temper ly sync o

8/8/2019 Temper Ly Sync o

1/23


2/23

Popular Music (1999)Volume 18/1. Copyright O1999CambridgeUniversityPress.Printed n the United Kingdom

Syncopation i n r o c k : aperceptual perspectiveDAVID TEMPERLEYIntroductionWhilestudy of the social and culturalaspects of popular musichas been flourishingfor some time, it is only in the last few years that serious efforts have been madeto analyse the music itself: what Allan Moore has called 'the primary text' (1993,p. 1). These efforts include general studies of styles and genres (Moore, 1993;Bowman, 1995);studies of specific aspects of popular styles such as harmony andimprovisation(Winkler1978;Moore1992,1995;Walser 1992),as well as more inten-sive analyses of individual songs (Tagg 1982; Hawkins 1992).In this paper I willinvestigate syncopation, a phenomenon of great importancein many genres ofpopularmusic and particularly n rock.One issue that arises frequently in the studies cited above is the question ofwhether it is valid or useful to apply methods from mainstreammusic theory andanalysisto popular music. In this paper I will make the case that such an approachcan indeed be valuable. However, my concern here is not with well-establishedparadigms such as Schenkeriananalysis and set theory, but rather with a recentlydeveloping branch of music theory, what might be called cognitive music theory.Cognitive music theory is an interdisciplinary ield, lying in the overlap betweenmusic theory and psychology, which brings together the perspectives of the twodisciplines in studying problems of music perception and cognition. In large part,cognitive music theory has been concerned with aspects of music which are oftentaken for granted in mainstreammusic theory, but which closer study has shownto be highly complex, for example, the mental representationof chords and keys(Krumhansl1979, 1990;Lerdahl1988), the detection of meter (Lerdahland Jacken-doff 1983;Parncutt1994),the way melodies are perceived and remembered(Meyer1973;Dowling 1978;Rosner and Meyer 1986;Narmour 1990),and the grouping ofmusical events into segments and phrases (Lerdahland Jackendoff1983; Deliege1987;Palmer and Krumhansl1987). This work parallels work in cognitive psy-chology and cognitive science which investigates the mental processes and rep-resentations underlying other commonplace abilities such as vision and language.Cognitivemusic theory is allied with psychology, also, in that it generally concernsitself with the hearing of a broadpopulation, rather han just trained experts:often,the goal is to describe the hearing of an 'experiencedlistener', namely someonewho is familiar with a genre of music through exposure but not necessarily musi-cally trained. As such, cognitive music theory does not seem open to the criticism,sometimes levelled against theory and analysis by students of popular music, thatit is merely concerned with pieces as autonomous formal objects, divorced fromexperience and culture.l It is true that the kinds of musical structure studied in

19


3/23

20 David Temperleymusic cognition tend to be, in a sense, fairly basic or low-level aspects, and areassumed to be explicable on theirown termswithout referenceto higher levels ofculturalcontext and 'meaning'. But to study lower levels of musical structure nthis way is not to assume that the higher levels are unimportant,any more thanstudies of phonology and syntax deny the importanceof higher-levelunderstand-ing in language.A full understandingof popular music,or any otherkind of music,must surely address both the higher levels of musical experience and the lower-level processesand representations hat supportthem.As in muchmusic theory,my evidence in this paper is simply pieces of music,along with my own intuitions, as an experiencedlistener of the style, about theirstructural(particularlymetrical) implications. In the present case, however, the'pieces of music' at issue are representednot by scores, as in conventionalmusictheory,but by recordings,which I have transcribed.There is no reason why thisshould be problematic;however, it does raise some methodological questions,which I will address later.SyncopationThe OxfordCompaniono Music defines syncopation as 'the displacement of thenormalmusical accent from a strong beat to a weak one'. While I will propose aslightly different understanding of the term, this definition does point to twoimportantaspects of syncopation. First,syncopationinvolves a deviationfrom the'normal'placement of an accent:usually, accentsoccur on strong beats, but in asyncopation,a weak beat (or rather an event on a weak beat) is accented.Second,syncopationinvolves displacement; n a syncopation,an accent that belongs on aparticulartrong beat is shifted or displaced to a weak one (I will suggest that it isactual events, ratherthan accents, that are displaced, but the idea of somethingbeing displacedfromwhere it belongs is essentiallyright). Thedistinctionbetweenthese two points - the accenting of weak beats,and the displacement of an accentfrom one beat to another - is important.Simply understood as the accenting ofweak beats, syncopation is commonplace n many kinds of music, includingclassi-cal music. Two instances are shown in Example 1, both fromBeethovenpiano son-atas: n each case, severalweak beats areaccentedby notes in the righthand. How-ever, it is not at all clear that these passages involve displacement;are the accentsreally heard as belonging on some other beat? In the case of rock, however, thesense of displacement s much more apparent,and there are strong constraintsonthe ways in which this displacementoccurs. In this way, I shall argue, the natureof syncopation in rockis fundamentallydifferentfromthat in classicalmusic.The terms rockand classicalare,of course,vague and problematic. shall notattemptto define them here, but shall rely on their common-sensemeanings. Myconceptionof rockis roughlyindicatedby my choice of musicalexamples,althoughperhaps not everyonewould considerall of these examplesto be rock.The phrase'classicalmusic' is meant in its broadersense: i.e., art music of the eighteenthandnineteenthcenturies,ratherthan merelymusic of the classicalperiod.The clearestevidence of syncopation in rock is found in the setting of text.Consider the melodies in Example2. The firstis fromHandel'sMessiah;he secondand third are from Beatles songs, 'Here Comes The Sun' and 'Let It Be'. It is awell-known fact that text tends to be set to music in a way which matches thestressesof the text with the metricalaccentsor 'strong beats' of the music. In most


4/23

. i o r - r r + Y C s . b._L; b;; L D.^s-.^ lwf^ ,^ . _MS'- ;' 'J2I Hiz A ; r t ; W r S X r ;- s - l - 0 l l x s x l w - l u l Rw X r w | } w X V ^

'9?Xr #C": :8," i #w # "* * * s s

% e # : 2 - : - - ; D r IG zExample2a. Handel, 'For Unto Us a Child is Born',from Messiah

5 0 # # $ f f i $ I Y J :J

t22122


5/23

22 David Temperley* *

* * * * ** * * * * * * * * * ************************

i i i i i i i i i i i i i i i| | l | | | | | | | | | | | |! ! ' ! ! ! ! ! ! ! ! ! ! ! 'ThemightyGod theev- er-lastingath-er thePrinceof PeaceFigure la. 'For Unto Us a Child is Born',metricalstructure

* * * a* * * * * * * a* * * * * * * * * * * * * * * @*******************************@i i i i . i i i i i i i i. . . . . . . . . . . . .. . I . . . . | | : | | :Lit tledar-ling It's been a | cold loneZ Iy winter

Figure lb. 'HereComesTheSun', metricalstructure

* * ** * * * * * *

* * * * * * * * * * * * * * *******************************i i i i i i i i i .' i i i .l i i i. . . . . . . . . . . . . . . .WhenI find my-self in times of trou-ble Moth-erMar- - - ycomes to meFigure 1c. 'Let it Be', metricalstructurestresspatternof the lyricsis also shown in an informalway, with stressedsyllablesemboldened and highly stressed syllables emboldened and underlined. In 'HereComes The Sun', 'been', 'long', 'lone-' and 'win-' are stressed; the syllables inbetween are unstressed.But in the music, each of these syllables falls on a weakmetricalbeat; thus they are all equal in metrical terms. In 'Let It Be', '-self' isstressedwhile 'my-' is unstressed;musically,however, 'my-' is on a strongerbeatthan'-self'.Yetthe Beatlesmelodiesdo not sound anomalousor metricallyambigu-ous; indeed, they are quite typical of the style in their use of rhythm. At firstthought,this might suggest that the rules for text-setting n rockare fundamentallydifferentfromthose in classicalmusic.Notice, however, that if certainnotes in themelodies are shifted over by a quaver or semiquaver,the metricalgrid becomesnicely aligned with the stresspatternof the words. It seems reasonable o suggest,therefore, hat,in the internalrepresentationwe formwhen listeningto rockmusic,we are understandingthe metric grid and the stress pattern as coinciding. Weretain,on principle,the assumptionthat stressedsyllables should occur on strongbeats, but we understandcertainsyllables as 'belonging'on beats other than theones they fall on. I will begin by proposing a simple way of formalisingthis rule,using the frameworkof Lerdahland Jackendoff'sGenerativeTheoryof TonalMusic.I will then use this model to exploresome of the specificways that syncopationisused in rock.2


6/23


7/23


8/23

Syncopationn rock 25over one in which they arenot,but it says nothingabouthow this will be measured.This is in keeping with the spirit of the other rules in GTTM,which are generallystatedin a fairlyinformalfashion.We have also said nothinghere about the highlycomplex matter of stress in language. There is in fact some disagreementabouthow linguistic stress should be represented.One theory proposes a tree structure,representingpairsof syllablesas strongor weak relativeto one another,with eachstrong syllable then forming a strong-weak relationship with another syllable;anothermodel proposes a metricalgrid (similarto Lerdahland Jackendoff'sgridsfor music, except that beats at each level need not be evenly spaced), with morestressedsyllablesbeing markedby higher-levelbeats. (Fora review of work in thisarea,see Dogil 1984).We will not pursue this matterfurtherhere;eitherrepresen-tation would provide the informationnecessary for present purposes, that is, therelativestress levels of syllables in a phrase.A final point: it seems clearthatMPR11 is strongestat fairlylow levels (bothof stressand of metre).At higherlevels, oneoften finds conflictsbetween metre and stress;for example, in the Handel phrasein Example2, 'FATHer' s stressed relative to 'EV',but is metricallyweaker;butthis does not seem wrong or unnatural.Thus MPR11 operatesprimarilyat locallevels.Incorporating rock syncopation into the G5M theoryLet us now apply GTTM's heory of metre (includingthe new rule just added) tothe melodies in Example2. In all three cases, the metricalstructurecorrespondingto the notated metre is clearlyimplied by the accompaniment which is not shownhere). This is accounted for nicely by GTTM'smetrical preferencerules; by thenotated metre's strong beats are well aligned with events ratherthan rests (MPR3), particularlybass note events (MPR6), as well as changes in harmony(MPR5f).In the Handel, the melody appearsto reinforcethe notated metre as well. As dis-cussed earlier (and as MPR11 now stipulates),we prefer to align stressedsyllableswith strong beats;given the notated metre, stressed syllables and strong beats inthe Handelcoincidealmostperfectly.In the Beatlesmelodies,however, the notatedmetre seems quite incompatible with the melody. In some cases stressed andunstressed syllables are equally metricallyweak (as in 'Here Comes The Sun');inother cases, unstressed syllables are actually metrically stronger than adjacentstressedones (as in 'LetIt Be'). (Sinceother syllables in each melody reinforcethenotated metre, such as the phrase 'LlTTle DARling', there is conflict not onlybetween the melody and the accompaniment,but also between the differentsec-tions of the melody.) Thereare conflictsbetween these melodies and the notatedmetre, with regard to other preferencerules as well. For example, MPR 3 wouldprefera structurewhere every event-onsetfalls on a strongbeat; yet in the secondphraseof the first example, 'It'sbeen a long cold lonely winter',exactlythe oppositeis the case. MPR5a prefersstructureswhere longer syllablesfallon metricalaccents;in the second example, however, '-self', 'times' and 'Mar-'are relativelylong butfall on weak beats. In the Handel melody, by contrast,both of these rules are fol-lowed fairlystrictly(as they are in most classicalmusic;I shall returnto this point).Thephenomenonof rocksyncopationwould seem to presenta problem,or atthe very least a question, for GTTMas it stands. For it is a case where conflictsbetween the metrical preferencerules appear to be extremely common, indeed,normative.Thereare several possible ways of explaining this situation.One pos-


9/23


10/23

Syncopationn rock 27Forthe sake of generality,we could propose a 'deep structure' n othertonalmusicas well, except that in classicalmusic the deep structureand surfacestructureareidentical. (In fact, some have proposed deep-structure/surface-structuredistinc-tions for classical music as well, as I discuss below.) How is it to be determinedwhetheror not a particularevent is to be shifted forward, or which level it is to beshifted at? The answer is simple: the solution is chosen which most satisfies themetricalpreferencerules. (All the rules are to be taken into accounthere, not justthe one pertainingto text.) We thereforeeither leave an event where it is, or shiftit forward to a following beat at some level, depending on which resulting deepstructurebetter satisfiesthe preferencerules.Two points of clarification re needed here.By an 'event',I mean a single notein a line;thus theremay be severalsimultaneousevents at a given moment.Thisisan importantpoint; it should be possible for one event to be de-syncopated (thatis, to be shiftedforward to a following beat),while othersimultaneousevents (suchas accompanimentchords)are not. Anotherquestionariseshere:when we shift anevent forward, do we shift the attack-pointforward (thereby making the eventshorter),or do we shift the entireevent forward?This is a ratherdifficultquestion.Shifting only the attack-pointforward might reduce the durationof the event tozero (or it might even make the attack-pointoccurlaterthan the end of the event);on the otherhand, shiftingthe entireevent forwardmight cause it to overlapwiththe following event in the melody. I believe the problem of getting the right dur-ations for events can be solved, but I will not address it here; for now we willsimply adjustthe durationsof events in an ad hocmanner.The idea of regardingmusicalsurfacesas transformations f underlyingstruc-tures is not new. In particular, t brings to mind Schenkerianapproachesto rhyth-mic structure,such as those of ArthurKomar(1971)and CarlSchachter 1980).Inthese approaches,rhythmic nformation s added to Schenkerian eductions,so thatevents in the reductionare given explicit rhythmicvalues. What is importantforpresentpurposes is that this reductiveprocessmay also involve the rhythmicalter-ation of events. For example, syncopated events may be shifted to metricallystronger positions, irregularphrases may be made regular (for example, a 5-barphrasemightbe 'normalised' o 4 bars),and so on. In a sense, the factthatrhythmictransformationshave been proposed elsewhere in music theory might providefurtherjustificationfor proposing them in rock. However, as some authors havenoted (Lester1986,p. 187-9;Lerdahland Jackendoff1983,p. 288), the idea of posit-ing transformations s a generalaspectof rhythmicstructure s in some ways prob-lematic. It is often difficult to decide what the underlying structurefor a passageshould be; particularlyat higher levels, it is not entirelyclear what the motivationis for positingnormalisedversionsof passagesat all. Indeed,reductionsof the kindproposedby Schachterand Komarmight best be regarded,and in fact may well beintended, as ways of enhancingone's hearing of pieces, ratherthan as models forgeneral perception.By contrast,the kind of normalisationthat I propose here isextremely local and quite constrainedin the way it may be applied. I shall arguethat there is quite strongand compellingevidence for the 'psychologicalreality'ofnormalisationsof this kind. Whethernormalisationcan legitimatelybe applied inmore generaland large-scaleways is an importantquestion,but we will not pursueit here.The scheme I am proposing is as follows. In hearinga melody, we first forma 'surface'representationof pitchesand durations;we then forma 'deep structure'


11/23

28 David TemperleySurfacestructure

X # l ted IYJ. - 11Listledarling It'sbeen a long cold one-ly win - ter

* * ** * * * * * ** * * * * * * * * * * * * * * -*****************--------------I I I I / / / / / / / / /Lit-tledarling It's been a ! cold lonf Iy winter

Deep structure

_

Littledar-ling It's been a long cold lonely win - terExample5. 'Here Comes The Sun', showing metricalshiftsrepresentation,which is similar,except that the syncopationshave been 'normali-sed'. It is then this deep structurewhich serves as the input for furtherstructuressuch as metre (and otherkinds of structureas well, as I explainbelow). However,there is feedbackin the system, in that the deep structurewe choose for a melodywill depend on the metricalpreferencerules:we will choose the deep structure hatallows the preferencerules to be maximallysatisfied.Letus apply this model to the phrasefrom 'HereComes The Sun', shown inExample5. This shows the way the deep structureof the phraseis generatedfromthe surfacestructure.The placementof the syllables in the metricaldiagram rep-resents theirposition in the surfacestructure; he lines show how they are shifted(in some cases) to differentpositions in the deep structure.In the first half of thephrase ('Littledarling'),the surface structureof the melody satisfies MPR11 well:the stresslevel of events correspondsnicely to theirmetricalstrength,so no shiftingof events is necessary.Now considerthe syllables'It's-been'.Accordingto MPR 11,stressedsyllablesshould fallon strongerbeatsrather hanunstressedsyllables; hus'been' should fall on a strongerbeat ratherthan 'it's'. If we leave the syllablesexactlyas they are, they have the same metricstrength:both are on beats at level 1(i.e., the lowest level). But now if we shift 'been'forwardby one beat, it now fallson a beat at level 3; now 'been' is stronger than 'It's'. In this case it makes nodifferencewhetherwe shift 'it's' forwardto the following beat (level 2) or leave itwhere it is; eitherway, 'been' is still stronger.By this technique,the syllables canbe shifted forwardsuch a way that every pair of adjacentsyllables satisfiesMPR11, as shown in Example5. As mentioned, if we consider only MPR11 here, it isnot necessaryto shift unstressedsyllables forward. But now recallMPR3, statingthat a metricalstructureis preferredin which event-onsets coincide with strongbeats. This rule favours not only associating 'been' with the following beat


12/23

Syncopationn rock 29Surface structure

l22122DD2229p p G p p ^ p m vWhen I findmyself in times of trouble Mother Mar - y comes to me

* * ** * * * * * *

* * * * * * * * * * * * * * *******************************

When I find my-self in times of trou-ble Moth-erMar - - ycomes to me

Deep structure

4 22 1 2 2 2 D D : pU p pXWhen I find my- self in times of trouSle Mother Mar- y comes to me

Example 6. 'Let it Be', showing metrical shifts

(reinforcingMPR 11), but also 'It's', and every other syllablein the line. This leadsto the deep structureshown in Example 5.'Let t Be' presents a slightly more complicatedexample(see Example6). Here,several events are shifted forward by a beat at the semi-quaver evel, such as 'self'and 'comes'; however, two events, 'Mar-'and 'me', are syncopated at the quaverlevel. This presents no difficulty for the current model. The syncopation shift rulesays only that an event may be shifted forward by one beat at a low metric level;there is nothing to prevent different metric levels being chosen at different pointswithin the same piece or even the same phrase. As in 'Here Comes the Sun', thereare several events whose placement in the surface structure ully satisfies MPR 11,but which are shifted forward so as to satisfy MPR3 and MPR 5a. It may be noted,however, that this was not done in every possible case. The second syllables of'trouble'and of 'mother'would satisfy the rules more if they were shifted forward(since that would place the syllables on stronger beats), but I have left themunchanged: why? My intuition here is that these cases do not represent synco-pations, perhaps because the first syllables of 'trouble' and 'mother' are naturallyshort and therefore fit well with the original rhythm.However, it is not clear thatsuch considerationsshould be taken into account.There s clearly some indetermin-acy here; it is not always obvious what the deep structure or a melody should be.While this is a difficulty, I do not believe it is intractable. t seems that we applythe syncopation rule only to resolve the most severe violations of the preferencerules; other minor conflicts are perhaps left as they are.However, I will not attemptto work out these details here.Now let us examine how the syncopation rule might work with the ungram-matical setting of the 'Here Comes The Sun' lyric, fromExample4 (assuming,again,that the notated metre is communicatedby the accompaniment).Consider just the


13/23


14/23

5b , 2 D p N G 8 H: SThenyou'll be- gi - n to make it bet - ter

Example . TheBeatles,HeyJude'are now equally stressed.Moreover,both syllables now occur on the same beat: aconceptuallytroubling situation (we returnto this point below). Clearly the way toresolve this is by associating'-ter'with the following weakbeat. This is the one casewhere a syllable may be associated with a beat weaker than the one it falls on.This situation is fairly common in rock, but it rarely occurs a number of timesin succession; that is, we rarely encounter a series of syllables in which eachsyllable is shifted forward to the beat of the following one. Thephrase fromMarvinGaye's 'I Heard It Through The Grapevine'shown in Example 9 is an extremeexample.

XglUDDDDLI bet - cha wond- er how I knew

Example . MarvinGaye, I Heard t ThroughTheGrapevine'A phrase from 'Daughter', by Pearl Jam, illustrates a related point (see

ExamplelOa).Here, 'daught-' s stressedrelative to 'me', but falls on a beat of equalmetricstrength;we can resolve this conflictby shifting it forward to the followingbeat at the quaver level. As the rules stand, there is no reason to shift the secondsyllable of 'daughter'.However, shifting the firstsyllable but not the second wouldcause the first syllable of 'daughter'to fall on a beat after the second syllable! Themore naturalsolution would be to shift the second syllable of 'daughter'forwardby a quaveras well, as shown in ExamplelOb. In general,it seems countentuitiveto have situationswhere two events in a melody fall on the same beat in the deepstructure,or where the orderof events in deep structurediffers fromtheirorder insurface structure.We canpreventthis with a simple changeto the syncopation shiftrule:Syncopation Shift Rule (revised). In inferringthe deep structureof a melody, any eventmay be shifted forward by one beat at a low metrical evel. The order of events in a line inthe deep structuremust be the same as theirorder in the surfacestructure.

We mentioned that it is quite possible for events within a piece or phrase toa.

t " l d : r D D l t 9Don't all me daught-er not Elt tob.

i t $ I D D : rDon't call me daught-r not fit to

Syncopationn rock 31

Example 10. Pearl Jam, 'Daughter'


15/23

Xb- :22+ S t :+, S 2 X L 2 j t

Figure . 'Go YourOwn Way',metricalstructure,showing two-stagemetricalshif:tsA9glb WS St

Example2. TheBeatles, 'Day Tripper'

Example3. TheBeatles, 'Hey Bulldog'beyncopatedat differentlevels. This raises a furtherquestion:do we ever find ainglevent that is syncopated at more than one level; for example, at both theuavernd semiquaver evel?Thiswould meanthatthe event was shiftedforwardobeat threesemiquaversafterthe beat it occurredon. Thisphenomenondoes inactccur,but only very occasionally.One exampleis found in the FleetwoodMacong,GoYourOwn Way'.Thisis shown in Example11.Inmy view, in ourhearingfhis phrase,we firstshifteachnote of the phraseforwardto the followingquavereat;henwe furthershift the finalnote of the phraseto the followingcrotchetbeat.Thiswo-stage shifting process is shown in Figure 3.) (Anotherexample is theNirvanaong 'Come as you Are', on the phrase 'No I DON'Thave a GUN';heregain,gun' seems to be syncopatedat both the quaver and crotchetlevels. Suchasesrerare,however,and one feels thatthey are 'stretching he rules'of the style.While all the examplesdiscussed so farhave used vocal music, syncopationsaylso occur in instrumental ines. Two examples of syncopated instrumentalinesre shown in Examples12 and 13, from the Beatlessongs 'Day Tripper'and

32 David Temperley

rGoYou can go your own way

your own way

Example1. 'FleetwoodMac,'GoYourOwnWay'

g * -

* * * * -

.

* * * * * * *

**************< ' ! ! ! !' O O O O O O1 / //',/t' /S / /Youcan go your own way


16/23

Syncopationn rock 33'Hey Bulldog'.The motivationfor consideringthese lines as syncopated is essen-tially the same as in the case of vocal melodies. As they stand, the lines appear tocreatemetricalconflictsbetween certainevents of the melody and the notatedmetre(implied, as usual, by the accompanimentand by other parts of the melodiesthemselves).In 'Day Tripper', or example,the D at the end of bar 1 and the follow-ing F# are on weak beats (and are followed by strongbeats with no event);MPR3would thus favour treatingthese events as metricallystrong.But again, these mel-odies do not seem incompatiblewith the notated metre; the logical explanationisthat the syncopatedbeats areunderstoodas occurringon the following beat.Syncopationsmay also have implicationsfor harmonicstructure.As I havediscussed extensively elsewhere (Temperley1997), the perception of tonal musicinvolves a complex process of harmonic analysis, which entails segmenting themusic into short spans (which I call 'chord-spans')and identifyingthe harmonyofeach one, based on the pitches present in the span and other contextualfactors.Metresis an important actor n the way this processis performed; here is a strongpreference orbeginningand ending chord-spanson strongbeatsof the metre.Sincethe root of a chord-spandepends on the pitches that are present in the span, theexact position of events in time is obviously a factorin the harmonicanalysis.Thisraisesa question,however:in this harmonicanalysis,is it the deep structureor thesurfacestructurethat is used? The formeralternativeseems more likely. Consider'Here Comes The Sun'. The harmonyin the second measure is clearlyA; strictlyspeaking, the F# could not be consideredpart of this chord-span,since it is not achord-toneof A (nor is it a 'legal' ornamentaldissonance).Clearly,the F# shouldbe consideredpartof the following span (with root D);but this would meanbegin-ning a span on the last quaverof a bar, and, as mentioned above, there is usuallya strongtendencyto avoid beginningchord-spanson very weak beats.Whatseemsto be happeningis that, for harmonicpurposesas well as metricalpurposes,the F#is reallyunderstoodas belongingon the followingbeat; f so, then it canbe includedin the following D span in bar 3 without requiringa harmonychange on a weakbeat.Notice that this provides a furthermotivationfor the idea of an unsyncopateddeep structure.Without it, we would have to assume that the basic rules of har-monic analysis, that is, the way chords are derived from pitches, were fundamen-tally different in rock than in classical music. Once a deep structureis assumed,however, the same principleswould appearto apply to both styles fairlyreadily.Of course, this is not to deny that there are also fundamentaldifferencesbetweenrockand classicalmusic in termsof harmony;see, for example,Moore (1992).Anothersourceof evidence forunsyncopateddeep structuresshould be men-tioned, althoughit is somewhatanecdotal.Thisrelatesto the way syncopatedmel-odies are remembered.I have noticed that when people reproducea syncopatedmelody, they often reproduceit with slightly differentsyncopationsfrom the orig-inal. (I have also noticed that, in trying on my own to rememberthe exact synco-pations of a song, even a song that I know extremelywell, sometimesfind it quitedifficultto do so.) This is in contrastto unsyncopatedmelodies, such as Christmascarolsand folk songs, which usually seem to be reproducedfairly accurately.Forinstance,in asking various people to sing the opening phrase of 'Let It Be', I haveheard various differentversions of the second half of the phrase,includingthe twoshown in Example14.Notice, however, that the two versionsof the melody shownin Example14 have the same deep structureas the original.Thisraises the possibil-ity that perhaps what is stored in memory is the deep structureof a melody; per-


17/23

Ex.14.a. b.

8 l^ 6 p 1 1 p S TMoth r Mar ycomes o me Moth-er Mar - y comes o meExample4haps the syncopationsare somehow encoded as alterationsof this melody, but arestored n a less reliable ormthanthe melody itself.Alternatively,perhaps hemelodyis simply stored as a deep structure,alongwith generalinstructions ike 'somewhatsyncopated'; n reproducing he melody, someone would then have to choose theirown syncopations.If it provesto be true that peoplereproducesyncopatedmelodiesina way whichpreserves he deepstructurebutnot thesyncopations, hiswould pro-vide strongevidencefor thepsychologicalrealityof thekind of 'deep structure' haveproposed(this mightbe aninterestingareafor experiment).Syncopation as a musical resourceEarlier arguedagainst the view that the functionof syncopationsin rock is to addmetricaltension or instability.One might ask, then, what functionsdo they serve?This is a very difficultquestion,but I wish to offera few thoughts.Onebasicattrac-tion of syncopationis that it allows for a great variety of surfacerhythms.I men-tioned earlierthat in classicalmusic, certain metricpreferencerules tend to be fol-lowed to a high degree,particularlyMPR3, which prefersevent-onsets(rather hanrests or continuationsof events) to occur on strong beats. (WhenI say the rule isfollowed, I mean that the music is usually writtenso that a metricalstructurecanbe formedfor it which satisfiesthe rule to a high degree.)Of course, the rule is notfollowed absolutely(so thatevent-onsetsarealways on strongbeats and rests neverare), but it is followed to a much greater degree than in rock. As a result of this,many rhythmicpatternsfound in rock simply never occur in classicalmusic. Forexample,I would challengeanyoneto find a melody with anythinglike the rhythmof the first line of 'Let It Be' in classicalmusic. Now, in rock, the preferencerulesmay well apply as strongly as they do in classicalmusic, but they apply only todeep structures; he surface rhythm is allowed considerablefreedom, due to thesyncopationshift rule. The variety of surfacerhythmsin rock is reflected in manyof the melodies cited in this paper; it is often also reflected in the interactionbetween melody and accompaniment ines. ConsiderExample 15, from 'TwoPrin-ces' by the Spin Doctors.The exampleshows the melody on the top staff, the bass

1t"#'' $ : ppW p v ;U LOne two Prin-ceskneelbe- fore you That'swhat said now

1 v "#('= Y yYt t a Y yY

34 David Temperley

Example15. Spin Doctors, 'Two Princes'


18/23

Syncopationn rock 35line on the bottom (the guitar and bass drum follow the rhythmof the bass); thisbasic patternrepeats a number of times. The deep structuresof the two lines arequite simple and indeed rather similar; without any syncopations, the texturewould seem quite dull and square.However, both lines are in fact highly synco-pated, and syncopatedin differentways (essentiallyeach one is syncopatedwherethe other is not), creatinga rich and satisfyingcounterpoint.The variety of rock rhythmis apparentalso in cases where a single melodicline is repeatedwith differentlyrics. Here, it is often the case that the syncopationsof the melody vary slightly fromone repetitionof the line to another(althoughthedeep structure normallystays the same); 'Let It Be' is an example of this. In suchcases, however, I would argue that the goal is not so much variety, but, rather,fittingthe melody to the rhythmof the lyrics in the optimalway. Even if two lineshave the same stress pattern,the most naturalrhythms for them may not be thesame. As linguists have shown, the normalrhythm of speech does not accord allsyllablesequal length; for example,vowels precedingvoiced consonants('bed') arelonger than those precedingunvoiced consonants('bet'),and some consonantclus-ters take longer to say than others (Garman1988,p. 222). Syncopationallows thesurfacerhythmsof a melody to be adjusted to each line of text.The basic model of syncopation proposed above is quite a simple one, andmany of its uses are quite straightforward.However, syncopation also has thepotentialfor more ambitiousand complex uses, and these are sometimesfound inrock as well. One such use occurs in cases of metricalshift from one structuretoanother. If we considerthe surfacerhythmof a melody such as 'Let It Be',we findthatin a numberof casesstressedevents occur threebeatsapart: or example,'FINDmy SELF'.We might think of this as implying a latent triple metre in the song,which may or may not be exploited. In some cases, this latent triple metre isexploited and there is an actualmove to triplemetre; and in cases where this hap-pens, the fact that the triple metrehas been anticipatedby the syncopatedrhythmsof the duple metre section makes the transition smootherthan it might otherwisebe. The Beatles'song 'Two of Us' provides an example (Example16). As usual, thesyncopations n the duple metresectionhere featurestressedsyllables threequaversapart ('SPEND-ingSOME-one's').When the transition to triple metre occurs, itseems rathernaturaland seamless;in part, I think,this is becauseit has been subtlyanticipatedby the syncopated rhythmsof the duple section.4 Threeother elegantexamples of duple-triple shift are found in Led Zeppelin's 'The Ocean', Rush's'EntreNous' and Soundgarden's SpoonMan'.)A further use of syncopation is in situations of metrical ambiguity. Whilemetrical ambiguity is not common in most kinds of rock, there are cases where itarises,and, here, syncopationscan play an importantpart.I would argue that theway metricalambiguityis perceived in rock is ratherdifferentfromotherkinds ofmusic. In Stravinsky,for example,we are not generallyexpectingsyncopationsofthe kind discussed here;that is, we are not expecting stressed events that conflictwith a continuingmetricalpattern(which is not to say that this phenomenonneveroccurs in Stravinsky).Thus we tend to simply realign the metrewith every stressedevent that comes along, as in the famous passage in Example17 from the RiteofSpring.It seems to me that the way one hears this is with a new bar startingonevery accentedchord.5With rock,however,where thereis a norm of stressedeventsoccurringagainst some continuing metrical structure,there is a tendency to hearirregularpassages in this way as well. An example of this is found in the melody


19/23

36 David Temperley

sglStl 11 r ITwof us rid- ingno - where spend-ingome - one's hard eamed

; " --- 1t >2 ' e . 1. >t . t .-1 . t tpay Two f us Sun-daydriv - ing not arr-iv - ing

b"l: r 1r r 1X; ' $ I D;;


20/23

a.

i " D D D : 2 1 R ; D u A J I 0 ;Notlingo do to save his life call hiswife in

b.

J 2 1 2 S g I i JNotSing to do to save his life callhiswife inc.

W J I S J 4 1 I X JNothngo do to savehis life call his wife in


Example18. TheBeatles, 'GoodMorning, GoodMorning'leading to the deep structureshown in Example18b.Or one might go furtherandhearthe syllable 'save'as syncopatedat the crotchet evel, implying the deep struc-ture in Example18c.I believe thatall threeof these hearingsarepartof my percep-tion of the passage. In any case, the main point here is that rockencouragesus totry to keep a metricalstructuregoing even when thereare stressedevents that (onthe surface)conflictwith it; in genuinely irregularpassages such as this one, thiscan createmetricalsituationsof considerablecomplexityand richness.In short,while the basic principlebehind syncopationin rockis fairlysimple,it offersa numberof possibilitiesfor extensionand development;as I have tried toshow, composersand performershave exploitedthese possibilities,therebyaddingto the vitality and richnessof the style.ConclusionsAt firstthought,rocksyncopationsappearto presenta challengeto GTTM's theoryof metre fill. Rock songs feature pervasive metricalconflicts,where the metricalimplicationsof the melodic rhythmand text-settingappearto conflictwith those ofthe accompaniment; et these passages do not seem metricallyunstableor ambigu-ous. I have argued,however, that it is not necessaryto posit a much greaterdegreeof metricalconflictin rock,nor is it necessaryto introducefundamentallydifferentpreferencerules.Rather,rocksyncopationscan be explainedthroughthe introduc-tion of one simple (though ratherfundamental)extensionof GTTM: a 'deep struc-ture' in which syncopationshave been resolved. Once this extension is accepted,the phenomenaof rocksyncopationfit nicely with GTTM, in thatrocksongs gener-ally featurea strong and consistentmetricalstructure n the mannerpredictedbythe theory.The fact that settingsof text which arejudged as metricallyunstablebythe model (such as Example 4) tend to sound 'ungrammatical'can be taken asfurtherconfirmationof the model. Insome ways, it is the deep structureof a melodythat seems more important; t is the direct input to metricaland harmonicpro-cessing,and may also serve as the representationof melodies in long-termmemory.On the otherhand, the surfacestructure s important oo;it contributes o the rhyth-


21/23

38 David Temperleymic interest and variety of melodies, allows for melodic rhythms to be closely tail-ored to each line of text, and plays an importantrole in cases of metricalshift andambiguity.

A few furtherpoints are warrantedabout the methodology and assumptionsof this paper. As mentioned earlier, he currentpaper differs from much traditionalmusic theory, in that the musical objects under investigation are recordings ratherthan scores. This does not, I believe, fundamentallyweaken the conclusions I havedrawn here, but it does lead to some furtherquestions. First,any transcriptionof arock song into conventional notation assumes that it implies an underlying 'score':a representation f quantised notes and durations.This seems a reasonableassump-tion, although given that there is generally no explicit score for rock songs, it isperhaps more debatable in rock than in classical music. Of course, the durationalvalues implied by my transcriptionsare somewhat simplified:the actual durationsmay deviate from these values, sometimes, no doubt, in musically importantways,just as rubatoand other aspects of expressive timing are importantaspects of classi-cal music. In that I am representing he syncopations in my transcriptions, am ineffect treatingthem as part of the 'piece' rather han an aspect of expressive timing;but this assumption might be questioned. In classical music, 'expressive timing'usually implies aspects of timing that are controlled by the performerratherthanthe composer. In rock, however, the composer-performerdistinction is much lessclear-cut han in classicalmusic. Determining he actualnotes of a piece, the implied'score', if there is one, is often a collaborativeeffort of several people, for example,the members of a band; and these individuals are also frequently the primaryper-formersof the song. In this case, the distinctionbetween composition and perform-ance, and the question of where syncopation lies in this scheme, seems ratherarbi-trary. Another, related, question concerns the stability of syncopations overdifferent performancesof a song. I have regarded the (most well-known) recordedversions of these songs as their definitive performances;but one wonders to whatextent the melody of a song is renderedconsistentlyby the singer, or whether thereare subtle variationsbetween performances n the way syncopations are applied. Ifso, it may be misleading, at least from the point of view of production, to regard asingle rendition of, for example, 'Let It Be' as the definitive performance;perhaps itis simply one performancewhich happened to be recorded. (Returning o the pre-vious point this might be a reason for considering syncopation more an aspect ofperformancerather han of composition.)But beyond that, this consideration s notreally problematic or my argumenthere; ndeed, if it turned out that singers tendedto vary the syncopations of a melody while preserving the deep structure, thiswould be furtherevidence for the model I have proposed.A question naturally arises as to whether the model proposed here might beapplicableto other kinds of music. In principle it seems relevant to other genres ofrecentpopular music where syncopation s important, uch as jazz and much Carib-bean and Latin American popular music, although this is an area for furtherresearch. A parallel might also be drawn with traditional African music. Manyrhythmicpatternscan be found in Africanmusic, particularly hose based on 8-beator 16-beat phrases, which seem rather similar to rock rhythms. For example, therhythm of the melody in Example 19 (from Nketia 1974) might easily be found ina rock melody. However, we must be cautious here. There is a broad consensusthat much African music is 'metrical' n the most basic sense, in that it is built onan (often implicit) frameworkof equally spaced beats; t is widely agreed, also, that


22/23

Example 19. African (Kasem) olk melody,from Nketia (1974, p. 153)much of the rhythmic richness of African music lies in the way rhythmic eventsconflict with this framework(Nketia 1974;Chernoff1979;Agawu 1995;Rahn 1996).But this does not necessarily mply thatevents on weaks beats are to be understoodas syncopations,that is, as displaced events which belong on some other beat. Somestudies have pointed to quite differentways of understandingAfrican rhythms; orexample, in terms of additive rather than divisive structures (Nketia 1974), or interms of their set-theoreticaland intervallic properties (Rahn 1996). Agawu sug-gests, furthermore, hat the alignmentbetween word stress and metrical accent, ofcentral importance n the currentmodel, is much less rigid in African music (1995,p. 192). Thorough discussion of this issue is well beyond the scope of this paper;suffice it to say that the superficialsimilarity between some African rhythms androck rhythms does not necessarily imply that the same models are applicable.By focusing on relatively basic and low-level aspects of musical perception, Icertainly do not wish to imply that the cultural meanings of rock music, and itsimplicationsfor issues of politics,economics, gender, race, and so on, are unimport-ant. It goes without saying that a full understandingof rock requiresconsiderationof these social and cultural aspects. Whether the observations about syncopationoffered here would yield any insight into these aspects is an open question. WhatI have tried to show is that low-level aspects of musical perception such as rhythmand metre can be examined and understood on their own terms, and that they areinteresting and complex enough to deserve careful study and attention.

Endnotes


1. For examples of such criticism, see Tagg (1982,pp. 40-3) and Shepherd (1982, pp. 145-9).2. The idea of explaining rock syncopations astransformations of a underlying structure isbriefly discussed by Middleton (1990, pp. 212-13).3. Halle and Lerdahl (1993) offer a proposal forincorporating text-setting into the GTTMtheory. Halle and Lerdahl's model is based onthe same principle as my MPR 11, that is, thatstressed syllables should coincide with strongbeats; however, their model is more complexthan is necessary for our purposes here.4. In fact, this example is somewhat complex.

The latent triple metre of 'SPEND-ing SOME-

one's' is really 6/8 rather han 3/4 (since theaccents are three quavers apart).This is madeexplicit by the guitar line in the first 3/4 bar,which likewise implies 6/8; this, in turn,smooths the way for the 3/4 bars that follow.5. The subsequentbars establishthe notated 2/4metre much more clearly. (This metre is alsosuggested by the previous passage, but onlyvery weakly.)6. The accompanimentgives us little guidancehere. The guitar and bass chords emphasisethe syllables 'Noth', 'do', 'save', and 'in'. Thedrums maintain a regular crotchet pulse,which seems to favour either hearing B or Cover hearing A.

ReferencesAgawu, K. 1995. African Rhythm (Cambridge)Bowman, R. 1995. 'The Stax sound: musicological analysis', Popular Music, 14, pp. 285-320Chernoff, J.M. 1979. African Rhythm and African Sensibility (Chicago)


23/23

40 DavidTemperleyDeliege, I. 1987.'Groupingconditions in listening to music: an approach o Lerdahland Jackendoff'sgrouping preference ules',MusicPerception,, pp. 325-60Dogil, G. 1984.'On the evaluationmeasurefor prosodicphonology', Linguistics,2, pp. 281-311Dowling, W.J.1978. Scaleand contour: wo componentsof a theoryof memoryfor melodies',Psychologi-

cal Review, 5, pp. 341-54Garman,M. 1988.PsycholinguisticsCambridge).Halle, J. and Lerdahl,F. 1993. 'A generative ext-settingmodel', CurrentMusicology,5, pp. 3-23Hawkins,S. 1992.'Prince:harmonicanalysis of AnnaStesia',PopularMusic,11, pp. 325-35Komar,A.J. 1971.Theory f SuspensionsPrinceton,NJ)Krumhansl,C.L. 1979. 'The psychologicalrepresentation f musical pitchin a tonal context',CognitivePsychology,1, pp. 346-741990.Cognitive oundationsf MusicalPitch(New York)Lerdahl,F. 1988.'Tonalpitch space',MusicPerception,, pp. 315-510Lerdahl,F. and Jackendoff,R. 1983.A Generative heory f TonalMusic (Cambridge,MA)Lester,J. 1986.TheRhythms f TonalMusic (Carbondale, L)Meyer, L.B.1973.ExplainingMusic(Berkeley,CA)Middleton,R. 1990. StudyingPopularMusic(MiltonKeynes)Moore, A. 1992.'Patterns f harmony'.PopularMusic,11, pp. 73-1061993.Rock:ThePrimaryText (Buckingham)1995. 'The so-called "flattened eventh" n rock',PopularMusic,14, pp. 185-201Narmour,E. 1990. TheAnalysis ndCognition f BasicMelodic tructures: he mplication-Realizationodel(Chicago)Nketia, J.H.K.1974. TheMusicof Africa New York)Palmer,C. and Krumhansl,C.L. 1987. 'Pitch and temporalcontributions o musical phrase perception:effects of harmony,performanceiming, and familiarity'.Perception Psychophysics,1 pp. 505-18Parncutt,R. 1994.'A perceptualmodel of pulse salienceand metricalaccent n musical rhythms',MusicPerception,1, pp. 409-64Rahn,J. 1996. Turninghe analysisaround:African-derivedhythmsandEurope-derivedmusic theory',BlackMusicResearchournal, 6, pp. 71-89Rosner,B.S. and Meyer, L.B. 1986. 'The perceptualroles of melodic process, contour,and form',MusicPerception,, pp. 1-39Schachter,C. 1980. Rhythmand linearanalysis:durational eduction', n TheMusicForum , ed. F. Salzer(New York)Shepherd,J. 1982.'A theoreticalmodel for the socio-musicological nalysis of popular musics',PopularMusic,2, pp. 145-77Tagg, P. 1982.'Analysingpopular music: theory, method,and practice',Popular.Music,2, pp. 37-68Temperley,D. 1997. 'An algorithm or harmonicanalysis',MusicPerception,5, pp. 31-68Walser, R. 1992. 'Eruptions: eavy metal appropriations f classical virtuosity',PopularMusic,11, pp.263-308Winkler,P. 1978.'Towardsa theoryof pop harmony.' n TheoryOnly,4, pp. 3-26

temper ly sync o

Documents