“What will our soldiers think when they return to the university and find that they are expected to learn at the feet of a woman?” A University of Göttingen faculty member, describing the German-born mathematician Emmy Noether, circa 1915

“Are photons gendered?” is the title of a chapter written by Yale University astrophysicist Meg Urry in Gendered Innovations in Science and Engineering. Physics is phenomenally successful at taking data on sexless, raceless objects such as photons and transforming that information into mathematical laws with highly accurate predictive power. To be a physicist is to love working in a field where this is possible. So why should there be a paucity of women in physics, at all professional levels and in virtually all industrialized nations?

Some readers may deny that low participation by women in physics is a problem requiring intervention. Some might think it a free choice by women, or else a legacy of past discrimination that will disappear over time. But how can one view data that reveal less job satisfaction, lower pay, slower progress through the ranks and fewer professional resources and opportunities as indicating anything but a problem that begs a solution?

There are many theories to explain this phenomenon, and it is a tenet of gender studies that one should avoid attributing complicated effects to a single cause. For example, biological determinism – that one sex or race is, from the moment of birth, better suited for a tenured professorship at Harvard – was one of several theories once favoured by Larry Summers, director of the US’s National Economic Council and a former president of Harvard University. Gender studies is a field guaranteed to annoy the physicist – where data not only stubbornly confound our ability to write laws but also where unifying principles, when they exist at all, offer much less in the way of accurate predictive power.

Here, I will explore only one cause that could explain the paucity of women in physics: a double standard of professional evaluation for men and women in the sciences. It is well documented in the social-science literature that gender can sometimes enter into professional evaluation. When it does, men fare better than women in high-status fields, like physics, that prize intellect, the ability to analyse and leadership. According to such studies, teaching evaluations and letters of recommendation are quantifiably different for men and women in the sciences. A name change, for example from John to Jane, on identical CVs or preprints can result in different ratings. Particularly interesting are studies finding that women do less well than men, even if they have equivalent experience and qualifications.

Yet while some studies on scientific grants and tenure rates find gender bias, it is certainly not the case that all such studies do so. Moreover, some studies find qualitative and quantitative differences in bias in different branches of science. Esther Duflo, an economist at the Massachusetts Institute of Technology, had applied randomized evaluation to political and economic issues, for example the willingness of voters to elect a woman, in the developing world. According to the New Yorker, one colleague speaking at a session of the World Bank in New Delhi said that Duflo’s methods could be taught and that they “are not nuclear physics”, in response to which a UNICEF official said “Studying human beings is much more complicated than nuclear physics.” Amen!

Caught on camera

With a grant from the Mellon Foundation, I, together with my colleagues Kris Lui and Etsuko Hoshino-Browne, conducted a study to determine how male and female physicists are evaluated in the classroom. We used videotaped lectures in which professional actors – two male and two female – played the role of a physics professor. They each gave a 10-minute physics lecture to a class of students that included blackboard work, a demonstration and a question-and-answer session. None of the actors were trained in physics but all received the same preparation and memorized the same script.

We then showed 126 physics students the lecture by one of the four “professors” and got them to fill out a survey in which they rated various aspects of the lecture using a five-point scale. The students supplied some personal information but not their own gender, which was noted covertly by the person collecting the surveys.

Our study resembles those in which subjects are asked to assign certain qualities, such as leadership, by looking at a photograph of a person. However, one would hope that viewing a lecture would produce a more meaningful evaluation – at least that is the hope of those of us who use student evaluations to decide on hiring, promotion and salaries.

Responses to 15 survey questions were combined to create a single score. We found that, on average, the male professors received higher scores than the female professors. But while female students gave slightly higher marks to the female professors than they did to the men, male students rated male professors vastly better – a result that is reminiscent of studies of science-teacher evaluations from colleges and secondary schools.

We also considered the students’ responses to questions that looked at gender-stereotypical attributes, such as if the lecturer had a “solid grasp of the material”, if they were knowledgeable, or good with equipment. These questions yielded a distinct gender bias, in that both male and female students rated male professors as better. (Female students were, however, more equivocal – the gender gap in their ratings was not significant, as it was for male students.) On the other hand, a group of questions asking whether the professor “teaches in a way that really helps students learn”, is well organized and interacts well with students produced quite a different effect: here, there was a clear own-gender bias, with female students rating the female professors as superior and the male students giving higher marks to the male professors.

We concluded that not only can the gender of the physics professor make a difference to how a lecture is received by students, and in what sorts of strengths and weaknesses students perceive, but also that student gender can play a role. So the male soldiers of Göttingen would, on average, prefer to learn physics at the feet of a man – not at those of a woman speaking the same words, writing the same equations on the board and giving the same answers to questions that the soldiers pose. Were there any female soldiers, their preference would be less clear cut.

A little bias goes a long way

The results we found are consistent with the theory that people have implicit beliefs that associate different genders with different aptitudes and predilections. Even people who consciously believe in gender equality cannot suppress these automatic associations. Indeed, an online test (available here) reveals that I too have a moderate, innate tendency to associate men with science. A set of associations forms a “schema”, which is understandably useful to our survival as a species. However, schemas of gender, race and disability also seep into our professional interactions and judgments. Psychologists like Virginia Valian have argued that an identity that violates such schemas, for example a female physicist or a male nurse, has a negative consequence in terms of the individual’s evaluation and perception.

Further, small disadvantages such as an inferior teaching evaluation or a smaller start-up grant can accumulate over time and have dramatic consequences on a career. To a physicist, this suggests an analogy with a Monte Carlo simulation. A random walker has many important, gender-independent terms in its Hamiltonian. There is also a tiny, implicit bias term that couples to its gender. In each interaction (with a teacher, thesis advisor or tenure committee, for example) a Boltzmann factor determines the likelihood of the walker proceeding in a given direction. The ensemble-averaged flux of male walkers will be greater, when projected onto directions indicating professional success and job satisfaction, than will the flux of female walkers, due to the existence of the gender-biased term. Indeed, no matter how tiny the gender-biased term, a biased random walk will drift according to the direction of the bias.

Recently we have seen huge strides in equity issues in science between men and women. For example, medical researchers Christine Wenneras and Agnes Wold from Göteberg University in Sweden analysed awards handed out in 1994 by the Swedish Medical Research Council and found sexism and nepotism (Nature 387 341). A follow-up study in 2004, however, found no evidence of sexism (though nepotism found in the original study was still present). In the US, we have gone from a “leaky pipeline” at all career stages, overt discrimination and tolerance of harassment to double-digit percentages of women on faculties, hiring and promotion at equitable rates and gender bias that is usually quite subtle.

Today, the big issues are acknowledging and correcting for implicit bias, reforming workplace-policy, bringing in students from under-represented minorities, retaining girls between school and college, and seeking equity in the developing world. I believe that progress will continue, as long as good people are willing to act, rather than saying “it is not my problem to solve”. It will also depend on continued effort by educational and funding institutions and professional organizations, so many of which have engaged in difficult introspection and institutional change. Hopefully, leaders in academia, government and industry will continue to fund research and disseminate results and recommendations – so that eventually women will not need to “swim against the tide” in the physical sciences.