Vol 30 Issue 5

Journal for Research in Mathematics Education
1999, Vol. 30, No. 5, 520–540
A Meta-Analysis of the Relationship

Between Anxiety
Toward Mathematics and
Achievement in Mathematics
Xin Ma, University of Alberta, Canada
In this meta-analysis I examined 26 studies on the relationship between anxiety

toward mathematics and achievement in mathematics among elementary and secondary
students. The common population correlation for the relationship is significant (–.27).
A series of general linear models indicated that the relationship is consistent across
gender groups, grade-level groups, ethnic groups, instruments measuring anxiety, and
years of publication. The relationship, however, differs significantly among instruments
measuring achievement as well as among types of publication. Researchers using stan-
dardized achievement tests tend to report a relationship of significantly smaller magni-
tude than researchers using mathematics teachers’ grades and researcher-made achieve-
ment tests. Published studies tend to indicate a significantly smaller magnitude of the
relationship than unpublished studies. There are no significant interaction effects
among key variables such as gender, grade, and ethnicity.
Key Words: Achievement; Anxiety; Meta-analysis; Review of research; Secondary, 5–12
There is an increasing recognition that affective factors play a critical role in the
teaching and learning of mathematics (McLeod, 1992, 1994). One affective factor
that “has probably received more attention than any other area that lies within the
affective domain” is anxiety toward mathematics (McLeod, 1992, p. 584). Aiken
(1960) considered mathematics anxiety a “relative” of the general attitude toward
mathematics, only being more visceral. Most researchers, however, consider math-
ematics anxiety to be a construct that is distinct from attitude toward mathematics.
For example, McLeod (1992) stated that the term attitude “does not seem adequate
to describe some of the more intense feelings that students exhibit in mathematics
classrooms” (p. 576), such as anxiety, confidence, frustration, and satisfaction.
Mathematics anxiety is often referred to as “the general lack of comfort that
someone might experience when required to perform mathematically” (Wood,
1988, p. 11) or the feeling of tension, helplessness, and mental disorganization one
has when required to manipulate numbers and shapes (Richardson & Suinn, 1972;
Tobias, 1978). Mathematics anxiety can take multidimensional forms including,
for example, dislike (an attitudinal element), worry (a cognitive element), and fear
(an emotional element) (see Hart, 1989; Wigfield & Meece, 1988).
Spielberger (1972) conceptualized anxiety as a state, a trait, and a process.
Through his model of anxiety-as-process, he explained anxiety as a result of a chain
reaction that consisted of a stressor, a perception of threat, a state reaction, cogni-
Copyright © 1999 The National Council of Teachers of Mathematics, Inc. www.nctm.org. All rights reserved.
This material may not be copied or distributed electronically or in any other format without written permission from NCTM.
Xin Ma 521
tive reappraisal, and coping. Cemen (1987) also defined mathematics anxiety as
an anxious state in response to mathematics-related situations that are perceived
as threatening to self-esteem. In her model of mathematics anxiety reaction, envi-
ronmental antecedents (e.g., negative mathematics experiences, lack of parental
encouragement), dispositional antecedents (e.g., negative attitudes, lack of confi-
dence), and situational antecedents (e.g., classroom factors, instructional format)
are seen to interact to produce an anxious reaction with its physiological manifes-
tations (e.g., perspiring, increased heart beat).
Although the search for causes of mathematics anxiety is often unsuccessful
(Gough, 1954), many researchers have reported the consequences of being anxious
toward mathematics, including the inability to do mathematics, the decline in
mathematics achievement, the avoidance of mathematics courses, the limitation in
selecting college majors and future careers, and the negative feelings of guilt and
shame (Armstrong, 1985; Betz, 1978; Brush, 1978; Burton, 1979; Donady &
Tobias, 1977; Hendel, 1980; Preston, 1986/1987; Richardson & Suinn, 1972;
Tobias & Weissbrod, 1980). Therefore, not only are the professional and economic
gains that would result from changing mathematics anxiety into mathematics
confidence indispensable, the psychological boost that individuals experience
when they are successful in mathematics is also important (National Research
Council, 1989).
Theoretical models of the relationship between anxiety toward mathematics and
achievement in mathematics are complicated to establish (Gliner, 1987; Mevarech
& Ben-Artzi, 1987), but researchers have demonstrated various characteristics of
the relationship. The traditional arousal theorists state that there exists an optimal
level of arousal around the middle of the arousal dimension—optimal both in terms
of performance and in the sense of hedonic tone (being most pleasant, see Hebb,
1955). This idea is often graphically represented as an inverted-U curve depicting
a curvilinear relationship between anxiety and performance. Thus, this arousal
theory indicates that some anxiety is beneficial to performance, but after a certain
point it undermines performance. Most researchers, however, start with the linear
notion that anxiety seriously impairs performance (see Lazarus, 1974). Specifically,
a higher level of anxiety is associated with a lower level of achievement. This nega-
tive relationship has been displayed across several age populations. For example,
mathematics anxiety is negatively correlated with mathematics performance among
adults in general (Quilter & Harper, 1988) and among college students in partic-
ular (Betz, 1978; Frary & Ling, 1983). In his meta-analysis, Hembree (1990)
reported an average correlation of –.31 between anxiety and achievement for
college students.
This negative relationship also appears at the elementary and secondary school
levels (e.g., Chiu & Henry, 1990; Lee, 1991/1992; Meece, Wigfield, & Eccles,
1990). Hembree (1990) reported an average correlation of –.34 for school students,
concluding that mathematics anxiety seriously constrains performance in mathe-
matical tasks and that reduction in anxiety is consistently associated with improve-
ment in achievement. Mathematics anxiety is usually associated with mathematics
522 Anxiety and Achievement
achievement individually but not necessarily collectively. For example, a student’s

level of mathematics anxiety can significantly predict his or her mathematics
performance (Fennema & Sherman, 1977; Hendel, 1980; Rounds & Hendel, 1980;
Wigfield & Meece, 1988). But when the effects of previous mathematics perfor-
mance, attitude toward mathematics, and mathematics self-concept are controlled,
the influence of mathematics anxiety becomes either nonsignificant or substantially
reduced (Betz, 1978; Brush, 1980; Fennema & Sherman, 1977; Rounds & Hendel,
1980; Siegel, Galassi, & Ware, 1985).
The theoretical explanation of the negative relationship stems mainly from the
theory of test anxiety. Some researchers have regarded mathematics anxiety as a
kind of subject-specific test anxiety (e.g., Brush, 1981). Although others have
argued that mathematics anxiety may be a psychologically different construct
from test anxiety (e.g., Richardson & Woolfolk, 1980), “a tacit belief has seemed
to prevail that test-anxiety theory can be used to support both constructs” (Hembree,
1990, p. 34). Two theoretical models have been influential in the research on math-
ematics anxiety. In the interference model, based on the work of Liebert and
Morris (1967), Mandler and Sarason (1952), and Wine (1971), researchers have
described mathematics anxiety as a disturbance of the recall of prior mathematics
knowledge and experience. Consequently, a high level of anxiety causes a low level
of achievement. In the deficits model, Tobias (1985) regarded mathematics anxiety
as the remembrance of poor mathematics performance in the past and believed that
poor performance causes high anxiety. According to this deficits model, a student’s
low level of mathematics achievement is attributed to poor study habits and defi-
cient test-taking skills instead of to mathematics anxiety. Some have made theo-
retical efforts to integrate these competing frameworks; one result was the “limited
cognitive capacity formulation” (see Tobias, 1985, p. 138).
Some researchers have attempted to introduce mediating variables, such as
gender, age, and race-ethnicity, into the theoretical models of the relationship. For
example, Eccles and Jacobs (1986) suggested that gender differences in mathe-
matics anxiety are attributable to gender differences in mathematics achievement.
Aiken (1970) stated that “no one would deny that sex can be an important moder-
ator variable in the predication of achievement from measures of attitudes and
anxiety” (p. 567). Hembree (1990) found that mathematics anxiety increases
during junior high grades, reaches its peak in Grades 9 and 10, and levels off during
senior high grades, implying that the relationship is a function of grade levels. Aiken
also concluded that the correlation between performance in mathematics and
anxiety toward mathematics is much stronger at the junior high level. Engelhard
(1990) showed that the correlation between mathematics anxiety and mathematics
achievement is higher among American students than among Thai students, indi-
cating that the same level of anxiety tends to be associated with achievement of
American students more strongly than with achievement of Thai students. McLeod
(1992) concluded that “it seems reasonable to hypothesize that affective factors are
particularly important to differences in performance between groups that come from
different cultural backgrounds” (p. 587). Lacking in the theoretical models,
Xin Ma 523
however, are variables that depict the genesis of mathematics anxiety, the nature
of treatment for mathematics anxiety, and the utility of mathematics anxiety to
improve mathematics performance (Betz, 1978).
There have been few systematic reviews of research on mathematics anxiety.
Leder (1987), McLeod (1992, 1994), and Reyes (1980, 1984) reviewed the affec-
tive domain as it relates to the cognitive domain in mathematics, but their discus-
sions on mathematics anxiety are quite limited. On the basis of a review of a number
of studies concerning gender differences in mathematics anxiety, Hunt (1985)
concluded that there are evident differences between males and females in math-
ematics anxiety and that researchers need to examine the reason females are more
anxious about mathematics than their male counterparts. Later, some researchers
conducted a meta-analysis of the same studies and reached a different conclusion;
they found that gender differences are small in size but that when differences do
exist, females show more anxiety than males (Hyde, Fennema, Ryan, Frost, &
Hopp, 1990). Wood (1988) reviewed research on mathematics anxiety manifested
among elementary teachers and suggested that mathematics teachers’ anxiety
toward mathematics was likely to be transmitted to their students.
There have been even fewer systematic reviews of the relationship between math-
ematics anxiety and mathematics achievement. Aiken (1970), in his narrative
review, highlighted a weak, though statistically significant, negative relationship;
he noticed that the magnitude of the relationship is usually somewhat smaller in
absolute value than that of the relationship between attitude toward mathematics
and achievement in mathematics. Hembree (1990), in his meta-analysis, did not
focus on the relationship between mathematics anxiety and mathematics achieve-
ment, although some analyses were done on that topic, and he concluded that there
exists a significant but negative relationship. However, because it was based on
58 studies on college students but only 7 studies on elementary and secondary
students, Hembree’s conclusion seems to apply more to college students than to
precollege students.
In his meta-analysis Hembree (1990) displayed a simplified picture of the rela-
tionship between mathematics anxiety and mathematics achievement. Many issues
basic to this relationship, however, remain unclear. There is a need to locate more
research on this relationship and display further the characteristics of the relation-
ship. Both theoretical and practical work can benefit from, for example, a compre-
hensive review of gender differences in the relationship and a study of changes in
the relationship as students progress through school. The main research questions
of my meta-analysis are (a) What is the magnitude of the relationship between
anxiety toward mathematics and achievement in mathematics? and (b) How does
the magnitude of the relationship fluctuate in response to various study features such
as gender, grade level, ethnicity, instruments used to measure anxiety and achieve-
ment, year of publication, and type of publication?
METHOD
Sample of Studies
Although Aiken and Dreger (1961) have demonstrated that it is possible to
construct an inventory to measure mathematics-related anxiety, instruments that
were developed for this purpose in the 1960s might be considered primitive. For
example, the item response theory (IRT), an important statistical method used to
examine the psychometric properties of response data, was not well developed until
the 1970s. Effective IRT software programs were developed even later. To increase
the accuracy of measurement in this meta-analysis, I examined studies reported after
1975 because attitudinal instruments have been greatly refined since the late 1970s
(Pedersen, Bleyer, & Elmore, 1985).
I used a three-step approach to search for relevant studies on the relationship
between mathematics anxiety and mathematics achievement. First, to promote a
broad search of several computerized databases for the years 1975 through the
present, I used the key topic-related descriptors (mathematics, achievement, and
anxiety) as independent words (see Dusek & Joseph, 1983). The databases searched
were (a) Educational Resources Information Center (ERIC), (b) Psychological
Abstracts (PSY), (c) Dissertation Abstracts International (DAI), and (d)
International ERIC, which comprises Australian Education Index (AEI), British
Education Index (BEI), and Canadian Education Index (CEI).
The next step was to find, on the basis of the same descriptors, both qualitative
and quantitative reviews published since 1975 as a means to enrich the pool of
studies. Reference lists from Aiken (1976), Hembree (1990), Hunt (1985), Hyde,
Fennema, and Lamon (1990), Leder (1987), McLeod (1992, 1994), Reyes (1980,
1984), and Wood (1988) were checked for relevant studies. Finally, I conducted
a manual search of seven leading journals in education, particularly in mathematics
education, for the years 1975 through the present. These journals were American
Educational Research Journal, Educational Studies in Mathematics, Journal for
Research in Mathematics Education, Journal of Educational Psychology, Journal
of Educational Research, Review of Educational Research, and School Science and
Mathematics.
Using this search procedure, I located many journal articles, dissertations, and
ERIC documents on various aspects of the topic, then studied the abstracts to
screen all the studies. I obtained and read promising studies and those that could
not be evaluated from their abstracts. A study was included in this meta-analysis
if it (a) was an investigation of the relationship between mathematics anxiety and
mathematics achievement, (b) did not have any experimental interventions on
either anxiety or achievement, (c) reported on students at the elementary or
secondary school level, and (d) reported quantitative data in sufficient detail for
calculation of effect size. On the basis of these criteria, I selected 26 individual
studies for this meta-analysis.
Xin Ma 525
Coding of Selected Studies

Each study was coded for several independent variables that depicted design
features as well as for author(s), year of publication, and type of publication.
Design features included gender, grade, ethnicity, sample size, and instruments used
to measure anxiety and achievement. The coding of the 26 individual studies is
reported in Table 1.
The dependent variable was effect size, indicating the relationship between
anxiety toward mathematics and achievement in mathematics. Because researchers
in all selected individual studies used correlation coefficients to describe the rela-
tionship, in this meta-analysis I used the common metric of r as its effect-size
measure (see Rosenthal, 1991). One effect size was obtained from each study unless
the study contained independent samples to measure the relationship. Gender
groups, grade levels (in cross-sectional designs), and ethnic groups in a single study
were considered separate primary studies (L. V. Hedges, personal communication,
1987, cited in Hyde, Fennema, & Lamon, 1990).
Table 1
Descriptive Information of Studies on the Relationship Between
Anxiety Toward Mathematics and Achievement in Mathematics (in Chronological Order)
Study Effect size Study feature Instrument
Sepie & Keeling, 1978 –.28 N = 132, Grade 6, male, MASC/PAT
New Zealander
Sepie & Keeling, 1978 –.30 N = 114, Grade 6, female, MASC/PAT
New Zealander
Sandman, 1979 –.47 N = 184, Grade 8 MAI/MAT
Sandman, 1979 –.17 N = 229, Grade 11 MAI/MAT
Brassell, Petry, & Brooks, 1980 –.30 N = 714, Grade 7 MAI/CTBS
Suinn & Edwards, 1982 –.59 N = 28, Grades 7–11 MARS-A/MTG
Suinn & Edwards, 1982 –.20 N = 1009, Grade 12 MARS/MTG
Saigh & Khouri, 1983 –.60 N = 73, Grades 9–12, male, MARS/MTG
Lebanese
Saigh & Khouri, 1983 –.48 N = 60, Grades 9–12, female, MARS/MTG
Lebanese
Eccles & Jacobs, 1986 –.17 N = 164, Grades 7–9 RMQ/SAT-M
Donnelly, 1987 –.85 N = 177, Grade 10 MARS/CAT
Gliner, 1987 –.12 N = 95, Grades 9–12 MARS/CTBS
Wahl, 1987 –.31 N = 59, Grade 8 MARS/MTG
Suinn, Taylor, & Edwards, 1988 –.28 N = 105, Grades 4–6, Hispanic MARS/SAT
Wigfield & Meece, 1988 –.22 N = 564, Grades 6–12, Year 1, MAQ/MTG
White
Wigfield & Meece, 1988 –.26 N = 564, Grades 6–12, Year 2, MAQ/MTG
White
Wither, 1988 –.51 N = 271, Grade 10, male, MARS/MTG
Australia
Wither, 1988 –.24 N = 245, Grade 10, female, MARS/MTG
Australian
Reavis, 1989 –.28 N = 407, Grades 9–12, White MAS/SAT
Suinn, Taylor, & Edwards, 1989 –.31 N = 1119, Grades 4–6 MARS/SAT
Baya’a, 1990 –.42 N = 418, Grades 9–12, Israeli MARS/MTG
Chiu & Henry, 1990 –.37 N = 50, Grade 5 MASC/MTG
Table 1, continued
Descriptive Information of Studies on the Relationship Between
Anxiety Toward Mathematics and Achievement in Mathematics (in Chronological Order)
Study Effect size Study feature Instrument
Engelhard, 1990 –.24 N = 4091, Grade 8 SIMS-AS/SIMS
Engelhard, 1990 –.14 N = 3613, Grade 8, Thai SIMS-AS/SIMS
Meece, Wigfield, & Eccles, 1990 –.13 N = 250, Grades 7–9, Year 1, MAQ/RMT
White
Meece, Wigfield, & Eccles, 1990 –.21 N = 250, Grades 7–9, Year 2, MAQ/RMT
White
Bieschke & Lopez, 1991 –.51 N = 289, Grade 10 MAS/SAT-M
Bush, 1991 –.49 N = 584, Grades 4–6, concept MARS/ITBS
Bush, 1991 –.19 N = 584, Grades 4–6, application MARS/ITBS
Cooper & Robinson, 1991 –.47 N = 290, Grade 12, White MAS/MMPT
Green, 1991 –.35 N = 496, Grades 7–12, Australian MAS/MTG
Mevarech, Silber, & Fine, 1991 –.44 N = 149, Grade 6, Israeli RMQ/AAT
Thorndike-Christ, 1991 –.47 N = 1516, Grades 6–12 MAS/MTG
Hadfield, Martin, & Wooden, –.23 N = 358, Grades 8–10, MARS/CTBS
1992 Native American
Lee, 1992 –.27 N = 255, Grade 6 MAS/MTG
Note. In the column of instrument, the slash separates anxiety instrument from achievement instrument.
For anxiety instrument, MAI = Mathematics Attitude Inventory, MAQ = Mathematics Anxiety
Questionnaire, MARS = Mathematics Anxiety Rating Scale, MAS = Mathematics Anxiety Scale,
MASC = Mathematics Anxiety Scale for Children, RMQ = researcher-made questionnaire, and
SIMS-AS = Second International Mathematics Study-Anxiety Scale. For achievement instrument,
AAT = Arithmetic Achievement Test, CAT = California Arithmetic Test, CTBS = Comprehensive
Test of Basic Skills, ITBS = Iowa Test of Basic Skills, MAT = Mathematics Achievement Test, MMPT
= Missouri Mathematics Placement Test, MTG = mathematics teacher grading, PAT = Progressive
Achievement Test, RMT = researcher-made test, SAT = Stanford Achievement Test, SAT-M =
Scholastic Aptitude Test-Mathematics, SIMS = Second International Mathematics Study.
Meta-analytic methodology literature is not explicit on the use of longitudinal

studies. Longitudinal data can be viewed as being within a single study in which
correlations are aggregated to represent the effect size of the study. Willett and
Singer (1991) argued, however, that “a complex longitudinal time-dependent
process cannot be adequately summarized by a single statistic” (p. 430). In line with
this argument, longitudinal data in a study were treated as several independent
primary studies based on different grade levels. Using this treatment, I maintain a
clear identity or an explainable background for each effect size at the primary-study
level so that statistical results are easy to interpret and understand.
Characteristics of the Sample

The sample of 26 studies included 18 published articles, 3 unpublished articles,
and 5 dissertations. The published studies appeared in 14 journals. Among the 26
studies, 21 were published after 1985. The median year of publication was 1991.
In total, 6 instruments were used to measure anxiety, with the Mathematics Anxiety
Rating Scale (MARS) as the most frequently used one (in 12 of 26 studies), and 9
instruments were used to measure achievement in mathematics. A total of 18 279
Xin Ma 527
students across Grades 5 to 12 participated in these studies. The largest sample size
was 4 091, and the smallest sample size was 28. The studies had an average of 703
students per sample. Most studies were mixed in terms of gender, grade, and
ethnicity. The 26 studies generated 37 effect sizes for this meta-analysis.
Statistical Procedure
Variation among effect sizes was examined through Hedges’s Q test of homo-
geneity (Hedges & Olkin, 1985). This test was used to determine whether popu-
lation effect sizes were relatively consistent across unweighted effect sizes. If the
test showed a nonsignificant Q value, which indicated that effect sizes were homo-
geneous, effect sizes were then combined following procedures of combining
estimates of correlation coefficients (Hedges & Olkin, 1985). If the test showed a
significant Q value, which indicated that effect sizes were heterogeneous, I took
Hembree and Dessart’s (1986) suggestion to delete outliers repeatedly until the
remaining effect sizes became homogeneous.
When effect sizes are homogeneous, the population correlation can be determined
by predictor variables. The substantive rationale is that studies differ because of
different research design characteristics (Hedges & Olkin, 1985). In this study I
fitted homogeneous effect sizes into a general linear regression model through
which I examined the effects of a number of independent variables on the rela-
tionship between mathematics anxiety and mathematics achievement. I employed
the weighted least-squares procedures for fitting general linear models to correla-
tions as outlined by Hedges and Olkin. Sample size was used to create weight for
the regression analysis (see Hedges & Olkin, 1985) but was not entered into the
regression equation (see Schram, 1996).
To better illustrate the overall effect (population coefficient) in this meta-
analysis, I used the percentage of distribution nonoverlap, or the U3 statistic
(Cohen, 1977), to denote the percentage of participants in the group with the
larger anxiety mean whose achievement scores were exceeded by the achievement
scores of half the participants in the group with the smaller anxiety mean. If the
dependent variable was standardized, the U3 statistic also indicated the change in
scores or percentiles when a participant moved from one group to the other.
RESULTS
Overall Effects
The test of homogeneity of the 37 effect sizes was significant. Three evident outliers
were deleted: –.84 in Donnelly (1986/1987) as well as .49 and .19 in Bush (1991).
The removal of these outliers significantly improved the homogeneity of the
remaining effect sizes. The homogeneity test was not significant at the .01 level (Q
= 55.82, df = 33). Effect sizes were then combined with respect to their sample sizes.
The weighted estimator of population correlation was –.27, which was significant at
the .01 level, indicating that it was unlikely (likelihood less than 1 time in 100) that
the observed significance of this population correlation occurred by chance alone.
The U3 statistic corresponding to a population correlation of –.27 is .71. To under-

stand the meaning of the U3 statistic, one should imagine that all the students are
classified into two groups according to their levels of mathematics anxiety. One
group has low mathematics anxiety; the other has high mathematics anxiety. Then,
the U3 statistic of .71 indicates that the average student in the group of low math-
ematics anxiety would have a score in mathematics achievement that is greater than
the scores of 71% of the students in the group of high mathematics anxiety.
Expressed another way, measures (or treatments) that resulted in movement of a
typical student in the group of high mathematics anxiety into the group of low math-
ematics anxiety would be associated with improvement of the typical student’s level
of mathematics achievement from the 50th to the 71st percentile.
Inasmuch as the effect sizes all shared the same population correlation, varia-
tion among effect sizes existed mainly because studies differed according to a
number of research design characteristics. A general linear regression was used to
model this variation. Using the general linear model enabled me to identify the
significant variables responsible for the variation among effect sizes and to gain
insight into several practical concerns, such as gender differences and age differ-
ences, regarding the relationship between mathematics anxiety and mathematics
achievement. Table 2 displays the models considered in this meta-analysis.
Table 2
Selected Models From the General Linear Regression Analysis of the
Relationship Between Anxiety Toward Mathematics and Achievement in Mathematics
Regression test Residual test
Model QR df QE df
Gender effect (2 vectors: male vs. female; mixed vs. female) 17.74*** 2 214.01*** 31
Grade effect 18.34*** 2 213.41*** 31
(2 vectors: Grades 4–6 vs. 10–12; Grades 7–9 vs. 10–12)
Race-ethnicity effect (1 vector: mixed vs. unmixed) 16.50*** 1 215.25*** 32
Anxiety instrument (1 vector: MARS vs. others) 4.43* 1 227.32*** 32
Achievement instrument 36.19*** 1 195.56*** 32
(1 vector: psychometric vs. nonpsychometric)
Publication type 78.07*** 2 153.68*** 31
(2 vectors: published and dissertation vs. unpublished)
Publication year (continuous variable) 0.28 1 231.48*** 32
Gender, grade, race, 133.58*** 10 98.17*** 23
anxiety/achievement instrument, publication type/year
* p < .05. ** p < .01. *** p < .001.
Effect of Gender
Effect coding (see Cohen & Cohen, 1983) was used to create two variables. The
female group was used as the baseline against which the male group and the
mixed group were compared. Table 2 shows that the Q statistic (the weighted sums
Xin Ma 529
of squares) for gender effects explains a statistically significant amount of the vari-
ability in the effect sizes (QR = 17.74, df = 2). However, this amount is a trivial
portion of the total variance (231.75); thus the Q statistic for error is statistically
significant (QE = 214.01, df = 31). The two gender-related variables did not have
appreciable effects on the magnitude of the relationship (regression coefficient, β,
= –0.12; standard error, SE, = 0.09 for the male vs. female comparison). This finding
indicates that the relationship between mathematics anxiety and mathematics
achievement was similar for males and females.
Effects of Grade Level

Three grade-level groups were formed in this meta-analysis: Grades 4 through
6, Grades 7 through 9, and Grades 10 through 12 (there were no studies examining
the relationship in Grades 1, 2, or 3). Effect sizes were effect coded so that the rela-
tionship in Grades 10 through 12 was the baseline against which the relationships
in Grades 4 through 6 and Grades 7 through 9, respectively, were compared. As
did gender effects, grade effects accounted for a very small, though statistically
significant, percentage of the total variance. The Q statistic for error remained
substantial (QE = 213.41, df = 31). Grade levels did not have statistically signifi-
cant effects on the relationship between mathematics anxiety and mathematics
achievement (for the Grades 4 through 6 vs. Grades 10 through 12 comparison:
β = –0.02, SE = 0.03; for the Grades 7 through 9 vs. Grades 10 through 12 compar-
ison: β = 0.05, SE = 0.04). Therefore, the relationship was consistent across the
three grade-level groups.
Effects of Ethnicity
Because few ethnic groups were involved in the selected studies, it was impos-
sible to classify effect sizes into conventional ethnic groups (White, Black, etc.).
Instead, in this meta-analysis, I compared the magnitude of the relationship between
mixed and unmixed ethnic groups. Dummy coding (see Cohen & Cohen, 1983) was
used to represent this variable, which, as shown in Table 2, explains a very small,
though statistically significant, amount of the total variance in the effect sizes (QE
= 215.25, df = 32). This small effect (β = –0.06, SE = 0.04) indicated that the rela-
tionship was consistent between mixed and unmixed ethnic groups.
Effects of Instrument Measuring Mathematics Anxiety

Because the Mathematics Anxiety Rating Scale (MARS) was employed to
measure mathematics anxiety in many studies, in this meta-analysis I created a
dummy variable to compare the effect of the MARS with that of other instruments
(as the baseline effect) used to measure mathematics anxiety. Table 2 shows the
effects of instruments on the relationship between mathematics anxiety and math-
ematics achievement. Instruments explained a trivial amount of variance (QR = 4.43,
df = 1) and had a very small effect (β = –0.04, SE = 0.05). This finding indicates
that the relationship was consistent between studies using the MARS and those
using other instruments to measure mathematics anxiety.
Effects of Instrument Measuring Mathematics Achievement

Quite diverse instruments were used to measure mathematics achievement in the
selected studies. However, there was a balance between use of commercially
developed instruments and other instruments including researcher-designed math-
ematics tests and mathematics teachers’ grades. A dummy variable was created with
instruments that are commercially developed as the baseline effect. This variable
seems to be a better predictor than previous variables (QR = 36.19, df = 1), although
the Q statistic for error is still substantial (QE = 195.56, df = 32). This variable also
had a statistically significant effect on the relationship between mathematics
anxiety and mathematics achievement at the .05 level (β = –0.09, SE = 0.04).
Therefore, studies in which standardized achievement tests (the commercially
developed instruments) were used showed a significantly smaller magnitude of the
relationship than studies in which researcher-made achievement tests and mathe-
matics teachers’ grades were used as achievement measures.
Effects of Type of Publication

The type of publication comprises three categories: published article, disserta-
tion, and unpublished article. Using published articles as the baseline effect, I used
effect coding to create two variables (see Table 2). The Q statistic shows that this
model explains a statistically significant amount of the total variation in the effect
sizes (QR = 78.07, df = 2). This model is the best individual model because it
explains the greatest amount of total variance. One of the two variables (unpub-
lished vs. published) was statistically significant at the .05 level (β = –0.12, SE =
0.04); the other variable (dissertation vs. published) was not statistically signifi-
cant (β = 0.04, SE = 0.06). Therefore, published articles showed a significantly
smaller magnitude of the relationship than unpublished articles, whereas published
articles and dissertations showed similar magnitudes of the relationship.
Effects of Year of Publication

The year of publication was used as a continuous variable in the general linear
model. It is the only variable that explained little (a statistically nonsignificant
amount) of the total variance (QR = 0.28, df = 1). Year of publication also showed
no effect on the relationship between mathematics anxiety and mathematics
achievement (β = –0.00, SE = 0.01). This finding indicates that studies from
different periods of time showed consistent effect sizes.
Interaction Effects
A series of general linear models containing interaction terms were tested to
examine whether the effect of one variable depends on the levels of another vari-
Xin Ma 531
able (these models are not shown in Table 2). The interactions between gender and
grade (QR = 8.19, df = 1), between grade and ethnicity (QR = 12.79, df = 2), as well
as between anxiety instrument and achievement instrument (QR = 4.91, df = 1) were
not statistically significant. However, the interaction between publication year and
publication type explained a statistically significant amount of the total variance
in the effect sizes (QR = 80.83, df = 2). Results show that the differences in effect
sizes between published and unpublished articles decreased over time in the period
examined in this meta-analysis.
The Final Model

The final model included all variables discussed in individual models. The Q
statistic indicates that this final model explains a statistically significant and prac-
tically substantial amount of variance in the effect sizes (QR = 133.58, df = 10). A
total of 58% of the variance among effect sizes was accounted for in this final model.
In terms of regression coefficients, the variable that compares unpublished with
published articles is the only statistically significant variable in the model (β = –0.16,
SE = 0.06, p = .011), indicating that published articles reported a significantly
weaker relationship between mathematics anxiety and mathematics achievement
than unpublished articles. Because of the statistically significant interaction between
year of publication and type of publication discussed earlier, I attempted to include
this interaction term in the final model. The resultant model, however, differed very
little from the final model in Table 2, in terms of the fitting statistics. For simplicity
that interaction term was then removed from the final model.
DISCUSSION
Principal Findings
In this meta-analysis I demonstrate that the common population correlation for
the relationship between anxiety toward mathematics and achievement in mathe-
matics was –.27. A series of general linear models were fitted to examine the major
research design characteristics that determine the variation among effect sizes.
Results show that the relationship between mathematics anxiety and mathematics
achievement is consistent across gender groups (male, female, and mixed), grade-
level groups (Grades 4 through 6, Grades 7 through 9, and Grades 10 through 12),
ethnic groups (mixed and unmixed), instruments used to measure anxiety (MARS
and others), and years of publication. The relationship, however, differs significantly
between types of instruments used to measure achievement as well as among
types of publication. Researchers using standardized achievement tests tended to
report a significantly weaker relationship than those using researcher-made achieve-
ment tests and mathematics teachers’ grades. Published studies tended to indicate
a significantly weaker relationship than unpublished studies. There were no statis-
tically significant interaction effects among key variables such as gender, grade,
and ethnicity.
Theoretical and Practical Implications

This meta-analysis shows support for the findings of significance of the rela-
tionship between mathematics anxiety and mathematics achievement for school
students, as previously reported (e.g., Armstrong, 1985; Eccles, 1985; Hackett,
1985; Wigfield & Meece, 1988). More important, in this meta-analysis I have quan-
tified the potential improvement in mathematics achievement when mathematics
anxiety is reduced. Hembree (1990) concluded that as psychological treatments,
systematic desensitization and “anxiety management training and conditional inhi-
bition were highly successful in reducing mathematics anxiety levels” (p. 43). The
current meta-analysis further indicates that such a reduction may be associated with
an improvement from the 50th to 71st percentile in mathematics achievement for
an average student highly anxious about mathematics. This amount of improve-
ment is substantial according to any threshold standards in behavioral sciences (see
Cohen & Cohen, 1983).
This meta-analysis has important implications for educational intervention.
Those advocating treatment programs, such as self-management of emotional
stress, for students with mathematics anxiety have traditionally emphasized affec-
tive factors only (see Richardson & Woolfolk, 1980). Although various treatment
programs appear to be effective in reducing mathematics anxiety (Hembree, 1990),
in few has attention been paid to the role of cognitive factors such as skill devel-
opment. The significant, negative association between mathematics achievement
and mathematics anxiety, as reported in this meta-analysis, indicates the potential
value of cognitively based treatments. That is, measures or treatments that help
students overcome their cognitive difficulties in the learning of mathematics may
be associated with an appreciable reduction in mathematics anxiety. There has been
effort to include a stronger cognitive orientation in the research of mathematics
anxiety because of the finding that cognitive style is associated with mathematics
anxiety (e.g., Hadfield & Maddux, 1988). For example, Hunsley (1987) found that
mathematics anxiety is related to cognitive factors such as pessimistic post-exam
appraisals and negative internal dialogues. Handler (1990) suggested that a cogni-
tive process approach reduces mathematics anxiety through (a) making knowledge
work for the learner, (b) joining skill and content, (c) linking motivation to cogni-
tion, and (d) using social communities. By taking cognitive factors into account,
program developers may well improve the effectiveness of various treatment
programs for mathematics anxiety.
Researchers recognize that the nature of the relationship between mathematics
anxiety and mathematics achievement is not clear (see Reyes, 1984). In this meta-
analysis I attempted to partially unfold the “mystery” by breaking the effect sizes
of the relationship into categories and examining the differences among those cate-
gories. I see implications related to several key issues in this area. First, in this meta-
analysis I found no significant gender differences on the relationship between math-
ematics anxiety and mathematics achievement. Ma and Kishor (1997) reported a
similar finding—that the relationship between attitude toward mathematics and
Xin Ma 533
achievement in mathematics is the same for males and females. These findings do
not support Aiken’s (1970) conclusion that “measures of attitudes and anxiety may
be better predictors of the achievement of females than of males” (p. 567).
Another important issue is the developmental characteristics of the relationship.
In this meta-analysis I studied three grade-level groups (Grades 4 through 6,
Grades 7 through 9, and Grades 10 through 12) and found that the relationship
between mathematics anxiety and mathematics achievement is significant from
Grade 4 on. This finding seems reasonable because mathematics anxiety can arise
at any time during schooling (see Lazarus, 1974). But this meta-analysis further
shows that once mathematics anxiety takes shape, its relationship with mathematics
achievement is consistent across grade levels. This finding, when considered
together with reports that uneasiness, worry, and anxiety associated with the
learning of mathematics increase during the early adolescent years (e.g., Brush,
1985; Hembree, 1990; Meece, 1981; Wigfield & Meece, 1988), implies that an
increasing decline in mathematics achievement during early secondary schooling
is possible for adolescent students with mathematics anxiety. For these students,
one way to avoid high mathematics anxiety and poor mathematics achievement is
to avoid mathematics courses, particularly advanced courses (e.g., Armstrong, 1985;
Cemen, 1987). Early detection of mathematics anxiety leads to implementation of
effective treatment programs (e.g., Betz, 1978; Lazarus, 1974; McMillan, 1976).
This meta-analysis indicates that screening and treatment programs should be
introduced during the upper elementary grades (Grades 4 through 6).
Comparisons of the relationship between mathematics anxiety and mathematics
achievement among ethnic groups are rare in the literature. Because of the sparse
data on this issue, I could not compare this relationship among ethnic groups in this
meta-analysis. I did, however, find that researchers studying participants of varied
ethnic backgrounds tended to find a relationship similar to that found by researchers
who studied participants with homogeneous ethnic backgrounds. This result indi-
cates that the ethnic formation of a sample does not bias the relationship. It also raises
the likelihood that there may not be significant ethnic differences in the relation-
ship. More studies are needed to examine the relationship from the racial-ethnic
perspective before one can conclude that measures of anxiety toward mathematics
may predict the level of mathematics achievement equally well across ethnic groups.
Comments on the Instruments

Instruments are especially important in the research of the relationship between the
affective and cognitive domains. Ma and Kishor (1997) used a meta-analysis to
examine the relationship between attitude toward mathematics and achievement in
mathematics. Reporting weak mean effect sizes, they suggested that current instru-
ments measuring attitude toward mathematics seem unable to capture what constitutes
a “true” attitude. The average population correlation in the current meta-analysis was
relatively much stronger (see Cohen & Cohen, 1983). This finding indicates that current
instruments used to measure mathematics anxiety are more effective than those used
to measure attitude. Perhaps anxiety toward mathematics is easier to measure than atti-
tude toward mathematics in that it is more operationally definable for researchers and
more verbally expressible for students. But, “the preciseness with which pupils can
express their attitudes varies with level of maturity” (Aiken, 1970, p. 558).
This meta-analysis also shows that researchers using the MARS reported a rela-
tionship between mathematics anxiety and mathematics achievement similar to that
reported by researchers using other instruments. This consistency among instru-
ments with respect to anxiety may have something to do with their strong corre-
lations. For example, Dew, Galassi, and Galassi (1983, 1984) found that the corre-
lation is .68 between MAS and MARS and .78 between MAS and MAI. Note that
these correlations are considered fairly strong from the perspective of behavioral
sciences (see Cohen & Cohen, 1983). Thus, researchers’ decisions on what instru-
ments to use do not seem to affect analytic results found on the relationship. This
conjecture certainly does not mean that all the instruments measure the same
aspects of mathematics anxiety. I summarize the conceptual and operational
aspects of various instruments in Table 3.
Mathematics anxiety often originates during students’ early educational experi-
ences (Chiu & Henry, 1990). But this meta-analysis shows that studies are rare in
the early elementary grades, probably because of the lack of instruments that
measure mathematics anxiety of children at the lower elementary level. Almost all
existing instruments were originally designed for adults or secondary school students.
A version of the MARS has been developed mainly for upper elementary school
Table 3
Summary of Instruments Measuring Anxiety Toward Mathematics (in Chronological Order)
Number
Description of items of items Scale
Mathematics Anxiety Rating Scale (MARS) 98 5-point
(Richardson & Suinn, 1972) Likert
Measures students’ anxious reactions when they do mathematics in
ordinary life and in academic situations
Mathematics Attitude Inventory (MAI) (Sandman, 1974) 6 4-point
Contains six scales, one of which is Anxiety Toward Mathematics, which Likert
measures students’ mathematics anxiety in general academic situations
Mathematics Anxiety Scale (MAS) (Fennema & Sherman, 1976) 10 5-point
Measures students’ feelings of anxiety and nervousness as well as Likert
associated somatic symptoms when students use mathematics
Mathematics Anxiety Questionnaire (MAQ) (Meece, 1981) 22 7-point
Measures cognitive and affective components of mathematics anxiety Likert
parallel to those of text anxiety: dislike, lack of confidence, discomfort,
worry, fear and dread, and confusion and frustration
Second International Mathematics Study Anxiety Scale (SIMS-AS) 5 5-point
Measures the extent to which students feel afraid and scared of Likert
mathematics or feel calm and relaxed when they perform mathematical tasks
Mathematics Anxiety Scale for Children (MASC) 22 4-point
(Chiu & Henry, 1990) Likert
Describes various situations that can arouse mathematics anxiety—
from getting a new mathematics textbook to taking an important test in
a mathematics class
Xin Ma 535
students (see Suinn, Taylor, & Edwards, 1988). There is a need to develop new instru-
ments that measure mathematics anxiety in the early elementary grades.
In many studies involved in Hembree’s (1990) meta-analysis, mathematics
achievement results were reported for subscales such as concepts, computation, and
application. Because few studies in the current meta-analysis were reported this way,
a general measure of mathematics achievement was used. The results indicated that
studies using commercially developed achievement instruments reported a smaller
magnitude of the relationship between mathematics anxiety and mathematics achieve-
ment than studies using mathematics teachers’ grades and researcher-designed math-
ematics tests. This finding does not seem to support Hembree’s conclusion that
“grades in mathematics courses seemed depressed in relation to anxiety by about the
same proportion as the students’ test scores” (p. 38). Mathematics teachers’ grades
or researcher-designed mathematics tests appear to overestimate the relationship. One
possible reason is that these measures lack control of the difficulty level of the
items: The ceiling effect (items are so easy that many students perform well) or the
floor effect (items are so difficult that many students perform poorly) is more likely
to occur when mathematics teachers’ grades or researcher-designed mathematics tests
instead of commercial tests are used as achievement measures.
Unfolding the Anxiety-Achievement Dynamic in Mathematics

In a meta-analysis one usually does not pay much attention to outliers among effect
sizes. However, an examination of outliers may sometimes have important theoret-
ical or practical implications. Two of the three outliers in this meta-analysis are from
studies by Bush (1991), who found a positive, significant relationship between math-
ematics anxiety and mathematics achievement and argued that mathematics anxiety
tends to rise in students whose mathematics performance is improving. This result
“represented a contradiction of previous research on the relationship between math-
ematics anxiety and achievement” (p. 42) probably because his sample of students
had been extensively exposed to mathematics. They were either gifted students or
students, in academic tracks, with intentions to enter a career for which they would
need quantitative skills. These students are often able to control their anxiety and
channel it into the task because of their strong self-esteem and high levels of task-
related confidence (see Cemen, 1987). When this control occurs, students’ anxiety
actually facilitates their performance (Cemen, 1987). Instead of being an abnormal
result, Bush’s finding may be a hint that mathematics anxiety can be useful in
promoting mathematics achievement. Therefore, through examining sample char-
acteristics in studies by Bush, one can distinguish a special group of students whose
mathematics performance benefits from a certain level of mathematics anxiety.
Resnick, Viehe, and Segal (1982) found that a decrease in mathematics anxiety
is not associated with improvement in mathematics performance. This finding led
them to doubt that the reduction in mathematics anxiety improves mathematics
achievement. However, their sample was a group of college students with exten-
sive mathematics backgrounds. Not only might the level of mathematics anxiety
of those students be limited, but their mathematics achievement might also be high.
This study thus demonstrates a unique relationship between mathematics anxiety
and mathematics achievement for a specific group of students. Again, through
examining participant characteristics in studies by Resnick et al., one can distin-
guish another group of individuals for whom higher mathematics performance is
not associated with the reduction in mathematics anxiety.
The two examples above illustrate the dynamic nature of the relationship between
mathematics anxiety and mathematics achievement. That is, the relationship can
change dramatically for students with different social and academic background
characteristics. These social and academic characteristics of students appear to be
the key to unfolding this anxiety-achievement dynamic. When students’ charac-
teristics are diverse and unique, so are the relationships: Mathematics anxiety can
facilitate mathematics performance, can debilitate mathematics performance, or can
be unassociated with mathematics performance.
Results of this meta-analysis, considered together with the work of Buxton
(1981), can be used to derive a psychological framework in which the relationship
between mathematics anxiety and mathematics achievement can be understood as
a psychological function of emotional reaction. Emotion, belief, and attitude are
the major elements of the affective domain in the learning of mathematics (McLeod,
1992). Panic, fear, anxiety, and embarrassment have been identified as the results
of emotional reaction to mathematical tasks (Buxton, 1981). McLeod described
emotion as “hot” in that it “may involve little cognitive appraisal and may appear
and disappear rather quickly” (p. 579). Emotional reaction to academic situations
involving mathematics seems more likely to trigger mathematics anxiety that
relates to mathematics performance than do both beliefs and attitudes that are
considered “cold” and “cool” (see McLeod, 1992). Social and academic charac-
teristics broaden the focus beyond psychological perspectives to anthropological
and sociological perspectives. These characteristics mediate students’ emotional
reactions. In general, characteristics can be classified as personal (e.g., gender, age,
ethnicity, and social class), environmental (e.g., social stereotypes, mathematics
experiences, and parental encouragement), dispositional (e.g., attitude, confidence,
and self-esteem), and situational (e.g., classroom factors, instructional format,
and curricular factors) (see Cemen, 1987; Zaslavsky, 1994). In this meta-analysis
I have examined some personal characteristics and found their mediation to be
limited. This finding signals the need for inclusion of characteristics from other cate-
gories for further research. Overall, studies following this line of logic are likely
to provide theoretical appreciation of the complexity of the relationship as well as
practical implications for front-line educators.
REFERENCES
References marked with an asterisk (*) indicate studies included in the meta-analysis.
Aiken, L. R., Jr. (1960). Mathemaphobia and mathemaphilia: An analysis of personal and social
factors affecting performance in mathematics (Doctoral dissertation, The University of North
Carolina, 1960). Dissertation Abstracts International, 21, 1992A.
Aiken, L. R., Jr. (1970). Attitudes toward mathematics. Review of Educational Research, 40, 551–596.
Xin Ma 537
Aiken, L. R., Jr. (1976). Update on attitudes and other affective variables in learning mathematics. Review
of Educational Research, 46, 293–311.
Aiken, L. R., Jr., & Dreger, R. M. (1961). The effect of attitudes on performance in mathematics. Journal
of Educational Psychology, 52, 19–24.
Armstrong, J. (1985). A national assessment of participation and achievement in women in mathematics.
In S. F. Chipman, L. R. Brush, & D. M. Wilson (Eds.), Women and mathematics: Balancing the equa-
tion (pp. 59–94). Hillsdale, NJ: Erlbaum.
*Baya’a, N. F. (1990). Mathematics anxiety, mathematics achievement, gender, and socio-economic
status among Arab secondary students in Israel. International Journal of Mathematics Education in
Science and Technology, 21, 319–324.
Betz, N. E. (1978). Prevalence, distribution, and correlates of math anxiety in college students. Journal
of Counseling Psychology, 25, 441–448.
*Bieschke, K. J., & Lopez, F. G. (1991, August). A causal model of career aspirations. Paper presented
at the annual meeting of the American Psychological Association, San Francisco.
*Brassell, A., Petry, S., & Brooks, D. M. (1980). Ability grouping, mathematics achievement, and pupil
attitudes toward mathematics. Journal for Research in Mathematics Education, 11, 22–28.
Brush, L. R. (1978). A validation study of the Mathematics Anxiety Rating Scale (MARS). Educational
and Psychological Measurement, 38, 485–490.
Brush, L. R. (1980). Encouraging girls in mathematics: The problem and the solution. Cambridge, MA:
Abt.
Brush, L. R. (1981). Some thoughts for teachers on mathematics anxiety. Arithmetic Teacher, 29(4),
37–39.
Brush, L. R. (1985). Cognitive and affective determinants of course preference and plans. In S. F.
Chipman, L. R. Brush, & D. M. Wilson (Eds.), Women and mathematics: Balancing the equation
(pp. 123–150). Hillsdale, NJ: Erlbaum.
Burton, G. M. (1979). Getting comfortable with mathematics. The Elementary School Journal, 79,
129–135.
*Bush, W. S. (1991). Factors related to changes in elementary students’ mathematics anxiety. Focus
on Learning Problems in Mathematics, 13(2), 33–43.
Buxton, L. G. (1981). Do you panic about maths? Coping with maths anxiety. London: Heinemann.
Cemen, P. B. (1987). The nature of mathematics anxiety (Tech. Rep.). Stillwater: Oklahoma State
University. (ERIC Document Reproduction Service No. ED 287 729)
*Chiu, L. -H., & Henry, L. L. (1990). Development and validation of the Mathematics Anxiety Scale
for Children. Measurement and Evaluation in Counseling and Development, 23, 121–127.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences (Rev. ed.). New York: Academic
Press.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis in behavioral sciences.
Hillsdale, NJ: Erlbaum.
*Cooper, S. T., & Robinson, D. A. G. (1991). The relationship of mathematics self-efficacy beliefs to
mathematics anxiety and performance. Measurement and Evaluation in Counseling and Development,
24, 4–11.
Dew, K. M. H., Galassi, J. P., & Galassi, M. D. (1983). Mathematics anxiety: Some basic issues. Journal
of Counseling Psychology, 30, 443–446.
Dew, K. M. H., Galassi, J. P., & Galassi, M. D. (1984). Math anxiety: Relation with situational test
anxiety, performance, physiological arousal, and math avoidance behavior. Journal of Counseling
Psychology, 31, 580–583.
Donady, B., & Tobias, S. (1977). Math anxiety. Teacher, 95(3), 71–74.
*Donnelly, D. F. (1987). The effect of a high school computer studies course on achievement and anxiety
level in mathematics (Doctoral dissertation, Northern Arizona University, 1986). Dissertation
Abstracts International, 47, 2933A.
Dusek, J. B., & Joseph, G. (1983). The bases of teacher expectancies: A meta-analysis. Journal of
Educational Psychology, 75, 327–346.
Eccles, J. (1985). Sex differences in achievement patterns. In R. A. Dienstbier (Series Ed.) & T. B.
Sonderegger (Vol. Ed.), Nebraska symposium on motivation, 1984: Vol. 32. Psychology and gender
(pp. 97–132). Lincoln: University of Nebraska Press.
*Eccles, J. S., & Jacobs, J. E. (1986). Social forces shape math attitudes and performance. Signs: Journal
of Women in Culture and Society, 11, 367–380.
*Engelhard, G., Jr. (1990). Math anxiety, mother’s education, and the mathematics performance of
adolescent boys and girls: Evidence from the United States and Thailand. Journal of Psychology,
124, 289–298.
Fennema, E., & Sherman, J. (1977). Sex-related differences in mathematics achievement, spatial visu-
alization, and affective factors. American Educational Research Journal, 14, 51–71.
Frary, R. B., & Ling, J. L. (1983). A factor-analytic study of mathematics anxiety. Educational and
Psychological Measurement, 43, 985–993.
*Gliner, G. S. (1987). The relationship between mathematics anxiety and achievement variables.
School Science and Mathematics, 87, 81–87.
Gough, M. F. (1954). Mathemaphobia: Causes and treatments. Clearing House, 28, 290–294.
*Green, J. J. (1991). Mathematics anxiety and attitudes towards mathematics at the secondary level.
Unpublished doctoral dissertation, University of New England, Australia.
Hackett, G. (1985). Role of mathematics self-efficacy in the choice of math-related majors of college
women and men: A path analysis. Journal of Counseling Psychology, 32, 47–56.
Hadfield, O. D., & Maddux, C. D. (1988). Cognitive style and mathematics anxiety among high school
students. Psychology in the Schools, 25, 75–83.
*Hadfield, O. D., Martin, J. V., & Wooden, S. (1992). Mathematics anxiety and learning style of the
Navajo middle school student. School Science and Mathematics, 92, 171–176.
Handler, J. R. (1990). Math anxiety in adult learning. Adult Learning, 1(6), 20–23.
Hart, L. E. (1989). Describing the affective domain: Saying what we mean. In D. B. McLeod & V. M.
Adams (Eds.), Affect and mathematical problem solving: A new perspective (pp. 37–45). New York:
Springer-Verlag.
Hebb, D. O. (1955). Drives and the C.N.S. (conceptual nervous system). Psychological Review, 62,
243–254.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA: Academic
Press.
Hembree, R. (1990). The nature, effects, and relief of mathematics anxiety. Journal for Research in
Mathematics Education, 21, 33–46.
Hembree, R., & Dessart, D. J. (1986). Effects of hand-held calculators in precollege mathematics educa-
tion: A meta-analysis. Journal for Research in Mathematics Education, 17, 83–99.
Hendel, D. D. (1980). Experiential and affective correlations of math anxiety in adult women.
Psychology of Women Quarterly, 5, 219–230.
Hunsley, J. (1987). Cognitive processes in mathematics anxiety and test anxiety: The role of appraisals,
internal dialogue, and attributions. Journal of Educational Psychology, 79, 388–392.
Hunt, G. E. (1985). Math anxiety  Where do we go from here? Focus on Learning Problems in
Mathematics, 7(2), 29–40.
Hyde, J. S., Fennema, E., & Lamon, S. J. (1990). Gender differences in mathematics performance: A
meta-analysis. Psychological Bulletin, 107, 139–155.
Hyde, J. S., Fennema, E., Ryan, M., Frost, L. A., & Hopp, C. (1990). Gender comparisons of mathe-
matics attitudes and affect: A meta-analysis. Psychology of Women Quarterly, 14, 299–324.
Lazarus, M. (1974). Mathophobia: Some personal speculations. National Elementary Principal, 53(2),
16–22.
Leder, G. C. (1987). Attitudes towards mathematics. In T. A. Romberg & D. M. Stewart (Eds.), The
monitoring of school mathematics (pp. 261–277). Madison: Wisconsin Center for Educational
Research.
*Lee, B. Y. (1992). The effects of learner control and adaptive information in mathematics computer-
assisted instruction on achievement and task-related anxiety (Doctoral dissertation, Florida Institute
of Technology, 1991). Dissertation Abstracts International, 52, 3236A.
Xin Ma 539
Liebert, R. M., & Morris, L. W. (1967). Cognitive and emotional components of test anxiety: A
distinction and some initial data. Psychological Reports, 20, 975–978.
Ma, X., & Kishor, N. (1997). Assessing the relationship between attitude toward mathematics and
achievement in mathematics: A meta-analysis. Journal for Research in Mathematics Education, 28,
26–47.
Mandler, G., & Sarason, S. B. (1952). A study of anxiety and learning. Journal of Abnormal and Social
McLeod, D. B. (1992). Research on affect in mathematics education: A reconceptualization. In D. A.
Grouws (Ed.), Handbook of research on mathematics teaching and learning (pp. 575–596). New
York: Macmillan.
McLeod, D. B. (1994). Research on affect in mathematics learning in the JRME: 1970 to the present.
Journal for Research in Mathematics Education, 25, 637–647.
McMillan, J. H. (1976). Factors affecting the development of pupil attitudes toward school subjects.
Psychology in the Schools, 13, 322–325.
Meece, J. L. (1981). Individual differences in the affective reactions of middle and high school students
to mathematics: A social cognitive perspective (Doctoral dissertation, University of Michigan,
1981). Dissertation Abstracts International, 42, 2035A.
*Meece, J. L., Wigfield, A., & Eccles, J. S. (1990). Predictors of math anxiety and its influence on young
adolescents’ course enrollment intentions and performance in mathematics. Journal of Educational
Mevarech, Z. R., & Ben-Artzi, S. (1987). Effects of CAI with fixed and adaptive feedback on children’s
mathematical anxiety and achievement. Journal of Experimental Education, 56, 42–46.
*Mevarech, Z. R., Silber, O., & Fine, D. (1991). Learning with computers in small groups: Cognitive
and affective outcomes. Journal of Educational Computing Research, 7, 233–243.
National Research Council. (1989). Everybody counts: A report to the nation on the future of mathe-
matics education. Washington, DC: National Academy Press.
Pedersen, K., Bleyer, D. R., & Elmore, P. B. (1985). Attitudes and career interests of junior high school
mathematics students: Implications for the classroom. Arithmetic Teacher, 32(7), 45–49.
Preston, P. A. (1987). Math anxiety: Relationship with sex, college major, mathematics background,
mathematics achievement, mathematics performance, mathematics avoidance, self-rating of math-
ematics ability, and self-rating of mathematics anxiety as measured by the Revised Mathematics
Anxiety Rating Scale (RMARS) (Doctoral dissertation, University of Tennessee, 1986). Dissertation
Abstracts International, 47, 2494A.
Quilter, D., & Harper, E. (1988). “Why we didn’t like mathematics, and why we can’t do it.” Educational
Research, 30, 121–134.
*Reavis, P. S. (1989). Mathematics anxiety and the relationship between attitude, sex, ethnicity and
achievement in mathematics in three high school curriculum tracks (Doctoral dissertation, University
of Arizona, 1987). Dissertation Abstracts International, 49, 1742A.
Resnick, H., Viehe, J., & Segal, S. (1982). Is math anxiety a local phenomenon? A study of prevalence
and dimensionality. Journal of Counseling Psychology, 29, 39–47.
Reyes, L. H. (1980). Attitudes and mathematics. In M. M. Lindquist (Ed.), Selected issues in mathe-
matics education (pp. 161–184). Berkeley, CA: McCutchan.
Reyes, L. H. (1984). Affective variables and mathematics education. The Elementary School Journal,
84, 558–581.
Richardson, F. C., & Suinn, R. M. (1972). The Mathematics Anxiety Rating Scale: Psychometric data.
Journal of Counseling Psychology, 19, 551–554.
Richardson, F. C., & Woolfolk, R. L. (1980). Mathematics anxiety. In I. G. Sarason (Ed.), Test anxiety:
Theory, research, and applications (pp. 271–288). Hillsdale, NJ: Erlbaum.
Rosenthal, R. (1991). Meta-analytic procedures for social research (Rev. ed.). Newbury Park, CA: Sage.
Rounds, J. B., Jr., & Hendel, D. D. (1980). Measurement and dimensionality of mathematics anxiety.
Journal of Counseling Psychology, 27, 138–149.
*Saigh, P. A., & Khouri, A. (1983). The concurrent validity of the Mathematics Anxiety Rating Scale
for Adolescents (MARS-A) in relation to the academic achievement of Lebanese students.
Educational and Psychological Measurement, 43, 633–637.
*Sandman, R. S. (1979, April). Factors related to mathematics anxiety in the secondary school. Paper
presented at the annual meeting of the American Educational Research Association, San Francisco.
Schram, C. M. (1996). A meta-analysis of gender differences in applied statistics achievement. Journal
of Educational and Behavioral Statistics, 21, 55–70.
*Sepie, A. C., & Keeling, B. (1978). The relationship between types of anxiety and under-achievement
in mathematics. Journal of Educational Research, 72, 15–19.
Siegel, R. G., Galassi, J. P., & Ware, W. B. (1985). A comparison of two models for predicting math-
ematics performance: Social learning versus math aptitude-anxiety. Journal of Counseling
Spielberger, C. D. (1972). Conceptual and methodological issues in anxiety research. In C. D.
Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 2, pp. 481–493). New York:
Academic Press.
*Suinn, R. M., & Edwards, R. W. (1982). The measurement of mathematics anxiety: The Mathematics
Anxiety Rating Scale for Adolescent (MARS-A). Journal of Clinical Psychology, 38, 576–580.
*Suinn, R. M., Taylor, S., & Edwards, R. W. (1988). Suinn Mathematics Anxiety Rating Scale for
Elementary School Students (MARS-E): Psychometric and normative data. Educational and
Psychological Measurement, 48, 979–986.
*Suinn, R. M., Taylor, S., & Edwards, R. W. (1989). The Suinn Mathematics Anxiety Rating Scale
(MARS-E) for Hispanic elementary school students. Hispanic Journal of Behavioral Sciences, 11,
83–90.
*Thorndike-Christ, T. (1991). Attitudes toward mathematics: Relationships to mathematics achieve-
ment, gender, mathematics course-taking plans, and career interests (Tech. Rep.). Bellingham:
Western Washington University. (ERIC Document Reproduction Service No. ED 347 066)
Tobias, S. (1978). Overcoming math anxiety. Boston: Houghton Mifflin.
Tobias, S. (1985). Test anxiety: Interference, defective skills, and cognitive capacity. Educational
Psychologist, 20, 135–142.
Tobias, S., & Weissbrod, C. (1980). Anxiety and mathematics: An update. Harvard Educational
Review, 50, 63–70.
*Wahl, M. W. (1986). Mathematics anxiety in high school students: A study of gender and interrelated
factors (Doctoral dissertation, University of Wisconsin at Milwaukee, 1985). Dissertation Abstracts
International, 46, 2639A.
*Wigfield, A., & Meece, J. L. (1988). Math anxiety in elementary and secondary school students. Journal
of Educational Psychology, 80, 210–216.
Willett, J. B., & Singer, J. D. (1991). From whether to when: New methods for studying student dropout
and teacher attrition. Review of Educational Research, 61, 407–450.
Wine, J. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92–104.
*Wither, D. (1988). Influences on achievement in secondary school mathematics. Unicorn, 14, 48–51.
Wood, E. F. (1988). Math anxiety and elementary teachers: What does research tell us? For the
Learning of Mathematics, 8(1), 8–13.
Zaslavsky, C. (1994). Fear of math: How to get over it and get on with your life. New Brunswick, NJ:
Rutgers University Press.
Author
Xin Ma, Assistant Professor, Centre for Research in Applied Measurement and Evaluation,
University of Alberta, Edmonton, AB T6G 2G5, Canada; xin.ma@ualberta.ca
1999, Vol. 30, No. 5, 541–557
Cross-National Comparison
of Representational Competence
Mary E. Brenner, Sally Herman, Hsiu-Zu Ho, and Jules M. Zimmer
University of California, Santa Barbara
Flexible use of multiple representations has been described as a key component of competent
mathematical thinking and problem solving. In this study, 6th-grade American students are
compared to 3 samples of Asian (Chinese, Japanese, and Taiwanese) 6th graders to determine
if the well-documented mathematical achievement of students from these Asian nations may
be due in part to a greater understanding of mathematical representations. The results show that,
among all groups, Chinese students generally scored highest on the representation tasks and,
except on items about the visual representations of fractions, all Asian samples scored signifi-
cantly higher than the American sample. The results are discussed in terms of possible instruc-
tional antecedents and textbook differences.
Key Words: Cross-cultural studies; Middle grades, 5–8; Problem solving; Rational number/repre-
sentations; Representations, modeling
The well-documented national differences in mathematics achievement (Beaton

et al., 1996; McKnight et al., 1987; Mullis et al., 1997; Peak, 1996; Robitaille &
Garden, 1989; Stevenson & Lee, 1990; Stevenson et al., 1985) and instruction
(Stigler & Hiebert, 1997; Stigler, Lee, Lucker, & Stevenson, 1982; Stigler &
Stevenson, 1991) provide cognitive researchers with rich opportunities to explore
the competencies that underlie mathematical performance. Although some who
report on mathematical achievement emphasize how many “facts” students from
different countries know, in most recent research mathematical competence has
been defined more broadly. For instance, Schoenfeld (1992) identified the knowl-
edge base, problem-solving strategies, monitoring and control, beliefs and affect,
and practices as aspects of mathematical cognition. Other researchers have proposed
mathematical problem-solving models that show how different skills need to be
combined in a process of problem solving to produce competent performance
(Mayer, 1987).
We thank our collaborators in the participating countries: Sou-Yung Chiu, National Chan-
Hwa University of Education, Republic of China; Yasuo Nakazawa, Seisen Women’s Junior
College, Japan; and Chang-Pei Wang, Beijing Institute of Education, People’s Republic of
China. We acknowledge the contributions of the principals, teachers, and students of the
schools that participated in the study. Appreciation is extended to the school officials and
ministries of education in the respective nations for their support of the project. The data in
this article were first analyzed as part of the second author’s doctoral dissertation (Herman,
1994/1995). The first author was chair of that doctoral dissertation committee. However, the
presentation of the data here is from a somewhat different theoretical perspective. This
research was funded by a grant from the Pacific Rim Project of the University of California
to Drs. Hsiu-Zu Ho and Jules Zimmer.
542 Cross-National Representational Competence
In this study we examined one component of problem-solving competency for

samples of students from four nations that differ substantially in mathematics
achievement and mathematics instruction. In particular, we focused on the problem-
representation skills of students from the United States, China, Taiwan, and Japan.
In prior studies, students from the three Asian nations have demonstrated superior
mathematical achievement when compared to American students. Most recently
the Third International Mathematics and Science Study, TIMSS, has confirmed the
high level of mathematics achievement in Japan and other Asian countries (Beaton
et al., 1996; Mullis et al., 1997). However, there are significant differences in the
ways mathematics is taught in these countries (Cao & Cai, 1989; Stigler & Hiebert,
1997; Stigler & Stevenson, 1991), and students’ mathematical competencies might
differ as a consequence.
BACKGROUND
When trying to make sense of national achievement differences in mathematics

in order to improve educational practices, researchers must determine whether
higher achievement is related to students’ having been taught more mathematics
or to their having received better instruction. Evidence indicates that students
from high-achieving nations have been taught more mathematics than American
students of a comparable age. Many countries have longer school years and
students may receive more hours of mathematics instruction (Stevenson & Lee,
1990), although the TIMSS study showed that American students and students in
most other countries receive comparable amounts of instruction (Beaton et al., 1996;
Mullis et al., 1997). In addition, students in high-achieving Asian nations cover
advanced mathematics topics earlier than American students, as revealed in
comparisons of mathematics curricula and textbooks (Fuson, Stigler, & Bartsch,
1988; Mayer, Sims, & Tajika, 1995; Peak, 1996; Stevenson, 1985; Stevenson &
Bartsch, 1992; Westbury, 1992). However, even when curricular coverage appears
to be equal (Baker, 1993), American students score lower than many of their high-
achieving Asian counterparts. Baker’s analysis suggests that at least some Asian
students are taught in a way that enhances their mathematical competency and
fosters development of their repertoire of facts and skills. Comparisons of teaching
methods (Stigler & Hiebert, 1997; Stigler & Stevenson, 1991) and textbook mate-
rials (Fuson et al., 1988; Mayer et al., 1995; Stevenson, 1985; Stevenson & Bartsch,
1992; Westbury, 1992) support this proposition.
Cognitively oriented researchers have contributed to this debate by looking at
national differences in light of models of mathematical competence rather than in
terms of lists of classroom mathematics topics. In this article we work from a
problem-solving model in which good problem representation is posited as a key
characteristic of mathematical understanding (Mayer, 1987). According to Mayer
and Hegarty (1996), the competent problem solver engages in a comprehension
process that has three phases—translation, integration, and planning. During the
translation phase, the problem solver interprets the problem statements in internal
M. E. Brenner, S. Herman, H. -Z. Ho, and J. M. Zimmer 543
propositional statements. In the integration phase, the problem solver creates a

problem model from these propositional statements, drawing from schematic
knowledge about problem types. A rich problem model is a representation that inte-
grates both the numerical information presented in the problem and the relations
among the objects that are relevant to the problem. An example of a relation
between two numbers is the relation greater than to indicate a quantitative differ-
ence between the numbers. Or the two numbers might have an equivalence rela-
tion by having different symbolic forms that are mathematically equivalent. In the
planning phase, the problem solver uses the problem model to determine the rele-
vant strategies for solving the problem. Having a rich problem model enables the
problem solver to choose an appropriate computational representation.
Although the theoretical model describes the internal creation of problem models,
supporting students with instruction on different kinds of external representation
enhances problem-solving skill (Brenner et al., 1997; Lewis, 1989; Willis & Fuson,
1988; Yerushalmy, 1997). Presumably, knowledge of external representations is
part of the knowledge base important during the integration phase. Lewis found
that teaching students to create a problem model based on a number line led to
increased skill in solving word problems. Willis and Fuson taught students to use
a schematic drawing for addition and subtraction word problems and found that
use of such drawings enhanced students’ performance. Brenner et al. compared
experimental classes of students who were taught multiple representations in the
context of a prealgebra class to students receiving instruction primarily on solu-
tion strategies. When compared to the solution-oriented classes, experimental
classes exhibited more skill with diverse problem representations and enhanced
problem-solving success. Yerushalmy analyzed how creating multiple represen-
tations also enhanced students’ understanding of functions.
There are only scanty and equivocal data about whether the higher Asian math-
ematical achievement may be due to better problem-solving skills, and to repre-
sentation in particular. In a study by Silver, Leung, and Cai (1995), Japanese and
American students were asked to provide multiple explanations about how to
solve a nonroutine problem about the number of marbles in an array. Although they
had many similarities in their explanations, the Japanese students were more likely
to use what Silver et al. labeled verbal/symbolic representations whereas the
American students were more likely to use visual ones, such as marks on a diagram.
The Japanese students were also more successful than American students in
providing totally correct explanations (96% vs. 66%). A qualitative examination
of the data revealed that the Japanese used what the authors called “more sophis-
ticated mathematical ideas” (p. 35) as shown by a greater use of mathematical
expressions and multiplication instead of addition.
In three cross-national studies, the problem-solving framework described by
Mayer and Hegarty (1996) was used to compare students’ skills at the three
problem-solving stages of translation, integration, and planning. Studying groups
of students matched for computational skill, Mayer, Tajika, and Stanley (1991)
found that American fifth graders did significantly better on integration tasks than
Japanese fifth graders, and there was a trend in this direction for translation and
planning as well. Tajika, Mayer, Stanley, and Sims (1997) replicated these results
with other samples of Japanese and American children. Studies of similarly
matched samples of Chinese and American sixth graders showed similar results
with significant differences favoring the American samples in all three stages
(Cai, 1995). Although on the basis of these results American students seemed to
have better representation skills, the authors themselves and other commentators
(Stigler & Miller, 1993) have noted that the matching procedures created
mismatches along other dimensions and masked the fact that, on average, American
students had much lower computational skills and as a group were no better at
solving complex problems in some of these studies.
Cai (1995) has provided other data comparing the representations generated by
American and Chinese students during problem solving. When the American
students used symbolic representations, they were based on arithmetic. The Chinese
students were more likely to use advanced representations, particularly algebraic
notation. In contrast, American students were more likely to use visual and picto-
rial representations; for most problems, the Chinese students used no visual or picto-
rial representations at all.
In sum, these few studies are consistent in showing that American students as
well as Asian students can competently use representations to solve problems.
However, both Japanese and Chinese students chose to use what other authors have
called more advanced symbolic forms of representation whereas Americans seemed
to prefer visual ones. None of the studies directly addressed whether students of
any nation were flexible in moving across representations. Nor did they assess
whether students of different nations were competent when compared on the same
kinds of representations as opposed to self-generated representations. This study
directly addresses both of these issues. We did not seek to examine whether
students’ internal representations differed across nations; rather we examined
whether or not students exhibited different skill levels with standard mathematical
representations.
FLEXIBILITY WITH REPRESENTATIONS
Dreyfus and Eisenberg (1996) argued that flexibility in moving across repre-
sentations is a hallmark of competent mathematical thinking. Each representation
highlights specific aspects of a concept and can help students in problem solving.
This flexibility can entail moving within a representational mode, such as using a
diagram with different arrangements of components to represent different propor-
tion problems (e.g., Thompson, 1990). It also entails moving between quite different
representations, such as between an equation and a graph. Each representation
reveals a different structural aspect of the problem. Solving difficult problems may
require using several representations. Lesh, Post, and Behr (1987) made a similar
argument to define mathematical understanding.
Part of what we mean when we say that a student “understands” an idea like “1/3” is
that (1) he or she can understand the idea embedded in a variety of qualitatively
different representational systems, (2) he or she can flexibly manipulate the idea
within given representational systems [i.e., perform transformations], and (3) he or she
can accurately translate the idea from one system to another. (p. 36)
They further stated, “Good problem solvers tend to be sufficiently flexible in their
use of a variety of relevant representational systems that they instinctively switch
to the most convenient representation to emphasize at any given point in the solu-
tion process” (p. 38).
In this article we focus particular attention upon visual representations because
in prior cross-cultural studies American students have been found to use them (Cai,
1995; Silver et al., 1995). Visual representations are becoming more prominent in
the work of professional mathematicians (Dreyfus & Eisenberg, 1996), and such
representations have been shown to help poor problem solvers in problem solving
(Bondesan & Ferrari, 1991). The Curriculum and Evaluation Standards of the
National Council of Teachers of Mathematics (1989) in the United States also
emphasized the importance of visual representations in use of both manipulatives
and written forms such as graphs and diagrams. However, students who use visual
representations may also deemphasize analytical reasoning (Presmeg, 1986) or
substitute visual images for algebraic or other formal representations that may be
more appropriate for solving the problem. Dreyfus and Eisenberg suggested that
overreliance on visual representations in place of other representations may
diminish mathematical competency.
For this study, three aspects of representation use are examined. The first two
involve transformations within the written-symbol representation system (Lesh et
al., 1987). We make a distinction between Flexibility Within a Representation and
Flexibility Across Formal Symbolisms (but still within a single representation
system). We tested the former flexibility by asking students to express fractions
as the combination of other fractions, as one example. This type of skill has been
linked to mathematical competence in areas such as mental computation; people
skilled at mental computation have been found to use many regrouping strategies
(e.g., converting the problem 65 + 47 into 60 + 40 + 5 + 7, Sowder, 1992). The
second type of flexibility, ability to translate among different kinds of written repre-
sentations, is believed to contribute to greater conceptual knowledge and enhanced
problem solving. To test this ability, we asked students to, for example, express a
fraction as a decimal, a division expression, and a proportional relationship.
The third aspect we examine is skill in translating between visual representations
and written symbols, different representational systems (Lesh et al., 1987). In earlier
studies (Cai, 1995; Silver et al., 1995) the use of visual representations by American
students and those from Asian nations differed significantly. By looking at trans-
lations between visual representations and written symbols, we sought to determine
if Asian mathematical competence also entailed competence in this area across the
three Asian samples. In addition, we examined whether the American sample
displayed any particular competence with visual representations and whether they
showed corresponding competence with other representations related to the same

concepts.
METHODS
Participants
The participants in this study included 895 sixth-grade students from the four
nations: 223 from the People’s Republic of China, 224 from Taiwan, 177 from
Japan, and 271 from the United States. The data included in this study were
collected as part of a cross-national longitudinal project in which we investigated
factors that influence mathematics achievement. Data were collected from students
during fourth and sixth grades, although only data collected in sixth grade are used
in this report. Measures of attitudes and beliefs of parents and teachers were
collected as well (see Tuss, Zimmer, & Ho, 1995, for more information). We chose
to study students from the nations of China, Japan, Taiwan, and the United States
in our effort to replicate and extend the findings of Stevenson and his colleagues
(Stevenson & Lee, 1990; Stevenson et al., 1985; Stigler et al., 1982). The People’s
Republic of China was included because of the similarity in its traditions and
customs to those of the Taiwanese culture, yet it was different in terms of its polit-
ical, educational, and economic systems.
In an attempt to obtain a somewhat representative sample from each nation (e.g.,
approximately 80% of China’s population reside in rural areas), we selected
schools from both rural and urban settings. The schools were selected on the basis
of educational, economic, institutional, and residential characteristics and recom-
mendations of the educational authorities and researchers in each nation. In each
nation, local authorities identified one high-achieving urban school and one low-
achieving urban school with similar economic and residential characteristics. The
urban sites were located in Beijing, China; Taipei, Taiwan; and Claremont,
California. The rural sites were located in the Men Tou Gou district, China; Miao-
Li and Yang Ming Shan, Taiwan; and Cuyama and Santa Ynez, California. The
Japanese classes were all located in the Nagano Prefecture, where the two main
economic activities were agriculture and the manufacturing of electronic products.
Rural and urban distinctions were not made for these Japanese areas. In each
participating school, all sixth-grade students who were not absent at the time of data
collection (more than 95% in each selected classroom) participated in the study.
The gender and age characteristics of each sample are given in Table 1.
Table 1
Gender and Age Characteristics by Nation
Nation n Male Female Mean age (years)
China 223 124 99 12.8
Japan 177 83 94 12.0
Taiwan 224 115 109 12.1
United States 271 148 123 12.0
Instrument
A test of mathematical achievement was developed for this study. Two parts
were administered to students in separate sessions, spaced about 2 months apart.
The test included solution items—either straightforward computations (A2, B3,
B5) or complex problems (A1, B4)—and a representation item corresponding to
each solution item (all items are shown in Figure 1). In addition, there was a repre-
sentation item for 4 × 9, but that solution item was judged as too easy for sixth
graders. We developed two solution items, and the others were selected from text-
books of the participating nations. For each solution item, we wrote a represen-
tation item comprising five to seven subitems that included both numerical/symbolic
representations and pictorial representations. Participants were asked to judge
whether each representational subitem was a correct or incorrect representation for
the corresponding solution item. Students were asked to complete a solution item
at one session and to judge the corresponding representational items at the other
session.
Procedure
Students were administered two forms of a mathematics achievement test that
included the items shown in Figure 1 and others not analyzed in this article. In this
article we analyze only those items that included both solution items and corre-
sponding representation items. At the first testing session, the students were given
two solution items and four of the representational items on Form A of the test. At
the second session they were given the other three solution items and the remaining
representational items on Form B of the test. Classmates completed the tests indi-
vidually at their desks during the same class period. No time constraints were
imposed, and students did not have difficulty finishing the items within the class
period allotted for the exam.
Analysis
The solution items were scored as either right or wrong. Each representational
subitem was scored 1 if the item was answered correctly and -1 if it was answered
incorrectly; thus a group score of 0 indicates that half of the students chose the
correct answer. In addition, a total score was obtained for each representational set
of items by summing the subitem scores for that question. Scores for two subitems
(B2f, A6d) from Figure 1 were excluded from their respective total scores because
when analyzing the results, we did not reach consensus on whether these items were
good alternative representations of the original items. In addition, we decided that
the reference to a dozen in Subitem B2f was not equally culturally relevant to each
sample. For other items (e.g., A1) measurement units had been converted to the
appropriate measurement system for each country.
Mean scores for each solution and representational item were calculated for each
national group. An ANOVA using nation as the grouping factor was run on all indi-
(Asterisks indicate significant sample differences at .05 level of significance.)

Solution Items Corresponding Representational Items
(Student solves the problems) (Student marks as right or wrong)
*A1 A pole 2 yards high casts a shadow B1 (The same problem is stated.) “Now
3 yards long. The shadow of a tree is decide if each of the following statements is
9 yards long. How tall is the tree? right or wrong. For each statement check
the appropriate box.” (There are two boxes
labeled Right and Wrong for each item.)
*a. 23 = 9h
2 yds *b. 2 : 3 = h : 9
? yds *c. 3 + 2 = 9 + h
3 yds *d. 29 = 3h
9 yds
*e. 2 × 9 = h × 3
*A2 Give the answer to the following B2 The fraction 36 may be represented in
problem in decimals. several different ways. Decide if each of the
following examples is a right or wrong
3 ÷ 6 = _______
representation of 36.
*a. 3 ÷ 6
*b. .50
*c.
:
*d.
e.
*f. Paul has half a dozen doughnuts. He
wants to share them equally with his friend
Joani.
*g.
*B3 1 34 + 2 12 = A3 Given the problem below, decide if

each of these statements is right or wrong.
1 34 + 2 12 =
*a. (1 34 ) + (2 12 )
4
*b. 4 + 34 + 42 + 12
*c. (1 × 34 ) + (2 12 )
*d. 1 + 2 + 34 + 12
*e. 1 + 0.75 + 2 + 0.50
*B4 A4 (The same figure is given.)

In the rectangular diagram of a garden two *a. The total area of the figure is 20 × 32
white paths were made. What is the total *b. The area of the white paths is
area of the shaded sections below? (2 × 20) + (2 × 32)
Answer ____________ c. The area of the shaded sections is
2 yds (20 × 32) – (2 × 20) – (2 × 32)
*d. The area of the shaded sections is
2 yds (20 × 32) – (2 × 20) – (2 × 32) + (2 × 2)
20 yds *e. The area of the shaded sections is
18 × 30
32 yds
Figure 1. Test items and results of analysis of variance for national differences.
2
*B5 3 ÷ 21 A5 Read the problem below. Paul and his
friend Mathew love to eat pizza. They are
given two-thirds of a whole pizza to share
equally. What portion of the whole pizza
would each person get to eat?
Now decide if each of the following state-
ments is right or wrong.
*a. 23 + 12
*b. 2 ÷ 23
2
*c. 3 ÷2
*d. 2
3 × 12
2
*e. 3 ÷ 12
No solution item. A6 There are different ways of representing

4 × 9. Decide if each of the statements below
is a right or wrong representation of 4 × 9.
*a. 0 10 20 30 40
*b. 2(3) + 2(3) + 2(3) + 2(3)

9 ft
4 ft
*c.
*d. 2(9) + 2(9)
*e. 9 + 9 + 9 + 9
Figure 1 (cont.). Test items and results of analysis of variance for national differences.
vidual items (both solution items and representational subitems), and post-hoc
contrasts were conducted for each significant ANOVA result. Bonferroni family-
wise alpha for each set of comparisons was tested at p < .05.
RESULTS
Scores and Rank Orders on Solution Items

For the total sample, scores ranged from a low of 36% on finding the area of the
garden divided by two paths (B4) to 80% on the calculation of “2/3 divided by 2/1”
(B5). Among the four national samples, the Chinese students ranked first on three
of the five solution items, and the Taiwanese students ranked second on four of the
five. The Japanese sample ranked third on three of the five. The U.S. sample ranked
last on all these items. Most Chinese students were able to accurately solve all but
one solution item, and the Taiwanese students showed a similar rate of accuracy.
In contrast, a slim majority of the U.S. sample answered two questions correctly,
with small percentages of the students answering the other three correctly. Table 2
shows the results for each solution item by nation.
Table 2
Percentage Correct (and Rank Order) on Solution Items
Item China Japan Taiwan U.S. Total sample Description
A1 79 (1) 47 (3) 70 (2) 18 (4) 52 Proportion
A2 94 (1) 84 (3) 90 (2) 26 (4) 70 3÷6=
B3 93 (1) 66 (3) 79 (2) 53 (4) 72 1 3/4 + 2 1/2
B4 43 (3) 45 (2) 52 (1) 10 (4) 36 Area of garden
B5 87 (3) 92 (1) 90 (2) 57 (4) 80 2/3 ÷ 2/1
Mean 79 (1) 67 (3) 76 (2) 33 (4) 62
Scores and Rank Orders on the Representational Items

Overall, students had more difficulty with the representational items than the solu-
tion items. However the pattern among samples was similar to the pattern for the
solution items: The Taiwanese and Chinese samples ranked first or second on most
items, with the Japanese sample showing the third strongest performance overall.
Students in the U.S. sample had strikingly low performances on the representational
subitems, scoring with less than 50% accuracy on more than 1/3 of the questions
(subitem average scores between 0 and -1) and ranking last among the nations on
about 70% of the subitems. In comparison, a majority of Chinese and Taiwanese
students correctly answered the representational subitems (scores between 0 and
+1) on all but one set of items. Table 3 shows the pattern of results, by nation, for
each set of representational items.
Table 3
Scores (and Rank Order) on Representational Items
Item China Japan Taiwan U.S. Total sample Description
B1 total .61 (1) .20 (2) .50 (2) .09 (4) .34 Proportion
B2 totala .47 (2) .13 (4) .53 (1) .26 (3) .36 3/6
A3 total .39 (2) .26 (3) .48 (1) -.12 (4) .23 1 /4 + 2 1/2
3
A4 total .11 (3) .25 (1) .21 (2) -.03 (4) .12 Area of Garden
A5 total .67 (1) .35 (3) .57 (2) .03 (4) .39 Sharing Pizza
A6 totalb .39 (2) .67 (3) .65 (1) .39 (4) .51 4×9
Mean .44 (2) .31 (3) .49 (1) .10 (4) .33
aThe total score for B2 does not include Item B2f. bThe total score for A6 does not include Item A6d.
ANOVA on Individual Items

Each of the five solution items had significant sample differences. Twenty-eight
of the 30 representational subitems also had significant sample differences.
Flexibility Within a Representation

There were nine subitems that required the students to judge the equivalence of
different fraction representations: A3a–d and A5a–e. On these subitems, the
Chinese sample consistently scored highest (with scores of more than .5 with two
exceptions). Interestingly, the Japanese sample had mixed performance on these
subitems: For five subitems the Japanese students’ performance statistically equaled
that of the Chinese and Taiwanese students whereas for four subitems they
performed significantly worse than the Chinese and Taiwanese students and
approximately equal to the American students. Fewer than half of the American
students correctly answered most of these subitems. On only one subitem was their
average above .2 (A5a, .23).
Flexibility Across Formal Symbolisms

The subitems related to this type of flexibility were mixed and included (a) trans-
formations of fraction to decimal, fraction to proportion, and multiplication to addi-
tion and (b) translations of multiplication to a number line and simple number
sentences to algebraic number sentences. The 12 subitems analyzed for this kind
of flexibility were A4c, A4d, A6a, A6b, A6d, B2b, B2d, B1a, B1b, B1c, B1d, and
B1e. These were the most difficult representational tasks, as shown by the gener-
ally lower scores, and performance within (to us) similar subitems was uneven for
all samples. The Chinese and Taiwanese samples showed the highest scores on these
difficult subitems, although even their performances were often little better than
would be achieved by guessing. They tied for highest scores on 9 subitems. The
Japanese sample scored the highest on 1 subitem, and there were no significant
sample differences on 2 other subitems.
In a number of cases, the students, despite showing computational skill with items
in these forms, seemed to reject the very idea that different forms of formal nota-
tion could be equivalent. For instance, except for the U.S. sample, all national
samples did very well calculating a decimal answer for 3 ÷ 6 (87% to 97% accu-
racy on A2). However, they rejected the idea that .50 was an appropriate repre-
sentation for 3/6 (scores ranged from -.57 to .17 on Item B2b) although they saw
the equivalence between the fraction and the division computation (scores ranged
from .54 to .85 on Item B2a).
The use of algebraic notation in many of these items may have contributed to
the students’ difficulties. For this analysis, a subitem was judged to use algebraic
notation when it involved a variable or used parentheses for grouping different oper-
ations. There were 10 subitems of this sort (B1a, B1b, B1c, B1d, B1e, A6b, A6d,
A4b, A4c, A4d). The 3 most difficult representational subitems in the study (B1e,
A6d, A4d), each of which had less than 50% accuracy for all four national samples,
used algebraic notation. Four of the algebraic subitems on which students showed
strong performance (B1c, B1d, A6b, and A4b) were not appropriate representa-
tions for the problems, and the students marked them as wrong. This may indicate
a rejection of the algebraic notation as much as a judgment about the appropriate-
ness of the representation. The pattern of difficulty holds for all four national
samples. The only 2 subitems that had no significant national sample differences
(A6b and A4c) used algebraic notation.
Visual Representations
Two of the solution items (A1 and B4) included diagrams that could facilitate
solving the problems. Despite American students’ preference for visual represen-
tations in earlier studies (e.g., Cai, 1995), in this study their scores on these items
were significantly lower than scores of the children from the other three nations.
In fact, American students’ scores on these two items were their lowest among all
solution items; they scored only 18% and 10% correct on A1 and B4, respectively.
Of course, because they are fairly complex problems, these items required more
than good visual representation skills for proper solution.
Five of the representational subitems (A6c, B2c, B2d, B2e, and B2g) were
visual, that is, diagrammatic representations of the solution items. On B2c and B2g,
each of which was a rectangle with regions shaded to represent a fraction, the U.S.
sample scored .70 and .69, respectively, significantly higher than the samples of
the other nations. However, on two other visual subitems, the U.S. sample had the
lowest accuracy score among the four samples. In contrast to their performance on
other subitems, on two of the visual subitems the Chinese students had the lowest
score or tied for the lowest score with either the Japanese or the U.S. students (with
total scores of .40 and .28 on Subitems A6c and B2c, respectively). The Chinese
sample also scored significantly lower than all other national samples on two of
the representations for the visual-solution item about finding the area of a garden
cut by two paths (on subitems A4c and A4e, on which they scored -.28 and .14,
respectively), and this sample was not the highest scorer for any of the represen-
tational subitems for this problem.
DISCUSSION AND SUMMARY
As in many earlier achievement studies, the Asian students in this study outscored
the American students on most test items. The gap in performance between the
American students and the three Asian groups was greater for the representational
items than for the solution items. On the solution items, the samples from Japan,
Taiwan, and China were twice as likely as samples from the United States to
correctly answer an item. On the representational items, Japanese students were
almost three times as likely as American students to get an item correct, and
Chinese and Taiwanese students nearly five times as likely as American students
to answer correctly. These results indicate that, in addition to having stronger basic
skills, the Asian students in this study also had conceptual skills that might give
them an advantage in solving more difficult problems. Because of the nonrandom
and nonrepresentative sampling design of this study, other studies are needed to
explore whether these differences are true in general for each nation. In addition,
other curricular areas should to be explored to determine if the results hold for topics
other than rational numbers, proportional reasoning, and multiplication.
In terms of the relative strengths of the samples from each nation, these results
about students’ representational skills tend to be in agreement with results of other
studies in which different methodologies were used. Whereas other researchers have
examined students’ preferences for representational strategies when they had a
choice about which to use (e.g., Cai, 1995; Silver et al., 1995), in this study we
examined performance in judging the appropriateness of a range of representations
chosen by the researchers to include a variety of standard mathematical represen-
tations. As in other studies (Cai, 1995; Silver et al., 1995), the American students
showed some strength with visual representations. But inasmuch as the American
students were unsuccessful on the other fraction subitems, the results seem to
support Dreyfus and Eisenberg’s (1996) caution that visual representations have
limited utility if not supported by other kinds of representational competency.
As in previous studies, the Asian samples were more successful than American
students on items that required understanding of formal symbolism (Cai, 1995;
Silver et al., 1995). However, each sample seemed to have difficulty on some
subitems, particularly those requiring flexibility across different forms of repre-
sentation. Other researchers have also found this type of task to be difficult. In a
study by R. E. Reys, Reys, Nohda, and Emori (1995), a Japanese sample had diffi-
culty in linking decimals and fractions, and although the Japanese sample had the
lowest of any scores on this subitem, all samples had such difficulty. In some cases
national samples may have scored very low on a particular subitem because of unfa-
miliarity with the notation; for example, number lines (in Subitem A6a) may be
unfamiliar to Chinese students and the proportion notation may be unfamiliar to
U.S. students (in Subitems B1b and B2d). Students may have experienced diffi-
culty with translations because they felt that key features of the original represen-
tation or problem were lost in other representations. For instance, using the decimal
.50 obscures information about the number of pieces into which an item was
divided in the original problem. So although 3/6 and .50 are equivalent, they may
not be seen as equally good representations of the original problem information.
There are some discrepancies between results from this study and from other
research. The most anomalous result relates to the Japanese sample. Although
Japanese students were high performers in most other studies (Beaton et al., 1996;
McKnight et al., 1987; Mullis et al., 1997; Peak, 1996; Robitaille & Garden, 1989;
Stevenson & Lee, 1990; Stevenson et al., 1985), the Japanese sample in this study
usually scored midway between the American and Chinese samples and often
scored significantly lower than the Chinese and Taiwanese samples. This effect was
most marked for the representational items. Sampling differences from other
studies could explain this discrepancy inasmuch as none of these samples is a
random, representative sample. However, evidence from other studies shows some
similar patterns for Japanese samples. In one study a lack of flexibility was noted
in a Japanese sample. R. E. Reys et al. (1995) found that when doing mental compu-
tation tasks, Japanese students predominantly used a mental version of the written
algorithm instead of regrouping strategies that had been used by other samples. In
addition, the authors noted, “Most students were unable to describe any other
approach to the problems and seemed surprised that they would be asked for
another method” (p. 323). In the TIMSS study (Beaton et al., 1996; Mullis et al.,
1997), Japanese students were high performers overall at the fourth-grade and
eighth-grade levels, but, like U.S. students, they had mixed performance on specific
items similar to those in this study. For instance, both Japanese and U.S. samples
at fourth grade were above the international average of 61% correct (scoring 89%
and 80%, respectively) on an item in which students were to match a fraction and
a shaded figure (Beaton et al., 1996). Both samples scored less than the interna-
tional average of 53% correct (45% and 43%, respectively) on an item about
proportionality. On a third item, matching a decimal and a shaded figure, the
Japanese sample was well above the international average of 40% correct with a
score of 71% correct whereas the U.S. sample was slightly below average at 32%.
Because there are no similar studies of the four nations we compared, conclusions
about the Japanese sample’s performance relative to that of students in other coun-
tries need to be treated with caution. Although TIMSS did not include Taiwan or
China, several other Asian nations participated, including Singapore, Hong Kong,
and Korea. On the three items described above, the samples from these nations had
consistently high performance in contrast to Japan (Beaton et al., 1996). The combined
results of our study, the TIMSS study, and the studies by Mayer et al. (1991), Tajika
et al. (1997), and R. E. Reys et al. (1995) warrant further exploration of the nature
of the representational and problem-solving skills of American and Japanese samples,
when basic skills are either irrelevant or controlled for in the research design.
There was one striking area of weakness in the Chinese sample. Although the
Chinese sample scored higher than the other nations on most items, they were the
lowest scoring sample on half of the items relating to visual representations. The
Taiwanese did not exhibit weakness in this area, one of the few areas in which the
Chinese and Taiwanese samples showed a different pattern of results.
The differences in the performances among the four national samples on repre-
sentational items as described in this study may be due in part to the instructional
materials used in each country. Several studies have compared American and
Japanese textbooks for possible explanations of superior Japanese performance.
Mayer et al. (1995) found that seventh-grade textbooks in Japan emphasized coor-
dination of multiple representations, integrating verbal, pictorial, and symbolic
representations in the explanations, much more than did American textbooks.
There were many illustrations in the U.S. textbooks, but they were not always rele-
vant to the mathematical topic, and seldom were the linkages among representa-
tions made explicit. B. J. Reys, Reys, and Koyama (1996) found integration of illus-
trations and mathematical concepts in primary-grade textbooks in Japan but not in
the United States. Stevenson (1985) also reported that Japanese textbooks placed
greater emphasis on proportions and ratios than American texts and that conver-
sions from one mathematical form to another, including conversion of decimal
numbers to fractions, were more common than in U.S. texts. Carraher (1996)
analyzed conceptual difficulties that students have in understanding rational
numbers when they are taught with part-whole representations such as in the
shaded-squares and shaded-circle examples used in this study. According to
Carraher, the exclusive use of these representations increases students’ difficulties
in understanding the relation of fractions to proportions. Such part-whole repre-

sentations indeed appear to be the most common fraction representations in
American textbooks according to Post, Cramer, Behr, Lesh, and Harel (1993).
Carraher’s analysis fits with the pattern of results for the American sample in this
study. The American sample was very competent with part-whole representations
when compared to the Asian samples but was low in other areas of performance.
Although a number of researchers have compared instruction across nations, most
have studied the use of different teaching techniques or coverage of curricular topics
rather than the ways in which particular topics such as rational number are taught.
The existing evidence indicates that students in classrooms in the United States may
have less systematic exposure to multiple representations than Asian students
typically receive. Cai (1995) noted that concrete examples and manipulatives are
used differently in Chinese classrooms than in American classrooms, with more
linkage to abstractions made in Chinese classrooms. Stigler and Stevenson (1991)
reported less use of manipulatives in U.S. classrooms compared to Japanese and
Taiwanese classrooms and confirmed Cai’s point that manipulatives are used
more consistently to build abstract concepts in Taiwanese and Japanese classrooms.
They gave an example to illustrate that use of concrete manipulatives enables
teachers to introduce proportional reasoning in fourth grade in Japan, much earlier
than this topic is taught in the United States.
The results of this study and others based upon cognitive models of mathemat-
ical competence suggest that examining classroom activities in terms of the compo-
nents of skilled mathematical performance would be productive. Advanced math-
ematical thinking depends upon a solid base of conceptual knowledge and flexible
integration of different mathematical representations (Dreyfus & Eisenberg, 1996).
We suggest that by the sixth grade, Asian students may have advanced beyond their
American peers in this area, at least in the area of rational number. Knowing how
this happens in Asian classrooms would be useful.
REFERENCES
Baker, D. P. (1993). Compared to Japan, the U.S. is a low achiever ... really. Educational Researcher,
22(3), 18–20.
Beaton, A. E., Mullis, I. V. S., Martin, M. O., Gonzalez, E. J., Kelly, D. L., & Smith, T. A. (1996).
Mathematics achievement in the middle school years: IEA’s Third International Mathematics and
Science Study (TIMSS). Chestnut Hill, MA: Boston College.
Bondesan, M. G., & Ferrari, P. L. (1991). The active comparison of strategies in problem-solving: An
exploratory study. In F. Furinghetti (Ed.), Proceedings of the 15th international conference on the
Psychology of Mathematics Education (Vol. I, pp. 168–175). Genova, Italy: U Press.
Brenner, M. E., Mayer, R. E., Moseley, B., Brar, T. , Durán, R., Reed, B. S., & Webb, D. (1997). Learning
by understanding: The role of multiple representations in learning algebra. American Educational
Research Journal, 34, 663–689.
Cai, J. (1995). A cognitive analysis of U.S. and Chinese students’ mathematical performance on tasks
involving computation, simple problem solving, and complex problem solving. Journal for Research
in Mathematics Education Monograph Series, Number 7. Reston, VA: National Council of Teachers
of Mathematics.
Cao, C., & Cai, J. (1989). Mathematics pedagogy. Nanjing, People’s Republic of China: Jiangsu
Educational Press.
Carraher, D. W. (1996). Learning about fractions. In L. P. Steffe & P. Nesher (Eds.), Theories of math-
ematical learning (pp. 241–266). Mahwah, NJ: Erlbaum.
Dreyfus, T., & Eisenberg, T. (1996). On different facets of mathematical thinking. In R. J. Sternberg
& T. Ben-Zeev (Eds.), The nature of mathematical thinking (pp. 253–284). Mahwah, NJ: Erlbaum.
Fuson, K. C., Stigler, J. W., & Bartsch, K. (1988). Grade placement of addition and subtraction topics
in Japan, Mainland China, the Soviet Union, Taiwan, and the United States. Journal for Research
in Mathematics Education, 19, 449–456.
Herman, S. J. (1995). Connecting math performance to the representation of mathematical ideas: A study
of 6th grade students in China, Japan, Taiwan and the U.S.A. (Doctoral dissertation, University of
California, Santa Barbara, 1994). Dissertation Abstracts International, 55, 2275A.
Lesh, R., Post, T., & Behr, M. (1987). Representations and translations among representations in math-
ematics learning and problem solving. In C. Janvier (Ed.), Problems of representation in the teaching
and learning of mathematics (pp. 33–40). Hillsdale, NJ: Erlbaum.
Lewis, A. B. (1989). Training students to represent arithmetic word problems. Journal of Educational
Mayer, R. E. (1987). Educational psychology: A cognitive approach. Boston: Little, Brown.
Mayer, R. E., & Hegarty, M. (1996). The process of understanding mathematical problems. In R. J.
Sternberg & T. Ben-Zeev (Eds.), The nature of mathematical thinking (pp. 29–53). Mahwah, NJ:
Erlbaum.
Mayer, R. E., Sims, V., & Tajika, H. (1995). A comparison of how textbooks teach mathematical problem
solving in Japan and the United States. American Educational Research Journal, 32, 443–460.
Mayer, R. E., Tajika, H., & Stanley, C. (1991). Mathematical problem solving in Japan and the United
States: A controlled comparison. Journal of Educational Psychology, 83, 69–72.
McKnight, C. C., Crosswhite, F. J., Dossey, J. A., Kifer, E., Swafford, J. O., Travers, K. J., & Cooney,
T. J. (1987). The underachieving curriculum: Assessing U.S. school mathematics from an interna-
tional perspective. Champaign, IL: Stipes.
Mullis, I. V. S., Martin, M. O., Beaton, A. E., Gonzalez, E. J., Kelly, D. L., & Smith, T. A. (1997).
Mathematics achievement in the primary school years: IEA’s Third International Mathematics and
Science Study (TIMSS). Chestnut Hill, MA: Boston College. (ERIC Document Reproduction Service
No. ED 410 120)
National Council of Teachers of Mathematics. (1989). Curriculum and evaluation standards for school
mathematics. Reston, VA: Author.
Peak, L. (1996). Pursuing excellence: A study of U.S. eighth-grade mathematics and science teaching,
learning, curriculum, and achievement in international context. Washington, DC: U.S. Department
of Education, National Center for Education Statistics. (ERIC Document Reproduction Service No.
ED 400 209)
Post, T. R., Cramer, K. A., Behr, M., Lesh, R., & Harel, G. (1993). Curriculum implications of research
on the learning, teaching and assessing of rational number concepts. In T. P. Carpenter, E. Fennema,
& T. A. Romberg (Eds.), Rational numbers: An integration of research (pp. 327–362). Hillsdale,
NJ: Erlbaum.
Presmeg, N. C. (1986). Visualisation in high school mathematics. For the Learning of Mathematics,
6(3), 42–46.
Reys, B. J., Reys, R. E., & Koyama, M. (1996). The development of computation in three Japanese
primary-grade textbooks. Elementary School Journal, 96, 423-437.
Reys, R. E., Reys, B. J., Nohda, N., & Emori, H. (1995). Mental computation performance and strategy
use of Japanese students in grades 2, 4, 6, and 8. Journal for Research in Mathematics Education,
26, 304–326.
Robitaille, D. F, & Garden, R. A. (1989). The IEA study of mathematics II: Contexts and outcomes of
school mathematics. New York: Pergamon.
Schoenfeld, A. H. (1992). Learning to think mathematically: Problem solving, metacognition, and sense
making in mathematics. In D. A. Grouws (Ed.), Handbook of research on mathematics teaching and
learning (pp. 334–370). New York: Macmillan.
Silver, E. A., Leung, S. S., & Cai, J. (1995). Generating multiple solutions for a problem: A compar-
ison of the responses of U.S. and Japanese students. Educational Studies in Mathematics, 28, 35–54.
Sowder, J. (1992). Estimation and number sense. In D. A. Grouws (Ed.), Handbook of research on math-
ematics teaching and learning (pp. 371–389). New York: Macmillan.
Stevenson, H. W. (1985). An analysis of Japanese and American textbooks in mathematics. Washington,
DC: Office of Educational Research and Improvement. (ERIC Document Reproduction Service No.
ED 271 394)
Stevenson, H. W., & Bartsch, K. (1992). An analysis of Japanese and American textbooks in mathe-
matics. In R. Leetsman & H. J. Walberg (Eds.), Japanese educational productivity (pp. 103–134).
Greenwich, CT: JAI Press.
Stevenson, H. W., & Lee, S. (1990). Contexts of achievement: A study of American, Chinese, and
Japanese children. Chicago: University of Chicago Press.
Stevenson, H. W., Stigler, J. W., Lee, S. -Y., Lucker, G. W., Kitamura, S., & Hsu, C. (1985). Cognitive
performance and academic achievement of Japanese, Chinese, and American children. Child
Development, 56, 718–734.
Stigler, J. W., & Hiebert, J. (1997). Understanding and improving classroom mathematics instruction:
An overview of the TIMSS video study. Phi Delta Kappan, 79, 14–21.
Stigler, J. W., Lee, S. -Y., Lucker, G. W., & Stevenson, H. W. (1982). Curriculum and achievement in
mathematics: A study of elementary school children in Japan, Taiwan, and the United States.
Journal of Educational Psychology, 74, 315-322.
Stigler, J. W., & Miller, K. F. (1993). A good match is hard to find: Comment on Mayer, Tajika, and
Stanley. Journal of Educational Psychology, 85, 554–559.
Stigler, J. W., & Stevenson, H. W. (1991). How Asian teachers polish each lesson to perfection. American
Educator, 15, 12–20, 43–47.
Tajika, H., Mayer, R. E., Stanley, C., & Sims, V. (1997). Mathematical problem-solving processes of
students in Japan and the United States: A cross-cultural comparison. Psychologia: An International
Journal of Psychology in the Orient, 40, 131–140.
Thompson, P. W. (1990). A theoretical model of quantity-based reasoning in arithmetic and algebra
(Tech. Rep. No. 101). Normal: Illinois State University.
Tuss, P., Zimmer, J., & Ho, H. -Z. (1995). Causal attributions of underachieving fourth-grade students
in China, Japan, and the United States. Journal of Cross-Cultural Psychology, 26, 408–425.
Westbury, I. (1992). Comparing American and Japanese achievement: Is the United States really a low
achiever? Educational Researcher, 21(5), 18–24.
Willis, G. B., & Fuson, K. C. (1988). Teaching children to use schematic drawings to solve addition
and subtraction word problems. Journal of Educational Psychology, 80, 192-201.
Yerushalmy, M. (1997). Designing representations: Reasoning about functions of two variables.
Journal for Research in Mathematics Education, 28, 431-466.
Authors
Mary E. Brenner, Associate Professor, Department of Education, University of California, Santa
Barbara, Santa Barbara, CA 93106; betsy@education.ucsb.edu
Sally Herman, Department of Education, University of California, Santa Barbara, Santa Barbara, CA
93106
Hsiu-Zu Ho, Associate Professor, Department of Education, University of California, Santa Barbara,
Santa Barbara, CA 93106; ho@education.ucsb.edu
Jules M. Zimmer, Professor, Department of Education, University of California, Santa Barbara,
Santa Barbara, CA 93106; jules@education.ucsb.edu
Brief Report
The Meaning of Randomness for

Secondary School Students
Carmen Batanero and Luis Serrano, University of Granada, Spain
In the experimental study reported here we intended to examine possible differences

in secondary students’ conceptions about randomness before and after instruction in
probability, which occurs for the Spanish students between the ages of 14 and 17.
To achieve this aim, we gave 277 secondary students a written questionnaire with
some items taken from Green (1989, 1991). With our results we extend Green’s
previous research to 17-year-old students and complement his results with the
analysis of students’ arguments to support randomness in bidimensional distributions.
Our results also indicate that students’ subjective understanding of randomness is close
to some interpretations of randomness throughout history.
Key Words: Assessment; Conceptual knowledge; High school, 9–12; Probability;
Statistics and probability; Stochastics
Randomness has been interpreted in various ways during different periods in

history (Batanero, Serrano, & Green, 1998; Bennett, 1993). From ancient times to
the beginning of the Middle Ages, randomness was considered to be the opposite
of something that had known causes. With the first theoretical developments of
probability, for example in the Liber de Ludo Aleae by Cardano, randomness was
related to equiprobability because these developments were closely linked to
games of chance, for which the principle of equal probabilities is reasonable. In
later conceptions randomness was linked to the frequentist and to the subjective
approaches to probability (Bennett, 1993).
The advent of tables of random numbers at the end of the 19th century indicated
the need for examining the sequence of random results regardless of the process
by which they had been generated. Von Mises (1928/1939) based his study of this
topic (selection algorithms) on the intuitive idea that a sequence is considered to
be random if we are convinced of the impossibility of finding a method to predict
that sequence. In Kolmogorov’s approach to formalize the concept (complexity
approach), which was described by Fine (1973), a sequence would be considered
random if it could not be codified in a more parsimonious way, and the absence of
This research was supported by Grant PB96-1411 from the Ministry of Education and
Culture, Madrid, Spain.
Carmen Batanero and Luis Serrano 559
pattern would be its essential characteristic. Each of these characterizations includes

only part of the complex meaning of randomness, explaining why they have been
debated from the philosophical point of view.
Previous Research
Research into children’s perceptions of randomness started with Piaget and
Inhelder (1951), who investigated children’s understanding of patterns in
two-dimensional random distributions. They designed a piece of apparatus to
simulate raindrops falling on paving stones. When asked where the next raindrop
would fall, young children (6 to 9 years old) allocated the raindrops in approxi-
mately equal numbers on each pavement square, thereby producing a uniform distri-
bution. With older children, proportional reasoning begins to develop, and Piaget
and Inhelder believed that children understood the law of large numbers, which
simultaneously explains the overall regularity and the particular variability of
each experiment.
Green’s (1983, 1989, 1991) investigations with large samples of children aged
7 to 16, using paper-and-pencil versions of Piaget’s tasks, showed, however, that
ability to recognize randomness does not improve with age. In his study, Green
confronted the children with tasks relating to a random sequence of heads and tails
representing the results of flipping a fair coin. The study showed that children were
able to describe what was meant by equiprobable. However, they did not appear
to understand the independence of the trials, and they tended to produce series in
which runs of the same result were too short when compared to those that we would
expect in a random process. Other authors, such as Toohey (1995), Fischbein and
Gazit (1984), and Fischbein, Nello, and Marino (1991), have also documented chil-
dren’s difficulties in differentiating random and deterministic situations and their
beliefs in the possibility of controlling random experiments.
There has also been a considerable amount of research into adults’ subjective
perceptions of randomness (see Falk & Konold, 1997, for a survey); systematic
biases have consistently been found. People tend to reject sequences with long runs
of the same result (such as a long sequence of heads), and they consider sequences
with an excess of alternation of different results to be random. In the case of
two-dimensional tasks, clusters of points seem to prevent a distribution from being
perceived as random.
AIMS AND METHODOLOGY OF THE EXPERIMENTAL RESEARCH

In this report we present the results from 277 secondary school students’ written
responses to test items taken from Green (1991). About half the students (n = 147)
were in their third year of secondary school (14-year-olds) and had not studied prob-
ability. The rest of the students (n = 130) were in their last year of secondary educa-
tion (17- and 18-year-olds, preuniversity level). This second group had studied prob-
ability with a formal, mathematical approach for about a month when they were
15 years old and for another month during their previous school year. Our main
560 The Meaning of Randomness
purpose was to contribute to the investigation of the evolution of probabilistic intu-

itions with age, which according to Fischbein and Schnarch (1997) has not been
extensively studied.
Eight items were analyzed; four referred to random sequences, and the remaining
four referred to random two-dimensional distributions. For each of the two types
of items, we first compared the percentages of students who considered the situa-
tion to be random. The students’ arguments supporting or rejecting randomness
were then analyzed.
In Items 1 to 4 (see Figure 1), we studied the students’ capacities for discrimi-
nating random models in sequences of results from flipping a fair coin (examples
of Bernoulli sequences). Two variables were changed in the sequences:
1. The proportion of heads was varied. In Items 1 through 3 the proportions of heads
were very close to the theoretical probability (i.e., P(H) = .53 in Item 1; P(H) = .48
in Items 2 and 3), whereas in Item 4 the proportion of heads was only .30.
2. The lengths of runs and, consequently, the proportion of alternations (changes
in the type of outcome from head to tail or from tail to head) were varied. This propor-
tion was very close to the theoretical value of .5 in Items 1 and 4 (i.e., P(A) = .54)
and in Item 3 (i.e., P(A) = .51). However, because P(A) = .74 in Item 2, that item
had too many alternations (runs that were too short) for the sequence to be consid-
ered random.
Some children were each told to toss a coin 40 times. Some did it properly. Others
just made it up. They put H for heads and T for tails.
Maria: TTTHTHHTTTHTHHHHHTHTHTTHTHHTTTTHHH
THHTHH
Daniel: HTHTTHHTHTHHTTHTTHHTTHTHHTTHTHTHTH
THTTHT
Martin: HTTTHTTHHHTHTTTTTHTHTHHTHTTHHHHTTT
HTTHHH
Diana: HTTTHTTHTHTTTHTTTTHHTTTHTTHTTHTTTTH
TTTHT
Item 1: Did Maria make it up? How can you tell?
Item 2: Did Daniel make it up? How can you tell?
Item 3: Did Martin make it up? How can you tell?
Item 4: Did Diana make it up? How can you tell?
Figure 1. Items 1 to 4.
From a normative point of view, we considered that the correct response to Items
1 and 3 is that the child was playing within the rules and for Items 2 and 4 that the
child was cheating. Because in a test of randomness there is always a small prob-
ability that the sequence is not random in spite of having passed the randomness
tests, the students’ answers themselves were not enough to judge whether their
conceptions were right or wrong. Therefore, we were interested not just in whether
the students’ responses coincided with the normative answer but also in the prop-
erties of sequences that they associated with randomness.
We included four additional items for which students were asked to indicate
whether they thought the points in a two-dimensional array were distributed
randomly. To make the stochastic model clear for the students, we gave them the
introductory activity in Figure 2. After the teacher read aloud this text, the students
were told to play this game until they understood the rules. We then gave the chil-
dren two items to explore their perceptions of what might happen if the experiment
were continued over 16 and 30 selections (generation tasks). The last four ques-
tions (the recognition tasks) are shown in Figure 2.
Introductory activity: Paul plays a game using 16 counters numbered 1, 2, 3, 4,

…, 16. Paul puts all the counters in a tin. He shakes the tin well. Rachel shuts her
eyes and picks out a counter. It is number 7. Paul puts an x in box 7. The 7 counter
is put back into the tin and someone else picks out a counter.
1 2 3 4
5 6 7x 8
9 10 11 12
13 14 15 16
Items 5 to 8: Some children were told to play the counters game by themselves
using 16 real counters. Did some cheat and make it up?
x x x x
x xxx
x x x x x
x x x x xx xx
x
x x xx x x x
xx x xx x
x
Jaime Laura
5. Did Jaime cheat? How can you tell? 6. Did Laura cheat? How can you tell?
xx x x x x
x x
xx x x x xx xxx x
x x x x x
xxx xx x
xx xx
xx xx x xx
Miguel Luis
7. Did Miguel cheat? How can you tell? 8. Did Luis cheat? How can you tell?
Figure 2. Introductory activity and Items 5 to 8.

In a theoretical random distribution of the number of counters, about 6 empty

squares, 6 squares with only one counter, 3 with two counters, and 1 with three or
more counters would be expected. The chi-squared test of goodness of fit between
the theoretical and the observed distribution in each of the items yielded a value
of χ2 = 1.83 for Item 5 (Jaime), χ2 = 2.39 for Item 6 (Laura), χ2 = 15.50 for Item
7 (Miguel), and χ2 = 24.36 for Item 8 (Luis). In addition, Laura’s pattern had too
many adjacent empty squares, all of them on the main diagonal. Consequently, only
Jaime’s distribution would be considered random from a normative point of view.
RESULTS AND DISCUSSION
The percentages of students considering each item to be random are presented in

Table 1. For 14-year-old students, the percentages were similar to those obtained
by Green (1991) with students of the same age. There were no 17-year-old students
in Green’s research. Toohey’s (1995) study did not include such comparative data.
Table 1
Percentages of Students Who Considered the Situations to Be Random
Age
14 17
Items (n = 147) (n = 130)
Linear sequences
1. Maria P(H) = .53, P(A) = .54 60 58
2. Daniel P(H) = .48, P(A) = .74 58 63
3. Martin P(H) = .48, P(A) = .51 53 56
4. Diana P(H) = .30, P(A) = .54 36 37
Bidimensional arrays
5. Jaime χ2 = 1.83 85 85
6. Laura χ2 = 2.39 57 58
7. Miguel χ2 = 15.50 36 38
8. Luis χ2 = 24.36 25 22
Most students considered all Bernoulli sequences (Items 1 to 4) to be random

except for Item 4, for which the frequency of heads (12) was quite different from
the theoretical frequency expected in a random sequence. In Item 2, most students
considered the sequence to be random even though the number of alternations (and
consequently the lengths of runs) had been biased. This result indicates that
students had greater difficulty in recognizing run properties than frequency prop-
erties. The consistency in the percentages of students considering the sequences
to be random in Items 1, 2, and 3 indicates that the similarity between the observed
and expected frequencies may be more important than the run lengths in students’
deciding whether a sequence is random.
There was a greater spread in the number of students considering the bidimen-
sional point distributions to be random (Items 5 to 8). Here the students also
seemed to concentrate on the differences between the theoretical and observed

frequencies because the percentage of students considering an item to be random
decreased when this difference increased.
Analysis of Students’ Arguments

There were no significant differences by age group among the proportions of
students considering the items to be random (as shown in Table 1), which might
indicate that age and instruction have little influence on students’ conceptions of
randomness. This implication is difficult to assess on the basis of only these
percentages because the concept of randomness is very complex. To clarify which
specific properties of randomness the students used to make their judgments and
whether these properties would depend on age and the items’ variables, we asked
the students to justify their answers. Their arguments were classified according to
a scheme similar to one developed by Green (1991) to analyze students’ responses
for Bernoulli sequences (Items 1 to 4). Green did not classify students’ arguments
in justifying the randomness of bidimensional arrays such as in Items 5 to 8. We
used the following categories for responses for all the items, whether a response
was used to justify or reject randomness:
1. There is a regular pattern. This reasoning refers to the order in which heads
and tails appear in the sequence and the regularity of this pattern or the regularity
in the spatial arrangement of the results presented: “He might have made it up,
because it is all too correct, all in its place, well ordered”; “He did not cheat, because
there is a very regular sequence of heads and tails, which is most probable to
happen.”
2. There is an irregular pattern in the sequence. For example, “He did not make
it up, because the sequence does not follow any order.” In this example the student
associates the lack of pattern with randomness; this approach is consistent with the
complexity approach to randomness (described earlier) in which the absence of
patterns is an essential characteristic.
3. The frequencies of the possible results are quite similar. For example, “He
did not cheat, because the numbers for heads and tails are very balanced” or “He
is cheating; it is very difficult that all the squares should appear the same number
of times.”
4. The frequencies of the possible results are quite different. This is the oppo-
site of the previous argument. The previous two reasons are based on a compar-
ison between the observed frequencies and the theoretical probability distribution.
Underlying such reasoning might be the frequentist approach to probability and
randomness, an approach in which an object is considered a random member of a
class if one can select it using a method that provides a given a priori relative
frequency in the long run. Some students’ responses relate randomness to equiprob-
ability, using a classical approach in which an object is considered a random
member of a class if there is the same probability of occurrence for this object and
for any other member of its class: “There must be equal probability for heads and
tails.”
5. There are long runs (there is a cell with too many points). Students with this
view show misconceptions concerning the idea of independence between succes-
sive trials: “Because there are many consecutive tails” or “Some squares have too
many crosses.”
6. There are no runs (there are no cells with several points). This is the reversal
of the previous reason: “There are too few sequences of heads.”
7. It is unpredictable; it is random. Students giving this response might say,
for example, “That is luck” or “Though correctly flipped, it can give those
results because it is a game.” Such students could be reasoning following the
outcome approach, described by Konold (1991), in which people interpret ques-
tions about probability in a nonprobabilistic way and rely on the unpredictability
of random events to avoid a decision. Belief in an underlying causal mechanism,
a belief that parallels the earlier meaning of randomness, might be implicit in some
of these categories of students’ explanations.
In Table 2 we present the percentages of students giving these arguments for each
item. To complement the study by Green (1991), who did not approach variation
by items in this way, we classified the percentages according to whether the
students considered the situations to be random. In general, the highest frequen-
cies appear for responses based on unpredictability. Certain reasons were commonly
given for decisions about randomness for certain items: (a) regular patterns for Items
Table 2
Percentages of Students’ Arguments in Supporting or Rejecting Randomness
Argument
Pattern Frequencies Runs/Clusters
Unpre- No
Response Item Regular Irregular Similar Different Long None dictable reasons
Random 1. Maria 5 14 10 5 1 1 52 12
2. Daniel 12 8 19 3 0 4 43 11
3. Martin 1 23 7 4 2 2 48 13
4. Diana 13 7 5 5 1 3 51 15
5. Jaime 1 31 1 11 0 0 39 17
6. Laura 2 28 0 4 6 0 46 14
7. Miguel 9 12 0 4 0 0 62 13
8. Luis 14 0 24 3 0 1 40 18
Not 1. Maria 6 2 6 18 50 3 12 3
random 2. Daniel 81 1 6 1 0 6 5 0
3. Martin 13 5 2 18 54 0 5 3
4. Diana 3 7 1 35 30 3 18 3
5. Jaime 3 20 0 5 33 0 25 14
6. Laura 17 6 0 5 47 0 20 5
7. Miguel 74 6 3 2 0 0 10 5
8. Luis 27 0 63 0 0 0 8 2
Note. Items 1 through 4 were linear sequences; Items 5 through 8 were bidimensional arrays.
2, 7, and 8; (b) irregular patterns for Items 3, 5, and 6; (c) differing frequencies for
the possible outcomes for Item 4; (d) the long runs in linear sequences for Items
1, 3, and 4 and, similarly, the cells with too many points for the arrays in Items 5
and 6. This clustering of responses indicates that the students were able to discern
the items’ features.
Table 2 shows that students’ arguments differed depending on whether the
sequence was considered to be random or not (p values in the chi-squared test of
independence between argument and type of response were less than .01 for all the
items). The main arguments given to support randomness were unpredictability and
irregular patterns. Regular patterns and long runs or clusters of points were asso-
ciated with lack of randomness.
We also compared the arguments used in the two groups by carrying out the
chi-squared test of independence between the argument and group of students (p
values were less than .01 for all items except Item 6). Younger students made more
reference to runs and clusters and to the regularity or irregularity of patterns,
whereas 17-year-old students seemed to favor frequency-based reasoning and
unpredictability.
FINAL REFLECTIONS
Our results show that students associated both correct and incorrect properties
with randomness. On the one hand, they perceived the local variability, lack of
patterns in the linear or spatial arrangements of outcomes, and unpredictability of
the random processes underlying the tasks we gave them. In many cases, our
students carried out informal statistical analyses of the frequencies for the different
events in the random sequences and compared these frequencies with an under-
lying equiprobability model. For two-dimensional distributions, they rejected
items with only one or two counters per square as well as the item with several adja-
cent empty squares. Educators should consider all these correct intuitions when
organizing the teaching of probability because, when learning something new,
students construct their own meanings by connecting the new information to what
they already believe to be true.
On the other hand, we must also consider the students’ misconceptions in
reviewing our ways of teaching. Students overemphasized unpredictability and luck
to justify their attributions of randomness, and this tendency seemed greater in the
older students when they analyzed the two-dimensional distributions, tasks that are
not considered in the Spanish curriculum. The existence of runs of 5 and even 4
identical results in the sequences and cells with clusters of 3 or 4 markers in two-
dimensional distributions was mistakenly associated with lack of randomness,
though after instruction a smaller proportion of students provided these types of
arguments. Because accepting the existence of runs and clusters underlies under-
standing of independence, our results indicate that independence is not an intuitive
idea and that students continued to experience difficulties with the idea of inde-
pendence even after instruction, even when some improvement was noticeable.
The students’ arguments and responses also indicate underlying conceptions that
parallel some of the meanings attributed to randomness throughout history. In
particular, students relate randomness to luck and unknown causes and to proba-
bility in its classical and frequentist approaches. Some of the features of the
complexity approach (lack of pattern) and selection algorithms (unpredictability)
are also found in their responses. All these results are very close to Green’s (1991),
and, therefore, his results could be extended to our group of 17-year-old students.
In addition, we have extended the classification of students’ arguments, which
Green carried out for linear sequences, to two-dimensional distributions.
In summary, our experimental results as well as our previous analysis reveal the
complexity of the meaning of randomness; to understand randomness one needs
to understand several properties. It may in fact be preferable to consider the term
randomness as a label with which we associate many concepts, such as experiment,
event, sample space, or probability (Konold et al., 1991). In this sense, the word
randomness refers us to a collection of mathematical concepts and procedures that
we can apply in many situations. We should think about the orientation we take
toward the phenomenon that we qualify as random instead of thinking about a
quality thereof. We apply a mathematical model to a random situation because this
model is useful to describe it and to understand it, but we do not believe that the
situation will be identical to the model. Deciding, in a given situation, when prob-
ability is more suited than other mathematical models is part of the work of
modeling that we should encourage among our students.
REFERENCES
Batanero, C., Serrano, L., & Green, D. R. (1998). Randomness, its meanings and educational impli-
cations. International Journal of Mathematical Education in Science and Technology, 29, 113–123.
Bennett, D. J. (1993). The development of the mathematical concept of randomness: Educational impli-
cations (Doctoral dissertation, New York University, 1993). Dissertation Abstracts International,
54, 449A. (University Microfilms No. 93–17, 657)
Falk, R., & Konold, C. (1997). Making sense of randomness: Implicit encoding as a basis for judgment.
Psychological Review, 104, 301–318.
Fine, T. L. (1973). Theories of probability; an examination of foundations. London: Academic Press.
Fischbein, E., & Gazit, A. (1984). Does the teaching of probability improve probabilistic intuitions?
Educational Studies in Mathematics, 15, 1–24.
Fischbein, E., Nello, M. S., & Marino, M. S. (1991). Factors affecting probabilistic judgements in chil-
dren and adolescents. Educational Studies in Mathematics, 22, 523–549.
Fischbein, E., & Schnarch, D. (1997). The evolution with age of probabilistic, intuitively based miscon-
ceptions. Journal for Research in Mathematics Education, 28, 96–105.
Green, D. (1989). School pupils’ understanding of randomness. In R. Morris (Ed.), Studies in mathe-
matics education: Vol. 7. The teaching of statistics (pp. 27–39). Paris: UNESCO.
Green, D. R. (1983). A survey of probabilistic concepts in 3000 students aged 11–16 years. In D. R.
Grey, P. Holmes, V. Barnett, & G. M. Constable (Eds.), Proceedings of the First International
Conference on Teaching Statistics (Vol. 2, pp. 766–783). Sheffield, England: University of Sheffield.
Green, D. R. (1991). A longitudinal study of pupils’ probability concepts (Tech. Rep. ME90/91).
Loughborough, England: Loughborough University of Technology.
Konold, C. (1991). Understanding students’ beliefs about probability. In E. von Glasersfeld (Ed.),
Radical constructivism in mathematics education (pp. 139–156). Dordrecht, The Netherlands:
Kluwer.
Konold, C., Lohmeier, J., Pollatsek, A., Well, A., Falk, R., & Lipson, A. (1991). Novices views on
randomness. In R. G. Underhill (Ed.), Proceedings of the thirteenth annual meeting of the North
American Chapter of the International Group for the Psychology of Mathematics Education (Vol.
1, pp. 167–173). Blacksburg: Virginia Polytechnic Institute and State University.
Piaget, J., & Inhelder, B. (1951). La genése de l’idée de hasard chez l’enfant [The origin of the idea of
chance in the child]. Paris: Presses Universitaires de France.
Toohey, P. G. (1995). Adolescent perceptions of the concept of randomness. Unpublished master’s thesis,
University of Adelaide, Adelaide, Australia.
Von Mises, R. (1939). Probability, statistics and truth (J. Neyman, D. Sholl, & E. Rabinowitsch, Trans.).
New York: Macmillan. (Original work published 1928)
Authors
Carmen Batanero, Profesora Titular de Universidad, Departamento de Didáctica de la Matemática,
Facultad de Ciencias de la Educación, Campus de Cartuja, Universidad de Granada, 18071 Granada,
Spain; batanero@goliat.ugr.es
Luis Serrano, Profesor Titular de Escuela Universitaria, Departamento de Didáctica de la Matemática,
Escuela de Formación del Profesorado, Alfonso XIII, S.N., Melilla, Spain; lserrano@goliat.ugr.es
1999, Vol. 30, No. 5, 568–578
Revitalizing and
Refocusing Our Efforts
Glenda Lappan, President, National Council of Teachers of Mathematics
Michigan State University
We are currently standing at one of the most

important crossroads we have faced as a Council
and as a profession in the past 20 years. We have
more than a decade of direction setting and
progress under our belt. And we have learned a
lot. We have every right to celebrate our
successes even as we look to the future and the
continuing need for improvement.
CELEBRATE OUR SUCCESS
As an organization, we have had a substantial

influence on setting standards as a way to guide
the building of excellent instructional programs in
all subject areas. We are updating our Standards
Glenda Lappan in mathematics to build on our experience with
our first wave of Standards documents.
The Standards have influenced more than a decade of curriculum development
that takes seriously the following distinctive features and goals:
• Higher mathematical expectations for students
• An inquiry-investigative approach to teaching
• Appropriate use of technology in learning mathematics
• A central role for applications or uses of mathematics
• A more comprehensive view of assessment
Figures 1 and 2 are mathematics problems that illustrate raising expectations and
letting students investigate intriguing relationships. Try these problems before
reading the answers at the end of the article.
Also, we should celebrate our progress toward our teaching goals. Our new
teaching goals attempt to reach all students. To make progress in attaining these
The remarks published here have been adapted from the Presidential Address at the 77th
Annual Meeting of the National Council of Teachers of Mathematics in San Francisco,
California, April 1999.
Glenda Lappan 569
A
B
Directions: Make the figure above from two squares of colored acetate joined as
shown so that square B rotates around the center point of square A. As square B rotates,
what is the largest possible area of the overlapping shaded region?
A A A
B
B
B
Figure 1. The overlapping-squares problem.
Part A. If you and your friend each make a quadrilateral from the same four lengths, two
sides of length 5 cm and two sides of length 3 cm, will you get exactly the same shape?
Part B. If you and your friend each make a triangle from the same three lengths—5 cm,
3 cm, and 3.5 cm, will you get exactly the same shape?
Part C. Explain in what ways the two situations, part A and part B, are the same and in
what ways they are different, and why.
Figure 2. The quadrilateral problem.
goals, we recognize that students differ in substantial ways. Not all students reach
the same levels of cognitive growth at the same time. The mathematics classroom
teacher therefore must create an environment that supports students’ mathematics
growth at many different levels. Accommodating students who mature in different
ways and on different timetables takes all the following:
Mathematics must be connected to students’ interests. Students in Grades K–12
deal with many social, emotional, and physical challenges. To capture their atten-
tion and their growing cognitive powers, many mathematical tasks posed for
students should focus on material that they find interesting and important. Things
that adults think will interest students are not always on target. Students should
570 Revitalizing and Refocusing Our Efforts
therefore have opportunities for choice within the mathematics classroom. Choice
of project topics, choice of problems of the week on which to work, and choice of
solution strategies for tackling complex problem situations are ways of allowing
students to feel the power of choice. Figures 3 and 4 show an example of a problem
that young students enjoy because it relates to their own real world.
(a)
The problem
Reprinted from Karen Economopoulos and Tracey Wright. Investigations in Number, Data, and
Space: Collecting and Representing Data: How many pockets? How many teeth? ”Dale Seymour
Publications. Used by permission.
After the students worked to summarize, represent, and discuss what they
observed, the teacher gave the mystery missing-teeth data presented in
Figure 4. The classes are kindergarten, second grade, third grade, and fourth
grade. The students were then asked which is which and how they knew that
they were correct.
(b)
An extension
Figure 3. A real-world problem to the students.
Mathematics should be considered as an experimental science. For students to

make sense of concepts, ways of reasoning, and productive procedures; use
problem-solving strategies in mathematics; and develop skill with number and
symbolic operations and procedures, they need opportunities to explore, investi-
gate, invent, generalize, abstract, and construct arguments to support their ideas and
Glenda Lappan 571
Reprinted from Karen Economopoulos and Tracey Wright. Investigations in Number, Data, and
Space: Collecting and Representing Data: How many pockets? How many teeth? ”Dale Seymour
Publications. Used by permission.
Figure 4. Mystery missing-teeth data.

prove their conjectures. Newer mathematics materials are problem oriented and
pose interesting contexts for investigation. Within these contexts, students bump
into mathematics that they need but do not yet have in their tool kits. This need
drives students to invent and create strategies for solving problems. Out of these
student ideas come the conversations that the teacher skillfully uses to help make
the underlying mathematical ideas, procedures, skills, and arguments more explicit
for the students.
Creating a tool-rich environment for learning mathematics is important.
Technology is engaging to students. It also allows access to mathematics that could
not be explored in the past. It helps create environments in which students can see
changing relationships among variables and can engage in a dynamic way with
mathematical conjecturing. It allows students to tackle real problems with messy
data and gives them control over different forms of representation of mathemat-
ical relationships.
Unfortunately, we face unacceptable inequities in students’ use of technology.
Although the percent of schools with Internet hookups has increased, hookups lag
within classrooms. A recent study (U.S. Department of Education, National Center
for Education Statistics, 1999a) indicated that 78% of public schools had Internet
access but only 27% had that access in an instructional setting. Using the Internet as
a classroom resource is not possible in many schools. In addition, the rate of increase
in Internet access in poor schools is much smaller than in more affluent schools. We
have to be concerned that the rich schools get richer in what they can offer while chil-
dren in poor communities fall even further behind. This inequity of access to tools
for doing mathematics is a serious societal problem that we must all face.
Teachers must understand the sense that students are making of mathematics.
Since students are at many stages of cognitive, physical, and social development,
the teacher needs to understand where all students are in their mathematical growth.
Typical paper-and-pencil computation-driven tests do not lead to specific under-
standing of students’ thinking. Thus the teacher needs to create and use many oppor-
tunities for assessing students’ understanding. This need to better understand the
sense that students are making of mathematics has led to new forms of assessments
that range from partner or group problem solving to portfolios.
STAY THE COURSE
We have learned that setting careful directions and staying the course can help
improve mathematics programs. Radical starts and direction changes are detrimental
to progress. Many teachers simply do what they have always done in teaching math-
ematics and say, “This too shall pass.” To make progress, we must learn as we go
and make course corrections but not radical swings. We are not wrong in our basic
set of commitments. Every child deserves a high-quality mathematics program
taught by a knowledgeable, caring teacher. Research shows that if we value under-
standing and being able to think and reason with what we know, the kinds of engage-
Glenda Lappan 573
ments with mathematics that students have must include opportunities to muck
about with ideas; to struggle to make sense of new mathematics situations; to try out
ideas; and with the help of classmates and the teacher, raise those ideas to powerful
mathematics strategies and ways of thinking. We have data to support our optimism.
• Scores on the National Assessment of Educational Progress (NAEP) have risen
slowly during the 1990s in all subgroups of students in schools. The alignment of
the NAEP framework with our Standards gives us a way to monitor whether the
direction of change is producing the desired results. Of course we are not satisfied,
but the direction of change is positive (Wilson & Blank, 1999).
• During the 1990s, professional development opportunities for teachers, time
spent in professional development, and the perceived helpfulness of these activi-
ties have risen (U.S. Department of Education, National Center for Education
Statistics, 1999b). Of particular interest to me is the percent of teachers involved
in professional development who participated in an in-depth study in their subject
field. This change was from 29% in 1993–1994 to 73% in 1998. We also have an
indication that teachers’ classroom instruction is greatly influenced by intense,
extended professional development experiences.
• The establishment of the National Board for Professional Teaching Standards
has helped us gain recognition for the professionalism of teachers. It is exciting to
see the tremendous increase in the numbers of mathematics teachers who seek Board
certification and the growing recognition in communities and schools of those
teachers who become Board certified. Figure 5 gives information related to this effort.
• More than 1836 teachers are currently Board certified.

• At least 38 states have nationally certified teachers.
• North Carolina has the most—nearly one third of the teachers certified by the National
Board for Professional Teaching Standards (NBPTS).
• In 1998–1999, more than 6000 teachers applied for national certification—the number
who passed is not yet known.
• Approximately 45% of applicants pass.
Figure 5. Facts about National Board Certification.
• Let us make our celebration personal. I invite everyone who has taken the risk
of trying one new teaching, technology, content, or assessment idea in one of your
classes in this school year, whether it succeeded or failed, to stop reading and pat
yourself on the back!
We have much to celebrate. We are a brave and caring bunch—willing to take
risks so that we can continue to grow in our skill and knowledge in meeting the
mathematical needs of our students. But we face formidable challenges in our quest
to serve our students well.
TIME: CHANGING THE SOCIETAL DEMANDS ON SCHOOLS
One of our huge challenges is time! With thanks to Jerry Becker, I would like
to use the summary of Jamie Vollmer, a former CEO of the Great Midwestern Ice
Cream Company and an attorney, to remind us of how demands on teachers and
school programs have increased in this century (Vollmer, 1996). Here is his
summary of added new demands on school programs:
• From 1900 to 1920—nutrition, immunization, and health
• From 1920 to 1950—vocational education, practical arts, physical education,
and school lunch programs
• During the 1950s—addition of safety education and driver education, strength-
ened foreign-language education, and introduction of sex education
• During the 1960s—consumer education, career education, peace education,
leisure education, and recreational education
• During the 1970s—special education, character education, drug- and alcohol-
abuse education, and school breakfast programs
• During the 1980s, the floodgates opened—keyboarding, computer education,
global education, ethics education, multicultural and nonsexist education, English-
as-a-second language and bilingual education, early childhood programs, full-day
kindergartens, preschool programs for at-risk students, stranger-danger educa-
tion, sexual-abuse-prevention programs, child-abuse monitoring
• Finally, the 1990s—HIV and AIDS education, death education, and gang
education in urban centers; bus safety; and bicycle safety
To these demands we can add all the changes in teaching and learning mathe-
matics! No wonder that our jobs have become more demanding. Schools have
assumed responsibility for many aspects of educating children that were once
considered the responsibility of parents. Although all these programs are needed,
the school day and year have not been adjusted to meet these needs. Academic
programs are getting short shrift in the grander scheme.
Perhaps an even greater need for time comes from teachers’ perspectives. The
issue of time—to teach, to study, to interact with colleagues, to prepare, to eval-
uate students, to build programs—is one of the big challenges of the next decade.
We must lobby in our communities for time for mathematics and time for such plan-
ning and inquiry, and we must be willing to deliver when we are given that time.
TEXTBOOKS—ANOTHER CHALLENGE
Another continuing challenge is textbook materials. Although we have made

progress, we have by no means carried the day in this area. We can all quote the
Glenda Lappan 575
characterization of the U.S. curriculum from the Third International Mathematics

and Science Study (TIMSS), “a mile wide and an inch deep,” or from the Second
International Mathematics Study (SIMS), “the underachieving curriculum.” But the
system and its current traditions make change to more focused, slimmer curriculum
textbooks extremely difficult. We are part of the problem. We all want to see our
favorite topics in the textbooks. But the union of all our favorite topics leads to the
500- to 600-page concrete slabs that bend children’s spines and break their spirits.
We also are part of the problem when our students proceed from one grade level to
the next while retaining little or none of what was taught the previous year.
But I place on the states partial blame for the current dilemma about textbook
materials. Pick a grade level and look at the world through the eyes of the textbook
publishers who have so many competing voices to satisfy. The union of all state
standards for that grade level promotes the cover-everything-every-year syndrome.
Publishers are also part of the problem. As the rhetoric has heated up over math-
ematics reform, many stories that have been reported to me about publishers’ sales
representatives make me sad. If the stories that I hear are correct, publishers are
guilty of disseminating unwarranted and unsubstantiated bad press about the
competition. For the sake of our children, we should be offering quality products—
presenting our best evidence that the products are effective with children—and not
attempting to “win” by working to undermine other good products. In the end we
all win by bringing statesmanship into the process. Publishers are in business to
make money, and we all respect that goal. However, we need to be allies in
producing excellent material for children and teachers. I suppose that this attitude
is naïve, but the students should be our conscience. At the very least, we should
be fair to one another and do no harm!
WORKING TOGETHER TO BUILD A NEW SOCIETY
The future of both the United States and Canada depends on a strong, competi-
tive workforce and a citizenry equipped to function in a complex world. To para-
phrase the National Science Board (1999), our interests encompass what every
student in a grade level should know and should be able to do in mathematics and
in science. Educational excellence improves not just the health of mathematics and
science but everyone’s life chances through productive employment, active citi-
zenship, and continuous learning.
This statement certainly gives credence to the important stake that communities,
parents, administrators, school boards, business and industry, and—above all—
teachers have in building mathematics programs. Teachers have the ultimate respon-
sibility for teaching children. As we struggle to build ways of interacting with the
broad range of stakeholders, two things are clear to me. First, the system of decision
making must respect the experience and knowledge of teachers and appreciate the
challenges that we face in teaching children. And second, teachers must be open and
welcoming to parental concerns and input. When respect is a part of the dialogue,
we can make progress. However, that respect is missing in far too many exchanges.
I am not complaining about differences of opinion. We need to have dialogue.

We need to exchange differing ideas. But we need to listen and learn. The future
needs to be built on the best of the past. But the past is not good enough! If we
respect research, we must recognize that the longest running large-scale experiment
in mathematics education is with the curriculum of the past 50 years and the
American teaching script. The evidence that they do not work for a huge propor-
tion of our children is substantial. Our children deserve better. We must learn to
listen to, and work with, those with whom we disagree even as we steadfastly work
to build better mathematics programs.
Remember that the stakes are very high indeed. During the past decade the math-
ematics-education community has examined the changing mathematical needs of
society. Instead of adults able to function smoothly and effectively in a stable
economy geared to mass production, society now needs people who are sufficiently
independent in their thinking to be adaptive, innovative, and inventive.
A recent publication of the Educational Testing Service, Education for What?
(Carnevale & Rose, 1998), reports that the new office economy is changing the face
of work. In the United States, 41% of workers have office jobs. Most of these jobs
require a college education, unlike the less skilled office jobs of the past. Other
aspects of the new office worker’s work life include flexibility, learning on the job,
and keeping up-to-date. Many of these jobs require quantitative reasoning and other
mathematical skills.
The workforce of the future will require a high level of technical skill—using
computers for tasks ranging from word processing to controlling machines, analyzing
complicated sets of data, and ensuring quality control in production processes. Our
present academic preparation in mathematics barely touches on these skill areas. As
stated in the National Research Council report Mathematical Preparation of the
Technical Workforce (Mathematical Sciences Education Board, 1995), “Mathe-
matics in the workplace is quite different from mathematics in school. It is more
concrete and more intuitive, yet at the same time more exacting and more unpre-
dictable. It is rich in data and inextricably linked to technology” (p. 5). To become
adults who are capable of thriving in the new workplace, students must be active
learners and collaborative problem solvers.
The result of this decade of work, beginning with the publication of the
Curriculum and Evaluation Standards for School Mathematics in 1989, is a
national vision of a powerful mathematics education for all students, regardless of
color, ethnicity, socioeconomic status, primary language, or gender.
To create mathematics programs that can foster powerful mathematical learning
for all students, we must make fundamental changes in students’ opportunities to
learn. These changes include modernizing the curriculum, improving classroom
instruction, and assessing student progress in a way that informs and supports the
continued mathematical learning of each student.
A while ago, I summarized how I think about what we are trying to accomplish
for students. This summary “standard” has become an important touchstone for my
own work and that of my close colleagues. I hope that you will agree that this state-
Glenda Lappan 577
ment captures the spirit of reform even if you disagree with the coverage or the
particulars.
Students should be able to communicate proficiently in mathematics. Communication
includes knowledge and skill in using the vocabulary, forms of representation, mate-
rials, tools, techniques, and intellectual methods of the discipline of mathematics,
including the ability to define and solve problems with reason, insight, inventiveness,
and technical proficiency.
The reform of mathematics education in the United States and Canada focuses
on helping us move our mathematics programs toward a curriculum that
• represents significant, powerful mathematics for all students;
• emphasizes topics that are relevant to students’ present and future needs; and
• emphasizes the full use of such tools as calculators, computers, measuring
devices, tiles, and cubes for making sense of and doing mathematics.
We have made progress, and yet we recognize that we have a long way to go to
build a high-quality mathematics program in every classroom in the United States
and Canada. In these high-quality classrooms, students learn to use their mathe-
matical knowledge to adapt, innovate, and invent new strategies and solutions.
We have every right to be proud of our profession and of our accomplishments.
A character in Pat Conroy’s The Prince of Tides (1989) said, “There’s no word in
the language I revere more than teacher. My heart sings when a kid refers to me
as his teacher, and it always has. I’ve honored myself and the entire family of man
by becoming one” (p. 557).
ANSWERS
For the overlapping-squares problem, the area of the overlapping region is the
same no matter where Square B stops in its rotation around Square A.
For the quadrilateral problem, the figures may be different. If the short lengths
are opposite each other, a parallelogram would be formed; if the short lengths are
attached to each other, a kite would be formed. The quadrilaterals may also be
different even if the lengths are placed in the same order. The angles can vary
because the figures are not rigid. The perimeters would be the same for any of the
quadrilaterals, but the areas may vary. In the case of the triangle, the two resulting
figures will be congruent. Having three identical sides is sufficient to guarantee
congruence of triangles, whereas having four congruent sides does not guarantee
congruence of quadrilaterals.
In the follow-up problem in Figure 4, A is Grade 3, B is Grade 2, C is Kinder-
garten, and D is Grade 4. One key point is recognizing that younger children have
lost fewer teeth, so when the mystery data are organized and compared, you can
tell the order of the age of the children in the four samples.
REFERENCES
Carnevale, A. P., & Rose, S. J. (1998). Education for what? The new office economy. Princeton, NJ:
Educational Testing Service.
Conroy, P. (1986). The prince of tides. Boston: Houghton Mifflin.
Economopoulos, K., & Wright, T. (1998). Investigations in number, data, and space: Collecting and
representing data: How many pockets? How many teeth? White Plains, NY: Dale Seymour.
Mathematical Sciences Education Board. (1995). Mathematical preparation of the technical work force:
Report of a workshop. Washington, DC: National Academy Press.
National Council of Teachers of Mathematics. (1989). Curriculum and evaluation standards for school
mathematics. Reston, VA: Author.
National Science Board. (1999, March 2). Preparing our children: Math and science education in the
national interest. NSB 99-31. Arlington, VA: Author.
U.S. Department of Education, National Center for Education Statistics. (1999a). The condition of educa-
tion 1998. NCES 1999-005. Washington, DC: U.S. Government Printing Office.
U.S. Department of Education, National Center for Education Statistics. (1999b). Teacher quality: A
report on the preparation and qualifications of public school teachers. NCES 1999-008. Washington,
DC: U.S. Government Printing Office.
Vollmer, J. (1996, October 17). Talk given at the annual meeting of the Illinois Council of Teachers
of Mathematics, Springfield, IL.
Wilson, L. D., & Blank, R. D. (1999). Improving mathematics education using results from NAEP and
TIMSS. Washington, DC: Council of Chief State School Officers.
Review
Knowing and Teaching Elementary Mathematics
Knowing and Teaching Elementary Mathematics: Teachers’ Understanding of

Fundamental Mathematics in China and the United States. (1999). Liping Ma.
Lawrence Erlbaum Associates, xxv + 166 pp. ISBN 0-8058-2909-1(pb) $19.95.
ISBN 0-8058-2908-3 (hb) $45.00.
Reviewed by ROGER HOWE, Yale University
(The following review has been reprinted, with permission, from the
Notices of the American Mathematical Society, Vol. 46, No. 8, pp. 881–887)
Notation: The reviewer will refer to the book under review as KTEM.
For all who are concerned with mathematics education (a set which should
include nearly everyone receiving the Notices), KTEM is an important book. For
those who are skeptical that mathematics education research can say much of value,
it can serve as a counterexample. For those interested in improving precollege math-
ematics education in the U.S., it provides important clues to the nature of the
problem. An added bonus is that, despite the somewhat forbidding educationese
of its title, the book is quite readable. (You should be getting the idea that I recom-
mend this book!)
Since the publication in 1989 of the Curriculum and Evaluation Standards by
the National Council of Teachers of Mathematics [NCTM], there has been a
steady increase in discussion and debate about reforming mathematics education
in the U.S., including increased attention from university mathematicians (cf.
[Ho]). Many mathematicians who take time to consider pre-college education form
an intuition that it would help the situation if teachers knew more mathematics. If
these mathematicians get more involved in mathematics education, they are likely
to be surprised by how little this intuition seems to affect the agenda in mathematics
education reform.
Partly this noninterest in mathematical expertise reflects an attitude widespread
among educators [Hi] that “facts,” and indeed all subject matter, are secondary in
importance to a generalized, subject-independent teaching skill and the develop-
ment of “higher-order thinking.” Concerning mathematics in particular, the study
[Be] is often cited as evidence for the irrelevance of subject matter knowledge. For
this study, college mathematics training, as measured by courses taken, was used
as a proxy for a teacher’s mathematical knowledge. The correlation of this with
student achievement was found to be slightly negative. A similar but less specific
method was used in the recent huge Third International Mathematics and Science
Study (TIMSS) of comparative mathematics achievement in forty-odd countries.
580 Knowing and Teaching Elementary Mathematics
For TIMSS, U.S. students demonstrated adequate (in fourth grade) to poor (in
twelfth grade) mathematics achievement [DoEd1–3]. To analyze whether teacher
knowledge might help explain TIMSS outcomes, data on teacher training was gath-
ered. In terms of college study, U.S. teachers appear to be comparable to their coun-
terparts in other countries [DoEd1–3].
How can this intuition—that better grasp of mathematics would produce better
teaching—appear to be so wrong? KTEM suggests an answer. It seems that
successful completion of college course work is not evidence of thorough under-
standing of elementary mathematics. Most university mathematicians see much of
advanced mathematics as a deepening and broadening, a refinement and clarifi-
cation, an extension and fulfillment of elementary mathematics. However, it seems
that it is possible to take and pass advanced courses without understanding how
they illuminate more elementary material, particularly if one’s understanding of
that material is superficial. Over the past ten years or so, Deborah Ball and others
[B1–3] have interviewed many teachers and prospective teachers, probing their
grasp of the principles behind school mathematics. KTEM extends this work to a
transnational context. The picture that emerges is highly instructive—and sobering.
Mathematicians can be pleased to have at last powerful evidence that mathemat-
ical knowledge of teachers does play a vital role in mathematics learning. However,
it seems also that the kind of knowledge that is needed is different from what most
U.S. teacher preparation schemes provide, and we have currently hardly any insti-
tutional structures for fostering the appropriate kind of understanding.
The main body of KTEM (Chapters 1–4) presents the results of interviews with
elementary school teachers from the U.S. (23 in all) and China (72 in all). The U.S.
teachers were roughly evenly split between experienced teachers and beginners.
Ma judged the group as a whole to be “above average.” In particular, although “math
anxiety” is rampant among elementary school teachers, this group had positive atti-
tudes about mathematics: they overwhelmingly felt that they could handle basic
mathematics and that they could learn advanced mathematics. The Chinese teachers
were from schools chosen to represent the range of Chinese teaching experience
and expertise: urban schools and rural, stronger schools and weaker.
The teachers’ grasp of mathematics was probed in interviews organized around
four questions. In summary form, the questions were as follows:
1) How would you teach subtraction of two-digit numbers when “borrowing” or
“regrouping” is needed?
2) In a multiplication problem such as 123 × 645, how would you explain what is
wrong to a student who performs the calculation as follows?
123
×645
615
492
738
1845
(The student has correctly formed the partial products of 123 with the digits of 645,
but has not “shifted them to the left,” as required to get a correct answer.)
Roger Howe 581
1 43
3) Compute 1 . Then make up a story problem which models this computation, that
2
is, for which this computation provides the answer.
4) Suppose you have been studying perimeter and area and a student comes to you
excited by a new “theory”: area increases with perimeter. As justification the student
provides the example of a 4 × 4 square changing to a 4 × 8 rectangle: perimeter
increases from 16 to 24, while area increases from 16 to 32. How would you respond
to this student?
These questions are in order of increasing depth. The first two involve basic issues
of place-value decimal notation. The third involves rational numbers and also
involves division, the most difficult of the arithmetic operations. It further requires
“modeling” or “representation”—connecting a calculation with a “real-world”
situation. The last problem, which was originally stated in terms of perimeter and
area of a “closed figure,” potentially involves very deep issues. Even if one replaces
“closed figure” with “rectangle,” as all the teachers did, one must still compare the
behavior of two functions of two real variables.
On sheepskin the American teachers seemed decidedly superior to the Chinese:
they all were college graduates, and several had MAs. The Chinese teachers had
nine years of regular schooling, and then three years of normal school for teachers—
in terms of study time, a high school degree. However, measured in terms of mastery
of elementary school mathematics, the Chinese teachers came out better.
The rough summary of the results of the interviews is: the Chinese teachers
responded more or less as one would hope that a mathematics teacher would, while
the American teachers revealed disturbing deficiencies. In more detail, on the first
two problems, all teachers could perform the calculations correctly and could
explain how to do them, that is, describe the correct procedure. However, even on
the first problem, fewer than 20% of the U.S. teachers had a conceptual grasp of
the regrouping process—decomposing one 10 into 10 ones. By contrast, the
Chinese teachers overwhelmingly (86%) understood and could explain this decom-
position procedure. On the second problem, about 40% of the U.S. teachers could
explain the reason for the correct method of aligning the partial products, while over
90% of the Chinese teachers showed a firm grasp of the place value considerations
that prescribe the alignment procedure.
On the third problem, a gap appeared even at the computational level: well
under half of the American teachers performed the indicated calculation correctly.
Only one came up with a technically acceptable story problem. Even this one was
pedagogically questionable, since the units for the answer (3 1/2) was persons,
which children might expect to come in whole numbers. The Chinese teachers
again all did the calculation correctly, and 90% of them could make up a correct
story problem. Some suggested multiple problems, illustrating different inter-
pretations of division.
On the fourth problem, the U.S. teachers did exhibit some good teaching
instincts, and most, though not all, could state the formulas for area and perimeter
of rectangles. However, when it came to analyzing the mathematics, they were
lost at sea. Although most wanted to see more examples, over 90% were inclined
to believe that the student’s claim was valid. Some proposed to look something
up in a book. Only three attempted a mathematical investigation of the claim, and
again a lone one found a counterexample. The Chinese teachers also found this
problem challenging, and most had to think about it for some time. After consid-
eration, 70% of them arrived at a correct understanding, with valid counterex-
amples. Of the 30% who did not find the answer, most did think mathematically
about the problem, though not sufficiently rigorously to find the defect in the
student’s proposal.
The contrast between the performances of the two groups of teachers was even
more dramatic than this summary reveals. Some Chinese teachers gave responses
that more than answered the question. They sometimes offered multiple solution
methods. In the integer arithmetic problems, some indicated that, if the student was
having trouble here, it meant that something more fundamental had not been
learned properly. These comments point to a deeper layer of teaching culture that
simply does not exist in the U.S. For example, American teaching of two-digit
subtraction is usually based on “subtraction facts,” the results of subtracting a one-
digit number from a one- or two-digit number to get a one-digit number. These are
simply to be learned by rote. The Chinese base subtraction on these same facts, but
they refer to this topic as “subtraction within 20” and treat it as one to be under-
stood thoroughly, since they regard it as the link between the computational and
the conceptual basis for multidigit subtraction. In answering question 3, some
Chinese teachers suggested that the given problem was too easy and offered harder
ones. Also, the Chinese teachers were comfortable with the algebra that is implic-
itly involved in performing arithmetic with our standard decimal notation—for
example, many explicitly invoked the distributive law when discussing multidigit
multiplication. No such awareness of the algebraic backbone of arithmetic was
shown by the American teachers.
In these first four chapters, KTEM also discusses issues of teaching methods.
Without going into detail about this, I will report that the same limitations that
teachers showed in giving a conventional explanation of a topic also prevented them
from getting to the conceptual heart of the issue when using teaching aids such as
manipulatives.
Thus, KTEM suggests that Chinese teachers have a much better grasp of the
mathematics they teach than do American teachers. The hard-nosed might ask for
evidence that this extra expertise actually produces better learning. Since Ma’s work
did not extend to a simultaneous study of the students of the teachers, KTEM cannot
address this question. However, the substantial studies of Stevenson and Stigler [SS]
do document superior mathematics achievement in China. (The Stevenson-Stigler
project provided part of the motivation for Ma’s work.) KTEM itself also provides
some evidence of superior learning in China and of a sort directly related to the
knowledge of teachers, as indicated in the interviews. The four interview questions
were presented to a group of Chinese ninth-grade students from an unremarkable
school in Shanghai. They all (with one quite minor lapse) could do all the calcu-
lations correctly and knew the perimeter and area formulas for rectangles. Over 60%
Roger Howe 583
found a counterexample to the student’s claim about area and perimeter, and over
40% could make up a story problem for the division of fractions in question 3. These
Chinese ninth-grade students demonstrated better understanding of the interview
problems than did the American teachers.
One should also entertain the possibility that Ma was overly optimistic in judging
her group of American teachers to be “above average.” However, this rating is
broadly consistent with evidence from a much larger set of interviews conducted
by Deborah Ball [B1–3] and also with the study [PHBL] of over two hundred
teachers in the Midwest. In that study, for example, only slightly over half the
subjects could provide an example of a number between 3.1 and 3.11. The portion
of satisfactory responses to questions testing pedagogical competence was consid-
erably smaller. The results of KTEM are also consistent with massive informal testi-
mony from serious workers in professional development for teachers. The remark-
able thing is that this problem—the failure of our system to produce teachers with
strong subject matter knowledge and the negative impact of this failure—is not more
explicitly recognized. Furthermore, solving this problem is not a major focus of
mathematical education research and of education policy. I hope that KTEM will
provide impetus for making it so.
KTEM gives us new perspectives on the problems involved in improving math-
ematics education in the U.S. For example, it strongly suggests that without a radical
change in the state of mathematical preparedness of the American teaching corps,
calls for teaching with or for “understanding,” such as those contained in the
NCTM Standards, are simply doomed. To the extent that they divert attention from
the crucial factor of teacher preparedness, they may well be counterproductive.
KTEM also indicates that claims that the traditional curriculum failed are misdi-
rected. The traditional curriculum allowed millions of people to be taught reliable
procedures for finding correct answers to important problems, without either the
teachers or the students having to understand why the procedures worked. At the
same time, students with high mathematical aptitude could learn substantially
more mathematics, enough to support various technical or academic careers. This
has to be counted a major success.
However, times have changed. The success of the traditional curriculum has
fostered a mathematically based technology, which in turn has created conditions
in which that curriculum is no longer appropriate. There are at least two reasons
for this. First, we have cheap calculators that will do (at least approximately) any
calculation of the elementary curriculum (and much more) with the push of a couple
of buttons. These machines are typically much faster and more reliable than we are
in doing these calculations. We also have “computer algebra” systems that will do
more kinds of calculations than any single human knows how to do. It has always
been one of the strengths of mathematics to seek reliable and systematic methods
of computation, which has often meant creating algorithms. Anything that has been
algorithmized can be done by a computer. Automation of calculation means that
actually performing a calculation is no longer a problem working people usually
have to worry about.
At the same time, it means that calculation is much more prevalent than before.
Hence, people have to spend more time determining what calculation to do. That
is the second reason that mathematics education needs to change. My daughter was
a solid mathematics student but had no enthusiasm for the subject and did not expect
to use it in whatever career she might choose. Now she works in management
consulting, and she finds that her high school algebra comes in handy in creating
spreadsheets. Simply learning computational procedures without understanding
them will not develop the ability to reason about what sort of calculations are
needed. In short, to function at work, people now need more understanding and less
procedural virtuosity than they did a generation ago. (Who knows what they will
need in another generation!)
The good news from KTEM is that there is no serious conflict between proce-
dural knowledge and conceptual knowledge: Chinese teachers seem to be able to
develop both in their students. (This is another intuition of most mathematicians I
know who have been studying educational issues: it should be the case that proce-
dural ability and conceptual understanding support each other. The Chinese teachers
had a traditional saying to describe this learning goal: “Know how, and also know
why.”) The bad news is that our current teaching corps is not capable of delivering
this kind of double understanding: we can only reasonably ask them for procedural
facility. Let us be clear that this is not a matter of teachers lacking certification or
teaching outside their specialty, which are both frequent problems that aggravate
the situation. The certification procedures, the teaching methods courses, most
college mathematics courses, the recruitment processes, the conditions of employ-
ment, most current teacher development—none of these is geared to ensuring that
U.S. mathematics teachers have themselves the understanding needed to teach for
understanding. In short, virtually the whole American K–12 mathematics educa-
tion enterprise is out of date.
How might the U.S. create a teaching corps with capabilities more like those of
the Chinese teachers? To begin to answer, we should try to be precise as to what
the differences are between the two groups. From the evidence of KTEM, I would
list three salient differences:
1. Chinese teachers receive better early training—good training produces good
trainers, in a virtuous cycle.
2. Chinese mathematics teachers are specialists. Making mathematics teaching
a specialty can be expected to increase the mathematical aptitude of the teaching
corps in two ways: it reduces the manpower requirements for mathematics educa-
tion by concentrating it in the hands of the mathematically most qualified teachers,
and it raises the incentives for mathematically inclined people to become teachers.
Beyond its recruitment implications, it means that Chinese teachers have more time
and motivation for developing their understanding of mathematics. This self-
improvement is amplified by a social effect: specialization creates a corps of
colleagues who can work together to deepen the common teaching culture in
mathematics. Thus, making mathematics teaching a specialty works in multiple
ways to increase the quality of mathematics education.
Roger Howe 585
3. Chinese teachers have working conditions which favor maturation of under-

standing. U.S. teachers spend virtually their whole day in front of a class, while
the Chinese teachers have time during the school day to study their teaching mate-
rials, to work with students who need or merit special attention, and to interact with
colleagues. New teachers can learn from more experienced ones. All can study
together the key aspects of individual lessons, an activity they engage in system-
atically. They can also sharpen their skills by discussing mathematical problems.
Stevenson and Stigler [SS] have observed that time for self-development is a
general feature of mathematics education in East Asia, which, to go by TIMSS
[DoEd1–3] as well as [SS], has the most successful systems of mathematics educa-
tion in the world today.
The combination of training, recruitment, and job conditions that prevails in
China helps produce a level of teaching excellence that Ma calls PUFM, “profound
understanding of fundamental mathematics.” PUFM and how it is attained is the
concern of Chapters 5 and 6. It is important to understand that PUFM involves more
than subject matter expertise, vital as that is; it also involves how to communicate
that subject matter to students. Education involves two fundamental ingredients:
subject matter and students. Teaching is the art of getting the students to learn the
subject matter. Doing this successfully requires excellent understanding of both.
As simple and obvious as this proposition may seem, it is often forgotten in discus-
sions of mathematics education in the U.S., and one of the two core ingredients is
emphasized over the other. In K–12 education the tendency is to emphasize
knowing students over knowing subject matter, while at the university level the
emphasis is frequently the opposite. (This cultural difference may well be part of
the reason some university mathematicians have reacted negatively to the NCTM
Standards. The emphasis on teaching methods over subject matter is prominent
in the recommendations and “vignettes” of this document.) Both these views of
teaching are incomplete.
What educational policies in the U.S might promote the development of a
teaching corps in which PUFM were, if not commonplace, at least not extremely
rare? This question is discussed in Chapter 7, the final chapter of KTEM. I would
like to add my own perspective on the issue. The differences (1), (2), and (3) listed
above suggest part of the answer.
Differences (2) and (3) are primarily matters of educational policy. No revolu-
tion in American habits is required to create mathematics specialists or to give them
opportunity for study and collegial interaction. What is mainly required is polit-
ical will.
Regarding difference (2), the manpower considerations which favor mathematics
specialists beginning in the early grades are much stronger in the U.S. than in
China. The U.S information society has much higher demand for mathematically
able people than does the predominantly rural economy of China. Hence, schools
face much heavier competition for mathematically competent personnel, and every
policy that could lower their manpower requirements or improve their competitive
position would benefit mathematics education. The difference in technological

level also makes the need for coherent mathematics education greater in the U.S.
than in China. Simply partitioning the present cadre of elementary teachers into math
specialists and nonmath would already offer the average child a better-qualified
(elementary) math teacher while relieving many others of what is now an onerous
duty, all without raising overall personnel requirements. Some educators have for
some time been calling for mathematics specialists even in the elementary grades
[Us]. Perhaps the evidence from KTEM that having teachers who understand math-
ematics can make a difference already in the second grade (the usual time for two-
digit subtraction) can convince education policymakers to heed this call.
Regarding difference (3), testimony from interviews of teachers with PUFM indi-
cates that having time for study and collegial interaction is an important factor in
developing PUFM. Such time would be most productive in the context of mathe-
matics specialists—both study and discussion would be more focused on mathe-
matics. Scheduling this time might be more controversial than creating specialists
because it requires resources. In fact, in East Asia classes are larger than here, so
a given teacher there handles about the same number of students as does a teacher
in the U.S.[SS]. The improvement in lessons promoted by study and interaction
with colleagues seems to more than make up for larger class size. There is currently
in the U.S. a call to reduce class size. On the evidence of KTEM and [SS], I believe
that the resources required for such a change would be better spent in eliminating
difference (3).
What will be hardest is eliminating difference (1), that is, establishing in the U.S.
the virtuous cycle, in which students would already graduate from ninth grade or
from high school with a solid conceptual understanding of mathematics, a strong
base on which to build teaching excellence. I expect that movement in that direc-
tion will, at least at the start, require massive intervention from higher education.
New professional development programs, both preservice and in-service, that
focus sharply on fostering deep understanding of elementary mathematics in a
teaching context will need to be created on a large scale. Current university math-
ematics courses will not serve; as KTEM makes clear, the needs of teachers at
present are of a completely different nature from the needs of professional math-
ematicians or technical users of mathematics, for whom almost all current offer-
ings were designed.
I would recommend that these programs be joint efforts of education departments
and mathematics departments to guarantee that the two poles of teaching, the
subject matter and the pedagogy, both get emphasized. These departments have
rather different cultures, and developing productive working relationships will not
be a simple task; but with sufficient backing from policymakers who understand
the current purposes and needs of mathematics education and the shortfall between
current capabilities and these needs, some beneficial programs should emerge.
While the greatest need for improvement is probably at the elementary level,
middle school and secondary teachers should not be neglected in the new profes-
Roger Howe 587
sional development programs. Undoubtedly they know more mathematics than the
typical elementary school teacher, but they too must have suffered from the lack
of attention to understanding during their early education. Moreover, they need to
deal with a larger body of material than do elementary teachers.
There is also the issue of texts. The Chinese teachers have materials, texts, and
teaching guides that support their self-study. American texts tend to be lavishly
produced but disjointed in presentation [Sc, DoEd1-3], and the teacher’s guides do
not help much either. Thus, the intervention programs should also work to create
materials which will help teachers both learn and transmit a coherent view of math-
ematics. Eventually, these might be the basis for new texts.
At least at the start, these programs should be multiyear in scope, both so that
teachers who do not have the favorable working conditions of Chinese teachers can
nevertheless refresh and progressively improve their understanding of mathe-
matics and so that those teachers who do obtain such working conditions can get
to the level where self-directed study can be a reliable mode of improvement. One
of the most outmoded ideas in education is that a teacher can reasonably be
expected to know all that he or she needs to know, of subject matter or teaching,
at the start of work. Continued study, especially of subject matter, since teaching
itself will provide plenty of opportunities for learning about children, should
become the norm. If a program of this sort is implemented successfully, it should
gradually become less necessary. The step-by-step improvement in education
provided by teachers with better understanding and the gradual deepening of
teaching culture by teachers interacting collegially among themselves should
allow elaborate development programs to shrink and eventually disappear or to shift
to study of more sophisticated topics, becoming, in subject matter at least, more
like standard college mathematics courses. This would constitute truly satisfying
progress in our system of mathematics education. However, it will require great
effort and resolve to achieve.
In summary, KTEM has lessons for all educational policymakers. Legislators,
departments of education, and school boards need to understand the potential
value in creating a corps of elementary-grade mathematics specialists who have
scheduled time for study and collegial interaction. University educators need to
understand teacher training in mathematics as a distinct activity, different from but
of comparable value to training scientists, engineers, or generalist teachers. I
believe that these mutually supportive changes would give us a fighting chance for
successful mathematics education reform.
Getting the Mathematics to the Students

Ma’s notion of “profound understanding of fundamental mathematics
(PUFM),” involves both expertise in mathematics and an understanding of how
to communicate with students. Teacher Mao, one of the teachers Ma identified
as possessing PUFM, eloquently expressed the need for both types of under-
standing:
I always spend more time on preparing a class than on teaching, sometimes

three, even four times the latter. I spend the time in studying the teaching mate-
rials; what is it that I am going to teach in this lesson? How should I introduce
the topic? What concepts or skills have the students learned that I should draw
on? Is it a key piece on which other pieces of knowledge will build, or is it built
on other knowledge? If it is a key piece of knowledge, how can I teach it so
students grasp it solidly enough to support their later learning? If it is not a
key piece, what is the concept or the procedure it is built on? How am I going
to pull out that knowledge and make sure my students are aware of it and the
relation between the old knowledge and the new topic? What kind of review
will my students need? How should I present the topic step-by-step? How will
students respond after I raise a certain question? Where should I explain it at
length, and where should I leave it to students to learn it by themselves? What
are the topics that the students will learn which are built directly or indirectly
on this topic? How can my lesson set a basis for their learning of the next topic,
and for related topics that they will learn in their future? What do I expect the
advanced students to learn from this lesson? What do I expect the slow students
to learn? How can I reach these goals? etc. In a word, one thing is to study
whom you are teaching, the other thing is to study the knowledge you are
teaching. If you can interweave the two things together nicely, you will succeed.
We think about these two things over and over in studying teaching materials.
Believe me, it seems to be simple when I talk about it, but when you really do
it, it is very complicated, subtle, and takes a lot of time. It is easy to be an elemen-
tary school teacher, but it is difficult to be a good elementary school teacher.
I would like to highlight the concern in Teacher Wang’s statement for the
connectedness of mathematics, the desire to make sure that students see math-
ematics as a coherent whole. This is certainly how mathematicians see it, and
to us it is one of the major attractions of the field: mathematics makes sense and
helps us make sense of the world. For me, perhaps the most discouraging aspect
of working on K–12 educational issues has been confronting the fact that most
Americans see mathematics as an arbitrary set of rules with no relation to one
another or to other parts of life. Many teachers share this view. A teacher who
is blind to the coherence of mathematics cannot help students see it.
—R. H.
REFERENCES
[B1] D. BALL, The Subject Matter Preparation of Prospective Teachers: Challenging the Myths, Na-
tional Center for Research in Teacher Education, East Lansing, MI, 1988.
[B2] ———, Teaching Mathematics for Understanding: What Do Teachers Need to Know about the
Subject Matter?, National Center for Research in Teacher Education, East Lansing, MI, 1989.
[B3] ———, Prospective elementary and secondary teachers’ understanding of division, J. Res. Math.
Ed., 21 (1990), 132–144.
Roger Howe 589
[Be] E. BEGLE, Critical variables in mathematics education: Findings from a survey of empirical lit-
erature, MAA and NCTM, Washington, DC, 1979.
[DoEd1] U.S. Department of Education, Pursuing excellence: A study of U.S. fourth-grade mathematics
and science achievement in an international context, U.S. Government Printing Office, Washing-
ton, D.C., 1997.
[DoEd2] U.S. Department of Education. Pursuing excellence: A study of U.S. eighth grade mathematics
and science teaching, learning, curriculum, and achievement in an international context, U.S. Gov-
ernment Printing Office, Washington, D.C., 1996.
[DoEd3] U.S. Department of Education, Pursuing excellence: A study of U.S. twelfth-grade mathe-
matics and science achievement in an international context, U.S. Government Printing Office,
Washington, D.C., 1998.
[Hi] E. D. HIRSCH, The Schools We Need and Why We Don’t Have Them, Doubleday, New York, 1996.
[Ho] R. HOWE, The AMS and mathematics education: The revision of the “NCTM Standards”, No-
tices of the AMS 45 (1998), 243–247.
[NCTM] Curriculum and Evaluation Standards for School Mathematics, National Council of Teach-
ers of Mathematics, Reston, VA, 1989.
[PHBL] T. POST, G. HAREL, M. BEHR, and R. LESH, Intermediate teachers’ knowledge of rational num-
ber concepts, E. Fennema, T. Carpenter, and S. Lamon (eds.), Integrating Research on Teaching
and Learning Mathematics, SUNY, Albany, NY, 1991, pp. 177–198.
[Sc] W. SCHMIDT et al., A Summary of Facing the Consequences: Using TIMSS for a Closer Look at
United States Mathematics and Science Education, Kluwer Academic Publishers, Dordrecht, 1998.
[SS] H. STEVENSON and J. STIGLER, The Learning Gap, Why Our Schools Are Failing and What We
Can Learn from Japanese and Chinese Education, Paperback Reprint edition, Touchstone Books,
January 1994.
[Us] Z. USISKIN, The beliefs underlying UCSMP, UCSMP Newsletter, no. 2 (Winter 1988).

Vol 30 Issue 5

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vol 30 Issue 5

Uploaded by

Copyright:

Available Formats

Journal for Research in Mathematics Education

1999, Vol. 30, No. 5, 520–540

A Meta-Analysis of the Relationship

In this meta-analysis I examined 26 studies on the relationship between anxiety

achievement individually but not necessarily collectively. For example, a student’s

Coding of Selected Studies

Meta-analytic methodology literature is not explicit on the use of longitudinal

Characteristics of the Sample

The U3 statistic corresponding to a population correlation of –.27 is .71. To under-

Effects of Grade Level

Effects of Instrument Measuring Mathematics Anxiety

Effects of Instrument Measuring Mathematics Achievement

Effects of Type of Publication

Effects of Year of Publication

The Final Model

Theoretical and Practical Implications

Comments on the Instruments

Unfolding the Anxiety-Achievement Dynamic in Mathematics

The well-documented national differences in mathematics achievement (Beaton

In this study we examined one component of problem-solving competency for

When trying to make sense of national achievement differences in mathematics

propositional statements. In the integration phase, the problem solver creates a

FLEXIBILITY WITH REPRESENTATIONS

showed corresponding competence with other representations related to the same

(Asterisks indicate significant sample differences at .05 level of significance.)

*B3 1 34 + 2 12 = A3 Given the problem below, decide if

*B4 A4 (The same figure is given.)

No solution item. A6 There are different ways of representing

*b. 2(3) + 2(3) + 2(3) + 2(3)

Scores and Rank Orders on Solution Items

Scores and Rank Orders on the Representational Items

ANOVA on Individual Items

Flexibility Within a Representation

Flexibility Across Formal Symbolisms

DISCUSSION AND SUMMARY

in understanding the relation of fractions to proportions. Such part-whole repre-

The Meaning of Randomness for

In the experimental study reported here we intended to examine possible differences

Randomness has been interpreted in various ways during different periods in

pattern would be its essential characteristic. Each of these characterizations includes

AIMS AND METHODOLOGY OF THE EXPERIMENTAL RESEARCH

purpose was to contribute to the investigation of the evolution of probabilistic intu-

Introductory activity: Paul plays a game using 16 counters numbered 1, 2, 3, 4,

Figure 2. Introductory activity and Items 5 to 8.

In a theoretical random distribution of the number of counters, about 6 empty

RESULTS AND DISCUSSION

The percentages of students considering each item to be random are presented in

Most students considered all Bernoulli sequences (Items 1 to 4) to be random

seemed to concentrate on the differences between the theoretical and observed

Analysis of Students’ Arguments

We are currently standing at one of the most

CELEBRATE OUR SUCCESS

As an organization, we have had a substantial

Figure 1. The overlapping-squares problem.

Figure 2. The quadrilateral problem.

Figure 3. A real-world problem to the students.

Mathematics should be considered as an experimental science. For students to

Figure 4. Mystery missing-teeth data.

STAY THE COURSE

• More than 1836 teachers are currently Board certified.

Figure 5. Facts about National Board Certification.

TIME: CHANGING THE SOCIETAL DEMANDS ON SCHOOLS

Another continuing challenge is textbook materials. Although we have made