You are on page 1of 8

P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ

0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

Gender Differences in Math


Cognitive Processes in an Expanded Framework

James P. Byrnes

Scientists and nonscientists alike construct theories to explain variations in


the environment (Byrnes, 2001a). For example, zoologists devise theories
to explain observable differences in the physical appearance of species,
and developmental scientists create theories to explain changes in perfor-
mance that occur between early childhood and adulthood. The authors
in this book are chiefly concerned with variations in math performance
that are evident when one compares boys with girls or men with women.
Explaining these gender-based variations is not an easy task because the
size and direction of differences change with age, content, measure, and
context (Hyde, Fennema, & Lamon, 1990). The purpose of this chapter is to
present a comprehensive account of gender differences that explains most
of the variations that have been revealed to date. As I will argue later, the
primary virtue of such an account is that it could form the basis of highly
effective forms of intervention.
The rest of this chapter is organized as follows. In the first section, I
provide a brief overview of the pattern of gender differences that have
been reported in the literature. This pattern represents the phenomenon
that needs to be explained by any theory of gender differences. In the
second section, I summarize and critique existing explanations of these
findings (including a Cognitive Process approach that my colleagues and I
proposed in the mid-1990s). In the third section, I present a new explanatory
model that was created to extend and integrate the existing explanations
(called the Three Conditions model). In the final section, I consider the
theoretical, empirical, and practical implications of the Three Conditions
model.

patterns of gender differences in math


Before considering the merits of existing theories of gender differences in
math, it is first important to consider the phenomena that these theories
73
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

74 James P. Byrnes

are trying to explain. In my view, a theory gains both scientific credibility


and practical value to the extent that it can explain an increasing array
of related phenomena. For example, consider the following three findings
from the literature on gender differences in math:

1. High school males perform much better than high school females on
the SAT-math.
2. Prior to high school, gender differences are usually not found on
measures of math problem solving.
3. Boys rate their math competence higher than girls in both elementary
school and high school (Hyde et al., 1990; Wigfield, Eccles, Yoon,
Harold, Arbreton, Freedman-Doan, & Blumenfeld, 1997).

In my view, theories that can explain all three of these findings are prefer-
able to theories that can explain only one or two of them for the following
reason: important developmental outcomes are nearly always produced
by the confluence of multiple factors that work in concert. If so, then in-
terventions must necessarily target multiple factors to have any hope of
being effective. To know which factors to target, however, one must have an
accurate and comprehensive theory of the outcome in question. Theories,
after all, identify the causal factors responsible for an outcome.
With this argument in mind, we can consider the primary findings in
the literature on gender differences in math. For more detailed summaries
of these and other findings, see Byrnes (2001a), Halpern (2000), or Hyde
et al. (1990):

1. Below the age of 15, girls tend to perform better than boys on tests
requiring computational skill (e.g., Green, 1987; Newman, 1984). No dif-
ferences are found for tests that measure math concepts or math problem
solving (e.g., Fennema & Sherman, 1978; Lewis & Hoover, 1987). Among
gifted 7th graders, however, a moderate difference favoring boys (d = 0.41)
appears for overall scores on SAT-like tests (e.g., Benbow & Stanley, 1980).
On SAT items that require one to know when one has enough information
to provide an answer, however, gifted girls perform better than gifted boys
(e.g., Becker, 1990).
2. At around age 15, a small to moderate gender difference appears for
problem solving (d = 0.29, favoring boys), especially when college-bound
students are at issue (Hyde et al., 1990). Moderate differences also emerge
on timed standardized tests such as the SAT-math (d = 0.40) and the GRE-
Quantitative (d = 0.70). Both of these findings illustrate how the gender
gap is larger in select samples than in the general population.
3. No gender differences emerge for students older than 14 for measures
of math concepts or computations, however (Hyde et al., 1990). In addi-
tion, male and female 12th graders have obtained nearly identical scores
on the last four National Assessments of Educational Progress (NAEPs)
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

Gender Differences in Math 75

for math (e.g., means of 299 and 303 on the 2000 NAEP). The latter involve
large, nationally representative samples (N > 17,000) and assess compu-
tational and problem-solving skills in algebra, geometry, probability, and
arithmetic. Students are given 45 minutes to solve 40 to 60 items.
4. Girls tend to obtain either the same grade as, or higher grades than,
boys in mathematics at all grade levels (ds ranging from −0.09 to −0.35;
Kimball, 1989).
5. Boys tend to give higher ratings of their math ability than girls at all
grade levels (Wigfield et al., 1997).

In essence, then, theories of gender differences have to answer questions


such as (1) why does the computational advantage of girls exist prior to
age 15 but disappear after age 15?; (2) why does the gap in problem solving
grow over time (i.e., nonexistent in most ability groups prior to age 15 but
apparent after age 15?); (3) why do boys rate their abilities higher than
girls even at ages when they fail to demonstrate more ability?; (4) why
do girls demonstrate similar or better performance on various indicators
of achievement at all grade levels (e.g., grades in courses or scores on the
NAEP), whereas boys demonstrate better performance on SAT-like tests
mainly after age 15?

existing theories and their shortcomings


Various theories have been proposed over the years to explain some of
these findings. In what follows, four representative theories are briefly
explained and critiqued in turn. As noted above, more detailed treatments
of these theories can be found in sources such as Byrnes (2001a) or Halpern
(2000).

Genetic and Other Physiological Views


One way to explain gender differences in math is to appeal to brain-based
differences between males and females (e.g., Benbow, 1988). To be consis-
tent with what we currently know about the brain (and to also be suffi-
ciently precise and plausible), this perspective would have to consist of the
following component propositions:

1. In the brain, there are specific networks of neural assemblies ded-


icated to particular cognitive processes (e.g., comparing the magnitude
of two numbers). When someone engages in a given cognitive process,
the network dedicated to this process tends to be active (around 70% of
the time). Networks dedicated to other, unrelated, or uninvolved pro-
cesses tend not to be active. The locations of various networks can be
observed using technologies such as functional magnetic resonance imag-
ing. These technologies show that all major cognitive processes (including
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

76 James P. Byrnes

math) activate widely distributed assemblies from all four lobes and both
cerebral hemispheres of the brain.
2. There are specific, still-to-be-determined configurations of neural as-
semblies within these networks that lead to optimum mathematics perfor-
mance (e.g., problem types are quickly recognized, relevant knowledge is
accessed quickly). Other configurations lead to average or below average
performance. Presumably, the key differences among these configurations
have to do with such things as the total number of neurons (too many or too
few can cause problems), the proportion of certain kinds of neurons in the
assembly (e.g., inhibitory vs. excitatory), which assemblies are recruited (in
what order), and patterns of interconnections within and among neurons
and neural assemblies.
3. Each configuration of neural assemblies is the end product of normal
processes of brain development (e.g., creating enough cells of different
types, cells migrating to the right locations after they are produced, cells
growing in size and projecting axons, preprogrammed and experience-
based synaptogenesis, cell death).
4. These normal processes of brain development operate somewhat dif-
ferently in males and females (according to advocates of this perspective).
As a result, (a) neural assemblies used during mathematical thinking tend
to be configured differently in males and females, and (b) males are more
likely than females to acquire the optimal configuration for assemblies
dedicated to math.

Whereas most researchers who advocate an important role for brain


physiology would probably agree with assertions 1 to 4 above, they often
disagree about the mechanisms that are responsible for achieving the pre-
sumed differences between male and female brains. For example, some
researchers appeal to genetics and suggest that male genes specify a dif-
ferent brain morphology than female genes. Several facts about brain de-
velopment are consistent with this claim. First, there is good evidence to
suggest that processes such as cell proliferation (i.e., creating brain cells),
cell migration, cell differentiation, cell growth, and initial, preexperience
synaptogenesis are largely determined by genetic instructions (although
local signaling among cells also affects these processes; see Byrnes, 2001b).
Second, male brains are about 9% larger than female brains, which suggests
that cell proliferation and growth operate differently in the two gender
groups (e.g., the proliferation phase lasts a little longer in males, giving
them more brain cells; cell growth also lasts longer, giving them larger
brain cells).
Other researchers agree that genes play a role in the process, but argue
that genes mainly contribute to the creation of hormone-producing organs
(e.g., gonads) and receptors for sex hormones in brain cells. Research with
rats shows that male and female hormones alter the morphology of brain
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

Gender Differences in Math 77

structures such as the hypothalamus and corpus callosum; therefore, some


researchers assume that prenatal exposure to sex hormones would proba-
bly also affect the growth and development of human brains.
To illustrate one variant of the hormonal perspective, consider the ac-
counts proposed by the research groups of Benbow, Casey, and others to
explain gender differences in math (e.g., Benbow, 1988; Casey, Pezaris, &
Nuttall, 1992). Researchers in these groups sought a physiological account
that would emphasize the right hemisphere because

The right hemisphere is traditionally considered specialized for non-verbal tasks


and the left for verbal, although these differences may not be qualitative but quanti-
tative. Mathematical reasoning ability, especially in contrast to computational abil-
ity, may be more strongly under the influence of the right hemisphere. (Benbow,
1988, p. 180.)

Some advocates of this perspective found what they needed in the work
of the late Norman Geschwind and colleagues on the anatomical basis of
dyslexia. Geschwind and Galabura (1985) proposed a model to explain
three sets of findings: (1) a higher incidence of language problems in boys
than in girls, (2) symmetry or reversed asymmetry in the size of certain
brain areas in dyslexic children, and (3) unexpected empirical links be-
tween left-handedness, language disorders, and immune disorders. To
explain these findings, Geschwind and colleagues proposed that, during
prenatal development, testosterone levels affect the growth of the left cere-
bral hemisphere in such a way that an anomalous form of dominance
develops. Instead of being right-handed and having language lateralized
in the left hemisphere, affected individuals become left-handed with lan-
guage lateralized in the right or both hemispheres. This altered physiology,
in turn, leads to problems such as developmental dyslexia, impaired lan-
guage development, and autism. Testosterone levels also affect the thymus,
resulting in disorders of the immune system (e.g., allergies, colitis). To ex-
plain asymmetries in the size of the left and right hemispheres that are
found in most normal individuals, Geschwind and colleagues suggested
the testosterone may either retard the growth of the left hemisphere, or
interfere with normal reductions in the right. With respect to the latter,
note that most children produce many more cells than needed prenatally.
If children avoid brain insults postnatally, excess cells are believed to be
eliminated through cell death (but this claim is somewhat controversial).
To test these speculations about right hemisphere development, Benbow
and colleagues considered whether gifted children were more likely than
nongifted children to be left-handed and have immune disorders (e.g.,
allergies). To assess handedness, Benbow and colleagues gave the Edin-
burgh Handedness Inventory (Oldfield, 1971) to two kinds of children
who were drawn from their sample of more than 100,000 gifted students:
(1) an extremely precocious group of seventh-grade children (N = 303)
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

78 James P. Byrnes

who scored above 700 on the SAT-math or above 630 on the SAT-verbal,
and (2) a less precocious group who scored closer to 500 on the SAT-math
(N = 127). Whereas the norms for Edinburgh Handedness Inventory sug-
gest that 8% of Scottish adults use their left hands occasionally or often
to perform everyday tasks, 13% of children who were extremely preco-
cious for math and 10% of the less precocious group were left-handed in
this way (Benbow, 1986). Whereas the incidence of left-handedness was
found to be significantly higher in the extremely precocious children than
in the Scottish adults (p < 0.04), two other comparisons revealed no signif-
icant differences: (1) less precocious children vs. the Scottish adults, and
(2) extremely precocious children vs. less precocious children.
As for gender differences in the extent of left-handedness in extremely
precocious students, Benbow (1988) reports that more males (16%) than
females (11%) were left-handed in the study (p < 0.05, using an unspecified
test). The present author, however, applied the standard test for comparing
frequencies (i.e., the chi-square test) to Benbow’s (1986) data and found that
the difference between 16% and 11% is not significant (p = 0.17). In addition,
the key difference between mathematically precocious males (14%) and
mathematically precocious females (6%) was also not significant (p = 0.33).
With respect to immune disorders, Benbow (1988) reports that students
with extremely high mathematical ability are twice as likely to have aller-
gies as children in the general population (53% vs. 25%). A comparison
between extremely precocious males (53%) and females (54%), however,
showed no significance difference in the incidence of allergies.
Thus, the preliminary findings based on handedness and allergies were
not terribly supportive of the idea of greater right hemisphere involvement
in gifted children, in general, and gifted males, in particular. It could be
argued, however, that these studies really do not test the right-hemisphere
proposal directly because indices such as handedness and allergies are
fairly imprecise. A more direct approach would be to look at patterns of
activation in the right and left hemisphere using either neuroimaging or
gross electrical recording techniques. In their review of the literature using
the latter, O’Boyle and Gill (1998) report that gifted adolescents appear to
engage their right hemispheres more than nongifted adolescents when they
listen to auditory stimuli or process facial expressions. In addition, gifted
adolescents show a pattern of resting neural activation that is similar to that
of college students and significantly different from nongifted adolescents
(i.e., greater activation in the frontal and occipital lobes).
Other studies conducted by Benbow and colleagues revealed gender
differences with respect to the involvement of the right hemisphere for the
processing of faces and mental rotation (more involvement for males), but
not for verbal stimuli. In addition, graphs of data presented in Alexander,
O’Boyle, and Benbow (1996) also suggest greater resting activations in the
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

Gender Differences in Math 79

parietal and possibly frontal lobes in gifted males than in gifted females, but
these specific comparisons were not reported in the text. The one study that
had the potential to consider whether greater right hemisphere involve-
ment was associated with higher SAT-math scores (i.e., O’Boyle & Benbow,
1990) failed to report this correlation because the authors expressed con-
cerns over a restricted range problem with the SAT-math scores (i.e., most
students scored over 500). The authors did report a correlation of r = −0.29
between laterality scores and total SAT scores (i.e., greater bias to process
faces in the right hemisphere corresponded to higher total SAT scores).
Whereas Benbow and colleagues suggested in their earlier work that
the right hemisphere is associated with math skill in some unspecified
way, Casey and colleagues suggest that the link has to do with spatial
ability (Casey, Nuttall, Pezaris, & Benbow, 1995). This idea seemed rea-
sonable because there are rather substantial gender differences in spatial
skills such as mental rotation (ds on the order of 0.7 to 1.5). Math is often
alleged to require spatial skills (e.g., to imagine solutions), so Casey and
colleagues wanted to see if (1) spatial ability would predict performance
on the SAT-math (controlling for other factors), and (2) gender differences
on the SAT-math would disappear once one controlled for spatial ability.
As for the first hypothesis, Casey et al. (1995) found that spatial ability
did predict SAT-math scores after one controlled for SAT-verbal scores
(an average of 9% of the variance in four female age groups and 8% in
four male age groups). However, verbal skills explained two to three times
as much variance as spatial skills (26% in females and 15% in males). In
another study of 8th graders, spatial ability only predicted math skills for
non-right-handed females. It did not predict math skills for right-handed
females, right-handed males, or left-handed males (Casey et al., 1992). As
for the second prediction, Casey et al. (1995) found that the significant gen-
der difference in SAT-math scores can be eliminated when one controls for
spatial ability. However, whereas such an effect was found for college stu-
dents and high-ability high school students, it was not found for precocious
students.
Besides the lack of strong or consistent support in the studies conducted
by advocates of the right hemisphere/spatial ability account, several other
problems also exist. First, other researchers have not found any correlation
between spatial scores and math performance (e.g., Fennema & Sherman,
1977). Second, some have argued that it is not clear that spatial skills would
even be required to solve SAT-like items where the gender differences are
largest (e.g., Royer, Tronsky, Chan, Jackson, & Marchant, 1999). Third, most
neuroscientists assume that the frontal lobes are the sites of higher-order
reasoning (Luria, 1973; Waltz et al., 1999). More posterior regions of the
right hemisphere could be associated with certain aspects of conceptual
knowledge in math or certain types of spatial reasoning (but not all), but
P1: GDZ/FFX P2: GDZ/FFX QC: GDZ/FFX T1: GDZ
0521826055c04 CB717-Gallagher-v2 July 2, 2004 14:47

80 James P. Byrnes

these regions are also active when working memory and attention are en-
gaged (Byrnes, 2001b). Thus, even if evidence suddenly did accumulate to
suggest that extremely talented mathematicians engage their right hemi-
spheres more than less talented individuals, this difference could reflect
the former’s greater reliance on math concepts, spatial skills, or working
memory. These capacities may relate to the kind of reasoning required to
do well on the SAT, but the core processes of problem comprehension and
strategic planning are likely to be associated with the frontal lobes.
Fourth, speculations about size differences in the right hemisphere be-
tween genders have not been borne out in neuroimaging studies or autop-
sies (Byrnes, 2001b). Whereas these studies have shown that the average
female brain tends to be 9% smaller overall than the average male brain
(as noted earlier), the difference is not limited to the right hemisphere.
Moreover, researchers have not found gender differences in the degree of
asymmetry of the left and right hemispheres.
Fifth, it is not at all clear why a theory designed to account for reading
disabilities (i.e., the model of Geschwind and colleagues) would even be
appropriate for explaining high levels of math talent. If the Geschwind
model really did apply, one would expect to find reading disabilities in
many of the extremely precocious children or people with high spatial
ability. In fact, however, most precocious children have a great deal of
verbal ability in addition to having considerable math ability (Benbow,
1986, 1988). Sixth, comprehensive meta-analyses of the literature reveal
that there is little evidence in support of the proposals of Geschwind and
colleagues, even where it is meant to apply (e.g., Bryden, McManus, &
Bulman-Fleming, 1994).
Seventh, many studies have shown that experience can alter brain mor-
phology (Byrnes, 2001b). Experiences can promote the growth of dendrites
and also the retraction or pruning of axons. So, if the genders are found
someday to differ in cytoarchitecture (number, types, and patterns of con-
nectivity among neurons), this difference could be due to genes, hormones,
or experience. Eighth, males and females may have the same morphology
but use different strategies when they solve tasks (Byrnes & Takahira, 1993;
Halpern, 2000). These strategies might show up in brain scans as different
regions of the brain being active when problem solving is underway, but
such a difference in activity does not mean that male and female brains
are “naturally” more or less lateralized (or better organized, etc.). If all stu-
dents were taught to use the same strategy, then the same regions of the
brain would probably be active in all students (given the correspondence
between networks of assemblies and particular tasks).
Finally, perhaps the biggest problem with existing physiological views
is that they do not provide a comprehensive story that can account for
the entire pattern of gender differences reported here. For example, if the
right hemisphere account is true, why are gender differences in problem

You might also like