select acoustic and perceptual measures of choral formation · 11.5 years of solo vocal training;...

Aspaas/McCrea/Morris/Fowler (2004) International Journal of Research in Choral Singing 2(1). Choral Formation 11

Select Acoustic and Perceptual Measures of Choral Formation

Christopher Aspaas, Christopher R. McCrea,Richard J. Morris, and Linda Fowler

Florida State University

Abstract

The purposes of this study were (1) toacoustically compare three contrasting choralformations (block sectional, mixed, and sectionalin columns) in performance of two contrastingpieces of choral literature (homophonic andpolyphonic) by a university graduate choir(N=30), and (2) to compare singers’ perceptionsof and preferences for the contrasting choralformations across vocal part and gender. Bothlong-term average spectrums (LTAS) andperceptual chorister ratings were examined fordifferences. Acoustic results demonstrated nodifferences between the formations. However,LTAS differences attributable to the homophonicand polyphonic literature selections wereobserved. Chorister ratings indicated significantgender differences across many of theperception and preference items surveyed,particularly with respect to the sectional incolumns formation.

Choral music educators have long soughtthe means to create a blended sound within theirensembles. Knutson (1987) defines this blendas “…a product of sound where each elementbecomes unified or homogenized,” asdetermined by the conductor. In an articlecomparing the acoustical properties of choraland solo singing, Goodwin (1980) stated,“Achieving choral blend is often a fundamentalobjective in ensemble singing, but there exists adiversity of concepts concerning how thephenomenon occurs, resulting in a profusion of

pedagogical techniques.” Researchers haveexamined a variety of possible influences,including choral formation, in an effort to identifyor better understand the elements of choralblend.

Many conductors believe that the formationor arrangement of singers greatly influences theblend of the ensemble. One of the earliestexperiments in this area was conducted byLambson (1961), who placed members of anintact choral ensemble in several contrastingformations (block sectional, quartets/sextets,scrambled, and random distribution) and had thechoir perform homophonic and polyphonicworks. Ten judges were directed to listen to thechoruses live and via audiotape. The judgeswere directed to determine the choral formationsfor each example and to rank each formation inorder of preference. Judges had difficultyidentifying the formations from the sound of thelive and recorded performances for bothhomophonic and polyphonic selections.Moreover, they exhibited mixed preferences forthe choral arrangements. In fact, Lambson(1961) concluded that the formation of the choirdid not result in clearly perceived differences. Hestated that other, practical factors should beused in selecting choral arrangements.

The Lambson (1961) study also exhibitedcertain methodological problems. First, therewas no mention of how long all of the examplestook to perform, and whether or not there wasany consideration of fatigue for either performersor judges. Second, judges indicated that both

http://www.choralresearch.org


the lack of familiarity with the process ofidentifying choral formation from sound aloneand the requirement of written commentsbetween examples hampered their ability tofocus their attention on the task at hand. In spiteof these issues, however, the judges did agreewith the findings of the study in a post-experiment questionnaire. It would have beeninteresting if Lambson (1961) had examined thechoristers’ perceptions of the differentarrangements to compliment the perceptions ofjudges listening to the choruses.

Tocheff (1990) compared how mixed andsectional formations may have influencedlistener perception for a variety of perceptualaspects of singing, including intonation andoverall blend. The sectional formation was ratedhigher than the mixed for intonation and forchoral blend. In addition, Tocheff (1990)examined the effect of “acoustic placement” ofsingers on overall choral performance ofhomophonic and polyphonic selections. Theplacement of voices by choral conductors, alsoknown as voice matching, is a highly subjectiveprocedure that uses pitch, intensity, timbre,vowel formation, and vibrato rate to placesingers of the same voice part next to othersingers throughout their section (Giardiniere,1991). In contrast to Lambson’s (1961) findings,Tocheff’s (1990) results suggested that theacoustical placement of voices had a greatinfluence on the dependent variables of overallblend, intonation, and achievement of a choralensemble. Again, however, this studydemonstrated some methodological problems,as more than thirty performances were judgedand fatigue may have been an influence.

In his 1996, 1999, and 2003 studies,Daugherty examined the effects of choralformation and spacing on the perceptions ofchoristers and listeners. Daugherty (1999)placed an intact ensemble of high schoolstudents (N=46) in one of two formations,sectional blocks or mixed, as well as threedifferent levels of spacing; close, lateral (morespace side to side), and circumambient (bothincreased lateral and front-to-back space). Boththe auditors (N =160) and the choristerspreferred the sound produced with widerchorister spacing. In addition, although auditorsexpressed no clear formation preference, thesingers preferred the mixed formation over theblock sectional formation. One of the mostinteresting points about this research was that95.6% of the choristers reported that they feltspacing influenced choral sound. Such

measures of chorister preferences may help usunderstand the effects of differences related tochoral formations. Hence, further study ofchorister preferences and perception iswarranted.

Moreover, previous research regarding theeffects of formation on choral blend has beenbased upon perceptual measures of sound.Examination of the acoustical signal producedby a chorus may provide researchers withvaluable information regarding how to arrangean ensemble to generate the best choral blend.For example, spectrograms have been used tomeasure both individual voices (Hunt, 1970) andan ensemble demonstrating high- and low-formant modes of singing (Ford, 2003). In anearly spectrographic inspection of choralsingers, Hunt (1970) examined “good” and “bad”examples of singing by individuals in the choir.He discovered that the examples rated byjudges as “good” contained formant frequenciesthat were tuned to the natural overtone series forthe given fundamental frequency. Conversely,“bad” examples were found to contain formantfrequencies that were misaligned, implying thathearing and tuning vowel formant frequenciesbetween singers may be an important part of theblending process. One may infer from thisresearch that measurements of well-blendedchoral ensemble singing may have more clearlyaligned and reinforced formant frequencies,whereas less blended examples would contain adifferent and more scattered spectrographicsignature.

The single instance of correlatingspectrographic analysis of a choral ensemblewith perceptual measures was performed byFord (2003). He examined auditor preferencesfor either strong or weak presence of a singer’sformant across mode of singing. Cleardifferences between the spectrograms emerged,with auditors preferring the weak singer’sformant examples of singing. This difference inspectrograms demonstrated this tool’s potentialefficacy for measuring different modes of singingwithin ensembles. Unfortunately, the singerswere moved between examples to allow formore evenly biased recordings for the auditors’perceptual evaluation. These changes in thedistance from the microphone to the singersbetween the singing modes could alter thespectrographic results. In addition, it should benoted that Ford (2003) reported that thealteration of singing modes affected pitchaccuracy and vibrato rate. These factors may


have influenced auditor preference for the weaksinger’s formant examples.

The studies cited above have not providedacoustic information that compares a chorusacross different choral arrangements for avariety of musical selections. Although someprevious studies indicate that different choralarrangements and different musical selectionsmay affect listeners’ perceptions of the quality ofthe choral sound, there are little data onchorister perception of formation, with theexception of Tocheff (1990) and Daugherty(1996; 1999; 2003). Such information couldassist conductors in their use of differenttechniques of choral arrangement and singerplacement. This study was designed to makesuch acoustic comparisons and to assess theperceptions of the choristers concerningdifferent choral arrangements.

Thus, the first purpose of the present studywas to acoustically compare three contrastingchoral formations: block sectional, mixed, andsectional in columns. It was hypothesized that amode of singing more close to the solo modewould be present in the mixed formation incomparison to the block sectional and sectionalby columns formations. The expected soloisticmode for the mixed formation would be evidentas a relative increase of acoustic energybetween 2500 and 3000 Hz when comparingpieces sung at similar dynamic levels (i.e., pp,mf, ff). This frequency region of increasedenergy has been reported as a commonacoustic feature of solo singers (Sundberg,1973; 1974; 1977; Schutte & Miller, 1983).

The second purpose of this study was tocompare the choristers’ perceptions of andpreferences among the contrasting choralformations. Previous research has shown thatchoristers exhibit a preference for singing inmixed or synergistic formation over a blockformation (Daugherty, 1999; 2003). However,Daugherty (2003) also suggested that choristersmay simply prefer the formation that they aremost accustomed to. A chorister’s preferencesmay influence an individual member’s quality ofsinging (Ford, 2003), which could alter theacoustic silhouette of the ensemble sound.Based on Daugherty’s (1999) finding, it washypothesized that the choristers would prefer themixed formation over the two block formations.

METHOD

Participants

The participants for this study included anintact, functioning choir of 30 choristers,consisting of 18 female and 12 male singers,ranging in age from 21-53 years (M=32.5 years).The choir was divided into four singing parts,which included sopranos (N=8), altos (N=10),basses (N=7), and tenors (N=5). All participantswere trained singers who reported an average of11.5 years of solo vocal training; the reportedyears of training ranged from 5 to 15 years. Allparticipants reported being in good generalhealth with no vocal difficulties at the time ofrecording.

Equipment

Recordings of the choir were completed in aconcert hall with a seating capacity of 480occupants. Room reverberation for this studywas determined using a Goldline GL-60 ReverbTime Meter. A sound source was introduced andreverberation was measured using the – 0 dBSPL reference point at 2000 Hz. A stablereverberation measurement was achieved at thissetting, allowing other frequencies to be reliablymeasured. All measurements were obtainedwith the meter directly under the position of themicrophone used for the recordings and thesound source produced at the position of thechorus on the stage. Repeatable reverberationmeasurements were made at the followingfrequencies and levels: 1.3 second at 125 Hz,1.85 second at 250 Hz, 2.4 second at 500 Hz,2.5 seconds at 1000 Hz, 2.3 seconds at 2000Hz, and 2.4 seconds at 4000 Hz. The volume(V) of the hall was calculated to be 1844 cubicmeters and the average reverberation time (T)was measured at 2.4 seconds. Using theequation rr=.056↔�V/T, the reverberation radius(rr) was 1.55 m.

The choir was digitally recorded using twoSchoeps CMC-5 omni condenser microphonessuspended from the ceiling of the concert hallapproximately 4 meters above the floor and 10meters in front of the chorus. The choir’sproductions were recorded on an Ampex 467(R-60) DAT recorder through a Soundcraft 220-B mixing board. The DAT recordings were thendigitized to a Dell Optiplex GX1P computer via aCSL 4300B (Kay Elemetrics) at a samplingfrequency of 44.1 KHz. The files were stored


and acoustically analyzed using the CSL 4300Bcomputer hardware/software system.

PROCEDURE

The chorus was recorded when singing thefollowing choral selections: Jubilate Deo byOrlando di Lasso (Theodore Presser #352-0080)in its entirety, and rehearsal numbers six to eightof Glor ia by Lars Edlund (Walton MusicCorporation WH-103). These compositions wereselected because they were familiar to the choirand they represented two contrasting textures:the polyphonic selection, Jubilate Deo, becauseof the independent lines of music for eachsection and the homophonic, Gloria, for thehomophonic singing by all parts together.

The musical selections were sung acappella with the choristers on stage facing theseats and standing on five sections of portableWenger standing risers, which were 18 incheswide and four feet long. The choristers stood inthree different choral formations: blockedsectional, mixed, and sectional in columns (seeFigure 1). These formations are commonlyutilized by choral directors and have been thefocus of previous research dealing with theperceptual effects of choral formation(Daughtery, 1999; Lambson, 1961; Tocheff,1990). The block sectional formation involvedplacing the singers in three rows. The front tworows consisted of sopranos to the left of theconductor and altos to the right. The back rowconsisted of basses placed behind the sopranosand tenors behind the altos. The mixedformation also involved placing the singers inthree rows; however, unlike the block sectionalformation the singers were grouped together inquartets, each containing one soprano, bass,tenor, and alto. Because of the unequal numberof choristers in each part, only five of the sevenquartets included all four voice parts. Theremaining singers were asked to stand next tosomeone of a different voice part. The sectionalin column formation involved grouping thesingers into two or three rows, depending on thenumber of singers in each category. Then,according to voice part, the groups wereorganized so that each vocal part was next toeach other, with no section behind another. Thesopranos and basses were to the left of theconductor, and the tenors and altos were to theright of the conductor.

Figure 1. Three formations according to voicepart.

The choristers sang both of the musicalselections three times, once in each formation.The order of the pieces in each formation wasrandomized. At the end of the hour-longrecording session, the choristers tookapproximately 10-15 minutes to complete a five-item questionnaire utilizing an anchored 10-cmscale designed to determine the choristers’perception of the various formations. Beloweach item were the anchor words and three linesmarked ‘A’, ‘B’, and ‘C’ for the three choralformations, block, mixed and column,respectively (See Figure 2). The perceptualitems were designed to determine the chorister’srating of the particular choral arrangement. Thefive perceptual items and their anchor wordswere as follows. The first item was “In whichchoir formation did you find it easiest for you tosing your part?” with anchor words ‘Difficult and‘Easy.’ The second item was “How would yourate the sound of each choir formation?” withanchor words ‘Worst’ and ‘Best’. The third itemwas “How would you rate each choir formationregarding the ease of hearing/blending with your

B T

BlockSectional

S A

S A

SATB SATB SATB SATB

Mixed BTAS BTAS BTAS BTAS

SATB SATB SATB SATB

S B T A

Sectional inColumns S B T A

S B T A


section?” with anchor words ‘Worst’ and ‘Best’.The fourth item was “How would you rate thechoir formations regarding the ease of hearingother sections?” with anchor words ‘Worst’ and‘Best’. The fifth item was “Which choir formationdid you prefer singing in?” with anchor words‘Worst’ and ‘Best’. The choristers wereinstructed to mark an ‘X’ on the anchored linesto rate each of the formations across each of thequestions.

1. In which choir formation did you find iteasiest for you to sing your part?

Difficult Easy

A _________________________

B _________________________

C _________________________

Figure 2. Example question from the choristersurvey.

Data Analyses

For acoustic data, the long term averagespectra (LTAS) for each choral formation andmusical selection were visually inspected todetermine any differences in energy pattern.Specifically, energy peaks at the fundamentalfrequency, first formant, and 2-3 kHz werevisually examined for differences among thethree formations. Any noticeable differences inLTAS pattern underwent significance testingusing a one-way analysis of variance (ANOVA).

The length of the samples for LTAS analysisvaried slightly across formation and musicalselection. For example, the Gloria excerpt lengthwas 76.79 seconds for the block arrangement,75.69 seconds for the mixed formation, and74.07 seconds for the sectional by columnformation. On the other hand, the lengths of theJubilate Deo selection were 87.44 seconds inthe block formation, 84.54 seconds in the mixedformation, and 83.73 seconds in the sectional bycolumn formation. For the perceptual data a Chi-squared analysis of variance by ranks (χ2) wasperformed to determine whether or not thesingers’ perceptions significantly differed among

the choral formations by vocal part (Siegel &Castellan, 1988).

Reliability

To minimize conductor variability, anexperienced conductor was asked to conduct inthe same manner over all six experimentalconditions. His performance was videotaped.Three experienced choral conductors viewed thevideotape and completed the ConductorConsistency Observation Form (Madsen andYarbrough; 1980). Several items on this formwere adapted so that the focus of the evaluationwas on conductor constancy across conditions.

Results from the conductor consistencytests showed an overall average of 87%interrater agreement (SD=0.07). These resultsare similar to those of Ekholm (2000), whoadapted this method to assess choir conductorreliability. Agreement among the raters was highacross all of the categories, including 100%agreement for magnitude and expressiveness ofconducting gestures, eye contact, and facialexpression.

RESULTS

Acoustic Findings

The LTAS for each of the choral formationsexhibited strong agreement in their distributionof power across the frequency band. As shownin Figures 3 and 4, the slopes of the three linesare almost identical. Each formation showedmaximum energy in the bandwidth range of 500-600 Hz, followed by a decline in energy until the2-3 kHz bandwidth range where anapproximately 5 dB increase of energy wasobserved. These findings are consistent withprevious acoustical analyses associatedexamining the LTAS patterns during choralsinging (Rossing et al., 1986, 1987; Ternström,1991). The greater energy in the 2-3 kHz regionwas not great enough to be considered asinger’s formant.

Further visual inspection of the LTASrevealed differences in the energy distributionacross frequency bands between thehomophonic and polyphonic selections used inthis study. Figure 5 depicts the two separateLTAS of the homophonic and polyphonicselections. The LTAS of the homophonicselections displayed higher energy levels thanthe polyphonic selections for the frequenciesbelow 650 Hz and lower energy levels for the


frequencies above 650 Hz. For example, theLTAS for the polyphonic selection displayedmaximum energy in the frequency bandcentered at approximately 517 Hz, followed by asteep decrease in energy, while the homophonic

selection’s peak energy was centered atapproximately 603 Hz. In addition, the LTAS forthe homophonic selection displayed less energyin the singer’s formant region of 2-3 kHz rangethan the LTAS for the polyphonic selection.

0 1000 2000 3000 4000 5000

Frequency (Hz)

Po

wer

level

10 d

B p

er

div

isio

n

Block

Mixed

Column

Figure 3. Long-term average spectre (LTAS) of the block sectional, mixed and sectional in columnsformations for the monophonic selection, Gloria

0 1000 2000 3000 4000 5000

Frequency (Hz)

Po

wer

level

10 d

B p

er

div

isio

n

Block

Mixed

Column

Figure 4. Long-term average spectre (LTAS) of the block sectional, mixed and sectional in columnsformations for the polyphonic selection, Jubilate Deo


0 1000 2000 3000 4000 5000

Frequency (Hz)

Po

we

r le

ve

l1

0 d

B p

er

div

isio

n Polyphonic Block

Polyphonic Mixed

Polyphonic Column

Homophonic Block

Homophonic Mixed

Homophonic Column

Figure 5. Long-term average spectre (LTAS) of the block sectional, mixed and sectional in columnsformations for both the homophonic and polyphonic selections.

Perceptual Findings

In general, all choristers gave fairly highratings for all of the items, with average scoreswithin the vocal parts above 5.0 (See Table 1).Exceptions to that trend were the sopranosrating of the sectional by columns formation forsound of the choir (M=4.08), ease of hearingothers (M=2.39), and preferred formation forsinging (M=3.73). In addition, the altos rating ofthe sectional by columns formation for ease ofhearing other vocal parts (M=4.80, and preferredformation for singing (M=4.55), were also belowaverage. Finally, the basses’ rating of the mixedformation for ease of blending within section wasbelow average with a score of 2.82.A trend observed across the perceptual itemsoccurred for the sectional by columns formation.Analysis of these data revealed that the

chorister ratings for this formation variedsystematically across vocal part. Thesedifferences in average ratings of sectional bycolumns formation were significant across thevocal parts (χ2 (3)=35.18; p < .001). Vocal partdifferences for the perceptual items acrossformation are displayed in Figure 6. Thesopranos gave the lowest average rating of 4.56(SD=1.74) for the sectional by columnsformation, accompanied by the altos with anaverage rating of 5.95 (SD=1.36). The tenors, onthe other hand, rated the sectional by columnsformation rather favorably at 7.53 (SD=1.09),followed by the basses, who gave the highestaverage rating of 8.47 (SD=1.43).


Table 1

Choristers’ mean preference ratings and standard deviations for perceptual questions by choralformation and vocal part.

Note. Standard deviations are enclosed in parentheses.

Figure 6. Choristers’ average ratings of choral formations across perceptual questions by vocalsection.

ChoralFormation Vocal Part (N)

Ease ofProduction

Sound ofChoir

Ease ofBlendingWith Section

Ease ofHearingOthers Prefer Singing

BlockSoprano (8) 8.07 (2.10) 7.08 (1.61) 6.76 (3.04) 5.78 (2.61) 5.89 (2.18)Alto (10) 8.13 (1.83) 7.10 (1.88) 8.39 (1.71) 5.75 (2.30) 6.20 (2.76)Tenor (5) 8.77 (1.06) 6.05 (2.88) 6.71 (3.45) 5.23 (4.16) 6.22 (3.56)Bass (7) 7.99 (1.42) 6.80 (1.70) 7.62 (1.62) 5.83 (1.83) 6.63 (1.55)

MixedSoprano (8) 5.28 (2.60) 7.69 (2.22) 6.00 (2.89) 7.10 (3.52) 7.21 (3.35)Alto (10) 7.23 (2.55) 6.80 (2.34) 6.20 (2.93) 9.20 (0.73) 8.68 (1.60)Tenor (5) 6.42 (3.52) 6.51 (3.64) 6.05 (3.34) 8.13 (3.88) 7.09 (3.77)Bass (7) 5.01 (1.94) 6.26 (3.38) 2.82 (2.58) 8.63 (1.60) 6.94 (3.43)

ColumnSoprano (8) 6.81 (2.99) 4.08 (2.13) 5.78 (3.40) 2.39 (2.20) 3.73 (3.88)Alto (10) 7.56 (2.38) 5.68 (1.69) 7.18 (2.18) 4.80 (2.78) 4.55 (2.61)Tenor (5) 8.38 (1.6) 7.81 (1.96) 8.68 (2.33) 6.22 (3.21) 6.57 (3.91)Bass (7) 9.64 (0.23) 8.66 (1.62) 9.36 (0.75) 6.04 (2.74) 8.66 (0.97)


As indicated by Figure 6, soprano and altomean responses displayed similar patternsacross formations, with higher responses for theblock sectional and mixed formations comparedto the sectional in column formation. Thesopranos average rating across questions forthe block sectional formation was 6.72(SD=0.94) and 6.66 (SD=0.99) for the mixedformation, whereas the average rating for thesectional in columns formation was 4.56(SD=1.74). Likewise, the altos average ratingsacross questions were 7.11 (SD=1.16) for theblock sectional, 7.62 (SD=1.27) for the mixedformation, and 5.95 (SD=1.36) for the sectionalin columns formation.

The tenor and bass mean responsesdisplayed a pattern that differed from that of thefemale vocal parts across the formations. Malechoristers rated the sectional in columnsformation higher than the other formations at8.47 (SD=1.43). The tenors’ average ratingsacross questions were 6.60 (SD=1.33) for theblock sectional, 6.84 (SD=0.81) for the mixedand 7.53 (SD=1.09) for the sectional in columnsformation. Similarly, the basses’ average ratingsacross questions were 6.97 (SD=0.85) for theblock sectional, 5.93 (2.17) for the mixed and8.47 (SD=1.43) for the sectional in columnsformation.

Given the differences in ratings for thesectional by columns formation across vocalparts, further analysis was conducted todetermine if the differences reported acrossvocal part displayed any trends. Overallreference ratings across voice part were similarfor the block sectional and mixed formations, butdiffered along gender lines with respect to thesectional in column formation. As reported inTable 2, the sopranos and altos rated thesectional in columns formation significantly lowerthan basses and tenors (χ2 (1)=30.93; p <.001).No significant gender differences were observed

for the mixed (p >.05) or block formations (p>.05).

In terms of average ratings for each item,the first question related to “ease of production”was rated fairly high for the block sectionalformation across all of the vocal parts.Responses ranged from 7.99 for the basses to8.77 for the tenors (See Table 1). The lowestmean ratings for “ease of production” occurredfor the mixed formation. Scores ranged from5.01 for the basses to 7.23 for the altos. Themean ratings for the sectional by columnsformation across the ‘ease of production’ itemwere fairly high, from 6.81 for the sopranos to9.64 for the basses.

As depicted in Figure 7a, the male andfemale choristers rated the block sectional andmixed formations similarly, but the sectional bycolumns formation differently. Although the malechoristers favored the sectional by columnsformation more than the females, no significantgender differences were observed here (p >.05)(See Table 2).

Next the “sound of the choir” was rated. Thevocal part means were more similar across theformations. For the block sectional formation themeans varied from 6.05 for the tenors to 7.10 forthe altos (See Table 1). For the mixed formationthey varied from 6.26 for the basses to 7.69 forthe sopranos. For the sectional in columnsformation means varied from 4.08 for thesopranos to 8.66 for the basses.

A gender based difference for “sound of thechoir” again occurred for the sectional incolumns formation (χ2 (1)=12.06; p=.001).Figure 7b displays that the males rated thesectional by columns formation higher than thefemales, whereas the females rated the blockand mixed slightly higher than then males (SeeTable 2).


Table 2

Choristers’ mean preference ratings and standard deviations for perceptual questions by choral formationand gender.

Note. Standard deviations are enclosed in parentheses

For the “ease of hearing/blending with theirown vocal section” item, gender baseddifferences occurred for both the mixed andsectional by columns formations (χ2 (1)=7.82;p=.005). As shown in Figure 7c, the averagefemale chorister ratings were fairly consistent at7.58 for the block sectional, 6.10 for the mixed,and 6.48 for the sectional in columns formation.In contrast, the average male chorister ratingsvaried more widely among the formations from7.17 for the block sectional to 4.44 for the mixedto 9.07 for the sectional in columns formation(See Table 2). The low rating for the mixedformation was strongly based on the lowresponses of the basses at 2.82. The othervocal sections exhibited similar ratings of 6.00for the sopranos, 6.20 for the altos, and 6.05 forthe tenors. The average ratings of the blocksectional varied from 6.71 for the tenors to 7.62for the basses. For the sectional in columnsratings the range was from 5.78 for the sopranosto 9.36 for the basses.

The trend for responses to the “ease of hearingother sections” item differed from thoseobserved for the other items reported thus far.The average responses for both males andfemales to this question showed that the highestaverage ratings occurred for the mixed formation(male=8.38, female=8.15) and the lowestaverage ratings occurred for the sectional incolumns formation (male=6.13, female=3.59)(See Table 2). As can be seen in Figure 7d, thegender differences for the ratings of thesectional in columns formation were significant(χ2 (1)=4.22; p < .040). Furthermore, theaverage ratings of the “ease of hearing othersections” item for the different choral formationswere from 5.23 among the tenors to 5.83 amongthe basses for the block sectional formation,from 7.10 among the sopranos to 9.20 amongthe altos for the mixed formation, and from 2.39among the sopranos to 6.04 among the bassesfor the sectional in columns formation

ChoralFormatio

n Gender (N)Ease of

ProductionSound of

Choir

Ease ofBlending

With Section

Ease ofHearingOthers

PreferSinging

Block

Male(12) 8.31 (1.29) 6.49 (2.18) 7.24 (2.44) 5.58 (2.87) 6.46 (2.44)

Female(18) 8.10 (1.89) 7.09 (1.71) 7.66 (2.46) 5.76 (2.37) 6.06 (2.45)

Mixed

Male(12) 5.60 (2.66) 6.36 (3.33) 4.17 (3.23) 8.42 (2.63) 7.00 (3.40)

Female(18) 6.36 (2.68) 7.20 (2.27) 6.11 (2.83) 8.26 (2.56) 8.02 (2.56)

Column

Male(12) 9.11 (1.18) 8.31 (1.74) 9.08 (1.55) 6.11 (2.80) 7.79 (2.69)

Female(18) 7.23 (2.61) 4.97 (2.01) 6.56 (2.79) 3.73 (2.75) 4.19 (3.16)

.


Sound of the Choir

0

2

4

6

8

10

Block Scatter Column

Ra

tin

gs

a) Male (red square) and female (bluediamond) average ratings of choralformation for the question, “In whichformation did you find it easiest for youto sing your part?”

b) “How would you rate the sound of eachchoir formation?”

Ability to Blend

0

2

4

6

8

10


Rati

ng

s

Ability to Hear Others

0

2

4

6

8

10


Ra

tin

gs

c) “How would you rate each choir formationregarding the ease of hearing/blending with yoursection?”

d) How would you rate the choir formationsregarding the ease of hearing other sections?”

Chorister Preference

0

2

4

6

8

10


Rat

ing

s

e)”Which choir formation did you prefer singing in?”

Figure 7a-7e Choristers’ average rating of choral formations across perceptual questions by gender.

Ease of Production

0

2

4

6

8

10


Ra

tin

gs


Finally, the responses to the ‘preferredformation for singing’ followed a pattern similarto most of the other questions with similarresponses by the male and female choristers forthe block sectional and mixed formations, butsignificant differences by gender for thesectional in columns formation (χ2 (1)=6.68;p=.01) (See Table 2 & Figure 7e). The ratings ofthe block sectional formation ranged from 5.89among the sopranos to 6.63 among the basses.For the mixed formation the ratings ranged from6.94 among the basses to 8.68 among the altos,and for the sectional in columns formation theratings ranged from 3.73 among the sopranos to8.66 among the basses.

DISCUSSION

The purpose of this experiment was tocompare LTAS data and chorister perceptionsacross different choral formations. Choralconductors place a great deal of emphasis onthe selection of choral formation, often switchingbetween several formations during one concert.Hence, the acoustic and perceptual data fromthis study could indicate the nature of anydifferences among the formations.

The acoustic data from this study did notreveal spectral differences among the blocksectional, mixed and sectional in columnsformations. It appears that choral formation doesnot significantly affect the LTAS of choral soundwhen recorded in the diffuse field out in theauditorium. Apparently, the blend of the singers’voices maintained a similar pattern across thedifferent formations. This result was contrary tothe hypothesized result of the mixed formationproducing different spectral patterns than theother two formations. Recording the voices frommicrophones located near the singers or at theposition of the conductor may result in differentLTAS patterns.

Recording from this position might provideacoustic data that also reinforce the differentperceptions of the choral formations betweenthe male and female singers. Visual inspectionof the LTAS for each piece demonstrated that nospectral differences were evident in the sourceor resonance spectra by altering formation. Thedistribution of energy across the frequencybands is similar to data reported from previousresearch examining the LTAS during choralsinging, in which most of the energy was seen inthe fundamental frequency and first formant, and

another increase in energy around 3 kHz(Rossing et al., 1986, 1987; Ternström, 1991).

However, the increase in energy at 3 kHzwas much less noticeable in the current resultsthan in previous studies, even for the choristers.This reduction in energy may indicate that thechoristers were attempting to produce a more“choral” sound (Goodwin, 1980), which displaysa concentration of energy at the fundamentaland first formant frequencies, in contrast to thesolo mode of singing which displays moreenergy around the singer’s formant or 3 kHz(Rossing et al., 1986; 1987).

The difference in energy increase at 3 kHzbetween the present study and previous studies(Rossing et al., 1986; 1987) could also berelated to the room’s diffuse sound field.Whereas Rossing et al. (1986; 1987) analyzedrecordings from an anechoic chamber, thepresent recordings were measured in a largeauditorium with the microphone a good distanceaway from the choir. The diffuse field of theroom may have dampened or smoothed out thedirectivity of the acoustic signal; resulting in therelatively low values observed presently.

The LTAS results did demonstratedifferences between the musical selections.Overall, the polyphonic selection was moreintense than the homophonic selection forfrequencies above 650 Hz. This difference wasparticularly noticeable at the singer’s formant,where a difference of nearly 6 dB is shownbetween the selections. A more “soloistic” modeof singing may have been used by the choristersbecause in polyphonic music each sectioncontains its own independent line of musicalmaterial (Goodwin, 1980). Were this the case,however, one might then expect differencesbetween the formations, especially for the mixedformation where singers are separated frommembers of their own section. Anotherpossibility is that differences in pitch range anddurations between the selections resulted in thediffering LTAS spectra.

In future research, the same selection, acanon or round that can be sung bothhomophonically in unison and polyphonically inany number of parts without changing the pitchset, should be used for all tests in order tocontrol for variability and duration of pitches.Also, future researchers may want to examinethe auditory processing of the choral signal fordifferences, given formation. For example,Bregman and Pinker (1978) examined theeffects of auditory streaming and the building of


timbre which revealed possible auditory effectson the processing of sound.

However, the LTAS did successfullydemonstrate differences between the musicalexamples and will be an important part of futureresearch. Measurement of source andresonance spectra provides valuable informationregarding the type and intensity of production.Unfortunately, it accounts neither for instancesof individual noticeable voices (Tocheff, 1990),nor for the tuning of formant frequencies (Hunt,1970), which are considered important factors indetermining blend.

It is also possible that formation affectsvariables that are not measured by LTAS, butrather may affect variables in the auditory orcognitive pathways that are time dependentrather than pitch or loudness dependent. In thatvein, perceivable differences among the choralformations may relate to the listener’s ability toextract streams of auditory information(Bregman, 1990). For example, when hearingchoral singing the listener may follow the pitchrelated stream of each vocal part whilesimultaneously appreciating the auditory scenecreated by the blend of the four vocal parts.Variations in the separation of the four streamsmay create perceptible variations in the auditoryscenes. This combination of perceived auditorystreams across time with the perceivedchanging momentary auditory scenes cannot becaptured by acoustic equipment that measureseither the acoustic changes over time or theacoustic scene at one moment.

In addition to the acoustical analyses,perceptual data were collected and analyzed todetermine differences in preference across thethree formations. The major perceptual finding ofthe present investigation was a strong genderdifference in preferences and perceptions ofsinging in the three choral formations. The malesingers clearly preferred the sectional incolumns formation. They rated it highly in termsof ease of production, desirability of choralsounds, and the ability to blend with others inthe section, and generally favored it more thanthe block sectional and mixed formations. Incontrast, the female singers varied their highestperceptual ratings between the mixed and blockformations, and indicated a strong dislike for thesectional in columns formation.

The male preference for the sectional incolumns formation may have been secondary tothe novel acoustic experience of singing at thefront of the choir. The choristers in this studyroutinely sang in the classical block sectional

formation illustrated in Fig. 1. In the sectional incolumns formation, most male singers hadmembers of their vocal section standing directlybehind them, which appears to have enhancedtheir self-reported ability to blend with others intheir section. The reception of vocal input frombehind appeared to impact the bass and tenorsinger’s perceptions of ease of production andtheir overall impressions regarding the sound ofthe choir.

Most of the female singers in this particularchoir were accustomed to receiving vocal inputfrom at least two other vocal sections behindthem, and in some cases, three vocal sectionsby virtue of routinely singing in the blocksectional formation. Their marked dislike for thesectional in columns formation may be attributedto a perceived decrease in vocal input from theother vocal sections. Interestingly, femalesingers also indicated their ability to blend withsingers in their own section remained relativelystable across all three choral formations. Thiseffect could be directly attributable to the largernumber of female singers in the choir and theconsistency of having people singing their vocalpart behind them in all the configurations.

A possible factor in the choral arrangementpreference differences between the males andfemales might be the typical differences inacoustic output power between them. Ternström(1994) reported that soprano singers in fortetypically produce stronger signals than do basssingers in forte. He stated that this effect wasrelated to the increase in sound pressure level ofthe voice with increases in fundamentalfrequency (Ternström 1994; 1999). In twostudies Ternström (1994; 1999) discussed theSelf-to-Other Ratio, a measure of choristers’experience of their own sound in comparison tothat of other singers. He indicated that therelative strength of a chorister’s sound incomparison with those around him or her is animportant factor in choral singing. He found thisratio to be typically lowest for basses andhighest for sopranos.

Quite possibly, when the bass singers go toa mixed formation, they will be standing closer tosingers that are louder than themselves, and sothe ratio of their sound in comparison to thesingers around them drops even lower, making itharder for them to hear their own voice. Incontrast, when sopranos go to a mixedformation, they get to stand closer to singerswith softer voices, and so their ability to hearothers increases.


This concept relates to the perceptual datadepicted in Figure 6. The preference of basssingers for the overall sound and blend of thecolumn formation and dislike of the mixedformation for overall sound and preferencecontrasts with the sopranos ratings of theseformations. The basses were better able to hearthe singing of the other basses and thus couldbetter appreciate the blend of the entire chorus.The sound of the sopranos would still be heardbecause of the greater signal strength comingfrom them. In contrast, the sopranos could nothear the other, weaker vocal parts, so they couldnot hear the overall choral sound as well.

Inherent in the above discussion is theimplication that singers’ voices send adirectional signal such that the signal projectedlaterally has less of an effect on chorister abilityto hear other singers than the signal goingforward to these choristers in front of eachsinger. Further evidence to support thisassertion is the fact that all of the singers,divided into vocal sections, rated the sectional incolumns formation as being the least effective inallowing them to hear singers in the other vocalsections. In the column formation, many of thesingers were standing directly next to one singerfrom a different vocal section, and were twospaces away from another vocal section on theiralternate side, yet this proximity did not increasethe ability to hear these other parts.

This finding is consistent with the fact thatgreatest energy in sound waves emanating fromthe vocal tracts of singers are in front of thesinger (Kent, 1997; Kent & Read, 1996). Thehead will dampen the projection of the sound indirections other than directly in front of themouth. Consequently, singers in this studyappeared to hear other sections better whenthey were standing directly in front of singersfrom those sections.

Finally, the singers in this study indicatedthat being able to hear the singers from differentvocal sections is an important variablecontributing to overall improved choral blend.They also indicated that the mixed formationbest provided that acoustic atmosphere, which isconsistent with the earlier findings of Daugherty(1999). When in this configuration, singers werereceiving both lateral and posterior vocal inputfrom other vocal sections, which appeared toinfluence the ease of hearing the other parts.Additionally, because all of the music wasperformed a cappella, there was greater relianceon listening to the other parts while singing thanwhen singing with instrumental accompaniment.

This increased ability to hear the other sections,however, did not significantly influence thechorister’s perceptual rating of the overall soundof the choir. Male singers continued to display astrong preference for the sectional in columnsformation, whereas female singers gave a lowerrating for this formation. As noted above, thisfinding indicates that the experience of beingable to hear other singers in their section and tohear the other vocal sections varied among thevocal parts. The weaker sound of the sectionsmade the column formation more desirable asthey could hear both their own section and theother sections well. However, the perception ofthe column formation was opposite for thestronger sound of the female sections.

The primary limitation to the data concerningthe overall sound of this choir was the lack ofusing outside auditors to rate the sound.However, when it is considered that the singer’sratings on this question across the threeformations were generally in the range of 6 to 8on a 10 point scale, these perceptions supportthe lack of significant findings on the LTAS. Theapparent agreement between the acoustic andperceptual findings of this study suggest that thepitch and loudness features essential toestablishing a balanced choral blend result insimilar LTAS. Perceived chorister differencesamong the choral arrangements may be moreinherent in the time based aspects of the soundof the filtering functions of the human ear, ratherthan resulting from differences in the intensity byfrequency aspect of the acoustic signal. Otherlimitations that may have impacted thesefindings were that only one choir was included,the relative small size of that choir, and theunequal balance of members in each vocalsection.

Alternately, it can be argued that LTAS issimply not a sensitive enough measure to detectdifferences in the acoustic signals produced bythese different formations. For instance,because the LTAS averages frequency energyover time produced by both the vocal sourceand vocal tract, it would not measure instancesof noticeable individual voices, i.e. instanceswhen an individual singer’s voice could be heardabove the rest of the choir. Tocheff (1990),reported that critics of mixed choral formationscite noticeable individual voices as being acommon problem, however no such increasedincidence was detected by the LTAS in thisstudy due to the averaging effect of theprocedure. LTAS also does not account for thetuning of formant frequencies (Hunt, 1970),


which is considered another important facet ofblend. Even so, the LTAS did successfullydemonstrate a difference between musicalliterature examples, and will be an important partof future research in this area.

Choristers reported that differences in soundoccurred for the three formations; however,these differences were not seen acoustically inthis experiment. It is possible that the acousticfiltering of the auditory system and/or thecognitive processing of the signal may beresponsible for these differences. Futureresearch should include auditory filtering of theacoustic data before measurement in order toobserve any differences. Also, additionalcontrols for the musical selection should beperformed as mentioned above, as should thespacing of the ensemble (Daugherty, 1999;2003), which was approximated but notspecifically controlled in this study. Finally, theuse of a videotaped conductor would ensure aconstant pace and increase the reliability offuture research.

Perceptual measures can be effective indescribing a subjective art as such choralsinging. Even so, the discovery anddevelopment of acoustic measures of choralblend would be advantageous to the choralmusic community, if only to better understandhow each unique voice works with others tocreate a blended sound.

REFERENCES

Bregman, A. S. (1990). Auditory scene analysis:the perceptual organization of sound.Cambridge, MA: MIT Press.

Bregman, A. S., & Pinker, S. (1978). Auditorystreaming and the building and timbre.Canadian Journal of Psychology, 32, 19-31.

Daugherty, J. F. (1996). Spacing, formation, andchoral sound: preferences and perceptionsof auditors and choristers. UnpublishedPh.D. Dissertation, Florida State University,Tallahassee.

Daugherty, J. F. (1999). Spacing, formation, andchoral sound: preferences and perceptionsof auditors and choristers. Journal ofResearch in Music Education, 47(3), 224-238.

Daugherty, J. F. (2003). Choir spacing andformation: Choral sound preferences inrandom, synergistic, and gender-specificchamber choir placements. InternationalJournal of Research in Choral Singing, 1 (1),48-59.

Ekholm, E. (2000). The effect of singing modeand seating arrangement on choral blendand overall choral sound. Journal ofResearch in Music Education, 48(2), 123.

Ford, J. K. (2003). The preference for strong orweak singer's formant resonance in choraltone quality. International Journal ofResearch in Choral Singing, 1, 29-47.

Giardiniere, D. C. (1991). Voice matching: aperceptual study of voice matches, theiraffect on choral sound, and procedures ofinquiry conducted by Weston Noble.Unpublished Ph.D. Dissertation, New YorkUniversity.

Goodwin, A. (1980). Acoustic study of individualvoices in choral blend. Journal of Researchin Singing, 3, 25-35.

Hunt, W. A. (1970). Spectrographic analysis ofthe acoustical properties of selected vowelsin choral sound. Unpublished D.E. Thesis,North Texas State University.

Kent, R. D. (1997). The Speech Sciences. SanDiego, London: Singular Publishing.

Kent, R. D., & Read, C. (1996). The AcousticAnalysis of Speech. San Diego: SingularPublishing.

Knutson, B. J. (1987). Interviews with selectedchoral conductors concerning rationale andpractices regarding choral blend.Unpublished Ph.D. Dissertation, FloridaState University, Tallahassee.

Lambson, A. R. (1961). An evaluation of variousseating plans used in choral singing. Journalof Research in Music Education, 9, 47-54.

Madsen, C. K., & Yarbrough, C. (1980).Competency-based music education.Englewood Cliffs, NJ: Prentice-Hall.

Rossing, T. D., Sundberg, J., & Ternström, S.(1986). Acoustic comparison of voice use in


solo and choir singing. Journal of theAcoustical Society of America, 79, 1975-1985.

Rossing, T. D., Sundberg, J., & Ternström, S.(1987). Acoustic comparison of soprano soloand choir singing. Journal of the AcousticalSociety of America, 82, 830-836.

Siegel, S., & Castellan, J. N. J. (1988).Nonparametric statistics for the behavioralsciences (2nd Ed.). Boston: McGraw-Hill.

Sundberg, J. (1977). The acoustics of thesinging voice. Scientific America, March, 82-91.

Ternström, S. (1991). Physical and acousticalfactors that interact with the singers to

produce the choral sound. Journal of Voice,5, 128-143

Ternström, S. (1994). Hearing myself with theothers - sound levels in choral performancemeasured with separation of the own voicefrom the rest of the choir. Journal of Voice,8, 293-302.

Ternström, S. (1999). Preferred self-to-otherratios in choir singing. Journal of theAcoustical Society of America, 105, 3563-3574.

Tocheff, R. D. (1990). Acoustical placement ofvoices in choral formations. UnpublishedPh.D. Dissertation, Ohio State University.

select acoustic and perceptual measures of choral formation · 11.5 years of solo vocal training;...

Documents