On the first day of a large introductory chemistry or calculus course at many American universities, some version of the same speech is given. Look to your left. Look to your right. By the end of the semester, one of you will not be here.
The line is delivered as a warning, sometimes as a challenge, occasionally as a point of pride. It reflects an idea that runs deep in science and engineering education: that the early courses are supposed to be hard, that their difficulty serves a purpose, and that the students who leave were never going to make it anyway. Students call these courses “weed-out classes.” Faculty rarely use the term in public, but many recognize it.
The trouble is that the idea rests on an assumption that, when researchers have looked closely, has not held up well.
How Many Leave
The best-known national figure comes from the U.S. National Center for Education Statistics, which followed students who began college in 2003–04 for six years. About 28 percent of students pursuing a bachelor’s degree entered a STEM field at some point. By 2009, 48 percent of those STEM entrants had left. Roughly half of the leavers switched to a non-STEM major. The rest left college without any degree at all.
For students in associate’s degree programs, the loss was larger still: 69 percent of STEM entrants had left the field.
Those numbers are now old, and the landscape has shifted since. But the pattern they describe has been remarkably stable over decades of study. An earlier analysis found that 44 percent of freshmen who entered college in 1987 intending to major in a STEM field had switched out by 1991. The share of freshmen who say they want to study STEM has long sat well above the share of degrees that actually go to STEM graduates. Something in between is draining the pipeline.
The students who stayed and the students who left often described the same problems. What differed was not their ability.
The Twist in the Data
Here is the part of the NCES report that gets quoted less often. STEM was not unusual. Attrition in humanities, education and health sciences was actually higher, between 56 and 62 percent. Business and the social sciences were roughly comparable to STEM. Students leave majors all the time, in every field.
That finding cuts two ways. On one hand, it suggests that the STEM exodus is not some unique catastrophe. On the other, it undermines the usual justification for weed-out culture. If students leave every field at similar rates, then the extra difficulty of introductory science and math courses is not obviously doing the special sorting work it is claimed to do. And STEM loses students in a particular way: many of those who leave are capable, high-achieving students who arrived with strong preparation and genuine interest.
What the Leavers Said
The most influential study of why students leave science is a book called Talking About Leaving, published in 1997 by the sociologists Elaine Seymour and Nancy Hewitt. They interviewed hundreds of students at several universities, including both those who switched out of science, math and engineering and those who stayed.
Their most cited finding is that about 90 percent of the students who switched out described poor teaching as a concern. But the more revealing finding was quieter. Many of the students who stayed described the same problems: overloaded curricula, impersonal lectures, a competitive grading culture, faculty who seemed uninterested in undergraduates. The persisters and the leavers were not cleanly separated by ability. Many of those who left were as capable as those who stayed. What often differed was how they responded to an environment both groups found discouraging.
If that is true, then the weed-out course is not functioning as a fair test of who can do science. It is functioning, at least in part, as a test of who will tolerate a particular style of teaching.
The Evidence on Teaching
There is now strong evidence that the style of teaching matters a great deal. In 2014, a team led by the biologist Scott Freeman published a meta-analysis in the Proceedings of the National Academy of Sciences covering 225 studies of undergraduate science, engineering and math courses. They compared traditional lecturing with “active learning”: classes where students work problems, discuss in small groups, answer questions in real time, and get frequent feedback.
The results were striking. Students in traditional lecture courses were about 1.5 times more likely to fail than students in active-learning courses. Average failure rates were roughly 34 percent under lecturing and 22 percent with active learning. Exam performance improved as well. The authors noted that if the experiments had been medical trials, the control condition might have been stopped early for the benefit of the patients.
That finding does not mean every lecture is bad or that every active-learning class is good. Implementation matters, and doing it well takes faculty time and institutional support. But it does mean that a large share of failure in introductory courses is not simply a property of the students. It is partly a property of the course.
If a medical treatment cut failure rates by a third, we would not call the old treatment rigorous. We would call it outdated.
Who Pays the Price
The costs of weed-out culture are not distributed evenly. Students who arrive without strong high school preparation, without family members who have navigated college, or without the financial cushion to retake a course are more likely to be pushed out by a single bad semester. First-generation students, students from under-resourced schools, and students who are the only person like them in a room of three hundred all face extra friction. When the course is designed to lose a third of its students, those are often the students it loses.
That matters for more than fairness. Engineering and science need people who will design water systems for small towns, keep bridges standing, write safe software and run laboratories for decades. Every capable student who leaves because of how a first-year course was taught is a loss the profession pays for later.
What Rigor Should Mean
None of this is an argument for easy courses. Engineering is hard because the world is hard: loads are uncertain, materials are imperfect, and mistakes can hurt people. Students need to be held to high standards, and some will discover that the field is not for them. That is normal and healthy.
But there is a difference between a course that is rigorous and a course that is designed to lose people. A rigorous course sets high expectations and then gives students what they need to meet them: practice, feedback, chances to fail cheaply and improve. A weed-out course sets high expectations and treats the resulting casualties as proof of its standards.
The universities that have moved away from weed-out culture have not lowered the bar. They have redesigned the path toward it, with active learning, peer support, early warning systems and faculty who are rewarded for teaching well. The results, where they have been studied, are encouraging.
The old first-day speech assumed that one in three students did not belong. A better version would say something different: this course is hard, and it is our job to help you through it. Look to your left and to your right. Our goal is for all three of you to be here in May.
❦