Further evidence for an autonomous processing of pitch in auditory short-term memory Catherine Semal

Laboratoirede Psychoacoustique, Universit•de Bordeaux2, 146 rue L•o-Saigna•F-33076Bordeaux ½edex, France

LaurentDemany Laboratoirede Psychoacoustique, Universit•de Bordeaux2. 146 rue L•-Saignat, F-33076 Bordeaux Cedex;Franceand Laboratoired',4udiologie Exp•rirnentale,Unit• INSERM 229, H•pital Pellegrin,Place Am•!ie Raba Ldon, F-33076 Bordeaux Cedex, France

(Received2 June1992;accepted for publication21 April 1993) In four relatedexperiments,subjectshad to discriminatebetweenthe presenceor absenceof a frequencydifferencebetweentwo puretonesseparatedby 4.3 s. The interferenceeffectsof other tones(I}, insertedduringthe retentioninterval,wereinvestigated. A previousstudy[C. Semal and L. Demany, J. Aeoust. Sac. Am. 89, 2404-2410 (1991)] had shown that subjects' performancestronglydependedon the pitchrelationsbetweenthe testtonesand the I tones,but not on the spectralcompositionof the I tones. This suggestedthat the mnemonicsystem processing pitch is deafto sharpness of timbre. In the presentstudy,the I tonescoulddiffer or not from the testtonesin intensity(by + 6 or 15 dB) or in amplitudeenvelope(periodicaswell as aperiodicenvelopes were used). It wasfoundthat suchdifferences had very little effecton performance, suggesting that the mnemonicsystemprocessing pitch is deafto loudness and to dynamicaspects of timbre.However,for I toneswith a densespectrum in thevicinityof thetest tones'frequencies, wideningtheI tones'spectrumimprovedperformance, probablybecause this spectralwideningdecreasedthe salienceof the I tones'pitches.

PACS numbers: 43.66.Hg,43.66.Jh,43.66.Mk [WDW]

INTRODUCTION

In a previousarticle (Semaland Demany, 1991), we reportedevidencethat pitch is processed independently of timbre (spectral composition) in auditory short-term memory.This evidencecamefrom experiments basedon a procedureoriginally usedby Deutsch (see, e.g., Deutsch, 1970, 1982). The subjects'task was to detecta pitch differencebetweentwo test tonesseparatedby an intervalof 4.3 s, duringwhicha sequence of six "interfering"tones(I tones) was presented.The I toneswere similar or dissimilar to the test tones with regard to pitch (periodicity) and/or spectral composition (harmonic content). In agreementwith Deutseh (1972), it was observedthat subjectsperformedmuchbetterwhenthe I toneswereremote in pitch from the test tonesthan when all toneswere close in pitch. However,we also found that performancewas essentiallyunaffectedby the relation betweenthe I tones and test tonesfrom the point of view of speetralcomposition. Thus the results suggestedthat the "pitch memorizer" is basicallydeaf to timbre, or at leastto the aspects of timbrewhichare dependenton the spectraldistribution of energy,such as "brightness"or "sharpness"(yon BismareL 1974).

In the presentstudy, the propertiesof the pitch memorizer were further investigatedwith the sameprocedural paradigm.Again, the two test tonesto be comparedon eachtrial wereeitheridenticalin everyrespector different in frequency.One of our aims was to investigatethe infiu1315

d. Acoust.Soc. Am. 94 (3), Pt. 1, Sept. 1993

enceof an intensitydifferencebetweenthe test tonesand theI tones.In thisregard,threehypotheses cameto mind. First, it could be hypothesizedthat loud I toneswould yield poorer performancesthan soft I tones,all other thingsbeingequal.A secondhypothesis was that the absoluteintensityof theI toneswouldbe lessimportantthan the absolutevalueof the intensitydifferencebetweenthe I tonesand test tones:If pitch is not dissociatedfrom loudnessin auditory memory,then I tonesclosein loudnessto the test tones should have a more deleterious effect on the

pitch comparison taskthan muchlouderas well as much softerI tones.Finally, if the pitch memorizeris deaf not only to sharpness of timbrebut alsoto loudness, then the intensityof the I tonesshouldhaveessentiallyno effecton performance(provided,of course,that the I tonesare at least detectable).

A secondgoal of this studywas to examinethe influence of a differencein amplitudeenvelope(or temporal profile) betweenthe testtonesand the I tones.In principle, the perceptualqualityassociated with the amplitudeenvelope of a tone is to be consideredas a dynamicaspectof its timbre, and thus somethingquite different from pitch. However,periodicenvelopeswith a frequencybelow,say, 50 Hz, may giveriseto "infrapiteh" sensations which have somerelation with pitch sensations(Warren, 1982, Chap. 3 }. We previouslyfound that the pitch memorizeris deaf to static (i.e., spectral) aspectsof timbre. Is it alsodeaf to dynamic aspectsof timbre, and to infrapiteh?

0001-4966/93/94(3)/131518156.00

¸

1993 AcousticalSociety of Amedca

1315

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.

I. EXPERIMENT 1: EFFECT OF INTENSITY CASE OF PITCH PROXIMITY

IN THE

Interferingtones: absent

45 dB

54 dB

60 dB

66 dB

75 dB

(61.0)

(62.1)

(57.4)

(57.6)

(60.1)

3

4

5

6

A. Method •

1. Stimuli and procedure

5o

o

On eachtrial, the subjecthadto makea same/different judgment on two test tones,$1 and S2, which were pure and separatedby 4.3 s. The frequencyorS1 wasrandomly selectedwithin an octave geometricallycenteredon 1000 Hz (the probabilitydistributionbeing rectangularon a log-frequency axis). When S2 differedfrom S1, the differenceconsistedof a 4-4% frequencyshift. Trials were organizedin blocksof 24, including 12 "same" pairs (SI ----S2)and 12 "different"pairs,randomlyordered;for each "different"pair, the directionof the frequencyshift was

• 40

2O

10

I

2

Conditions

selected at random.

The subjectswere run in six conditions.In condition 1--a

control

or "baseline"

condition

the two test tones

had a sound-pressure level (SPL) of 45 dB and no extra tonewaspresentedbetweenthem. In eachof the otherfive conditions, the $PL of the two test tones was 60 dB and a

FIG. 1. Resultsof experiment1, averagedacrosssubjects.For eachcondition, the four bars stand for the four successivetrial blocks. The two

numbersabove each set of bars are respectivelythe overall error rate, expressedas a percentage,and (in parentheses)the proportion of

"misses"(incorrect"same"responses) as a percentage of total errors.

sequence of six pureI toneswasinsertedduringthe 4.3 s retentioninterval. The frequencyof each I tone was randomly selectedamong four possiblevalues, differing by 4-3% or 6% from the frequencyof Sl. The SPL of the I toneswas respectively45, 54, 60, 66, and 75 dB in conditions 2 to 6.

All toneshad a total duration of 300 ms, including linear rise/fall

discardedimmediately after their first block of trials (in condition 1) becausethey made more than three errors in this block. The same selection criterion

had been used in

our previousstudy (Semaland Demany, 1991). Five of the

eightsubjects had alsoparticipatedas subjects in this previousstudy.

times of 10 ms. Consecutive I tones were

separatedby 300-ms silent pauses,and a 500-ms silent pauseseparatedthe two test tonesfrom their respective neighbors,the first and last I tones.The stimuli were generated in real time by a digital signalprocessor(OROS AU22, basedon a TMS320C25 chip), usinga generalsoftware systemfor psychoacoustic research("SON", L.C.I., Paris). They were presentedto the left ear via a TDH 39 earphone.Subjectswere individually testedin a doublewalled soundproofbooth, where they sat in front of a keyboard connectedto the computercontrollingthe experiment. "Same"and "different"responses were givenusing two labeledkeysof this keyboard.Any responseinitiated the next trial after a 1-sdelay. No feedbackwas provided. Subjectsreceivedno preliminaryinformationaboutthe differencesto be detected.They wereinstructedto controlthe intertrial intervalsas they wished (through the delay of their responses),and to ignorethe I tones. Two experimentalsessions were run, on differentdays for eachsubject.They wereorganizedin the samemanner and included two blocks of trials in each of the six condi-

tions. Within every session,the 12 blocks were ordered as follows: (a) condition I (one block); (b) conditions2-6 in a random order (five blocks); (c) condition I (one block); (d) conditions 2-6 in a redetermined random order (five

blocks). Between(b) and (c), subjectstook a restof about 10 rain outside of the booth.

B. Results

In condition1, the performanceof five subjectscould not be measuredin terms of d' (Green and Swets, 1974)

becausethesesubjectsnever responded"different" on a "same"trial. In conditions2-6, by contrast,individuald' statisticsweri alwayscomputable.However,this was not the casefor importantconditionsof the other experiments reportedhere. Thus, as before (Semal and Demany, 1991), performances weremeasuredin termsof error ratesfor all conditions.

Figure1 showsthe meanerrorratefor eachof thefour blocksof trials (two blocksper session X two sessions) run in the six conditions.Not surprisingly,performancewas much better in condition

1 than in the other conditions.

For conditions2-6, an analysisof variance (with subjects as the random factor) revealed no main effect of conditions

[F(4,28) < 1] and blocksof trials [F(3,21)= 1.9, P> 0.10], and no interactionbetweenthesetwo factors [F(12,84) = 1.6, P>0.10]. Thus the intensityof the I toneshad no detectableeffecton performance,within a 30-dB range.It is interestingto note that condition4 of this experiment replicatedalmost exactly a condition used in a previous experiment(Semaland Demany, 1991,experiment1, condition 2), and that the results were extremely similar

(36.3% errors here, versus36.7% in the previousexperiment).

2.. Subjects

Eight subjects,with a meanageof 25 years,provided completedata. Three additionallistenerswere testedbut 1316

d. Acoust. Soc. Am., Vol. 94, No. 3, Pt. 1, Sept. 1993

Figure 1 alsoshows,for eachcondition,the proportion of errors correspondingto "misses"(incorrect "same" re-

sponses).Misseswere systematically more frequentthan C. Semal and L. Demany: Pitch in memory

1316

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.j

falsealarms.Similarasymmetries werefoundby Semaland Demany (1991) and in the other experimentsreported

Interfering tones:

here.

absent

II. EXPERIMENT 2: EFFECT OF INTENSITY CASE OF REMOTE PERIODICITY

H. 1-5

[{. 1-5

H. 28-32

H. 28-32

H. 28-32

45 dB

60 dB

75 dB

4.5 dB

60 dB

75 dB

IN THE

The negativeresultsof experiment1 couldbe ascribed to a ceilingeffectinsofaras performancewasquitebad for all I tones.In experiment2, we re-examinedthe influence of I tones' intensityusingthe samepure test tonesbut complexI tones.Someof the newI toneshad periodsthree timeslongerthan in experiment1, which correspondedto

a musicalpitchabouta twelfthbelowthe pitchof the test tones.For suchI tones,we expectedfrom previousdata (Semal and Demany, 1991) that overall performance wouldb• betterthan in experiment1, thuspossiblyletting "more room" for the emergenceof intensityeffects.However, we also employeda secondclassof I tones,with periodsevenmore remote (by an additionalfactor of 10) from the periodsof the test tones.Their fundamentalfrequencies,always below 50 Hz, were in the infrapiWh

range. ] Although these I tones wereverydistant fromthe test toneswith respectto periodicity,they had a narrow spectrum(lessthan a critical band) in the frequencyvicinity of the test tones. Thus, in addition to an atonal infrapiteh sensation,they elicited a tonal pitch sensation closeto the pitchof the testtones.It washypothesized that because of this pitch,and in spiteof their infrapitch,they would producemore interferenceeffectsthan the I tones with shorterperiods. A. Method

Exceptfor the useof newI tones,the methodwasthe sameas in experiment!. However, sincethere were six types of I tones insteadof five, subjectswere run in six conditions in addition to the control (condition 1).

Each I tone was made up of five consecutiveharmonics,equalin amplitudeand addedin sinephase.In conditions 2, 3, and 4, the I tonesconsistedof harmonics 1 to 5 of a fundamentalfrequencyequal to 1/3 of the frequency of $1, + 3% or 6%. Thus the frequencyrelationbetween S1 and the median harmonic of the I tones (their third

harmonic) wasexactlythe sameas the frequencyrelation betweenSI and the I tonesin experiment1. However,the I tones had a much lower pitch here. The SPL of their spectralcomponentswas respectively45, 60, and 75 dB in conditions 2, 3, and 4. In conditions 5, 6, and 7, the I tones consisted of

harmonics28 to 32 of a fundamentalfrequencyequal to 1/30 of the frequencyof $1, 4-3% or 6%. Thus the frequencyrelationbetween$1 and the medianharmonicof the I tones was the same as in conditions 2-4, but the

presentI tones had a much narrower power spectrum, with unresolvable components, and the perceptualcorrelate of their period--with a minimumvalueof 20.0 ms--was no longer a pitch sensation.Their overall SPL was respectively45, 60, and 75 dB in conditions5, 6, and 7. (In conditions 2 to 7, therefore, all I tones had the same SPL

per critical band.) 1317

a. 1-5

J. Acoust.Soc. Am., Vol. 94, No. 3, Pt. 1, Sept. 1993



25.622.421.726.323.8 19.9 ao............ !?'?:?)--.. 2O

................... 0

:

I

2

3

4

5

6

7

Conditions

FIG. 2. Sameas Fig. I, but for experiment2.

Eight subjects,with a meanageof 21 years,provided completedata. Nine other listenerswere discardedafter their first block of trials, in consequence of the selection rule alreadyusedin experiment1. None of the eight subjectshad previouslyparticipatedin a relatedexperiment.

B. Results

and discussion

Figure2 displaysthe results.The error ratesobtained in conditions2-7 weresubmittedto an analysisof variance with the followingsystematic factors:classof/tones (conditions2-4 vs 5-7), SPL (45, 60, and 75 dB), and block of trials (four blocksper conditionfor each subject,as in experiment1). Thesethreefactorsdid not interactsignificantly (P>0.10 for each F statistic) and SPL was the only factorwith a significantmain effect[F(2,14}=3.7, P=0.05]. The SPL effecthas a paradoxicaldirectionsince the error rate is inverselyrelated to the I tones'intensity. However,as shownby Fig. 2, the magnitudeof thiseffect is rather small.

It was expectedthat fewer errors would be made in conditions2-4 of experiment2 than in conditions2-6 of experiment1. This is confirmedby a Studentt test[t(14) =2.4, P<0.05]. The improvementshouldnot be ascribed to a selectionof "better"subjects i.e., a samplingbias• becausethe presentsubjectsdid not performbetter than the previousonesin condition1 [t(14) < 1]. Subjects' error rate in condition2 (25.6%) is quite consistentwith previous data obtainedin a comparablecondition (Semal and Demany, 1991, experiment1, condition5: 24.0% errors). The positiveinfluenceof SPL is surprising.Maybe the subjectswere simply more alert when the I toneswere loud. But another surprisewas that the two classesof I tonesproducedequivalenteffects.This resultcouldbe interpretedin three differentways. A firsthypothesis wasthat thepitchmemorizeris sensitive to infrapitch as well as pitch. In other words, it would actuallystoreperiodicityinformation.In this memory store,an I tone would stronglyinteractwith a previous test tone if and only if the two toneshave similar periods, C. Semal and L. Demany:Pitchin memory

1317

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.

irrespective of the perceptual quality--pitch or Interferingtones: infrapitch---corresponding to eachperiod. absent Modulation period pure A secondhypothesiswas that the pitch memorizer 33.3 ms 75 ms 150 ms 600 ms doesnot processinfrapitch or periodicityper se, but does 31.8 32.0 34.8 34.5 35.0 processdynamic aspectsof timbre in addition to pitch. (77.1) (74.41 [75.6) [68.1) [6•.8) This meansthat I toneswith a givenamplitudeenvelope wouldproducelittle interferenceon the memorytraceof a test tone with a markedly differentamplitudeenvelope, irrespectiveof periodicityfactors.Such could be the case 2O for the I tonesfrom conditions5-7 sincetheyhad a peaky envelope,elicitinga sensationof roughness,whereasthe 10 test toneswere steadyburstsof sinusoids. 0 Third, it could be hypothesizedthat the I tonesfrom 1 2 3 4 5 6 conditions5-7 yieldedweak interferenceeffectsonly beConditions causethey did not elicit salientpitch sensations in the vicinity of the test tones'pitches.Sincethey were similarto narrow bandsof noise,their pitchescertainlyhad lesssaFIG. 3. Sameas Fig. 1, but for experiment3. liencethanthe pitchesof puretonesin thesamefrequency region (Fastl and Stoll, 1979). This lesserpitch salience mighthavebeenfunctionallyequivalentto a pitchsepara- was respectivelyequal to 33.33, 75, and 150 ms (corretion. If such was the case, it could be maintained that the spondingfrequencies:30, 13.33, and 6.67 Hz). Hence, pitchmemorizerprocesses pitchand nothingbut pitch. giventhat all toneshad a durationof 300 ms, there were The goalof experiment3 wasto confrontthe first and respectively nine,four, and two modulationcyclesfor each secondhypotheses. The third onewastestedin experiment I tone in conditions 2, 3, and 4. Note that the modulation 4. period used in condition2 (33.33 ms} was closeto the averageperiodof the I tones'envelopein conditions5-7 of III. EXPERIMENT 3: EFFECT OF DIFFERENCES IN experiment2 (30 ms). AMPLITUDE ENVELOPE In condition5, the periodof the sinusoidalmodulation function was increasedto 600 ms, i.e., twice the duration of ConsiderI toneswith clear pitchesin the vicinity of was the test tones'pitches.Accordingto the first hypothesis eachI tone.Therefore,theI tones'amplitudeenvelope actually aperiodic. The initial phase of the modulation stated above, the test tones should be easier to discriminate to themaximumpower,andits final when the I tones'amplitudeenvelopegivesthem a clear functioncorresponded phase to the minimum power (40 dB below).The soundof periodicityin the infrapitchrangethan when suchis not the I tones was similar to that of tappedcrystalglass. the case.Accordingto the secondhypothesis, in contrast, In condition 6, the depth of the modulationfunction it will not matterthat the I tones'envelopeis periodicor was reduced to zero. Thus the I tones wereactuallysteady not, the only importantfactorbeingthat thisenvelopedifpure tones, identical to those used in condition 4 of experfersor not from that of the test tones.In the presentexperiment,thesetwo conflictingpredictionswere testedusing I tonesconsistingof sinusoldswhich were amplitude modulatedin a periodicor nonperiodicmanner. A. Method

iment 1.

Sevenlisteners,with a meanageof 28 years,provided completedata.Two of themhad previouslyparticipatedas subjectsin a relatedexperiment.Nine additionallisteners were discarded after their first block of trials.

Six conditionswere run, exactly as in experiment 1 exceptthat the SPL of the testtoneswasnow 60 dB in all conditions and that new I tones were used. The I tones

were amplitude-modulatedsinusoidswith the same rms pressureas the test tones.Their carrier frequencieswere selectedin the same way as the I tones' frequenciesin experiment 1. In conditions2-4, the amplitudemodulationwas periodic. The instantaneousamplitude of each I tone could be written

A(t) = 10sin(2•rgt) sin(2rrft),

( 1)

g being the modulation frequencyand f the carrier frequency.Thus the logarithmof the amplitudeenvelopevaried sinusoidallyand the spanof powervariationwas40 dB. The modulationfunctionstartedand endedat a positivegoingzero crossing.In conditions2, 3, and 4, its period 1318

J. Acoust.Soc.Am.,Vol.94, No. 3, Pt. 1, Sept.1993

B. Results

and discussion

The errorratesobtainedin conditions 2-6 (seeFig. 3) did not differ significantlyfrom eachother, as shownby an overallanalysisof variance[F(4,24)= 1.7, P> 0.10]. Performance was poor in each of these five conditions. An examination of Fig. 3 suggeststhe presenceof a weak monotonictrend from condition2 to condition6; however, the error rate for condition2 did not differ reliably from the error rate for condition6 [F(1,6) =3.4, P>0.10]. Interestingly, performance was significantlyworse [t(13)

=2.2, P<0.05] in condition2 of this experimentthan in condition 6 of experiment2, where the I tonesalso had a

narrow spectrumin the vicinity of the test tones'frequencies,the sameintensityas the testtones,and a periodicity C. Semaland L. Demany:Pitchin memory

1318

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.j

closeto 30 ms. In condition6 of the presentexperiment, the errorratewasequivalentto that foundin the identical conditionof experiment1 (condition4). In usingI toneswith amplitudeenvelopes whichdifferedin variouswaysfrom that of the test tones,our goal here was to determinewhether subjects'discrimination performance wouldbe influenced by the presence of a periodicityin the I tones'envelope.Accordingto the results, not only is this periodicityfactor unimportant,but it is unimportantthat the I tones'envelopediffersor not from that of the test tones.

Interfering tones: absent

pure

H. 59-{il

H. 56-64 H. 45-75 H. 45-75 -SdB



40

37.5

35.4

(76.81

(91.4[

............... I-'.......i-........ ...... ......

20

10

In Sec.II B, threehypotheses wereput forth to explain why the two classes of I tonesusedin experiment2 had equivalenteffectson performance. In the light of experiment 3, the first two hypotheses appearto be inadequate. The third hypothesis wasthat the I tonesusedin conditions5-7 of experiment2 yieldedweakinterference effects becausetheir pitches(in the vicinity of the test tones' pitches)werenot very salientdue to the width of their spectrum.Their spectrumwasmadeup of fivecomponents with equalamplitudes;for the averageI tone,the spacing of the components was 33.3 Hz (1000/30 Hz) and the spectrumthus extendedfrom 933.3 to 1066.7Hz. In experiment3, by contrast,the I tones'spectralenvelope was not rectangular and had a peakcorresponding to the carrier frequency(doseto 1000Hz); for the widestspectra-thoseusedin condition2--the spacingof the spectralcomponents was30 Hz, whichis closeto 33.3Hz; however,for the averageI tone (carrier frequency1000 Hz) of this condition,the 970- and 1030-Hz componentswere about 2.6 dB belowthe 1000-Hz component,and the 940- and

1060-Hz components wereabout 9.0dBbelow it.2Therefore, it seemslikely that the pitchesof the I tones were

1

2

3

4

5

6

Conditions

FIG. 4. Sameas Fig. 1, but for experiment4.

eachcondition,the relativephasesof the harmonicswere fixed within each session,but randomly varied acrossses-

sionsand subjects.Note that the I tonesof condition4 (harmonics56-64) had exactlythe samebandwidth--i.e., 0.193 octave--asthe I tonesof conditions5-7 in experiment 2 (harmonics 28-32). In conditions3 and 4, where the I tones' bandwidth did not exceedone critical band, the SPL of the I tones was

equalto that of the testtones,namely,60 dB; the I tones were thus similar in loudness to the test tones. In condi-

tions 5 and 6, the I tones'bandwidthexceededone critical

band;for condition5, we usedagaina 60 dB $PL; for condition 6, the I tones' SPL was decreasedby 5 dB in order to match the I tones and test tones for loudness (see

Zwicker and Fastl, 1990, p. 189, Fig. 8.7).

more salientin experiment3 than in conditions5-7 of The data were obtainedfrom eight listenerswith a experiment 2, whichwouldexplainwhy moreinterference meanageof 24 yearsandno previousexperience in related effectswerefound in experiment3. experiments. Five additionallisteners werediscarded after IV. EXPERIMENT

4: EFFECT OF SPECTRAL

WIDTH

The purposeof experiment4 wasto checkthat for I toneswith densespectrain the vicinity of the test tones' frequencies, subjects' performance depends on thewidthof the I tones' spectra. A. Method

As in experiment3, the testtoneswereat 60 dB SPL in all conditions.

In condition2, the I toneswere pure. This condition wasidenticalto condition4 of experiment1 and condition 6 of experiment3. In conditions3-6, each I tone was made up of equalamplitudeharmonicsof a fundamentalfrequencycorrespondingto 1/60 of the frequencyof S1, 4-3% or 6%. Thus the I tones' periods randomly varied, in the infrapitehrange,between40 and 90 ms.The numberof har-

their first block of trials.

B. Results

and discussion

An inspectionof Fig. 4 showsthat the error rate was largerin conditions2 and 3 than in conditions4-6. An overallanalysisof variancefor conditions2-6 confirmed that these conditionsyielded reliably different performances[F(4,28)=28.0, P<0.001]. Subsequent contrasts gavenegativeresultswhencondition2 wascomparedto condition3 [F( 1,7)- 1.8, P>0.10] and when conditions 4-6 were comparedto eachother [F(2,14)=3.1, 0.10>P > 0.05], whereasthe differencebetweenconditions2 and 3 andconditions 4-6 appeared to bequitesignificant [F(1,7) =57.0, P<0.001]. It shouldbe noted,however,that the proportion of "misses" wasmarkedlylargerin condition3 than in condition2, and markedly smaller in condition6 than in conditions 4 and 5. For each condition, we com-

monies varied from condition to condition, but the rank of the median harmonic was always 60. In condition 3, we

puteda d' statisticfrom the two percentages givenin Fig. 4; d' was respectivelyequal to 0.75, 1.17, 1.82, 2.01, and

used three harmonics, with ranks 59-61; in condition 4, there were nine harmonics,with ranks 56-64; in conditions

1.75 in conditions2 to 6. Thus, when assessedin terms of

d', performance did not improveabruptlyfromcondition3

5 and 6, 31 harmonies were used, with ranks 45-75. For

to condition 4; its variation was more gradual.

1319

d. Acoust.Soc. Am., Vol. 94, No. 3, Pt. 1, Sept. 1993

C. Semal and L. Demany:Pitchin memory

1319

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.

memory.It now appearsthat the "pitch memorizer"is deafnot only to the "static,"i.e., spectral,aspectsof timbre (Semal and Demany, 1991), but also to its dynamic asperiments1 and 3. Another satisfactioncomesfrom the pects,and to loudness(althoughloudnessmay havea nonsimilarityof the error ratesobtainedin condition4 of the effecton alertness andthusmemoryperformance). presentexperiment andcondition6 of experiment 2, where specific the I toneshad the sameSPL and, more importantly,the Other resultssupportingthe ideathat the pitch memsame bandwidth. orizer is deaf to timbre were recentlypublishedby KrumTherefore,the resultsfully confirmedour hypothesis: hansland Iverson (1992). In their experiment3, subjects Wideningthe I tones'spectraimprovedperformance. We werepresentedwith two melodicsequences of seventones believethat the causeof this improvementis that the saand had to detecta pitch differencebetweenthe fourth lieneeof the I tones'pitcheswas inverselyrelatedto the components of thesesequences. The melodiccontextof the width of their spectra.Admittedly,an alternativehypoth- fourth tone (i.e., the pitchesof the othersix tones)could esiscanbeproposed. From condition2 to conditions5 and be the sameor differentfor the two sequences. In the "dif6, the I tones'pitch saliencedecreasedbut, in addition, ferent context" condition,subjectswere forced to focus

As expected, theerrorrateobtainedin condition2 was

similar to that obtained in the identical conditions of ex-

these tones differed more and more from the test tones with

their attention on the fourth tones since the contexts did

respectto timbre;theirtimbrewasmoreandmore"noisy." The improvementin performancemay be due to the changein timbre rather than to the changein pitch salienee.However,this is not very likely in the light of experiment3 andof our previousstudy(SemalandDemany, 1991). Experiment3 showedthat largedifferences in temporal envelope--i.e.,in the dynamicaspectof timbre--

not provideusefulmelodiccues.In the "samecontext" condition,however,melodiccueswereavailableand it appearedthat the subjectsusedthem becauseperformance

between the test tones and I tones were not sufficient to

simulationsof notesproducedby variousmusicalinstru-

reduceinterference effects.Our previousstudyshowedthat largedifferences in spectralcomposition---i.e., in the static aspectof timbre--were alsoineffective(when the I tones as well as the test toneshad very salientpitches).

ments:piano,trumpet,etc.Equivalentperformances were

V. GENERAL

DISCUSSION

In the four experimentsreportedhere, it was found that the frequencydiscriminationof two pure testtones,a few seconds apart,is impairedto variousdegreesby I tones inserted between the two test tones. The overall results

may be summarizedas follows.

( 1) The amountof impairmentis not cruciallydependent on the I tones'intensity.In any case,the impairment is not a monotonicallyincreasingfunction of their inten-

was better. In both conditions,the timbre of the tones was

eitherconstantthroughouteachtrial or variedwithin the sequences; in the latter case,the timbrai differences between tones were multidimensional

since the stimuli were

measuredfor constantand variedtimbres.Thus pitch ap-

pearedto be processed independently of timbrein the memorizationof melodiesas well as singletones.

Krumhansland Iversonpoint out that the tempoof their sequences wasrelativelyslow (aboutthreetonesper second) and that probably different resultswould have been obtainedat much faster tempi, due to streamsegregationphenomena. A fast sequence of toneswidelydifferent in timbre tendsto be perceptuallysegregatedinto separate streams,each containingtoneswith similar timbres (see,e.g.,van Noorden,1975;Singh,1987). In suchconditions, the interferenceeffect of a tone on the memory

traceof a previoustone'spitch may dependon the timbre of the interferingtone as well as its pitch. Of course,our sity. Moreover,the impairmentis not larger when the I own resultsshouldalsonot be overgeneralized with regard tonesand testtonesare similar in loudnessthan when they to timing parameters. Moreover, it should be notedthat markedlydiffer in loudness(experiments1 and 2). pitch did not appear to be perfectly dissociable from timbre (2) Similarimpairmentsare foundwhen the I tones in reaction time studies where subjects had to categorize have the sameamplitudeenvelopeas the test tonesand quicklythe pitch of a singletone (Crowder, 1989;Melara when their envelopeis different (experiment 3). In this regard,thepresence or absence of an infrapitchperiodicity and Marks, 1990; Krumhansl and Iverson, 1992) or in in the envelopedoesnot matter:Whereasweakerinterfer- somestudieswherethe task consistedin pitch comparisons between two consecutive tones with the same timbre or enceeffectsare observedfor I tonesremotein pitch from different timbres (Ritsma, 1966; Moore and Glasberg, the testtonesthan for I tonesclosein pitch, the presenceof 1990; Moore et al., 1992; Singh and Hirsh, 1992; but see an infrapitchperiodicityin the I tones'envelopeis not also Hafter and Richards, 1988). sufficientto yield weak interferenceeffects(experiments3 Is the pitch memorizerdeafto infrapitch?Point (2) of and 4). the summaryof resultsgivenaboveis consistent with this (3) For I toneswith densespectrain the vicinity of view. However, for pitchescorresponding to frequencies the test tones'frequencies, the amountof interferenceis between,say,70 and 300 Hz, it remainsto be determined inverselyrelatedto spectralwidth (experiment4). This is "atonal"pitches(induced probablydueto the effectof spectralwidth on the salience if the pitchmemorizerprocesses of pitch: A lossof pitch salienceseemsto have the same by periodicamplitudeenvelopesof unresolvablesound complexes)as well as "tonal" pitches(inducedby comeffectas a pitch separation. which are resolvableby This set of resultssignificantlyaugmentsthe previous plexeswith Fourier components thecochlea)? Presently, onedoesnotknowwhether these evidence (Semal and Demany, 1991) that pitch is processedin an autonomous manner in auditory short-term two kinds of pitchesare analyzedby one and the same 1320

J. Acoust.Soc. Am., Vol. 94, No. 3, Pt. 1, Sept. 1993

C. Sernaland L. Demany:Pitchin memory

1320

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms

sensorydeviceor by two completelydifferentsensorydevices(see,e.g.,Houtsmaand Smurzynski,1990;Carlyon et al., i[992;Langneret al., 1992).In thelattercase--orin either case--twodifferentmemorydevicesmight be involved..

Is the mnemonicprocessing of pitchcompletely autonomous'? A previousstudyby Deutsch(1978) suggests that the an.,iwer is negative.Usingpure test tonesand I tones, shefoundthat her subjects'performancewasbetterwhen theI tonesandtesttoneswereheardthroughdifferentears than when all toneswere presentedto the sameear. This wouldsuggest that the pitchmemorizeris not deafto spatial location.However,the spatiallocationeffectobtained by Deutschwasquite small.In a similarbut more recent study (Kallman et al., 1987), a small advantagefor the "different ears" condition was found when the I tones were

presentedto a fixed ear during blocksof trials, but this advantagedisappeared whenthe ear presentedwith the I tonesvariedfrom trial to thai in an unpredictable manner. Finally,it remainspossible thatspeech sounds areprocessedby a specificmemorysystemwhich would be sensitive, amongother things,to their pitches;the pitchesof

nonspeech sounds wouldbe processed by a separate memory system.This is not unlikely because,whereasthe left

only two additional components("side tones") are introducedin the

powerspectrum.In experiment3, more sidetoneswere producedbecausewe modulatedsinusoidally the logarithmof the amplitudeenvelope.

3Unresolvable butperiodic (or quasiperiodic) sound complexes areable to inducesensations of pitch sinceit is possibleto identify, in musical terms,the melodicintervalformedby the periods(or quasiperiods)of two suchsounds(Burnsand Viemeister,1976;Houtsmaand Smurzynski, 1990). However,as shownby Burnsand Viemeister(1976) with sinusoidally amplitudemodulatednoise,the identification of musicalintervalsbecomes muchmoredifficultfor periodsexceeding15 or 20 ms; this supportsthe notionthat 60 Hz can be considered as a frequency boundarybetweenpitch and infrapitch (cf. footnote 1).

Bismarck,G. von (1974). "Sharpnessas an attribute of the timbre of steady sounds," Acustica 30, 159-172.

Burns, E. M., and Viemeister,N. F. (1976). "Nonspectralpitch," J. Acoust. Soc. Am. 60, 863-869.

Carlyon,R. P., Demany,L., and Semal,C. (1992). "Detectionof acrossfrequencydifferences in fundamentalfrequency,"J. Acoust.Soc.Am. 91, 279-292.

Crowder,R. G. (1989). "Imageryfor musicaltimbre,"J. Exp. Psychol.: Human Percept.Perform. 15, 472-478. Deutsch,D. (1970). "Tonesand numbers:Specificityof interferencein immediatememory," Science168, 1604-1605. Deutsch,D. (1972). "Mapping of interactionsin the pitch memory store," Science 175, 1020-1022.

Deutsch,D. (1978). "Interferencein pitchmemoryasa functionof earof input," Q. J. Exp. Psychol.30, 283-287. Deutsch,D. (1982). "The processingof pitch combinations,"in The

cerebralhemisphere is generallydominantfor the processing of speech,the neuralsubstratumof short-termmemory Psychology of Music,editedby D. Deutsch(Academic,London),pp. for the pitchof nonspeech soundsseemsto be localizedin 271-316. the rig,«themisphere(Zatorre and Samson,1991). In one Fastl,H., andStoll,G. (1979). "Scalingof pitchstrength,"Hear. Res.1, of her studieson pitch memory,Deutsch (1970) showed 293-301. that, indeed,the frequencydiscrimination of two puretest Formby,C. (1985). "Differentialsensitivityto tonalfrequencyand to the rate of amplitudemodulationof broadbandnoiseby normallyhearing toneswas much lessaffectedby I soundsconsistingof listeners," J. Acoust. Soc. Am. 78, 70-77. spokennumbers thanby I sounds consisting of puretones.

However, these twosetsOfI sounds probably differed in pitch range,and the differencein pitch rangemay have been the only causeof the observedeffect. In our laboratory, new expehmentsare now underwayin order to investigatethis issue. ACKNOWLEDGMENTS

Thtiswork was supportedby the ConseilR6gional d'Aquitaine.Author LD is affiliatedwith the Centre National de la RechercheScientifique.We are grateful to AnneDucouraufor her cooperation in experiment 4 andto Ken McAnally and Kazuo Ueda for commentson a previous versionof the typescript.Thanks are also due to Jean-MarcDeshouillersfor his help in the computationof some power spectra.

•Studies suchas thosereported by Formby(1985) andGuttmanand Pruzansky(1962) suggest that sensations inducedby periodsexceeding 17 ms--which correspondsto a frequencyof 60 Hz--should not be consideredas "pitch" sensations.Formby (1985) measureddifferential

thresholdsfor pure tonesthat variedin frequencyand for broadband noisethat variedin the frequencyof sinusoidalamplitudemodulation;he found lower thresholdsfor the pure tonesthan for the noisewhen the

Green, D. M., and Swets,J. A. (1974). Signal Detection Theory and Psychophysics (Krieger, Huntington, NY). Guttman, N., and Pruzansky,S. (1962). "Lower limits of pitch and musicalpitch," J. SpeechHear. Res. $, 207-214. Hafter, E. R., and Richards, V. M. (1988). "Discrimination of the rate of

filteredimpulses,"Percept.Psychophys. 43, 405-414. Houtsma,A. J. M., and Smurzynski,J. (1990). "Pitch identificationand discriminationfor complextoneswith many harmonics,"J. Acoust. Soc. Am. 87, 304-310.

Kallman, H. J., Cameron,P. A., Beckstead, J. W., and Joyce,E. (1987). "Ear of inputas a determinantof pitch-memoryinterference,"Memory Cognit. 15, 454-460. Krumhansl,C. L., and Iverson,P. (1992). "Perceptualinteractions be-

tweenmusicalpitch and timbre,"J. Exp. Psychol.:Human Percept. Perform. 18, 739-751.

Langner,G., Schreiner,C., and Albert,M. (1992). "Tonotopyandperiodotopyin the auditorymidbrainof cat and guineafowl," in Auditory Physiology and Perception, editedby Y. Cazals,L. Demany,and K. Horner (Pergamon,Oxford), pp. 241-248. Melara, R. D., and Marks, L. E. (1990). "Perceptualprimacyof dimensions:Supportfor a modelof dimensional interaction,"J. Exp. Psychol.: Human Percept.Perform.16, 398-414. Moore, B.C. J., and Glasberg,B. R. (1990). "Frequencydiscrimination of complextoneswith overlappingand non-overlapping harmonics,"J. Acoust. Soc. Am. 87, 2163-2177.

Moore, B.C. J., Glasberg,B. R., and Proctor,G. M. (1992). "Accuracy

of pitchmatchingfor puretonesand for complextoneswith overlapping or nonoverlappingharmonics,"J. Acoust. Soc. Am. 91, 34433450.

standardfrequencyexceeded60 Hz, but identicalthresholdsat or below 60 Hz. Guttman and Pruzansky(1962) askedtheir subjectsto adjust

Noorden,L. P. A. S. van (1975). "Temporalcoherence in theperception of tone sequences,"Doctoral dissertation,TechnischeHogeschool,

two successive periodicpulsetrainsonemelodicoctaveapart;theyfound that "downwardfrom about 60 Hz, octaveperceptionis weak," and concludethat 60 Hz is the lowerfrequencylimit of "musical"pitch.

Eindhoven, The Netherlands. Ritsma, R. J. (1966). "The 'octave deafness'of the human ear," Ann. Prng. Rep. 1, 15-17, Institute for PerceptionResearch,Eindhoven,The

:Whent]heamplitude envelope of a puretoneis sinnsoidally modulated, 1321

J. Acoust.Soc.Am.,Vol. 94, No. 3, Pt. 1, Sept.1993

Netherlands.

C. Semaland L. Demany:Pitchin memory

1321

nloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms

Semal,C., and Demany,L. (1991). "Dissociationof pitch from timbrein "auditory short-termmemory," J. Aeoust. Soe. Am. 89, 2404-2410. Singh,P. (1987). "Perceptualorganizationof complex-tone sequences: A tradeoffbetweenpitch and timbre?,"J. Acoust.Soc.Am. 82, 886-899. Singh,P. G., and Hirsh, I. J. (1992). "Influenceof spectrallocusand F0 changeson the pitch and timbre of complextones,"J. Aeoust.Soc.Am. 92, 2650-2661.

1322

J. Acoust.Soc. Am., Vol. 94, No. 3, Pt. 1, Sept. 1993

Warren,R. M. (1982). AuditoryPerception (Pergamon,New York). Zatorre, R. J., and Samson,S. (1991). "Role of the right temporalneocortexin retentionof pitchin auditoryshort-termmemory,"Brain 114, 2403-2417.

Zwicker, E., and Fastl, H. (1990). Psychoacousties (Springer-Verlag, Berlin).

C. Semal and L. Demany:Pitch in memory

1322

wnloaded 29 Jun 2010 to 193.50.102.40. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms

Further Evidence for an Autonomous Processing of ...

For such I tones, we expected from previous data .... ous data obtained in a comparable condition (Semal and ... license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.jsp ..... "Mapping of interactions in the pitch memory.

761KB Sizes 2 Downloads 102 Views

Recommend Documents

Further evidence of male trimorphism
Here I discuss male trimorphism in this beetle in detail. This paper is dedicated to Professor Yasuaki WATANABE in com- memoration of his retirement from Tokyo University of Agriculture. ... to 0.1 mm with a slide caliper. To analyze the polymodal fr

pdf-54\an-alien-harvest-further-evidence-linking-animal-mutilations ...
... apps below to open or edit this item. pdf-54\an-alien-harvest-further-evidence-linking-animal ... bductions-to-alien-life-forms-by-linda-moulton-howe.pdf.

Embedded Hardware Design For An Autonomous Electric ... - GitHub
Mar 9, 2011 - Department of Electrical and Computer Engineering. University of .... At peak load, the steering motor can draw up to 10Amps continuously.

Evidence for an additional effect of whole-body ...
Sep 3, 2010 - Summary The addition of whole-body vibration to high- load resistive ...... Conflicts of interest Dieter Felsenberg acts as a consultant to the.

Mariokart An Autonomous Go-Kart - GitHub
Mar 9, 2011 - Make a robust platform for future projects ... Nice hardware platform for future years. - Project almost stuck to time ... Automation Software. 45d.

Further Evidence That Congenitally Blind Participants ...
Congenital blindness is one of the rare human models to explore the role of experience-driven cross-modal compensation after early sensory deprivation.

012 Further Evidence On External Auditors' Reliance On Internal ...
Retrying... Whoops! There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. 012 Further Evidence On External Auditors' Reliance On Internal Auditors.pdf. 012 Furthe

Semantic and perceptual processing of number symbols-evidence ...
Furthermore, processing of the visual simi- larity of both digits and ideographs was associated with activa- tion of the left fusiform gyrus. Using culture as an independent. variable, we provide clear evidence for differences in the brain. regions a

An Evidence Framework For Bayesian Learning of ...
data is sparse, noisy and mismatched with test. ... In an evidence Bayesian framework, we can build a better regularized HMM with ... recognition performance.

Design and Field Testing of an Autonomous ...
sors for estimating the vehicle's state with respect to its environment and with reference to some .... position with respect to the locally consistent submaps defined during route profiling. The inputs .... Given the dynamic and unstructured nature.

A position-sensitive Stroop effect: Further evidence for a ...
2109, Australia (e-mail: [email protected]). ...... batch faith rays wash ghost whisk bath path rich kiss gift skip bit zip riff cows goats stick boat sash.

Learning of Tool Affordances for Autonomous Tool ...
plan a strategy for target object manipulation by a tool via ... through its motor actions using different tools and learning ..... Robotics and Automation (ICRA).