You are on page 1of 18

Perceiving Emotion in Melody: Interactive Effects of Pitch and Rhythm

Author(s): E. Glenn Schellenberg, Ania M. Krysciak and R. Jane Campbell


Reviewed work(s):
Source: Music Perception: An Interdisciplinary Journal, Vol. 18, No. 2 (Winter, 2000), pp. 155-
171
Published by: University of California Press
Stable URL: http://www.jstor.org/stable/40285907 .
Accessed: 20/12/2012 19:22

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 is collaborating with JSTOR to digitize, preserve and extend access to Music
Perception: An Interdisciplinary Journal.

http://www.jstor.org

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
Music Perception © 2000 by the regents of the university of California
Winter 2000, Vol. 18, No. 2, 155-171 all rights reserved.

PerceivingEmotion in Melody: InteractiveEffectsof


Pitch and Rhythm

E. GLENN SCHELLENBERG
Universityof Toronto

ANIA M. KRYSCIAK
Universityof Windsor

R. JANE CAMPBELL
Universityof Toronto

We examinedeffects of pitch and rhythmon the perceivedemotional


contentof shortmelodies.We initiallydevelopedexemplarsof melodies
thatwerejudgedconsistentlyto conveya singleemotionalcategory:happy,
sad, or scary.We subsequentlymanipulatedthe pitch and rhythmpa-
rametersto derivethreealteredversionsof each exemplar:a pitch-only
version(pitchdifferencesintactbutall tonesof equalduration),a rhythm-
only version(durationaldifferencesintact but all tones of equal pitch),
and a baselineversion(all tones of equal durationand pitch). Listeners
rated how well each exemplarand alteredversionconveyedits corre-
spondingemotion.Effectsof pitch and rhythmvariedacrossmelodies.
In all cases,ratingswere influencedmoreby differencesin pitchthan by
differencesin rhythm.Wheneverrhythmaffectedratings,it interacted
with pitch.
ReceivedJuly22, 1999, acceptedJuly 5, 2000.

the presentinvestigation,we examinedwhether the perceivedemo-


tionalcontentof shortmelodiesvariesas a functionof the melodies'pitch
and rhythmicstructures.1Our primarygoal was to use emotionalratingsof
1. We also examinedthe possibilityof differentialrespondingas a functionof the sex of
the listener.No differenceswere found.

Addresscorrespondenceto E. G. Schellenberg,Departmentof Psychology,Universityof


Toronto at Mississauga, Mississauga, Ontario, Canada L5L 1C6 (e-mail:
g.schellenberg@utoronto.ca)
ISSN:0730-7829. Send requestsfor permissionto reprintto Rights and Permissions,
Universityof CaliforniaPress,2000 CenterSt., Ste. 303, Berkeley,CA 94704-1223.

155

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
156 E. GlennSchellenberg,Ania M. Krysciak,&;R. JaneCampbell

melodiesas a means of contributingto the ongoing debateabout the way


in which pitch and rhythm are processedand representedduring music
listening.At present,there is a definite lack of consensusabout whether
thesetwo factorsareadditiveor interactivein musicperception(e.g.,Jones,
1987, 1993; Palmer& Krumhansl,1987a, 1987b). If they are additive,
listeners'perceptionof pitch relationsin a musicalpiece would not be in-
fluencedby the piece'srhythmicproperties,and, moreover,rhythmicper-
ception would be independentof pitch structure.By contrast,if pitch and
rhythmareinteractive,the rhythmof a piecewould influenceperceptionof
its pitch relations;likewise,pitch structureswould influencelisteners'per-
ception of rhythm."Perception"here should be construedto includerela-
tively low-level representationsof the pitch and durationof tones, as well
as higher-levelrepresentationsof pitch and rhythmicstructures,grouping
processes,expectancies,the emotionalcontent of a musicalpiece, and so
on. As such, the issue of whetherpitch and rhythmare additiveor interac-
tive is centralto the field of music psychology.

Perceiving Emotion in Music

Previousresearchon the emotional expressivenessof music has exam-


ined the influenceof a varietyof musicalparameters,includingpitch and
rhythm.Hevner's(1935, 1936, 1937) studies are seminal in this regard.
When listenersare presentedwith pieces that vary in tempo (fast or slow)
or pitch (high or low) (Hevner, 1936; see also Rigg, 1940; Scherer&
Oshinsky,1977; Wedin, 1972), they select descriptorssuch as "exciting"
and "happy"for fast tempi,and "serene"and "dreamy"for the samepieces
playedat slower templ. "Sad"and "vigorous"are selectedmore often for
low-pitchedpieces, whereas judgmentssuch as "sprightly"and "humor-
ous" aremorecommonfor the samepiecesplayedat a higherpitch.Hevner
(1936; see also Crowder,1985; Wedin, 1972) also found that consonant
harmoniesare judgedto be happy,graceful,and serene,whereasdissonant
harmoniesareconsideredexciting,agitating,andvigorous.Althoughpieces
in the majormode tend to be categorizedas happy,light, sprightly,cheer-
ful, joyous, and gay, the same pieces playedin the minormode are judged
to be pathetic,melancholy,plaintive,yearning,mournful,and sad (Hevner,
1935). These major-happyand minor-sadassociationshave been success-
fully replicated(Crowder,1984, 1985; Scherer& Oshinsky,1977; Wedin,
1972), evenwith youngchildren(Gerardi&cGerken,1995; Gregory,Worral,
& Sarge, 1996; Kastner& Crowder,1990; Kratus, 1993). There is also
some speculationthat particularpitch sequenceshave a one-to-onecorre-
spondencewith specificemotions (Cooke, 1959).
When Hevner (1936) manipulatedthe rhythmof the harmonicaccom-
panimentof variouspieces, she found that "firm"rhythms(i.e., full chord

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 157

on each quarter-notebeat) tended to be perceivedas vigorous and digni-


fied, whereasthe samemelodiesaccompaniedby "flowing"rhythms(bro-
ken chords with successive16th notes) were consideredhappy,graceful,
and dreamy.Indeed,rhythmicpatternswith no pitch variationsvaryin the
degreeto which they are perceivedto be vital, exciting, rigid, or solemn
(seereviewby Gabrielsson,1993). More recentstudiesrevealthat complex
rhythms(i.e., tones of manydifferentdurations)areperceivedto be happy,
whereasjudgmentsof sadness or boredom prevailwhen rhythmis more
regular(Scherer& Oshinsky,1977; Wedin, 1972). By contrast,the influ-
ence of melodiccontour (i.e., risingor fallingmelodic lines) on emotional
expressiveness remainsunclear(Hevner,1936). SchererandOshinsky(1977)
found that upwardcontours are indicativeof fear and surprise,whereas
downwardcontoursareboring,pleasant,or sad. In anotherstudy(Gerardi
& Gerken,1995), adultsbut not 5- or 8-year-oldsfound upwardcontours
to be happierthan downwardcontours.In yet anotherstudy,upwardcon-
tourswere perceivedas relativelyassertivecomparedwith downwardcon-
tours, which were consideredpassive (Gabriel,1978).
Juslin(1997b) presentedhis listenerswith varioussynthesizedrenditions
of a singlemelody (StephenFoster'sNobody Knows). Eachrenditionwas
synthesizedspecificallyto conveya singleemotion (angry,sad, happy,fear-
ful, or tender).Forexample,the happyrenditionhad a relativelyfast tempo
and staccatoarticulation,whereasthe sad renditionwas slow and legato.
Overall, the various renditionswere identifiedwith 82% accuracy.In a
secondexperiment,the samemelodywas manipulatedin a factorialdesign
(with factorsrepresentingtempo, sound level, spectrum,articulation,and
attack), and listenersrated each renditionusing five differentscales (i.e.,
one for each of five emotionalcategoriesfrom the first experiment).Mul-
tiple regressionanalysisrevealedthat ratingsfor each emotion were well
explainedby the five manipulatedvariables,which accountedfor between
55% and 88% of the variancein the ratings.Whereastempo was the best
predictorof happiness,sadness,and tenderness,spectrumand articulation
were the best predictorsof angryand fearfulratings,respectively.In sum,
the literaturemakesit clearthat the emotionalstatusof a piece of musicis
determinedby multiplefactors,of which pitch and rhythmare two.

Pitch and Rhythm: Additive or Interactive?

We now turn to studies that examinedwhether pitch and rhythmare


additiveor interactivein musicperception.Palmerand Krumhansl(1987a,
1987b) examinedthe influenceof pitch and rhythmon listeners'ratingsof
how complete a musical phrase sounds. They reportedthat an additive
model providedthe best account of the data. For example,relativelylong
tones at the end of a phrasewere ratedas soundingcomplete,as were tones

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
158 E. GlennSchellenberg,Ania M. Krysciak,ôc R. JaneCampbell

that had a stable pitch (e.g., dob) in the scale of the phrase.Phrasesthat
ended with a long and stable tone soundedparticularlycomplete, but no
more so (or less so) than one would predictby simply addingthe relative
contributionsof durationand stability.
In a relatedstudy (Monahan& Carterette,1985), listenersheardseveral
pairsof melodiesand madea similarityratingfor eachpair.As attendingto
pitchor rhythmincreased,attentiondirectedtowardthe otherfactortended
to decrease.In other words, limited attentionalresourcesappearedto be
allocatedin an additivemannerbased on the pitch and rhythmicinforma-
tion availablein the melodies. Similarly,when Pitt and Monahan (1987)
requiredlistenersto judge the similarityof pairs of tone sequenceswith
differentpolyrhythms,differencesin pitch betweenstandardand compari-
son sequencesreducedperceivedsimilaritiesbut there was no interaction
betweenthe pitch and rhythmmanipulations.
Otherresearchersemphasizethe integrationof pitch and rhythm,which
is consistentwith an interactiveview of musicperception(Cooper&cMeyer,
1960; Jones, 1987, 1993; Lerdahl& Jackendoff,1983). Becauseimpor-
tant pitches often occur at points of rhythmicimportance,listenersmay
perceivethese events as unified instead of perceivingpitch independently
fromrhythm.InJones'(1987, 1993) concepto( joint accentstructure,pitch
and rhythmicpatternsjointly determinea higher-orderaccent structure
(see also Monahan, 1993). Pitch accents (e.g., changes in pitch contour,
largeintervals,stabletones) aresaidto interactwith rhythmicaccents(e.g.,
longer tones, tones afterpauses)in their influenceon listeners'perception
of melodies. Jones and her colleagues (Jones & Ralston, 1991; Jones,
Summerell,& Marshburn,1987; Kidd, Boltz, & Jones, 1984) tested this
hypothesisby examininglisteners'ability to recognizea previouslyheard
melody.Listenersfound it difficultto ignorerhythmvariationswhen asked
to distinguish"target"from "decoy"melodieson the basis of pitch. Simi-
larly,pitch changes that occur at points of rhythmicemphasisare more
easilyidentifiedthan changesat otherpoints (Jones,Boltz, & Kidd, 1982).
Severalother studieshave also reportedthat rhythmicvariationsaffectthe
perceptionof pitch (e.g., Boltz, 1989a, 1989b, 1991, 1993; Deutsch,1980;
Schmuckler& Boltz, 1994). Conversely,variationsin pitch can affect the
perceptionof tone durations(Boltz, 1992) and the perceiveddurationof a
silent period betweentones (Crowder& Neath, 1995).
In sum, effectsof pitch and rhythmappearto be additivein some experi-
mentalcontexts but interactivein the majorityof cases. This apparentdis-
crepancycould be methodologicalin origin. One possibilityis that pitch
andrhythmareindependentin tasksinvolvingsubjectiveratings(Monahan
& Carterette,1985; Palmer& Krumhansl,1987a, 1987b;Pitt& Monahan,
1987), but interactivein more objectivetasks, such as those that measure
recall or recognition(e.g., Boltz, 1991; Deutsch, 1980; Jones et al., 1987;
but see Monahan,Kendall,& Carterette,1987).

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 159

Althoughpitch and rhythmusuallyvarysimultaneouslyin realpiecesof


music,only two studiesof perceivedemotion in musicexaminedhow mu-
sical parameterswork jointly in determiningemotionsexpressedby music
and whethersuch effectsare additiveor interactive.Schererand Oshinsky
(1977) reportedan interactionbetweentempo and pitch patterns,but they
did not examine the influenceof "rhythmproper" (i.e., durationalpat-
terns;Jones, 1993, p. 73). Juslin (1997b) found no interactionsbetween
the factorshe manipulatedbut these did not includepitch or rhythm.

The Present Study

Weexaminedhow listenersperceivethe emotionalcontentof briefmelo-


dies as a functionof their pitch and rhythmicproperties,with a focus on
whetherthese two factors interact.Our experimentis the first to explore
the natureof pitch and rhythmprocessingin the context of listeners'per-
ception of the emotions expressed by music. Our first objective was to
form a set of shortmelodiclines that unequivocallyexpressedone of three
emotions:happy, sad, or scary. Our motivation for selectingthese three
particularemotions was twofold: (1) each is consideredto be a "basic"
emotion (Ekman,1992), and (2) adults from a common culturegenerally
show broad agreementwhen associating such emotions with particular
piecesof music (Hevner,1935, 1936, 1937; Kratus,1993; Sloboda, 1991;
Terwogt & Van Grinsven,1991; Thompson & Robitaille, 1992), as do
young children(Cunningham&cSterling,1988; Dolgin 8c Adelson, 1990;
Giomo, 1993; Kastner&cCrowder,1990; Kratus,1993; Terwogt& Van
Grinsven,1991). We reasoned,then, that it should be relativelystraight-
forwardto find melodiesthat would be emotionalexemplars(e.g., Brooks,
1978), or specificexamplesof unambiguouslyhappy,sad, or scary melo-
dies.
We subsequentlyexaminedwhetherthe emotionalstatusof the melodic
exemplarsvariedas a functionof the melodies'pitch and rhythmicstruc-
tures. The exemplarswere manipulatedto derive three new versions, al-
teredsuchthat: (1) all tones were identicalin pitch, (2) all tones were iden-
tical in duration,or (3) all tones were identicalin pitch and duration.Our
stimulus manipulationwas identical to that of Palmer and Krumhansl
(1987a, Experiment1; 1987b, Experiment1; see also Hébert & Peretz,
1997; White, 1960), except for our inclusionof a "baseline"condition (all
tones of equal pitch and duration).The listeners'task was to rate the de-
-
gree to which each tone sequence- in originaland alteredforms evoked
its correspondingemotion. Based on our review of the literature,which
revealedthat reportsof interactiveeffectsfar outnumberedreportsof addi-
tive effects, we predictedthat pitch and rhythm would interact in their
influenceon ratings.Nonetheless,our methodprovidedfor a conservative

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
160 E. GlennSchellenberg,Ania M. Krysciak,Se R. JaneCampbell

test of this hypothesis.Specifically,we used a task involvingsubjectiverat-


ings, and our stimulusmanipulationmirroredthat of studies (Palmer&
Krumhansl,1987a, 1987b) in which pitch and rhythmwere found to be
independent.

Method

PARTICIPANTS

The listenerswere 57 undergraduatesin psychologycourses (primarilyyoung adults)


who receivedpartialcourse credit for participatingin the study.Twenty-sevenwere as-
signedto a preliminarystimulus-selectiontask;the remaining30 participatedin the actual
experiment.The listenerswere recruitedwithout regardto musicaltraining.All of the lis-
tenersspecifiedthat theywereexposedto Westerntonal musicthroughouttheirchildhood,
althoughone was also exposedregularlyto musicfrom China.

APPARATUS

Stimuliwere initiallycreatedas musicalinstrumentdigitalinterface(MIDI)files using


Cubasesequencingsoftwareinstalledon a PowerMacintosh7100/66 AVcomputer.MIDI
fileswereoutputthrougha MIDIinterface(Markof theUnicornMIDIExpress)to a Roland
JV-90synthesizerand digitallyre-recordedback onto the same computeras 16-bit digital
soundfiles usingSoundEdit16 software.A customizedprogramcreatedwith PsyScope1.1
software(J.D. Cohen,MacWhinney,Flatt,& Provost,1993) was usedto controlpresenta-
tion of the sound files and to recordlisteners'responses.The monauralstimuliwere pre-
sentedbinaurallythroughlightweightpersonalstereoheadphones(SonyCD550). Listeners
used a mouse connectedto the computerto initiatetrials and to recordtheir responses.
They were tested individuallyin a sound attenuatingbooth (EckelIndustries)and viewed
the computermonitorthrougha window in the booth.

STIMULUS MATERIALS

Our initial set of stimulicontained30 short melodies, 10 for each of the three target
emotions(happy,sad, and scary).Eachmelodywas less than 30 s in durationandconsisted
of four completemeasures.Some of the melodieswere excerptstaken from folk-songcol-
lections that were unlikelyto be familiarto the listeners(e.g., EasternEuropeancollec-
tions). Othermelodieswere used previouslyby Dolgin and Adelson (1990) or composed
specificallyfor the presentexperiment.Althoughtones in the stimulusmelodiesvariedin
pitch and duration,amplitudeswere held constantfor all tones. Thus, "rhythm"was op-
erationallydefinedsimplyas variationsin tone durations,which is consistentwith Jones'
(1993) definitionof "rhythmproper."For each of the threeemotionalcategories,a timbre
was selectedfrom the set of factory-installedtimbreson the RolandJV-90,with experi-
menters'judgmentsdeterminingwhich was the most evocativeof the correspondingemo-
tion. The chosen timbreswere a classicalguitar (UserA32) for happy excerpts,a violin
section(W-ExpA24) for sad excerpts,and an organ(PresetC41) for scaryexcerpts.Careful
selectionof the melodiesand timbresensuredthat each stimulushad the structuralproper-
ties characteristicof its specifictargetemotion(i.e., the happymelodieswerequickin tempo
with staccatoarticulation,the sad melodieswere slow and legato, and the scarymelodies
had a broadpitch range).
Twentylistenerswere assignedto a task designedto find two exemplarsfor each of the
threetargetemotions.The 30 stimulusmelodieswerepresentedin a differentrandomorder

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
Perceiving Emotion in Melody 161

for each listener. Listeners made a single emotionality rating for each of the melodies. For
the 10 melodies from the happy set, they rated how "happy" the melody sounded on a scale
from 1 (not at all happy) to 7 (extremely happy). Likewise, they made sadness ratings for
the 10 sad melodies and scary ratings for the 10 scary melodies. For each of the emotional
categories, the two melodies with the highest average ratings were selected as the exem-
plars. The melodies are illustrated in musical notation in Figure 1. To ensure that the se-
lected melodies were unequivocally indicative of their respective emotions without any over-
lap, an additional seven listeners heard the six exemplars in random order and were asked
which of the three labels (happy, sad, or scary) they would assign to each melody. They were
informed that more than one label could be selected. Interrater agreement on the appropri-
ate emotional label proved to be 100%, with complete consensus that each exemplar con-
veyed its corresponding emotion and no others. Although such agreement does not guaran-
tee that our target emotions were the "best" descriptors of the selected melodies, it confirms
that our target emotions and our stimulus melodies had a one-to-one correspondence.
The exemplars were subsequently manipulated to produce three altered versions. Thus,
there were eight stimulus sequences (4 versions of each of the 2 exemplars) for each of the
three emotions, for a total of 24 melodies. In the pitch and rhythm condition, the melodies
were simply the exemplars with pitch and duration values corresponding to those specified
in Figure 1. In the pitch-only condition, the pitch of the component tones of the melody was
the same as in the exemplars, but individual tone durations were altered such that all tones
were of equal duration yet the overall duration of the exemplar remained unchanged (e.g.,

Fig. 1. The stimulus exemplars in musical notation.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
162 E. Glenn Schellenberg, Ania M. Krysciak, & R. Jane Campbell

if an exemplar had 20 component tones and a total duration of 15 s, each tone in the pitch-
only condition would be 0.75 s). In the rhythm-only condition, tone durations were the
same as those of the exemplars but pitches were altered so that all tones had the same pitch
(i.e., the median pitch of the exemplar). Finally, in the baseline condition, component tones
had uniform pitch and duration. Each of the modified excerpts had the same overall dura-
tion and timbre of the exemplars, as well as the same constant amplitudes. Thus, the condi-
tions formed a 2 x 2 factorial design, with two levels of pitch (varying or equal) and two
levels of rhythm (varying or equal), with all other factors (i.e., timbre, overall length, ampli-
tude, tempo) held constant. Because the baseline conditions were monotonie with a regular
rhythm, they provided a baseline measure of the perceived emotionality of the selected
timbres.

PROCEDURE

Each of 30 listeners (15 men, 15 women) heard all 24 stimulus sequences blocked ac-
cording to emotional category. The order of the blocks was randomized separately for each
listener, as were the stimuli within blocks. Each emotional category was announced imme-
diately before the corresponding eight trials. After each trial, listeners were asked to rate
how happy/sad/scary the musical sequence was on a scale from 1 (not at all happy/sad/
scary) to 7 (extremely happy/sad/scary). Listeners made a single "happy" rating for each of
the eight happy stimuli, a "sad" rating for the sad stimuli, and a "scary" rating for the scary
stimuli. In order to make ratings as normally distributed as possible, listeners were urged to
use the full range of the scale, reserving ratings of 1 and 7 for extreme cases.

Results

HAPPY MELODIES

Mean ratings from each of the four versions of Happy 1 and Happy 2
are illustrated in Figure 2. An analysis of variance (ANOVA) with three
repeated measures (pitch, rhythm, and exemplar) revealed a significant three-
way interaction, F(l, 29) = 16.60, p < .001, which rendered the other main

Fig. 2. Mean ratings as a function of the pitch and rhythm manipulations for Happy 1 and
Happy 2. Error bars are standard errors.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 163

effects and two-way interactions uninterpretable because response patterns


varied across the two exemplars. Accordingly, both exemplars were ana-
lyzed separately by using two-way (pitch and rhythm) ANOVAs. The analy-
sis for Happy 1 uncovered highly significant main effects of pitch, F(l, 29)
= 160.75, p < .001, and of rhythm, F(l, 29) = 9.78, p = .004. Happiness
ratings were higher for excerpts with pitch changes than for excerpts with
pitch held constant and for excerpts with rhythmic changes than for ex-
cerpts in which tone durations were held constant. As predicted, the two-
way interaction between pitch and rhythm was also significant, F(l, 29) =
4.72, p = .038, which motivated closer inspection of the main effects. Fol-
low-up tests of simple effects revealed that rhythm made a significant con-
tribution to happiness ratings when pitch varied £(29) = 5.06, p < .001, but
not when pitch was held constant.
An identical two-way ANOVA for Happy 2 also revealed highly signifi-
cant main effects of pitch, F(l, 29) = 173.60, p < .001, and of rhythm, F(l,
29) = 52.48, p < .001. Happiness ratings were higher for excerpts with
pitch changes than for excerpts without pitch changes and for excerpts
with rhythmic variation than for excerpts without such variation. Although
the interaction between pitch and rhythm was statistically reliable, F(l, 29)
= 12.56, p = .001, it differed markedly from the interaction for Happy 1.
Specifically, rhythm made a significant contribution regardless of whether
pitches varied or not, and the effect was larger when pitches were held con-
stant, *(29) = 7.03, p < .001, than when pitches varied, t(29) = 3.10, p = .004.

SAD MELODIES

Mean ratings from the four testing conditions for Sad 1 and Sad 2 are
illustrated in Figure 3. An ANOVA revealed that the main effect of the

Fig.3. Meanratingsas a functionof the pitchand rhythmmanipulationsfor Sad 1 and Sad


2. Errorbarsare standarderrors.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
164 E. Glenn Schellenberg, Ania M. Krysciak, & R. Jane Campbell

exemplar variable was nonsignificant and that it did not interact with the
pitch or rhythm variables. In other words, patterns of responding were
similar across the two sad exemplars. A significant main effect of pitch
revealed that sadness ratings were higher for excerpts with pitch changes
than for excerpts without pitch changes, F(l, 29) = 37.37, p < .001. All
other main effects and interactions were nonsignificant. In short, rhythm
had no effect on listeners' ratings. Indeed, Figure 3 shows that ratings were
virtually identical for excerpts with and without variation in tone dura-
tions. Thus, for sad exemplars, only the pitch manipulation provided a
reliable explanation of the variation in emotionality ratings.

SCARYMELODIES

Mean ratings from each of the four conditions for both of the scary
exemplars are illustrated in Figure 4. An ANOVA confirmed that response
patterns were similar across the two exemplars (no main effect of exem-
plar, no interactions involving exemplar). The main effect of pitch was sig-
nificant, F(l, 29) = 109.07, p < .001, as were the main effect of rhythm,
F(l, 29) = 6.13, p = .019, and the interaction between pitch and rhythm,
F(l, 29) = 8.26, p = .008. In general, scariness ratings were higher for
excerpts with pitch changes than for excerpts without pitch changes and
for excerpts with rhythm changes than for other excerpts. Nonetheless, the
interaction between pitch and rhythm necessitated clarification of the main
effects. Follow-up tests of simple effects revealed that scariness ratings were
higher for the two pitch-and-rhythm conditions than for the pitch-only
conditions, F(l, 29) = 16.26, p < .001, and that the rhythm-only and baseline
conditions did not differ. In other words, rhythmic variation affected lis-
teners' ratings only when differences in pitch were also present in the ex-
cerpts.

Fig. 4. Mean ratings as a function of the pitch and rhythm manipulations for Scary 1 and
Scary 2. Error bars are standard errors.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 165

Discussion

The presentstudyexaminedthe effectsof pitch and rhythmon listeners'


ratingsof the emotionalcontentof shortmelodies.For all of the melodies,
differencesin pitch madea strongcontributionto theirperceivedemotion-
ality. By contrast,the influenceof tone durations(rhythm)was relatively
weak, evident for some of the melodies but not for others. More impor-
tantly,wheneverrhythmhad an effect, it interactedwith pitch. Finally,the
natureof the interactionbetweenpitch and rhythmvariedacrossstimulus
melodies.Thus, the joint influenceof pitch and rhythmon the perceived
emotionalcontent of melodiesappearsto vary across musicalcontexts.
Similaritiesbetweenour experimentand those of Palmerand Krumhansl
(1987a, 1987b)- in the experimentaltask and in the stimulusmanipula-
tions- leadone to askwhy we foundinteractiveeffectsof pitchand rhythm
when such effectswere found to be additivepreviously.Hébertand Peretz
(1997) and Thompson (1994) suggestthat melody and rhythmare inde-
pendentin the initiallow-levelstagesof musicprocessingbut interactivein
later, higher-levelstages. Although our ratings of emotional content in-
volved a relatively"high"(postperceptual)level of analysis,one could ar-
gue that ratingsof completion(Palmerand Krumhansl,1987a, 1987b) are
similarlyhigh level. Nonetheless, Palmerand Krumhansl'slistenersmay
have focused almost exclusivelyon a single tone (i.e., the last tone of the
stimuli)when makingtheircompletionjudgments,whichwould maketheir
task relativelylow level in comparisonto ours.
Anotherpossibilityis that processingof emotional valence in music is
relativelyindependentof other forms of music processing.For example,
exposure effects differ for recognition and liking judgments (Peretz,
Gaudreau,& Bonnel, 1998). Studiesof a brain-damagedpatientrevealed
that she could identify "happy"and "sad" music in a mannerthat was
indistinguishablefrom normal(control)listeners,yet she found it difficult
to discriminateone musicalpiece from anotherpiece (e.g., two composi-
tions by Mozart)or to identifygrossdeviationsin pitch and rhythm(Peretz
& Gagnon,1999;Peretz,Gagnon,& Bouchard,1998). Zajonc(1980, 1984)
proposes that feeling (emotion) and thinking (cognition)are psychologi-
cally independenteven in nonmusicalcontexts. WhereasZajone's theory
concernsparticipants'actual emotions or preferences,the patient studied
by Peretzand her colleaguesmadejudgmentsof emotionalexpressiveness.
Nonetheless, this "emotivist"-"cognitivist"distinction (Kivy,1990) may
be somewhat Listeners'
artificial. subjectiveevaluations suggestthat emo-
tional respondingto "real"musical stimuli is reliable and often intense
(Goldstein,1980; Panksepp,1995; Sloboda, 1991, 1992). Suchresponses
also vary with the mood of the music (Krumhansl,1997; Stratton &

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
166 E. GlennSchellenberg,Ania M. Krysciak,& R. JaneCampbell

Zalanowski,1991), and objectivephysiologicalresponsesvaryaccordingly


(Krumhansl,1997).
The relative importanceof pitch comparedwith rhythm implies that
listenersextract more emotional meaningfrom pitch patternsthan from
rhythm patterns (at least for our target emotions), despite the fact that
rhythmcan be moresalientthan pitch (Monahan,1993) and moreinfluen-
tial in melody recognition(Boltz, 1993). When Hébert and Peretz(1997)
tested listeners'long-termmemoryfor melodies,however,they also found
that pitch variationswere a bettercue than rhythmvariations.The relative
importanceof pitch over rhythmcould stem from listeners'exposure to
Westernmusic. As noted, melodiesthat are composed in majoror minor
keys typicallyconveypositiveor negativeemotions,respectively,and these
associations become strongerover development(Crowder,1984, 1985;
Gerardi& Gerken,1995; Gregoryet al., 1996; Hevner,1935; Kratus,1993;
Kastner& Crowder,1990; Scherer& Oshinsky,1977;Wedin,1972). Hence,
Westernlistenerslearnover time that differentpitch structuresare indica-
tive of differentemotions (Gerardi& Gerken,1995; Gregory&cVarney,
1996; Gregoryet al., 1996). Theselearnedassociationsmaygeneralizeacross
listening contexts, such that listenersattend disproportionatelyto pitch
when consideringemotionalvalence.Moreover,simpleWesternmelodies,
such as those used in this study,often implyharmonies(simultaneouscom-
binationsof tones).In otherwords,becauseWesternmusiccontainsa highly
developed system of harmony,adult listenersabstracthigher-orderpitch
relationsbasedon simultaneouscombinationsof tones from simplemono-
phonic sequences (Holleran,Jones, &cButler,1995; Trainor& Trehub,
1992, 1994). Evenchildrenas young as 7 years of age demonstratesensi-
tivity to such "impliedharmonies"(Trainor&cTrehub,1994). Listeners
from other cultures (e.g., native African cultures),in which harmonyis
relativelyless developedand rhythmmore developed (Nettl, 1986), may
learnto attendmoreto the rhythmicratherthan the pitchcomponentsof a
melody.
Althoughour manipulationsinvolvingrhythmmade a significantcon-
tributionto the ratingsof the emotionalcontent of happyand scarymelo-
dies, they failed to do so for sad melodies. In a relatedstudy, deviations
fromexact timingin musicalperformancehad little effecton the perceived
sadnessof melodies,but they influencedhappyand scaryjudgments(Juslin
& Madison, 1999). In the present study, equalizingtone durationsmay
have given the sad exemplarsa special "funereal"qualitythat made them
seemas sad as they were in theiroriginalform.The happyand scaryexem-
plars also tended to have more notes per unit of time than the sad exem-
plars had (see Figure 1) and, hence, a faster tempo. Although one could
arguethat effects of rhythmmight be more apparentwith "faster"melo-
dies, a slow tempo is one of the distinguishingfeaturesof a sad melody

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 167

(e.g.,Cunningham& Sterling,1988;Dolgin& Adelson,1990;Juslin,1997b;


Scherer&cOshinsky,1977). Thus,the variousfactorsthatdeterminewhether
a melody conveys a particularemotion are correlated(e.g., Kratus,1993,
Table2; Wedin,1972, Table7), as are prosodiefeaturesthat communicate
emotion in speech (Bachorowski,1999; Frick, 1985). Variouscues that
musiciansuse to convey specificemotions in their performancesare simi-
larly intercorrelated(Gabrielsson&cJuslin, 1996; Juslin, 1997a). In sum,
makingtempo orthogonalto "sadness"would yield stimulithat are not as
sad as they would be otherwise.
One of the most notablefindingsof the presentstudywas that effectsof
pitch and rhythm differed across contexts, even between those that ex-
pressedthe same emotion. Althoughboth of the happy melodiesrevealed
main effects for pitch and rhythmas well as an interactionbetweenthese
variables,the interactiondifferedfor the two sets of stimuli.For one of the
happymelodies,rhythmmade a greatercontributionwhen pitchesvaried,
whereasfor the other melody,rhythmmade a greatercontributionwhen
pitchesdid not vary.Again, these discrepantresultsmost likely stem from
other differencesin the stimuluscontexts. For example, one of the melo-
dies (Happy1) had a triplemeterwhereasthe other (Happy2) had a duple
meter.Becausetriple meters are less common than duple meters and be-
cause the amplitudeof all tones was equalized,establishinga sense of the
meter- and consequentlythe rhythmof Happy 1- may have beenparticu-
larlydifficultfor first-timelisteners.Bycontrast,Happy2 had a clearrhyth-
mic motif (long/short/short)repeatedeight times that was likely to be rec-
ognizedquickly.Becauseof the rhythmicambiguityof Happy Jf,changesin
pitch may have been necessarybefore rhythm'scontributionto the emo-
tionality of the melody could be perceived.For Happy 2, however, the
regular"jointaccentstructure"(redundancyof the pitch and rhythmpat-
terns) meant that either pitch or rhythm could have been sufficient for
evokingthe emotion.Hence,when both factorswere present,the observed
increasein happinessratingswould be smallerthan one would expectif the
melodyhad less redundancyin pitch and rhythm.
The melodicexemplarsof the presentstudy were selectedsolely on the
basis of theiremotionalexpressiveness,which was determinedempirically
by the resultsof the stimulus-selectionprocedure.Althoughthis criterion
ensuredthe validityof the stimuli,it also meantthat the variousexemplars
differedon a numberof dimensions(e.g.,relativesalienceof pitchor rhythm,
numberof contour changes,pitch range, mean pitch level, etc.). In other
words, our focus on emotional validity came at a cost, and differences
amongexemplarson these otherdimensionscould accountfor some of the
findingswe observed.Nonetheless,we would arguethat such extraneous
varianceamongexemplarsmadethemmorerepresentativeof "realmusic"
than they would be otherwise.Moreover,the findingthat differentparam-

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
168 E. GlennSchellenberg,Ania M. Krysciak,& R. JaneCampbell

eters work differentlyin differentmusical contexts seems intuitivelyrea-


sonable.
The presentfindingsneed to be replicatedwith a largerset of melodies,
possibly with manipulationsmore subtle than the all-or-noneapproach
usedhere.Furtherresearchin this areacould investigatethe relativecontri-
butions of pitch and rhythmon the perceptionof other emotionsin melo-
dies, such as anger,surprise,and disgust. One could test non-Westernlis-
teners(orWesternlistenerswith non-Westernmusic)to investigatecultural
differencesand similaritiesin the effectsof pitchand rhythmon the percep-
tion of emotion in music (see Balkwill &cThompson, 1999; Gregory&c
Varney,1996). Such examinationscould help to determinewhich factors
are culture-general(i.e., biologically based) and which are the result of
culture-specificexposureand learning(see Schellenberg& Trehub,1999;
Trehub,Schellenberg,& Kamenetsky,1999). A replicationof this study
with childrenand with older adultswould also prove usefulfrom a devel-
opmentalperspective.Moreover,comparingnaive musiclistenerswith ex-
periencedmusiciansand musicalexperts could help to distinguisheffects
of passivelearningand simpleenculturationfromthose garneredfrom for-
mal training.Carefullycontrolledmanipulationsof additionalmusicalele-
ments, such as timbre,meter,harmony,tempo, and amplitude,could pro-
vide a more complete understandingof how music conveys emotional
meaning.Finally,becauseour resultsmake it clearthat effectsof pitch and
rhythmdifferacrosscontexts, futureresearchcould examinesuch contex-
tual differenceswith closer consideration.2

References
Bachorowski,J.-A. (1999). Vocal expressionand perceptionof emotion. CurrentDirec-
tions in PsychologicalScience,8, 53-57.
Balkwill,L.-L.,& Thompson,W.R (1999). A cross-culturalinvestigationof the perception
of emotionin music:Psychophysicaland culturalcues. MusicPerception,17. 43-64.
Boltz,M. (1989a). Perceivingthe end:Effectsof tonalrelationshipson melodiccompletion.
Journalof ExperimentalPsychology:Human Perceptionand Performance,15, 749-
761.
Boltz,M. (1989b). Rhythmand "good endings":Effectsof temporalstructureon tonality
judgments.Perception& Psychophysics,46, 9-17.
Boltz, M. (1991). Some structuraldeterminantsof melody recall.Memory& Cognition,
19,239-251.
Boltz, M. (1992). The rememberingof auditoryevent durations.Journalof Experimental
Psychology:Learning,Memory,and Cognition,18, 938-956.

2. This researchwas supportedby a grantawardedto the firstauthorfromthe Natural


Sciencesand EngineeringResearchCouncilof Canada.PaulPilonwrotethe computerpro-
gram;MargaretMcKinnonhelpedin recruitingand testingparticipants;and Bill Thomp-
son, PatrikJuslin,and two anonymousreviewersprovidedconstructivecommentson an
earlierversionof the manuscript.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 169

Boltz,M. G. (1993). The generationof temporaland melodicexpectanciesduringmusical


listening.Perception& Psychophysics,53, 585-600.
Brooks,L. R. (1978).Nonanalyticconceptformationandmemoryfor instances.InE. Rosch
& B. B. Lloyd(Eds.),Cognitionandcategorization(pp.169-21 1). Hillsdale,NJ:Erlbaum.
Cohen,J. D., MacWhinney,B., Flatt,M., & Provost,J. (1993). PsyScope:A new graphic
interactiveenvironmentfor designingpsychologyexperiments.BehaviorResearchMeth-
ods, Instruments,& Computers,25, 257-271.
Cooke,D. (1959). The languageof music.London:OxfordUniversityPress.
Cooper,G., & Meyer,L. B. (1960). Therhythmicstructureof music.Chicago:Universityof
ChicagoPress.
Crowder,R. G. (1984). Perceptionof the major/minordistinction:I. Historicaland theo-
reticalfoundations.Psychomusicolozy,4, 3-12.
Crowder,R. G. (1985). Perceptionof the major/minordistinction:III. Hedonic, musical,
and affectivediscriminations.Bulletinof the PsychonomicSociety,23, 314-316.
Crowder,R. G., & Neath, I. (1995). The influenceof pitch on time perceptionin short
melodies.MusicPerception,12, 379-386.
Cunningham,J. G., & Sterling,R. S. (1988). Developmentalchangein the understandingof
affectivemeaningin music.Motivationand Emotion, 12, 399-413.
Deutsch,D. (1980). The processingof structuredand unstructuredtonal sequences.Per-
ception& Psychophysics,28, 381-389.
Dolgin,K. G., & Adelson,E. H. (1990). Agechangesin the abilityto interpretaffectin sung
and instrumentally-presented melodies.Psychologyof Music, 18, 87-98.
P.
Ekman, (1992). Are therebasic emotions?PsychologicalReview,99, 550-553.
Frick,R. W. (1985). Communicatingemotion:The role of prosodiefeatures.Psychological
Bulletin,97, 412-429.
Gabriel,C. (1978). An experimentalstudyof DeryckCooke'stheoryof musicandmeaning.
Psychologyof Music,6, 13-20.
Gabrielsson,A. (1993). The complexitiesof rhythm.In T.J. Tighe& W.J. Dowling (Eds.),
Psychologyandmusic:Theunderstanding of melodyandrhythm(pp.93-120). Hillsdale,
NJ:Erlbaum.
Gabrielsson,A., & Juslin, P. N. (1996). Emotionalexpressionin musical performance:
Betweenthe performer'sintentionand the listener'sexperience.Psychologyof Music,
24, 68-91.
Gerardi,G. M., & Gerken,L. (1995). The developmentof affectiveresponsesto modality
and melodiccontour.MusicPerception,12, 279-290.
Giomo, C. J. (1993). An experimentalstudy of children'ssensitivityto mood in music.
Psychologyof Music,21, 141-162.
Goldstein,A. (1980). Thrillsin responseto musicand otherstimuli.PhysiologicalPsychol-
ogy, 8, 126-129.
Gregory,A. H., SeVarney,N. (1996). Cross-culturalcomparisonsin the affectiveresponse
to music.Psychologyof Music,24, 47-52.
Gregory,A. H., Worral,L., ôc Sarge,A. (1996). The developmentof emotionalresponsesto
musicin voungchildren.Motivationand Emotion,20, 341-348.
Hébert,S., & Peretz,I. (1997). Recognitionof music in long-termmemory:Are melodic
and temporalpatternsequalpartners?Memory& Cognition,25, 518-533.
Hevner,K. (1935). The affectivecharacterof the majorand minormodesin music.Ameri-
canJournalof Psychology,47, 103-118.
Hevner,K. (1936). Experimentalstudiesof the elementsof expressionin music.American
Journalof Psychology,48, 246-268.
Hevner,K. (1937). The affectivevalue of pitch and tempo in music.AmericanJournalof
Psychology,49, 621-630.
Holleran,S., Jones,M. R., & Butler,D. (1995). Perceivingimpliedharmony:The influence
of melodic and harmoniccontext. Journal of ExperimentalPsychology: Learning,
Memory,and Cognition,21, 737-753.
Jones,M. R. (1987). Dynamicpatternstructurein music:Recenttheoryand research.Per-
ception& Psychophysics,41, 621-634.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
170 E. GlennSchellenberg,Ania M. Krysciak,& R. JaneCampbell

Jones, M. R. (1993). Dynamicsof musicalpatterns:How do melody and rhythmfit to-


gether?In T.J. Tighe8c W.J. Dowling (Eds.),Psychologyand music:Theunderstanding
of melodyand rhythm(pp. 67-92). Hillsdale,NJ: Erlbaum.
Jones,M. R., Boltz,M., & Kidd,G. (1982). Controlledattendingas a functionof melodic
and temporalcontext. Perception& Psychophysics,32, 211-218.
Jones,M. R., & Ralston,J. T. (1991). Someinfluencesof accentstructureon melodyrecog-
nition. Memory& Cognition,19, 8-20.
Jones, M. R., Summerell,L., Se Marshburn,E. (1987). Recognizingmelodies:A dynamic
interpretation.QuarterlyJournalof ExperimentalPsychology,39A, 89-121.
Juslin,P.N. (1997a). Emotionalcommunicationin musicperformance:A functionalistper-
spectiveand some data. MusicPerception,14, 383-418.
Juslin,P.N. (1997b).Perceivedemotionalexpressionin synthesizedperformancesof a short
melody:Capturingthe listener'sjudgmentpolicy.MusicaeScientae,1, 225-256.
Juslin,P.N., &cMadison,G. (1999). The role of timingpatternsin recognitionof emotional
expressionfrommusicalperformance.MusicPerception,17, 197-221.
Kastner,M. P., &cCrowder,R. G. (1990). Perceptionof the major/minordistinction:IV.
Emotionalconnotationsin young children.MusicPerception.8, 189-201.
Kidd, G., Boltz, M., & Jones, M. R. (1984). Some effectsof rhythmiccontext on melody
recognition.AmericanJournalof Psychology,97, 153-173.
Kivy,P. (1990). Music alone: Philosophicalreflectionson the purelymusicalexperience.
Ithaca,NY: CornellUniversityPress.
Kratus,J. (1993). A developmentalstudyof children'sinterpretationof emotionin music.
Psychologyof Music,21, 3-19.
Krumhansl,C. L. (1997). An exploratorystudyof musicalemotionsandpsychophysiology.
CanadianJournalof ExperimentalPsychology,51, 336-352.
Lerdahl,E, & Jackendoff,R. (1983). A generativetheoryof tonal music.Cambridge,MA:
MIT Press.
Monahan,C. B. (1993). Parallelsbetweenpitchand time and how theygo together.In T.J.
Tighe&:W.J. Dowling (Eds.),Psychologyand music:Theunderstandingof melodyand
rhythm(pp. 121-154). Hillsdale,NJ: Erlbaum.
Monahan,C. B., & Carterette,E. C. (1985). Pitchand durationas determinantsof musical
space.MusicPerception,3, 1-32.
Monahan, C. B., Kendall,R. A., & Carterette,E. C. (1987). The effect of melodic and
temporalcontouron recognitionmemoryfor pitch change.Perception& Psychophys-
ics. 4L 576-600.
Nettl, B. (1986). Africa.In D. M. Randel(Ed.),The new Harvarddictionaryof music(pp.
16-24). Cambridge,MA: Belknap.
Palmer,C, & Krumhansl,C. L. (1987a). Independenttemporaland pitch structuresin
determinationof musicalphrases.Journalof ExperimentalPsychology:HumanPercep-
tion and Performance,13, 116-126.
Palmer,C, &C Krumhansl,C. L. (1987b).Pitchandtemporalcontributionsto musicalphrase
perception:Effectsof harmony,performancetiming,and familiarity.Perception& Psy-
chophysics,41, 505-518.
Panksepp,J. (1995). The emotionalsourcesof "chills"inducedby music. MusicPercep-
tion, 13, 171-207.
Peretz,I., & Gagnon,L. (1999). Dissociationbetweenrecognitionand emotionaljudge-
mentsfor melodies.Neurocase,5, 21-30.
Peretz,I., Gagnon, L., & Bouchard,B. (1998). Music and emotion:Perceptualdetermi-
nants,immediacy,and isolationafterbraindamage.Cognition,68, 111-141.
Peretz,I., Gaudreau,D., & Bonnel,A.-M. (1998). Exposureeffects on music preference
and recognition.Memory& Cognition,26, 884-902.
Pitt, M. A., ÔCMonahan,C. B. (1987). The perceivedsimilarityof auditorypolyrhythms.
Perception& Psychophysics,41, 534-546.
Rigg, M. G. (1940). Speedas a determinerof musicalmood.Journalof ExperimentalPsy-
chology,27, 566-571.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions
PerceivingEmotionin Melody 171

and culture-specificfactorsin
Schellenberg,E. G., ôc Trehub,S. E. (1999). Culture-general
the discriminationof melodies.Journalof ExperimentalChildPsychology,74*in press.
Scherer,K. R., & Oshinsky,J. S. (1977). Cue utilizationin emotional attributionfrom
auditorystimuli.Motivationand Emotion, 1, 331-346.
Schmuckler, M. A., & Boltz,M. G. (1994). Harmonicand rhythmicinfluenceson musical
expectancy.Perception& Psychophysics,56, 313-325.
Sloboda,J. A. (1991). Music structureand emotionalresponse:Some empiricalfindings.
Psychologyof Music, 19, 110-120.
Sloboda,J. A. (1992). Empiricalstudiesof emotionalresponseto music.In M. R. Jones &
S. Holleran(Eds.),Cognitivebasesof musicalcommunication(pp. 33-46). Washington,
DC: AmericanPsychologicalAssociation.
Stratton,V. N., & Zalanowski,A. H. (1991). The effectsof musicand cognitionon mood.
Psychologyof Music, 19, 121-127.
Terwogt,M. M., & Van Grinsven,F. (1991). Musicalexpressionof moodstates.Psychol-
ogy of Music, 19, 99-109.
Thompson,W. F. (1994). Sensitivityto combinationsof musical parameters:Pitch with
duration,and pitch patternwith durationalpattern.Perception& Psychophysics,56,
363-374.
Thompson,W. E, & Robitaille,B. (1992). Can composersexpressemotionsthroughmu-
sic?EmpiricalStudiesof the Arts, 10, 79-89.
Trainor,L. J., & Trehub,S. E. (1992). A comparisonof infants'and adults'sensitivityto
Westernmusicalstructure.Journalof ExperimentalPsychology:HumanPerceptionand
Performance,18, 394-402.
Trainor,L. J., & Trehub,S. E. (1994). Key membershipand impliedharmonyin Western
tonal music:Developmentalperspectives.Perception& Psychophysics,56, 125-132.
Trehub,S. E., Schellenberg,E. G., & Kamenetsky,S. B. (1999). Infants'and adults'percep-
tion of scale structure.Journalof ExperimentalPsychology:Human Perceptionand
Performance,25, 965-975.
Wedin,L. (1972). A multidimensionalstudy of perceptual-emotional qualitiesin music.
ScandinavianJournalof Psychology,13, 241-257.
White,B. W. (1960). Recognitionof distortedmelodies.AmericanJournalof Psychology,
73, 100-107.
Zajonc,R. B. (1980). Feelingand thinking:Preferencesneed no inferences.AmericanPsy-
chologist,35, 151-175.
Zajonc,R. B. (1984). On the primacyof affect.AmericanPsychologist,39, 117-123.

This content downloaded on Thu, 20 Dec 2012 19:22:58 PM


All use subject to JSTOR Terms and Conditions

You might also like