You are on page 1of 11

A Generative Theory of Tonal Music by Fred Lerdahl; Ray Jackendoff Review by: Joseph P.

Swain Journal of the American Musicological Society, Vol. 37, No. 1 (Spring, 1984), pp. 196-205 Published by: University of California Press on behalf of the American Musicological Society Stable URL: http://www.jstor.org/stable/831166 . Accessed: 16/03/2012 20:08
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

University of California Press and American Musicological Society are collaborating with JSTOR to digitize, preserve and extend access to Journal of the American Musicological Society.

http://www.jstor.org

196

JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY

writing make his conclusions appear entirely persuasive, but Walker is not always as careful as he should be with the data, as several examples have demonstrated. Walker'saccomplishment in this work must be applauded nonetheless. His eye for color makes his narrativeof the driest episode enjoyable, and there is no question that he has recreated the period for the modem reader with great flair. He has tried not to approach his subject too loftily; yet sometimes one can become impatient with the almost messianic tone he assumes when he is at pains to refute an apochryphal story. Still this is a small defect when one compares his work to the idolatry or negativism to which Liszt has been subjected previously. Walker could have been less chary with his credits to other scholars in the field,20 but doubtless he will remedy this in future volumes. In sum, this is a major contribution, one that deserves a place on the shelf of every scholar of the period; but in certain respects it must be used with caution.
RENA CHARNIN MUELLER

New YorkUniversity

Fred Lerdahl and Ray Jackendoff. A Generative Theory TonalMusic. of Cambridge, Mass.: MIT Press, 1983. xvi, 368 pp. THEAIM THISBOOK madeclearin its openingsentence: OF is Wetake goalof a theory music beaformal the of to intuitions ofthe description musical in a idiom. ofa listener is experienced musical who The authors'description is hierarchical,so that for the first time, the levels is defined entirely conceptionof music as a system of interrelated according to principles of listening, with an eye toward "the largely unconscious knowledge . . . that the listener brings to his hearing-a knowledgethat eqables him to organize and make coherent the surface of patterns pitch, attack,duration,intensity,timbre,andso forth" 3). For (p. the authors,FredLerdahl,composer,and RayJackendoff, linguist,this goal is a mostambitious to one, even thoughthey limitthemselves thoseintuitions that are hierarchical. The book does fall short in some ways, but the effort has producedrefinedconceptionsof basic musical elements, a significant contribution cognitivescience,and, most importantly, powerfulinvestito a gativemethodfor music theory. Their theory is madeup of four main components:
20 Edward N. Waters, the dean of American Liszt scholars, is not given proper acknowledgment for his discovery in the British Library of Liszt's autograph manuscript of "De la situation des artistes et de leur condition dans la societ6"; see Waters, "Sur la piste de Liszt," Notes, XXVII (1i970-7 i), 668. This very important discovery in part laid to rest the charge by Haraszti that no literary work dating from the I830s was written by Liszt himself.

REVIEWS

197

structure As an initialoverviewwe may say thatgrouping expressesa hierarchical of the piece into motives, phrases,and sections.Metrical structure segmentation expresses the intuition that the events of the piece are related to a regular of alternation strongand weak beatsat a numberof hierarchical levels. Time-span reduction of assignsto the pitchesof the piece a hierarchy "structural importance" with respectto their position in groupingand metricalstructure.Prolongational reduction assignsto the pitches a hierarchythat expressesharmonicand melodic tensionand relaxation, continuityand progression (pp. 8-9). The hierarchicalmodels assigned to each component are rather strict. They concern only contiguous events and are recursive, that is, the analytical processes that generate each layer operate again on the newest one to generate the next. The processes are expressed in two sets of rules for each component, called "well-formedness rules" and "preference rules." All together, they make up a formal description of musical intuition, which, when applied to compositions, generates structural analyses. The title of the book comes from this method, which is best illustrated in the context of one of the components. The grouping structure of the theory begins by stating, in five wellformedness rules, what can qualify as a group: drumbeats,or the likecan constitutea Any contiguoussequenceof pitch-events, group,and only contiguoussequencescan constitutea group. A piece constitutesa group. A groupmay containsmallergroups. If a groupG1 containspartof a groupG2, it must containall of G2. If a group G, containsa smaller group G2, then G, must be exhaustively into partitioned smallergroups(p. 345). Clearly, these ensure the type of hierarchical description that the authors want, but they do not generate acceptable analyses. Certain groupings will be intuitively preferable to others, and these are selected by preference rules. The preference rules are developed by applying Gestaltperception principles to music, in the manner begun by Leonard Meyer,I and by using "experimental examples" designed by the authors to isolate effects of intuitions of the local levels. The rules are then applied in concert to the musical surface in order to determine which of the many possible groupings are preferred by the listener. As theorists have long noted, the Gestalt principles, and thus the preference rules, will not always agree about where grouping boundaries should be drawn. The authors here have eschewed any system of weighting the rules and instead simply resort to judgment about the outcome. In such situations they often note certain musical effects resulting from the conflict between the preference rules, and these are often illustrative. Unfortunately, however, their notation does not permit one to see where conflicts have occurred, and this is true for the other three components as well. In any case, even though the preference rules cannot

and Meyer first applied these principles to a significant extent in his Emotion Meaningin Music (Chicago, 1956).

198

SOCIETY JOURNAL OF THE AMERICANMUSICOLOGICAL

generateanalyseswithout sensitivejudgment,they do providea reasonable for decisions. procedure arrivingat structural The grouping structure, then, is a relatively straightforward interpretation for of the Gestalt principles.There is a distinctpreference binaryorganization in the hierarchy, whichoften reflectsintuitionbut at othertimesdoes not. It rule is surprising that there is no well-formedness regulating numberof the groupsto be combinedinto the groupat the next level, especiallyin view of such organizations.2 evidencefor limits on humancapacityfor perceiving The other componentsare developed similarly, beginning with wellformedness rulesand a characteristic notation. rules, then addingpreference Only groupinganalysis uses as evidence a well establishedpsychological as theory;the othersrely exclusivelyon "experimental examples" basesfor rules. After groupingcomes metricalstructure. preference The originalaspect metricalstructureis that the authorsconceiveof beatsnot as durations.of as pointsin time, stronglyor weaklyaccented,and but in a hierarchical of organized way that is independent grouping: It is theinteraction different of levelsof beats theregular alternation strong of (or andweakbeats a givenlevel)thatproduces sensation meter 68). at the of (p. The authors' effortto presentthis componentas completelyindependent of in groupinghas advantages method, since it is easierto isolatethe musical effectsof a particular elementif it is not complicated otherelements.The by formulationis in most cases convincing, althoughit is doubtful whether metercan be entirelydivorcedfrom grouping,since it is, afterall, derived from surface gestures, which can be created by group boundaries.The authorsunderstatethis relationshipin their discussionof the opening of Beethoven's Hammerklavier Sonata.They explainthat the senseof arrival in measure17 resultsfromthe coincidence a hypermetrical and a highof beat level group boundary(p. 34). Since arrivals,stresses, and other surface featurescreatethe sense of meter, however,it seemsa circularargument to the effect in this way. explain Metricalstructureis the only one of the four componentsthat is not completelyrecursive.The authorsbelieve,rightlyI think, thatthe intuition of meter fades at high levels. The real importance metricalstructurein of their theory is that it combineswith groupingstructureto form the next component,time-spanreduction. Of the four, time-spanreductionis most closely relatedto Schenkerian and analysis,since its aim is to selectcertainpitcheventsas structural others as mereelaborations any given level and to expressthese relationships on in reductional all and notation,assuming its inherentadvantages disadvantages. The authorsbeginwith a lengthyapologyfor reduction, two addressing of its most important drawbacks: first,thatviolenceis doneto the actualmusic,
2 For evidenceaboutsuchlimits,see GeorgeA. Miller,"TheMagical Number psychological Seven, Plus or Minus Two: Some Limits on Our Capacityfor ProcessingInformation,"

Review, LXIII (1956), 82; and Neal F. Johnson, "The Role of Chunking and Psychological Organization in the Process of Recall," The Psychology Learningand Motivation:Advancesin of Research Theory,IV, ed. Gordon H. Bower (New York, I970), I71-247. Concerning the and implications of these limits for music, see Joseph P. Swain, "Limits of Music Structure"(Ph.D. diss., Harvard Univ., 1983).

REVIEWS

199

and second, that exciting, sometimes essential musical events are eliminated at early stages of reduction. In response to the first, the variations in the finale of Beethoven's PastoraleSymphony are cited as clear elaborationsof the

Variations: governingtheme. Then the authorsturn to the Goldberg

Why is the listenerable to recognize,beneaththe seeminglyinfinitevarietyof its musicalsurface,that the ariaand 30 variations all variations one another? are of Why do they not sound like 3i separatepieces?It is becausethe listenerrelates them, more or less unconsciouslyin the process of listening, to an abstract, structure commonto them all (pp. Io5-Io6). simplified This example is not the best since these are variations on a bass line, not a soprano melody, and the melodic progression of the bass in each variation, resorting to a more abstract reduction.

of is alongwith its pattern modulations, easilyheardin eachof them, without

The authors are forced to make a crucial distinction between structural and salience: importance surface In assessingone's intuitions about reductions,it is importantnot to confuse structural with surface salience.. . . Perhaps moststriking the moment importance in the firstmovementof Beethoven's Eroica climaxin Symphonyis the dissonant measures this 276-279. But in termsof structural importance, eventresolvesinto less of (i.e., is less stablethan,andhencestructurally important than)the dominant E minorin measures to 280-283, whichin turnis subordinate the E minorchordat of the beginning measure 284 .... Thus the chordin measures 276-279,despiteits would be deeply subordinate within a reductionof the whole conspicuousness, movement... We do not deprecate auralor analyticimportance salientevents;it is just the of thatreductions designedto captureother,grammatically are morebasicaspectsof musicalintuition(pp. 1i08-109). This is indeed an important admission about a theory whose goal is "a formal description of musical intuitions," since the listener's appreciation of that moment in the Eroica as climactic is certainly one of the most important intuitions about that piece. If the authors argue that this is not a hierarchical intuition, it should be remarkedthat the particulardissonances in that chord are developed throughout the first half of the movement, although their various occurrences may not be related in a strict hierarchy as they define it. If they claim that they are concerned only with "grammaticallymore basic aspects" of intuition, then the theory risks describing intuitions so basic that the analyses produced by them are uninteresting. This danger arises again at the end of the book. While similar, then, to Schenkerism in concept, time-span reduction is significantly different in procedure. Working from the lowest level up, the music is divided into time-spans according to the metrical and grouping structure already assigned. In each span, regardless of level, a single, most important event, the head, is chosen according to preference rules. On the next level up, the resulting heads compete with one another to yield a higherlevel head, and so on, until the tonic chord results as the head of the entire

The secondpoint is Schoenberg's complaintabout Schenkerian analysis, that one'sfavoritemomentshavelittle structural in importance the analysis.

200

JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY

piece. The great breakthrough is the use of grouping and metrical analyses to set the bounds for choosing structurally important events. In this, time-span reduction is intuitively more appealing than a Schenkerian reduction because it is much more consistent in procedure. The preference rules that determine selection of the head seem to conflict more often in this component, so that judgment plays an even more important part. One problem with the selection criteria is their virtually total reliance on rather faded ideas about the relative stability of melodic and harmonic events: Broadly,the relativestabilityof a pitch-eventcan be thoughtof in termsof its relativeconsonance dissonance.For example,a localconsonance morestable or is than a localdissonance,a triadin root positionis morestablethan its inversions, the tonicis the most stableharmony,the relative stabilityof two chordsis a factor of the relativecloseness to the tonic . .. of their roots on the circle of fifths, conjunctlinearconnectionsare more stablethan disjunctones, and so forth(pp.
I17-18).

This is all right as a first approximation, and perhaps for the sake of simplicity this is as much as one can reasonably expect, but there are too many instances of the musical facts contradicting the rules for this formula to suffice. The harmonic system under discussion responds flexibly to the context of a work; at times, even the tonic chord can be unstable.3 Another problem in the preference rules for reduction arises with cadences, because the authors, like Schenkerians, want to retain both harmonic members of the V-I cadence at high levels. In wrestling with the problem, they arrive at the following postulate: We observethatintuitivelya cadencemustbe a cadenceof something; groupthat a consistedonly of the articulation its endingwould be unsatisfying 168). of (p. Although there is no cognitive evidence presented, this is an interesting idea that seems to gain some validity in the authors'analyses, especially that of the Chopin A-Major Prelude, Op. 28, no. 7: For instance,measures1-8 of the samepreludeconsistof two V-I progressions. . . The firstof the V-I progressions, since it completelyoccupiesthe firstfourmeasure group,is not reallyheardas a cadence.Therefore, only the V or the I may be retained the next level. By contrast,the V-I progression measures is a at in 5-8 cadencefor the entireeight-measure shouldbe retainedas a group, and therefore whole in the reduction(p. 168). Unfortunately, this example is somewhat confusing since the last chord in measure eight is a I4, which the authors have characterized as inherently unstable, and is indeed difficult to hear as a cadence. Generally, although the authors' assessment may have empirical support, there are times when the effect of a movement depends upon the perception of an isolated group as a

cadence. Haydn's String Quartet, Op. 33, no. 5, in G major begins with a

' Charles Rosen illustrates this point with a remarkable passage from Beethoven's Fourth Piano Concerto in The Classical Style:Haydn, Mozart, Beethoven (New York, i97i), pp. 387-88.

REVIEWS

201I

fully harmonized cadential gesture that is clearly separate in its grouping. Unless this gesture is understood as a cadence by the listener, however, most of the humor of the movement, deriving from various musical puns using the gesture both as beginning and as ending, would vanish. Prolongational reduction is both the most original and the most problematic component of this generative theory, for the authors intend it to be a formal description of tension and resolution, perhaps the most essential and complex intuition of music. hereis the moreelasticsortwhoseopposite The kindof tensionwe wish to address is relaxation-the incessantbreathingin and out of music in responseto the of juxtaposition pitch and rhythmicfactors.We wish to be ableto speakof points of relativetensionand reposeand the way musicprogresses fromone to the other. This is the functionof prolongational reduction 179). (p. Once more, the procedure for this analysis depends upon previously derived components. The time-span reduction defines "prolongational regions," from which the significant tensing or relaxing events are selected, and the analysis proceeds from the top level down. This is a most attractive aspect of the theory, for it attempts to describe the properties of musical events not only in terms of their pitch construction, as dissonant or consonant, but also in terms of their occurrence in the time-span of the piece, of their place within the high-level rhythm. The close dependence prolongational of reduction time-span on reduction constitutes a majorclaimof our theory. It assertsthat the perceived of patterns tension and relaxation pitch structure in of dependcruciallyon the hierarchy structurally events within time-spansas definedby meter and grouping.In other important of in words,the listener's understanding pitch connections a piece is a functionof how he segmentsits surface.This claimentailsthe unification pitchandrhythm of within one overarching i88). theory(p. As with other components, the authors derive the preference rules from "experimentalexamples" and design a notation to describe the results: We begin by definingtensionand relaxation termsof right and left prolongain
.... motionby a left branch.These arestructural relaxing interpretations placedon the elaborational hierarchies prolongational of trees(p. 181).
8.13

tional branching

A tensing motion will be represented by a right branch, a

V-I

I
(p'onaw

V-I
I duo)

(irme - span ro~c~on)

Althoughthe trees in 8. 13 look similar,they have contrasting interpretations: 8. 13aexpressesthe arcof tonalmotionfromthe piece'sstructural to beginning its

202

SOCIETY JOURNAL OF THE AMERICANMUSICOLOGICAL

8. a of cadence; I 3b saysfirstthatthe pieceforms relaxing prolongation thetonic thatthe opening tensesinto the V, whichin turnrelaxes-more I andsecond than strongly the previous tensing-intothefinalI (p. I89). But when does this relaxingbegin?The V-I progression described is with a left branch,therefore relaxingmotion.Does the musicrelaxwith the onset a of the V? This wouldappear be counterintuitive, to sincethe I is supposed to be the most stable, and the dominantchordis amongthe most tense. If the between V and I is one of relaxing,is that between the two relationship tonicseven morestronglyrelaxingbecauseit occurson a higherlevel?If so, how can the initial tonic be heard as having greater tension than the penultimateV? Finally, how is one to interpret the highest-levelleft Is branching? the entirepiece essentiallya processof relaxing? Notationalproblemsin simple exampleswill compoundin others with many levels because of the recursive quality of the hierarchy. These in uncertainties the authors'notationbecome most troublesome when the issue of musical climax is considered. One would expect that a formal of description tensionwould show clearlythe climacticmomentsof the piece underanalysis.However, the authors'readingsof the prolongational trees produce at least two ways of finding the climactic moment. The first implicatesa shift in a multileveledtree from predominantly right to left branches: Suchjumps prolongational have special in trees a to meaning: point moments they in the musical wherethe "structural flow action" to begins takeplace.. . . More a followed a left-branching will generally, right-branching pattern by pattern in alwayscreatea jump. Such jumpsrepresent majorreversals tensionand from closure 197). toward relaxation, ongoing progression (p. Although it is not explicitly stated that this means musical climax, the authors' Preludewouldindicatethatit does: analysisof the ChopinA-Major The maximal frommeasure i to i pointof tensionin the piece,the transition measure emerges in from to 12, clearly thetreeasa transition right leftbranching (p. 238). Their analysisof the Bach C-MajorPreludefrom Volume I of the WellClavier,however, indicates another source of musical climax, Tempered locatedwell afterthe high-level"jump."
Thus the uncommon circumstance arisesthatthe momentof greatest tension(that is, the most deeply embeddedright branch,measure29) takesplace within the

elaboration of the final cadence (pp. 263-64).

The difficulty of interpreting the notation causes significant problems when the authors take on large-scale forms like the sonata (p. 247):

REVIEWS

203

Sv

V I
theme 2d& development \.

V I roup
2nd therne

V I

itt theme

ift theme

group

group
eposition

grDoup rempttvlttion

According to the criterion of "most deeply embedded right branch," the dotted branch in the reduction (indicating a less favored interpretation) leading from the V preceding the recapitulation would here seem a reasonable approximation of the climax in a sonata, since many pieces can be cited in which this moment is indeed powerful, and critics have often pointed to this transition to the tonic as a crucial gesture of the form. However, the authors prefer the attachment of this chord as a relaxing motion into the recapitulation. Judging by the other criterion of climax, the "jump" from right to left branching, results in nonsense. The greatest tension would be placed in the second theme group of the recapitulation, which contradicts not only intuition of the music but most critical opinion of the last century. Even the motion into the second theme is described as a tensing one on a high level. Only the tree of the exposition seems to mirror the tension patterns of the sonata. What is the cause of these questionable results? One could be the assumption, never stated but clearly present, that preference rules derived from experiments on a local scale will continue to be accurate at high levels of the hierarchy. Since little cognitive evidence other than the "experimental examples" accounts for the preference rules (with the exception of grouping analysis), trusting their validity on high levels is really an act of faith, and one that is perhaps unjustified, as in the case of metrical analysis. Another problem is that the formal description of tension and resolution is just too simple. Because prolongational reduction depends largely upon timespan reduction, the criteria for tension and resolution derive from those of stability and instability discussed earlier, and these are generally harmonic. Reliance on pitch events rather than patterns and on the on-off tree notation, produces an impoverished image of musical tension. First, the binary

204

JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY

patterns of the tree force distortions in choosing between either tensing or relaxing. This is seen even at the highest level, with the interpretationthat a tonal piece is one great relaxation, despite the authors' reference to Edward Cone's image of a ball being thrown and caught. Second, as stated before, the notion of harmonic stability is too constrained, limiting the flexibility and richness of the tonal idiom. Third, there is no mention of tension resulting from the interruption of an established pattern, such as consistent phrase lengths. Fourth, there is no accounting for the speed of tensing and relaxing. Some pieces create tension slowly, others in an instant. Would there be a visible difference in the trees? Fifth, there is no accounting for the differences in tension among styles. Surely intuition about tension is different in a Bach prelude than in a Mozart symphony, yet the trees look much the same in their outline (pp. 259-63). Certainly, it is possible that the incorporation of these aspects would hopelessly complicate the theory, which is already very complex. But the issue of tension and resolution is an immensely complicated one, and in claiming a great deal for this aspect of their theory, the authors may perhaps be underestimating this aspect of music. The book concludes with two essays that use the theory as a point of departure. The first, on musical universals, is an intriguing discussion of which characteristics of music are learned by the listener when becoming experienced, and which ones are innate. This leads to an outline of several important musical characteristics-pitch, collection, tonic, and relative stability-which are basic to musical comprehension and are found in musical idioms the world over. The chapter concludes with some tough and inevitable conclusions about serialism and other twentieth-centurystyles (but I will not spoil the suspense for prospective readers). The second essay explores the parallels between a generative music theory and other theories developed in linguistics and psychology. The authors also demonstrate the utility of their preference-rule method in areas of cognitive science other than music theory. The connections are all very impressive, but the whole matter brings to mind the danger of designing a theory of music too closely upon cognitive principles. The chief advantage in using cognitive principles in music theory is that the postulates thereby gain an independent and scientific source of supporting evidence, and this makes the postulates more convincing, so long as they also make sense in musical terms. However, if the resulting structures derive from the most basic cognitive processes, that is, those that are not specifically concerned with music but with a wide variety of cognitive stimuli, then the usefulness of the theory in addressing interesting musical questions may be limited. Early in the book the authors write: In orderto appreciate poetic or dramatic the of structure a poem in French,one must first understand Frenchlanguage.Similarly,to appreciate Beethoven the a the quartetas art, one must understand idiomof tonalmusic(p. 7). True enough, but one would not study linguistics to grasp the subtleties of Racine, because that discipline would only describe the most basic grammatical relationships of the poetry in a way that would not change even if words, even lines, were changed around. Similarly, a music theory founded on the

REVIEWS

205

most basic principles of cognition may be unable to address the particular concerns of musicians, critics, and listeners because the musical descriptions are too basic to be of interest. Would a generative analysis of a Beethoven quartet differ significantly from one of Brahms or Wagenseil? It should, for intuition tells us that Beethoven is different from Brahms and better than Wagenseil. It is not a question of accuracy, but of focus. It is conceivable that a music theory that reduces out the climactic moments of the Eroicacould be cognitively correct, but such a theory will not help us understand such moments. And unless one values a structural description for its own sake, as an end rather than a means, we must look for something more specifically tailored to musical needs. I am not ready to state that this theory is troubled by this problem, but it could explain one puzzling aspect of the book: the authors do not seem to draw many conclusions from their numerous analyses, at least of the type that would explain the effect of a particular moment or a short piece. There are some exceptions: a fine analysis of a Bach chorale that reveals a double retrograde symmetry in the melody (p. 145) and some enlightening comments on Beethoven sonatas (pp. 253-56). But other analyses are simply left standing without reward for the listener. The results of future work will have to be more effective if the fear mentioned above is to be dispelled. One should not get the impression that, because of reservations about the theory, I do not recommend this book. On the contrary, A Generative Theory of TonalMusicis required reading for theorists and critics. A comprehensive theory such as this will inevitably raise a lot of hackles, especially in its particulars, but there are general features that unquestionably deserve attention and admiration. The authors have largely succeeded in isolating four important characteristics of music, each with its own structural description, and in showing their interrelationships in a reasonable way. Their method of solving this problem, the preference rules in conjunction with well-formedness rules, is a brilliant insight that may well open a new approach to theoretical investigation. These rules are worked out in the text before the reader'seyes, which makes not only for more convincing rules but for fascinating reading. Whatever the fate of the theory as a whole, be it acceptance or revision, the authors' method should stand as a virtuoso piece of thinking and reasoning about music. In turn, the reader'sown thinking is constantly encouraged and rewarded, and there is no better recommendation for a book. PhillipsAcademy,Andover,Massachusetts
JOSEPHP. SWAIN

You might also like