When I look around the offices of the Duke Biology building where I work, I see graduate students working fervently on data analysis, finishing papers, or writing grant proposals—and the majority of these students are female. Throughout my academic training, never have I felt out of place because of my gender, even though the majority of my courses were in math and statistics—classes, as new research confirms, that are dominated by men.

And yet, I am often insecure about my abilities as a female scientist.

I’m not alone. Female scientists across the country are leaving prestigious paths. At each stage of the scientific ladder—undergraduate to graduate to postdoc—more women than men leave the academic sciences, a phenomenon that has been termed “the leaky pipeline.” In 2010, The National Science Foundation reported, women earned 49 percent of doctoral degrees in all science, engineering, and health fields. But they held just 39 percent of postdoctoral positions and 32 percent of full-time faculty positions. This national trend is echoed in the Duke biology department. From 2003–2013, 54 percent of Ph.D. graduates were female, yet 28 percent of these women left academia, compared to only 20 percent of men.

The reason for the “leaky pipeline” is a combination of social, cultural, and psychological factors—but they all contribute to a confidence gap that plagues female scientists, just as acutely as it plagues other professions. A recent survey of 193 graduate students in STEM fields at Duke showed that women consistently underestimate their abilities compared to men. In this study, Psychologist Lindsey Copeland found that 41 percent of men indicated that it is “definitely” true they have good technical skills (defined as the knowledge and abilities needed to accomplish mathematical, engineering, scientific, or computer-related duties), compared to only 11.5 percent of women.

I’ve noticed a similar distinction in the way men and women talk about their work. Men talk about how exciting their projects are; if they’re stuck on a project, they blame it on resources or lack of support from their advisor. Women talk about how they don’t know how to move on, the grant they didn’t get, how their results aren’t interesting. For men, the next step after graduating is obviously a postdoc. Women are already considering alternatives in case academia does not work out.

Take Mariana Gomez-Schiavon, a Duke Ph.D. student I met my first week of graduate school. Gomez-Schiavon says she comes off confident because she constantly applies the motto, “fake it until you make it.” She has taught herself computer programming and advanced mathematics, and convinced a reluctant professor to fund her and a group of like-minded students to compete in a prominent research competition when she was an undergrad. Yet Gomez-Schiavon downplays her accomplishments. Unless explicitly asked, she rarely mentions them. And though she knows she wants a career as a researcher, she says she is much less ambitious than her boyfriend, also a science Ph.D. student. She is quick to point out her weaknesses—that she is not results-oriented, that her research area is too theoretical—and contrasts them with her boyfriend’s strengths: He is better at discussing his work with others, he cares more about publishing papers. “It’s easier for me to compromise [my career] than it is for him,” Gomez-Schiavon says.

Or take Kate Hertweck, a postdoc at the National Evolutionary Synthesis Center. Hertweck says that she would quickly sacrifice her career for her partner. She recently accepted a position in the biology department at the University of Texas at Tyler, where she is now the first female tenure-track faculty in the department. Yet instead of feeling elation or relief, Hertweck was hesitant to accept, fearing that her boyfriend, who works from home, would not want to move. “I would hate to sacrifice his personal happiness in favor of a career that is stressful,” she says. “There are many other jobs that I think I would be happy doing.”

The persistence of this type of attitude largely derives from the persistence of gender-based norms relating to career and family. Female scientists, says psychologist Alysson Light, may be more inclined to sacrifice their careers for the sake of their partners’ because women tend to define themselves in terms of their relationships more than men do. And research by psychologist Bernadette Park has shown that women perceive conflict between their work roles and parenting roles, while men, comparatively, do not.

In many of the interviews I conducted for this piece, I saw evidence of these gender-based norms among my colleagues. When Duke biology professor Kathleen Pryer was a graduate student at Duke in the early ’90s, she was told by a male professor that women should be at home baking cookies, not messing around in the lab. While much improved, Pryer says the academic environment for women in the sciences is still more difficult than for men. Hertweck says that colleagues complain that she is too harsh when she gives strong feedback. Yet, if Hertweck is emotional, she says, people don’t know how to react. Beth Sullivan, a professor of Molecular Genetics and Microbiology at Duke, told me she was close to leaving academic science at a moment of personal distress; it was only because of a fierce and supportive mentor that she remained. Yet Sullivan points out that this is rare. “Women are not mentored as well as men,” Sullivan says. “Men get supported very differently. Women have to ask, but men typically get mentored without asking.”

These norms don’t just play out on an anecdotal level. A 2012 study by Corinne Moss-Racusin and co-authors backs up Sullivan’s assertion that male scientists are mentored more readily than female ones. Moss-Racusin gave 127 science faculty members an application for a laboratory manager position and randomly changed the name of the applicant to be either unambiguously male or female. Both male and female faculty were significantly less likely to mentor—to encourage the applicant to stay in science, for example—the female applicant compared to the male applicant. Moss-Racusin argues that because of implicit gender bias, faculty invest less in female scientists, which results in women more readily leaving science.

The lack of self-confidence among female scientists ultimately stems from a conflict between the stereotypes associated with a woman’s role in society and a woman’s perception of herself as a scientist. Acknowledging this conflict is crucial; only once we take note of all the consequences of this conflict will we be able to repair the leaky pipeline. On the wall of the office of Debbie Silver, a professor of Molecular Genetics and Microbiology at Duke, hangs her daughter’s drawing of a scientist—a woman with long, wavy hair, smiling. Hopefully, the perception of scientists will soon shift from that of the iconic image of wild-haired Einstein to that of Silver’s daughter’s—crazy hair optional.

Rotem Ben-Shachar is a fourth-year student in the Computational Biology and Bioinformatics graduate program at Duke University.