Neuroethics

, Volume 6, Issue 2, pp 369–409

Is There Neurosexism in Functional Neuroimaging Investigations of Sex Differences?

Authors

    • Psychological SciencesUniversity of Melbourne
    • Centre for Ethical LeadershipMelbourne Business School
Original Paper

DOI: 10.1007/s12152-012-9169-1

Cite this article as:
Fine, C. Neuroethics (2013) 6: 369. doi:10.1007/s12152-012-9169-1

Abstract

The neuroscientific investigation of sex differences has an unsavoury past, in which scientific claims reinforced and legitimated gender roles in ways that were not scientifically justified. Feminist critics have recently argued that the current use of functional neuroimaging technology in sex differences research largely follows that tradition. These charges of ‘neurosexism’ have been countered with arguments that the research being done is informative and valuable and that an over-emphasis on the perils, rather than the promise, of such research threatens to hinder scientific progress. To investigate the validity of these contrasting concerns, recent functional magnetic resonance imaging (fMRI) investigations of sex differences and citation practices were systematically examined. In line with the notion of neurosexism, the research was found to support the influence of false-positive claims of sex differences in the brain, to enable the proliferation of untested, stereotype-consistent functional interpretations, and to pay insufficient attention to the potential plasticity of sex differences in both brain and mind. This, it is argued, creates a literature biased toward the presentation of sex differences in the brain as extensive, functionally significant, and fixed—and therefore implicitly supportive of a gender essentialist perspective. It is suggested that taking feminist criticisms into account would bring about substantial improvement in the quality of the science, as well as a reduction in socially harmful consequences.

Keywords

Sex/genderfMRIGender stereotypesPublication biasGender essentialismCitation bias

Introduction

A now notorious claim of Victorian sexual science was that men’s intellectual superiority could be explained by the larger male brain—a phenomenon described by one scientist as the “missing five ounces” of female brain [1, p. 23]. Developed under the assumption of a positive correlation between brain size and intelligence, it was only abandoned some time after the absence of such a relation became clear [see 2]. Historical examples of erroneous hypotheses regarding the neurological differences between the sexes and their functional implications are readily found [24], and it is not controversial to suggest that such scientific claims had political influence. Russett, for instance, has argued that the theories of the Victorian sexual scientists were not simply garnered as support by those who opposed greater social equality for the sexes, but were a “key source” of that opposition [2, p. 191].

In recent decades, a new research tool has been made available to those interested in neurological differences between the sexes: functional neuroimaging (FNI). While neuroscientists have long been able to compare structural features of the male and female brain, technologies such as functional magnetic resonance imaging (fMRI) and (now less commonly) positron emission tomography (PET) can indirectly index neuronal activity in the working brain. Following insightful and important critiques of structural neuroscientific research by feminist neurobiologists [e.g., 58], the use of FNI to investigate sex differences has also recently begun to be subjected to feminist scrutiny. A number of scholars have argued that, as in the past, FNI research currently reinforces and legitimates traditional gender stereotypes and roles in ways that are not scientifically justified. The growing number of charges of ‘neurosexism’ (as it was first termed in this journal [9]) have argued for bias in the way FNI research on sex differences is conducted and interpreted, detrimental effects for understanding the complex phenomenon of gender, as well as harmful social and psychological effects from the reification of gender roles [e.g., Fine, C. unpublished. Neurosexism in functional neuroimaging: from scanner to pseudoscience to psyche. In The Sage handbook of gender and psychology, eds. M. Ryan and N. Branscombe. Thousand Oaks: Sage., 1018].

However, a stark counterpoint to this perspective is that there are “deeply entrenched, harmful, and generally implicit biases against sex influence research among neuroscientists” [19, p. 38]. Recent criticisms of research investigating neurobiological contributions to sex differences [10, 20] have been countered with arguments that there is “an overemphasis on criticism of a few errant studies or errant interpretations”, and that they do not make sufficient acknowledgment of the careful and valuable neuroscientific research into sex differences that is being conducted [21, 22, p. 3]. From this perspective, the foremost concern is that failure to investigate sex influences on the brain impedes scientific progress in understanding sex differences in behavior and vulnerability to psychiatric and developmental disorders [e.g., 19, 23, 24], and that an over-emphasis on the perils, rather than the promise, of such research will hinder scientific progress. Thus McCarthy & Ball [22, p. 3] expressed fear that recent feminist critiques “threaten to severely hamper or even reverse the progress being made in this field”. It is worth noting that a perception endorsed by some within the scientific community is that the hypothesis that there are intrinsic sex differences in brain and behavior is ‘politically incorrect’: it has been described by a number of scientists as “politically dangerous” [25, p. xi], “taboo” [26, p. 19, 27] and, politically, “not a permissible hypothesis” [Haidt, quoted in 28]. In a similar vein, at a recent Social Issues Roundtable on the ‘The Promise and Peril of Research on Sex Differences’ at the 2011 meeting of the Society for Neuroscience, Cahill [quoted in 29] was reported as raising “the opposite concern [to the abuse of neuroscience to justify sexism]: His colleagues are so afraid of being called ‘neurosexists’ that they’ve refused to study or acknowledge differences.”

This article seeks to gain insight into the validity of these differing perspectives by using recent two year literature samples and a case study to investigate how researchers address three substantial obstacles to gaining replicable and valuable insights into sex differences from FNI data. The first difficulty is the scope for false-positive errors that arises when sex comparisons are made. The second obstacle lies in understanding what, if anything, a sex difference in brain activity means in terms of differences in mental processes. Group differences in brain activity are not readily translated into psychological differences and this gap in knowledge of brain-mind relations creates a danger that, as in the past, gender stereotypes will be drawn upon to putty-fill in the gap [6]. The third issue is that sex differences in both brain and behavior are potentially malleable, or plastic, meaning that insight into the developmental or situational origins of sex differences in the brain, and the extent to which social factors increase, reduce, eliminate or even reverse them, cannot be gained through single “snap-shot” comparisons of male and female brains [18]. Inadequate treatment of these difficulties creates potential for a bias toward an exaggeration of the extensiveness, functional importance, and fixedness of sex differences in the brain. Such a presentation would implicitly support gender essentialism; that is the notion of deep-rooted, permanent, and distinct male versus female “essences” that make gender divisions seem natural, inevitable, and desirable [for overview of theory and data on psychological essentialism, see 30]. Alternatively, if charges of neurosexism are incorrect, then the overall research picture (excepting the possibility of a few isolated ‘bad apples’) should reveal methods and practices that take care to reduce the influence of false-positive errors on the field, that involve the systematic testing of hypotheses about the functional implications of sex differences in the brain in order to reduce scope for gendered assumptions to creep into interpretations, and that show careful attention to the influence of gendered experiences on sex differences in brain activity.

Exaggeration of Extent of Male/Female Brain Activation Differences? Sex Comparisons and False-Positive Errors

There can be legitimate reasons to compare the sexes in FNI work [see 24]. For example, if male and female samples differ behaviorally, for whatever reason, sex differences in brain activity can be expected. It is also possible that males and females may reach the same behavioral end via different neural means [3133]. If so, conflating the two populations, or only investigating one sex and extrapolating to the other sex, may be inappropriate. Indeed, one rationale for making sex comparisons in neuroscientific research is to redress an historical tendency for biological research to be conducted mostly on males, with findings and implications for pathology then extrapolated, perhaps incorrectly, to females [24]. However, publication bias towards statistically significant findings is a long-noted problem within sex differences research [34]. When a single experiment establishes a “significant difference” between the sexes, this does not necessarily reflect a real and reliable result. False-positive errors and publication bias is a major issue in behavioral science generally [for recent discussions, see 35, 36]. However, this problem can be especially exacerbated in sex differences research. The primary reason is the ease and obviousness of testing for sex differences, even in the absence of an a priori reason to do so. Since by convention one in twenty “significant” results occur by chance, if 20 researchers routinely test for sex differences then, even if there is no real difference between the populations, one researcher will find a statistically significant difference.

There is no way of knowing whether researchers who do not report sex comparisons have nonetheless tested for them, and it is not known how common such practices are. However, Kaiser and colleagues have argued that because sex is a primary and ubiquitous social category, classifying participants by sex is a “natural default” and “seemingly effortless and obvious in brain research” [12, p. 54]. They have also noted that, in comparisons of pathological groups with controls, “introducing sex/gender as a supplementary factor augments the possibilities for differentiating the groups from each other”, offering additional opportunities for publishing.

A second reason for concern regarding false-positive findings in FNI investigations of sex differences arises from the small sample sizes that are common in FNI research. As Wallentin [37] has noted in relation to sex difference findings, FNI studies are especially vulnerable to spurious results due to the difficulty of balancing participants on nuisance variables (like breathing rate and caffeine intake) that affect the imaging signal, and that this is particularly an issue when sample sizes are small. An analysis of the effect of sample size on the reliability and sensitivity of fMRI studies concluded that the large inter-subject variability is a key issue, and recommended a sample size of at least twenty for adequate reliability in group fMRI studies [38]. To the extent that sex comparisons are made post hoc rather than a priori, the issue of sample size is likely to be exacerbated further, as researchers will not have considered in advance the number of male and female participants necessary for adequate statistical power (sensitivity) or reliability.

In addition, the choices researchers make about how to analyse their results can increase the scope for false-positive errors in sex comparisons. For the purposes of analysis, the brain is divided into tens of thousands of tiny regions (or voxels), and tested for blood flow changes as a function of the psychological construct of interest. Brain regions that show changed activity become plausible (although far from certain) candidates for involvement in the psychological process of interest. As Bluhm [15] and Kaiser et al. [12] have noted, researchers may sometimes inappropriately make use of within-group analyses to make claims about male/female difference. That is, researchers may analyze data for males and females separately and find that while a particular brain region was, say, significantly activated in males in the experimental task (compared with a control condition), this difference was not statistically significant in females. However, within-group analyses do not allow claims to be made about between-group differences. (In this hypothetical example, it is possible that male/female activity in that particular brain region does not significantly differ.) Despite this, within-group analyses have been found to be surprisingly common in neuroscience research [39].

The reality of the problem of false-positive errors in FNI studies of sex differences is well-illustrated by investigations of the long-standing idea that the male brain is more lateralized than the female brain for language processing: that is, that the male brain tends to engage the left hemisphere when processing language stimuli, while the female brain tends to engage both. This alleged sex difference in lateralization is often suggested to underlie female language superiority, and a similar lateralization difference for visuo-spatial processing to underlie male superiority in that domain [see 8, 10, 40] In an oft-cited fMRI study of 19 men and 19 women, Shaywitz and colleagues found that for phonological processing (although not semantic or orthographic processing), neural activity in language regions was lateralized in men but not women [41]. Yet an investigation of the generalizability of sex differences observed with 13 males and 13 females found that those differences failed to generalize to similar language tasks within a second group of 10 men and 10 women. Moreover, identical analyses of the same participants ‘discovered’ brain activation differences between randomly created groups matched on sex, performance, and obvious demographic characteristics [42]. Importantly, two recent large meta-analyses of FNI studies of language lateralization failed to find evidence overall for sex differences [43, 44], and the earlier meta-analysis reported markedly smaller sample sizes for studies reporting sex differences (a mean of 31) than for studies reporting no differences (a mean of 76). These findings underline the need for scepticism towards any one particular finding of a sex difference, particularly when it arises from a small sample.

The seriousness of false-positive errors was recently emphasized by Simmons et al. [45, p. 1], who suggested that they are “[p]erhaps the most costly error” in science. They noted that false-positive results tend to persist because failures to replicate are inconclusive, and unappealing both to attempt by researchers and to publish by journals. In addition, they can lead to wasted resources in fruitless research enquiries, as well as misguided practical applications. Thus scientists have general reasons to avoid false-positive errors. If, contra claims of neurosexism, scientists perceive additional political pressures to avoid false-positive errors (or, indeed, true-positive results) in sex differences research, then this will be reflected in the literature. This was systematically investigated in three ways: first, by looking at typical sample sizes in studies reporting sex differences in brain activation; second, by assessing the standards of evidence deemed sufficient for a hypothesis of male/female brain difference to be presented as supported in the scientific literature; and third, by examining carefulness in citing a potentially spurious result.

Typical Sample Sizes

Typical sample sizes for FNI studies reporting sex differences offer an insight into the treatment of scope for false-positive errors in this area of research. In particular, a literature in which studies reporting positive results are predominantly underpowered, with group sample sizes smaller than twenty [as recommended by 38] or, more liberally, sixteen [as recommended by 46] should create concern that such findings will not prove to be replicable. Therefore to investigate typical sample sizes in this area of research, the Medline, Web of Science and PsycINFO databases were searched for fMRI studies published in 2009 and 2010 in which sex differences were referred to in the article title.1 Thirty-nine studies fulfilled the necessary criteria (see Table 1 and Appendix for further details).2 Over the entire sample, the mean number of males was 19.0 (median = 16, sd = 10.6) and the mean number of females was 18.5 (median = 18, sd = 9.5). However, a number of the studies in the sample also made sex-by-group or sex-by-age comparisons, both of which require larger sample sizes than for sex-only comparisons, for similar statistical power. For studies that made only sex comparisons (n = 22) the mean number of males was 13.5 (sd = 6.4) and the mean number of females was 13.8 (sd = 6.5). It is worth noting that the second largest study in this group (with a total sample size of 50) reported a null finding [47]. For studies reporting sex-by-group comparisons (n = 14) the mean number of males per subgroup was 11.9 (sd = 3.8), and females per subgroup was 11.4 (sd = 4.0). The largest sample sizes were observed in the three studies exploring age effects, for which the mean number of males was 32.7 (sd = 9.2) and the mean number of females was 26.3 (sd = 2.3). The distribution of sample sizes is shown in Fig. 1, which shows that the clear majority of studies had (a mean of) fifteen or fewer participants of each sex in each of their groups of interest, and that studies were as likely to have fewer than ten participants as they were to have more than twenty. It is worth emphasizing that these are not studies in which sex differences are mentioned incidentally, but studies in which sex differences are the primary focus of research, as indicated by the article title.
Table 1

Numbers of male and female participants in 2009 and 2010 functional MRI studies of sex differences

Study (by first author name)

Total males

Total females

Subgroups

Mean no. males in subgroups

Mean no. females in subgroups

Sex only comparison

Aikins

12

12

   

Cornier

21

22

   

Domes

16

17

   

Fine

10

10

   

Frank

6

6

   

Garn

13

13

   

Goldstein

13

12

   

Ino

10

10

   

Keller

24

25

   

Killgore

8

8

   

Krach

12

12

   

Lee

12

10

   

Li

30

30

   

Mak

12

12

   

Ohrman

8

12

   

Owens

9

10

   

Qiu

19

19

   

Riedl

10

10

   

Schmidta

25

25

   

Straube

12

12

   

Sveljo

6

6

   

Wang

10

10

   

Mean (sd)

13.5 (6.4)

13.8 (6.5)

   

Sex by group comparison

Bitan

17

22

Low/High VIQ using median split

8.5

11

Clements-Stephens

16

16

Plus/Minus 10 years of age

8

8

Coman

25

18

Genotype

9

12.5

Derntl

12

12

Luteal phase (females)

12

6

Eisenberger

18

18

Treatment versus placebo conditions

9

9

Elsabagh

30

20

Clinical status

15

10

Felmingham

42

44

Trauma exposure and clinical status

14

14.7

Gauthier

22

22

Low/High verbal fluency

11

11

Kempton

40

34

Genotype

13.3

11.3

Klucken

20

20

Awareness of conditioning contingency

10

10

Mather

24

23

Stress versus control conditions

12

11.5

Merz

20

19

Cortisol versus control conditions

10

9.5

Rumberg

12

24

Oral contraceptive status for females

12

12

Valera

46

47

ADHD versus control participants

23

23.5

   

Mean (sd)

11.9 (3.8)

11.4 (4.1)

Age effects

Christakou

38

25

   

Rubia

38

25

   

Zuo

22

29

   

Mean (sd)

32.7 (9.2)

26.3 (2.3)

   

Mean (sd) for entire sample

19.0 (10.6)

18.5 (9.5)

   

aThis study reported a null result

https://static-content.springer.com/image/art%3A10.1007%2Fs12152-012-9169-1/MediaObjects/12152_2012_9169_Fig1_HTML.gif
Fig. 1

Frequency of male/female group or mean subgroup sample sizes in 2009 and 2010 functional MRI studies of sex differences

Case Study: Standards of Evidence

It might be objected that the above literature survey misrepresents the situation, by looking at a sample made up of largely exploratory and isolated findings rather than an established research program. As a test of the validity of this objection, a currently prominent program of sex differences research was chosen as a case study to assess the standards of evidence required to claim a neurological sex difference as supported. This program investigates the proposal that there is a sex difference in the modulation of memory consolidation by the amygdala for emotionally salient material [see 19, 48]. Specifically, Cahill and colleagues have suggested that there is sex-dependent lateralization of this modulation effect, with the right amygdala more strongly involved in men, and the left in women. Further theorizing and empirical work has suggested functional implications, such as more enhanced memory for the ‘gist’ of events in males, but of peripheral information in females [see 19]. For three reasons, this claim would appear to be an appropriate and fair test-case. First, the hypothesis is based on a well-developed neurocognitive model of emotional memory enhancement, and predictions regarding sex-modulation have been tested in numerous studies. In other words, it is one of the few examples of a claim of a sex difference from FNI data that is not based on an isolated or post hoc finding. Second, Cahill and colleagues have concluded that there is “no reasonable doubt about the existence” of this sex-related laterality in humans [49, p. 262]. Additionally, the finding has been cited without caveat in some subsequent general reviews of emotional memory as well those concerned specifically with sex influences [e.g., 19, 5052], although there are also more cautious review references [e.g., 53]. Finally, this research program was described as an example of “high quality research by scientists who take care in … the scope of their claims”, for which feminist criticisms of the field are misplaced [21, p. 1321], reflecting its status as an apparently well-supported finding. It would therefore appear to be well-positioned in the literature to have gained good empirical support, and thus to be a fair test case study for standards of evidence required to claim a finding as supported.

The claim for humans is principally based on FNI research showing stronger correlations between enhanced emotional memory assessed at a delay (generally 2–3 weeks) and right amygdala activity in males, but left amygdala activity in females. Cahill and colleagues’ (2004) conclusion that no reasonable doubt that this sex-related laterality exists was based on the finding of the study reported in that article, together with six previous studies [5459]. These seven studies are summarized in Table 2, and there are a number of observations to be made. The first point is that four of the seven studies included only one sex [54, 56, 58, 59]. Further, given that the methodologies of these four studies differed in a number of ways (such as imaging technology, stimuli, length of delay to memory test, and construction of the dependent variable), these studies cannot provide solid insights into sex differences since there may be methodological reasons for different findings. A fifth study provided only within-group analyses which, as noted earlier, also do not enable statements about between-sex differences to be made [55].
Table 2

Studies presented in support of a sex difference in the modulation of memory consolidation by the amygdala for emotionally salient material

Study

PET/fMRI

Males

Females

Stimuli

Memory test/delay

Reported correlations between amygdala activity and delayed enhanced emotional memory measures

Cahill et al. (1996)

PET

8

0

Emotional (-ve) vs neutral films

Recall at 3 weeks

Positive correlation found between right amygdala activity and number of emotional (but not neutral) films recalled.

Hamann et al. (1999)

PET

10

0

Emotional (+ve and -ve) versus control (neutral and interesting) pictures

Recall and recognition at 4 weeks

No findings for recall (possibly due to restricted range). Positive correlations found between bilateral amygdala activity and recognition memory enhancement for both positive and negative pictures. For negative pictures, this was described as “more right-lateralized”, although no statistical comparisons were presented.

Canli et al. (1999)

fMRI

0

10

Positive and negative pictures

Recognition memory 2–14 months later

Left amygdala activity correlated with both positive and negative recognition memory, right amygdala activity correlated with only negative recognition memory.

Canli et al. (2000)

fMRI

0

10

Negative and neutral pictures

Recognition memory at 3 weeks

Positive correlation between left amygdala activation and recognition memory for pictures rated as highly emotional. Data for other trials not included in the analysis.

Cahill et al. (2001)

PET

11

11

Emotional (-ve) versus neutral films

Recall at 3 weeks

Within-group conjunction analyses identified regions showing increased activity for emotional compared with neutral films that correlated with enhanced emotional memory. In men, the right amygdala was identified. In women, the left amygdala activity was identified.

8 previously reported in Cahill et al. (1996)

Canli et al. (2002)

fMRI

12

12

Negative and neutral pictures

Recognition memory at 3 weeks

Analyses were performed for highly emotionally arousing pictures only. Within-group analyses found a positive correlation between right amygdala activity and recognition memory in men, and between left amygdala activity and memory in women. The significance levels of the correlations were then used to construct a “laterality index” based on maximal amygdala activity-memory correlations, used to statistically test for the effects of sex on laterality. This found a significant group difference.

10 previously reported in Canli et al. (2000)

Cahill et al. (2004)

fMRI

8

7

Negative and neutral pictures

Recognition memory at 2 weeks

Within-group conjunction analysis identified regions where greater activity was associated with both better memory performance and increasing arousal ratings. In men, the right amygdala was identified. In females, the left amygdala was identified. Between-sex laterality compared by using β representing the modulation of memory by arousal for the voxel of maximal activation in the amygdala and β for activation in the homotopic voxel of the opposite hemisphere as the dependent variables in a sex-by-hemisphere ANOVA. The significant sex x hemisphere interaction was ‘replicated’ using data from Canli et al. 2002.

The last two studies provided direct statistical tests of sex differences in laterality of amygdala modulation of emotional memory [49, 57], but the dependent variable used to compare laterality in the sexes was constructed differently in the two studies (although the analysis of [49] was repeated on [57], see last column of Table 2). This is not surprising since it is currently unclear how lateralization should be assessed [12]. However, the lack of established techniques increases researcher degrees of freedom [see 45], and it has been found that different techniques and thresholds for testing laterality differences can yield markedly different results [12, 60].

The final point to be made concerns sample sizes. The largest sample (24 participants in total) introduced selection bias by including the data of ten female participants from a previous (female only) study [58]. Thus, the sample was biased towards supporting the hypothesis, since it was in part on the basis of those data that the hypothesis under test was derived. This was also the case for the next largest study [55]: again, data for eight of the 22 participants came from an earlier (male only) study [54] from which the hypothesized sex difference was partially derived. The one remaining study that included both males and females [49] had only eight men and seven women.

Casting further reasonable doubt on the hypothesis is the absence of evidence of sex differences in the effects of right versus left amygdala damage on emotional memory in a small patient study [61], and null findings in some recent FNI studies of this phenomenon [62, 63]. Interestingly, Ritchey and colleagues [63, p. 2502] argued that their null finding for sex (noted briefly in the Discussion) “should be treated with caution … due to relatively small sample sizes for males [n = 6] and females [n = 7].” This is a valid point, but as noted earlier, such caution is also appropriately applied to positive FNI findings from small samples.

It may be worth stating explicitly that the points made here should not be taken to imply that the possibility of a sex difference, suggested by the earlier findings with male-only and female-only samples, should not have been investigated at all. Rather the issue is that the hypothesis arguably remains to be put to fully rigorous testing and so, despite the costs of false-positive errors, the confidence with which it is presented by some experts seems premature.

Care in Citations of Potentially Spurious Results

Once a reported sex difference is in the published literature, the manner in which it is cited reflects the extent to which the scientific community is sensitive to the possibility of false-positive results. Two basic ways in which scientists can demonstrate carefulness in their treatment of findings of sex differences are to acknowledge any similarities also observed, and to take care to cite larger studies or meta-analyses that failed to demonstrate a particular sex difference in favour of (or at the very least as well as) smaller studies with positive findings. Recent (2009 and 2010) citations of Shaywitz et al. [41] were chosen to investigate this, by assessing how many noted the sex similarities in brain activity observed for semantic and orthographic processing and how many also noted the existence of contradictory evidence. This study was chosen due to the considerable counter-evidence to the positive Shaywitz finding including the publication of two meta-analyses [43, 44], and the time period was one for which (at time of analysis) citing authors would have had the greatest opportunity to become apprised of those meta-analyses and the data included in them. The Medline, Web of Science and PsycINFO databases were searched for all citations of the Shaywitz paper in 2009 and 2010, and these were assessed to see how many researchers citing that study: first, did so in a way that acknowledged the absence of sex differences in brain activation for two of the three language tasks; and second, acknowledged subsequent null findings (see Table 3 and Appendix for details).
Table 3

Studies citing Shaywitz et al. (1995) in 2009 and 2010, categorized according to the extent to which contradictory data are also cited, including two meta-analyses

Context in which Shaywitz et al. (1995) is cited

Studies (by first author)

Excluded: incorrect/inaccurate/non-gender specific citations (n = 12)a

Binder, Boyd, Chen, Gillies, Hansen, Hazlett, Limbrick, Locascio, Maruyama, Merz, Ohuchida, Stevens-Smith

No contradictory data cited (n = 43)

Allez, Badcock, Blakemore, Breier, Brugger, Byrnes, Caparelli, Capone, Costafreda (a), Cousin, Draca, Feinstein, Fink, Ford, Galambos, Gorbet, Hartwigsen, Henderson, Hines, Hockenbury, Isman, Iwabuchi, Jahanshad, James, Kempe, Lenroot, Lopes, Lv, Majovski, Sereno, Sousa, Sternadori, Strong, Tong, Tranel, Van Dyke, Van Strien, de Vries, Van Wormer, Weis, Wheldall, Wilde, Yousem

Contradictory data cited

No citation of Sommer/Sommer cited incorrectly (n = 11)

Cahill*, Darlington, Gauthier, Kivilevitch, Li, Logan*, Lyttelton, Rumberg, Semrud-Clikeman, Jancke, Mbwana

Sommer as if of equivalent warrant (n = 9)

Bitan, Costafreda (b), Geary, Gordon, Hausmann, Lebel, Liu, Sava, Reiterer

Sommer cited as holding greater warrant (n = 12)

Andreano, Chiarello, Cowell, Garn, Ihnen, Kaiser, Kherif, McManus, Pinel, Stelnikov, Wallentin, Zaehle

aAlso excluded were four non-English articles and a book chapter by Iris Sommer (see main text)

Asterisked (*) studies cite a critical review paper [37]

There were 92 citations of the article during the 2 year time period. Non-English language articles were excluded from the sample (n = 4), as were citations that were erroneous (e.g., relating to the comorbidity of language disorders and ADHD), inaccurate or vague (e.g., misrepresenting the finding as being in a non-language domain, or as providing a possible neurobiological explanation for greater numbers of boys having reading difficulties), or that referenced the findings of the article in a non-gender-specific way (n = 12). In addition, a book chapter by Sommer (2010) was excluded for obvious reasons. Citations that referred to the findings as being in the wrong language domain (e.g., “listening to speech”) or in a non-specific way (e.g., “high-level cognitive processing”) were included. This resulted in a sample of 75 citations of the Shaywitz study as demonstrating a neurological sex difference in brain activation either during language processing or in processing in general. Of these 75 citations, only three referred explicitly to the absence of findings of sex differences in brain activity for orthographic and semantic processing within the Shaywitz study itself [12, 42, 64],3 although several other citations noted more generally that lateralization differences applied specifically to phonological processing.

Each citation was then categorized according to whether, and how, it referenced the existence of contradictory findings (see Table 3 and Fig. 2). Forty-three of the seventy-five (57 %) cited the Shaywitz study without referring to the existence of any contradictory data. Of the remaining 21 citations, eleven (15 % of the total) either failed to cite Sommer and colleagues’ work or did so in a way that was misleading, for example, describing the findings as showing more bilateral activation in women [65]. A further nine studies cited one or both meta-analyses, but in a way that gave no indication that, through virtue of its status as a meta-analytic study, it lay claim to being a more reliable source of information than Shaywitz et al. Thus, just twelve of the 74 (16 %) studies that cited the Shaywitz finding also cited the work of Sommer and colleagues in a fully informative way (even if the latter’s conclusions were disputed).
https://static-content.springer.com/image/art%3A10.1007%2Fs12152-012-9169-1/MediaObjects/12152_2012_9169_Fig2_HTML.gif
Fig. 2

Studies citing Shaywitz et al. (1995) in 2009 and 2010, categorized according to citation of contradictory data

Exaggeration of Functional Significance of Male/Female Brain Activation Differences? Functional Obscurity

A second major difficulty for neuroscientists is to conduct research in ways that can yield understanding of the functional implications, if any, of sex differences in brain activation [see 10]. FNI can be used to make ‘forward inferences’ about which regions of the brain are candidates for involvement in a particular psychological process. Sex comparisons can then further establish whether and how these neural correlates differ between the sexes. Making further inferences about mental processes from differences in neural activity between experimental and control conditions requires making ‘reverse inferences’; that is, claims of the form ‘Brain Region X showed increased activation therefore Mental State Y was present.’ However, mental processes arise from the complex and dynamic interaction of multiple brain regions that themselves comprise a staggering complexity of connections. Conversely, any particular population of neurons will be recruited by multiple mental processes [see 66]. The absence of neat one-to-one mapping between brain regions and mental processes renders reverse inferences logically invalid (although the probability that brain activity in region X indicates the presence of mental state Y increases to the extent that X is exclusively activated by Y [67, 68]). Also, not only can activity be linked only extremely speculatively with a particular mental process, but correlation does not demonstrate causation. A number of studies have shown significant activity in regions that, from other techniques, appear not to be crucially involved in task performance [e.g., 69, 70]. Moreover, greater activation doesn’t imply ‘more’ mental process, since development or practice can lead to ‘streamlining’ of neural activation [7173].

How, then, to interpret sex differences in brain activity in a particular region? As Bluhm [15] has noted, researchers interested in sex differences have to interpret “a difference in a difference” in the brain activity associated with a psychological function compared with a control task, the psychological meaning of which is even more obscure than the difference itself. For example, a sex difference in brain activity could reflect a difference in either degree of activation of the same neural circuitry, or the use of different neural circuitry to perform the same task. It is also possible that sex differences in brain activity have no functional significance, and merely reflect a different neural means to the same behavioral ends [31]. The danger to researchers is that gender stereotypes can be readily drawn upon to interpret these highly ambiguous data. For example, it has been noted [10] that the absence of empirical or theoretical grounding for speculations about brain-mind relations in visuo-spatial processing means that precisely opposite hypotheses can be equally plausible [74], and that hypotheses can be revised post hoc in ways that maintain the over-riding assumption that the male brain is superior [75, 76]. Bluhm [14] has likewise documented how three studies respectively showing greater [77], lesser [78], and similar [79] activity in the male prefrontal cortex were all interpreted as suggesting greater cognitive control of emotion in males, in line with a long-standing gender stereotype.

Poldrack [68] has argued that reverse inferences are a useful research tool when used to generate hypotheses to put to test in further empirical work. Thus while there is certainly arguably a place for post hoc speculation based on exploratory research, it should form part of a larger strategy of the systematic development and testing of predictions derived from increasingly well-specified neurocognitive accounts of the sex-modulated processes involved in the behavior under investigation. This would reduce the scope for false-positive errors, as well as the scope for untested stereotype-infused speculations about the functional significance of neurological findings. The importance of such an empirically and theoretically grounded approach was recently stressed by Park and Huang [80] in their discussion of the perils of investigating cultural influences on the brain, noting that it is “critical that specific hypotheses grounded in knowledge of neural structures and behavioral data be tested. … [This enables us] to test specific hypotheses in regions of interest and limit the amount of neural “real estate” under investigation, increasing the prospects of finding interpretable, replicable differences that are related to cultural values and beliefs [rather than other factors not of interest].”

To what extent, then, does recent FNI research on sex differences, through the use of research driven by well-specified neurocognitive models, constrain the potential for neurosexist speculations? This question was investigated with the set of thirty-nine 2009 and 2010 fMRI studies described previously (see Table 4). First, studies were examined for the presence or absence of a prediction regarding expected sex differences in brain activity. Studies categorised as ‘exploratory’ presented no such predictions. Studies categorised as ‘empirically-driven-vague’ made imprecise predictions of the brain regions in which they expected sex differences to be observed, based on previous empirical findings. Studies categorised as ‘empirically-driven-precise’ made more precise predictions of this kind. A liberal definition was used, in which even very broad-brush predictions of differences in large regions of the brain (e.g., “parietal lobe”) were categorised as precise, so long as the direction of difference was specified. Studies categorised as ‘hypothesis-driven’ drew on neurocognitive models to make precise predictions of the brain regions in which they expected sex differences to be observed, corresponding to specific mental processes thought to differ between the sexes. Where a study included predictions of more than one type (e.g., both ‘vague’ and ‘precise’ empirically-derived predictions), they were placed in the more precise category. Eight of the studies were exploratory. A further eight studies were categorised as ‘empirically-derived-vague’, and the remaining twenty-three were categorised as empirically-derived-precise. No study was categorised as ‘hypothesis-driven.’
Table 4

Predictions, reverse inferences and consistency with available behavioral data for 2009 and 2010 fMRI studies of sex differences

Study (by first author)

Title

Prediction

Reverse inference

Relevant behavioral data available?

Behavioral data consistent with reverse inference?

Aikins

Sex-related differences in amygdala activity influences immediate memory

Empirically-based-vague: “examined whether the [predicted] relationship between lateralized amygdala activation and false-positive rate differed as a function of participant sex” (273).

Suggest that increased amygdala activation increases long-term retention of arousing material in men, but produces increased coherence in unpleasant pictures in women due to semantic encoding, that may result in difficulty discriminating between unpleasant pictures in women.

False positive rates measured. No measures of semantic encoding.

Inconsistent in that false positive rates did not differ between the sexes at the group level.

Bitan

Bidirectional connectivity between hemispheres occurs at multiple levels in language processing but depends on sex

Exploratory. Notably, findings are directly contradictory with standard accounts of the functional effects of sex differences in language lateralization.

Christakou

Sex-dependent age modulation of fronto-striatal and temporo-parietal activation during cognitive control

Empirically-driven-vague: “we hypothesised that gender would affect prefrontal, striatal and parietal brain activation, as well as the age modulation of these regions.” (224)

“the superior reliance on functional frontal mechanisms in females, and on functional parietal mechanisms in males may underlie aspects of the well-documented differences in cognitive strategies and relative abilities between the sexes.” (234)

No measures of hypothesised strategy differences (e.g., visual-spatial strategies in males versus top-down inhibition strategies in females).

Clements-Stephens

Developmental sex differences in basic visuospatial processing: Differences in strategy use?

Empirically-driven-precise: “hypothesized that with increasing age, males would develop a bilateral representation of visuospatial processing in parietal regions shown by the utilization of right hemisphere regions in younger-aged participants and the engagement of left hemisphere regions later in adolescent males. For females, we predicted that we would see a consistent right hemisphere network in frontal and parietal regions regardless of age.” (156).

The differences in activation patterns in females “could be mediated by a verbal strategy used early in development. Additionally, differences seen in males are consistent with visually based strategy use in which it is suggested that males rely on imagery and are ‘more hands-on’ during task completion. Moreover, males’ reliance on a visuomotor network could account for the traditional advantage on visuospatial tasks.” (160, all references removed)

No measures tapping strategies, no sex difference in performance on task, nor male advantage on other visuospatial tasks established for this sample.

Coman

The effects of gender and catechol O-methyltransferase (COMT) Val108/158Met polymorphism on emotion regulation in velo-cardio-facial syndrome (22q11.2 deletion syndrome): An fMRI study

Empirically-driven-vague: “we hypothesize that COMT genotype will have a sexually dimorphic effect on brain activation, especially in areas such as the amygdala and other limbic structures that have been previously shown to have differential activation during emotional processing in the general population.” (1044)

Cornier

Sex-based differences in the behavioral and neuronal responses to food

Empirically-based-vague: “hypothesized that women would have a greater cognitive or frontal response to food-related visual stimuli than men which would be related to the behavioral measures [greater sensitivity to hunger and satiety responses, resulting in greater capacity to maintain energy balance].” (539)

“women are more sensitive or ‘reactive’ to food-related cues than men … have greater attention (parietal response) and cognitive processing (prefrontal response) related to food stimuli. The greater DLPFC activation in women may also suggest a greater inhibitory response to the food cues.” (542)

Some loosely relevant measures.

Consistent. Women showed increased post-meal satiety ratings, and were more likely to maintain isocaloric intake during ad libitum feeding, which was correlated with DLPFC response to visual food cues.

Derntl

Multidimensional assessment of empathic abilities: Neural correlates and gender differences

Empirically-based-precise: “we assume that during emotional perspective taking and emotion recognition females will recruit more emotion-related regions such as the inferior frontal and superior temporal gyrus, while males will exhibit stronger activation in the temporo-parietal junction. Finally, for affective responsiveness we hypothesize overall stronger neural activation in female subjects, and particularly in the superior temporal and medial-frontal regions as well as the amygdala.”

Emotion recognition: “females rely on autobiographical memory to correctly label the emotional expressions.” (77) Emotional perspective-taking: “males might rely more strongly on a perceptual-analyzing network and mentalizing abilities than do females who rather recruit regions associated with emotional contagion and affective responsiveness when assessing the emotional expression of another person supporting the view that men probably have a lower tendency to share their emotions with others than females.” (77–8, reference removed)

No measures of ‘emotional’ versus ‘cognitive’ strategies to the various tasks. No measures of tendency to share emotions with others.

 

“we also conducted exploratory analyses concerning the influence of cycle phase on the activation pattern and suggested more pronounced responses in emotional networks during the follicular phase.” (69)

Affective responsiveness: “females relied more on emotional regions to experience the emotion, while males show a different neural strategy … supporting assumptions of a more cognitive route.” (78)

Empathy network: While females recruited regions “associated with emotion imitation and evaluation, males rather relied on a region known to be involved in semantic retrieval … thus indicating a rather cognitive approach.” (79)

“we also observed a strong trend towards a significant general task independent impact of menstrual cycle phase on amygdala activation further supporting the assumption that the hormone status during the follicular phase facilitates sensitivity and behavior in socio-emotional-situations. This may be traced back to the evolutionary advantage of higher attention and responsiveness to social-emotional interactions thereby improving mating chances during times of increased fertility.” (78, references removed)

No behavioral sex differences observed on any task (other than self-report).

Inconsistent. No performance differences between women in different menstrual phases.

Domes

The neural correlates of sex differences in emotional reactivity and emotion regulation

Empirically-derived-precise: “We hypothesized that women show enhanced initial emotional responding to aversive stimuli compared to men, associated with enhanced activity in emotionally relevant brain areas, in particular the amygdala. In addition, women were expected to show attenuated activity in the areas subserving the cognitive regulation of emotional responses, namely in the dlPFC, OFC, and ACC when attempting to decrease their initial emotional reactions to negative stimuli.” (760)

“the recruitment of prefrontal areas during the initial processing [in women] might reflect effortful cognitive processing, such as the allocation of attentional resources to the emotional aspects of the stimulus, which have been enhanced in women compared to men. Alternatively, women might have attempted to decrease their emotions as soon as the aversive stimuli appeared. However, as women showed enhanced amygdala responding in the initial viewing phase, these attempts might have been less effective compared to men.” (767)

Post-scan valence and arousal ratings of stimuli measured, as well as subclinical symptoms of depression and trait anxiety. No measures of emotional response post-regulation taken. No measures of attention to emotional stimuli or aggressive tendencies taken.

Inconsistent. No sex differences in post-scan valence and arousal ratings of stimuli. No sex differences on subclinical symptoms of depression and trait anxiety.

“our data in part support the idea that women may be more vulnerable to depression because they tend to be more reactive to emotional stimuli and are less effective in regulating their emotional response.” (767)

“increased amygdala activity to negative stimuli in men during cognitively increasing emotional responses might also have implications for the neural basis of maladaptive behaviors associated with enhanced emotional responding to aversive interpersonal stimuli that is more prevalent in men, e.g., aggressive behavior. The present results suggest that emotionally laden aggressive behavior might not only be due to difficulties in impulse control, but might also be promoted by the relative ease of voluntary emotional up-regulation in men.” (767)

Eisenberger

An fMRI study of cytokine-induced depressed mood and social pain: The role of sex differences

Exploratory

Elsabagh

A longer duration of schizophrenic illness has sex-specific associations within the working memory neural network in schizophrenia

Empirically-based-precise: “hypothesised that a longer illness duration would be associated with decreased activation of the PFC during performance of a WM task and may be accompanied by compensatory effects in other brain regions. Furthermore, we hypothesized that these effects would be more pronounced in male patients than in female patients.” (42)

Felmingham

Neural responses to masked fear faces: Sex differences and trauma exposure in posttraumatic stress disorder

Empirically-based-precise: “predicted that (a) trauma-exposed women would have greater activity in fear and arousal networks than would non-trauma-exposed women, trauma-exposed men, and non-trauma-exposed men; (b) both PTSD men and women would display greater fear and arousal network activation than trauma-exposed and non-trauma-exposed women and men; (c) women with PTSD would display greater activity in fear networks than would PTSD men.” (242)

“greater brainstem activation to fear may contribute to the greater prevalence of PTSD in women [through heightened arousal and orienting to threat], greater hippocampal activation in men may subserve an enhanced capacity for contextualizing fear-related stimuli.” (246).

No measures of arousal or orienting to threat, or capacity to contextualise fear-related stimuli

Fine

Gender differences in BOLD activation to face photographs and video vignettes

Empirically-based-precise: “hypothesized greater right lateralization in males compared to females to photos in the social network formed by the amygdala, frontal gyri and anterior cingulate. … hypothesized gender differences in the right and left temporal regions that process the various aspects of language [specifically greater left lateralization in males].” (138)

Two “plausible” inferences are that “the males in our study interpreted the stimuli differently than did the females” or “that males are actually less efficient at processing the stimuli … greater, more widespread activation may suggest that males recruit less expert neural systems when processing emotional stimuli.” (143). Also speculate that “male responses included additional checking to be sure that the positive stimuli were genuine and without threat at the preconscious level”, in reference to supposed greater male responsiveness to threatening stimuli (143–4)

No behavioral measures taken

Frank

Processing of food pictures: Influence of hunger, gender and calorie content

Exploratory

“higher [superior medial frontal lobe] activation in women while viewing high-caloric pictures when hungry compared to being not hungry. … These might be related to different self perception of men and women concerning food processing”. (164)

No measures of self perception concerning food.

Garn

An fMRI study of sex differences in brain activation during object naming

Empirically-driven-vague: that “men and women might differ in terms of hemispheric processing distribution during object naming” (611) and that there may be sex differences in regions related to either choice selection or word retrieval, corresponding to contrasting hypotheses about the reason for behavioral sex differences in naming living versus nonliving objects.

Suggest that activation differences may be due to differential lexical selection versus retrieval demands for men versus women, stemming from sex differences in vocabulary size.

No behavioral measures taken (silent naming task).

Gauthier

Sex and performance level effects on brain activation during a verbal fluency task: A functional magnetic resonance imaging study

Empirically-based-vague: “The aim of this study was to test whether (1) difference in neural correlates between sexes exist in a covert fluency task, irrespective of high or low level performance, (2) performance differences are related to neural activity regardless of sex for this same task and (3) sex and levels of performance act in an additive or interactive manner on neural area activation … our attention was drawn to understanding (4) whether the precuneus, occipital gyri and cerebellum would be differentially engaged to resolve the task according to performance and/or sex factors” (165).

“men seem to make greater use of visual mental strategies … than the usual phonological and semantic strategies” (175).

No measures of strategy taken.

Goldstein

Sex differences in stress response circuitry activation dependent on female hormonal cycle

Empirically-based-precise: “predicted that men would look more similar to women in [early follicular] than during [late follicular/midcycle phase], given the greater similarity of men’s hormonal status to women’s hormonal status at [early follicular] compared to [late follicular/midcycle phase].” (432).

“From an evolutionary point of view, it is important for the female during midcycle to have a heightened cortical capacity, unencumbered by excessive arousal, to optimally judge whether a potentially threatening stimulus, such as an approaching male, is an opportunity for successful mating or for fight or flight. Thus, females have been endowed with a natural hormonal capacity to regulate the stress response that differs from males.” (437)

No measures of judgment of threat level of relevant stimuli taken.

Inconsistent. No group differences in mood or anxiety measures before or after viewing aversive images. See main text for further discussion.

Ino

Gender differences in brain activation during encoding and recognition of male and female faces

Exploratory

“the reduced activation observed in women compared to men during encoding suggests that the relevant cerebral regions were recruited more efficiently in women. Therefore, this fMRI finding provides the neuronal basis for the superiority of women over men regarding facial recognition, which has been indicated by previously behavioral studies but was not duplicated in the present study, probably due to the ceiling effect.” (2)

Facial recognition performance measured.

Inconsistent. No sex differences in performance (although this is attributed to a ceiling effect).

Keller

Gender differences in the functional and structural neuroanatomy of mathematical cognition

Empirically-based-precise: “we predicted that … (2) gender differences in brain responses would exist even in the absence of overt gender differences in behavior, (3) males and females would show extensive overlap in activation of the left and right IPS and angular gyrus … and (4) males would show greater activation in the PPC whereas females would show greater responses in the left and right frontal cortex.” (343)

Kempton

The effects of gender and COMT Val158Met polymorphism on fearful facial affect recognition: a fMRI study

Empirically-based-precise: “that females homozygotes for the Met allele, would … show amplified task-induced activations within limbic regions … possible that a genotype-associated prefrontal-associated mechanism can override the effect of the genotype within the limbic regions.” (372)

Findings suggest that “emotional mimicry is perhaps minimal in males whereas in females there may be increased inhibition in this region” (377).

No relevant behavioral measures taken.

Killgore

Sex differences in cerebral responses to images of high versus low-calorie food

Empirically-based-precise: “hypothesized that females would show greater overall activation in response to high-calorie food images, particularly within prefrontal inhibitory and self-monitoring regions compared with males. Furthermore … it was hypothesized that women would show greater activation than men within [the insula and amygdala].” (354)

“Though speculative, [greater activation patterns in particular regions in women in response to calorie-rich food] may reflect increased engagement of prefrontal evaluative, decision-making, inhibitory, and self-referential cognitive systems” (357). Other sex differences in activation “reinforces the notion that women may process food imagery at a more complex cognitive/somatic level, whereas men may process food stimuli at a less complex hedonic approach-withdrawal level.” (358)

Self-reported appetite and control of eating behavior measured. No other relevant behavioral measures taken.

Inconsistent. No sex differences in self-reported appetite or control of eating behavior.

Klucken

Neural activations of the acquisition of conditioned sexual arousal: Effects of contingency awareness and sex

Empirically-based-vague: “hypothesized that … men would show increased responses” to conditioned stimuli relative to neutral stimuli, compared with women (3072).

Results “in line with current findings suggesting that men are more receptive to conditioning of sexual arousal than women” (3081),

Subjective sexual arousal ratings of conditioned stimuli measured, as well as skin conductance responses (SCRs).

Inconsistent. No sex differences in subjective arousal ratings of conditioned stimuli. (No effect of conditioning on SCRs.)

Krach

Are women better mindreaders? Sex differences in neural correlates of mentalizing detected with functional MRI

Empirically-based-precise: “we expected … that women would be better perspective takers and therefore display stronger signal changes in the medial prefrontal cortex.” (2)

Lee

Sex-related differences in neural activity during risk taking: An fMRI study

Empirically-based-precise: “hypothesized that the male and female participants would show differential patterns of neural activation associated with risk taking. … that when the female and male participants showed comparable rates of risky selection, the neural activations in the insula and the OFC would be stronger for the female participants than for the male participants. … We hypothesized that stronger correlations [between right insula and the OFC with the rate of risky selection and punishment] would be obtained for the female participants than for the male participants.” (1304)

“Our findings regarding the stronger activation in the OFC for the female participants than for the male participants suggests that women need to partake in a higher degree of mental consideration before making a risky response.” (1308)

No relevant behavioral measures other than risk-taking itself.

Inconsistent. No sex differences in either rate of making risky responses, or response time to make risky responses.

Greater “responsiveness [in particular brain regions in females] may arise because the female participants are more mentally alert than the male participants when updating and valuating their subsequent actions during the task.” (1308)

“it is likely that the male participants, relative to their female counterparts, were more active in encoding the positive experience from the reward feedback. The encoded positive experience may render men than women more ready for committing to subsequent risk-taking behaviors.” (1310)

“The stronger OFC and insula activities in women than in men may reflect the sensitivity of women to situations of ambiguity … [This] could make women more risk averse than men.” (1310).

Li

Gender differences in cognitive control: An extended investigation of the stop signal task

Exploratory

“These opposite patterns of cerebral responses between [stop success] and [stop error] trials suggested there were perhaps fundamental differences in the way men and women perform the stop signal task.” (270).

No relevant behavioral measures taken, and no behavioral differences on the task.

“This [neurological] gender difference … along with greater error-related activity in women, as compared to men, may indicate fundamental differences in the temporal dynamics with which men and women respond to errors during a cognitive task that requires moment-to-moment monitoring of performance.” (270)

“PCC activity may also reflect that, compared to men, women were involved to a greater extent in ‘mental reflection’” (270).

Mak

Sex-related differences in neural activity during emotion regulation

Empirically-based-precise: “we hypothesized that during emotion regulation, males would mainly recruit the dorsolateral prefrontal regions that are implicated in cognitive processes, while females would mainly recruit the orbitofrontal regions that are implicated in affective processes.” (2901).

The “pattern of findings is consistent with the higher emotional reactivity and emotion-focused coping style commonly observed among females. … [Observed male activity] is probably related to the cognitive effort and specific regulatory strategies that they had employed.” (2907).

Self-reported regulatory strategies measured.

Consistent. Marginally significant (p = .066) trend for females to report more emotion-focused regulatory strategies and males to report more cognitive-focused strategies.

Mather

Sex differences in how stress affects brain activity during face viewing

Empirically-based-precise: “for men observing other’s emotions, stress will decrease interactions between the amygdala and brain regions such as the insula and temporal pole that help people understand others’ state of mind and simulate others’ emotions, whereas for women, stress will increase interactions among these regions. In addition, it is hypothesized that, for men, stress will decrease coordination between a brain region engaged in basic visual processing of faces (the fusiform face area or FFA) and regions engaged in simulating and interpreting facial emotions (the temporal pole and insula) whereas, for women, stress will increase coordinated activities among these regions.” (933, references removed)

“it seems that coordination of basic face processing by the FFA and interpretation and simulation of emotional expressions by the extended temporal pole region increased under stress for women but decreased for men. This pattern is consistent with behavioral findings that stress promotes social affiliation for women but disrupts it for men.” (936, reference removed)

No relevant behavioral measures taken, such as tendency to social affiliation in response to stress.

Merz

Investigating the impact of sex and cortisol on implicit fear conditioning with fMRI

Empirically-based-precise: “predicted that cortisol would reduce the CS+/CS- differentiation in men, while enhancing them in women in [the amygdala, the hippocampus, or the thalamus] and additionally in the frontal cortex.” (35)

“Women were more prone to a facilitation of [the insula] during enhanced [glucocorticoid] levels possibly revealing an unconscious shift towards the acquisition of potential danger cues in a stressful situation.” (42)

SCRs measured.

Consistent. Cortisol enhanced mean SCR to the conditioned stimulus in women but reduced it in men.

Ohrmann

Effect of gender on processing threat-related stimuli in patients with panic disorder: Sex does matter

Empirically-based-precise: “postulated a stronger activation of the neural fear circuitry in response to threat-related stimuli in female [panic disorder] patients, as compared to the male patient group.” (1035)

“strongest amygdala activation in women was observed after exposure to angry and neutral faces, implying that these two facial expressions are the most threatening to women.” (1041)

No threat judgment measures taken.

Owens

An fMRI study of self-reflection about body image: Sex differences

Empirically-based-precise: that “females might show stronger activation in mPFC when engaged in self-reflection about their bodies” (850).

“when females [but not males] are confronted with fat images and asked to apply this criterion to their own bodies, they are more inclined to seriously engage processes of self-reflection and (re)evaluate relevant self-representations given the cultural directive of thinness for female bodies.” (853)

Measures of body image concern and post-experiment reports of self-reflection taken.

Inconsistent. Both sexes scored extremely low on a measure of body image concern.

“results provide converging evidence of a stronger relationship between body size and self-representation for women than for men and they provide further support for the hypothesis of a number of theorists that a typical, non eating disordered woman in US society shows a relatively high level of concern for body shape.” (853)

Qiu

The effects of acupuncture on the brain networks for emotion and cognition: An observation of gender differences

Empirically-based-vague: “we hypothesize that women and men may have different brain activation/deactivation patterns at the [limbic–paralimbic–neocortical network] and the [default mode network] in response to acupuncture procedure.” (57)

Reidl

Are there neural gender differences in online trust? An fMRI study on the perceived trustworthiness of eBay offers

Empirically-based-precise: eight specific hypotheses made.

“Our fMRI data (i.e., the striatum activation) provide empirical support that women consider socio-psychological and emotional concerns most relevant in both conventional shopping and online shopping.” (416)

Perceived trustworthiness of offers measured, but no behavioral measures of functional versus social/emotional motivations in shopping, of emotional versus cognitive processing, or of perception of uncertainty or risk.

Inconsistent. Women gave higher ratings of trustworthiness than men.

Larger cluster size of dorsal anterior cingulate cortex activation in men “supports the notion that men, in contrast to women, usually process information in a cognitive rather than affective manner” (417, references removed).

Task “triggered activation in more brain areas for females than for males … This pattern of female brain activation may occur because women perceive greater levels of uncertainty and risk” (419, reference removed).

Rubia

Effects of age and sex on developmental neural networks of visual-spatial attention allocation

Empirically-based-precise: “expected to observe enhanced frontal activation in females and enhanced parietal activation in males during visuospatial attention allocation. Furthermore, … we hypothesised that brain regions that differed in activation between males and females would be related to underlying sex differences in the neurofunctional maturation of these brain regions, with more pronounced maturation of parietal regions in males and frontal regions in females.” (818).

“Males seem thus to have relied more on bottom-up parietal visual-spatial perception mechanisms for stimulus saliency processing, while females appear to have relied more on top-down fronto-striato-temporal executive control of selective attention, which may be slower and less effective than more automatic bottom-up processes used by males.” (824).

No behavioral measures of processing style differences. Mean reaction time and error rates measured.

Inconsistent. Males and females did not differ in mean reaction time on either congruent or oddball trials, although the increase in reaction time on oddball trials compared to congruent trials was greater for females than for males.

Rumberg

Cycle and gender-specific cerebral activation during a verb generation task using fMRI: Comparison of women in different cycle phases, under oral contraception, and men

Empirically-based-precise: “hypothesize that the lateralization of the activation patterns will differ in women dependent on their hormonal status, leading to significant activation in the frontal and/or temporal areas for women under oral contraception compared to women without oral contraception in two different cycle phases. Concerning the gender differences, we assume that the previously described differences between men and women in the temporal cortex [higher in males] will be verified, but hypothesize that this finding will also depend on the cycle time.” (367).

Schmidt

No gender differences in brain activation during the N-back task: An fMRI study in healthy individuals

Empirically-based-precise: “initial tentative hypothesis was that men would show greater activation in frontal and parietal regions than women [in a verbal working memory task].” (3610)

Straube

Sex differences in brain activation to anticipated and experienced pain in the medial prefrontal cortex

Exploratory

“On the basis of the putative role of the anterior MPFC, increased activation in this area [in females] suggests stronger self-related processing of anticipated and experienced, clearly painful stimuli in women, as compared to men.” (697).

No relevant behavioral measures taken.

Sveljo

Gender differences in brain areas involved in silent counting by means of fMRI

Exploratory

“Our results may suggest that gender related differences in language processing reflect specific cognitive and executive strategies as a response to certain stimuli” (7).

No behavioral measures taken. Silent counting task.

Valera

Sex differences in the functional neuroanatomy of working memory in adults with ADHD

Empirically-based-precise: “predicted that … functional differences [between adults with ADHD and comparison subjects, including frontal hypofunction and right frontal-striatal-cerebellar abnormalities] would be smaller for the women with ADHD than for the men” … “we also conducted exploratory analyses to assess the correlation between ADHD symptoms and neural activation associated with performance on the working memory task for men and women separately.” (87)

Wang

Is the contribution of the amygdala to the sex- and enhancement-related effects of emotional memory time-dependent?

Empirically-based-precise: to “explore whether the sex-related lateralization of amygdala function is time-dependent in [emotional enhancement of memory]” (2)

Zuo

Growing together and growing apart: Regional and sex differences in the lifespan developmental trajectories of functional homotopy

Exploratory

Second, studies were examined for the presence of reverse inferences in interpretation of findings: that is, speculations about the psychological meaning of observed sex differences in brain activity. If reverse inferences were made, it was noted first whether any relevant behavioral data were available. If so, it was then noted whether those data were consistent or inconsistent with the reverse inference. Twenty-seven of the 39 studies (69 %) suggested reverse inferences, speculating either a quantitative difference in male/female responsiveness to the experimental stimuli (e.g., “greater, more widespread activation may suggest that males recruit less expert neural systems when processing emotional stimuli” [81, p. 143]), and/or a qualitative difference in mental processing (e.g., the “pattern of findings is consistent with the higher emotional reactivity and emotion-focused coping style commonly observed among females … [observed male neural activity] is probably related to the cognitive effort and specific regulatory strategies that they had employed.” [82, p. 2907]). Behavioral data relevant to the reverse inference (e.g., performance data if greater male/female expertise was speculated from brain activation differences, or behavioral measures of strategy differences if those were inferred) were available in 14 of the 29 studies in which reverse inferences were made.

Strikingly, however, in 11 of these 14 studies the relevant available behavioral data were inconsistent with, or unsupportive of, a reverse inference made (see last column of Table 4). For example, Klucken et al. [83] looked at brain responses to conditioned sexual stimuli, and found greater neural response in males in the amygdala, brainstem, thalamus and occipital cortex. They suggested that these neural differences reflect the greater capacity of men to be conditioned for sexual arousal. However, their behavioral data contradicted this suggestion. Despite the fact that the female participants found the unconditioned stimuli (erotic pictures) less positive and sexually arousing than did men, they nonetheless rated the sexually conditioned stimuli similarly to men on arousal, valence and sexual arousal. If anything, this indicates a greater female capacity for sexual conditioning. Similarly, Lee and colleagues speculated that observed male/female brain activation differences suggested that women “need to partake in a higher degree of mental consideration before making a risky response, and render women more risk averse” [84, p. 1308], even though they found no sex differences in either response time to make, or number of, risky responses. In both cases, it appears that, in interpreting their highly ambiguous neurological data, assumptions about likely sex differences in behavior (in sexual conditionability and taste for risk) had greater influence than the similarity actually observed.

Other interesting examples of psychological speculations unsupported by behavioral data were provided by two studies investigating hormonal influences on female brain activity. Derntl et al. [85] looked at brain activity during the performance of a number of empathy tasks and found a non-significant trend for greater amygdala activity in six women in the follicular phase of the menstrual cycle, compared with six women in the luteal phase. Given the role of the amgydala in socio-emotional processing they suggested that “the follicular phase facilitates sensitivity and behavior in socio-emotional situations [references removed]” and that this facilitation “may be traced back to the evolutionary advantage of higher attention and responsiveness to social-emotional interactions thereby improving mating chances during times of increased fertility.” [85, p. 78]. However, no differences in empathizing ability were observed between women in different phases of their cycle. A second study investigated the effect of menstrual cycle phase on brain responses to aversive stimuli [86]. The authors reported greater attenuation of brain activity in subcortical regions, including the amygdala, relative to males, in twelve women when in the late follicular/mid-cycle phase compared with the early follicular phase. From this finding the authors suggested that hormonal changes in women may influence stress responses such that the mid-cycle female can “optimally judge whether a potentially threatening stimulus, such as an approaching male, is an opportunity for successful mating or for fight or flight.”[86, p. 437] However, if the neural changes wrought by mid-cycle hormonal status reduce stress responses in a way that has behavioral influence, one would expect these women to have displayed less anxiety or other negative emotions than when in the early follicular phase. Yet this was not the case, either before or after viewing seventy-two highly arousing aversive pictures.

These two speculations, both based on extremely small sample sizes, are of particular note given the long history of prejudice against women based on the supposed psychological effects of their hormones [8]. Considered together, they also illuminate the problematic nature of a piecemeal approach to research. How, for example, would the amygdala of the fertile heterosexual female be expected to respond to an attractive but potentially dangerous male? Would it be predicted to show a greater response to facilitate more effective flirtation, or reduced response to facilitate boldness in the social interaction?

Exaggeration of Fixedness? Plasticity of Brain and Mind

A third important issue when it comes to interpreting sex differences in the brain is the potential plasticity of sex differences in both brain and mind. Neural circuitry develops through, and is altered by, experience [e.g., 8790], and this has been shown to be the case even for low-level sex differences in the rat brain [91, for discussion of this point and its neglect in research, see 92]. As a number of feminist scientists have pointed out, gendered life experiences and social constructions of gender (such as leisure activities, educational interests, poverty and status) have material effects on the body, including the brain [e.g., 12, 9395]. Meanwhile, behavioral sex differences have often been shown to vary by historical period, culture, demographic factors, ethnicity, and even in response to subtle social cues. To take the example of mathematics, in the United States the gender gap in mean mathematical performance at school has reduced and closed over the past few decades, while the sex difference at very high levels of mathematical performance has been found to vary across time and place as well as across ethnicities within North America [9698]. In addition, social psychological research indicates that mathematical performance can vary within the same individual across different social contexts. For example, investigations of the ‘stereotype threat’ phenomenon find that female mathematical performance is improved when, for instance, tests are presented in a way that de-emphasizes or makes irrelevant the stereotype that females are poor at maths [99, 100, although for a third meta-analytic study that eliminates all studies without male comparisons and/or that use preexisting mathematics scores as a covariate, and comes to a different conclusion, see 101]. This is just one example of the general phenomenon that social context affects gendered behavior in myriad ways [for numerous other examples, see 10]. Clearly, such social influences have neural correlates.

The implication of this for researchers is that the effects of gendered experiences or situations on brain development or function are (literally) incorporated into a sex difference in neural activity. Thus a single ‘snap-shot’ comparison of male/female brain activity is uninformative as to the extent to which any neurological sex differences (and any concomitant behavioral differences) are due to the effects of experience on brain development and function, versus the manifestation of fixed, stable and universal male/female neural signatures [18]. Importantly, while a ‘snap-shot’ approach to research doesn’t explicitly endorse the idea of fixed sex differences in the brain (and behavior), such an approach can neither produce data to challenge such an interpretation, nor illuminate or support any other account of how behavioral sex differences might arise. Conversely, a ‘plasticity’ approach that devotes attention to the neural correlates of the known malleability of sex differences in behavior could potentially do so. Thus, if charges of neurosexism are incorrect, researchers should avoid over-reliance on simple ‘snap-shot’ comparisons.

To investigate whether this was the case, each study from the previously used sample was categorised as taking a ‘plasticity’ approach if it examined whether potentially relevant gender-related experiences (such as masculine or feminine occupations, educational history or past-times), social factors (such as socio-economic background or marital status) or context (such as presence or absence of stereotype threat in the portrayal of the task) moderated sex differences in brain activity. Otherwise, it was categorised as ‘snap-shot’. All 39 studies took a ‘snap-shot’ approach to research.

Discussion

The brain science of male/female difference has an unfortunate history of neurosexism, in which scientifically unjustified claims furnished support for traditional gender stereotypes and roles. Some have recently argued that FNI investigations of sex differences largely follow that tradition. Conversely, others have responded to such criticisms with claims that they over-emphasize the peril of such research, at the expense of the current and future promise. This article examined the navigation of three obstacles to obtaining valuable information about sex differences, as a way of illuminating the validity of these two differing perspectives.

First, treatment of the scope for false-positive errors in FNI sex differences research was systematically explored in three ways: typical sample sizes, a case study, and citation practices. Analysis of a recent two year (2009 and 2010) sample found that findings of sex differences in fMRI investigations are predominantly reported in studies with inadequate sample sizes. Nearly three quarters of studies had 15 or fewer participants of each sex in each of their subgroups of interest, and for approximately a fifth of the studies there were fewer than ten. Empirically based analyses of group differences arising from samples of this size indicate low reliability, sensitivity, and generalizability [38, 42], a point underlined by the example of reports of sex differences in language lateralization discussed previously [37, 43, 44]. That the majority of reported sex differences involve studies with low statistical power that compare groups with fewer than 16 participants suggests that there is not great concern, in general, as to the possibility that claims of sex differences in brain activations may be spurious. Second, treatment of the scope for false positive errors was also investigated with an examination of standards of evidence for a recent claim of sex differences in the brain described as responsibly made [21], beyond reasonable doubt [49] and of sufficient warrant to be cited without caveat in some expert reviews [e.g., 52]. This found that only two supporting studies directly compared the sexes, and one of these suffered from sample selection bias. Both had small sample sizes and reflected ambiguity as to how to construct the dependent variable. Finally, concern with respect to false-positive errors was investigated in a third way by looking at recent citations of a sex difference in language lateralization for which considerable contradictory data exists, including two meta-analyses. Fifty-seven per cent of citing authors did not refer to the presence of any contradictory data, and fewer than a third cited one or both meta-analyses, approximately half doing so in a way that did not indicate its greater scientific warrant. In short, the picture revealed from these three investigations was one in which evidentiary standards are tolerant of both the production and persistence of false-positive claims of sex differences.

The second approach to judging the validity of the two perspectives of the literature looked at how researchers typically navigate the functional obscurity of brain differences, and the room for gender stereotype-infused speculations this allows. The extent to which researchers speculated on, rather than systematically investigated, the psychological significance of sex differences in brain activity was investigated in the same recent sample of FNI studies. None of the thirty-nine studies tested a prediction based on an explicit neurocognitive model of sex-modulated mental processes subserving the behavior of interest. Over two-thirds of the studies speculated on psychological differences in the sexes on the basis of neurological findings, all but three doing so either in the absence of supporting behavioral data or in the presence of contradictory data. While caution and restraint were in evidence in a number of studies, it was still the case that post hoc reverse inferences were more common than not. This is not surprising since, ultimately, such research is intended to illuminate psychological differences between the sexes. Yet when such speculative reverse inferences are not part of systematic model building, the scope for influence of erroneous gender stereotypes is high. This was especially clear in cases in which speculations were consistent with gender stereotypes but inconsistent with the researchers’ own behavioral data.

Finally, using the same sample, researchers’ attention to the potential plasticity of sex differences in brain and mind was investigated. Studies were categorised on the basis of whether they explored the potential plasticity of sex differences in brain activity by looking at how factors such as systematically gendered experiences, demographic factors or social context might contribute to, or moderate, sex difference in brain activation; or whether they relied on a single ‘snap-shot’ of male/female difference. All 39 studies took a ‘snap-shot’ approach.

These findings seem to render untenable claims that charges of neurosexism in scientific research are misplaced; at least with regards to the FNI literature. The long-noted publication bias toward positive findings of sex differences is exacerbated by tolerance for positive findings from sample sizes that yield low reliability and generalizability, as well as less than ideal citation practices. While a report of a sex difference in the brain, even if erroneous, does not necessarily support gender stereotypes (particularly given the possibility that neural differences may have no behavioral significance), functional interpretations of apparent neurological sex differences are common, and little effort appears to be devoted to the development of neurocognitive models to constrain and test such speculations to avoid gender stereotype-infused interpretations. Lastly, a predominantly static ‘snap-shot’ approach to research guarantees that no data that challenge the implicit assumption of fixed male/female neural signatures will be found, and that the processes underlying experiential contributions to, and plasticity of, neurological or behavioral gender differences remain unexplored. Together, these common neuroscientific research practices bias the scientific literature toward a presentation of sex differences in the brain as “essential”; that is extensive, functionally significant in gender stereotypically consistent ways, and fixed.

It might be objected that the conclusion to be drawn from the information presented here is that cognitive neuroimaging, like all science, is imperfect—not that it is sexist. It is certainly true that issues such as false-positive errors, within-group analyses, citation bias, and logically and empirically unpersuasive reverse inferences are hardly unique to sex differences research [39, 45, 102, 103]. However, that these problems affect FNI investigations and behavioral science more generally is not grounds to dismiss charges of neurosexism. The reason this objection fails is a feature of the research so obvious that it is all-but invisible: theoretically, the science could carry exactly the same flaws but be biased in precisely the opposite way. Thus, in theory, it could be false-positive claims of the effects of gender socialization (rather than biological sex) on brain activity that are facilitated through common research practices. Similarly researchers could, on the basis of ambiguous neurological findings, frequently speculate about the psychological effects of socialization or context on gendered behavior in a piecemeal fashion that is not developed into explicit neurocognitive models and systematically tested. Lastly, researchers could routinely neglect to investigate the influence of biological sex on neural activity during the performance of gendered tasks, guaranteeing that no data are produced that could contradict the notion that gender is entirely socially constructed.

There are social, as well as scientific, reasons for the neuroscientific community to acknowledge the validity of feminist criticisms of the research: arguably, the relatively subtle forms of neurosexism identified here contribute to the more egregious versions that appear in the popular literature. Recently, a number of scientists have drawn attention to the harmful effects of popular neurosexism in reinforcing gender stereotypes [e.g., 9, 10, 21, 104, 105]. Yet research is currently being conducted in a way that biases the literature towards presenting neurological sex differences as numerous, functionally significant, and fixed. Thus, even adequately accurate popular dissemination of the neuroscientific research literature would be biased towards interpretations that implicitly reify the status quo.

Moreover, neuroscientists could also usefully consider other ways in which research may contribute to, and license, popular forms of neurosexism. First, the greater the volume of false-positive errors that exist in the literature, the more scope there is for popular writers to make claims about sex differences in the brain that will ultimately prove to be false. Clearly, false-positive results are inevitable, but the scope for them is greatly increased by current common practices. In addition, the scientific community itself does not always set a good example in terms of how such findings should be treated. Eliot [104], for example, has recently rightly criticized popular writers for cherry-picking spurious findings, using as an example references in the popular media to Shaywitz et al.’s finding of greater male lateralization, despite the null conclusion of the two meta-analytic investigations of the hypothesis. Yet as noted earlier, the same phenomenon is present in more than half of recent citations in the scientific literature itself. A similar argument applies to functional interpretations of supposed male/female brain differences. While some claims made by popular writers appear to be the products of enviably potent imaginations, in other instances they are recognisable (albeit exaggerated and over-confident) versions of hypotheses, interpretations and speculations made by researchers themselves [see 10]. Anecdotally, scientists have a reputation for over-cautiousness in the statements they are prepared to make, thus it seems plausible that the propensity of scientists to make such speculations, to draw on gender stereotypes to do so, and to fail to test them systematically (and thus potentially reject them), will increase the ease and apparent appropriateness for popular commentators to make such speculations themselves. Finally, a number of scientists have appropriately criticized popular writers for false assertions that any sex differences observed in the brain must be ‘innate’ or fixed [e.g., 10, 21, 104]. Yet the ‘snap-shot’ research approach implicitly reinforces—or at the very least fails to challenge—the notion of fixed male/female neural signatures. If the neuroscientific literature were full of studies exploring the plasticity of sex differences in the brain—how, for example, they vary across time, culture, socioeconomic status, gendered experience, social context, and so on—popular writers would be less able to assume or assert that supposed neurological differences are ‘hardwired’. In line with this point, Eliot [105] has recently argued that neuroscientists should pay more attention to gender enculturation processes in the brain. The current findings strongly reinforce this suggestion.

It is of course not suggested that scientists bear blanket responsibility for popular representations of their work, and their responsibility at the popular interface is a difficult question beyond the scope of this article. However, in the face of evidence of the manifestation of neurosexism in the scientific research itself, it behoves the scientific community to consider their contribution to popular scientific understanding of gender through the proliferation of false-positive errors [see also 106], untested stereotype-consistent reverse inferences, and lack of attention to the potential plasticity of sex differences.

Then, there are scientific reasons for the neuroscientific community to acknowledge the validity of feminist critiques. Concerns that the influence of such criticisms on neuroscientists may “hinder” or even “reverse” [22] scientific progress seem entirely misplaced. FNI investigations of sex differences can improve in their service of the internal goal of science, which is to establish reliable and valuable knowledge about nature. It is worth noting that none of the feminist critics cited here have suggested, as a solution to this, the termination of such research: to argue that the research is often flawed is not equivalent to arguing that it should not be done at all. To the contrary, taking heed of charges of neurosexism in each of the three domains discussed would serve to improve the quality of the science.

First, scientific progress is not served by the proliferation of false-positive errors, the premature acceptance of hypotheses, or the perpetuation of false-positive findings through citing practices. Second, although exploratory research and post hoc speculation clearly have important roles in science, to reduce scope for unsubstantiated reverse inferences and, more generally, better serve scientific progress, this should lead to the systematic development and testing of neurocognitive models of sex differences. In an interesting observation that spans both these issues, Bluhm [14] has pointed out that a statistical technique known as region-of-interest (ROI) analysis, in which brain activity differences are sought in pre-specified regions of the brain, is common in this area of research. Because ROI analyses involve far fewer comparisons than whole brain analyses, researchers are able to lower statistical thresholds for significance. While the aim is to increase the chance of finding a group difference, it also increases the risk of false-positive error. Moreover, because ROI analyses examine only isolated regions of the brain, Bluhm notes that they serve to act against the development of sophisticated neurocognitive models, which require an understanding of the functional relationships of the entire brain network involved in such tasks, and the relation of different parts of the network to differences in behavior. Thus, the decision to use this analysis technique implicitly prioritizes the finding of a sex difference over theoretical advance and the avoidance of false-positive errors.

Third, similar arguments can be made as to the benefit of taking into account criticism of the lack of attention to the plasticity of sex differences. In humans especially, gendered behavior shows substantial within-group variation that is influenced by factors such as historical period, nationality, experience and social context. Neuroscientific research will only contribute to rich explanations of gender if research is guided by a sophisticated conceptual understanding of the phenomenon. As Springer, Stellman and Jordan-Young have recently argued, “measures of sex are not pristine, but include effects of gender” [93, p. 1] meaning that insights into gender differences require going “[b]eyond a catalogue of differences” to instead investigate the biological mechanisms that interact with gendered experiences, by which differences between sexes arise. Jordan-Young & Rumiati [13] have recently made recommendations along such lines for brain research, suggesting greater exploration of neural plasticity, as well as into the ways variation arising from other social categories—such as social class, occupation, and nationality—impact gender differences. As they point out, such a research approach “might provide much more illumination on the concrete mechanisms through which the social world shapes behavior, and even becomes embodied (brain) difference” (p. 23).

In short, it is hard to see how taking heed of charges of neurosexism could possibly impede scientific progress in this area. To the contrary, it should hasten it. In the absence of changes in current practices, FNI investigations of sex differences will continue to underperform in service of the internal goal of science, at the cost of contributing to erroneous and socially harmful scientific and pseudo-scientific claims and speculations. It is hoped that the neuroscientific community will respond to the arguments presented here, and other such critiques, with openness as to how research practices might be changed so as to improve the quality of the science in this important and socially sensitive area of research.

Footnotes
1

This was done using the search terms “sex” or “gender” in title, and “fMRI”, “functional magnetic resonance imaging” or “functional MRI” in the title, abstract or keywords. ‘Difference’ was not included as a search term so as to not exclude studies reporting sex/gender similarities. Studies that were not full reports of original findings that investigated sex differences in brain activation in humans were then excluded.

 
2

Excluded from the analysis was one study, referred to earlier, that was specifically testing the generalizability of sex differences based on a total sample size of 20 [42].

 
3

It’s noteworthy (or at least footnoteworthy) that two of these three articles are specifically concerned with problems arising from neuroimaging studies of sex differences in language lateralization.

 

Acknowledgments

My warmest thanks to Martha Farah, Fiona Fidler, Kit Fine, Nick Haslam, Anelis Kaiser, Neil Levy, Carsten Murawski, and Danielle Pogos for their very helpful feedback on earlier versions of this paper. This research was supported in part by an Australian Research Council Future Fellowship.

Copyright information

© Springer Science+Business Media Dordrecht 2012