Race and gender differences in how sense of belonging influences decisions to major in STEM

A first-year student walks into a large introductory physics lecture hall, scans the room, and in just a few seconds of pure pattern recognition, she notices that almost no one there looks like her. That moment — quick, wordless, and automatic — turns out to matter enormously. Rainey and colleagues interviewed two hundred one students who lived exactly that experience, and what they found is that those few seconds of scanning do lasting, measurable damage to whether someone stays in science at all. The concept at work here is sense of belonging. Researchers define it, following Goodenow, as students' sense of being accepted, valued, included, and encouraged by others in an academic setting — feeling like an important part of the life and activity of the community. It sounds soft, but the evidence is solid. Prior work consistently links belonging to academic achievement, retention, and persistence. What Rainey and colleagues wanted to know was whether it works the same way for everyone — or whether race and gender, and crucially their intersection, change the picture entirely. The study design was deliberate about who got included. Rainey, Dancy, Mickelson, Stearns, and Möller drew from a larger project called the Roots of STEM Success, selecting two hundred one in-depth interviews with college seniors from sixteen University of North Carolina campuses. Half were current STEM majors; the other half had started in STEM and switched out — the authors call them "leavers." The sample was purposely constructed to oversample underrepresented groups: sixty-six percent of interviewees identified as women, and fifty-two percent identified as students of color. Interviews ran between thirty minutes and an hour, were recorded and transcribed, and participants were paid twenty-five dollars. The analysis used what the authors call partial grounded theory. Two researchers independently coded responses under broad categories, some specified in advance and others allowed to emerge from the data. They then compared and refined those codes iteratively until a final scheme was stable. Every interview was double-coded, and discrepancies were resolved through discussion. Crucially, the team used an intersectional framework, analyzing race and gender together rather than separately. This matters because treating these as isolated axes, Rainey and colleagues argue, can erase the experiences of women of color, whose situation ends up invisible in analyses that collapse across one category or the other. The results are stark. White men were the most likely group to report a sense of belonging in their STEM major. Women of color were the least likely. These two poles frame everything else the study finds. Among students who had stayed in STEM, female majors were significantly less likely than male majors to report belonging — with a z-score of 2.41. Students of color who majored in STEM were significantly less likely than white majors to feel they belonged — with a z-score of 2.23. And at the intersection, women of color showed the largest gap of all, with a z-score of 3.62. What makes that last number particularly striking is that even women of color who stayed in STEM — who by definition had persisted — just over half reported consistent feelings of belonging. They were making it through a major while feeling, much of the time, like they didn't belong in it. Across all demographic groups, the study identified four recurring factors that contributed to sense of belonging: interpersonal relationships, perceived competence, personal interest, and science identity. Think of these as the four levers. Interpersonal relationships were the most commonly cited — the peer conversations outside class, the connections with instructors, and the small moments of being included in a group. Perceived competence is the confidence that a student can actually do the work. Personal interest is a sustained intrinsic draw to the subject. And science identity is perhaps the most loaded of the four: whether a student sees themselves as a person who does science, as someone to whom that label naturally applies. These four factors shaped belonging for everyone. But access to them was not equal. White men benefited from being surrounded by people who looked like them, from fitting the stereotype of who does STEM, and from the frictionless interpersonal connections that tend to follow from that fit. Students from underrepresented groups often described real friction — difficulties forming working groups when those groups were composed entirely of white men, fewer signals that they were seen as competent, and barriers to claiming a science identity at all. No white male in the study — not a single one, major or leaver — reported lacking a science identity. Every report of missing that identity came from students in marginalized groups. Among students who left STEM, roughly one quarter cited a lack of science identity as a reason. Among leavers, women were more likely than men, and students of color more likely than white students, to report that gap. Science identity, in other words, isn't just a nice psychological bonus. It appears to be a separator between students who stay and students who go. There’s a layer of specificity in the findings worth noting. The study shows that STEM is not a monolith — which sub-discipline you’re in changes the belonging calculus, particularly for women. Women in biology reported different belonging experiences than women in the physical sciences. The paper links this directly to visible representation: external cues, like looking around a room and seeing few people who share your gender or background, influence belonging especially in male-dominated fields. Biology, at the undergraduate level, has moved closer to demographic parity. The physical sciences have not. And students can read that difference. They read it in the first few minutes of class. This connects back to that opening scene in the lecture hall. The student scanning the room isn’t overreacting or misreading herself. She is accurately reading her environment. And her environment is providing her with accurate information. Rainey and colleagues are careful about what their data can and can't support. The sample is self-selected, limited to two axes of identity — race and gender — and drawn entirely from one state, which limits how far the findings can generalize. The analytic team was primarily composed of white women with social and physical science backgrounds, a limitation the authors name explicitly. These are real constraints. But the core findings are consistent enough to warrant the conclusion the team reaches: this is not a story about individual students who lacked the resilience to persist. It is a story about institutional cultures and organizational structures that systematically privilege white males and that shape the environments in which belonging either develops or doesn't. The four factors — relationships, competence, interest, and identity — are real, and they matter for everyone. The problem is that the environments STEM programs create make those factors easier to access for some students and harder for others in ways that track directly onto race and gender. What would actually change this? The study points toward systemic interventions rather than individual supports. Increasing visible representation of women and people of color in departments and curricula changes what students see when they scan the room. Reforming pedagogy to create structured opportunities for peer relationships addresses the factor students most commonly cited as central to belonging. Attending explicitly to how classroom norms and curricula build or erode science identity — and tracking who gets to claim that identity and who doesn't — would address the most consequential gap the study found. Departments need to audit their norms and build conditions where meaningful interpersonal connections can form, because the presence or absence of those connections was the single most common reason students gave for feeling they did or did not belong. Return to that student in the lecture hall. According to this research, what she experiences in that moment of scanning is not a private psychological event. It is a response to structural signals that her environment is broadcasting. The composition of the room, the norms of the discipline, who gets called on, and who forms study groups with whom — these are features of an institution, not features of a person. Changing who stays in STEM means changing those features. That is the finding. It is not complicated, and it is not small. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.