3. The melody, too, has changed, now peacefully circulating
around B4, with unimpeded ascents to E5. After m. 45, the intensity of the section
increases with the cessation of the repeated notes and the introduction of the chro-
matic second alto line of mm. 46 47. In m. 48, however, the melodic line that
descends from E5 once again comes to rest on a clear dominant har mony, which
leads immediately to the reprise of the A mater ial.
Although section A3 opens with the same figure in parallel tenths that began the
song (adjusted to fit the major mode), the figure now leads toward a dominant-
seventh in m. 50 and then to a relatively extensive prolongation of the tonic in
mm. 51 55. This prolongation prepares the way for the cadence of mm. 56 58
with which the vocal part concludes. A brief piano postlude follows; it is based on
the opening of the A3 section, but it includes a motion from Fn to E that recalls the
minor mode of the first half of the song.
Thus, Kleins music falls into two distinct parts. The first part, through m. 28, sets
out an intense and unstable A minor. Relatively lengthy and obvious prolongations
of large-scale har monic dissonances (such as those associated with the dominant
and leading-tone) and sudden inter ruptions mark the texture, as do persistent
repeated notes. The second part, from m. 29 onward, disperses the intensity of A
minor and replaces it with a goal-or iented and (for the most part) orderly A major.
Here there are also prolongations of the dominant, but of a sort that seem to affir m
rather than challenge the viability of tonic.What had been difficult in the first part
becomes easy in the second; what had been troubled becomes calm.
This music, together with Mllers text, sets up the conceptual integration net-
work (CIN) diagrammed in figure 6.1. Two observations should be made about the
network represented by this diag ram, observations that extend to all of the CINs
discussed in this chapter. First, as was the case in chapter 2, the diagram represents
a kind of analytical snapshot of the elements and relations that make up the net-
work, and that contr ibute to the emergent structure of the conceptual blend. A
254 analysis and theory

F* B*.#(,5#-.2$

K0LC#1)+E0# M51*E#1)+E0#
F* !"#$#%&N$/&-.2$:+*>"(5?$ F* +*,-%."N$/-3-.-5%"/A
#",(5?$%*'"+"55."55$ 5"9".#%5?$OM$B"-+-.2?$
F* '(&("#)*5'(-.2?$ +,B@$*:$B,/".B"$
)+**3-.2$:+*>"(5?$%*'"$ F* +*,/0."N$B,/".B"5?$
:*($#%"$:1#1("$ PM?$<QM?$+-3-#"/$

F* *)5"55-9"?$3"(B1(-,+$
F* *(/"(+&?$B,+3$:1#1("$

figure 6.1 CIN for Kleins Trockne Blumen

fuller account of the way concepts combine within this network would require a
series of such snapshots. Second, what I am concer ned with here is the overall dis-
course set up by the text and the overall discourse set up by the music. Although
there are interesting details at more local levels, I am most intr igued by what con-
ceptual blending can tell us about song and by what song can tell us about concep-
tual blending. Let me expand on this for just a moment before returning to Kleins
Trockne Blumen.
The theory of conceptual blending assumes that there are structural invariances
between the input spaces of a blend: these invariances, encapsulated in the elements
and relations of the gener ic space, are what make conceptual blends possible. In the
case of songs, the invariances are between the mental space set up by the text and
the mental space set up by the music. Put another way, the fact that combinations of
text and music can give rise to conceptual blends suggests that there are syntactic
correspondences between linguistic and musical discourse. Although the syntax of
most nineteenth-century German poetry is usually clear at or below the level of the
stanza, the syntax of songs from the same milieu needs to be conceived in ter ms of
the song as a whole, which was construed as a single, if often extended, musical
utterance. Because of this, I find it most useful to focus on conceptual blends that
encompass the entirety of a given song that is, blends that ar ise from the overall
discourse set up by the text and the overall discourse set up by the music.16

16. This perspective also leads me to regard the poem set by the composer as a text, rather than a
poem. Susan Youens, for different reasons, advocates a similar approach; see Youens, Retracing a Winters
word s, mu s i c, and s ong 255

What nineteenth-centur y songs can tell us about conceptual blending stems

from the character istic of musical syntax I have just descr ibed. The blend set up by
a song (assuming that there are structural cor relates between the mental spaces set
up by the text and the music) spans at least two or three minutes, involves two dif-
ferent media (with two different, if related, kinds of syntax), and engages us on both
emotional and intellectual levels. All these features have the potential to add immea-
surably to our understanding of conceptual blending as a general cognitive
With these thoughts in mind, let us return to the CIN set up by Kleins song.
The gener ic space for the CIN is structured around the notion of two linked but
contrasting ontological states. Such states are often temporally indexed (as past and
present or as present and future), but they can be more abstractly represented as
theme and variation, or prototype and copy. The delineation of two ontological
states often provides the basis for a more extensive interpretive framework. This
framework then fur nishes an account of the cause for the change of state or an
explanation for why one state is to be preferred over another.
Contrast between the two ontological states is supported in Mllers poem by an
emphasis on the present in the first four stanzas and an emphasis on the future in
the second four stanzas. There is also far less activity descr ibed in the first four stan-
zas than in the second four. Of course, there are exceptions to both of these gen-
eralizations. In the first half of the poem, the future is summoned at the end of the
first stanza and the past is alluded to in the fourth; the gaze of the flowers and the
tears of the miller suggest that the scene is not wholly devoid of activity. In the sec-
ond half of the poem, the past is evoked briefly in the last two lines of the sixth
stanza, the first two lines of which call up the embodiment of lifelessness.
In the mental space of the music, the linked but contrasting ontological states of
the gener ic space are set up by the A minor and A major music of the two halves
of the song. These states are linked less by temporal index than by conceptual frame:
the A minor music functions as the theme, as it were, and the A major music as the
variation. Nonetheless, the A major music rewrites the character and syntax of the
A minor music. It exchanges the third and sixth scale steps of minor for those of
major, provides the first cadences on tonic, avoids the fully diminished seventh
chords omnipresent in the first half of the song, and allows the melody to ascend
to E5 and FS5. As a result, it enacts a specific shift of time frame by placing the A
minor music fir mly in the past and announcing the A major music as the new pres-
ent. This shift is underlined by the mir rored processes that occupy each half of the
song: over the course of the first half of the song, the music becomes more intense

Journey: Schuberts Winterreise (Ithaca, N.Y.: Cornell University Press, 1991), 307 12. As will also become
clear, I am typically concer ned with the story told by the text as a whole, a parabolic level discussed by
Mark Turner in The Literary Mind (New York: Oxford University Press, 1996), esp. chap. 1.
17. Although it should be clear from the foregoing, let me emphasize that not all songs will set up
conceptual blends. Among those songs that do set up blends, some blends will be more successful (that
is, lead to more extended elaborations) than others. One example of the latter situation is provided by
the third phrase of Brahmss In der Fremde; see the discussion of the song later in this chapter.
256 analysis and theory

and darker; over the course of the second half, the music becomes less intense (with
the exception of the poignant recollection of mm. 46 47) and br ighter.
Correlations between these idealized representations of contrasting ontological
states provide an opportunity for the combination of concepts in the blended space,
which draws upon elements of the input spaces to create a portrayal of the millers
transfor mation from a tortured soul into a hopeful and ecstatic one. As was the case
in chapter 2, the emergent structure of conceptual blends (including the present
one) can be descr ibed in ter ms of the processes of composition, completion, and

composition The combining of elements from the text-space and the

music-space is quite straightforward. The enervated present of the first four stanzas
and the A minor music combine to create a representation of one ontological state
in the blended space: a stark image of the present as dominated by obsessive
thoughts, scant opportunity for rest, and a mercurial play of emotions. The active
future of the second four stanzas and the A major music combine to create a rep-
resentation of the second ontological state: what was shadowy has become clear;
there is an orderly progress toward goals; and the sur roundings are, almost without
exception, calm and soothing. The transfor mation of one state into the other occa-
sions a strange conceptual inversion in this part of the song: the flowers that blos-
som after the change to A major in m. 29 seem to be the same ones that the miller
predicts will lie in his grave (cf. mm. 31 33 and mm. 34 40), even though the lat-
ter event, according to the logic of the poem, must precede the for mer.

completion Building on the context provided by this blend of concepts, we

can imagine outward signs of the millers torment in the first half of the song a
sluggish, depressed body language, accompanied by sudden expressions of anguish
as his agony increases and how, in the second half of the song, these give way to
gestures and behavior associated with peacefulness and happiness as the miller con-
templates his future.We can also imagine that, overall, the miller grows happier and
more at peace with himself as he becomes more distant from his earlier tor ment.
This inference is supported in part by the music of the concluding A3 section,
where the melody breaks through to the high FS5 of m. 49, a moment followed by
the long prolongation of tonic in mm. 51 55 that prepares the singers final

e laboration Elaboration of the blend suggests that the miller will act in
order to make the future he imagines a reality. His Heraus, heraus! thus calls forth
not only the futures flowers but also the future itself. Elaboration may also provide
an explanation for how the state of affairs portrayed in the first half of the song is
transfor med into the state of affairs portrayed in the second half . Here the music
provides a useful guide, for it suggests that the present can be transfor med into the
future by reinterpreting it, just as the musical mater ials from A minor were reinter-
preted in A major. Thus, the dead love of the present will become the eter nal love
of the future; the dying blossoms will become living ones. The strange celebration
that is enacted in mm. 34 40, as the miller contemplates the flowers that will be laid
word s, mu s i c, and s ong 257

in his grave, then becomes more comprehensible, for in the second half of the song
the withered flowers that he holds are transfor med into symbols of his eter nal love.

Franz Schuberts Trockne Blumen

Although the exact circumstances of Franz Schuberts encounter with Mllers Die
schne Mllerin are not known, it appears that Schubert came upon the poetic cycle
sometime in the first part of 1823 and set twenty of the twenty-five poems dur ing
the second half of the same year. The completed cycle was published the following
year by the fir m of Sauer and Leidesdorf.Trockne Blumen, the score for which is
given in example 6.2, is the eighteenth song in the cycle. Schuberts setting divides
the eight stanzas of the poem into a 3 + 3 + 2 grouping. The music in E minor for
the first three stanzas is repeated for the second three, after which there is a shift to
the parallel major for the final two stanzas. Schubert does not repeat any text in his
setting of stanzas 1 6, but he does repeat the last two stanzas in the patter n 7 8,
7 8, 8.
The song begins with the most minimal of piano introductions: two bars of a
repeated E minor chord, played piano, each eighth-note chord followed by an eighth
rest. This patter n continues as the voice enters with a bare melody outlining E
minor. With the shift of the bass from E3 to D3 in m. 5, the E minor chord is trans-
formed into a G chord in position. The latter leads to a cadence on G on the sec-
ond beat of m. 6, coincident with the conclusion of the first stanza of text. The
music of mm. 3 6 is then repeated for the second stanza in mm. 7 10, sustaining
the spare effects created by the minimal melodic and har monic mater ials, but
adding the barest hint of forward motion when the voice closes on B4 instead of on
G4, as it had in m. 6.
In m. 11, the music becomes more active and its har monic mater ial somewhat
more variegated: the rhythmic patter n of the accompaniment is livelier; the har-
monies change every beat; and the melody, as it follows these har monic changes,
becomes more animated. The initial har monic strategy for the setting of the third
stanza is, in compressed for m, similar to that used for the first two stanzas: E minor,
now preceded by its dominant in m. 11, leads again to G in m. 12. Measure 11 is
repeated in m. 13, but this time it is followed by an Italian augmented-sixth chord
that leads to the dominant of E minor, while the singer reaches past E5 for the first
time to grasp a high FS5. The piano then echoes this gesture, sustaining the domi-
nant chord for a half cadence on B that lasts through m. 16. Schuberts setting of
stanzas 4 6 in mm. 17 29 is virtually identical to the music he provided for the first
three stanzas; the only significant change occurs in mm. 28 29, where the E5-DS5-
FS5 of the melody is restated as FS5-E5-DS5.
With the shift to the parallel major in m. 30, the accompaniment becomes yet
more active: Schubert adds a lightly tr ipping bass melody to the animated rhyth-
mic patter n of the right hand. In m. 31, the singer takes up the bass melody, in
modified for m.With the momentar y shift toward CS minor in mm. 33 34, how-
ever, the melody becomes streamlined once again. The CS minor chord in m. 34
yields in the following measure to another augmented-sixth chord. This leads (in
m. 37) to a cadential , which prepares the first cadence on E in the song. Schubert
example 6.2 Franz Schubert, Trockne Blumen, from Die schne Mllerin, D. 795/18

Ziemlich langsam

Ihr Blm lein al le, die sie mir gab, euch


soll man le gen mit mir ins Grab. Wie seht ihr al le mich an so weh, als

ob ihr w
tet, wie mir ge scheh? Ihr Blm lein al le, wie welk, wie bla? ihr


Blm lein al le, wo von so na? Ach Tr nen ma chen nicht


mai en grn, ma chen to te Lie be nicht wie der blhn, und Lenz wird kom men, und

example 6.2 (continued)


Win ter wird gehn, und Blm lein wer den im Gra se stehn, und Blm lein lie gen in


mei nem Grab, die Blm lein al le, die sie mir gab!


Und wenn sie wan delt am H gel vor bei und


denkt im Her zen, der meint es treu! dann Blm lein al le, her aus, her aus, der


Mai ist kom men, der Win ter ist aus. Und wenn sie wan delt am


example 6.2 (continued)

H gel vor bei und denkt im Her zen, der meint es treu! dann


Blm lein al le, her aus, her aus, der Mai ist kom men, der


Win ter ist aus, dann Blm lein al le, her aus, her aus, der


Mai ist kom men, der Win ter ist aus.


word s, mu s i c, and s ong 261

immediately repeats the whole of mm. 30 38 in mm. 39 47 and then, without

pause, repeats the last four bars, with a change of bass and a slight thickening of tex-
ture. A six-bar piano postlude, reprising the accompanimental figure of m. 30, con-
cludes the song. In the course of the postlude, the register drops an octave, and the
mode shifts to minor for the last four bars.
Although this descr iption cannot do it justice, the effect of the relatively spare
musical mater ials that Schubert employs in this setting is profound. A somewhat
more nuanced view of the overall musical discourse can be gained through a mid-
dleground sketch of the voice-leading of mm. 1 16 and 30 38, given in example
6.3. As the sketch shows, in the first portion of the song the repeated motions
toward G major simply prolong E minor and the initial B4 of the Anstieg. It is only
in m. 14 that the music finally begins to move, wrenched from its enervated reiter-
ation of tonic (or the tonic-substitute mediant) by the augmented-sixth chord that
leads to the dominant. The augmented-sixth chord also provides the impetus for
the continuation of the Anstieg, which is temporar ily left hanging on the FS5 over
the dominant. Given the long delay of the continuation of the Anstieg, almost the
entirety of mm. 1 29 remains stalled on a prolongation of tonic and B4. Only at the
half cadences of mm. 15 16 and mm. 28 29 does the music move away from B4
and momentar ily struggle upward toward the Kopfton. With the shift to the paral-
lel major in m. 30 the oppressive harmonic stasis of the first half of the song lifts, and
the music begins to move forward, pulled along by the extended dominant prepa-
ration of mm. 33 37. Finally, in m. 37, the voice achieves the Kopfton, but as GS5
rather than Gn5. This accomplishment is followed immediately by the descent of
the fundamental line, supported by the first cadence on tonic of the entire song.
Schuberts music falls into two distinct parts in a way that parallels Kleins set-
ting of Mllers poem. The first part, up to m. 29, presents a minimal, washed-out

example 6.3 Middleground sketch of mm. 1 16 and 30 38 of Schuberts Trockne


3 7 11

30 35

8 7
10 10 6 5
5 6 4 3
262 analysis and theory

F* B*.#(,5#-.2$

K0LC#1)+E0# M51*E#1)+E0#
F* !"#$#%&N$/&-.2$:+*>"(5?$ F* 5',("?$5#,#-B$P$3-.*(?$
#",(5?$-."::"B#1,+-#&?$ +1(B%$#*$%,+:$B,/".B"?$
-."9-#,)-+-#&$*:$5",5*.5$ ,((-9,+$*.$<QM$
F* '(&("#)*='(-.2?$ F* :+*>-.2?$/-("B#"/$P$
)+**3-.2$:+*>"(5?$ 3,D*(?$B*3'+"#-*.$*:$
#(,.5B"./,.B"$#%(*12%$ :1./,3".#,+$+-."$


figure 6.2 CIN for Schuberts Trockne Blumen

E minor inter rupted twice by rather overwrought cadences on the dominant. The
second part, from m. 30 on, moves forward, and without so much as a backward
look culminates in author itative and emphatic cadences on E major.
This music, together with Mllers text, sets up the CIN diag rammed in figure
6.2. The gener ic space, like that for Kleins Trockne Blumen, is structured around
the notion of two linked but contrasting ontological states. However, where in
Kleins setting the emphasis is on connections between the ontological states, in
Schuberts setting the emphasis is on their incommensurability. Kleins approach
takes the point of view of an observer: the undesirable situation represented by the
first ontological state can be improved by reconstructing it in ter ms of the second.
Schuberts approach takes the point of view of a participant: recognizing that the
first ontological state offers no hope, it is simply abandoned in favor of the second.
The change of emphasis in the gener ic space reflects a reinterpretation of the
elements of the textual space, prompted by Schuberts division of the text into a
group of six stanzas followed by a group of two stanzas (which are subsequently
repeated). While the ontological states can still be character ized as present and
future, the index is now emotional rather than temporal. The present is still occu-
pied with dying flowers, thoughts of the grave, tears, and dead love, but it is also suf-
fused with a sort of numbness, born of the millers inability to br ing about any
change in his circumstances. He barely realizes that the tears that moisten the dying
flowers are his own; he is powerless to stop them, just as he is powerless to stop the
word s, mu s i c, and s ong 263

coming of the seasons, or of his own death. Thus the future of Und Lenz wird
kommen / Und Winter wird gehn is contained within the present, a leaden emo-
tional domain circumscr ibed by the echoing rhymes of the first and sixth stanzas.
By contrast, the future state called forth by the last two stanzas of the poem is
devoid of such dark thoughts: here the miller maid wanders past the resting place of
the miller; here spring has truly arrived, and the winter of despair has been ban-
Contrast between ontological states is also apparent in the musical space. Schu-
berts E minor music is, for the most part, spare and static. After m. 11, it gradually
pulls itself into an agitated state, lurches into a half cadence, and then returns to its
barren beginning. His E major music, in contrast, exploits and adds to the quick-
ening of pulse that precedes the half cadences of the E minor music, and it dr ives
toward a consummatory cadence that effaces the equivocation and ir resolve of the
first half of the song. This is music of possibility and completion, of action.
Correlations between these idealized representations of contrasting ontological
states provide an opportunity for the combination of concepts in the blended space,
which draws on elements of the input spaces to create a portrayal of two very dif-
ferent personae for the miller.

composition Composition provides the concepts basic to this portrayal.

The E minor music and the first six stanzas of text combine to summon a miller
who sits in numb contemplation of what seems to be a hopeless predicament.When
the gravity of his situation seizes him, he stirs for a moment (responding to the
augmented-sixth chords), but even the intensity of torment is denied him and he
sinks back into his tor por. The E major music and the last two stanzas portray a
quite different miller who, with growing confidence, reaches out toward goals
most particularly, achieving immortality through true love beyond the grasp of
the millers first persona.

completion Through completion we can build on what we know about

human behavior in such extreme states: we can imagine that in his first persona the
miller is virtually inert until, at the moment of his anguish, he half rises, a look of
panic on his face; similarly, we can imagine the expansive gestures of the miller in
his second persona and perhaps also see the somewhat fevered look in his eyes as
he proclaims Der Mai ist kommen / Der Winter ist aus.

elaboration While the emergent structure produced by composition and

completion offers a wealth of material to the imag ination, it is elaboration that
offers the most str iking inter pretive possibilities. Elaboration of the blend suggests
that the transition from the first persona to the second marks a decisive turn in the
millers story.With this transition, the miller has moved beyond merely entertaining
the thought of death and has now accepted the fact of it. The miller will overcome
his ineffectuality, he will transfor m himself, and he will do so through suicide. Sup-
porting this inter pretation are the almost frenetic repetitions of the second half of
the song (as though the miller still needs to convince himself of the necessity of his
264 analysis and theory

course of action) and the sad tur ning toward minor of the postlude as the enor mity
of his decision begins to sink in.18

Two Settings, Two Millers

These two settings of Mllers Trockne Blumen give us two rather different
millers.We first come upon Kleins miller tur ning over the shards of his shattered
love affair and becoming more distraught as he realizes the hopelessness of his
predicament. Then there is a sudden shift, and what was dark and foreboding
becomes light and promising. The millers burden has been lifted, for he has found
a way to breathe life into dead flowers and dead love: he will welcome death rather
than fear it. In contrast, Schuberts miller has two str ikingly different personae. The
first is emptied out, nearly paralyzed by despair, numbly aware of his misery and his
inability to escape it. The second is effusive, expansive in welcoming a future in
which the miller is no more. Schuberts miller is, in short, psychotic: at first immo-
bile in the face of death and then feverishly and joyously embracing it.
Are both millers present in Mllers poem? Yes and no.Yes, in that his poem can
support these different inter pretations, as the CINs above have shown. And yes, in
that, from the perspective of the poetic cycle as a whole, the miller is a truly com-
plex character, a quiet psychopath destined for destruction. But no, in that it
requires the agency of the music to br ing forth these specific millers.
The differences between these two millers suggests that a review of the differ-
ent compositional strateg ies that brought them into being is warranted so that we
may better understand the role of musical syntax in breathing them to life. The sim-
ilarities between these two settings of Mllers poem are str iking: both employ a
shift from minor to major mode; both avoid any definitive cadence on the tonic in
the first portion of the song; and both make use of relatively delicate textures, espe-
cially in the first half of the song.19 More distinctive are Kleins and Schuberts
respective treatments of dominant and tonic. Klein, through extended prolonga-
tions, employs the dominant as a tension-generating device, and in consequence the
dominant receives more emphasis than the tonic. By contrast, Schubert is spar ing
with the dominant of his home key, reserving it as a device to pull the music tem-
porarily out of its stasis (in the first half of the song) or using it as part of a nor ma-
tive cadence (in the second half of the song). Tonic, by compar ison, is given quite
a bit of emphasis, so much so that the first portion of the song becomes threatened
with immobility.
The conceptual blends created by Kleins and Schuberts songs, then, are not sim-

18. Further evidence for the moment of this decision is provided by the poetic cycle as a whole. In
all of the poems up to Trockne Blumen, the millers presence is marked by the use of the first person.
After the shift to the third person in Trockne Blumen, the miller never appears in the first person
19. In addition to the similar ities between Kleins and Schuberts settings of Mllers poem, both also
share features with Ludwig Bergers 1818 setting of the poem, which was mentioned in n. 15; the com-
mon features are discussed in Zbikowski, The Blossoms of Trockne Blumen, 335. Although such
similar ities are provocative, no research of which I am aware traces specific patter ns of influence between
Berger, Klein, and Schubert.
word s, mu s i c, and s ong 265

ply a matter of words and music happening at the same time (and thus being com-
bined by dint of sharing the same temporal frame). For blending to take place, the
mental spaces set up by the text and the music must have a shared topography,
reflected in part in commonalities between the syntactic structures of the two
media. This shared topography does not exhaust the resources of the input spaces.
Indeed, the difference in syntax important to Nicholas Cooks conception of mul-
timedia (discussed at the beginning of this chapter) is no less important in concep-
tual blending, for it ensures that the stor ies told by both music and text are com-
pelling in themselves. Nonetheless, common structural features (which may obtain
at a different level of structure than those that lead to contest between media) are
necessary for conceptual blending to take place.
In the settings of Trockne Blumen, words and music work together to tell sto-
ries that enhance and expand on structural features inherent in the source spaces for
each conceptual blend. I now tur n to a song by Robert Schumann in which the
mental space set up by the music makes possible a conceptual blend whose emer-
gent structure includes a nar rative considerably more extensive than that suggested
by the text alone.

im rhein
In 1840, having established himself as a composer of works for the piano, Robert
Schumann tur ned his energies to song. And for midable energies they were, for by
the time the year had ended, he had wr itten over 140 songs. Schumann quite obvi-
ously worked at speed: the twenty songs in the or iginal version of his Dichterliebe
cycle (four were cut before its publication) were written from 24 May to 1 June.
Nonetheless, among the songs wr itten that year are some of the most profound and
haunting of the nineteenth century, in part because Schumann found ways to com-
bine words and music to create fantastic worlds that extend far beyond those that
can be summoned by either words or music alone.

Heines Im Rhein
Im Rhein, the text and translation of which are given here, is the sixth of six-
teen songs that make up Schumanns Dichterliebe, first published by C. F. Peters in
1844.20 The text for all of the songs was taken from Heinr ich Heines Lyriches Inter-
mezzo, a collection of sixty-five poems published in 1823. Although usually ter med
a song cycle, Dichterliebe is held together not so much by common nar rative as by
theme: all of the poems deal with romantic love, which is transfor med, splintered,
or distorted by reminiscence or overheated emotions.

im rhein in the rhine

Im Rhein, im heiligen Strome, In the Rhine, in the holy r iver,
Da spiegelt sich in den Welln, There, reflected in its waves,

20. The text for Im Rhein given here is from the song as it appears in Clara Schumanns edition
of the Lieder of Robert Schumann, but with text repetitions omitted.
266 analysis and theory

Mit seinem groen Dome With its great cathedral,

Das groe, heilige Kln. Is the great holy city, Cologne.

Im Dom, da steht ein Bildnis, In the cathedral there hangs a portrait

Auf goldenem Leder gemalt; Painted on gilded leather;
In meines Lebens Wildnis Into my lifes wilder ness
Hats freundlich hineingestrahlt. It has shone down kindly.

Es schweben Blumen und Englein Flowers and cherubs float

Um Unsere Liebe Frau, Around Our Dear Lady.
Die Augen, die Lippen, die Wnglein, Her eyes, her lips, her cheeks
Die gleichen der Liebsten genau. Are exactly like those of my love.

The three stanzas of Im Rhein can be divided into two parts. The first part,
which compr ises the first stanza, summons a broad view of Cologne centered on its
cathedral and reflected in the waves of the Rhein. There is a sense of vastness to the
perspective (suggested by the repetitions of gro ), and of the holy (provided by the
cathedral and the descr iption of Cologne as the holy city).21 The second part of
the poem, which compr ises the second and third stanzas, moves inside the cathedral
to focus on a portrait within. It is at this point that the presence of a nar rator is
made explicit and the shifting perspective of the poem acquires an anchor and a
rationale.We are here in this city, inside this cathedral, to view a painting of singu-
lar importance to the nar rator, for its radiance has proved a balm to the troubles of
his soul. In the third stanza, the poetic lines become longer, and the nar rator takes
us inside the painting itself . He begins by describing a heavenly vision of the
Madonna, but as he takes us deeper into the painting, the image of the Madonna
becomes ever more worldly, until it becomes apparent that the nar rator savors this
image because it is the exact likeness of his beloved. The poem ends with this sud-
den shift from the spir itual to the car nal, and we hang in mid-air, suspended both by
and within the nar rators vision.
The elegance and power of Heines unsettling poem stem from two parallel
processes. The first is perspectival: we start from the largest scale the city of
Cologne and its cathedral and then nar row the scope, first to the inter ior of the
cathedral, then to a painting within, and then to specific features of the face por-
trayed in the painting. The second process is conceptual and affectual: we begin
with ideas about vastness and holiness; proceed to the active solace provided by the
contemplation of a religious work of art; and then are suddenly confronted with an
intense, carnal passion for a lover.

Schumanns Im Rhein
Schumanns setting of the poem Im Rhein, the score for which is given in exam-
ple 6.4, is in ABA' form. The first sixteen measures, which make up the A section,
present a forceful, leaden E minor. Beginning forte, the voice (accompanied by the

21. Schumann provided additional emphasis on the holy by changing the end of Heines first line
from im schnen Strome to im heiligen Strome.
example 6.4 Robert Schumann, Im Rhein, from Dichterliebe, Op. 48

Ziemlich langsam


Im Rhein, im hei li gen Stro me, da spie gelt


sich in den Welln, mit sei nem gros sen Do


me, das gros se, hei li ge Kln. Im Dom, da


steht ein Bild niss, auf gol de nem Le der ge malt; in


mei nes Le bens Wild niss hats freund lich hin ein ge strahlt.

example 6.4 (continued)

Es schwe ben Blu men und


Eng lein um un sre lie be Frau; die Au gen, die Lip pen, die

38 ritard.

Lip pen, die Wng lein, die glei chen der Lieb sten ge nau.




word s, mu s i c, and s ong 269

left hand of the piano in octaves) makes a stepwise ascent from E4 to G4 in m. 4,

answered by a stepwise descent from C5 back to E4. At this point, voice and piano
part ways.While the voice repeats its melodic mater ial almost exactly (save for the
crucial change of the penultimate FS4 for Fn4), the piano continues the descent ini-
tiated in m. 4, sinking to the G1-G2 octave of mm. 11 13. In the second half of
m. 13, the bass echoes the voices descent from C to G of mm. 12 13 but, prompted
by the important Fn4 in the voice, it arrives on a G1S-G2S octave instead. This
arrival coincides with the final E4 in the voice part and with the completion of the
first stanza of the poem. The right hand of the piano fills out, if minimally, the bare
texture created by the voice and the left hand; it does so with an accompaniment
made up of cascading gestures in a halting patter n of alternating eighth- and dotted-
The dominant of mm. 15 16, with which the A section concludes, leads to the
A minor chord and piano of m. 17 that starts the B section. Here, the str iding bass
of the left hand disappears, replaced by a more fluid and g ravitationally liberated
adaptation of the eighth/dotted-quarter patter n of the right-hand accompaniment.
In mm. 17 21, the voice also abandons its relentless descents and floats gently
between D5 and G4; both voice and piano momentar ily cadence on G major in
m. 21. Although there is a br ief struggle in mm. 22 25 as the voice counters the
pianos descending figures with an ascent from A4 to E5, once the E5 is achieved
the air clears, and the equanimity of the preceding phrase is restored, an effect
confir med by another cadence on G major in m. 27. There follows a br ief piano
interlude that beg ins a chromatically har monized and rhythmically displaced
descent through the sixth G2 -B1 (the voice-leading of which is shown in ex. 6.5).
The voice joins the piano in m. 31, and the completion of the sixth-descent in
m. 34 is followed by a cadence on A minor in the following measure. An A minor
chord, prolonged through a chromatically inflected voice-exchange (as shown in
ex. 6.5), prepares the way for the return to E minor effected by the ar rival on the
B chord of mm. 42 43.
The return to E minor in m. 44 also marks a return to the mater ial of the A sec-
tion, which (other than the mezzo forte dynamic) is exactly the same as in its first
appearance until the left hand of the piano reaches the FS2-FS3 octave in m. 49 (as
it did in m. 6 of the A section). At this point, the rhythmic displacements of
mm. 27 41 reappear, and the bass continues, now in augmentation, to the A1-A2
octave of m. 54. The A1-A2 octave then leads, through AS, to the dominant of
mm. 56 57 and the final cadence.

example 6.5 Linear analysis of mm. 27 42 of Schumanns Im Rhein

27 notated 32 35 42

7 7 4
270 analysis and theory

Schumanns music thus enacts a threefold process. The first part of this process
(in the A section of mm. 1 16) establishes an austere, weighty E minor. It is well to
note that the tonality is established with minimal reference to the sort of harmonic
progressions and disjunct bass motions that are typical of tonal music of the per iod:
there is no cadence on E minor in the entire section, and, moreover, the dominant
of E minor is completely absent, its place taken by the leading-tone seventh-
chord.22 The second part of this process (in the B section of mm. 17 43) moves
completely, if only temporar ily, away from E minor. Here there are definitive
cadences, but on G major and A minor. The large-scale instability of this section (a
product of its avoidance of the tonal center so unequivocally established in the A
section) becomes apparent only g radually. Ultimately, E minor cannot be avoided
forever, and the section concludes (in mm. 42 43) with the very first appearance
of the dominant of E minor in the entire song. The third part of the process (in the
A' section of mm. 44 58), initiated by the definitive return to E minor, reprises the
opening music. However, the music is now colored by rhythmic displacements that
recall the latter half of the B section and that dr ive inexorably forward to a final
cadence in the lowest practicable register.
This music, together with Heines text, sets up the CIN diag rammed in figure
6.3. The gener ic space for the CIN is structured around the notion of exterior and
interior domains. In the textual space, the exter ior domain is that summoned in the
first stanza of the poem: the city of Cologne, with its r iver and cathedral. The inte-
rior domain is that summoned in the second two stanzas: the cathedral and the
painting with which the nar rator is obsessed. In the musical space, the exter ior
domain is that of E minor, with its stark octaves and str iding phrase rhythm. The
interior domain is that of the contrasting (or not-E-minor) music of the B sec-
tion, with its more flowing accompanimental patter n, softer dynamic, rhythmic dis-
placements, and large-scale instability.
Correlations between these idealized representations of exter ior and inter ior
spaces provide an opportunity for the combination of concepts in the blended
space, which draws on elements of the input spaces to create a r ich portrayal of the
actions of the nar rator and the environment within which they take place.

composition Composition combines the exter ior view of the river, city,
and cathedral with the austere E minor music of the A section and suggests a
sweeping and somewhat foreboding perspective, surveyed with a deliberative and
relentless gaze. In turn, the inter ior view of the cathedral and the painting within
combines with the soothing music of the B section to create a sense of refuge, a
surcease from the harsh realities without.

22. It is interesting to speculate on whether Schumann himself would have differentiated between
the dominant and the leading-tone chord. The distinction is of more concer n to theor ists of the period
than to composers. Nonetheless, Schumann would most certainly recognize the shar per dissonance cre-
ated by the leading-tone chord (with or without an added seventh), and it is worth noting that in the E
minor portions of Im Rhein he favored these over the relative consonance of the dominant chord
until m. 42.
word s, mu s i c, and s ong 271

F* "H#"(-*($,./$

K0LC#1)+E0# M51*E#1)+E0#
F* #1&#"-."*2-#3$*:$(-9"(?$ F* P$3-.*($315-B$*:$N*
B-#&?$,./$B,#%"/(,+$ 5"B#-*.$C,15#"("?$
F* -%&#"-."*2-#3$*:$ :*(")*/-.2E$
B,#%"/(,+?$',-.#-.2?$ F* .*#APA3-.*($315-B$*:$H*
'*(#(,-#$*:$6,/*..,$ 5"B#-*.$C5**#%-.2?$

F* 2(,&?$B*+/$C'"(%,'5$
F* -.#-3,#"?$5"B1("$

figure 6.3 CIN for Schumanns Im Rhein

completion Completing the image, we can imagine details of each domain:

an exter ior world gray and cold and in some measure threatening; an inter ior world
intimate and secure, if somewhat frag ile. We can also imag ine something of the
physical and emotional life of the nar rator. Outside the cathedral, he is almost cer-
tainly self-absorbed, a disheveled figure hurrying to escape the demons that tor ment
him. Inside, he stands rapt, even as the tumult in his mind gradually increases to the
point of becoming unbearable.

elaboration Our elaboration of the conceptual blend produced by Schu-

manns setting of Heines poem might well seek to develop a nar rative that could
explain the forces that dr ive the nar rator into the cathedral: a personality or per-
spective that makes life in the outside world intolerable, or a deep obsession with
the beloved that renders her both an object of worship and something to be owned
and controlled. However, we could also follow the lead of Schumanns music,
which, in response to the denouement of the poem, returns to the E minor music
of the opening. We are thus returned to the world outside the cathedral and there
can see the nar rator, now more disturbed than ever, stagger ing under the weight of
his obsession with his lover, just as the music staggers with the rhythmic displace-
ments of mm. 49 55. This furtherance of the nar rative by the music suggests that
it is the psychic tor ment of the nar rator that is behind the sense of foreboding sum-
moned by the exter ior domain, as well as for the ir reality of the inter ior domain.
It is an oppressive torment that absorbs all, and among whose fires the nar rator
272 analysis and theory

wanders, knowing no rest. The elaboration of the blend, then, suggests a profoundly
troubled mind to whom the Rhein represents more than a mir ror for a g reat and
holy place: it also offers the means to extinguish the fires that consume him.

The dark twists and tur ns of love explored in Heines Lyriches Intermezzo are, in their
way, not so very far removed from those Mller captured in his Die schne Mllerin.
However, where Mller develops the topic and its ramifications over the course of
twenty-three poems, it is Heines gift to be able to accomplish this in the space of just
one. This suggests a mastery of psychological detail both in understanding his sub-
jects and in what can be asked of his reader that is altogether impressive. Equally
impressive is Schumanns ability to exploit this level of detail by combining Heines
words with music that supports, expands on, and even offers a completion for the
story told by the poem. It might be argued that Schumanns setting str ips some of the
ambiguities from Heines poem, ambiguities that offer even more possibilities for
interpretation than those provided by the song. The point is well taken, for the con-
ceptual blend set up by the song emphasizes certain aspects of the topography of the
text space at the expense of others. And yet the gains are considerable, for the con-
ceptual world that emerges from syntactic invariances between Heines text and
Schumanns music is one r ich in detail and str iking in its immediacy. The challenges
and rewards of creating such worlds no doubt fueled the burst of creativity that led
to Schumanns amazing output of songs in 1840 and that produced a transfor mation
in the way composers thought about how to organize their musical mater ials.

in der fremde
The similar ities between Kleins and Schuberts settings of Trockne Blumen were
striking in that Schubert apparently did not know Kleins setting, and Schubert was
occupied with developing a song cycle as opposed to creating a single Lied: not
only are the musical details of each setting similar, but the gener ic spaces behind the
conceptual blends that each song sets up are almost exactly the same. In Schumanns
and Brahmss settings of Joseph von Eichendorff s In der Fremde, nearly the exact
opposite set of conditions holds. Brahms knew and respected Schumanns setting,
but he seems to have wanted to show his fr iend and mentor what he could do with
the same mater ial. The result is two songs every bit as different as those of Klein and
Schubert, but whose divergences stem from the gener ic spaces that structure the
blends produced by each song.

Eichendorff s In der Fremde

In der Fremde, the text and translation of which are given here, first appeared as
the text of an untitled song introduced in the course of Eichendorff s 1833 novella
Viel Lrmen um nichts.23 The scene in the novella is a huge and gloomy hall within

23. Joseph Freiherr von Eichendorff , Viel Lrmen um nichts, ed. Gerhart Baumann and Siegfr ied
Grosse, Neue Gesamtausgabe der Werke und Schr iften (Stuttgart: J. G. Cottasche Buchhandlung Nachf.,
word s, mu s i c, and s ong 273

a castle; outside, a storm rages, and, tucked away in a window niche, a young woman
sings to her own accompaniment on the guitar. Her song is a deeply sad one, con-
cerned with loneliness and thoughts of a solitary death far from home. The song
over, however, she springs up, full of life, to meet her beloved and his fr iends, who
have chanced upon her musings. The scene within which this poem first occurs is
thus one of sharp contrasts between the cold and forbidding inter ior of the cas-
tle, which makes for scant refuge from the stor m outside, and the blushing and lively
young woman who rushes to embrace her fr iends.

in der fremde in foreign lands

Aus der Heimat hinter den Blitzen rot From home, behind the red lightning
Da kommen die Wolken her, There come clouds,
Aber Vater und Mutter sind lange tot, But Father and Mother are long dead;
Es kennt mich dort keiner mehr. No one there knows me anymore.
Wie bald, wie bald kommt die stille Zeit, How soon, how soon will that quiet time come,
Da ruhe ich auch, und ber mir When I too shall rest, and over me
Rauscht die schne Waldeinsamkeit, Shall rustle the forests beautiful solitude,
Und keiner mehr kennt mich auch hier. And no one here shall know me anymore.

The poem itself offers images and impressions rather than any specific nar rative
structure. Dominant are the images of nature (in the for m of the red lightning,
approaching clouds, and promise of the forests loneliness). Of equal importa