A while ago, I posted an overly-long discussion of recent research on the ‘math gap’ between boys and girls on standardized testing (Girls closing math gap?: Troubles with intelligence #1). That posting discussed several studies published in Science that have shown the gap in average math scores between boys and girls is not set in stone. In one paper, an increase in the test pool brought on by the No Child Left Behind program, with mandatory universal tests instead of exams only for those wishing to go to college, caused the gap in average scores to disappear; in the other paper, a decrease in the ‘math gap’ was found to correlate with other measures of greater gender equality in European states.
As I pointed out in the previous post, however, many commentators suggest that it is not the gap in average test scores that really matters; rather, these critics argue that the different variance in boys’ and girls’ scores explains the disproportionate number of boys who produce exceptional scores (as well as exceptionally bad scores), and thus the marked gap of men and women in PhD math programs, in prestigious prizes for physics and related subjects, and in related fields like engineering. In the earlier post, I argued that even if this greater variance showed up reliably across all testing populations, what exactly was being illuminated was still not clear; that is, many other explanations–other than that men had better ‘math modules’ in their brains, or greater ‘innate’ mathematics ability, or something like that–could explain even very stable differences in math performance. At the time I suggested a number of other possibilities, such as sex differences in stress response during testing, as other possible explanations for even a universal ‘math gap’ (which still had to contend with studies like the two in Science which severely undermined the assertion of universality).
As if on cue, I stumbled upon a video and accompanying article in Science Daily on differences in stress responses among men and women: Neuroscientists Find That Men And Women Respond Differently To Stress (but don’t click on that link — keep reading!). Stress is a good candidate to explain a test-taking gap because the observable physiological processes offer abundant evidence that men and women don’t respond to stress in exactly the same way (although there are underlying commonalities). For example, stress causes different diseases in men and women, and some long-term psychological disorders that demonstrate sex-linked disparities seem to emerge from stress.
Unlike the ‘black box’ explanation that boys and simply better at math or evidence greater variability in innate ability, with no observable neural correlate or plausible explanatory mechanism, in variation in stress response we have a clear candidate for male-female difference that plausibly affects their performance and even physiology (for example, in different stress-related diseases).
Ironically, when I tracked down the original article that the Science Daily piece was likely based on (there’s no citation, so I can’t be certain), I had to delete all the quotes from the Science Daily article from the draft I was writing for this post (that’s why you shouldn’t link to it).
I give science writers a bit of stick from time to time, but in this case, the explanation of the research was not merely misunderstood, it was simply wrong, not even consistent with the abstract from the article I think the popular piece is based on (Wang et al. 2007). So even though the erroneous Science Daily article put me onto this thread, I’m only going to work from the piece published by Jiongjiong Wang and colleagues at the end of 2007 in the journal, Social, Cognitive, and Affective Neuroscience.
Stress responses in men and women to arithmetic tasks
In the abstract, Wang and colleagues explain that they tested 32 subjects using both fMRI and endocrine salivary screening. In the experiment:
Psychological stress was elicited using mental arithmetic tasks under varying pressure. Stress in men was associated with CBF [cerebral blood flow] increase in the right prefrontal cortex (RPFC) and CBF reduction in the left orbitofrontal cortex (LOrF), a robust response that persisted beyond the stress task period. In contrast, stress in women primarily activated the limbic system, including the ventral striatum, putamen, insula and cingulate cortex.
The researchers also found that the men tended to have more intense responses in the hypothalamic-pituitary-adrenal (HPA) system (e.g., the release of cortisol). According to the researchers, the increased activity in the men’s right prefrontal cortex (RPFC) under the acute stress condition was the most significant finding of their experiment, forming a biomarker for a distinctly masculine acute stress response.
What to make of all this? The researchers use a general description of these two different stress responses, first proposed by Taylor et al. (2000), as ‘flight-or-fight’ in men and ‘tend-and-befriend’ in women. At first read, I just groaned, and I’m still opposed to the essentialist and evolutionary mythology being touted as explaining an observable difference in performance. I’m not even going to start down the well-trod path I’ve beaten criticizing ‘evolutionary psychologists’ for naturalizing observed differences between men and women, simultaneously conjuring away the problem of explaining these differences by assuming that they are ‘inherent’ and attributing them to some dramatic fantasy of evolutionary selection — we’ve been here before (say at Girls gone guilty: Evolutionary psych on sex #2 or Chicks dig jerks?: Evolutionary psych on sex #1).
In fact, their data is much more interesting. For example, Wang and colleagues point out that the difference in brain activation patterns might result from different stress coping strategies or from different response to high- and low-stress situations (see Wang et al. 2007: 237-8). Ironically, as Wang and colleagues discuss, the difference in men’s and women’s stress responses depended upon the stimulus being used to produce stress: some experiments used social rejection; others, like the Wang-led team, use arithmetic problems to create stress. Likely, different sorts of stressful situations produce subtle distinctions in stress responses, some evoking more social anxiety, for example, and others creating a greater sense of physical peril. The group offers a path for future research, suggesting that neuroimaging studies of stress responses on different sorts of cognitive tasks might help sort out what’s specific to mathematical problems or might be more general difference between men and women: ‘Given the sensitivity of stress responses to specific context and intensity, we are cautious to generalize the current finding to different types of stress’ (ibid.: 238).
These different stress responses likely affect other mental activities in a variety of ways; we know that not all responses, in parallel fashion, affect health or cortisol production or other physiological correlates of stress. In addition, it’s quite likely that men and women don’t read the situations as equally stressful, either for innate reasons or for encultured ones–the two would be very difficult to disentangle in adults as the physiological effects would be identical. For example, girls and boys might interpret a testing situation in different ways because of peer, family, social, or other dynamics.
But I’m just going to hold my nose about the ‘ev psych’ part of this and plow onward (my objection being the ontogenetic simplifications of how a trait might emerge rather than a phylogenetic objection to saying that men and women might have been subjected to different evolutionary pressures — someday I’m going to have to do a post on this). So, onward with nose held…
Differing stress responses: ‘fight-or-flight’ or ‘tend-and-befriend’
The ‘fight-or-flight’ response ‘invokes resources that increase focus, alertness and fear, while inhibiting appetitive goals to cope with the threat or challenge’ (Wang et al. 2007:236). This pattern shows up in the increasingly active RPFC, associated with vigilance and negative emotion, and the suppression of activity in the LOrF, linked to hedonic behaviour and positive feedback. In other words, in the ‘fight-or-flight’ response, according to this interpretation, an individual becomes very alert to potential dangers and anticipating dire consequences, much less capable of focusing on pleasure-seeking.
Female response, in contrast, ‘primarily involves the limbic system including ventral striatum, putamen, insula and cingulate cortex’ (ibid.). This pattern was labeled by Taylor and colleagues ‘tend-and-befriend,’ and included parts of the brain receptive to oxytocin, vasopressin, dopamine and endorphin, systems that have been linked in previous research to social relations, attachment, and maternal behaviour. The researchers suggest that this social rewards system may blunt the acute stress response, leading women to respond in similar fashion to both high- and low-stress situations.
Wang and colleagues do point out that there are a lot of parallels between male and female responses, including very similar endocrine response, in spite of the predictions of the ‘fight-or-flight’/’tend-and-befriend’ contrast. The point being that, as in many human traits, male and female differences tend to appear in some lights as oppositions, but upon closer examination often reveal instead a largely common, underlying pattern (although the RPFC response was distinctive of their male subjects). Even the differences that do exist, such as a divergence in cortisol feedback due to the effects of reproductive hormones, may or may not be linked to observable differences, such as patterns of ‘ruminative thinking,’ as the researchers discuss. And the study itself didn’t turn up some patterns of activation that the researchers expected, such as a more prominent role for the amygdala (see, for example, Paul Mason’s discussion of the Role of Emotions in Brain Function).
Stress and cognitive function
Turning away from the more general effects of stress, focusing instead on the parts of the stress response that might affect must profoundly cognitive processing, we find that:
Activation of RPFC and right parietal regions [the pattern more pronounced in men] has been associated with various cognitive control tasks, including working memory, response selection and task switching, as well as inhibitory functions …. Ventral striatum along with several limbic regions [both pronounced in women under stress] have also been involved in learning in addition to tasks related to reward, motivation and emotion…. The different computational roles subserved by these brain regions may contribute to the observed gender differences in central stress responses. Although somewhat controlled in the regression analyses, this possibility (e.g. inhibiting incorrect responses in males and updating task strategies in females) cannot be completely ruled out, especially in the direct comparison of average stress responses between men and women.
Wang and colleagues specifically indicate recent research on mathematics and science ability (Hyde and Linn 2006) and suggest that the variation in stress response might underwrite pronounced differences in test results.
The RPFC, for example, is especially associated with executive functions (such as inhibiting emotional responses) and with strategic thinking, which might suggest a particular pattern of responding to stress. If women were more likely to focus on updating the possibility of reward, their own emotional states, and their motives while under stress, this might lead to lower scores when they were stressed by a time-constrained testing format. I’m still not persuaded that this is innately male or inherently impossible for women to achieve. If this is a pattern of brain activity especially likely to lead to certain test scores, and if some women are able to achieve extremely high test scores, than perhaps some women are able to learn this cognitive strategy. After all, it’s not like women don’t have these same parts in their brains. Likewise, I suspect that you could train patterns of response to stress in boys; certainly, the people I work with in sports training are convinced this is the case.
Implications for test-taking
First off, the experiment was ideal for exploring a possible stress-related contributor to the ‘math gap’ because it actually used mental arithmetic as the stimulus. Although not a perfect fit for a test-taking environment (where you don’t do the arithmetic in your head while holding still in a giant fMRI scanner), I doubt we’re going to get much better than this until the imaging technologies make some major jumps.
Second, in general, women after puberty have a lower threshold for perceived stress. The irony is that perceiving that one is stressed can often exacerbate one’s stress response. If you think you are anxious, you can make yourself more anxious. One can easily see how this might affect women’s performance on standardized tests. I never recall being all that stressed out on things like the SAT or GRE (I think I fell asleep during the GRE when I finished one section early), but for some people, this significantly influences their performance. I need to point out that Wang and colleagues specifically controlled for this effect in their research, so it’s not such a factor in their data, but it’s not hard to imagine that, especially given the length of a standardized test and the possibility that difficult questions might heighten stress during the course of an exam, this might become a factor in test score differences.
But the bottom line is that, if boys’ and girls’ brains respond differently to stress, this might disproportionately affect tests exploring various subjects. Female stress response might make their mathematics problem solving drop off more than men’s in timed tests (of course, we still have the much-less-discussed ‘reading gap’ to deal with, too). This sort of pattern does seem to show up in the gap between women’s performance on standardized testing in relation to men’s, and their across-the-board higher averages in marks on university courses (since broader admission of women into universities).
But stress might also lead to a narrowing of girls’ variance on standardized tests. It might diminish ability to perform at an extremely high level, but it also might lift the scores of the lowest scoring, least-motivated individuals. Lack of stress on a standardized math test — the situation boys might be more likely to find themselves in — might improve some young men’s scores, but it also might lead low performers to be even more blasé about their situation. Even the high RPFC activation in men might show up in boys as a clear-headed strategizing about the irrelevance of doing well on a standardized test if they know that they are not high performers in mathematics.
In other words, even a different pattern of brain activation (elevated activity in the RPFC) may simply provide the emotional environment in s stressful environment in which a boy might perform especially well, or calmly assess that the exercise was pointless given what he already knew about his ability. The result would be indistinguishable from an innatist argument that ‘boys have higher variance in innate math ability,’ but the underlying causal mechanism would be subtly different (and thus require different remedial projects if someone tried to address the variation).
One indicator of the possibility of these sorts of subtle mechanism that a Scientific American article by Halpern and colleagues cites is the fact that preschool children score similarly on cognitive tests of quantitative thinking and geometrical reasoning. The start to diverge when the children get to school. Innatist explanations suggest that the ‘true nature’ of boys and girls emerges when they enter school, but it’s just as likely that peer dynamics, including strong sex-stereotyping among kids, starts to really kick in when they are exposed to school. My point is not to argue that an innatist position is untenable, only that the pattern we see is equally consistent with other ways of thinking about how differences might arrive.
More on higher variance arguments
The Scientific American article also looks at the ‘higher variance’ of math ability argument that a number of proponents of innate ability gaps put forward (which I discuss at length in the previous post on math tests). The gap is profound, but the trend in that gap is also interesting. The authors reflect on data that was first assembled in the early 1980s on SAT scores:
There were twice as many boys as girls with math scores of 500 or higher (out of a possible score of 800), four times as many boys with scores of at least 600, and 13 times as many boys with scores of at least 700 (putting these test takers in the top 0.01 percent of 12- to 14-year-olds nationwide).
Although it has drawn little media coverage, dramatic changes have been occurring among these junior math wizards: the relative number of girls among them has been soaring. The ratio of boys to girls, first observed at 13 to 1 in the 1980s, has been dropping steadily and is now only about 3 to 1. During the same period the number of women in a few other scientific fields has surged. In the U.S., women now make up half of new medical school graduates and 75 percent of recent veterinary school graduates. We cannot identify any single cause for the increase in the number of women entering these formerly male-dominated fields, because multiple changes have occurred in society over the past several decades.
Although 3 to 1 is still a very large gap, it’s also startling to see that the gap can close from 13 to 1 to as little as 3 to 1 in a bit more than two decades. There may be an innate gap in math ability, but with all the change in these figures, it seems a bit premature to suggest that we know for certain that we’ve ascertained it and cannot affect change in the performance gap any further.
In addition, specialized courses designed to remedy the women’s specific deficits in visuospatial skills at the Michigan Technological University led to marked improvement among women in this area, one of the abilities considered to be strongly sex-linked, and to higher retention of women in university science and math programs. (I don’t have specifics on this intervention yet, but I will post more information when I get it.)
Again, this doesn’t prove that there are not innate differences between men and women: the origin of the gap in visuospatial skills is not at all clear. Because I’m more of a developmental systems theorist than an innatist, I would tend to look in the developmental trajectory of boys and girls for the difference rather than assume math ability springs from a gene or hormone. Thinking of the child as a developmental system, the gap may arise in an odd, indirect way; for example, boys relatively lower verbal abilities might lead them to compensate by developing visuospatial skills, or girls play patterns — whether due to innate tendencies or socialization — may give them less experience with visuospatial manipulation. Because the gap is mutable and the skills deficits at least partially remediable, I’d say that the burden of proof starts to fall on the innatists; show us where the innate visuospatial ability actually lives in the brain and how it comes into the world pre-destined if a whole host of studies are showing gaps are mutable.
In the end, I suspect that there are biological differences in girls’ and boys’ brains that contribute to differences in test score variance, but these differences may not be where we expect them. For example, they may have more to do with something that indirectly affects math testing like stress response or motivational structure in education. Innatist thinking is too easy, too inconsistent with the actual way that brains and cognitive abilities develop in an unfolding of the human organism in relation to a social and learning environment. There’s a lot of ‘mights’ in my account, but the fact that there are other plausible explanations for something like the math gap — even if it is universal (which the Science papers question) — shows overly glib assertions of innate difference to be a sloppy way out of what are really a whole set of interesting theoretical and empirical questions.
Halpern, Diane F., Camilla P. Benbow, David C. Geary, Ruben C. Gur, Janet Shibley Hyde and Morton Ann Gernsbacher. 2007 (November). Sex, Math and Scientific Achievement: Why do men dominate the fields of science, engineering and mathematics?. Scientific American (available online here)
Hyde, Janet Shibley, and Marcia C. Linn. 2006. Gender similarities in mathematics and science. Science 314 (5799): 599–600. (pdf available here)
Phipps, Alison. 2008. Women in Science, Engineering and Technology: Three Decades of UK Initiatives. Trentham Books.
Taylor, Shelley E., Laura Cousino Klein, Brian P. Lewis, Tara L. Gruenewald, Regan A. R. Gurung, and John A. Updegraff. 2000. Biobehavioral responses to stress in females: tend-and-befriend, not fight-or-flight. Psychological Review 107(3): 411–29. (abstract on Pub Med, pdf available here)
Wang, Jiongjiong, Marc Korczykowski, Hengyi Rao, Yong Fan, John Pluta, Ruben C. Gur, Bruce S. McEwen and John A. Detre. 2007. Gender difference in neural response to psychological stress. Social Cognitive and Affective Neuroscience 2(3):227-239. doi:10.1093/scan/nsm018