You are on page 1of 19

Music and Mathematics: Modest Support for the Oft-Claimed Relationship

Author(s): Kathryn Vaughn


Source: The Journal of Aesthetic Education, Vol. 34, No. 3/4, Special Issue: The Arts and
Academic Achievement: What the Evidence Shows (Autumn - Winter, 2000), pp. 149-166
Published by: University of Illinois Press
Stable URL: https://www.jstor.org/stable/3333641
Accessed: 05-12-2018 12:24 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

University of Illinois Press is collaborating with JSTOR to digitize, preserve and extend
access to The Journal of Aesthetic Education

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics: Modest Support for the
Oft-Claimed Relationship

KATHRYN VAUGHN

According to conventional wisdom, music and mathematics are r


and musical individuals are also mathematically inclined. After all
rhythm is based upon mathematical relations, and it is certainly
to assume that an understanding of music requires some understa
ratios (e.g., 3/4 time vs. 4/4 time) and repeating patterns. In addition
experiences enhances spatial-temporal reasoning,1 then music may
hance understanding of those aspects of math that involve spatial
reasoning, such as geometry and proportional reasoning.2 Igor S
noted the links between music and mathematics, claiming that m
"something like mathematical thinking and mathematical relation
short, if music is based on mathematical principles, and if an unde
of music requires some understanding of these principles, then it
that music education can lead to an improved understanding. Wha
is a report on three meta-analyses investigating the relationship
music and mathematics.

General Method

In order to locate all published and unpublished studies on the relationship


between music and mathematics, seven electronic databases were searched
from their inception through 1998: Arts and Humanities Index (1988-1998),
Dissertation Abstracts International (1950-1998), Educational Resource In-
formation Clearinghouse (1950-1998), Language Linguistics Behavioral
Abstracts (1973-1998), MedLine (1966-1998), PsychLit/PsychINFO (1984-
1998), and Social Science Index (1988-1998). The initial search used six search
strings: 1) music, piano, vocal, instruction/instruct*, train, education/educat*,
and program; 2) learn, academic/academ*, cognition/cognit*, achieve-
ment/achiev*; 3) measure, find, outcome, effect; 4) transfer; 5) band or chorus
or instrument*/instrumental or string/strings; and 6) geometry or algebra
or arithmetic or calculus.

Journal of Aesthetic Education, Vol. 34, Nos. 3-4, Fall/Winter 2000


@2000 Board of Trustees of the University of Illinois
This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
150 Kathryn Vaughn

These searches yielded over 4000 initial references. Only studies testing
an association between music and mathematics were included. Advocacy
pieces and program descriptions (the vast majority) were excluded. In ad-
dition, studies had to have a measured math outcome, a control group, and
sufficient statistical information for an effect size to be computed (or the
authors had to be willing and able to supply needed missing statistical
information).
Three types of empirical studies linking music and math were excluded.
First, studies in which music was used as a reward for good performance in
math were excluded because they do not test whether there is anything
about music learning per se that improves mathematical aptitude. There
were many of these. Studies in which musical jingles were used as memory
aids to teach math were excluded because these studies did not provide au-
thentic music instruction.4 Finally, studies assessing whether individuals
who have a high musical aptitude also have a high mathematical aptitude
(or the reverse) were excluded because my primary interest was in the ef-
fects of music instruction/exposure, rather than on individual differences
in aptitude. However, a review of studies assessing the aptitude question
would be interesting. If aptitude in music predicts aptitude in math, such a
finding would suggest that training in music might result in improvements
in math, or the reverse (training in math leading to improvement in musical
ability). However, in what follows, only studies providing a direct test of
the hypothesis that training in music results in improvements in math were
included.

A total of 25 studies remained. These were of three types. Eight correla-


tional studies (yielding 20 effect sizes) examined whether students who
choose to study music have high math outcomes. A correlation between
music study and math achievement is a necessary but not a sufficient
condition for a causal relationship between music and math. Five experi-
mental studies (yielding 6 effect sizes) provided a test of the hypothesis that
music instruction causes mathematical improvement. These studies assessed
whether learning to play an instrument results in improved math out-
comes. Finally, 12 experimental studies (yielding 15 effect sizes) assessed
the hypothesis that performance on math tests is facilitated when certain
types of music are played in the background (rather than when silence pre-
vails or when the music played was deemed distracting by the experiment-
ers). For ease of reporting, from now on I refer to each effect size as a study.
Effect size rs were calculated, based on one of the methods listed in
Table 3 of Winner and Cooper (this issue), along with the associated Z level
and the significance level (p) associated with the Z. In one study (Kvet), four
effect sizes were calculated, but each was based on a separate experiment
and set of participants. In no case were the same participants used for more
than one effect size.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 151

Correlational Studies

A meta-analysis was first performed on 20 correlational studies who


characteristics are summarized in Table 1.5 Ten of these studies were from
data made available by the College Board on SAT math scores (betwe
1987-1998, excluding 1993, for which standard deviations were unavailable)
Ten other studies include data from comparatively small populations. In a
cases, the studies compared some form of math achievement in student
with and without self-selected music study. In all of the studies, studen
were trained to perform music either instrumentally or vocally. For the 1
College Board studies, the music experience consisted in having taken at
least one course in instrumental or vocal performance in high school.7 Th
comparison group consisted of students who had taken no arts courses o
any kind. Studies were coded according to the categories listed below.

Coding Characteristics
Year. Dates of publication/appearance ranged from 1950-1999.
Sample Size. The total sample size of the studies analyzed was n=
5,788,132. Sample sizes ranged from n=34 to n=648,144, with a mean of
n=286,907 and a median of n=273,858.
Outcome. All outcomes but one were measured by standardized math
tests. One outcome was based on grade point average in math.
Age of Participants. Thirteen studies assessed students at the high schoo
level, and six studies assessed elementary-school-aged children (3rd - 6th
grades); one study assessed students at the college level.
Duration of Music Training. Duration of music training varied widely
Ten studies assessed the association between math performance and 1-4
years of music instruction; four assessed the association between math an
1-6 years of music; and one assessed the association between math and 1
years of music.
Outlet. Most of the studies were in the form of "fugitive" (unpublished
literature. Seven studies appeared as unpublished doctoral dissertations,
one was in the form of a conference presentation, and ten appeared in th
form of unpublished tabulated data (the College Board studies). Only two
studies were in published form: of these, one appeared as a chapter and on
appeared in a in peer-reviewed journal.

Results

As shown in the stem and leaf display in Table 2, the 20 effect size rs rang
from r = -.05 to r = .37. The mean effect size was r = .15. When weighted b
size of study, the mean effect size r= .14. Thus there is little difference betwe
the mean weighted and unweighted effect size, despite the fact that the SA
studies had far larger sample sizes (mean n over the ten years = 573,227
than did the other ten studies (mean n= 586).

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Table 1: Correlational Studies

Za Duration of Effect
Study/Yr. N r (p) Outcome b Age Music Training

Anello (1972) 326 .16 2.81* Math GPA 9th-12th gr 1- 4 years A-meth
Catterall (1999) 1476 .17 6.62* NAEP 12th gr 1 - 5 years X2 met
Ciepluch (1988) 80 .37 3.33* CTBS 9th - 12th gr 1 - 4 years X2 metho
College Board (1998) 362,853 .18 105.81* SAT Math 12th gr 1 - 4 years A-met
College Board (1997) 354,886 .21 122.52* SAT Math 12th gr 1- 4 years A-met
College Board (1996) 349,032 .18 103.50* SAT Math 12th gr 1- 4 years A-meth
College Board (1995) 346,737 .18 105.05* SAT Math 12th gr 1- 4 years A-meth
College Board (1994) 343,270 .15 85.13* SAT Math 12th gr 1- 4 years A-meth
College Board (1992) 356,258 .12 71.03* SAT Math 12th gr 1- 4 years A-meth
College Board (1991) 361,998 .11 68.17* SAT Math 12th gr 1- 4 years A-meth
College Board (1990) 361,272 .11 63.27* SAT Math 12th gr 1- 4 years A-meth
College Board (1989) 385,943 .10 61.13* SAT Math 12th gr 1- 4 years A-meth
College Board (1988) 437,206 .08 54.25* SAT Math 12th gr 1- 4 years A-meth
Engdahl (1994) 598 .11 2.59* CTBS 5th-6th grade 1- 2 years t-metho
Kvet (1982a) 34 .27 1.60 CAT 3rd - 6th gr 1- 6 years t-method
(p=.05)
Kvet (1982b) 84 .15 1.34 Stanford 3rd - 6th gr 1- 6 years t-metho
(p=.09) Ach. Test
Kvet (1982c) 142 .08 .91 CAT 3rd - 6th gr 1- 6 years t-method
(p=.18)

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Za Duration of Effect
Study/Yr. N r (p) Outcome b Age Music Training
Kvet (1982d) 90 .14 1.33 MAT 3rd - 6th gr 1- 6 years t-method
(p=.09) CAT
McCarthy (1992) 1061 .10 3.28* SRA 9th -12th gr 1- 4 years A-method
Wheeler & Wheeler 1969 -0.05 -2.39* ACE College Music Majors (At A-met
(1950)c least 2 years)

Notes:
a. * means p<.05.
b.
Test Name Abbreviation

American Council on Education ACE


California Achievement Test CAT
Comprehensive Test of Basic Skills CTBS
Metropolitan Achievement Test MAT
National Assessment of Educational Progress NAEP
Science Research Associates SRA
Scholastic Assessment Test SAT
Stanford Achievement Test Stanford Ach. Test

c. No amount of musical training was specified. However, perfor


of a minumum of 2 years of study is a conservative estimate.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
154 Kathryn Vaughn
Table 2: Stem and Leaf Display of 20 Effect Size rs from Correlational Studies

Stem Leaf

+.3 7

+.2 1, 7

+.1 0, 0, 1, 1, 2,4, 5,5, 6, 6, 7, 8, 8, 8

+.0 8,8

-.0 5

The Stouffer's Z test f


cant Z=192.59, p<.0001,
new subjects who migh
servative t-test of the
that these results can
therefore, that it is hig
due to chance. The 95%
zero, ranging from r=.
studies averaging null r
Z down to the just-sign
the effect size is robust.

However, effect sizes were significantly heterogeneous, X2= 6472, df=19,


p< .00001.8 When I examined only the ten College Board studies, heteroge-
neity remained significant, X2= 6378, df=9, p< .00001. As can be seen from
Table 1, effect sizes from the College Board studies increased fairly steadily
with each year. I was not able to identify a factor that accounted for the het-
erogeneity in these effect sizes other than the influence of the large sample
sizes on the chi-square statistic.9

Discussion

This meta-analysis demonstrates a modest positive association between the


voluntary study of music, on the one hand, and mathematical achievement,
on the other hand. While correlation is a necessary condition for causality, it
is not sufficient. Thus, while it is conceivable that the music education re-
ceived by the students in these studies actually led to improvement in math
performance, other explanations for this correlation have not been ruled
out. Students who attend schools in which music classes are a part of the
curriculum may come from a stronger socioeconomic background than
those who attend music-poor schools; music-rich schools may be better
academically than music-poor schools because these are schools with better
resources; and students who study music and do well in math may come
from families that value both study of the arts and academic achievement.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 155

The claim that involvement in music improves math achievement is c


sistent with-but not proven by-the positive effect size found here. F
test of the causal power of music to improve math we must turn to t
experimental studies.

Experimental Training Studies

A meta-analysis was next performed on six experimental training studie


which students were given instruction in instrumental or vocal music p
formance for at least four months and then were tested on math skills
These studies and their characteristics are listed in Table 3.11

Coding Characteristics
Year. Dates of publication/appearance ranged from 1959-1997.
Sample Size. The total sample size of the studies analyzed was n= 357.
Sample sizes ranged from n=28 to n=128, with a mean of n=60 and a median
of n=48.

Design. Two studies were true-experimental studies in which students


were randomly assigned at the individual level to a music or a control
group. Four were quasi-experimental studies in which students were either
assigned to groups at the classroom level, or in which convenience samples
were used. In addition, four of the studies matched music and control groups
in terms of potentially relevant factors such as IQ, grade point average,
social class, or musical experience.
Outcome. Three studies used a standardized math achievement test as

the outcome, and three used a test designed by the researcher. In the study
by Graziano and her colleagues, a computerized interactive test was used
as the assessment. Children were given 16 nonverbal items assessing their
understanding of proportions and ratios. Children saw pictures of shapes
and objects on a computer screen and were asked how many of certain
shapes (e.g., blocks or triangles) would fit into a larger shape. In the two
studies by Neufeld, children were given a test assessing prenumber con-
cepts related to proportions and ratios. The test consisted of tasks such a
seriating objects by size or length (showing an understanding of a progres-
sion from less to more), and completing patterns given a choice of missing
pieces.
Age of Participants. All six studies assessed students at the preschool or
elementary school level.
Duration of Music Training. The amount of time spent on music training
ranged from four months to two years.
Instrument. In two studies, children were trained on the keyboard; in two
they were trained vocally (Kodaly method); in one study, children were
trained on the violin (Suzuki method); and in one they were trained on a

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Table 3: Experimental Training Studies

Duration of
Za Math Music Eff
Study/ Yr. N r (p) Design Outcome Age Training Instrument N

Costa-Giomi 128 .20 2.24* True- Canadian 4th gr. 2 Years Keyboard Trad
(1997) Experimental/ Achievement Notat
Unmatched Test

Friedman 28 .09 .46 True- Stanford 5th - 6th gr. 1-2 years School Band Tr
(1959) (p=.32) Experimental/ Achievement Instruments
Matched Test

Graziano, 55 .31 2.32* True- Researcher 2nd gr. 4 Months Keyboard Tra
Peterson, & Experimental/ Designed Not
Shaw (1999) Unmatched Math T

Neufeld 40 .04 .25 Quasi- Researcher K 2 Years Kodaly Method Non-Tra


(1986) (p=.40) Experimental/ Designed (Primarily Vocal
Matched Math Test Hand Sy

Neufeld 40 -.04 -.25 Quasi- Researcher Pre-K 1 Year Kodaly Method Non-Tr
(1986) (p=.40) Experimental/ Designed (Primarily Vocal
Matched Math Test Hand Sy

Weeden 66 .17 1.40 Quasi- Stanford 1st - 2nd gr. 4 Months Violin (Suzuki No
(1971) (p=.08) Experimental/ Early School Me
Matched Achievement
Test

Notes

a. * means p<.05.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 157

variety of instruments (school band instruments). In the study by Gra


and her colleagues, the music training consisted of group piano lesson
well as training in computerized spatial-temporal games and practice
fractions and proportions. The control group received the same spatial-t
poral and math practice, but instead of piano lessons, they received g
English lessons on the computer.
Notation. In three studies, children were exposed to traditional music
tation. In two studies, the Kodaly method was used, in which stud
learn to read hand movements that stand for pitches but in which no w
ten notation is involved. And in the study using the Suzuki method, no
tation of any kind was used. Unfortunately, as shown in Table 3, typ
instrument, type of instructional method, and presence or absence of e
sure to notation were confounded so that I could not test the impact of
measures separately.
Outlet. Three effect sizes came from studies published in peer-review
journals (two of these were actually from the same study, from which I co
puted two effect sizes). Two studies were doctoral dissertations, and one
a conference presentation.

Results

As shown in the stem and leaf display in Table 4, the effect sizes ran
from r = -.04 to r =.31. The mean effect size was r=.13. When weighted
size of study, the mean effect increased to r=.16, indicating that larger s
ies had higher effect sizes. The combined Zs yielded a Stouffer's Z= 2.
p=.004, indicating that it is highly unlikely that the positive effect size fo
was due to chance. The more conservative t-test of the mean Zr = 2.49,
nearly significant at p=.06. The 95% confidence interval did not span zero,
but ranged from r=.03 to r=.23. This tells us that in another sample of six
similar studies, the mean effect size is not likely to be at zero or below. A
file drawer analysis revealed that nine studies would need to be found av-
eraging null results to bring the Stouffer's Z down to just significant at
p=.05. Effect sizes were not significantly heterogeneous, X2= 3.79, df=5, p = .58.

Table 4: Stem and Leaf Display of 6 Effect Size rs from Experimental Training Studies

Stem Leaf

+.3 1

+.2 0

+.1 7

+.0 4, 9

-.0 4

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
158 Kathryn Vaughn

A linear contrast test was performed to determine whether studies in


which students learned to read music notation were associated with higher
effects sizes than studies without notation learning. This prediction was
made because it is possible that practice in reading symbols might general-
ize to practice in reading math symbols. Contrast weights correlated with
notation involvement at r = .62. The linear contrast Z, however, proved
nonsignificant (Z = .61, p = .27). Thus, it appears that, in this sample, music
training that included learning to read notation was no more effective at en-
hancing math performance than was music training without music reading.
This result contrasts with that reported in Hetland (this issue).

Discussion

One final comment is important to make. It is noteworthy that the study by


Graziano and her colleagues produced the largest effect size. In this study
both the experimental and control groups received the same instruction in
math, but the experimental group also received music. Perhaps it was the
combination of music instruction and the particular spatial-temporal math
instruction students in that study received that led to the improved math
performance. Clearly, further research is required to test the possibility that
concurrent but separate math instruction using spatial-temporal methods
improves math performance.

Background Music Studies

Fifteen experimental studies were identified which investigated whether


listening to soothing background music while taking a math test resulted in
elevated math scores. These studies were combined to test the prediction
that soothing (the researchers sometimes called this "mood-calming") mu-
sic enhances math performance. The types of music considered soothing in
these studies were: classical, instrumental music played at medium volume,
folk music, disco-instrumental music, soft "pop" music, and muzak. In vari-
ous control conditions, participants listened to music considered to be dis-
tracting and hence predicted to interfere with math performance: "stimulat-
ing" or loud pieces of classical music, opera, rock, and rap. In addition,
speech (without music) and industrial noise were predicted to interfere with
math.12 Table 5 summarizes the characteristics of these studies.

Coding Characteristics
Year. Dates of publication/appearance ranged from 1958-1998.
Sample Size. The total sample size of the studies analyzed was n= 1,652.
Sample sizes ranged from n=10 to n=320, with a mean of n=110 and a
median of n=95.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Table 5: Background Music Studies

Za
Special Condi
Study N r (p) Outcome Design Age Needs (weights
Abikoff 20 .05 .22 Wide Range Repeated 2nd - 6th Attention Deficit/ Sile
(1996a) (p=.41) Achievement Measures gr Hyperactivity Speech
Test Disorder Rap/Rock (
Abikoff 20 -.05 -.22 Wide Range Repeated 2nd - 6th No Silence
(1996b) (p=.41) Achievement Measures gr Speech (-1
Test Rap/Rock (-
Anno 63 .19 1.52 California Unmatched/ 3rd gr No Silence (-1)
(1958) (p=.06) Achievement Between Classica
Test Subjects Muzak (+
Hallam & 10 .82 2.59* Teacher Repeated 9-10 yrs Emotionally/ Silence
Price Designed Test Measures Behaviorally "Mood
(1997a) ing" music (+
Hallam & 31 .05 .29 Teacher Repeated 10-11 yrs No Silence (-1
Price (p=.39) Designed Test Measures "Mood ca
(1997b) ing" music (+
Hardie 200 .20 2.81* College Board Unmatched/ College No Silence
(1990) Comparative Between Classical (+
Guidance Exam Subjects
Kopp 95 .05 .51 Unspecified Unmatched/ 4th gr No Silence (-
(1958a) (p=.30) Standardized Between Class
Test Subjects ("relaxing")
Kopp 96 -.15 -1.44 Unspecified Unmatched/ 4th gr No Silence (
(1958b) (p=.07) Standardized Between Class
Test Subjects ("stimulating"
Manthei & 66 .03 .24 Customized Repeated College No Silence (-1)
Kelley (p=.41) Math Placement measures Classical
(1998) technical

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
za
Special Conditi
Study N r (p) Outcome Design Age Needs (weights
Miller & 198 .08 1.13 Differential Unmatched/ College No Silence
Schyb (p=.13) Aptitude Test Between Classical (Mozart) (+
Battery Subjects Disco (+1)
Rock (

Mowesian 167 .08 1.03 Basic Skills in Unmatched/ 10th gr No Silenc


& Heyer (p=.15) Arithmetic Between Folk (-1) m
(1973) Subjects Classical (Instrumental)
Classical (Voca
Rock (-1)

Norton 320 .07 1.25 Stanford Unmatched/ 6th gr No Classical (sof


(1971) (p=.11) Achievement Between Classical (loud) (-1)
Test Subjects Pop (soft) (+1)
Pop (l
Tucker & 151 -.18 -2.27* Customized from Unmatched/ College No Silenc
Bushman ACT & SAT Between Rock
(1991) Math Subjects J
Wolf & 15 .35 1.36 Researcher Repeated College No Silence
Weiner (p=.09) Designed Measures Classical (+2) m
(1972) Arithmetic Test Speech (-
Industrial Nois

Wolfe 200 .12 1.70* Researcher Unmatched/ College No Silence


Film Soundt
Film Soundtrac
(med) (+1)
Film Soundtrac
(loud) (-2)

Notes:

a. * means p<.05
b. MPC-F = Maximum Possible Contrast F

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 161

Outcome. In ten studies, some form of standardized math achieveme


test or portions taken from standardize tests were the outcome measu
The remaining five studies used a researcher-designed math test.
Design. Nine studies were "between-subjects" designs. These involved
comparison between a group exposed to music predicted to enhance ma
and one or more groups exposed to a condition predicted to interfere w
math. In each case, a math test was taken during exposure to various kin
of auditory stimulation. Six studies were designed as "within-subje
studies. In these studies, students listened to music predicted to enhan
math and also to one or more conditions predicted to interfere with ma
Again, during each condition participants took a math test. Condit
were always separated by a time period of at least one day, and the m
tests given during the different conditions were parallel but never ident
In all but one case, order of conditions was counterbalanced to rule ou
practice effects. In one study (Manthei and Kelley, 1998), the classical mu
condition always preceded the control (silence) condition. Thus, this or
works against the hypothesis that classical music should enhance math
since the math test following silence was the second test taken and mig
have benefited from a practice effect. I retained this study since it was
ased against rather than in favor of the hypothesis and so would contrib
to a more conservative average effect.
Age of Participants. The age groups ranged from 2nd grade to colleg
level.

Special Needs. Two of the studies assessed the effect of music on spec
needs students: attention deficit/hyperactivity, and emotionally/behavi
ally disturbed. All other studies assessed the effect of music on typica
children.

Conditions. The various combinations of conditions in each study are


shown in Table 5.

Outlet. Four studies appeared in peer-reviewed journals, two in non-peer-


reviewed journals, four as doctoral dissertations, one as a masters thesis,
one as a conference presentation, and one as a technical report available on
the internet. From these, 15 effect sizes were calculated.

Calculation of Effect Sizes


Where more than two conditions were present, it was necessary to calculate
the effect size using the Maximum Possible Contrast F test.13 To calculate
the F for the hypothesis that soothing music helps while distracting condi-
tions hurt, a positive weight was assigned to those conditions predicted to
enhance, and a negative weight was assigned to those conditions predicted
to interfere with math. Weights always sum to zero.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
162 Kathryn Vaughn
Results

As shown in the stem and leaf display in Table 6, effect sizes ranged from
r = -.18 to r = .82. The mean effect size was r = .14 (Stouffer's Z= 2.77, p=.003;
t-test of the mean Zr=1.71,, p=.11, only nearing significance). The weighted
mean effect size was r=.07. The 95% confidence interval ranged from r=.02
to r=.26, and thus did not span zero. The file drawer measure indicated that
27 additional studies averaging null results would need to be found to con-
clude that the findings overestimate the effect of soothing or relaxing back-
ground music. Effect sizes were significantly heterogeneous, however, X2=
37.56, df=14, p=.0006, suggesting that the sample may come from more than
one population of studies.

Table 6: Stem and Leaf Display of 15 Effect Size rs from Background Music Studies

Stem Leaf

+.8 2

+.7

+.6

+.5

+.4

+3 5

+.2 0

+.1 2, 9

+.0 3, 5, 5, 5, 7, 8, 8

-.0 5

-.1 5, 8

Two studies (Abikoff, Hallam and Pric


ground music on special-needs childre
turbed). The study with children with
fect size of only r = .05. However, the
and behavioral disorder yielded the hig
beneficial calming/focussing effect of
disturbed children merits further research.

When I eliminated these two studies from the analysis, in order to ren-
der studies more similar to one another (all based on typical children), a
mean effect size of r=.07 was found (Stouffer's Z= 2.19, p=.014; t-test of the
mean Zr=1.71, p=.11, only nearing significance), and a weighted mean effect
size of r= .06. The 95% confidence interval spanned zero, ranging from r =
-.01 to r = .15. The file drawer measure dropped to 11, meaning that only 11
additional studies averaging null results would need to be found to negate

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 163

the findings here. With these studies excluded, the effect sizes were
significantly heterogeneous, X2= 21.77, df=12, p=.04.

Discussion

The experiments analyzed in this third meta-analysis were conducted to


test the hypothesis that background music can enhance math performance.
The underlying hypothesis is that an optimum level of music or back-
ground sound causes just the right amount of arousal to enhance perfor-
mance. This hypothesis was not supported by the empirical evidence. The
results of this meta-analysis show that playing music in the background
while students are taking math tests has only a small positive effect, at best.

General Discussion

What are we to make of the oft-cited relationship between music and math?
This analysis did not explore the possibility that individuals with high lev-
els of musicality are also particularly able in mathematics; nor did it explore
the reverse possibility that individuals with high mathematical ability are
particularly musical. Instead, this analysis explored three other questions:
1) Do individuals who voluntarily choose to study music (and these may be
individuals with high musical ability but I cannot say for sure) show higher
mathematical achievement than those who do not so choose? 2) Do indi-
viduals exposed to a music curriculum in school (not voluntarily selected)
show higher mathematical achievement as a consequence of this music in-
struction? 3) And does background music heard while thinking about math
problems serve to enhance mathematical ability at least during the music
listening time?
The meta-analyses of the existing data allow the following conclusions.
The answer to the first question is yes. A small association between the vol-
untary study of music and mathematics achievement was found when 20
studies with correlational designs were combined.
The answer to the second question is also yes. A small causal relation-
ship was demonstrated when six studies were combined, showing that mu-
sic training enhances math performance. However, it is worthy of note that
six studies is a very small number. Of these few studies, three produced
modest effect sizes, and three produced essentially zero effect sizes (two
positive and one negative). Only nine more studies averaging null results
would be needed to overturn the significant finding here. In short, there is a
dearth of existing evidence testing the hypothesis that music training en-
hances performance in mathematics, and I conclude that the hypothesis has
not yet been adequately put to the test.
Finally, the answer to the third question, when only studies that include
non-special-needs children were considered, is a very small and shaky yes.
The confidence interval included zero, and the effect size found was trivially

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
164 Kathryn Vaughn

small. Thus, we cannot conclude with any certainty that soothing music en-
hances math performance. Moreover, even if such a conclusion could be
drawn, we would need to examine the effect of soothing music on other
kinds of academic performance in order to determine whether background
music has a particularly enhancing effect on math, or whether it enhances
any academic task as a function, perhaps, of its relaxing effect. There does
exist a body of studies investigating the effect of music listening on reading
achievement, and a meta-analysis of these findings would help put the
background music and math meta-analysis into proper context.

NOTES

1. See Lois Hetland, this issue, for evidence that music enhances spatial temp
reasoning.
2. For such an argument, see Amy Graziano, Matthew Peterson, and Gordon
Shaw, "Enhanced Learning of Proportional Math through Music Training and
Spatial-Temporal Training," Neurological Research 21, no. 2 (1999): 139-52.
3. Igor Stravinsky, Conversations with Robert Craft (London: Pelican Books, 1971), p.
34, cited in Howard Gardner, Frames of Mind: The Theory of Multiple Intelligences
(New York: Basic Books, 1983), p. 126.
4. Three such studies were found: K. Benes-Lafferty,"An Analysis of Using Musi-
cal Activities in a Second Grade Mathematics Class" (Doctoral diss., Indiana
University of Pennsylvania, 1995); A. Gregory, "The Effects of a Musical Instruc-
tional Technique on the Mathematical Achievement of Third-Grade Students"
(Doctoral diss., University of Alabama, 1988); and A. Traver-Holder, "At-Risk
Students and Math Achievement: The Effects of Manipulative-Based and
Music-Based Teaching Approaches" (Doctoral diss., Regent University, 1993).
5. Full reference for these studies can be found in Studies Used in Meta-Analyses.
6. Adjustments were made for the recentering of SAT scores that occurred starting
in 1996 such that all means were based on the original unrecentered scale.
7. Students taking the SAT were asked to respond to a questionnaire asking them
about courses taken. Further discussion of the SAT data can be found in the
article by Vaughn and Winner, this issue.
8. If the sum of the absolute value of the sum of the weighted squared differences
between each effect size (Zr) and the mean effect size Zr exceeds the upper-tail
critical value of chi-square at degrees of freedom equal to the number of studies
minus 1, then the effect sizes are heterogeneous. (See William R. Shadish and
Keith Haddock, "Combining Measures of Effect Size," in The Handbook of Re-
search Synthesis, ed. Harris Cooper and Larry Hedges (New York: Russell Sage
Foundation, 1994).
9. In calculating heterogeneity, the Fisher's Zr for a given study is treated as the
observed value, and the overall mean Fisher's Zr as the expected value for each
study. Even a small difference between the two values, when multiplied by
such large samples (n>500,000), yields a large value that adds to the weight of
the chi-square statistic.
10. One potentially relevant study (Madeleine Zulauf, "Three-Year Experiment i
Extended Music Teaching in Switzerland: The Different Effects Observed in a
Group of French-Speaking Pupils," Bulletin of the Council of Research in Music
Education 119 (Winter 1993/1994:111-12) could not be included because the ex-
perimental group received music training during regular math class time, while
the control group continued with its usual math studies. Hence, the experimen-
tal group received less instruction in math. This group yielded a negative effect

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
Music and Mathematics 165

size (r=-.12), and this poor showing of the music students may well have be
due to the fact that they actually studied less math than did the control group
11. Full references for these studies can be found in Studies Used in Meta-Anal
12. In one study (by Abikov and his colleagues, 1996), rock and rap were compa
to silence and were predicted to enhance rather than depress math performa
However, since this was in conflict with the hypotheses of all the other studies
used this study only to test the hypothesis that rock and rap (both coded as
silence coded as +2) would interfere with math. Thus, the hypothesis for th
study that was tested by this meta-analysis was not the hypothesis tested b
these authors.
13. See Robert Rosenthal and Ralph L. Rosnow, Contrast Analysis: Focused Compari-
sons in Analysis of Variance (Cambridge, U.K.: Cambridge Univ. Press, 1988), pp.
74-77.

Studies Used in Meta-Analyses

Abikoff, Howard, Mary E. Courtney, Peter J. Szeibel, and Harold S. Koplewicz, "The
Effects of Auditory Stimulation on the Arithmetic Performance of Children with
ADHD and Nondisabled Children," Journal of Learning Disabilities 29, no. 3 (1996):
238-46.
Annello, John A., "A Comparison of Academic Achievement between Instrumental
Music Students and Non-Music Students in the El Dorado and Valencia High
Schools of the Placentia United School District" (Doctoral Diss., Brigham Young
University, 1972).
Anno, Janet Winkle, "The Effects of Background Music on a Group of Third-Grade
Children" (Masters Thesis, Ohio State University, 1958).
Catterall, James S., Richard Chapleau, and John Iwanaga, "Involvement in the Arts
and Human Development: General Involvement and Intensive Involvement in
Music and Theater Arts. In Edward Fiske, ed. Champions of Change: The Impact of
the Arts on Learning (The Arts Education Partnership and The President's Com-
mittee on the Arts and the Humanities, 1999), pp. 1-18.
Ciepluch, Gary Michael," Sightreading Achievement in Instrumental Music Per-
formance, Learning Gifts, and Academic Achievement: A Correlational Study"
(Doctoral Diss., University of Wisconsin, 1988).
"College Bound Seniors Profile of SAT and Achievement Test Takers," 1987-1998
(College Board).
Costa-Giomi, Eugenia, "The McGill Piano Project: Effects of Piano Instruction on
Children's Cognitive Abilities, Academic Achievement, and Self-Esteem" (Paper
presented at the 12th National Symposium on Research in Music Behavior, Min-
neapolis, 1997).
Engdahl, Pherbia Mathis, "The Effect of Pull-Out Programs on the Academic
Achievement of Sixth-Grade Students in South Bend, Indiana" (Doctoral Diss.,
Andrews University, 1994).
Friedman, Bernard, "An Evaluation of the Achievement in Reading and Arithmetic
of Pupils in Elementary School Instrumental Music Classes" (Doctoral Diss., New
York University, 1959).
Graziano, Amy B., Matthew Peterson, and Gordon L. Shaw, "Enhanced Learning of
Proportional Math through Music Training and Spatial-Temporal Training,"
Neurological Research 21, no. 2 (1999): 139-52.
Hallam, Susan, and John Price, "Can Listening to Background Music Improve Chil-
dren's Behaviour and Performance in Mathematics?" (Paper presented at the Brit-
ish Educational Research Association Annual Conference, University of York,
September 11-14, 1997).
Hardie, Michael Howard, "The Effect of Music on Mathematics Anxiety and
Achievement " (Doctoral Diss., University of Nevada, 1990).
Kopp, George W., "The Effects of Stimulating and Relaxing Music on Children
Taking a Fourth Grade Arithmetic Test" (Doctoral Diss., Syracuse University,
1958).

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms
166 Kathryn Vaughn
Kvet, Edward J., "Excusing Elementary Students from Regular Classroom Activities
for the Study of Instrumental Music: The Effect on Sixth-Grade Reading, Lan-
guage, and Mathematics Achievement" (Doctoral Diss., University of Cincinnati,
1982).
Manthei, Mike, and Steve N. Kelly, Effects of Popular and Classical Background Music on
the Math Test Scores of Undergraduate Students (University of South Florida School
of Music, 1998). Available at http://arts.usf.edu/music/rpme/effects.html.
McCarthy, Kevin, "Music Performance Group Membership and Academic Success:
A Descriptive Study of One 4-Year High School." (Paper presented at the Colorado
Music Educators Association 1992).
Miller, Leon K., and Michael Schyb, "Facilitation and Interference by Background
Music," Journal of Music Therapy 26, no. 1 (1989): 42-54.
Mowesian, Richard, and Margaret R. Heyer, "The Effect of Music as a Distraction on
Test-Taking Performance." Measurement and Evaluation in Guidance 6, no. 2 (1973):
105-10.

Neufeld, K. Allen, "Understanding of Selected Pre-number Concepts: Relationships


to a Formal Music Program." Alberta Journal of Educational Research 32, no. 2
(1986): 132-39.
Norton, William Arnold, "Effects of Loudness and Type of Music on Figural Cre-
ativity. Reading, Coding, and Arithmetic Computation Tasks" (Doctoral Diss.,
University of Georgia, 1971).
Tucker, Alexander, and Brad J. Bushman, "Effects of Rock and Roll Music on Math-
ematical, Verbal, and Reading Comprehension Performance," Perceptual and Mo-
tor Skills 72, no. 3 Part 1 (1991): 942.
Weeden, Robert Edward. "A Comparison of the Academic Achievement in Reading
and Mathematics of Negro Children whose Parents are Interested, Not Interested,
or Involved in a Program of Suzuki Violin" (Doctoral Diss., North Texas State
University, 1971).
Wheeler, Lester T., and Viola D. Wheeler. "The Intelligence of Music Students," The
Journal of Educational Psychology 42 (1951): 223-230.
Wolf, Robert Hill, and Frederick F. Weiner, "Effects of Four Noise Conditions on
Arithmetic," Perceptual and Motor Skills 35 (1972): 928-30.
Wolfe, David, "Effects of Music Loudness on Task Performance and Self-Report of
College-Age Students," Journal of Research in Music Education 31 (1983): 193-201.

Acknowledgments. This research was supported by the Bauman Foundation. I thank


John Landrum Bryant of the Bauman Foundation for his support and guidance. I
thank Al Beaton and Ron Butzlaff for statistical advice. I also thank Ellen Winner and
Lois Hetland for comments on earlier drafts of this paper. Jenny Martin, Brian Moss,
Elisabeth Moriarty-Ambrozaitis, Lisa French, Kim Garris, Nandita Gosh, Leah
Okimoto, and Daniel Schneider helped in searching for and acquiring all of the po-
tential studies. Suzanne Bryce, Nancy Ervin, and Nelly Burd of the Educational Test-
ing Service generously provided me with the data from the Scholastic Assessment
Tests. Requests for reprints should be sent to Kathryn Vaughn, care of Ellen Winner,
Department of Psychology, Boston College, Chestnut Hill, MA 02147, or by email at
kvv@mediaone.net.

This content downloaded from 193.140.46.30 on Wed, 05 Dec 2018 12:24:45 UTC
All use subject to https://about.jstor.org/terms

You might also like