From Scanner to Sound Bite:Issues in Interpreting and Reporting Sex Differences in the Brain HTML version

Sex Differences in the Brain
researcher will find a statistically significant difference. The
concern is that, given that the publication process is geared
toward emphasizing difference rather than similarity, this
1-in-20 finding of difference will be reported while the 19 fail-
ures to find a difference will not. Indeed, Kaiser et al. provide
examples of how, in language research, even marginal sex
differences in the brain are given prominence in the published
The problem of spurious results for all sex-differences
research has been long-noted. However, the inevitable teething
problems of the new neuroimaging technologies may exacer-
bate the situation. Nuisance variables like breathing rate and
caffeine intake can influence the imaging signal and give rise
to spurious results; this is particularly the case when sample
sizes are small, as they often are in the studies that have
reported sex differences (Wallentin, 2009). Researchers have
found that sex differences in language lateralization fail to
generalize to a distinct but similar task within a second group
of men and women and that identical analyses of the same par-
ticipants can also ‘‘discover’’ brain-activation differences
between randomly created groups matched on sex, perfor-
mance, and obvious demographic characteristics (Ihnen,
Church, Petersen, & Schlaggar, 2009). As controversies over
statistical procedures are resolved, researchers may turn out
to have used inadequate or inappropriate techniques: It has
recently been argued that some reported sex differences in
language lateralization have not been put to adequate
statistical test or that they can come and go depending on how
the analysis is done (Kaiser et al., 2009). There is also an
as-yet-unresolved controversy regarding how—or even
whether—neuroscientists should control for the sex difference
in average brain size when attempting to establish whether the
sexes differ in the volumes of particular brain regions (see, for
example, Fausto-Sterling, 2000). A change in methodology can
transform what appears to be a sex difference into a difference
between people with smaller and larger brains (e.g., Im et al.,
These difficulties point to the importance of not placing too
much confidence in any single functional or structural neuro-
imaging study that seems to demonstrate a sex difference. The
wisdom of such caution is demonstrated by meta-analyses of
tests of the GML hypothesis. A recent meta-analysis of 26 func-
tional neuroimaging studies of language lateralization found no
significant sex difference (Sommer, Aleman, Somers, Boks, &
Kahn, 2008). Similarly, a meta-analysis of 49 postmortem and
structural neuroimaging studies of the corpus callosum found
no support for the hypothesis that this structure is larger in
females, even allowing for their relatively smaller brains
(Bishop & Wahlsten, 1997).
The failure of meta-analyses to support predictions of the
GML hypothesis, to which much research attention has been
devoted, highlights the importance of remaining skeptical
about other reported sex differences. Clearly, isolated reports
of sex differences in brain activation or regional brain volume
require replication and generalization before they can be
assumed to be reliable.
Interpretation: The Obscurity of
Structure–Function Relations
Further difficulties arise when it comes to understanding what,
if anything, brain differences might imply for psychological
function. Despite the extraordinary progress made in neu-
roscience, we still have minimal understanding of how neural
structures contribute to complex psychological phenomena.
Again, the GML hypothesis is an instructive example: There
is no a priori reason to think that a more lateralized brain would
be advantageous for visuospatial processing but disadvanta-
geous for language function. (Nor, to my knowledge, have any
such relations been demonstrated.) Bishop and Wahlsten
(1997) questioned the assumption that a modest size difference
in a structure as complex and massively interconnected as the
corpus callosum would have tangible implications for a spe-
cific psychological construct.
Nonetheless, functional speculations arising from the GML
hypothesis have recently taken new form in suggestions that a
male brain skewed toward more lateralized, intrahemispheric
processing may be advantageous for scientific disciplines that
supposedly require focused scrutiny of details rather than
integration of information (Baron-Cohen, Knickmeyer, &
Belmonte, 2005; Gur & Gur, 2007), and that a female brain
skewed toward more long-range processing may be advanta-
geous for empathizing (Baron-Cohen et al., 2005).
While speculation is an important part of the scientific pro-
cess, as Fausto-Sterling (2000, p. 118) has observed, the prob-
lem is that ‘‘despite the many recent insights of brain research,
this organ remains a vast unknown, a perfect medium on which
to project, even unwittingly, assumptions about gender.’’ As
noted earlier, it was once readily assumed that brain weight
correlated with intelligence, thus explaining women’s sup-
posed intellectual inferiority. With that history in mind, we
should be wary of suggestions that the typical female brain is
suboptimally designed for currently male-dominated pursuits
like science—suggestions made in the absence of adequate
knowledge of how the brain enables scientific thinking and
practice. There is no neuroscientific reason, for example, to
think that the shorter circuits of an intrahemispheric brain will
enable narrower focus in the mind.
Inferring a mental process from significant activation in a par-
ticular brain region (for example, inferring that the amygdala was
activated, therefore participants were anxious) is known as
reverse inference and is also fraught with difficulty. The statisti-
cal procedures of functional neuroimaging identify regions that
are differentially activated by the experimental task, compared
with a control task. However, while brain function involves spe-
cialization—the entire brain is not involved in all of its func-
tions—there is no simple one-to-one mapping between brain
regions and psychological processes.Mental processes arise from
the complex interaction of multiple areas, and any one regionwill
be involved in any number of mental processes. The anterior cin-
gulate, for example, is activated by so many tasks that one cogni-
tive neuroscientist known to the author refers to this region as
‘‘the on button’’ (Geoffrey Boynton, personal communication).
Downloaded from
at University of Melbourne Library on October 12, 2010