The question has been debated for decades: Are women less capable than men in science or mathematics?
It is tempting to answer with a simple yes or no, but the evidence is considerably more complicated. There are measurable average differences between males and females in some cognitive abilities and mathematical tasks. Recent research does find a small overall male advantage in mathematics, although the size of the difference varies substantially by mathematical domain, age and population.
Acknowledging this evidence, however, does not establish the much stronger claim that women are inherently less capable of mathematics or science. To reach that conclusion, we would have to move from population statistics to individual ability, and from observed differences to their causes. That is where the issue becomes much more interesting.
1. Average Differences Are Not Individual Ability
Whenever two large populations are compared, their averages may differ. But people within each population are spread across a wide range of abilities, so a difference between two averages does not divide humanity neatly into two groups of higher and lower ability.
Imagine two overlapping distributions of mathematical performance. Even if one group’s average is somewhat higher, there may still be considerable overlap between the two populations. A woman can therefore be substantially better at mathematics than the vast majority of men while belonging to a population whose average score is slightly lower.
An earlier meta-analysis covering 242 studies and more than 1.28 million participants found essentially no overall gender difference in mathematics achievement, with an effect size of only d = 0.05. More recent research points toward a small average difference, but it remains small relative to the enormous variation among individuals.
This distinction is often lost in public discussion. “Men perform somewhat better on average” gradually becomes “men are better at mathematics,” and eventually becomes “women are less capable of mathematics.” Each step goes further than the original statistic justifies.
2. Mathematics Is Not One Ability
There is also no single mental faculty called mathematical intelligence. Mathematics involves numerical reasoning, algebra, geometry, spatial reasoning, abstraction, pattern recognition, logical reasoning, working memory and problem-solving, among other skills.
Recent research illustrates this clearly. A 2026 meta-analysis found that gender differences vary considerably according to the mathematical domain, with some of the larger differences appearing in geometry while other areas show much smaller differences.
Science is similarly diverse. Physics, engineering, biology, medicine, computer science and mathematics require overlapping but different combinations of abilities, knowledge and interests. A difference observed in one particular cognitive task therefore cannot automatically be turned into a general statement about scientific capability.
3. The Bell Curve and the Problem of Extremes
A simple bell curve helps explain why population differences should not automatically be interpreted as individual differences.
In a normal distribution, approximately 68%, or roughly 70%, of people fall within one standard deviation of the mean. About 95% fall within two standard deviations and approximately 99.7% within three. For a general reader, the important point is simply that most people occupy the broad middle of the distribution.
Standard deviation provides the technical way of describing this variation. It tells us how widely individuals are distributed around the average, allowing us to judge whether a difference between two means is large or small relative to the natural variation within the populations.
This becomes particularly important at the extremes. A difference that looks modest around the middle can produce a considerably larger difference in the number of people found several standard deviations above the mean. The most exceptional performers are, by definition, statistical outliers, so the characteristics of this tiny group cannot automatically be applied to the much larger population.
This is one reason the composition of groups such as Nobel laureates, mathematical prodigies, elite scientists or world-class chess players needs to be interpreted carefully. They represent the extreme tail of a distribution rather than the ordinary population.
4. Chess Shows Why Elite Representation Is Complicated
Chess provides an unusually useful example because performance can be measured through competitive ratings rather than educational credentials.
Men substantially outnumber women in competitive chess, particularly at the highest levels. But researchers disagree about how much of this difference can be explained by participation and how much remains after other factors are considered.
One major study of more than 250,000 chess players found that boys entered competitive chess in much greater numbers and at higher initial performance levels. The authors argued that the enormous difference in participation could account for much of the difference observed at the top.
Another analysis of German chess data produced an even more striking result. There were approximately 16 male players for every female player, and the authors estimated that the difference in population size could explain about 96% of the difference between the top 100 men and women.
That interpretation was subsequently challenged. Another analysis estimated that participation accounted for only about two-thirds of the difference, leaving a substantial residual gap. More recent U.S. Chess Federation data have also found male-female rating differences across a broad range of ratings, while finding higher attrition among female players.
The important lesson is therefore not that chess has conclusively proved one side of the gender debate. Rather, it demonstrates how population size, participation, ability distributions, experience, retention and selection can interact to produce large differences at the elite level. For a related discussion of human performance against machine analysis, see The Stockfish Paradox.
A large imbalance among elite players is consequently not, by itself, a direct measurement of innate intellectual superiority.
5. My Experience Teaching Mathematics
My own experience also makes the simplistic argument difficult to accept.
I have taught mathematics to many students, including girls who were exceptionally good at it. Some were not merely competent; they were genuinely brilliant at mathematics. I have also seen female students achieve extremely high marks in competitive examinations and board examinations.
Of course, classroom experience cannot settle a population-level scientific question. It is not a controlled experiment and cannot establish the relative distribution of mathematical ability between men and women. But it does illustrate why population averages should not be turned into judgments about individuals.
If a particular girl is exceptionally good at mathematics, knowing that the average score of her gender is slightly higher or lower tells us very little about her actual ability. Her demonstrated performance is much more informative.
6. Interest Is Different From Ability
The discussion becomes more complicated when we move from ability to occupational choice.
Women are underrepresented in some STEM fields, particularly engineering, computing and certain physical sciences. But professional representation is the result of a long sequence of decisions and selection processes. Interest influences educational choices; those choices influence participation; participation affects who remains in a field; and training, family circumstances, persistence and selection further shape the eventual professional population.
There is strong evidence that men and women also differ, on average, in occupational interests. A major meta-analysis covering more than 503,000 respondents across 47 interest inventories found substantial differences, particularly between things-oriented and people-oriented interests.
Research examining individual STEM fields found the largest male-favouring differences in areas such as mechanics and engineering, while differences were considerably smaller in biological and medical sciences. Social sciences and medical services showed female-favouring patterns.
This is an important distinction because “STEM” is not one homogeneous category. A woman who has little interest in mechanical engineering may have a very strong interest in medicine or biology, both of which can require extensive scientific knowledge and high intellectual ability.
Preference therefore cannot be treated as a measurement of competence. A person may be highly capable of doing something without particularly wanting to do it, while someone else may be strongly interested in a field without being exceptionally talented at it.
7. Culture and Traditional Roles Can Shape the Outcome
Career decisions also take place within a cultural environment.
In many societies, girls and boys continue to encounter different expectations about adult life. Women may face stronger expectations concerning marriage, children, household responsibilities and caring for family members, while men may receive greater encouragement to prioritise income and career advancement.
These pressures do not have to take the form of an explicit prohibition. A girl may never be told that she cannot become an engineer or physicist, yet she may gradually receive the message that certain professions are more suitable for women, that family should take priority, or that an extremely demanding career is incompatible with the life expected of her.
Consider a profession requiring ten or fifteen years of training, international mobility, long working hours and intense competition before reaching senior positions. If women anticipate greater family responsibilities, the costs and attractiveness of that career may not be identical for men and women. Differences in participation can therefore emerge even when substantial mathematical ability exists in both populations.
This does not mean that every woman who chooses a family-oriented career has been forced into it, nor does it mean that every difference in occupational preference is caused by discrimination. People have genuine preferences, and those preferences may themselves be influenced by personality, biology, culture, experience and incentives.
The more reasonable conclusion is that culture can influence how individual abilities and interests are converted into career choices.
International education data support this broader interpretation. OECD analyses have found substantial gender differences in students’ career expectations even among students with comparable science performance. Girls are more likely to expect careers in health and medicine, while boys are more likely to expect careers in engineering, computing and other technical areas.
This is not simply a story of girls abandoning intellectually demanding careers. Often, boys and girls are choosing different intellectually demanding careers.
8. Biology Cannot Simply Be Dismissed
None of this requires us to pretend that men and women are biologically identical. They are not. There are biological differences between the sexes, including differences in hormones, physical development and brain characteristics, and it is reasonable to investigate whether some of these differences contribute to cognitive or behavioural differences.
The mistake is to move directly from “biological differences exist” to “therefore women are generally less capable of mathematics and science.” That conclusion requires considerably more evidence.
The existence of an average biological difference does not tell us how large its effect is on a particular cognitive ability, how much overlap exists between individuals, how the difference interacts with education and training, or whether it is sufficient to explain occupational outcomes.
Biology may be part of the explanation without being the whole explanation.
9. What the Evidence Actually Allows Us to Say
The evidence gives us a more complicated picture than either side of the traditional argument would prefer.
There are small average sex differences in mathematical performance, with larger differences in some domains than others. Men and women also show substantial average differences in occupational interests, particularly between things-oriented and people-oriented activities. Cultural expectations, traditional roles, education, family responsibilities and economic incentives can influence how those preferences become career choices.
At the same time, participation and selection can amplify differences when we look at elite performers, as the chess literature demonstrates. Biological differences may also contribute to some observed differences, and it would be just as unscientific to dismiss that possibility as it would be to assume that biology explains everything.
The crucial point is that none of these observations establishes that women are generally less capable of mathematics or science.
A population average is not an individual’s ability. An occupational preference is not a measure of competence. An imbalance among elite performers is not automatically proof of innate superiority.
10. Conclusion: Capability Is Individual
The original question—Are women really less capable than men in science or mathematics?—is therefore badly framed.
There may be genuine differences between the average distributions of men and women in some cognitive abilities. There are also differences in occupational interests, and biology may contribute to some of them. At the same time, culture, traditional expectations, education, family responsibilities, participation and selection can influence what eventually appears in universities and professions.
The broad middle of the distributions can overlap enormously even when their averages differ. The extreme tail is a different statistical environment, where relatively modest differences in distributions, population size and selection can become much more visible.
That is why the composition of the world’s best chess players, mathematicians or scientists cannot simply be used as a measurement of the mathematical or scientific capability of an entire sex.
My own experience teaching mathematics reinforces the simplest part of the argument. I have encountered female students who were exceptionally talented at mathematics. Their ability was not a theoretical possibility or a population statistic; it was visible in what they could actually do.
Statistics are useful for understanding populations. They can reveal genuine differences that deserve investigation, and they can also show us how misleading our intuitions about averages and extremes can be. But they do not assign an intellectual ceiling to an individual. For a broader discussion of how learning and capability develop, see The Apprentice Mind.
The distribution describes the population; the individual still has to be judged by what she—or he—can actually do.