Society for Music Theory

Functional Harmony Revisited: A Prototype-Theoretic Approach

Author(s): Eytan Agmon
Source: Music Theory Spectrum, Vol. 17, No. 2 (Autumn, 1995), pp. 196-214
Published by: University of California Press on behalf of the Society for Music Theory
Stable URL: http://www.jstor.org/stable/745871
Accessed: 03/03/2009 18:13

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

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucal.

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

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

University of California Press and Society for Music Theory are collaborating with JSTOR to digitize, preserve
and extend access to Music Theory Spectrum.

http://www.jstor.org
 Functional Harmony Revisited:
A Prototype-TheoreticApproach

EytanAgmon

I. PRELIMINARIES physical continuum, yet we hear them as three qualitativelydistinct

and discrete categories. In all three cases, perceptual boundaries
"How do we sort the objects, people, events, and ideas in have somehow arisen along the physical continuum, dividing it into
the world into their proper categories? What transforms the discrete regions, with qualitativeresemblanceswithineach category
'booming, buzzing confusion' that enters our eyes and ears and qualitativedifferences betweenthem. These bounded categories
at birth into that orderly world we ultimately experience and may provide the groundwork for higher-order cognition and lan-
interact with?" With these questions Stevan Harnad intro- guage.
duces an important collections of essays entitled Categorical Harnad's example of color categories aptly introduces the
Perception: The Groundwork of Cognition.' As Harnad im- ideas of prototype and prototypicality, quite often associated
mediately explains, with categorical perception. Consider the color category
"red." The perceptual class of all reds consists of a variety
Categorical perception occurs when the continuous, variable, and
confusable stimulationthat reaches the sense organs is sorted out by of different hues, the "redness" of which, intuitively, comes
the mind into discrete, distinct categories whose members somehow in varying strengths (indeed, there are borderline cases
come to resemble one another more than they resemble members which can be cross-classified as "orange" or "purple"); pro-
of other categories. The best-known example is color categories: totypicality is a technical term used to refer to these varying
Physically speaking, colors differ only in their wavelengths, which strengths of redness. When ordering all reds on a prototyp-
graduallyget shorter across the spectrumof visible colors. What we icality scale, an apex is reached in the form of a particular
see, however, are qualitative changes, from red to orange to yellow red (or perhaps a narrow band of reds) that exemplify "sheer
to green, and so forth. The same is true of musicalpitches: Gradually
redness"; prototype is a technical term used to refer to such
increasingfrequencies can come to be heard as categorical changes a red. Prototype theory is a theory of categorical perception
from C to C-sharpto D to E-flat. A lesser-knownexample is "stop-
cast in prototype-structural terms. All three examples of cat-
consonants": (synthesized) "ba," "da," and "ga" also vary along a
egorical perception cited by Harnad in this passage-that is
An early version of this paper was presented at the Annual Meeting of
to say, not only color perception, but also pitch perception
the Society for Music Theory, Kansas City, 1992. and speech perception-have received prototype-theoretic
'Ed. Stevan Harnad (Cambridge:CambridgeUniversity Press, 1987), ix. accounts. Prototype theory has also been applied to higher-
 Functional Harmony Revisited 197

level mental processes involving semantic analysis, memory The present article applies prototype theory to the domain
organization, and abstract thought.2 of harmonic theory. Specifically, it presents a prototype-
Music cognition is a relatively young discipline, formed in structural account of the music-theoretic construct known
the wake of the cognitive revolution of the late fifties. Like as harmonic functions. The hallmarks of functionalism are:
all branches of cognitive science, music cognition is inter- (1) the characterization of individual chords as tonic (T),
disciplinary in nature: philosophy, psychology, neurology, subdominant (S), or dominant (D) in function; and (2) the
linguistics, computer science, and of course, music theory, notion that the so-called primary triads I, IV, and V somehow
have all contributed to shaping the field. Given that prototype embody the essence of each of these functional categories.5
theories have 6een highly influential in cognitive psychology A theory of functions as such was first announced by Hugo
in the last two and a half decades, it is not surprising that such Riemann, who openly acknowledged his debts to other the-
theories have found their way into music-cognitive discourse. orists. Riemannian Funktionstheorie engendered consider-
A recent issue of Psychomusicology, for example, is devoted able controversy early on, and continues to do so even today.
in its entirety to the idea of music prototypes. According to Yet its significance is borne out by the many modern accounts
Mari Riess Jones, the guest editor, the issue shows that the of traditional harmony which incorporate one version or an-
idea "is useful in developmental approaches to music per- other of functionalism as an essential component. Possible
ception, in generating new approaches to rhythm perception examples are the harmony textbooks of William Mitchell,
and production, in understanding aspects of tonality, as well Allen Forte, and Joel Lester, from the last of which Example
as in expanding our understanding of complex music struc- 1 is reproduced.6 One conspicuous counterexample would
ture."3 Among the researchers who have referred to proto-
type theory in a musical context even before the appearance "TimbralHierarchies,"ContemporaryMusic Review2 (1987): 144-45; Anna
of this issue of Psychomusicology are Carol Krumhansl, Fred Unyk, "An Information-ProcessingAnalysis of Expectancy in Music Cog-
Lerdahl, and Anna Unyk.4 nition," Psychomusicology9 (1990): 236-37.
5For empirical studies of harmonic representation incorporating a
prototype-theoreticapproachsee Carol Krumhansl,JamshedBharucha, and
2Fora prototype-theoreticaccount of color perception see Eleanor Rosch EdwardKessler, "PerceivedHarmonicStructureof Chordsin Three Related
Heider, "Universalsin Color Naming and Memory,"Journalof Experimental Musical Keys," Journal of ExperimentalPsychology: Human Perceptionand
Psychology 93 (1972): 10-20; for a prototype-theoreticaccount of pitch per- Performance8 (1982): 24-36, and Jamshed Bharuchaand Carol Krumhansl,
ception see Eytan Agmon, "Towardsa Theory of Diatonic Intonation," In- "TheRepresentationof HarmonicStructurein Music:Hierarchiesof Stability
terface 22 (1993): 151-63; and for a prototype-theoretic account of speech as a Function of Context," Cognition 13 (1983): 63-102. These studies dif-
perception see Gregg Oden and Dominic Massaro, "Integrationof Featural ferentiate between a harmoniccore consistingof the triads I, IV, and V, and
Informationin Speech Perception,"PsychologicalReview 85 (1978): 172-91. the remainingfour diatonic triads, but do not otherwise partition the group
Eleanor Rosch has been highly influentialin prototype-theoreticaccounts of of seven diatonic triads into three functional categories.
higher-level mental processes; e.g., "Classificationof Real-World Objects: 6WilliamMitchell, ElementaryHarmony, 2nd ed. (New Jersey: Prentice-
Origins and Representations in Cognition," in Thinking:Readings in Cog- Hall, 1948), 65-66; Allen Forte, Tonal Harmony in Concept and Practice
nitive Science, ed. P. N. Johnson-Lairdand P. C. Wason (Cambridge:Cam- (New York: Holt, Rinehart and Winston, 1962), 120; Joel Lester, Harmony
bridge University Press, 1977), 212-22. in TonalMusic, vol. 1 (New York: Knopf, 1982), 21, 25, 251. See also Marion
3Mari Riess Jones, "Preface," Psychomusicology 10 (1991): 71. Guck, "TheFunctionalRelations of Chords:A Theory of MusicalIntuitions,"
4Carol Krumhansl, "The Psychological Representation of Musical Pitch In Theory Only 4 (1978): 29-42; Charles J. Smith, "Prolongationsand Pro-
in a Tonal Context," Cognitive Psychology 11 (1979): 349; Fred Lerdahl, gressionsas MusicalSyntax,"in Music Theory:Special Topics, ed. Richmond
198 MusicTheory Spectrum

Example1. A contemporaryview of harmonicfunctions.FromJoel Lester, Harmonyin TonalMusic, vol. 1, p. 251. Reproducedwith

permission.
Subdominant Tonic Dominant Modal Dominant
- A A---- A----. A of the Dominant

.1 Ir
t' eC J~~~~

C maj: II IV VI I I11 V VII 1I

seem to be Aldwell and Schachter's Harmony and Voice work which is intuitively appealing, and which also promotes
Leading, where the idea of harmonic functions is never ex- rigor and consistency. Second, this framework promotes re-
plicitly endorsed; yet careful study of this text reveals that examination of certain deep-rooted assumptions inherent in
beneath the surface there is more than a trace of function- a traditional functional approach. As an example, consider
alism in its approach.7 Riemann's derivation of the three primaryharmonictriads I,
Recruiting prototype theory in the service of functional IV, and V from a symmetrical,dualisticallymotivated under-
theory achieves, I believe, a numberof goals. First, prototype dominant and over-dominant relationship to a given tonic;
theory provides functional theory with a conceptual frame- this derivation (which harks back to Rameau) seems not to
have been superseded since in any essential respect.8 Yet the
special status of root relationships by fifth is by no means a
Browne (New York: Academic Press, 1981), 139-74; idem, "The Functional necessary assumption in the theory of harmonic functions.
Extravaganceof ChromaticChords," Music TheorySpectrum8 (1986): 94- This is a significantfinding, for it means that functional con-
139.
7EdwardAldwell and Carl Schachter, Harmony and Voice Leading, 2nd siderations, on the one hand, and chord-progressionalcon-
ed. (San Diego: Harcourt Brace Jovanovich, 1989). See, e.g., Chapter 9, siderations, on the other, can be separated from each other.
where IV, II, and II6 are treated together as "intermediateharmonies."See
also pp. 185 and 153, where VI is treated "as substitute for I" and as an
"intermediateharmony"(a term usuallyreservedfor IV and II), respectively.
Significantly,in their discussion of the I-III-V progression (p. 213) the au- 8See, for example, David Lewin's Generalized Musical Intervals and
thors take considerable pains to distinguishbetween the "intermediatehar- Transformations(New Haven: Yale University Press, 1987), 171-73, where
monies" (II, IV, and VI) and the III chord, which also "leadsfrom an opening the fifth-relatednessof IV and I, and V and I (represented by their roots)
tonic to a dominant." is depicted by a transformationgraph.
 FunctionalHarmonyRevisited 199

It is a central-albeit unproven-thesis in the present article issues in dispute are considered in light of the theory as pro-
that harmonic theory in general is best conceived in terms of posed in the present article. Although a clear line is drawn
two independent, interacting components, namely a theory between the notions of harmonic function and chord pro-
of functions (which describes the three harmonic categories) gression, some tentative suggestions as to how the two com-
and a theory of chord progression (which characterizes the ponents might possibly interact to produce a richer and in-
various root progressions, ascending third, descending fifth, tuitively more satisfying harmonic theory are included in
etc.). Indeed, one way of stating the core idea of the present Parts 4 and 5 of the article, which complete the theoretical
article is: given a separation of chord progression from har- discussion begun in Part 2.
monic function, the notions function and primary triad are
fully reducible to category and prototype, respectively.
It is my firmbelief that music theory, although a latecomer II. A PROTOTYPE-THEORETIC ACCOUNT OF HARMONIC FUNCTIONS
into the cognitive-scientificarena, is no second-rate player in
that field. The basis for my belief is twofold. First, music Translating functionalism into prototype-theoretic terms
theory has had a rich and impressive history-probably un- necessarily involves three operations: (1) selecting as pro-
paralleled in the theory of the arts; this history is replete with totypes the three primarytriads I, IV, and V; (2) associating
significantcognitive undertones. Second, the ontological sta- with each prototype a group of triads, corresponding (to-
tus of music is possibly unique, being a product of human gether with the prototype) to a conventional functional
cognition minimallyconstrainedby the structureof the outer category, as in T= {I,VI,III} (see Example 1); (3) associating
world. By this I mean that music seems to fulfill Claude Levi- with each categorized triad a prototypicality index corre-
Strauss'sdictum that "when the mind is left to commune with sponding to the triad's functional strength (e.g., the sense in
itself and no longer has to come to terms with objects, it is which I is the strongest tonic triad, followed by both VI and
in a sense reduced to imitating itself as object."9 Music, in III in the second place). First, however, some preliminaries
other words, offers a unique and invaluable window into are in order.
the inner workings of the human mind. It follows that a Let the notion degreeof triadicsimilaritybe defined for all
prototype-theoretic approach to harmonic functions could distinct pairs of triads formed from tones of a given diatonic
serve as an important impetus to theories of the mind which collection; in particular,let two (distinct) diatonic triads hav-
uphold the existence of categories and prototypes. ing two tones, one tone, or no tones in common be termed
No approach to functional harmony, however innovative, respectively maximally similar, intermediatelysimilar, and
can afford to ignore the controversy which has surrounded minimally similar.
the idea since its inception. Part 3 of this article, therefore, In Figure 1 the seven diatonic triads are ordered cyclically
is a critical discussion of the functional controversy;the main by third. In this representation, maximally similar triads oc-
cupy adjacent positions on the circle's circumference;if the
shortest distance along the circumference of the circle con-
9ClaudeLevi-Strauss, The Raw and the Cooked, trans. J. and D. Weight- necting two triads passes through one or two intervening
man (New York: Harper & Row, 1969), 10; quoted in Howard Gardner, The positions, the triads are intermediately similar or minimally
Mind's New Science (New York: Basic Books, 1985), 240-41. similar, respectively.
200 MusicTheory Spectrum

Figure 1. The circle of thirds, representingdegrees of triadic Figure2.

similarity
a) Degrees of triadicsimilarityin relationto the tonic triad

III maximal i
similarity VI

V IV
intermediate ; similarity
V-\ /IV

VII
VII ii11
II \ >
minimal m
similarity

VII II

b) The selection of the three harmonicprototypes

One might object that common tones are not the only
possible criterion by which similarity between triads might
be assessed; for example, triads may also be compared on
the basis of their intervallicstructure, in which case all triads
of a given quality (major, minor, etc.) would be regarded as / \
maximally similar. Yet there are compelling theoretical
grounds for grantingthe common-tone criterion a privileged
status in this case. After all, what makes a specific diatonic
triad (say, IV) a unique object is its pitch content, not its
intervallic structure. In other words, pitch content is the de-
fining property of specific diatonic triads; it follows that
pitch content must be the main criterion by which similarity
between such triads is assessed. vn
Figure 2a adds an important assumptionto the content of
Figure 1, namely, that the tonic triad is referential within the
diatonic triadicset. Thus, if Figure 1 depicts degrees of triadic
similarityin general, Figure 2a depicts, in addition, degrees
of triadic similarity in relation to the tonic triad. There are
 FunctionalHarmonyRevisited 201

c) The delineationof categories entails V). The principle of symmetry follows from the ref-
erential status of the tonic triad. A triad selected as prototype
is compared to the tonic triad; the comparison necessarily
evokes exactly one other triad, namely, the triad which cor-
III I responds to the selected prototype in terms of its degree of
/\
/ \TONIC/ \ similarity to the tonic triad.
The third and last principle states that two maximallysim-
ilar triads cannot, simultaneously, be selected as harmonic
prototypes (in other words, no two prototypes may occupy
adjacent positions on the circle of thirds depicted in Figure
1). Unlike the first two principles, which pertain solely to the
theory of harmony, this third principle is more universal.
Clearly, prototypes must be maximally dissimilar to each
other if categorization in general is to serve any useful pur-
pose; how could subjectsbenefit from categorizationif it were
difficult to differentiate one prototype from another? In the
three such degrees, namely maximallysimilarto the tonic (VI color domain, for example, to select both "yellow" and "yel-
and III), intermediately similar to the tonic (IV and V), and low with a slight orange tinge" as prototypes is inconceivable;
minimallysimilarto the tonic (II and VII). The obvious sym- similarlyin the harmonicdomain, it is inconceivable to select
metry of these relationships is highlighted in the figure.10 as prototypes, say, both II and IV.11
There are three principles on the basis of which the pri- These three principlesuniquely select the triads I, IV, and
mary triads I, IV, and V may be selected as prototypes. The V as prototypes of three harmonic categories (Figure 2b).
first principle is self-evident: the tonic triad, by virtue of its Two self-evident propositions (P1 and P2, below) determine
special, referential status, is a prototype. (Referential and the additionalmembers of each category and their respective
prototypical are synonymous for all practicalpurposes.) The prototypicalities:
second principle may be termed the principle of symmetry.It
states that the symmetrical structure depicted in Figure 2a P1. Additionaltriadscan only be admittedinto a given category
in an orderwhichcorrespondsto theirdegreeof similarityto the
must not be violated. In other words, if one selects any triad
as prototype, one must also select its mirrorimage (e.g., the
"A "principleof maximal differentiation"has been proposed by Andre
selection of VI entails the selection of III and vice versa, IV Martinetfor the domain of phonemic structure:"if by accident a given pho-
neme is not as much differentiatedfrom its neighbours in the system as the
m'Concerningthe criteriaby which the tonic is selected, see Eytan Agmon, organs could achieve, one might expect the articulationof the phoneme in
"Tonicityand the Tritone: Beyond the Rarity Issue," Proceedingsof the First question to be modified until such maximal differentiationis obtained" (El-
InternationalConferenceon CognitiveMusicology (Jyviskyla, Finland: Uni- ements of General Linguistics, trans. Elisabeth Palmer [London: Faber &
versityof Jyvaskyla,1993), 74-87. Common tones also play an importantrole Faber, 1964], 191). See also Eytan Agmon, "Towardsa Theory of Diatonic
in Riemann's theory, and indeed, in many earlier as well as later harmonic Intonation," 154, for an application of a similar principle to the domain of
theories. pitch perception.
202 MusicTheory Spectrum

category'sprototype(thatis, triadsmaximallysimilarto the given validity of the idea of harmonic functions in the first place,
prototypeare admittedfirst, etc.); moreover,a triad'sdegreeof a question which has generated, as the reader is probably
similarityto the givenprototypeis synonymousto thattriad'spro- aware, some rather heated debate.
totypicality.
P2. The processof admittancedescribedin P1 muststopjust short
III. HARMONIC FUNCTIONS: EMPIRICAL FACT
of anotherprototype,sinceonlyone prototypecanbe a memberof
OR THEORETICAL FICTION?
any givencategory.

In applying these two propositions, it remains to be de- It should be clear by now that the term "theory of har-
cided whether P2 is the sole constraint on admittance into a monic functions" is used in the present article with a very
category according to P1, or whether the process of admit- specific meaning. "Function," "primarytriad," and "func-
tance might possibly stop before P2 becomes applicable. Al- tional strength"are reduced to "category,""prototype,"and
though one could conceivably formulate some principle to "prototypicality," respectively. Moreover, the theory is
stop admitting triads into a given category earlier than P2 stripped of any chord-progressionalconnotations, and thus
dictates, parsimony urges us not to do so. Why burden the has no standing with regard to how the three functional cat-
theory with an additional assumption, one whose rationale egories may follow each other in time. Nonetheless, signif-
icant connections do exist between the proposed theory and
may be difficult to establish, if P1 and P2, which are self-
evident, suffice?Thus, I propose that P1 and P2 complete the the historical Funktionstheorie,usually attributed to Hugo
Riemann. Riemann's theory has been severely criticized on
theory, as shown in Figure 2c.
A comparisonof Figure 2c with Example 1 reveals that the several accounts, even to the point of questioning the ex-
istence of harmonicfunctions altogether. The present section
proposed theory accordsvery well with the conventional view
of harmonicfunctions. Yet the decision not to invoke a con- addresses two of the most prominent criticismsof Riemann:
one which stems from a rival approach to harmony known
straining principle on category membership other than P2
results in a somewhat curious feature: not only VI and III, as Stufentheorie, and another which stems from a theory
but also II and VII, have dual citizenship. That is to say, in which purportsto redefine altogether the notion of harmony,
addition to the primary(and conventionallywell-established) namely the Schenkerianapproachwith its emphasis on hier-
classification of II and VII as subdominant and dominant archical structure and voice leading.
triads, respectively, the function of II emerges as weakly Stufentheorieversus Funktionstheorie.A historically im-
dominant, and correspondingly, that of VII as weakly sub- portant alternative to Riemann's Funktionstheorieis an ap-
dominant (weakly in this context refers to the triad's proto- proach to harmonyknown as Stufentheorie(literally, "theory
of harmonic degrees," but more appropriately, "theory of
typicality, for a relation of intermediate similarity, rather
than maximal similarity, obtains between the triad and its degree progression").l2 Stufentheoriestresses the individu-
category'sprototype). The dual functionalstatus of II and VII
'2Simon Sechter is usually considered the most prominent proponent of
surely calls for separate discussion; however, since the issues
Stufentheorie.See Robert Wason, VienneseHarmonicTheoryfrom Albrechts-
involved are rather specialized, and moreover touch upon a
bergerto Schenkerand Schoenberg(Ann Arbor: UMI Research Press, 1985),
number of intricate side issues, such discussion appears in 33; Carl Dahlhaus, Studieson the Originof Harmonic Tonality,trans. Robert
Part 4 of this article. A more pressing question concerns the Gjerdingen (Princeton: Princeton University Press, 1990), 33 ff.; William
 FunctionalHarmonyRevisited 203

ality and independence of the seven harmonicdegrees. More- character,' does it not follow necessarily that all scale-
over, unlike Funktionstheorie,where the primary harmonic degrees-not only I, IV, and V-must be recognized as scale-
model is the I-IV-V-I progression, Stufentheorie leans degrees in their own right?"14
heavily on the cycle of descending fifths I-IV-VII-III-VI- The Schenkerianapproach: hierarchy and voice leading.
II-V-I. While Schenker's 1910 critique of functionalismis of a clear,
Both aspects of Stufentheoriefigure prominentlyin Schen- Stufentheorie origin, the subsequent development of his
ker's acrimoniousattack on functionalismin the first volume theories, with its well-known emphasis on voice leading and
of Counterpoint,echoes of which reverberate in the writings hierarchicalstructure, has inevitably led to friction of quite
of some of his followers.13For example, Schenker questions a different sort. Jonas, for example, states that functionalism
the explanatory value of an account of the major system "hadto fail, because it neglected the fact that two occurrences
where the "individualscale degrees, except for I, IV, and V, of the same chord could be worlds apartin meaning, and that
are deprivedof their independence and thus of their attractive everything depended on context."15In a footnote to that
capability of assuming various functions"; and in discussing statement, John Rothgeb concedes that functionalismallows
harmonic progressions by descending fifths, he asks: "Con- a secondary chord, such as III, to assume different functions
sidering that none of these passages manifests 'sequential in different contexts, but nevertheless insists that "the
theory fell far short of a recognition of the phenomenon of
composing-out, and, as a result, failed to discriminate be-
Caplin, "Harmonyand Meter in the Theories of Simon Sechter,"Music The- tween vertically and horizontally generated chords."
ory Spectrum2 (1980): 74-89.
Ernst Kurth, who used the term Fundamenttheorieto refer to Sechter's My response to all of these important objections rests on
three major points: (1) the proposed version of functional
system, seems to have been the first scholar to conceive of Sechter and Rie-
mann as two major opposing figures in the theory of harmony. See Ernst theory is not Riemann's; (2) the theory is not meant to ex-
Kurth, Die Voraussetzungender theoretischenHarmonik und der tonalen haust the harmonic domain; and (3) the proposed theory is
Darstellungssysteme(Bern: Max Drechsel, 1913), 6 ff. and 89 ff. See also Lee compatible with a hierarchicalapproach.
Rothfarb, "Ernst Kurth's Die Voraussetzungender theoretischenHarmonik
and the Beginningsof Music Psychology," Theoria4 (1989): 19-20. Needless Departuresfrom Riemann.Certaincontroversialaspects of
to say, Stufentheorieas well as Funktionstheorieare, to a considerableextent, Riemann's theory have no counterpartsin the proposed the-
abstractions,for neither exists in the work of a single theorist in as pure a ory. Most notably, the proposed theory abandonsRiemann's
form as their hypothesized rivalry suggests.
13See especially Oswald Jonas, Introductionto the Theory of Heinrich
Schenker,trans. and ed. John Rothgeb (New York: Longman, 1982), 127-28,
and Rothgeb's review of Hellmut Federhofer, Akkord und Stimmfiihrungin 14HeinrichSchenker, Counterpoint,vol. 1, trans. John Rothgeb and Jiir-
den musiktheoretischenSystemenvon Hugo Riemann, Ernst Kurthund Hein- gen Thym, ed. John Rothgeb (New York: Schirmer, 1987), 23, 27. As John
rich Schenker (Vienna: Verlag der OsterreichischenAkademie der Wissen- Rothgeb explains in a footnote on p. 348, "the application ... [of the theory
schaften, 1981) in Music TheorySpectrum4 (1982): 131-37, where "the rec- of tonal functions] is especially problematical in passages involving 'se-
ognition of only three principal functions" is dubbed "specious"(132). See quences' by descending fifths. This led Riemann to declare that 'as was
also Federhofer, Beitrage zur musikalischen Gestaltanalyse(Graz: Akade- recognized first of all by Fetis, however, the true harmonic movement-
mische Druck, 1950) and Akkord und Stimmfiihrung;and Matthew Brown, the cadential progression-remains stationary for the duration of the
"A Rational Reconstruction of SchenkerianTheory" (Ph.D. diss., Cornell sequence....' " The reference is to Hugo Riemann, Handbuch der Har-
University, 1989), 192-99. Both Brown and Wason (Viennese Harmonic monielehre, 7th ed. (Leipzig: Max Hesse, n.d.), 202.
Theory, 134-35) present synopses of Schenker's arguments. 15Jonas,Introduction, 127.
204 Music Theory Spectrum

notion of "apparent consonance" (Scheinkonsonanz), which quite apart from any attendant commitment to a dualistic
truly deprives the so-called secondary triads, right from the doctrine.19
start, of any potential independence.16 More generally, the The scope of the theory. As already indicated, harmonic
Riemannian sense in which a secondary degree is said to theory is viewed in the present article as a composite of
represent or substitute for a primary degree does not exist in two main subtheories: a theory of functions and a theory of
the proposed theory. Unlike Riemann's somewhat ambigu- chord progression. This view is indebted to the historical
ous use of tonic, subdominant, and dominant to refer to both FunktionstheorielStufentheorie opposition; however, the cor-
functions and/or chords, in the proposed theory these terms respondence is far from exact. Most notably, the proposed
are used exclusively to refer to functions, that is, categories functional theory is static, in the sense that it merely describes
of chords; reference to individual chords is made by Roman the three harmonic categories and their internal structure; the
numerals.17 II, for example, represents an abstract category T-S-D-T paradigm of functional succession (a venerable
(namely the subdominant), which is also represented by IV; component of Funktionstheorie) is not within its scope. The
at the same time, IV is of course the more prototypical exclusion of functional succession from functional theory
representative of that category. might seem a bit odd. However, although excluded from
The proposed theory largely bypasses another highly con- functional theory, the T-S-D-T paradigm remains a part of
troversial aspect of Riemann's theory, namely dualism.18 Af- harmonic theory in general-that is, the interaction of the
ter all, there is no necessary connection between function- theory of functions with a theory of chord progression.
alism and dualism, however essential it might have seemed Once the scope of functional theory is reduced, it cannot
to Riemann to view the rules of harmony and the major/ be faulted for failing to deal convincingly with the cycle of
minor polarity as stemming from a single source. Indeed, descending fifths. The priority of the descending-fifth rela-
there is no lack of evidence that functionalism can flourish- tionship is clearly an aspect of the theory of chord progres-
sion, not the theory of functions. When Schenker and his
followers point out, for example, that the II in II-V is not
merely a IV-substitute because a descending-fifth relationship
16SeeDahlhaus, Studies, 38: "the concept of functions can be separated obtains between the II and the V, they are certainly correct;
from Riemann'smethod of demoting secondarydegrees to dissonantvariants but their argument concerns the theory of chord progression
of primarydegrees, so that one can retain the concept of fundamentalpro- (as it interacts with the theory of functions), not the theory
gressions without giving up the concept of functions." For Dahlhaus's elab- of functions per se. It is surely senseless to reject a theory
oration of this statement, see especially pp. 57-59.
17Concerning the Riemannian chord/functionambiguity, see Dahlhaus,
Studies, 50. In a paper entitled "An Idea and its Politics: Hugo Riemann's
Treatmentof Harmonic Function," delivered at the 1992 annual meeting of 19Sincea number of American theorists have taken renewed interest in
the Society for Music Theory, Daniel Harrison has argued that the chord/ Riemann in general and dualism in particular,it might be clarifiedthat the
functionambiguityis not so much inherent in Riemann'stheory (as Dahlhaus intention here is not to rekindlethe dualismcontroversy;rather,the intention
implies), as it is an unfortunate result of Riemann's attempts to make his is merely to rebut criticismsof Riemann's dualism that might be addressed
theory more generally accessible. toward the proposed theory. An example of American Neo-Riemannism is
8sTheclosest counterpartin the proposed theory to Riemann's dualism David Lewin, "A Formal Theory of Generalized Tonal Functions,"Journal
is the principle of symmetry, discussed in connection with Figure 2. of MusicTheory26 (1982):23-60.
 FunctionalHarmonyRevisited 205

outright merely because its success is limited, to one degree because they deal with musical motion-that is, counter-
or another, in terms of the total domain under consider- point."23A number of theorists have attempted to draw the
ation.20 full consequences from such a view-with disappointing re-
The contextual role of hierarchy and voice leading. No sults, in my opinion.24 Moreover, as the appendix to this
obstacle bars combining a functional view of harmony with
Schenker's important insights into the hierarchicalnature of
tonal pitch structure, a hierarchy expressed, to a significant 23LeoKraft, Gradus:An IntegratedApproachto Harmony, Counterpoint,
and Analysis, vol. 1 (New York: W. W. Norton, 1976), 30. An extended
extent, throughvoice leading. For example, when Federhofer reference to Schenker on p. 117 (under the heading "Why Study Counter-
observes that in the progression II-IV-V (soprano: 2-i-?) point?") precedes the statement that "our definition of counterpoint, taking
the metrically weak IV-chord is subservient to the II, and a broad view, includes most of what is taught in courses called 'Harmony'."
not vice versa, there is no conflict with functionalism, for as 24PeterWestergaard'sAn Introductionto Tonal Theory (New York: W.
he himself points out, "undoubtedly the supertonic triad W. Norton, 1975) is probablythe most comprehensiveand consistent attempt
to date to implementa thoroughlylinear, quasi-Schenkerianapproachto tonal
assumes in this context a subdominant quality."21
theory. In Westergaard'ssystem not only is the seventh-chordconceptually
An altogether different (and much more loaded) question non-existent as a harmonic entity (on p. 325 one finds a reference to "a
is whether the concepts of voice leading and hierarchical collection of pitches-the so-called dominantseventh"),but also non-existent
structure are so powerful as to render "harmony"as tradi- are the diminished and augmented triads (p. 41; however, see p. 105). In
tionally conceived more or less obsolete, except, perhaps, as keeping with his thoroughly linear approach, Westergaardderives the V-I
cadence from a set of melodic fragments ending on a member of the tonic
the notion applies to some higher levels of tonal structure
triad, claimingthat some combinationsof these fragments"aremore generally
(Schenker's Stufen). The idea that harmony is reducible to useful than others" (325). With a truly remarkablesleight of hand, though,
counterpoint is surely connected in many people's mind with Westergaardincludes among his melodic fragments (which are otherwise
Schenker's name. In his review of Felix Salzer's Structural exclusively stepwise) the succession s-i.
A less extreme attempt to approach harmony in contrapuntalterms is
Hearing, for example, Milton Babbitt refers to "the more
William Benjamin's "Pitch-ClassCounterpoint in Tonal Music," in Music
autonomous nature of contrapuntal discipline as opposed
Theory: Special Topics, 1-32. In another article, "Models of Underlying
to harmonic discipline, the latter being almost completely Tonal Structure: How Can They Be Abstract, and How Should They Be
inferable from the former-involving at most a shift of Abstract?" Music Theory Spectrum4 (1982): 28-50, Benjamin claims that
emphasis-while the reverse is not possible."22 Leo Kraft "accordingto the theory [of pitch-class counterpoint], tonal harmonic pro-
states even more categorically that "the term harmony is gressionsare counterpointsof four pitch-classvoices, motion in each of which
is determined by motion in one or more of the others. Therefore, it is the
virtually useless ... If harmony books make any sense, it is complex of lines in a progressionwhich determines its identity, and not the
chordsof which it is comprised"(40). However, Benjamin'stheory is far from
20Cf.Brown ("A Rational Reconstruction," 192-99), who argues force- being harmonicallyinnocent. In particular, the voice-leading rationale un-
fully (after Schenker) againstfunctionalism,while at the same time admitting derlyingsome of his rules is not apparent(e.g., Rule 8 on p. 7, and the special
that "the notion of harmonicfunctions certainly accounts for a large amount "spawning"status of the ascending fourth). As a result, the appearance of
of tonal music" (195). a dependency of harmony on counterpoint is deceptive.
21". . . zweifellos nimmt der Dreiklang der II. Stufe in diesem Zusam- The most recent work in this lineage is Arthur Komar, Linear-Derived
menhang eine unterdominantischeFirbung an," Beitrdgezur musikalischen Harmony (Boston: Ovenbird Press, 1992). However, Komar refrains from
Gestaltanalyse,13. See also his Akkord und Stimmfihrung, 17-18. adopting a decisive theoretical stance on the harmony-versus-counterpoint
22Journalof the American Musicological Society 5 (1952): 264. issue, stating "it appears that the question of the priority of harmony or
206 Music Theory Spectrum

article shows, contrary to existing belief voice-leading con- In summary, harmonic functions are not easily dismissed
siderations do not suffice to account for certain well-known as theoretical fiction. To be sure, neither is it easy to prove
norms of doubling in four-part harmony. that harmonic functions empirically exist. But in as much as
Of course, Schenker did indeed canonize the horizontal the burden of proof in this case seems to weigh heavier with
dimension at the expense of the vertical, most notably in his those who question the existence of harmonic functions, I
highly polemic 1930 essay "Rameau oder Beethoven?"25 But shall assume henceforth that harmonic functions exist, and
while one may sympathize with Schenker's own need to dis- turn to a number of more tangential theoretical issues as-
tance himself from many of his predecessors (and contem- sociated with the present, prototype-theoretic approach.
poraries) so as to highlight the originality and boldness of
his ideas, today, when the significance of his achievement is
IV. THE DUAL FUNCTIONALSTATUSOF II AND VII
hardly open to question, such a need no longer exists.26 I am
convinced that for many musicians today the choice between
As discussed at the end of Part 2 of this article, II and VII
a Schenkerian approach and traditional harmonic theory is a
have a dual functional status in the proposed theory: the
painful one; and despite the ominous subtitle of Schenker's
essentially subdominant II is also weakly dominant, and sim-
above-noted essay, intellectually (and artistically) speaking it
ilarly, dominant VII is weakly subdominant. How can one
is not a choice between life and death.27
substantiate such a view? Since a reliable method for deter-
mining the functional status of a given chord empirically does
counterpointis comparableto the question of the precedence of the chicken not seem to exist, ultimately one must rely on one's intuitions.
or the egg" (12 n. 8). These intuitions, however, are rather subtle, and may not be
25HeinrichSchenker, "Rameau oder Beethoven? Erstarrungoder geis-
immediately available for introspection.
tiges Leben in der Musik?"Das Meisterwerkin der Musik, vol. 3 (Munich: Before presenting examples from the literature of II and
Drei MaskenVerlag, 1930;facsimile, Hildesheim:Georg Olms, 1974), 11-24;
English translationby SylvanKalib in "ThirteenEssays from the Three Year-
VII(7) in unconventional functional guises, it should be useful
books Das Meisterwerkin der Musik by Heinrich Schenker: An Annotated to draw a distinction between a given triad's functional po-
Translation"(Ph.D. diss., NorthwesternUniversity, 1973), vol. 2, 492-518. tential (e.g., the potential of II to assume subdominant or
26See Harald Krebs, "Schenker'sChanging View of Rameau: A Com- dominant function, as opposed to its inability to assume tonic
parison of Remarks in Harmony, Counterpoint, and 'Rameau or
Beethoven?'," Theoria 3 (1988): 59-72. function), and the realization of that potential in context. For
27Irecall hearing a pre-eminent Schenkeriandisciple describe the chord the most part, how a given functional potential is realized in
succession in the opening four measures of the C-majorPrelude from Bach's context is irrelevant to present purposes; nonetheless, in the
Well-TemperedClavier, Book I, solely in terms of neighboring motions. case of VII that question is tentatively addressed.
Should one deny the harmonic content of these measures? Could not one Consider first the claim that although the function of II is
rathersay that the emphasison stepwise motion here (especially in the bass)
is the means by which the opening progression'slow hierarchicalstatus within primarily subdominant, a weak dominant function never-
the overall harmonic structureis expressed? theless exists. Such a claim would be foreign to a functional
Although I believe in the existence of an essentiallyindependentharmonic
domain, I also believe that a dependency of the vertical dimension on the tween CyclicallyGenerated Chords,"Musikometrika3 (1991): 15-40, where
horizontaldimension indeed exists at a very deep level, deeper, in fact, than triad and seventh chord are defined on the basis of voice-leading consider-
that which Schenker posits. See Eytan Agmon, "LinearTransformationsbe- ations.
 Functional Harmony Revisited 207

theory where dominant and subdominant are seen, in Dahl- Example 2. II as dominant in major: Joseph Haydn, Piano So-
haus's words, as "antithetical extremes."28 Yet the dominant nata in D major, Hob. XVI:37
function of II, I believe, may be felt in certain contexts where
II (or II6) is followed by I (or I6). An example in major, from a) m. 1
the opening of a well-known piano sonata by Haydn, appears Allego con brio

in Example 2a. Interestingly, at the beginning of the devel-

opment section the opening idea is restated (in the dominant
key) with VII6 in the place of II6 (Example 2b). One could
argue, of course, that Haydn's reharmonization, necessitated
by the transfer of the thematic material to the bass, amounts
to a functional reinterpretation (that is, D-T instead of S-T). D--T D--T(?)

b) m. 41
28"Thenotion that ii before I should be taken as V9, and vi before V as
ii9, is absurd, since it leads to a confusion of antitheticalextremes-the sub-
dominant (subdominant parallel) with the dominant, and the tonic (tonic An IL_
parallel) with the dominant of the dominant"(Dahlhaus, Studies, 37). Dahl-
haus is responding(in part) to Sechter's interpretationof II-I(6) as II-V9-I, lI ~ i I T
I FI
which, ironicallyperhaps, is a possible precedent to the idea that II is weakly
dominant in function. See Wason, Viennese Harmonic Theory, 40-42, and
D--T D--T !
Caplin, "Harmonyand Meter," 76 (especially Example 2).
To the best of my knowledge, the closest Riemann ever comes to ac-
knowledging a dominant quality in II is when he identifies (possibly after The VII6, however, is so clearly a variant of the II6 chord,
Hauptmann)the root of II or II7in major with the fifth of V. See for example that to hear the two chords as contrasting in function seems
his Harmony Simplified,trans. H. Bewerung (London: Augener, 1896), 55. counterintuitive.29
To be sure, the theory of Klangvertretung(which Riemann credits to Helm-
holtz) allows Riemann to state that "each of the secondary triads of the key
represent simultaneouslytwo of the three primaryharmonies, one primary 29Alsorelevantin this connection are mm. 5-6, a variedrepetitionof mm.
harmonybeing comprehended as the main content (consonance), the other, 1-2. Replicating the II6 chord would have sounded awkwardwith the new
the foreign addition (dissonance)"(Hugo Riemann, Historyof Music Theory, sixteenth-note figurationin the left hand, and therefore Haydn departs from
vol. 3, trans. and ed. William Mickelsen [Lincoln: University of Nebraska mm. 1-2 harmonicallyby keeping the inner-voicea' stationary,thus implying
Press, 1977], 218). However, except for II in major, the principle of Klang- the succession V4-I6 rather than II6-I6.
vertretungdoes not seem to apply to II and VII. For Riemann, II in minor One could object that in comparingparallelpassagesin the Haydn Sonata
is the major triad 62-4-6 (the Leittonwechselklangof IV) and VII in major I am confusing the question of functional parallelismwith the independent
is the minor triad 1-2-#4 (the Leittonwechselklangof V), while VII in minor question of thematicparallelism.Yet I make no claimswhatsoeverconcerning
is the major triad (t)4-2-4 (the Parallelklangof the natural dominant). In a necessaryconnection between the two kinds of parallelisms.It is important
any event, I am not awarethat Riemannever actuallyanalyzedII as dominant to bear in mind that whereas thematic parallelismin the present case might
(even in major) or VII as subdominant.In particular,a VII triad (diminished, seem to supportthe dominantinterpretationof II, thematicparallelismcould
of course) is invariablyanalyzed by Riemann as a rootless dominant-seventh hardly support, say, a dominant interpretationof IV, for the simple reason
chord. that the latter possibility does not exist within functional theory.
208 MusicTheorySpectrum

An example in minor (from the opening of a Brahms in- in context the dominant potential of the leading-tone triad is
termezzo) appearsin Example 3a. For reasons to be clarified much more likely to be manifest than the weaker, subdom-
shortly in connection with VII, the dominant function of II inant potential.32 Nonetheless, the leading-tone chord can
in minor (a dissonant, diminishedtriad) is even more evident assume subdominantfunction in the form of a seventh chord;
than that of its major counterpart. (It is probably needless this is readily explained, for the seventh contributes to the
to remind the reader that the function of II, in minor as well chord another element in common with IV or II. Various
as major, is first and foremost subdominant.) Note that in contextual factors such as inversion, doubling, and chord suc-
the consequent phrase, which begins in m. 9, the opening II6 cession, can help in this respect, as the excerpts in Example
chord concludes the dominant prolongation which begins in 4 illustrate.
m. 6 (Example 3b).30
With respect to the leading-tone triad, the question of dual V. TWO ADDITIONAL THEORETICAL CONSIDERATIONS
functionalityis even more delicate. A subdominantleading-
tone triad is difficult to find.31Yet the leading-tone seventh Even though the proposed theory promotes an essentially
chord (VII7) is known to assume (in certain contexts) sub- symmetrical functional structure, the previous section sug-
dominant function. In what follows I suggest that although gests that contextual considerationscan modify that symme-
both VII and VII7 possess a weak subdominantpotential, in try to some extent; for example, the subdominant function
context only the latter chord's subdominantpotential is likely of VII is considerably less prominent than is the dominant
to be realized. My explanation involves equally the succes- function of II. The present section considers two additional
sional paradigm T-S-D-T and the special status of VII as factors which interfere with the system's underlying sym-
the only dissonant triad in the diatonic system. It should be metry.
emphasized that the T-S-D-T paradigm is evoked tenta- The effect of roots. The so-called root of any triad or sev-
tively, by way of suggesting how function and chord pro- enth chord is a tone of privileged status. The existence of
gression might interact in a more complete harmonictheory. roots disturbsthe balance among the so-called secondary tri-
VII is a dissonant, diminished triad (in minor, the har- ads of each functional category, since those secondary triads
monic form of the scale is assumed), and as such demands in each category which contain the prototype's root possess
resolution. In addition, the dominantpotential of the leading- an obvious advantageover those which do not. For example,
tone triad is stronger than its subdominantpotential. These VI has an advantage over III within the tonic category, and
two inherent attributesof VII are mutually supportive, since similarly, II over VI within the subdominant category. The
both the resolution of dissonance and the paradigmaticD-T dominant category is an exception, since VII is obviously a
functional succession are strongly goal-oriented. As a result,
32I do not assume that the dissonanttritone contained in the leading-tone
30Notealso that in the intermezzo'scontrastingmiddle section the opening triad must resolve by stepwise contrarymotion, or that the leading tone must
motive is restated (in D, major) with V7-I in the place of II6-I6 (e.g., mm. proceed to i. As arguedin the appendixto this article, to assumethe necessity
22-23). of these linear behaviors is to assume a D-T functional succession. It is not
31See, however, Yizhak Sadai, Harmony in its Systematicand Phenom- difficult to see that the same factors which suppress the weak subdominant
enological Aspects, trans. J. Davis and M. Shlesinger (Jerusalem: Yanetz, potential of VII enhance the (weak) dominant potential of II in minor, as
1980), 149. argued in connection with the Brahms example.
 FunctionalHarmonyRevisited 209

Example 3. II as dominantin minor:JohannesBrahms, Intermezzo,op. 117 no. 2

a) m. 1
Andantenontroppoe con molto mpreione

1bbbbb

p dolce

col v.
D--------T (?)
b) mm. 6-10

D------------------------------------------------------------------------------------------------T

more characteristicdominant triad than III, even though III, The norms of doubling in four-part harmony-e.g., the
and not VII, contains the prototype's root. As discussed norms by which one doubles the third of the secondary triads
above, context often enhances the dominant function of VII.
is the diatonic triad least frequentlyused. With the exception of a few clearly
Moreover, in comparison to VII-I, the progression III-I as defined circumstances,most of its usages are dependent upon context to a
D-T is not particularlyeffective, since III and I are maximally
greater extent than any other diatonic chord." In a similarvein, Smith ("The
similar triads, and thus offer very little contrast.33 Functional Extravagance,"113) states that "the mediant triad is notorious,
especiallyin major, where it has a leading-tone, but sounds perhapstoo much
33Itfollows that III is functionallythe most underprivilegedof all triads: like a tonic to be clearlyheard as a dominantin most contexts. Mediantsrarely
it is not only less of a tonic than VI, but also less of a dominant than VII. have any straightforwardfunction; sometimes they are followed by dominant
This conclusion, no doubt, is empiricallywell grounded. Lester (Harmony, preparationchords, sometimes by dominants, and sometimes by tonics." The
vol. 1, 251), for example, has commented on the "degree of functional am- dominant function of III can be enhanced through inversion and doubling,
biguity inherent in III," noting that it is "greaterthan is the case with any as in the progression III6-I (as discussed below). In minor, the use of #1 is
other diatonic chord in the key system. As a result, [Lester continues,] III highly effective in enhancing the mediant's dominant function.
210 MusicTheory Spectrum

Example4. VII7 as subdominant.After Aldwell and Schachter,Harmonyand VoiceLeading,2nd ed., Exx. 24-24 and 31-15 (pp. 392
and 534)

a) Robert Schumann,no. 5 from Kreisleriana,op. 16, mm. 51-53

[sehr lebhaft] *

S ----------- T

b) Franz Schubert, Moment Musical D. 780 no. 3 (op. 94 no. 3), mm. 34-38
*o modto
Allegro moderatol K I ] r L I

1~-':~f -j f -T
S------------------------------D-----------.---T

II (S), III (D), and VI (T)-are another empirical manifes- Perhaps this helps explain why II7, which has only one tone
tation of the functional significance of the prototype's root. in common with the tonic triad (cf. V7), almost seems to usurp
Given the common belief that these norms are better ac- the role of subdominant prototype in the domain of seventh
counted for purely in terms of voice leading, the issue of chords.
doubling is examined more closely in an appendix to this
article.
VI. CONCLUSION
Seventhchords. The symmetricalstructuredepicted in Fig-
ure 2a is disturbed by adding a seventh to all six non-tonic
"The report of my death was an exaggeration," quipped
triads, as shown in Table 1.34Particularlysuggestive in Table
1 vis-a-vis Figure 2a is that whereas V7, like V, shares only Mark Twain after an American newspaper announced that he
a single tone with the tonic triad, IV7, unlike IV, shares two. was dying in poverty in London.35 Sometimes it seems as

34Thetonic is by nature consonant, and thus would not normally appear 35DeLanceyFerguson, Mark Twain:Man and Legend (New York: Russell
as a seventh chord. See Agmon, "Tonicity and the Tritone." & Russell, 1965), 272-73.
 FunctionalHarmonyRevisited 211

Table 1. The tonic triad versus the six non-tonic seventh tionstheoriefrom a rationalperspective for the most part, late
chords: a common-tone hierarchy in life his general methodological outlook became overtly
mentalistic. It seems only fitting, therefore, to bring the pro-
Non-tonic seventh chord Number of tones in common
with the tonic triad posed, prototype-theoreticaccount of harmonic functions to
a close with one of Riemann's most uncompromisingstate-
VI7 3
ments in this regard:
III7, IV7 2
V7, II7 1 The"AlphaandOmega"of musicalartistryis notfoundin theactual,
VII7 0 soundingmusic, but ratherexistsin the mentalimageof musical
relationshipsthatoccursin thecreativeartist'simagination-amental
imagethatlivesbeforeit is transformed intonotationandre-emerges
in the imaginationof the hearer ... If one has grasped these fun-
though a similarlyunfounded announcementconcerninghar- damental ideas, then it is clear that the inductive method of tone-
monic theory has been posted. To be sure, tonal theory in physiologyandtone-psychology is headedin thewrongdirectionfrom
general has undergone a major reassessment in the past five the verybeginningwhenit takesas itspoint of departure the inves-
or six decades, due primarilyto the work of Heinrich Schen- tigationof theelementsof soundingmusic,insteadof theexamination
ker; however, it is a mistake to believe that harmony is nec- of the elementsof musicas it is imagined.In otherwords:neither
acoustics,nor tone-physiologynor tone-psychologycangivethekey
essarily one of the main casualties of the process. Harmony to theinnermost essenceof music,butratheronlya "Theory
is with us to stay; and with it, so it seems, is the idea of of Tonal
Imagination"-atheorywhich,to be sure,has yet to be postulated,
harmonic functions. muchless developedand completed.36
Harmonicfunction, in the present view, is only one of two
essential ingredients which ultimately should form part of a
theory of tonal harmony; the other ingredient is chord pro-
gression. A theory of chord progression is essential in order
to account for the privileged status of certain root relation- APPENDIX
ships, most notably by descending fifth. Although the ques-
tion of chord progressionlies outside the scope of the present
THE NORMS OF DOUBLING IN FOUR-PART HARMONY: A CRITIQUE
article, some tentative suggestions have been made as to how OF THE VOICE-LEADING ARGUMENT
harmonic theory in general may benefit from the interaction
of such a (hypothetical) theory with the proposed theory of
As is well known, it is customary to double the third of
functions.
Considerable liberty has been taken here with Riemann's II, III, and VI, in their capacity as subdominant, dominant,
and tonic triads, respectively. This finding, as already indi-
Funktionstheorie,not only for the sake of greater theoretical
cated, easily lends itself to a functional explanation. Indeed,
rigor and the removal of arbitraryfeatures, but also for the
sake of placing music-theoretic discourse within what I be- 36RobertWason and Elizabeth West Marvin, "Riemann's'Ideen zu einer
lieve should be its proper intellectual context, namely cog- Lehre von den Tonvorstellungen':An Annotated Translation,"Journal of
nitive science. Although Riemann himself developed Funk- Music Theory 36 (1992): 82-83.
212 MusicTheory Spectrum

Riemann stated very clearly that the doubled third in such Example5. The progressionV7-VI: VI with doubled root
cases is really the root of the primary triad which the sec-
a) with faulty b) without faulty voice lead-
ondary triad, presumably,represents.37Although it might be voice leading ing, not taking functional-
theoretic considerations
preferable to say that the secondary triad represents a func- into account.
tional category and the doubled third is the means by which 00-----O

the category'sprototype is emulated, either way it seems that { :o- 0,

O-V o 0
the norms of doubling in four-part harmony are compelling
empiricalevidence in supportof the functionalapproach.The
purpose of this appendix is to challenge the claim that these V7 VI V7 VI
norms are better explained purely in terms of voice leading.
The appendix thus supplements Part 3 of the article, par-
ticularlythe discussion of whether the Schenkerianapproach
with its emphasison hierarchyand voice leading rendersfunc-
tionalism, and possibly traditional harmonic theory in gen- tually, a diminishedfifth followed by a perfect fifth), as shown
eral, obsolete. in Example 5a.
The essence of the argumentby which characteristicdou- Even though Jeppesen's objectionable fifths may be re-
blings in four-part harmony are attributed to voice-leading moved through a different disposition of the upper voices (as
constraints may be found in a 1949 paper by Knud Jeppe- in Example 5b), many would surely argue that from a voice-
sen.38Jeppesen's point of departure is the discussion of the leading point of view the result remains objectionable since
deceptive cadence V7-VI in Rameau's Traitede l'harmonie. the tritone does not resolve properly, and, in particular,the
Rameau makes a strikingly Riemannian statement that "it leading tone does not move up to 1.
is preferable to place the octave of the third in the chord [of But should one assume a priori that a tritone must resolve
the sixth degree] rather than the octave of the bass . . . be- by contrary stepwise motion, or that in any context 7 must
cause the third implies the true fundamental sound, whose be followed by i? Surely not. In terms of dissonance treat-
replicate cannot be displeasing."39In an attempt to replace ment there are any number of possibilities besides contrary
this (presumably) functional explanation by a voice-leading motion to resolve a tritone (as in Example 6a). There are
one, Jeppesen seizes upon Rameau's subsequent reference even contexts (most notably, II6-V in minor) where the res-
to "gross errors" which may result from doubling the sub- olution of a tritone seems hardlyan importantissue (Example
mediant's root, interpretingthese to mean parallel fifths (ac- 6b). Furthermore,there are many contexts where the leading
tone properly moves down to 8 (Example 7).
Therefore, when one speaks of the tritone's inherent
37Riemann,Harmony Simplified, 71 ff. tendency to resolve by stepwise contrary motion, or of the
38KnudJeppesen, "ZurKritikder klassischenHarmonielehre,"Reportof
the Fourth Congressof the InternationalMusicologicalSociety (Basel, Baren-
leading-tone's inherent tendency to move to i, a particular
context is intended, namely, the functional succession D-T.
reiter, 1949), 23-34.
39Jean-PhilippeRameau, Treatise on Harmony, trans. Philip Gossett
Outside the context of the D-T functional succession these
(New York: Dover, 1971), 73. voice-leading constraints simply do not apply. Thus, in
 FunctionalHarmonyRevisited 213

Example 6. Some resolutionsof a tritone V(7)-VI the sense in which : must go to 1 is contingent upon
conceiving the progression functionally as D-T, that is, as
a)
referable to V(7)-I (the functional prototypes); the voice-
-8 1a leading argument, in other words, tacitly concedes the func-
: L - .0
o-_o i?o- o ,,. I -- II tional point of view.
?? -, 8,~~t;.& (-8
The functionalexplanationof the doubling norm in V7-VI
b) is therefore correct. This does not deny that voice-leading
considerationshere overlap with purely functional ones to a
significantextent (especially in minor). Nonetheless, the dou-
bling norm in V7-VI cannot be explained solely by voice-
leading considerations.
Jeppesen also purports to account in voice-leading terms
II6 V# for the normative doubled third in other instances; he con-
siders in particularthe II6 chord, set in a cadential context.
A cadential progression involving II6, Jeppesen notes, has
a much better tenor line if the third of the II6, rather than
the root, is the doubled tone (compare Example 8a with
Example 7. The leading tone in descendingcontexts Example 8b).
Even though Jeppesen's point here seems rather moot,
one may accept it for the sake of argument, noting the 6-6-
u f J J 6-8 intervallic relationship between soprano and tenor.
Jeppesen, however, ignores an essential question. Why is II
: 0 e characteristicallyemployed in the six-three position in the
r r
I
r r II r-'
I I first place, in major as well as in minor (particularlyat ca-
I III IIV
IV V VII6 III VI6 II ...
dences)? If the tenor line (or the relation soprano-tenor) is
really the main issue here, why not have a root-position II

Example 8. Progressionswith II6 and II

a) b) c) d) e)

Jd & J J J J J
i
rjas
n

F r
.
r r 0 .3
0 0

9 r
_W 43r u F-- r
430
v 4

r P -0

r Pr rr
214 MusicTheory Spectrum

with a doubled root, as in Example 8c? Once again the answer There is no basis for the claim that such an appeal renders
does not concern voice leading. In fact, voice-leading con- the more traditional functional explanation dispensable.
siderationscan even be sacrificedto some extent for the sake
of achieving a doubled third in II6. The voice leading in Ex-
ample 8d, for instance, is inferiorto that in Example 8e (there ABSTRACT
is a hidden octave between the bass and tenor, and the tenor Froma prototype-theoretic pointof viewthe threeharmonicfunc-
leaps unnecessarilyin moving to V). Yet Example 8d (II6with tions known as tonic, subdominant,and dominantare three
doubled third) is preferable to Example 8e (II6 with doubled (partially-overlapping) andthe threeprimarytri-
chord-categories;
root), other things being equal. ads, namelyI, IV, and V, are prototypesof these categories,re-
In summary, an appeal to voice leading does not suffice spectively.A theoryis proposed,whichaccountsforthesecategories
in explaining characteristicdoublings in four-part harmony. and prototypes.