Back in 2005, you might have been forgiven for thinking that the gender wars were largely a thing of the past. The extreme blank slate view of earlier decades was losing its foothold as evolutionary psychology, neuroscience, and common sense steadily eroded the Nurture-Only explanation of sex differences. The political correctness that peaked in the 1990s seemed to have relaxed to the point where you could discuss differences between the sexes without fear of the sky falling on your head. Sure, there were still some gender radicals out there. But they no longer had the kind of sway they had in earlier decades.
Or so one might have thought. But 2005 was the year we learned that there were still limits to what you can say about sex differences, and severe consequences for stepping over the line. For 2005 was the year that Lawrence Summers, then president of Harvard University, was invited to give a speech at a Harvard conference on Diversifying the Science and Engineering Workforce.
You probably know the story. The aim of the conference was to explore the contentious issue of why there are more men than women in top positions in the STEM fields (science, technology, engineering, and mathematics), and to look for solutions to the perceived problem. Summers decided to use his speech to discuss some of the possible reasons for the gender gap.
He laid out several hypotheses. One was the standard explanation: that the difference is due to a combination of gender socialization (teaching children that science is for boys not girls) and discrimination (hiring and promoting men more readily than women). Another was that, for various reasons, more men than women are willing and able to commit to the extreme work hours required to make it to the top. But Summers made a third suggestion as well, and it was this that landed him in hot water. The suggestion was that men are naturally more variable than women in cognitive ability, and thus that there are more men than women at the top levels of cognitive ability (as well as more men at the bottom). This, he suggested, might help explain the preponderance of males in the top tiers of the STEM hierarchy.
Summers didn't say he wanted this to be true. Indeed, he explicitly stated that "I would like nothing better than to be proved wrong, because I would like nothing better than for these problems to be addressable simply by everybody understanding what they are, and working very hard to address them." Summers also didn't say that nothing could be done to remedy the situation. On the contrary, he outlined several potential remedies.
Nonetheless, his suggestion that there might be more men than women at the highest levels of ability did not go down well, to put it mildly. One conference attendee, Nancy Hopkins, was so upset she stumbled out of the room, dizzy and nauseous. Similar reactions soon spread beyond the conference walls. There were protest marches at Harvard and outraged op-eds denouncing Summers in the national press. Summers was branded a sexist and accused of lowering the self-esteem of girls all around the nation. In the end, and despite repeated apologies, his hypothesis probably cost him his job. He resigned a year later.
Not everyone called for his head on a platter, though. Summers had various allies and defenders. Some pointed out that if he’d said that women had some kind of cognitive advantage over men—better language skills, for instance, or a more flexible thinking style—no one would raised an eyebrow. No one would have accused him of lowering boys’ self-esteem, and some people would have cheered him on. Other commentators lamented the anti-scientific attitude underlying the response to Summers' hypothesis. The lesson seemed to be that the only acceptable explanation for the dominance of men in science is discrimination. Anything else is met not with evidence but with outrage. As the psychologist Jonathan Haidt noted, “We psychologists should have been outraged by the outrage. We should have defended [Summers’] right to think freely.” And on top of all that, many suggested that, like it or not, Summers was quite possibly right.
The Great Debate
So, was he right or not? It’s not easy to find a balanced, calm discussion of a highly-charged issue like this. A good place to start, though, is a public debate staged soon after Summers' gaffe. The debaters were Harvard psychologists Steven Pinker and Elizabeth Spelke. Spelke argued that the sex difference is 100% nurture; Pinker argued that nature and nurture both contribute, but focused on possible biological contributions. The debate took place in 2005, but it’s still worth watching. In this post, I’ll summarize and evaluate the main arguments on both sides, adding my own thoughts and more recent data along the way.
The differences in the two speakers’ outlooks emerged early on. In the lead up to the debate, Spelke stated that there is “not a shred of evidence” that biology is involved in the STEM difference. As she put it, “The evidence against there being an advantage for males in intrinsic aptitude is so overwhelming that it is hard for me to see how one can make a case at this point on the other side.” This conclusion, she added, is “as conclusive as any finding I know of in science.”
Here is Pinker’s response to these statements:
“Well we certainly aren’t seeing the stereotypical gender difference in confidence here! Now, I’m a controversial guy. I’ve taken many controversial positions over the years, and, as a member of Homo sapiens, I think I am right on all of them. But I don’t think that in any of them I would say there is 'not a shred of evidence' for the other side, even if I think that the evidence favors one side. I would not say that the other side 'can’t even make a case' for their position, even if I think that their case is not as good as the one I favor. And as for saying that a position is 'as conclusive as any finding in science' —well, we’re talking about social science here! This statement would imply that the extreme nurture position on gender differences is more conclusive than, say the evidence that the sun is at the center of the solar system, for the laws of thermodynamics, for the theory of evolution, for plate tectonics, and so on. These are extreme statements—especially in light of the fact that an enormous amount of research…in fact points to a very different conclusion.”
What If There Are Innate Differences?
Before looking at some of this research, we need to get clear about a number of things. The first and most important is what it would imply if Summers turned out to be right and the Nurture-Only position on cognitive sex differences turned out to be false. Would this undermine our commitment to gender equality? The answer is that it would not. As Pinker points out, there are two distinct issues here:
Our commitment to the moral principle (“don’t discriminate”) should not be made dependent on the outcome of the empirical question. Otherwise, notes Pinker, if it turns out men and women aren’t biologically identical, we’ll then have to conclude that discriminating on the basis of sex is OK after all…or we'll have to suppress the facts. Neither of these options is desirable. And nor is either necessary. A much better option would be to say that, whether or not men and women are identical, we shouldn’t discriminate on the basis of sex—in science or in any other sphere of life.
A Matter of Stats
With that in mind, we next need to get clear about the phenomenon under discussion: the STEM sex difference. In the immediate aftermath of Summers' speech, Nancy Hopkins said: “It’s not appropriate for the man who holds in his hands the future of the brightest minds in America to say that 50% of them don’t have the right aptitude” for science. But this is a misunderstanding of Summers' position. And it points to a broader misunderstanding about women’s representation in science.
Let's start with a simple fact: Most women do not have the right aptitude to be professors at top STEM departments. This is unfortunate, perhaps, but it’s true. It’s also true, though, that most men don’t have the right aptitude! Only a small minority of people do. The phenomenon we’re trying to explain is not why half the population (men) can do it whereas half the population (women) can’t. Most of the population can’t, and of the tiny fraction who can, some are men and some are women. The only question is: Why is the tiny fraction of men working in STEM fields today somewhat larger than the tiny fraction of women?
Explaining the Differences
The fact that these fractions differ at all is often treated as direct evidence of discrimination against women (see here). It’s important to remember, though, that equal opportunities don’t necessarily result in equal outcomes, and that differences don’t necessarily imply discrimination. Discrimination is certainly a possibility, but it’s not the only one. There are at least three others.
These suggestions are not mutually exclusive. All three factors may have a role to play. Let's look at each in turn; then we’ll return to the question of discrimination.
Small Average Differences in Relevant Aptitudes
The first possibility is that the sex difference in STEM reflects average differences in certain aptitudes relevant to a STEM career. This sounds horribly sexist, so let’s be very clear about what’s being claimed.
First, the claim is not that all men are better than women at the relevant tasks, or even that most are. There is a wide range of abilities within both sexes, and a great deal of overlap between the distribution for men and the distribution for women. The claim is simply that the average score—that is, the central tendency of the distribution—is slightly higher for men than for women (see the first figure below).
Second, the claim is not that men do better on average in every area. Some sex differences favour men; others favour women. So, for instance, whereas men’s average score is higher for spatial tasks, women’s is higher for linguistic tasks (see here, here).
Third, the claim is not that average sex differences are necessarily very large. Even if men greatly outnumber women at the highest levels of ability, the sex difference will be much smaller nearer the mean—i.e., for most people. This reflects a fundamental property of normally distributed data: Even a small difference at the mean is associated with a much larger difference at the extreme of the distribution (as shown in the two figures below, both from Pinker's presentation). To illustrate, consider the sex difference in height. At 5-foot-10, there are 30 men for every woman. At six feet tall, however, there are 2,000 men for every woman.
The same principle applies to differences in STEM-relevant aptitudes. Even a trivial average difference between the sexes in a relevant trait will usually translate into a large sex difference at the tail of the distribution.
And there is good evidence for sex differences in aptitudes relevant to STEM fields. I've already touched on the difference in visual-spatial abilities. From puberty, men are better on average at mental rotation and other spatial tasks. Psychologists have known this for a long time, and it’s not an especially controversial finding. For some reason, though, people rarely suggest that it might have anything to do with the skewed sex ratio in jobs requiring exceptional visual-spatial skills.
In addition to the spatial sex difference, there may be relevant differences in certain mathematical abilities. As Pinker notes, the findings here are a bit complex. These days, girls get better school grades in mathematics (and indeed in almost all subjects). Furthermore, women are better at mathematical calculation than men, on average. However, in aptitude tests of mathematical reasoning, and of verbal mathematical problems, men’s average score is higher than women's. These skills are particularly relevant for math-heavy fields such as theoretical physics. So, although most men will not have the aptitude for these fields, somewhat more men than women will.
Of course, it might be argued that the differences result entirely from nurture, and that nature plays no role. Spelke points out, quite rightly, that complex mathematics is not part of our evolutionary endowment. Instead, it is a cultural invention that supervenes on lower-level abilities, such as the ability to represent small numerical quantities and to represent the geometrical layout of one's environment. But, notes Spelke, there are no sex differences in any of these basic mathematical competencies among infants or young children. Any differences appear only later. This, she argues, contradicts the idea that the differences trace to biology. It suggests instead that they emerge only once the forces of socialization sink their claws into us.
Unfortunately, the conclusion doesn't necessarily follow. The fact that sex differences are absent in infancy doesn't rule out a genetic explanation. Many sex differences emerge only at puberty. Indeed, that’s an important part of the definition of puberty. Furthermore, when we look closely, the usual explanations for the math differences—stereotype threat, math anxiety, Barbie dolls that say "Maths class is tough!"—are difficult to sustain. If girls are led to believe that they’re worse than boys at math, why do they get better grades in math class at school? If stereotype threat and math anxiety undermine their test-taking abilities, why does this happen on tests of some skills but not others? Is there a stereotype that girls are better at mathematical calculation but that boys are better at mathematical reasoning? Probably not. And in any case, social influences are not unanimous in painting girls as poorer at math or academics in general: Compare, for instance, Bart and Lisa Simpson. One recent study showed that, by four years of age, children tend to assume that boys are academically inferior to girls. So, there's no simple story about the nature of the social influences children are exposed to.
Further evidence against the Nurture-Only account comes from the intelligence researcher Jonathan Wai. Wai and colleagues looked at the ratio of males-to-females among the top .01% of seventh graders taking a test of mathematical reasoning between 1981 and 2010. In the early 1980s, the ratio was 13 boys to every girl. By the early 1990s, it had fallen to 4 boys to every girl, perhaps as girls gained greater access to a mathematical education. For the last two decades, however, the male:female ratio has remained stable, despite intensified efforts to eliminate it. (Meanwhile, girls have consistently outnumbered boys at the highest levels of verbal reasoning and writing ability.)
Similar difficulties beset Nurture-Only explanations of the spatial sex difference. As I mentioned above, men have better spatial abilities than women, on average, and women have better verbal abilities. In gay men and lesbians, however, these differences are reversed: Lesbians have spatial abilities comparable to those of straight men, and gay men have verbal abilities comparable to those of straight women (see here). Gay men and women are subjected to essentially the same socialization influences as their same-sex, straight peers. It seems unlikely, therefore, that the reversal of the usual pattern of aptitudes is due to socialization. More plausibly, other factors, such as prenatal hormonal exposure, are the primary cause.
Sex Differences in Variability
So, one possible contributor to the sex difference in STEM participation is average differences in relevant aptitudes. But even if there were no such differences, males might still outnumber women at the extreme right-hand tail of the distribution. Why?
There is good evidence now that, in many traits, we find greater variability among males than females. The distribution for men is flatter and stretches out further on both sides of the average. This variability reflects fundamental differences in the developmental program of the two sexes. These differences are found not only in humans but in a wide range of species. In barn swallows, for instance, there is more variance in tail length among males than females. Likewise, in red deer, there is more variance in body size among males than females. The same seems to be true in humans. In a wide range of traits, men exhibit greater variability than women. One example is height…
Is this pattern found for traits relevant to success in STEM? It seems to be. In a classic Science article, Novell and Hedges looked at six large, representative samples. They found that, in 35 out of 37 tests, the variability for men was greater than that for women. This included all the tests dealing with math and science. Another study, by Ian Deary and colleagues, showed that the same applies to general cognitive ability or IQ: There are more males at the top, but also more at the bottom—“more Nobels, more dumb-bells,” as Helena Cronin put it.
Could this pattern have a cultural explanation? Again, it seems unlikely. First, it’s not at all clear what kind of cultural influence or socialization practice could have the effect of making men consistently more variable across multiple, disparate traits. What could put more males at the top and more males at the bottom? And why does the greater variability of males emerge not only for psychological variables but for variables more impervious to social influence, such as height? Second, as noted, the greater variability of males than females is not unique to humans. It’s found in many species and for many traits. In other species, we don’t hesitate to attribute the pattern to biology. When we find the same pattern in humans, shouldn’t we attribute it to the same cause, rather than to an entirely unique cause that coincidentally replicates the pattern we see in other animals?
The finding that men are more variable than women in cognitive ability is controversial. Here are some questions to think about: (1) Is it sexist? Is it sexist even if it turns out to be true? Can facts be sexist? (2) If it’s sexist against women to say that there are more men than women at the highest levels of cognitive ability, is it sexist against men to say that there are also more men than women at the lowest levels? If not, why not? (3) Assume for a moment that it’s true. Should we suppress this knowledge? Could we suppress it even if we wanted to? (4) Might a better strategy be to try to educate people about avoiding exaggerating small average differences, not losing sight of the vast differences between individuals within each sex, and treating individuals as individuals, rather than as instantiations of the statistical properties of the groups to which they belong?
Sex Differences in Preferences and Priorities
So, average sex differences or sex differences in variability may be part of the explanation for the over-representation of males in STEM fields. Indeed, given how well established these differences are, it's hard to imagine that they're not part of the explanation. But even if there were no differences of either type, the over-representation of males in STEM would still not necessarily imply discrimination. It might instead be a result of average sex differences in preferences and priorities in life.
Several relevant differences have been documented. One concerns occupational preferences. A consistent finding, which has remained surprisingly stable across many generations, is that more women than men prefer to work with people than to work with things or abstract rules, whereas more men than women prefer to work with the things or rules—they prefer "people-free zones," as Camille Paglia put it. It is unlikely that this sex difference is purely a product of socialization. In addition to its long-term stability, research by Simon Baron-Cohen suggests that the first signs of the difference emerge within the first 24-hours of life: Newborn girls are more attentive to faces whereas newborn boys are more attentive to a mechanical stimulus. The preference for people vs. things may help to explain why more women than men do degrees in psychology or education, whereas more men than women do degrees in physics or engineering: It’s not discrimination; they’re just freely pursuing what interests them.
There's another reason that there may be more men than women at the very top of STEM fields or anywhere else. This relates to people’s priorities in life. Men are more likely than women to prioritize the pursuit of status above family, whereas women are more likely to give equal weighting to status and family. Uri Gneezy has a nice way of expressing this idea: There are more men at the top because women are more intelligent, and thus fewer women are willing to work 80 hours a week and sacrifice everything else. Are women wrong to hold such priorities? Should they prioritize money and status above everything else? If anything, one might want to argue the reverse: that many men might benefit from more balanced priorities!
Incidentally, we shouldn't forget that, although there are more men at the very top of the social ladder, there are also more men at the very bottom: more men in prison, more men on death row, more homeless men, more drug-addicted men, more men doing unpleasant jobs like garbage collection, more men killed on the job, etc. As Warren Farrell and Roy Baumeister have both pointed out, if we only look at the top layer of society, and assume that this reflects men’s position in general, we'll end up with a distorted picture.
Bias and Barriers
We’ve seen, then, three reasons that men could be overrepresented in STEM fields, even if there were no bias at all. This does not imply, of course, that there isn’t any bias or that bias plays no role in shaping the occupational landscape. On the contrary, some research suggests that it might. During the debate, Spelke discusses a study by Steinpreis and colleagues, in which university faculty evaluated hypothetical academic CVs. The researchers created two versions of the CV: an outstanding one and a middling one. The names on the CV were varied: For half the participants, the CV had a man’s name on it; for the other half, it had a woman’s. Thus, there were four experimental conditions: (1) male applicant, outstanding CV; (2) male applicant, middling CV; (3) female applicant, outstanding CV; and (4) female applicant, middling CV.
The results were unsettling. For the outstanding CV, it made no difference whether the hypothetical applicant was a man or a woman. However, for the middling CV, the “man” was perceived in a more favourable light. He was judged to have greater research productivity and more teaching experience, even though these sections of the CV were identical. When asked if they’d hire the hypothetical applicant, 70% of participants said yes to the man, whereas only 45% did to the woman. Note that it wasn’t just male professors who did this; the effect was just as strong for females. Other studies have reached similar conclusions (see here, here). Thus, it's reasonable to think that there's at least some bias in the academy. Whether this plays a role in real-life hiring decisions, when the decision-makers have more information about the candidates, isn’t so clear. But it certainly might, if only for middling applicants.
However, there are several things to remember. First, as I mentioned in an earlier post, bias and sexism presumably exist in all fields, but this hasn't stopped a flood of women going into prestigious non-STEM fields, such as law, medicine, and veterinary science. Why would it only throw women off the career path in some STEM-related areas? The bias-and-sexism hypothesis doesn't explain why women have gone into some fields in greater numbers than others; sex differences in aptitudes, variability, and preferences do.
Second, although several studies suggest that there may be bias in the sciences, not all the evidence points in that direction. The studies I mentioned above are one-off studies with relatively small samples. A much more systematic survey of the literature conducted by Stephen Ceci and Wendy Williams reached a very different conclusion. The researchers were interested in whether women face systematic bias in terms of getting academic positions, grants, and publications. Conventional wisdom says they do; however, since at least the late 1970s, the research provides scant evidence of this. The occasional study does seem to show sex bias. However, the effects in such studies are small, and the bias is just as likely to favour women as men. Ceci and Williams concluded that “the ongoing focus on sex discrimination in reviewing, interviewing and hiring represents costly, misplaced effort. Society is engaged in the present in solving problems of the past.”
Other studies have reached similar conclusions. This includes a non-partisan report by the US National Academy of Sciences. Christina Hoff Sommers highlights a typical finding from the report: In 2004-2005, only 20% of applications for faculty positions in mathematics were from women; however, 28% of the candidates interviewed were women, and 32% of those offered jobs were women. This is the opposite of what we’d expect if anti-female bias were pervasive in STEM departments.
At the very least, people in the bias-and-barriers camp should concede that the evidence is inconclusive. But this in itself suggests that any bias must be relatively weak—after all, if there were strong and pervasive bias, we would presumably have unambiguous evidence of it by now.
The Argument from Past Mistakes
This seems to me a reasonable conclusion. However, there’s one last issue to consider. As Spelke points out, there's always a temptation to conclude that the way things are today reflects the natural order of things. In the nineteenth century, she notes, most mathematicians were European. Few were Asian. At that time, it might have been tempting to conclude that European genes confer greater mathematical ability than Asian genes—something that no one would conclude today. Similarly, during the Scientific Revolution, it would have been tempting to conclude that Catholics are more innately disposed to scientific accomplishment than people of Jewish ancestry—again, something that no one would conclude today. We know that historical contingencies have determined the accomplishment of groups in the past in ways that do not reflect their innate aptitudes. It is reasonable to think that similar processes are at work today with respect to women in STEM. Indeed, it may be unreasonable to think otherwise. After all, no one denies that women faced discrimination in this arena until very recently. Is it realistic to think that this has evaporated entirely in a single generation?
This argument certainly deserves to be taken seriously. It is not, however, immune to criticism. First, the main takeaway of the argument is that people have been wrong in the past, and thus that we might be wrong today—but of course this applies to both sides of the debate. Furthermore, as Pinker points out, we could use Spelke’s examples to argue for the exact opposite conclusion regarding women in science. If arbitrary group differences are self-perpetuating, as many psychologists seem to assume, we would not have expected the European-Asian and Catholic-Jewish differences to disappear as they did. We would expect them to become more and more entrenched. Instead what happened is that, as soon as the barriers were removed, Asians and Jews started confounding the stereotypes and bucking the trends.
In the last century or so, we’ve removed a lot of the societal barriers holding women back. Despite many people’s low expectations for them—and despite an absence of role models and gender equity workshops— women have gone into every occupational field and thrived. But they haven’t gone into every field in equal numbers. The fact that relatively few women have gone into STEM fields, despite the removal of the societal barriers, provides at least tentative evidence that, unlike the European-Asian and Catholic-Jewish differences, the sex differences in STEM reflect deep-seated differences between the sexes, rather than just current social arrangements.
None of this is meant to suggest that there is no longer any bias against women or that we shouldn't do anything about it. It is only to suggest that, once the bias has been eliminated, we can't necessarily expect a 50:50 ratio in every area of life. If men and women were identical in their aptitudes and aspirations, we might expect gender parity in STEM fields. But we've known for a long time that they're not. Given this knowledge, we need to ask why a 50:50 ratio should be our goal. Most men don’t want a career in science and nor do most women. Why is it important that the small fraction of women who do want to do science should be the same size as the small fraction of men who want to do science? And why does it apply only to STEM fields? Why aren't we equally concerned that more women than men want to go into psychology, for instance? More to the point, why should we be concerned about the gender disparity in either case?
Certainly, if women were being unfairly deprived of the opportunity to go into certain fields, that would be something to remedy. But if it turns out that the disparities are due mainly to men and women's uncoerced choices, we need to ask whether balancing the sex ratio in STEM fields is a worthwhile goal. More than that, given that it might involve costly interventions aimed at curing a non-existent problem, we need to ask whether it is ethical.
Anyway, here's the Pinker-Spelke debate in full; hope you enjoy it:
My Previous Blog Posts on the STEM Debate:
Sex Differences: Proof of Sexism or a Sign of Social Health? "Why should we trample people’s career preferences in order to enact our own preference for a 50:50 sex ratio in science or any other area? Why do activists think that their preference for a 50:50 sex ratio should take precedence over women and men's preferences regarding their own lives and careers?" Click here.
Where Are All the Women? "No one is worried that there are fewer men than women going into psychology and other social sciences, or that there are fewer men going into the humanities. No one is worried that society gives boys the message that psychology is a girl thing or that male pharmacologists or veterinarians face a hostile environment. Why the myopic focus on the STEM fields?" Click here.
Affirmative Action for Women in Science? "Rather than trying to socially engineer men and women's preferences so as to create a 50:50 sex ratio in every occupation, we should respect people's right to make their own decisions about what they do with their lives." Click here.
Follow Steve on Twitter