musicians' working memory for tones, words, and pseudo words 2012

This article was downloaded by: [Mariana E. Benassi-Werke]On: 15 March 2012, At: 14:44Publisher: Psychology PressInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office:Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

The Quarterly Journal of ExperimentalPsychologyPublication details, including instructions for authors and subscriptioninformation:http://www.tandfonline.com/loi/pqje20

Musicians' working memory for tones, words,and pseudowordsMariana E. Benassi-Werke a , Marcelo Queiroz b , Rúben S. Araújo c ,Orlando F. A. Bueno a & Maria Gabriela M. Oliveira aa Department of Psychobiology, Universidade Federal de São Paulo(UNIFESP), São Paulo, Brazilb Computer Science Department, Universidade de São Paulo (USP), SãoPaulo, Brazilc São Paulo State Symphony Orchestra Choir (OSESP), São Paulo, Brazil

Available online: 01 Dec 2011

To cite this article: Mariana E. Benassi-Werke, Marcelo Queiroz, Rúben S. Araújo, Orlando F. A. Bueno &Maria Gabriela M. Oliveira (2011): Musicians' working memory for tones, words, and pseudowords, TheQuarterly Journal of Experimental Psychology, DOI:10.1080/17470218.2011.644799

To link to this article: http://dx.doi.org/10.1080/17470218.2011.644799

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantialor systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that thecontents will be complete or accurate or up to date. The accuracy of any instructions, formulae,and drug doses should be independently verified with primary sources. The publisher shall notbe liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever orhowsoever caused arising directly or indirectly in connection with or arising out of the use of thismaterial.

http://www.tandfonline.com/loi/pqje20

http://dx.doi.org/10.1080/17470218.2011.644799

http://www.tandfonline.com/page/terms-and-conditions

Musicians’ working memory for tones, words,and pseudowords

Mariana E. Benassi-Werke1, Marcelo Queiroz2, Rúben S. Araújo3,Orlando F. A. Bueno1, and Maria Gabriela M. Oliveira1

1Department of Psychobiology, Universidade Federal de São Paulo (UNIFESP), São Paulo, Brazil2Computer Science Department, Universidade de São Paulo (USP), São Paulo, Brazil3São Paulo State Symphony Orchestra Choir (OSESP), São Paulo, Brazil

Studies investigating factors that influence tone recognition generally use recognition tests, whereas themajority of the studies on verbal material use self-generated responses in the form of serial recall tests. Inthe present study we intended to investigate whether tonal and verbal materials share the same cognitivemechanisms, by presenting an experimental instrument that evaluates short-term and working mem-ories for tones, using self-generated sung responses that may be compared to verbal tests. This paradigmwas designed according to the same structure of the forward and backward digit span tests, but usingdigits, pseudowords, and tones as stimuli. The profile of amateur singers and professional singers inthese tests was compared in forward and backward digit, pseudoword, tone, and contour spans. Inaddition, an absolute pitch experimental group was included, in order to observe the possible use ofverbal labels in tone memorization tasks. In general, we observed that musical schooling has a slightpositive influence on the recall of tones, as opposed to verbal material, which is not influenced bymusical schooling. Furthermore, the ability to reproduce melodic contours (up and down patterns) isgenerally higher than the ability to reproduce exact tone sequences. However, backward spans werelower than forward spans for all stimuli (digits, pseudowords, tones, contour). Curiously, backwardspans were disproportionately lower for tones than for verbal material—that is, the requirement torecall sequences in backward rather than forward order seems to differentially affect tonal stimuli.This difference does not vary according to musical expertise.

Keywords: Short-term memory; Digit span; Tones; Melodies.

One important aspect of musical memory proces-sing, the storage of tones, has always been studiedusing recognition tests (Berti, Münzer, Schröger,& Pechmann, 2006; Brown & Martinez, 2007;Deutsch, 1970, 1973; Semal, Demany, Ueda, &

Hallé, 1996; Zatorre, Evans, & Meyer, 1994). Inthese studies, a tone is presented (standard tone)and is compared to another tone presented a fewseconds later (comparison tone). There are alsoother studies that use recognition tests for tone

Correspondence should be addressed to Mariana E. Benassi-Werke, Universidade Federal de São Paulo, Departamento de

Psicobiologia, Rua Botucatu, 862, Vila Clementino, 04023-062, São Paulo, SP, Brasil. E-mail: [email protected]

This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Conselho Nacional de

Desenvolvimento Científico e Tecnológico (CNPq), and Associação Fundo de Incentivo à Pesquisa (AFIP).

# 2012 The Experimental Psychology Society 1http://www.psypress.com/qjep http://dx.doi.org/10.1080/17470218.2011.644799

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY

2012, iFirst, 1–11

Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

sequences, such as Logie and Edworthy (1986) andWilliamson, Mitchell, Hitch, and Baddeley (2010).Studies investigating factors that influence tonerecognition usually focus on the question ofwhether verbal and tone material may (or maynot) have some underlying common processingmechanism. For instance, Logie and Edworthy(1986, p. 36) showed that verbal tasks, combinedwith tone sequences presentation, cause disruptionin tone sequence recognition, suggesting “somefunctional overlap in the mechanisms required forprocessing auditory verbal and auditory non-verbal material”. In Logie and Edworthy’s study,further evidence for the overlap is the susceptibilityof tonal material recognition to articulatory sup-pression. On the other hand, in an older but veryinfluential study, Deutsch (1970) verified thattone recognition tests are severely disrupted byintervening tones, but only a slight performancedecline occurs when the intervening stimuli aredigits. Recently, Koelsch et al. (2009) showedthat articulatory suppression interfered moreseverely with the storage of verbal material thanwith the storage of tonal material, although thereappears to be an overlap in the neural networkinvolved in both verbal and tonal storage asobserved by functional magnetic resonanceimaging (fMRI; Koelsch et al., 2009).

However, the majority of the studies on verbalmaterial use self-generated responses in the formof serial recall tests. Concerning short-termmemory, one of the most used paradigms is thedigit span test in its two forms: the forward spantest and the backward span test.

The forward digit span test is used extensivelyto evaluate verbal short-term system capacity(Gabrieli, Gabrieli, Stebbins, & Sullivan, 1998;Norman, Kemper, Kynette, Cheung, &Anagnopoulos, 1991; Olazaran, Jacobs, & Stern,1996; Saito, 2001; Shebani, Vijver, & Poortinga,2005). In this test, participants are asked to listento sequences of digits of increasing length,waiting until the end of each sequence to repeat itin forward order.

Miller (1956) proposed that in English thehuman ability to process information in span testshas a capacity limit of seven items, plus or minus

two items. However, Cowan (2000) proposedthat this high span is specific to linguistic materialand derives from our ability to chunk information.Thus, a more realistic number would be fourchunks, plus or minus one chunk.

Some factors are known to influence the abilityto repeat words after a single presentation (verbalspan), such as phonological similarity and wordlength effect (Baddeley, 1990). Another factorthat can influence this ability is the semanticcontext in which these words are inserted; forinstance, serial recall in bilinguals is higher forfirst-language words than for second-languagewords (Ardila, 2003; Thorn & Gathercole, 1999;Thorn, Gathercole, & Frakish, 2002).

When the sequence of verbal items to beremembered is presented, and the participant issupposed to repeat it backwards, the manipulationof mnemonic representations of information is gen-erally necessary. Therefore, these tasks are con-sidered to evaluate working memory processesand are not influenced by the same factors thataffect forward span tests, such as word-lengtheffect and phonological similarity, when correct-in-position scoring is used (Tehan & Mills, 2007).

In the present study, we used short-termmemory and working memory paradigms to evalu-ate the extent to which tonal and verbal materialmight share the same mechanisms.

For this purpose, we present an experimentalinstrument that evaluates short-term and workingmemories for tones, using self-generated sungresponses that may be compared to verbal teststhat use spoken responses. This paradigm wasdesigned according to the same structure of theforward and backward digit span tests. One limit-ation of the present paradigm is that the techniquecan only be used with participants that possesssome degree of musical training, in the same waythat a verbal test can only be applied to verballycompetent subjects.

The first purpose of the present study was toverify the performance profile of musicians inforward and backward tone, pseudoword, anddigit spans in serial recall tests. The tone sequenceswere built based on the tones of the chromatic scaleusing unusual tone intervals in order to minimize

2 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2012, 00 (0)

BENASSI-WERKE ET AL.

Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

the influence of previous schemata stored in long-term memory that might be analogous to verbalcontext. As verbal short-term memory is influencedby the meaning of words, a pseudoword span testwas also used in order to provide experimentalverbal material without the interference of semanticcontext, allowing us to verify possible differences inthe two types of material. If tonal and verbalmaterials shared the same mechanisms for proces-sing short-term memory and working memory,the same performance profile would be expectedfor meaningless verbal and tonal material.

A second purpose of the present study was toinvestigate the effect of musical expertise inforward and backward tone span tests, by compar-ing the spans of amateur singers, with little formalmusical education, with those of professionalsingers. It is considered that musical expertise mayinfluence long-term recall because professionalmusicians may use mnemonic strategies, based onstructural aspects of melodies such as harmonyand contour (Dowling, 1994; Schubert & Stevens,2006). Contour spans, also defined in analogy todigit spans, are easily obtained in the analysis ofself-generated sung responses, by modifying theacceptance criterion for a correctly recalledsequence: Only the up-and-down patterns ofmelodic contour are considered in this case, asopposed to strict tone comparison. In addition, anabsolute pitch experimental group was included,in order to observe the possible use of verbal labelsin tone memorization tasks (Itoh, Suwazono,Arao, Miyazaki, & Nakada, 2005; Zatorre, Perry,Beckett, Westbury, & Evans, 1998).

Method

ParticipantsFifty-three participants, aged from 19 to 52 years,were divided into three groups: amateur singers(n= 18), composed of 12 singers from the Choirof Universidade Federal de São Paulo (CoralUNIFESP) and 6 singers from other universitychoirs, who had been engaged in choir singing for4.54 years (SE= 0.96) and also had an averageexperience in music theory of 2.69 years (SE=0.69, see Table 1); professional singers (n= 20),

composed of 18 singers from the OrquestraSinfônica do Estado de São Paulo (OSESP) and2 singers from other two professional choirs fromSão Paulo, who reported an average music theoryexperience of 14.2 years (SE= 1.29); and the absol-ute pitch group (n= 15), which was composed ofprofessional singers from OSESP, singing studentsfrom a music school in São Paulo, and professionalinstrumentalists and composers, with an averagemusic theory experience of 12.93 years (SE=2.19). Absolute pitch was verified through theapplication of an absolute pitch test, which will bedescribed below. None of the participants fromthe first two groups possessed absolute pitch, sothat there was no overlap between groups. Thisstudy was approved by the UNIFESP ResearchEthics Committee.

EquipmentWe used a Samson C-15 microphone, an M-Audio® Mobile-Pre audio interface, and Sony’ssound editor and pitch analyser software SoundForge 8.0®, to record the stimuli material. Testswere reproduced using a Dell® notebook.Participants heard the stimuli through a Philips®professional headphone, and all sessions wererecorded with a Panasonic® portable digital recor-der. The pitch analyser of Sound Forge 8.0® wasalso used for the tone span assessment of eachparticipant.

StimuliThe stimuli were recorded in a professional studio.Two singers (male and female) were asked to singor talk the chosen tones/words/pseudowords.

Forward and backward digit span tests. Forward andbackward digit span tests were built based on theWechsler Adult Intelligence Scale–Third Edition(WAIS–III) Digit Span test. The singers recordedspoken digits from 1 to 9, and the sound signalswere edited in Sound Forge and built intosequences that varied in length from 2 to 9 digits.

In both the forward and backward digit spantests, the first two sequences had 2 digits, thenext two sequences had 3 digits, and so on up tosequences of 10 digits. The position of the digits

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2012, 00 (0) 3

WORKING MEMORY IN MUSICIANS

Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

in the audio sequences was similar to that of theWAIS–III Digit Span test. Each digit had theapproximate duration of 0.5 s and was followedby a silence interval of 0.5 s.

There were two lists of digit sequences, A and B,for the forward and backward digit span tests. Eachlist could be used either as a forward digit span testor as a backward digit span test according to eachexperimental participant, in such a way that forhalf of the participants, List A was used forforward digit spans, and List B was used for back-ward digit spans, and for the other half of theparticipants, the roles of these lists were reversed.

Forward and backward pseudoword span tests.Pseudoword span tests were built similarly todigit span tests, but pseudowords (words with nomeaning, but phonetically similar to realPortuguese words) were used instead of digits.Nine pseudowords were created based on thedigits, modifying some of their phonemes—forexample, timbo instead of cinco (five), daus insteadof dois (two)—and they were recorded by thesame singers. The items of digit span tests werereplaced by their corresponding pseudowords, pre-serving the same underlying sequences of increasinglength. The pseudoword audio sequences were builtusing the software Sound Forge. Each pseudowordhad the approximate duration of 0.5 s and was fol-lowed by a silence interval of 0.5 s. The role reversalof the sequence lists A and B was also applied topseudoword tests.

Forward and backward tone span tests. In order tobuild the tone stimuli, chosen tones were playedon a piano, and the singers had to reproduce themsinging the vowel [a] from the international pho-netic alphabet (Miller, 1986) for approximately 5sfor each tone. After that, the tuning of the recordedstimuli was checked and recorded again if necessary.The recorded tones were edited using Sound Forge,and all stimuli had their amplitudes rescaled inorder to produce equal intensity across all tonesequences. A 0.5-s sample was cut from approxi-mately the middle section of each recorded toneand was pasted into a new file, with very shortfade-in and fade-out ramps to avoid audible clicksin the sequences. After that, the tones were puttogether in predetermined sequences, always using0.5-s silence intervals between them. Nine toneswere chosen out of the 12 notes from the chromaticscale (C, C#, D, D#, E, F, F#, G, G#), rangingfrom C4 to G#4 (for male participants) and fromC5 to G#5 (for female participants). The C4–G#4 range was chosen because it lies in the intersec-tion of male vocal registers (tenors, baritones, andbasses), and the C5–G#5 range is also common tothe female vocal registers (alto, mezzo-sopranos,and sopranos; Miller, 1986).

Each of the nine chosen tones was identifiedwith a digit—for example, C= 1, C#= 2, . . . ,G#= 9, and after that the sequences of theWAIS–III Digit Span test were translated usingthe corresponding tones. In order to avoidmelodic structures that are easily recognized due

Table 1. Mean number of years of age, musical theory study, singing, and choir experience in amateur singers, professional singers, and

absolute pitch groups

Characteristic

Amateurs

(N= 4 M+ 14 F)

Professionals

(N= 7 M+ 13 F)

Absolute Pitch

(N= 8 M+ 7 F)

M SE M SE M SE

Age** 27.78 1.86 34.00 1.50 26.93 1.98

Musical theory* 2.69 0.69 14.2 1.29 12.93 2.19

Singing lessons*** 0.60 0.45 11.70 1.10 5.37 2.24

Choir*** 4.54 0.96 16.13 1.83 5.80 1.44

Note: All values in years. M=males. F= females. M=mean. SE= standard error.

*Amateur singers differ from the other two groups (p, .05). **Professional singers differ from the other two groups (p, .05). ***Three

groups differ among themselves (p, .05).



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

to their frequent use in western musical repertoire,whenever a tone sequence contained an ascendingor descending major triad, one of the notes was ran-domly changed in order to break the triadicstructure.

The tone sequences were built using the softwareSound Forge. Each tone had the duration of 0.5 sand was followed by a silence interval of 0.5 s.These tests followed the same pattern of the digitand pseudoword span tests, using two lists, A andB, of tone sequences of increasing length with rolereversal with respect to forward and backward spantests.

Tune test and absolute pitch test. For the tune testand the absolute pitch test, the nine previouslyrecorded tones (C5 to G#5 for female voices andC4 to G#4 for male voices) were randomly reor-dered to form a nine-tone random sequence. Inthese tests, each participant was supposed torespond to each isolated tone immediately afterhearing it (no serial recall), and for that reason a5-s interval was inserted between each tone.

ProcedureExperimental phase. All the experimental sessionswere thoroughly recorded, and all stimuli wereaurally presented to the participants through head-phones, which were worn throughout the entiresession.

After their consent, participants answered aquestionnaire about demographic characteristicsand musical experience, and with other questionsfor participant screening and classification in exper-imental groups. After that, the tune test and theabsolute pitch test were applied. Subsequently,forward and backward digit, pseudoword, andtone span tests were applied.

In the tune test, participants had to listen andreproduce one tone at a time, singing the samevowel [a] as that presented in the recordings.Participant tones were considered correct whenthe deviation was less than a quarter-tone down-wards or upwards with respect to the correct tone(F0), and they had to show 100% of correctanswers to be considered in tune.

The absolute pitch test was applied using thesame tones as those of the tune test, but in a differ-ent order. The participants had to hear one tone at atime and say the name of the notes, without pre-viously being given any reference note andwithout resorting to any tune-producing device(such as a tuning fork or a musical instrument).

In forward span tests, at the end of eachsequence, participants were asked to reproducethe sequence in forward order, by either speakingor singing, according to the nature of the sequence.

In backward span tests, the presentation of thesequences was similar to forward span tests, butnow the participants were asked to reproduce thesequences backwards.

All participants were exposed to the whole seriesof test sequences, and the test application was ran-domly reordered: After the tune test and absolutepitch test, the sequence of tests was randomlychosen for each participant. No volunteer testswere discarded until the end of the experimentalphase.

Data analysisThe recorded experimental sessions werethoroughly heard, and the span measures werecomputed from the recordings. When twosequences of the same length were incorrectlyrepeated, the analysis of the corresponding test setwas concluded, and the span was determined bythe number of items in the last correctly repeatedsequence.

For the recorded tone sequences, another type ofanalysis was conducted prior to span definition.Each tone sequence reproduced by the participantswas segmented into individual tones, and each tonewas compared to the correct reference tone used inthe original sequences. The spectrum analyser ofthe Sound Forge® software was used for thispurpose. In order to assess whether a recordedtone had been correctly reproduced by the partici-pant, we adopted a quarter-tone tolerance criterion.This corresponds to one half of the smallest tonedifference that exists in a chromatic (or 12-tone)scale and accounts for fluctuations of fundamentalfrequency that occur naturally both with trainedand with untrained singers.



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

The audio signal of each reproduced tone wasfirst trimmed of its initial and final sections, inorder to remove wavering frequencies that arecharacteristic of human vocal emission, especiallyin the attack and release portions of the tone. Thecentral, stable part of the sound was fed into the fre-quency analyser, where the first prominent partial(f0) was inspected. If this frequency fell withinthe range of a quarter-tone downwards orupwards with respect to the correct tone (F0), thesample was considered correct. The same compari-son was carried out between the second measuredpartial (f1) and its true theoretical value (2·F0), inorder to reconfirm the hypothesis and to accountfor possible oscillations (e.g., vibrato) that mightaffect measurements. So, if the first partial wasonly slightly off the tolerance range (a few Hz),and the second partial was near 2·F0 (up to aquarter-tone difference), the sample was also con-sidered correct.

For instance, if the reference tone was an E4=329.63 Hz, and the participant reproduced a waver-ing tone where f0= 340 Hz, and f1= 670 Hz, thenthis sample would be accepted as a correct reproduc-tion, since the range of acceptable f0 values would be[F0/1.0293, … , F0·1.0293] = [320.24 Hz, … ,339.29 Hz], and for f1, this range would be [2·F0/1.0293, … , 2·F0*1.0293]= [640.49 Hz, … ,678.57 Hz].

It should be noted that the above criterion fortone correctness applies only for obtaining tonespan measures. Contour span measures, which cor-respond to the largest sequence of up-and-downpatterns correctly reproduced, were obtained bycomparison of the corresponding binary patterns(e.g.,++ –+ –) in the reference sequence andthe one reproduced by the participant.

Statistical analysisWe used 4 one-way analyses of variance(ANOVAs) to compare each one of the followingparameters between the three experimentalgroups: age, musical expertise (measured in yearsof musical theory study), and singing experience(measured in years of choir practice). TheNewman–Keuls post hoc test was used whennecessary. Traditional schooling was analysed for

the three groups through the Kruskal–Wallis testand then pairwise compared between groupsusing the Mann–Whitney U test. We used thechi-square test to compare sex distributionbetween groups.

Forward and backward span tests were analysedseparately. Forward span results were analysed bytwo-way ANOVA (with one repeated measure),considering two factors: groups, which had threelevels (amateur singers, professional singers, andabsolute pitch musicians), and type of mnemonicmaterial, which had three levels (digit span, pseu-doword span, and tone span). The same analysiswas used for comparing tone and contour spans.The same was done for backward span results.The Newman–Keuls post hoc test was used for allanalyses when necessary.

The significance level adopted was p , .05.

Results

Groups differed in age, F(2, 50)= 5.04, MSE =274.64, p, .05. Newman–Keuls test establishedthat professional singers had a higher mean agethan the other two groups (p, .05). A chi-squaretest showed that there is no difference betweengroups regarding sex distribution. All the aboveresults are presented in Table 1. Groups weredifferent in terms of traditional schooling, H(2,N= 53)= 6.28, p , .05, and the Mann–Whitney test established that the absolute pitchgroup had lower mean schooling than the pro-fessional group, U = 80, p , .05.

Mean musical expertise (in years of musicaltheory study) was different between groups, F(2,50)= 20.24, MSE = 721.16, p, .05, and theNewman–Keuls test established that amateursingers had significantly fewer years of musicaltheory study than the other groups (p, .05).Choir experience (in years) also differed betweengroups, F(2, 50)= 18.64, MSE = 749.84,p, .05. Post hoc tests indicated that the meanchoir experience was higher for professionalsingers than for the other groups (p, .05).Finally, all three groups differed in years ofsinging lessons, F(2, 50)= 18.79, MSE= 589.68,p , .05, and the post hoc test indicated that



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

professional singers had more years of singinglessons than absolute pitch musicians, and thatboth had more years of singing lessons thanamateur singers (p, .05).

Forward test resultsThe first ANOVA comparing groups and the threetypes of material (digit, pseudoword, and tonespans) did not show a group effect, F(2, 50)= 0.24,MSE= 0.53, but there was an effect of type ofmaterial, F(2, 100)= 95.16,MSE= 84.73, p, .05.

There was a significant interaction between typeof material and group in forward tests, F(4, 100)=7.93,MSE= 7.06, p, .05. The analysis of post hoctests for this interaction showed that digit spanswere higher than pseudoword spans for amateursingers, and that spans for both were higher thantone spans (p, .05). For the other two groups,digit spans were higher than the other two spans(p, .05 for both groups), which were themselvesstatistically similar. Analysing each type offorward span separately, it was observed that thethree groups had similar digit spans (p. .05),pseudoword spans (p. .05), and tone spans(p. .05; Figure 1a).

The second ANOVA comparing groups withtype of material (tone and contour spans) showeda group effect, F(2, 45)= 5.081, MSE= 9.625,p, .05, and a Newman–Keuls test establishedthat professionals’ spans were higher than amateurs’spans (p, .05). There was also an effect of type of

material, F(1, 45)= 46.744, MSE= 51.470,p, .05, and a Newman–Keuls test establishedthat contour span was higher than tone span(p, .05). There was no interaction between thetwo factors, F(2, 45)= 1.789, MSE= 1.97(Figure 2a).

Backward tests resultsThe first ANOVA comparing groups and the threetypes of material (digit, pseudoword, and tonespans) did not show a group effect, F(2, 50)=1.12, MSE= 1.63. However, there was an effectof type of material, F(2, 100)= 87.58, MSE=94.59, p , .001.

Results showed an interaction between type ofmaterial and group, F(4, 100)= 6.40, MSE=6.91, p, .05, and a Newman–Keuls test estab-lished that digit spans were higher than pseudo-word spans for amateur and professional singers,and that these were both higher than tone spans(p, .05). For absolute pitch musicians, digitspans were also higher than the other two spans(p, .05), but pseudoword and tone spans for thisgroup were statistically similar (Figure 1b).

Comparing each backward span separately,groups presented similar digit spans (p. .05) andpseudoword spans (p. .05). Tone spans werehigher for absolute pitch musicians than for theother two groups (p, .05).

The second ANOVA comparing groups withtype of material (tone and contour spans) did not

Figure 1. Means and standard errors of digit spans, pseudoword spans, and tone spans of the three groups in (a) forward recall and (b)

backward recall.



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

show a group effect, F(2, 45)= 2.8058, MSE=6.7542. However, there was an effect of type ofmaterial, F(1, 45)= 52.9182, MSE= 55.9136,p, .05, and a Newman–Keuls test establishedthat contour span was higher than tone span(p, .05). There was no interaction between thetwo factors, F(2, 45)= 0.8452, MSE= 0.8931(Figure 2b).

Discussion

Verbal materialThe higher ratings of digit recall over pseudowordrecall may be related to the absence of semanticcontext (Ardila, 2003), or to the influence of fam-iliarity, since digits are likely to exist in long-termmemory whereas pseudowords are not (Baddeley,2000). These possibilities are reinforced by thefact that the length effect (Olazaran et al.,1996),the phonological similarity effect (Baddeley, 1990,2003), and the time for subvocal rehearsal werecontrolled.

Comparison of backward spans for these itemsmight lead us to conclude that working memoryfor meaningful material is better than that formeaningless material, but this conclusion wouldonly be supported by the relative loss of items dueto storage constraints rather than loss due to themanipulation itself, which is one of the main differ-ences in backward and forward recall. Althoughforward and backward recall may be influenced bydifferent factors (Tehan & Mills, 2007), the fact

that backward recall performance decreases simi-larly for digits and pseudowords could be inter-preted as a feature of working memory that iscommon to both types of material, despite theirdifferences in terms of familiarity and long-term-memory representations.

Verbal material× Tonal materialThe verbal stimuli (digits, pseudowords) were usedbecause they have progressively fewer componentsto aid storage: Digits possess meaning, phonetics,and sound; pseudowords do not possess meaning,only phonetics and sound. On the other hand,tones possess only sound (pitch), but their percep-tion also implies another dimension of represen-tation, with interval sizes and contour, which isrelated to up-and-down patterns (Dowling,1994). So it is not possible to equalize the phono-logical structure and the difficulty level of thesetwo kinds of stimuli. However, some overallfactors were controlled, such as the presentationtime, the interval between presentation of eachitem, and the presentation format (two sequenceswith the same number of stimuli).

Melodic contour is an important feature thatinfluences melodic recall, as observed by Dowling(1994). Small variations in contour have the poten-tial to confuse nonmusicians in recognition testsdue to pitch proximity, whereas musical trainingtends to reduce or even suppress this effect(Williamson, Baddeley, & Hitch, 2010). In par-ticular, the very nature of the sung melodic recall

Figure 2. Means and standard errors of tone spans and contour spans of the three groups in (a) forward recall and (b) backward recall.



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

task, which is not so much distant from everydaychoir practice, reinforces the need for some absolutepitch reference to be memorized by the participant,along with the sequence of intervals and generalcontour. Accordingly, participants in this exper-iment never produced transposed responses to anyof the tests.

Despite this fact, melodic contour alone seemsto be easier to memorize, as observed in the analysisof contour spans obtained in the same experiment.This might be explained by the fact that contour isrelational by definition, whereas the other types ofmaterial (digits, pseudowords, tones) are definedby absolute values.

Analysing the meaningless stimuli—that is,pseudowords and tones—in forward tests, weobserve subtle effects due to interaction betweengroup and material, since amateur singers presenta decrease in tone span performance compared topseudoword span performance, a characteristicthat is not observed in professional singers andabsolute pitch musicians. This phenomenon couldnot be due solely to differences in age or choirexperience, since professionals and absolute pitchparticipants were also different regarding thesefactors. However, professionals and AP partici-pants presented similar measures of musical exper-tise, as measured by formal musical education inyears (see Table 1: “years of musical theory”).This fact suggests that musical schooling may influ-ence tone recall, as suggested by Berti et al. (2006).These authors suggest that musicians have anadvantage in storing auditory information, duenot solely to their superior encoding of informationbut also to improved working memory processes,and that this advantage relies on specialized long-term structures built up during musical trainingand practice.

Forward tone spans for professional singers mayseem low, especially if we take into account thatmusicians are used to storing very long melodiesin memory. Surprise at this fact was reported bysome volunteers. The tone sequences of this studywere built from the chromatic scale, which dividesthe octave into 12 equal semitones, and were ran-domly generated and therefore not predictable.This was done to prevent musicians from using

tone schemes prestored in long-term memory.The definition of tone sequences after a randomnonmusical source (i.e., WAIS digit sequences)accounts for the lack of familiarity and predictabil-ity of these sequences, minimizing the impact ofboth schematic and veridical melodic expectations(Bharucha, 1994; Carlsen, 1981; Unyk &Carlsen, 1987). This supports the claim thatmelodic expectations were as controlled as theycould be in an experiment based on the digit spantest, considering the constraint that the set oftone sequences should be the same for allparticipants.

Western music, both classical and popular, islargely based on tonality, which is generally basedon the diatonic scale; this is the basis for thegeneral notion that diatonic sequences are morefamiliar to individuals with western musical train-ing. Although evidence to support the claim thatdiatonic sequences are more memorable than chro-matic sequences is scarce, Bartlett and Dowling(1988) were able to show that the scalar structuredoes influence the perception of tone similarity, aphenomenon they called asymmetric similarity. Itis possible that forward tone spans might be differ-ent if tone sequences were built using other types ofscales, such as the pentatonic or the diatonic scales.Further work is being developed to explore otherfactors that may influence short-term memoryspan for tones.

Absolute pitch participants possess the capacityto associate verbal codes with notes (Zatorreet al., 1998), and they also have the ability to uselong-term memory to recall tones without reverber-ating a pitch continuously to maintain it in short-term memory (Wayman, Frisina, Walton, Hantz,& Crummer, 1992). Those absolute pitch charac-teristics may have influenced the tone recall,explaining the similar profiles of tone and pseudo-word for these participants.

We may observe that manipulation of meaning-less verbal–acoustic items in working memory doesnot seem to be generally more difficult thanmanipulation of meaningful verbal–acoustic items.But the fact that both amateur and professionalsingers present a significant decrease in perform-ance for backward tone and contour recall is an



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

indication that the difficulty in reversing melodiesor contours is higher than the difficulty in reversingpseudowords, and this specific difficulty cannot beattributed exclusively to the meaninglessness ofthese items. This fact might be explained by therelational nature of tone and contour sequences,where the reversal of items also reverses musicalintervals and contour patterns.

In conclusion, we have introduced a novelmethodfor measuring tone spans, using self-generated sungresponses. Although this method requires theability to sing (but not necessarily the knowledge ofmusic notation) and therefore excludes considerationof naïve nonmusicians, it allows a more thoroughcomparison of paired results with other spanmeasures, such as digit span and word span.Despite the exclusion of nonmusicians, we wereable to verify the effect of musical expertise differ-ences in short-term memory through comparison ofamateur singers and professional singers.

Future experiments using the above methodmay address interesting and open issues relative tointerference tests such as articulatory suppressionor irrelevant sound effect, as well as investigationsinto familiarity or other factors that might influenceshort-term memory coding and processing.

Original manuscript received 29 November 2010

Accepted revision received 11 November 2011

First published online 24 February 2012

REFERENCES

Ardila, A. (2003). Language representation and workingmemory with bilinguals. Journal of Communication

Disorders, 36, 233–240.Baddeley, A. D. (1990). The role of memory in cogni-

tion: Working memory. In Human memory: Theory

and practice (pp. 67–95). London: LawrenceErlbaum Associates.

Baddeley, A. D. (2000). The episodic buffer: A newcomponent of working memory? Trends in Cognitive

Sciences, 4(11), 417–423.Baddeley, A. D. (2003). Working memory and language:

An overview. Journal of Communication Disorders, 36,189–208.

Bartlett, J. C., & Dowling, W. J. (1988). Scale structureand similarity of melodies. Music Perception, 5(3),285–314.

Berti, S., Münzer, S., Schröger, E., & Pechmann, T.(2006). Different interference effects in musiciansand a control group. Experimental Psychology, 53(2),111–116.

Bharucha, J. J. (1994).Tonality and expectation. InR.Aiello& J. Sloboda (Eds.), Musical perceptions (pp. 213–239).New York, NY: Oxford University Press.

Brown, S., & Martinez, M. J. (2007). Activation of pre-motor vocal areas during musical discrimination.Brain and Cognition, 63, 59–69.

Carlsen, J. C. (1981). Some factors which influencemelodic expectancy. Psychomusicology, 1, 12–29.

Cowan, N. (2000). The magic number 4 in short-termmemory: A reconsideration of mental storagecapacity. Behavioral and Brain Sciences, 24, 87–185.

Deutsch, D. (1970). Tones and numbers: Specificity ofinterference in immediate memory. Science, 168,

1604–1605.Deutsch, D. (1973). Interference in memory between

tones adjacent in the musical scale. Journal of

Experimental Psychology, 100(2), 228–231.Dowling, W. J. (1994). Melodic contour in hearing and

remembering melodies. In R. Aiello & J. Sloboda(Eds.), Musical perceptions (pp. 173–189).New York, NY: Oxford University Press.

Gabrieli, M. S., Gabrieli, J. D. E., Stebbins, G. T., &Sullivan, E. V. (1998). Working and strategic memorydeficits in schizophrenia. Neuropsychology, 2, 278–288.

Itoh, K., Suwazono, S., Arao, H., Miyazaki, K., &Nakada, T. (2005). Electrophysiological correlatesof absolute pitch and relative pitch. Cerebral Cortex,15, 760–769.

Koelsch, S., Schulze, K., Sammler, D., Fritz, T., Müller,K., & Gruber, O. (2009). Functional architecture ofverbal and tonal working memory: An fMRI study.Human Brain Mapping, 30, 859–873.

Logie, R. H., & Edworthy, J. (1986). Shared mechan-isms in the processing of verbal and musical material.In D. G. Russell, D. Marks, & J. Richardson (Eds.),Imagery II (pp. 33–37). Dunedin, New Zealand:Human Performance Associates.

Miller, G. A. (1956). The magical number seven, plus orminus two: Some limits on our capacity for processinginformation. The Psychological Review, 63(2), 81–97.

Miller, R. (1986). The structure of singing: System and art

of vocal technique. New York, NY: Schirmer.Norman, S., Kemper, S., Kynette, D., Cheung, H., &

Anagnopoulos, C. (1991). Syntactic complexity and



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

adults’ running memory span. Journal of Gerontology:Psychological Sciences, 46(6), 346–351.

Olazaran, J., Jacobs, D. M., & Stern, Y. (1996).Comparative study of visual and verbal short-termmemory in English and Spanish speakers: Testing alinguistic hypothesis. Journal of International

Neuropsychological Society, 2, 105–110.Saito, S. (2001). The phonological loop and memory for

rhythms: An individual differences approach.Memory, 9(4/5/6), 313–322.

Schubert, E., & Stevens, C. (2006). The effect of impliedharmony, contour and musical expertise on judg-ments of similarity of familiar melodies. Journal ofNew Music Research, 35(2), 161–174.

Semal, C., Demany, L., Ueda, K., & Hallé, P. A. (1996).Speech versus nonspeech in pitch memory. Journal ofAcoustical Society of America, 100(2, Pt. 1), 1132–1140.

Shebani, M. A., Vijver, F. J. R. V. D., & Poortinga, Y.H. (2005). A strict test of the phonological loophypothesis with Libyan data. Memory & Cognition,33(2), 196–202.

Sony Pictures Digital Inc. (2005). Sound Forge (version8) [computer software].

Tehan,G.,&Mills,K. (2007).Workingmemory and short-term memory storage. What does backward recall tellus?. The cognitive neuroscience of working memory

(pp. 153–164). Oxford, UK: Oxford University Press.Thorn, A. S. C., & Gathercole, S. E. (1999). Language-

specific knowledge and short-term memory in bilin-gual and non-bilingual children. The Quarterly

Journal of Experimental Psychology, 52A(2), 303–324.

Thorn, A. S. C., Gathercole, S. E., & Frakish, C. R.(2002). Language familiarity effects in short-termmemory: The role of output delay and long-termknowledge. The Quarterly Journal of Experimental

Psychology, 55A(4), 1363–1383.Unyk, A. M., & Carlsen, J. C. (1987). The influence of

expectancy on melodic perception. Psychomusicology,7(1), 3–23.

Wayman, J. W., Frisina, R. D., Walton, J. P., Hantz, E.C., & Crummer, G. C. (1992). Effects of musicaltraining and absolute pitch ability on event-relatedactivity in response to sine tones. Journal of

Acoustical Society of America, 91(6), 3527–3531.Williamson, V. J., Baddeley, A. D., & Hitch, G. J.

(2010). Musicians’ and nonmusicians’ short-termmemory for verbal and musical sequences:Comparing phonological similarity and pitch proxi-mity. Memory & Cognition, 38(2), 163–175.

Williamson, V. J., Mitchell, T., Hitch, G. J., &Baddeley, A. D. (2010). Musicians’ memory forverbal and tonal materials under conditions of irrele-vant sound. Psychology of Music, 38, 331–350.

Zatorre, R. J., Evans, A. C., & Meyer, E. (1994). Neuralmechanisms underlying tone perception and memoryfor pitch. The Journal of Neuroscience, 14(4),1908–1919.

Zatorre, R. J., Perry, D. W., Beckett, C. A., Westbury,C. F., & Evans, A. C. (1998). Functional anatomyof musical processing in listeners with absolute pitchand relative pitch. Proceedings of the National

Academy of Sciences, 95, 3172–3177.



Dow

nloa

ded

by [

Mar

iana

E. B

enas

si-W

erke

] at

14:

44 1

5 M

arch

201

2

musicians' working memory for tones, words, and pseudo words 2012

Documents