zatorre, r. j., evans, a. c., & meyer, e. (1994). neural mechanisms underlying melodic perception...

7/29/2019 Zatorre, R. J., Evans, A. C., & Meyer, E. (1994). Neural Mechanisms Underlying Melodic Perception and Memory f

1/12

The Journal of Neurosaence, April 1994, 14(4): 1908-1919

Neural Mechanisms Underlying Melodic Perception andMemory for PitchRobert J. Zatorre, Alan C. Evans, and Ernst MeyerMcConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada H3A 2B4

The neural correlates of music perception were studied bymeasuring cerebral blood flow (CBF) changes with positronemission tomography (PET). Twelve volunteers were scannedusing the bolus water method under four separate condi-tions: (1) listening to a sequence of noise bursts, (2) listeningto unfamiliar tonal melodies, (3) comparing the pitch of thefirs t two notes of the same set of melodies, and (4) com-paring the pitch of the first and last notes of the melodies.The latter two conditions were designed to investigate short-term pitch retention under low or high memory load, re-spect ively. Subtraction of the obtained PET images, su-perimposed on matched MRI scans, provides anatomicallocalization of CBF changes associated with specif ic cog-nitive functions. Listening to melodies, relative to acousti-cally matched noise sequences, resulted in CBF increasesin the right superior temporal and right occipital cortices.Pitch judgments of the firs t two notes o f each melody, rel-ative to passive listening to the same stimuli, resulted inright frontal-lobe activation. Analysis of the high memoryload condition relative to passive listening revealed the par-ticipation of a number of cortical and subcortical regions,notably in the right frontal and right temporal lobes, as wellas in parietal and insular cortex. Both pitch judgment con-ditions also revealed CBF decreases within the lef t primaryauditory cortex. We conclude that specialized neural sys-tems in the right superior temporal cortex participate in per-ceptual analysis of melodies; pitch comparisons are effectedvia a neural network that includes right prefrontal cortex, butactive retention of pitch involves the interaction of right tem-poral and frontal cortices.

[Key words: human auditory processing, music, pitch, au-ditory working memory, PET, temporal cortex, frontal cortex]

Perceiving a melody, or sequence o f pitches, requires relativelycomplex perceptual analysis, since short-term memory and ab-stract pattern-matching mechanisms must be involved (Deutsch,1982; Dowling and Hat-wood, 1986). It seems likely that suchprocessing would also demand a correspondingly complex set

Received July 8, 1993; accepted Octobe r 4, 1993.We thank the technical sta ff of the McConnel Brain Imagi ng Unit and of theMedical Cyclotron Unit for their invaluable assistance, and P. Neelin, J. Moreno-

Cantu, and S. Mllot for their technical expe rtwz. We also thank Dr. D. Perr) forhelpful discussions, and Dr. A. Gjedde for his guidance and support. Funding wasprovided in part by Gran ts MT 1 I541 and SP-30 fro m the Medical ResearchCounci of Canada, by an award to R.J .Z. fro m the Fends de la Recherche cnSante du OuCbcc . and bv the McDonnell-Pew Cognitive Neuroscience Center.

Correspondence should be addressed to Robert j. Zator re, Montreal Neurolog-lcal Inst itute , 3801 University Stre et, Montrea l, Quebec, Canada H3A 2B4.

Copyright 0 1994 Society fo r Neuroscience 0270-6474/94/ 14 190% 12$05.00/O

of neural computations, but at present, the cerebral substratesfor melodic processing remain poorly understood. Currentknowledge of complex auditory neural processing comes fromtwo principal sources: neuroanatomical and neurophysiologicalstudies of the auditory system of nonhuman vertebrates, andstudies of the effect s of human brain lesions on auditory tasks.Each of these will be brief ly reviewed before introducing thepresent investigation, which takes advantage of recent advancesin PET brain imaging techniques to study the funct ional cerebralsystems involved in melodic perception in the living humanbrain.

The superior aspect of the temporal lobe has long been knownto contain neurons responsive to auditory stimulation in bothmonkey (Ferrier, 1876; Merzenich and Brugge, 1973) and hu-man (Flechsig, 1896; Celesia, 1976). Several distinct corticalfields may be identified on the basis of the neurophysiologicalresponse properties of neurons within those fields (Brugge andReale, 1985), and also based on their cytoarchitectonic char-acterist ics. A general distinction may be drawn between theprimary auditory area-or koniocortex-lying deep withinHeschls gyri in the human brain (Galaburda and Sanides, 1980;Liegeois-Chauvel et al., I99 l), and surrounding regions, whichmay be termed secondary or associative auditory cortex. Someof these cortical fields are arranged tonotopically (Woolsey, 197 1;see Brugge and Rcale, 1985, for a review), suggesting that theyplay a role in frequency analysis. However, apart from the pri-mary and immediately surrounding regions, the frequency tun-ing of individual units can be very broad, and the tonotopicorganization can be dif ficult to discern (Manley and Mtiller-Preuss. 1978). Some neurons respond best to complex stimuli:such as tones modulated in frequency (Whitfield and Evans,1965) or patterns of ascending or descending tone sequences(McKenna et al., 1989).

The morphological and physiological distinc tions betweencortical regions raise important questions regarding the conse-quences of such an architecture for auditory processing. Candissociations be identified that would allow functional signifi-cance to be assigned to the structural heterogeneity? A possiblefunct ional role for human superior temporal neurons in higher-order auditory processes is suggested by the observations ofPenheld and Perot (1963) and of Penfield and Jasper (1954).They noted that complex auditory sensations (including voicesand music) could be elicited from electrical stimulation of theexposed cortex in this region, but not from Heschls gyri, whichinstead usually resulted in more elementary sensations, such asbuzzing.

The connec tivity of these regions indicates that both primaryand secondary areas receive input from various divisions of themedial geniculate nucleus (Burton and Jones, 1976) suggesting


2/12

The Journal of Neuroscience, April 1994, 14(4) 1909

a probable paralle l organization. This idea is supported by elec- age significantly impaired pitch retention, a finding mirrored introphysiological (Celesia, 1976) and behavioral (Tram0 et al., certain anim al studies of bilateral frontal abla tion (Gross and1990) observations indicating that secondary regions may con- Weiskrantz, 1962; Iversen and Mishkin, 1973). The latter resulttinue to function even after destruction of primary areas. In may reflect some nonspecific effect of frontal lesions on behavioraddition , several investigators have described extensive projec- (e.g., disturbances in inhibition of responses), but a more in-tions from the superior temporal lobe to the anterior frontal triguing possibility is that the disturbance is related to disruptioncortex (Chavis and Pandya, 1976; Petrides and Pandya, 1988) of the frontotemporal connections discussed above, which wouldwhich are topographically organized. Such connections imply imply a role for the interaction between these cortices in pitchthat there may be a functional interact ion between these cortices. retention (Perry, 1990; Marin and Perry, in press; see also Cha-Another important issue that is particularly relevant to me- vis and Pandya, 1976).lodic processing concerns hemispheric specialization. The hu- The recent advent of funct ional PET activation techniques inman neuropsychological literature remains somewhat contro- normal human subjects has added one more source of infor-versial on this point . Although there is some consensus that mation to the body of knowledge outlined above. To date, themany aspects of musical processing probably require contri- majori ty of these studies have investigated verbal auditory pro-butions from neural systems within each cerebral hemisphere cessing, with the principal findings indicating increased cerebral(Peretz, 1993) there is also considerable evidence indicating blood flow (CBF) bilateral ly in the superior tempora l gyri whilethat right-hemisphere mechanisms are particularly important listening to verbal stimul i, and activation of specific left-hemi-for some aspects of melody perception (for a more extensive sphere sites for certain aspects of phonological , lexical, or se-review, see Marin and Perry, i n press). Among the first to explore mantic processing (Petersen et al., 1988; Wise et al., 199 1; DC-this issue systematically, Milner (1962) and Shankweiler (1966) monet et al., 1992; Zatorre et al., 1992). Activation of primarydemonstrated decrements in melodic discrimination following auditory regions for simple auditory stimuli such as pure tonesright temporal lobectomy, but not after left-sided removals. This (Lauter et al., 1985) and noise bursts (Zatorre et al., 1992) hasfinding was subsequently confirmed by Zatorre (1985) and by also been reported. In one condition of our recent investigationSamson and Zatorre (1988) who also found that left temporal- (Zatorre et al., 1992) we investigated pitch processing by in-lobe removals could affect melodic discrimination, but only if strutting subjects to compare the pitch of a pair of syllables; thethe damage extended into portions of Heschls gyri. Further pattern of CBF was subtracted from a prior condition in whichconsistent evidence was provided by Zatorre and Halpem (1993) the subjects listened to the same syllables but made no overtwho demonstrated deficits in pitch judgments of both perceived judgment. The principal result of this comparison was a CBFand imagined pitches within well-known songs after right, but increase in two sites within the right infer ior frontal cortex,not left, temporal-lobe excision. which contrasted notably with CBF activation in left Brocas

Studies of single-tone pitch perception also generally support area when a phonetic judgment was required using the identicala preponderant role for right-hemisphere neural systems, but set of speech syllables. This result therefore implicates the rightonly in specific aspects of pitch processing. Thus, simple fre- prefrontal cortex in pitch processing.quency discrimination is affected only slightly or not at all by The present investigation was undertaken to explore melodicunilatera l cortical lesions (Milner, 1962; Zatorre, 1988), and can perceptual processes in normal human subjects via functionalprobably be accomplished via subcortical structures. The animal PET activat ion. Based on the studies described, we hypothesizedliterature is also consistent with this assertion (Evarts, 1952; that perceiving a novel tonal melody would entail neuronalJerison and Neff, 1953; Heffner and Masterton, 1978). However, processing in both left and right superior temporal regions, withif a pitch judgment requires spectral analysis, then right-side a possibly greater contribution from the right side. We alsoauditory cortical regions seem to play a special role. For ex- predicted that secondary auditory cortices would be primarilyample, perception of the missing fundamental is affected only involved, and therefore that activation of primary regions couldby right temporal-lobe lesions that invade portions of Heschls be subtracted out by using a nonspecific auditory stimulus,gyri, and not by more restricted anterior temporal-lobe damage matched for its acoustic characteristics with the melodies, as aor by left temporal excision (Zatorre, 1988). Simi lar findings control condition (Zatorre et al., 1992). Fina lly, we predictedhave been reported in tasks requiring processing of complex that right frontal-lobe mechanisms would be engaged when sub-harmonic structure (Sidtis and Volpe, 1988; Divenyi and Rob- jects make specific judgments of pitch changes within a melody,inson, 1989; Robin et al., 1990). Furthermore, timbre discrim- and that when the judgment requires retention of pitch over aination tasks involving changes in harmonic structure have also filled interval, additional right-temporal activity, beyond thatyielded consistent evidence favoring right-asymmetric process- ini tia lly present during passive listening, would be observed.ing, both with temporal-lobe lesioned patients (Milner, 1962;Samson and Zatorre, 199 la), as well as with commissurotom-ized subjects (Tram0 and Gazzaniga, 1989). Materials and MethodsShort-term retention is another aspect of pitch processing that PET scanning. PET scans ereobtainedusing he ScanditronixPC-apparently requiresasymmetric mechanisms.Bilateral ablationsin the superior temporal gyrus of the monkey result in deficitsin tonal retention (Stepien et al., 1960; Colombo et al., 1990).Single-unit data also implicate this region in auditory short-term memory (Gottlieb et al., 1989). In human patients,.Zatorreand Samson 1991) demonstrated hat right temporal-lobe ex-cision affected short-term memory for pitch when interferingstimuli werepresentedbetween he target and comparison tems

2048system,whichproduces 5brain mage lices t an ntrinsic es-olution of 5.0 x 5.0 x 6.0 mm (Evanset al., 1991a).Using he bolusH,150methodologyRaichleet al., 1983)without arterialbloodsam-pling (Fox and Raichle, 1984) the relative distributionof CBF wasmeasuredn baseline nd activatedconditions. ndividual hiah-reso-lution MRI studies63slices, mm hick) were btained romaPhilips1ST Gyroscan ndcoregisteredith thePETdata Evans t al., 199 ).An orthogonalcoordinate rame was hen established ased n theanterior-posterior ommissureine as dentified n the MRI volume(Evanset al., 1992).Thesecoordinates ereused o apply a linear(Deutsch, 1970); hey alsoobserved hat right frontal-lobe dam- resampling f eachmatchedpair of MRI and PET data sets nto a


3/12

1910 Zatorre et al. * Neural Mechanism s for Pitch PerceptIon

Frgzm I. A. Amplitude wa\-eform ofnoise stimulus . II. Waveform of onemelodic stimulus. Note slmllarity mduration and shape of noise bursts ascompared to melodic stimulu s C, ET-nmplrr of three melodies used as stim-uli, in music al notation. Cl 1

1 set I

standardized stereotaxlc coordinate system (Talalrach and Tvur~nour.1988). PET images were reconstruct4 using a 20 mm Hannlng filtcl-to ovcrcomc residual anatomical variability, normalized for global CUFand avcragcd across subjects for each actixatlon state. The mean statc-dependent change image volume was then obtained, and convex-ted toa t statistic \-olume b) dividing each boxrl by the mean standard dc-viation in normalized (ISF for all into-acercbral \ 0~1s (Worslcl et al.,1992). Jndividual MR images were subjected to the same averagingprocedure, such that compositr images volumes sampled at approxi-mat+ 1.5 mm in each dimension were obtained for both I statis llc andMRI volumes. Anatom ical and functional images were merged to allowdirect lo call/ation on the MR Imagch of rcglons \rith high t values.

The presence of significan t focal changes in CBF was tested by amethod based on 3-D Gaussian random field thcnry (tVorsle> et al..1992). Values equal to or exceeding a criterion of i = 3.5 were deemedstatistically significan t (i, -: 0.0002. on&ailed, uncorrected). Correctingfor multiple comparisons. a t x alue of 3.5 yields a false po siti\ c rate ofonly 0.58 in 200 resolution elements (each of which has dimension s 20x 20 x 7.6 mm). which appI-osimatcs the volume of cortex scanned.

.Sui~/c~ts. lwel\c normal right-handed volunteers. half of each sex.participated In the stud! after go\ mg Inform& consent. Subject5 Lvercunselected for music al training; most ofthem had received some music alInstruction, u\uall) a\ part of thclr elemcntar) or high school education.hut none WLI-c professional musicla ns.

Sfi,?~/i. Two types of stimu li were pl-eparcd: noise bursts and mel-odies. The noise bursts were constructed so a5 to approximate the acous-tic character-istics of the melodies in terms of number, dur-ation, inter--stimulus presentation rate. intensity. and onset off&t \hapc. White noisewas synthesized on a MAUI computer, and then scgmentcd to matchthe average duration ofeach note ofthc melodies (set below). producinga noisc melody consistin g of clght noise xgments with appropriatedurations (see Fig. 1). The inchvldual noise segments were fur-thermatchedto thr notes by shaping their onse ts and ufliets to approximate theamplitude en\clopcs of the music al tones, and the rntire sequence wasthen attenuated to an intensity level similar to that of the melodies.Total duration of the noise pattern was 5.0 see.

Sixteen different H-note tonal mclociie s were prepared, all identical intheir rhythmic conliguration (Fig. 1). with the aim of allowing pitch

Table 1. Summary of paradigm, showing stimu li presented and responses elicited during each of thefour experimental conditions

Reac-b tionConditlon Stlmulu\ Response coITect IlrneNoise Noise bursts Key press after each stimulu s -Passive melodlc s X-Note melodies Key press after each stimulu s -?-Note X-Note mclod ics Comparison of first two notes Y5.8 446First/Last S-Note mclodiea Comparison of first and last notes 89.1 917Th? stirnull were Identlm l for all but the firs t inoise) condltlon. and the mo tor response (Lx) presr) wa s the sanw in allfour conditions. The subjects Judgments diffcrcd across conditions. No specif ic dcclsio n \has required Tar either of thefirst two conditions. Howeve r, in the 2.note condition the pitch ol the firrt two notes was to bc compared, and theappropriate decl\lon signalled h! a kc) press. nherrac in the lirst, 1351 condltlon the prtch of the lir\t note of the melodywas to be compared lo the pitch of lhc last note, followed by the appropriate key prer\. Performance data (meanpercentage correct and mean reaction time. in mrllisccon ds. measured from stimulus offset) were collected on line duringPET scanmng . and mdlcate that \ubyct) succe>sfull> performed the task. and that the plrch companso,~ III the lirstla,tcon&ion was more dlffcult than in the 2-note conditm n.


4/12


Table 2. Passive melodies minus noise

Brodmann Coordinates (mm)Region area x Y z t ValueBlood flow increases

I. R Superior temporal gyms 22 62 -25 3 4.412. R Fusiform gyms 19 25 -78 -11 4.26

Blood flow decreases3. M Posterior cingulate gyrus 31 4 -49 29 4.404. M Posterior cingulate gyrus 31/7 0 -52 40 3.745. R Frontal operculum 44 44 10 12 3.646. R Inferior colliculus - 7 -26 -17 3.53

Data show significant activation foc i (blood flow increases and decreases, n order of decreasing significance) for sub-traction of passive melody condition minus noise condition. In this and subsequent tables, stereotaxic coordinates arederived from the human brain atlas of Talairach and Toumoux (1988), and refer, in millimeters, to medial-lateralposi tion (x) relative to mid line (positive = right), anterior-posterior posit ion (y) relative to the anterior com miss ure(positive = anterior), and superior-inferior position (z) relative to the comm issural line (positive = superior). Designationof Brodmann areas for cortical areas, based on this atlas, is approximate only. Significan ce level is given in t test units(see Materials and Methods for details). L, left; R, right; M, midline .

judgments of either the firs t two notes, or the firs t and last notes. Theduration of the firs t and last notes of each melody (approximately 1 set)was twice that o f each of the six intermediate notes, in order to facilitatetheir comparison. The last note had a higher pitch than the first in halfof the 16 melodies, and in the other half, the last note was higher inpitch. Within each of these subsets of eight melodies, half contained arising interval between the first and second notes, and the other halfcontained a falling interval (see Fig. 1). Thus, the pitch change betweenthe first two notes was independent of the pitch change between the firs tand last notes. Furthermore, the average pitch distance of the notes tobe compared in each condition was comparable across melodies: thefirst two notes differed in pitch by an average distance of 4.5 semitones(range, 1-l 2); the average pitch distance for the fir st and last notes was5.5 semitones (range, 3-12). The melodies were played on a YamahaPS4 electronic keyboard, using the guitar timbre, then digitized andstored on the computer for later playback. The average total durationof each melody was 4.7 sec.Procedure. Four separate conditions were run during each of the fourscanning periods (see Table 1). Although each scan lasted only 60 set,the tasks were always begun several seconds before scanning com-menced, and continued after scanning, until all 16 melodies had beenpresented. Performance data were collected on each subject on lineduring scanning. The total duration of each stimulus condit ion wasusually about 2.5 min.During the first condition, termed the noise condition, subjectslistened to the series of noise bursts described above, and afte r eachnoise melody depressed a key with their right hand, which resultedin the next stimulus sequence being played. In the second condition,termed passive melodies, the subjects were presented with each ofthe 16 tonal melodies, and depressed a key after each one, as before.No overtjudgments were required, but subjects were instructed to listencarefu lly to each melody as it was played. In the third condition, the

2-note pitch comparison, subjects listened to the same melodies asbefore, but this time were instructed to determine whether the pitch ofthe second note was higher or lower than that of the fi rst note. Theywere to respond accordingly on the computer keyboard, but were re-quested to withold the response until after the entire melody had beenplayed. Finally, in the first/ last pitch judgment, subjects were askedto compare the pitch o f the first and last notes, ignoring the notes inbetween, and to respond as before according to whether the pitch roseor fell.The order of the 2-note and first /last tasks was counterbalancedacross subjects, who were not instructed as to the nature of any of thejudgments until immediately prior to scanning; however, several prac-tice trials were given prior to starting each task. A different randomorder of the same 16 melodies was presented to each subject duringeach of the melody conditions. All stimuli were presented binaurallyvia insert earphones (Eartone type 3A). Subjects kept their eyes shutthroughout the scanning period.

ResultsThe experiment was set up to permit specif ic comparisons, ac-complished via subtraction of relevant conditions. The resultsof these subtractions, in terms of significant regions of CBFchange (increases or decreases), are given in Tables 2-4, togetherwith stereotaxic coordinates based on the brain atlas of Talair-ach and Toumoux (1988).

The firs t comparison, passive melodies minus noise, permitsexamination of the cerebral regions specifically active duringlistening to novel tonal melodies, as opposed to the activationthat might be present with any auditory stimulus with similaracoustic characteristics. The principal result, shown in Figure2A, indicates a large CBF increase in the right superior temporalgyrus, anterior to the primary auditory cortex, as predicted. Inaddition, and unexpectedly, a significant focus was also iden-tified in the fusi form gyrus of the right hemisphere, within area19 (see Fig. 2A). An area o f positive CBF change can also beseen in the lef t temporal lobe (Fig. 2A; x, y, z coordinates: -52,- 13, 2), although the strength of this signal (t = 3.35) wasinsuff icient to achieve statistica l significance by our criteria. Ar-eas ofblood flow decrease were also observed in this subtraction,including the posterior cingulate region and right frontal oper-culum.

The second and third comparisons both use the passive mel-ody condition as the baseline, so that any activation seen rep-resents neural responses beyond those already present duringinitial listening to the same stimulus materials. Subtraction ofthe 2-note condition from passive melodies results in the patternof cerebral activation shown in Figure 2B and Table 3. Mostprominent in this subtraction is the significant activation withinthe right frontal lobe, as predicted. Two separate foc i can bedistinguished within distinct cytoarchitectonic regions, includ-ing Brodmanns areas 47/l 1 (focus l), and 6 (focus 2). Otherregions that were also act ive include anterior cingulate and cer-ebellum. An area within the medial parietal lobe, 8 mm to theright of midline, fell at the borderline of significance (t = 3.47).

Most notable among the areas showing significantly decreasedCBF are two adjacent foc i within the lef t gyrus of Heschl (foc i6 and 8 of Table 3) or primary auditory cortex, shown in Figure


5/12

1912 Zatorre et al. - Neural Mechanism s for Pitch PerceptIon

Figure 2. Selected cortical regons activated in the various conditions. The averag ed PET subtraction images for the most pertinent foci are shownsuperimposed upon the corresponding averaged horizontal MRI scan. Subtraction of the control from activated state in each case yielded the focalchanges in blood flow shown as a t statist ic image, whose range is coded by the CIlor scale shown on each figure (see Table 1). A, passive melodyminus noise subtraction. The two images in this figure, taken at horizontal levels of 1 1 mm below and 3 mm above the commissural plane, illustrate


6/12

The Journal of Neuroscience, Apn 1994, 14(4) 1913

Table 3. 2-Note judgment minus pass ive melodiesBrodmann Coordinates (mm)

Region area x Y z t ValueBlood flow increases

I R Inferior frontal 41 11 40 46 -13 4.152. R Superior frontal 6 31 1 48 3.813. M Cerebellum - 4 -57 -11 3.814. L Anterior cingulate gyrus 32 ~8 18 29 3.66

Blood flow decreases5. M Anterior cingu late gyms 24 -1 27 -2 4.886. L Hcschls gyms 41 -43 -18 12 3.897. R Claustrum putamen - 27 -2 11 3.868. L Hesch ls gyrus 41 -44 -14 11 3.869. L Frontal pole 10 -5 58 0 3.82

10. L Frontal pole 9 -15 60 30 3.67I 1. L Lateral cerebellum - -35 -45 -20 3.6712. L Lateral cerebellum - -20 -47 -17 3.63

Data show significant activation foci (blood flow increases and decreases) for subtraction of 2-note cond ition minuspassive melody condition. See Table 2 note for further details.

2H.In addition we observed decreases in the anterior cingulatccortex, in two left frontal polar regions, and in the lef t lateralcerebellum.

The first /last passive melodies subtraction also yielded severalCBF increases within the right frontal lobe, consistent with thepredictions (see Fig. 2C, focus 3 visible at z = - 13 and z = -6,and focus I 1, at z = 22). Also in keeping with our predictions,we identified an area of significant CBF increase within area 2 1of the right temporal lobe (focus 6, 7 = -6). indicating that thiscondition rcsultcd in greater act iv-i ty within the right auditoryassociation cortex than already present during passive listeningto melodies.

The analys is of the data from this subtraction also yieldednumerous other regions of positive activation, notably withinthe parietal lobe bilaterally. and on the right side (foci 12, 13.and 14, respect ively, all visible in Fig. 2C at z = 38 and z =45) and medially (focus 7, visible at L = 45 in Fig. 2C). Otherregions visible in Figure 2C arc in the anterior cingulate gyrus(7 = 22, z = 38, 7 = 45) left precentral region (z = 45), andright inferior colliculus (z = ~ 13). Bilateral foc i were also ob-served deep within the Sylvian fissure. at the junction betweenthe frontal opercular region and the insula (\.isiblc in Fig. 2C,atz = 6).

As in the previous subtraction, the first /last minus passivemelody analysis also yielded a prominent area of CBF decrease

within Heschls gyrus in the lef t hemisphere (focus 19 in Table3. visible in the bottom row of Fig. 2C at z = ~ 12). In addition,se\-era1 other left-hemisphere regions demonstrated lower CBF.including an inferior temporal region (focus 21). and an area inthe posterior parietal lobe (focus 24). Finally, it is interesting tonote the presence of several f oci o f CBF decrease within themedial temporal region that may be related, including the uncusbilaterally (foci 23 and 25). and one in the right parahippocam-pal gyrus (focus 28). all of which arc visible in the bottom rowof Figure 2C at z = ~ 18.

The results o f the subjects performance on the 2-note andfirst /last conditions are given in Table 1, which shows the meanpercentage correct, together with the mean reaction time. Allsubjects performed both tasks \vc ll above chance (50%). but, asexpected, more errors were committed on the first /last conditionthan the 2-note condition (Wilcoxon signed rank test. W? = 45,p < 0.01) and reaction times wcrc consistently longer.DiscussionIn general. the patterns of CBF changes observed in this studyprovide support for the hypothesis that perceptual analysis andshort-term retention of tonal pitch information preferentiallyinvolve neural systems within the right frontal and temporalcortices. However. it also appears that the operations involvedin these complex tasks make demands upon a widely distributed

tthe significant CBF increases noted in the fusiform gyms (focus 2 in Table 2) and right superior temporal gyrus (focus l), respectively. Note alsothe presence of a nonsignificant CBF increase in the left superior temporal area, visible at z = 3 mm. n, Two-note minus passive melody subtraction.Shown are the two significa nt areas of CRF increase in the right frontal cortex. one more inferior (focus 1 in Table 3: plane of image z = -13)and one more superior (focus 2; plane of image z = G), as well as a significant CBF decrease (second TOW) in the left primary auditory cortex (foci6 and 8; plant of image z = 12). C. First/last minus passive melody subtraction. This figure illustrates the following significant CHF increases. TopYOW, Increases in the right inferior frontal lobe (focus 3 in Table 4. visible at z = -13 and z = ~6) in the right inferior collicu lus (focus 5. z= ~ 13). and in the right temporal lobe (focus 6. z = ~6). Second ruti, Increases bilaterally at the junction between anterior insular and frontalopercular regions (foci I5 and 16. both v-isible at : = 6). in the right m id-frontal region (focus 11. z = 22), and m the right anterior cingulate gyrus(focus 2. visible at : = 22). 7h/rd w~L, Increas es in the rrght and left inferior parietal lobe (foci I2 and 13, z = 38). in the right supe rior pat-ietalregion (focus 14. merging with the loner parietal-lobe focus at z = 45). in the anterior cingulate. midline (focus 1, visible at z = 38 and I = 45)in the left prcccntral region (focus 4. z = 45). and in the medial parictal cortex. midline (focus 7. visible at z = 45). Bottom row, Some of thesignificant CBF decreases detected in this subtraction: in the left and right uncus, and right parahippocampal gyrus (foci 23, 25, and 28, all visibleat z = - 18). and in the left primary auditory cortex (focus 19. z = 12); also visi ble at z = I2 is a portion of the decrease in the right o percularregion (focus 22).


7/12

1914 Zatorre et al. * Neural Mechanisms for Pitch Perception

Table 4. First/last note judgment minus passive melodies

Reeion BrodmannCoordinates (mm)

area x v % t ValueBlood flow increases

1. M Anterior cingulate gyrus2. R Anterior cingulate gyrus3. R Inferior frontal4. L Precentral frontal5. R lnferior colliculus6. R Middle temporal gyrus7. M Medial parictal8. R Mid-frontal9. R Cerebellum

10. L Mid-frontal1 1. R Mid-frontal12. R Inferior parictal13. L Inferior parietal14. R Superior parietal15. R InsuWfrontal operculum16. L Insula/frontal operculum17. M Cerebellum

Blood flow deco-eases18. M Anterior cingulate gyrus19. L Heschls gyrus20. L Posterior cingulate gyrus2 1, L Inferior temporal22. R Parietal operculum23. L Uncus24. L Posterior parietal25. R Uncus26. R Medial occlpltotemporal27. M Frontal pole28. R Parahippocampal gyrus

32/83247

219-91469/46

404040/74545

244131204034/283934/28371028136

538

-355536

8-36

3947

~363938

-31

-3~42

-7~46

46~29-48

232728

243651

-2-30-26~69

36-69

2941

-38-47-50

2022~5230 -2 5.21

-19 12 4.76-47 35 4.60-11 ~29 4.48~-.18 21 4.44

5 -18 4.23-69 24 4.19

5 -17 3.86~56 ~6 3.82

61 15 3.58-25 -20 3.50

4226

-945

-13~64731

-243022383945

8-11

4.944.944.714.414.294.044.043.803.803.763.723.683.683.633.633.593.51

Data show significant activation foci (blood flow increases and decreases) for subtraction of first/last note conditionminus passive melody condition. See Table 2 note for further details.

system, involving interactions between a number of distinctregions in both cerebral hemispheres.Some of the analysesyielded a largenumber of CBF change oci. not all of which canbe readily interpreted, given our presentknowledge. n this sec-tion, therefore, we discuss he most pertinent results rom eachsubtraction, and then turn to a more generaldiscussion.Passivemelodiesminus noiseAs predicted, this comparison yielded activation of the rightsuperior temporal cortex (focus 1 in Table 2, visible in Fig. 2Aat z = 3), in keeping with the recognized ole of this region inmelodic processing.The most surprising inding in this analysiswas the CBF increase n the right occipital cortex (fusiformgyrus) while subjects istened to melodies,as contrasted to lis-tening to acoustically matched noise segments visible in Fig.2A at z = ~ 11). The likelihood that this result represents implya statistical artifact would appear ow, since he t value of 4.26iswell above threshold; t is alsoworth mentioning that a secondfocus within area 19 wasobserved,and although t did not reachsignificance t = 3.25), its presence urther supports he finding.The possibility that the effect is due to someextraneous visualstimulation is alsoexcluded, sincescanningwascarried out withthe subjectseyesclosed.

Area 19 is typically described as extrastriate visual cortex(Diamond et al., 1985); here is but scantphysiologicalevidencefor its direct participation in auditory processingMorrell. 1973).Previous PET studies using melodies or musical scaleshavefailed to observe occipital changes Mazziota et al., 1982; Ser-gent et al., 1992); but the former study measuredglucoseme-tabolism with no specific task or stimulus controls, renderinginterpretation difficult, while in the latter study the baselinecondition included visual stimulation, rather than an acousti-cally appropriate control, soany visual-cortical activation mighthave beenobscured.However, there doesexist one brief reportof a PET study using frequency-modulated tones n which ac-tivation of left area 19was eported (Grifiths and Brown, 1991).These authors raise the possibility that a cross-modalspatialperceptualsystemwasbeing engaged y the apparentmovementof the frequency-modulated tones. Melodies. of course, nvolvefrequency modulation; it is therefore plausible to suggesthatat leastsomesubjectsmay have activated visual representations,consciously or not, and that our observations n area 19 reflectthis process.The notion that activation of this region in thepresentstudy is due to visual imagery processes ould be con-sistent with recent findings of blood flow increasesn area 19and other visual cortical regions during explicit visual image


8/12

The Journal of Neuroscience, April 1994, M(4) 1915

generation tasks (Kosslyn et al., 1993). This possibility mustremain a conjecture for the moment, until direct evidence canbe adduced in its favor. We also do not know at present whetherthe putative visual activation is a general effect, or if it is specificto some particular global or loca l features of our stimulus ma-terials.The fact that the occipita l CBF activity was confined to theright side is consistent with the general tendency for right-hemi-sphere processes to be particularly important in melody per-ception. This right-side predominance was also evident in thetemporal-lobe activation, which, as predicted, was observed inthe superior tempora l cortex anterior and inferior to Heschlsgyri. The much weaker activat ion in the left temporal cortex(visible in Fig. 2A at z = 3) is probably genuine, since bothtemporal lobes undoubtedly contain neurons responsive to theacoustic features present in the melodies. The asymmetry weobserve likely reflects the specialization of neuronal networkswithin the right associative auditory cortices for perceptual anal-ysis of tona l information, consistent with the human lesion ev-idence reviewed above (Milner, 1962; Zatorre, 1985; Zatorreand Halpern, 1993).

Note that, in this subtraction, no CBF change was present inthe primary auditory cortices beyond that elicited in the controlcondition. This result is expla ined by the use of noise sequencesas the control condition. Since CBF can change markedly withchanges to physical stimulus parameters (Fox and Raichle, 1984)it can be problematic to compare scanning conditions in whichthe stimul i are not physically similar. In the present experiment,by using the acoustically matched noise melodies (see Fig. l),nonspecific auditory processing could be dissociated from thatwhich is uniquely elic ited by listening to melodies. We previ-ously demonstrated that similar noise bursts result in primaryauditory cortical stimula tion when contrasted to a silent con-dition (Zatorre et al., 1992). These findings, together with find-ings from prior PET studies using speech sounds or tones (Lauteret al., 1985; Petersen et al., 1988; Wise et al., 1991; Demonetet al., 1992) point to differential activation of primary versussecondary auditory areas within the superior temporal gyrus,according to the nature ofthe processing elic ited by a given stim-ulus. Although the noise stimuli proved successful in demon-strating the intended dissociation, caution must be still exercisedin interpreting the results, for the noise bursts are clearly notphysically identica l to the melodic sounds. For example, thenoise stimul i contain no periodicity, whereas the tones do; theirspectral composition also is quite different. It remains to beestablished, therefore, which specific features of the melodiesmay lead to the observed pattern of activation.A word is in order at this point about the interpretation tobe given the passive melodies condition. Our use of the termpassive is meant only to describe the lack of overt behavioralresponse, and does not imply that listening to such complexmater ial is a passive cognitive process. On the contrary, it iscertain that listening to melodies implies quite active, complexmental operations, including perceptual and mnemonic pro-cessing. (The subjects were, after all , instructed to listen carefullyto the stimuli.) Apart from this purely semantic debate, how-ever, questions can be raised about the validity of any conditionin which the subject is not constrained to respond in a particularfashion. It might be argued, for example, that the lack of explicittask demands renders interpretation difficult, because it will notbe possible to determine what specific operations were per-formed during the so-called passive task. It is clear that using

specific behavioral tasks to identify particular aspects of pro-cessing is critica lly important, as will be discussed in the fol-lowing section, but in many instances the question of interestis precisely to understand the cognitive processes inherent tonondirected perceptual analysis. The advantage of functionalPET brain mapping is that answers to this question can beinferred from the pattern of CBF changes during such passiveperceptual conditions, as long as appropriate control conditionsare used. Thus, in addition to providing relative ecological va-lid ity, a passive cond ition allows one to address this crucialquestion. In the present case, for example, we were interestedin the cerebral mechanisms that allow extraction of perceptuallyrelevant information from a tonal pattern, something that occursautomatical ly (Dowling and Harwood, 1986). If only experi-menter-imposed judgments were to be analyzed, one might runthe risk of overlooking or confounding the cerebral correlatesof such automatic processes, which are likely to be of majorimportance in understanding music cognition in general.2-Note minus passive melodiesWe now turn to the results involving expl icit pitch comparisons,which were carried out in two ways (2-note and first/last con-ditions). Note that in both cases the stimuli are identical , andthat the judgment required (pitch rise or fall ) is also the same,the only difference being which notes are to be compared. Thus,these conditions should allow us to study the cognitive processesrequired for pitch decisions under conditions of low or highmemory load, respectively, using exper imentally closely matchedconditions. The performance data (Table 1) allow us to be cer-tain that the subjects followed the instructions, and were indeedperforming the appropriate comparisons, since performance rateswere generally high. At the same time, we were able to documentthe expected decrease in correct performance comparing the2-note to the first/last condition (Deutsch, 1970) which dem-onstrates that the latter comparison is indeed a more demandingcognitive task.

Considering first the 2-note minus passive melodies compar-ison, we had predicted that regions within the right frontal lobewould be activated, and this was indeed the case. Although themore superior frontal area (focus 2 in Table 3, visible in Fig.2B at z = 48) is not far (~2 cm) from one reported by Zatorreet al. (1992) the more inferior one (focus 1, z = - 13 in Fig.2B) is in a different cytoarchitectonic region (Brodmann area47/l 1).

In this same subtraction we also observed two adjacent areasof significant CBF decrease within the left primary auditorycortex, with no equivalent change in the homologous region onthe right side (Fig. 2B, z = 12). Decreases in CBF may be simplyinterpreted as reflecting increases in the baseline condition overthe experimental condition; that is, in this case the finding in-dicates that the left primary auditory cortex was significantlymore active during the passive listening than during perfor-mance of the pitch judgment task. We shall return to this poin tbelow.First/last note minus passive melodiesIn the first/last minus passive melody comparison, we observeda greater number of separate foci of CBF change over a widerswath of cortical and subcortical territory than was evident inthe low memory-load comparison, perhaps reflecting the com-plexi ty and increased cognitive demands o f the task, as mani-fested in the increased error rate and slower reaction time.


9/12

1916 Zatorre et al. - Neural Mechanisms for Pitch Perception

In particular. we note that our specific predictions were upheldby the presence of three lateral frontal peaks in the right hemi-sphere (Fig. 2c), only one of which (focus 8) is matched by asymmetric region in the left hemisphere (focus 10). In contrastto the 2-note condition, and also in accord with our predictions,we observed significantly increased CBF in the right temporallobe (focus 6, shown at z = - 6 in Fig. 2C). This finding impliesthat addi tional processing was taking place within the right tem-poral region during performance of the pitch retention task,above and beyond that already accounted for in the passivemelodies condition.In the context of this task, it is reasonable to expect activationof many different cortical regions, as the first/last comparisoninvolves a number of cognit ive processes, in addition to short-term retention. Thus, one may speculate that the numerousfrontal-lobe sites observed might be associated with successfulperformance of distinct aspects of the task. For example, main-tenance of pitch information in working memory might dependon a mechanism separate from that involved with the moreexecutive functions required to monitor the presentation ofthe tones and their temporal order, and to direct the appropriatepitch comparison (see Milner and Petrides, 1984). Similarly.aspects of sustained attention that may rely on frontal-lobemechanisms (Pardo et al., 199 1) may also have been implicatedin the present task. The right inferior colliculus, known to bean important auditory processing structure (Aitkin, 1986) wasalso activated in this subtraction (Fig. 2C, z = - 13) indicatingthat it too is a component of a specialized distributed networkinvolved in pitch memory.This subtraction also yielded a significant CBF decrease inthe left primary auditory region (bottom row of Fig. 2C), nearlyidentical to that found in the 2-note minus passive melodysubtraction. This unexpected finding strongly suggests a com-plex interaction in the physiological processes underlying per-formance of the tasks. As discussed in greater detai l below, wepropose that this finding reflects differential use of perceptualinformation computed at early stages of processing, within theprimary auditory cortices. Since the right primary region is prob-ably specialized for pitch extraction (Zatorre, 1988) the relativeCBF decrease in the left primary region may indicate that in-formation derived from this region is not required for either ofthe pitch-judgment tasks, whereas pitch information specificallyderived from processing in the right primary auditory cortex isutilized.The pattern of CBF change in the first/last condition as com-pared to passive lis tening also yielded significant CBF increasesbilaterally near the junction of the frontal operculum and an-terior insula (second row of Fig. 2(: at z = 6). We have recentlyobtained data on a vocal production task (Perry et al., 1993) inwhich we observed significant insular/opercular activation with-in 2-4 mm of those observed in the present investigation. Thesefindings may therefore reflect the involvement of these regionsin the control of vocal pitch, including possibly subvocal re-hearsal. The latter possibility is consistent with the results ofPaulesu et al. (1993) who report s ignificant CBF increases i nthe insula bilaterally (albeit more posteriorly than in our studies)during covert rehearsal of visually presented letters.Several arcas of significant CBF increase were also found tolie within the parietal lobe: in the inferior parietal lobule bilat-erally, and more superiorly on the right side (third row of Fig.2C, at z = 38 and z = 45) as well as in medial parietal cortex.area 7 (Fig. 2C, z = 45). These results clearly indicate a contri-

bution from a parietal system to the performance of this task,but its precise nature can only be guessed at for the moment. Itis tempting to speculate, given the widely acknowledged role ofinferior parietal regions in spatial processing, that a recoding ofpitch information may be taking place during the performanceof this task. If our observation of CBF changes in area 19 duringpassive lis tening does indeed reflect a visual component to theoriginal percept, then perhaps the hypothesized recoding mightinvolve integration of auditory and visual representations topermit some spatial comparison, which might facilitate the re-quired pitch judgment.Another possibility is that these parietal-lobe changes reflectsome aspect of vigilance or sustained attention, which wouldundoubtedly accompany the first/last comparison. Pardo et al.(199 1) have reported right parietal and dorsolateral frontal CBFincreases in somatosensory and visual vigilance tasks; one ofthe parietal-lobe peaks in the present study (focus 12 in Tab le2) is in close proximity (less than 6 mm) to three of the pointsdescribed by those investigators (one in each oftheir three tasks).Although mere similarity of locat ion is not sufficient to inferfunctional similarity, these comparisons permit us to hypoth-esize that the task used in the present study may have engagedpart of the parietal component of the sustained attention net-work described by Pardo et al. (1991). In addit ion to the rightparietal cortexs putative role in sustaining attention, its specificcontribution to pitch processing mechanisms is partia lly sup-ported by the recent study of Dcmonet et al. (1992) who foundright inferior parietal and temporal activation using a tonal pitchtask. Direct comparison of those data with our own is rendereddifficult by the very different subtraction conditions used; thepar&al sites identified in that study were also more inferiorlyand anteriorly located than those found in the present study.Nonetheless, the fact that right-sided parietal and temporalregions were identif ied suggests that some of the same psycho-logical processes may be impl icated.Finally. among the blood flow decreases in this subtractionwe note a pattern that may be of importance: three regionswithin the medial temporal lobe were ident ified, in the uncusbilateral ly. and in the parahippocampal gyrus on the right (seebottom row of Fig. 2C. z = - 18). Very similar CBF decreaseswere also revealed by a directed search of the 2-note minuspassive melody subtraction: bilateral ly in the uncus (coordinates19. 3, -20, and -29, 5, ~ 15: t values, 3.25 and 3.17, respcc-tively), and in the right parahippocampal region (coordinates29, ~ 19, ~ 18; t value, 2.98). Although these latter points didnot reach our level of statistical significance, their close prox-imity to the significant areas in the other subtraction renderthem unlikely to be spurious. The medial temporal lobe haslong been known to be associated with memory function in bothhumans (e.g.. Milner, 1978) and monkeys (e.g., Mishkin. 1982).We have also obtained evidence from work with temporal lo-bectomy patients that regions within both temporal lobes areimpl icated in melodic memory processes, but that under somecircumstances greater deficits are produced by lesions of theright temporal lobe than of the left (Samson and Zatorre, 199 1b,1992). We may therefore speculate that the CBF decreases ob-served reflect some aspect of mnemonic processing of thesemelodies. The CBF decrease indicates that greater CBF occurredduring passive listening than during either pitch judgment task;it seems plausible that more automatic mnemonic encoding mayoccur during ini tia l listening than when the subjects attentionis drawn to a specific judgment.


10/12


CBF increases within the right frontal cortex. Only in the highmemory-load condition, however, did we observe an additionalCBF increase in the right temporal lobe, beyond that seen inpassive listening. We interpret this result, together with the rightfrontal activation, as evidence that the high memory load im-posed by the first/last task engaged a specialized auditory work-ing memory system, and that this system is instatiated in thebrain via interaction of inferior frontal and superior temporalcortices in the right cerebral hemisphere. This conclusion wouldbe in accord with an earlier study (Zatorre and Samson, 199 l),in which deficits in pitch retention were observed after rightfrontal and/or temporal-lobe lesions. Further converging evi-dence favoring an asymmetric mechanism in pitch short-termmemory comes from recent data collected using magneteno-cephallography (MEG). Kaufman et al. (199 1) report that MEGsuppression time recorded over the right hemisphere correlatedwith stimulus set size for a short-term memory scanning taskusing tones. Moreover, they found an asymmetry in the am-plitude variation of the NlOOm component of the magneticevoked response, which originates in the auditory cortex, sug-gesting that early stages of cognitive processing, prior to memorysearch, are also linked to the right auditory region.

General discussion and conclusionThe tendency for right-asymmetric activation in frontal andtemporal-lobe siteswasobserved n all three tasks; ogether withthe relative decreasen the left primary region n the two-pitchjudgment tasks, hese findings strongly support our contentionof a functional specialization within the right cerebral hemi-sphere or tonal melodic processing,n accord with considerablehuman neuropsychologicalevidence. The data permit us o dis-tinguish between perceptual analysis mechanisms, nvolvingprimarily temporal neocortex, and auditory working memorymechanisms,nvolving complex temporofrontal interactions.A preliminary outline of a model to describe he perceptualprocessing tagemay be suggested.Basedon the results of thepresent study, together with the physiological and lesion iter-ature discussed arlier, we may speculate hat the primary au-ditory cortex is chiefly involved in early stages f processing(which might include computation of such signalparametersaspitch, duration, intensity, and spatial location), whereasmorecomplex feature extraction, involving temporally distributedpatternsof stimulation, is performed via populations of neuronswithin the secondarycortices. Neuronal systemsocated n bothtemporal lobes likely participate in higher-order perceptualanalysisof melodies,but those on the right seem o be partic-ularly important, perhapsbecausehey are specialized o extractthe features hat are most relevant for melodic stimuli (includ-mg, e.g., invariant pitch-interval relationships, and spectralcharacteristics mportant for pitch and timbre perception). Theexistence of temporal-lobe neurons with complex responseproperties e.g., McKenna et al., 1989)would be n keepingwiththis idea.Note that sucha hierarchical schemeneed not imply a serialorganization; indeed, both the anatomy and behavioral datareviewed earlier suggest hat various stages f processingmayoccur in parallel. Furthermore, the presenceof significant CBFdecreasesn the left primary auditory cortex implies that a sim-ple inear additive model, n which a smallnumber ofoperationsare added by each subsequentask without affecting operationsinvolved in earlier stagesPetersenet al., 1988) may be unten-able. Instead, the CBF decreases bserved n the two-pitch judg-ment conditions suggest hat neural processesn the primarycortices may differ, depending upon the ultimate use of theextracted information. Thus, we may hypothesize that duringactive listening, in which pitch information, specifically, mustbe acted upon, there is an interaction between higher-ordermechanisms perhaps involving frontal-lobe structures) andlower-order systems primary cortex), such that only the mostrelevant stimulus eatures are selected or further processing.thas previously been demonstrated that computation of com-plex-tone pitch dependscrucially upon the right primary au-ditory cortex (Zatorre, 1988).We may therefore tentatively con-clude that this feature of the stimulus s most important duringthe pitch judgment task, and that whatever information is de-rived from the left primary auditory region is less relevant,leading, therefore, to a relative decrease f CBF. Whether thisexplanation requires active suppressionof information or notremains to be established.Note that no CBF change positiveor negative) was observed in the right primary auditory regionduring the two pitch judgment tasks. Therefore, it would seemthat the perceptual nformation required wasalready extracted,probably automatically, during the passive istening stage.

In both pitch judgment conditions we observed significant

The presentdata are in accord with a detailed model for theinteraction of frontal and temporal mechanismsproposed byPerry (1990; Marin and Perry, in press),who suggestshat theconnectivity of the superior temporal gyrus with the frontalcortex (Petrides and Pandya, 1988) may be one component ofa distributed neural network that maintains auditory informa-tion in working memory. Thus, sensory nformation would beretained while someother process s carried out (in the presentcase the pitch of the first note is retained while the subjectmonitors for the occurrence of the final note). According to themodel, this processwould require neuronal nteractionsbetweenthe temporal neocortex, which has processed he pitch infor-mation, and portions of frontal cortex, which would activelymaintain the appropriate information until the right time tomake use of it. This conclusion is further supported by recentPET experiments Petrideset al., 1993) mplicating dorsolateralfrontal-lobe mechanisms n monitoring of verbal informationin working memory.

ReferencesAitkin LM (1986) The Auditory Midbrain. Cli fton, NJ: Humana Press.Brugge JF, Reale RA (1985) Auditory cortex. In: Cerebral cortex, Vol4 (Peters A, Jones EG, eds), pp 229-27 1. New York: Plenum.Burton H, Jones EG (1976) The posterior thalamic region and itscortical projection in new world and old world monkeys. J CompNeurol 168:249-302.Celesia G (1976) Organization of auditory cortical areas in man. Brain99:403-414.Chavis D, Pandya DN (1976) Further observations on corticofrontalpathways in the rhesus monkey. Brain Res 117:369-386.Colombo M, DAmato MR, Rodman HR, Gross CG (1990) Auditoryassociation cortex lesions impair auditory short-term memory inmonkeys. Science 247:336-338.Demonet J-F, Chollet F, Ramsay S, Cardebat D, Nespoulous J-L, WiseR, Rascal A, Frackowiak R (1992) The anatomy o f phonologicaland semantic processing in normal subjects. Brain 115: 1753-l 768.Deutsch D (1970) Tones and numbers: spec ifici ty of interference inshort-term memory. Science 168: 1604-l 605.Deutsch D (1982) The processing of pitch combinations. In: The psy-chology o f music (Deutsch D, ed). New York: Academic.Diamond IT, Fitzpatrick D, Sprague JM (1985) The extrastriate visualcortex. In: Cerebral cortex, Vol 4 (Peters A, Jones EG, eds), pp 63-

87. New York: Plenum.


11/12

1918 Zatorre et al. * Neural Mechanisms for Pitch PerceptIon

Divenyi PL. Robinson AJ (198)) Nonlingu istic auditory capabilitie s Milner B. Petrides M (I 984) Behavioural effects of frontal-lobe lesionsin aphasia. Brain Lang 37:290-326. in man. Trends Neurosci 7:403407.

Dowhng WJ, Harwood DL (I 986) Mu sic cognitio n. Orlando, FL: Mishkin M (1982) A memory system in the Monkey. Ph ilos Trans RAcadem ic. Sot Lond [Biol] 29885-95.

Evans AC. Thom pson CJ, Marrett S, Meyer E, Mazza M (I 991a) Morrell F (1973) Vis ual syste ms view of acou stic space. Nature 238 :Performance evaluation of the PC-2048: a new 15-slice encoded- 4445.crystal PET scanner for neurological studies. IEEE Trans Mcd Imaging Pardo JV. Fox PT. Raichle ME (199 1) Localization ofa human system10:90-98. for sustained attention by positron emissio n tomography. Nature 349:

6 l-64.vans AC , Marrett S, Torrescorzo J, Ku S, Collins L (199 1b) MRI-PET correlation in three dimension s using a volume-of-interest (VOI)atlas. J Cereb Blood Flow Metab I l:A69-A78.Evans AC, Marrett S, Neelin P, Collins L. Worsley K, Dai W. Milot S.Meyer E. Bub D (1992) Anatom ical mapping of functional acti-vation in stereotactic coordinate space. Neuroimage 1:43-53.

Evarts EV (1952) Effec t of auditory cortex abla tion on frequency dis-crimination in monkey. J Neurophysiol 15:443-448.

Ferrier D (1876) The functions of the brain. London: Smith Elder.Flechsig PE (1896) Gehim und Se&. Leipzig: Veit.Fox PT. Raichle ME (1984) Stimulu s rate dependence of regional

cerebral blood flow in human striate cortex, demonstrated by positronemission tomography. J Neurophysiol 5 1: 1 109-l 120.

Galaburda A, Sanides F (1980) Cytoarchitectonic organization of thehuman auditory cortex. J Comp Neu rol 190:597-610.

Gottlieb Y, Vaadia E, Abeles M (1989) Single unit activity in theauditory cortex of a monkey performing a short term memory task.Exp Brain Res 74: 139-148.Griffith5 TD, Brown WD (199 1) Focal activation of human left area19 during an auditory task. J Cereb Blood Flow Metab I 1 Suppl 21:s374.

Gross CC, Weiskrantr L (1962) Evidence for dissociatio n of impair-ment on auditory discrimination and delayed response following lat-eral frontal lesions in monkeys. Exp Neurol 5:453-476.

Heffner H E, Masterton B (1978) Contribu tion of auditory cortex tohearing in the monkey (Macaca mulutta). In: Recent advances inprimatology , Vol 1, Behaviour (Chivers DJ, Herbert J, eds). NewYork: Academic.

Iv-ersen SD. Mishkin M (1973) Com parison of superior temporal andinferior prefrontal lesi ons on auditory and non-auditory tasks in rhe-sus monkeys. Brain Res 55:355-367.

Jerison HJ, Neff WD (I 953) Effect of cortical ablation in the monkeyon discrimination of auditory patterns. Fed Proc 12:237.

Kaufman L, Curtis S. Wang J-Z. William son SJ (1991) Changes incortical activity when subjects sca n memory for tones. Electroenceph-alogr Clin Neurophysiol 82:266-284.

Kosslyn SM, Alpert NM, Thompson WL, Maljkovic V, Weise SB, Cha-bris CF, Hamilton SE, Rauch SL, Buonanno FS (1993) Visual men-tal imagery activ-atcs topographically organized visual cortex: PETinvestigations. J Cor Neurosci 5:263-287.

Lauter JL, Herscovitclh P, Formby C, Raichle ME (1985) Tonotopicorganization in human auditory cortex revealed by positron emissio ntomography. Hearing Res 20: 199-205.

Liegeois-Chauvel C, Musolino A, Chauvel P (199 1) Localization ofthe primary auditory area in man. Bra in 114: 139-l 53.

Manley JA, Miiller-Pre uss P (1978) Respo nse variab ility of auditorycortex ce lls in the squirrel monkey to constant acoustic stimu li. ExpBrain Res 32: 17 I-1 80.Marin OSM. Perry DW (in press) Neurological aspects of music per-ception and performance. In: The psychology ofmusic, 2d ed (DeutschD, ed), in press. New York: Academ ic.

Mazziota JC, Phelps ME, Carson RE, Kuhl DE (1982) Tomographicmapping of human cerebral metabolism: auditory stimulation. Neu-rology 3i:921-937.

McKenna TM, Weinberger NM, Diamo nd DM (1989) Respo nses ofsingle auditory cortical neurons to tone sequences. Brain Res 481:142-153.

Merzenich MM, Brugge JF (1973) Represe ntation of the coch lcar par-tition on the superior temporal plane of the macaque monkey. JNeurophysiol 24: 193-202.

Milner B (1962) Laterality effects in audition. In: Interhcmisphericrelations and cerebral dominance (Mountcastle VB, ed), pp 177-195.Baltimore: Johns Hopkins.

Milner B (1978) Clue s to the cerebral organization of memory. In:Cerebral correlates of consciou s experience @user P. Rougeul-BuserA. eds), pp 139-153. Amsterdam: Elsevier.

Paulesu E , Frith CD, Frackowiak RSJ (1993) The neural correlates ofthe verbal comp onent of working memory. Nature 36 2:342-345.Penfield W. Jasper HH (1954) Epilepsy and the functional anatomyof the human brain. Boston: Little, Brown.

Penfield W, Perot P (1963) The brains record of auditory and visualexperience: a final summary and discuss ion. Brain 86:595-696.

Peretz I (1993) Auditory agnosia: a functional analvsis. In: Thinkingin sound: the cognitive psychology of human auditjon (McAdams S.Bigand E. cds), pp 199-230. London: Oxford UP.

Perry DW (1990) Ear and hemisphere diffcrcnces in melody recall.Dissert Abstr Int 52:552B (University microfilms, 91-17914).

Perry DW. Alivis atos B, Evans AC , Meyer E, Petride s M. Zatorre RJ(1993) Neural network supporting auditory-vocal integratio n in sing-ing. J Acoust Sot Am 93:2403.Petersen SE, Fox PT , Posner MI, Mintun M, Raic hle ME (1988) Pos-itron emission tomographic studies of the cortical anatomy of singleword processing. Nature 33 1:585-589.

Petrides M, Pandya DN (1988) Associat ion fiber pathways to thefrontal cortex from the superior tempo ral region in the rhesus monkey.J Comp Neurol 273152-66.

Petride s M, Aliv isato s H, Meyer M, Evans AC (1993) Funct ionalactivation of the human frontal cortex during the performance ofverbal working memorv tasks. Proc Nat1 Acad Sci USA 90:878-882.

Raichle ME, M&in WRW, Herscovitch P, Mintun MA. Markham J(1983) Brain blood flow measured with intravenous H20 15. 11. Im-ulementation and validation. J Nucl Med 24:790-798.Robin DA, Tranel D. Damasio H (I 990) Auditory perception of tem-poral and spectral events in patients with foca l left and nght cerebrallesions. Brain Lang 39:539-555.

Samson S , Zatorre RJ (1988) Discrimination of melodic and harmonicstimu li after unilateral cerebral excisions. Brain Cogn 7:348-360.

Samson S, Zatorre RJ (199 la) Timbre perception after unilateral tem-poral lobcctomy in humans. Sot Ncurosci Abstr 17:867.

Sam son S. Zatorre RJ (I 99 1b) Recognition memory for text and mel-ody of songs after unilateral temporal-lobe lesion: evidence for dualencoding. J Exp Psycho1 [Learn Mcm Cogn] 17:793-804.

Samson S, Zatorre RJ (1992) Learning and retention of melodic andverbal information after unilateral temporal lobectomy. Neuropsy-chologia 30:8 15-826.

Sergent J, Zuck E, Terriah S, MacDonald B (1992) Distributed neuralnetwork und erlying mu sica l sight-readin g and keyboard performance .Science 257:106-109.

Shankweiler D (1966) Effects of temporal-lobe damage on the per-ception of dichotically presented melodies. J Corn Physiol Psycho162:l 15-l 19.

Sidtis JJ, Volpe BT (I 988) Selective loss of complex-pitch or speechdiscrimination after unilateral Icsion. Brain Lang 34:235-245.

Stepien LS. Cordeau JP, Rasmussen T (1960) The effect of temporallobe and hippocampal lesions on auditory and visual recent memoryin monkeys. Brain 83:470-489.

Talairach J, Toumoux P (1988) Co-planar stcreotaxic atlas of thehuman brain. New York: Thieme.Tramo MJ, Gazzaniga M (1989) Discrimination and recognition of

complex harmonic spectra by the cerebral hemispheres: differentiallateralization of acoustic-discriminative and semantic-associativefunctions in auditory pattern pcrccption. Sot Ncurosci Abstr 15: 1060.

Tramo MJ, Bharucha JJ, Musiek FE (1990) Music perception andcognition following bilateral lesions of auditory cortex. J Cogn Neu-rosci 2:195-212.

Whitfield IC, Evans EF (1965) Responses ofauditory cortical ncuronesto stimu li of changing frequency. J Neurophysiol 28:655-672.

Wise RJ, Chollet F, Hadar U, Friston K, Hoffner E, Frackowiak R(1991) Distrib ution of corti cal neural networks involved in wordcomprehension and word retries-al. Brain 1 14: 1803-l 8 17.

Woolse y CN (197 I) Ton otop ic organization of the auditory cortex.


12/12

The Journal of Neuromence, April 1994, 14(4) 1919

In: Physiology o f the auditory system (Sachs MB, ed), pp 271-282. Zatorre RJ. Halpern AR (1993) Eff ect of unilateral temporal-lobeBaltimore: National Educational Consultants. excision on perception and imagery of songs. Neuropsychologia 3 1:Worslcy KJ, Evans AC, Marrett S, Neelin P (1992) A three-dimen- 221-232.sional statistical analysis for CBF activation studies in human brain. Zatorre RJ, Samson S (199 1) Role of the right temporal neocortex inJ Cereb Blood Flow Metab 12:900-9 18. retention of pitch in auditory short-term memory. Brain 1 14:2403-Zatorre RJ (1985) Discrimination and recognition of tonal melodies 2417.after unilateral cerebral excisions. Neuropsychologia 23:3 14 1.Zatorre RJ (1988) Pitch perception of complex tones and humantemporal-lobe function. J Acoust Sot Am 84:566-572.Zatorre RJ, Evans AC, Meyer E. Gjedde A (1992) Lateralization ofphonetic and pitch processing in speech perception. Science 256:846-849.

zatorre, r. j., evans, a. c., & meyer, e. (1994). neural mechanisms underlying melodic perception...

Documents