De-anxietizing intro STEM.

The visit I get most often in office hours starts with an apology. "I'm so sorry, I know this is a stupid question, I'm just not a physics person." The student is usually a nursing or allied-health major. Physics is required for her program, she is certain it will end her plan, and she has waited three weeks longer than she should have to come see me, because walking through my door felt like admitting something. What follows is a composite of many such visits, and if you teach intro STEM anywhere, you have hosted it too.

Here is what I have learned to believe, from the research and from the chair across from mine: she is almost never bad at physics. She is afraid of it, behind in something underneath it, or both. Those are three different problems, they have three different fixes, and a large lecture course is built to address none of them.

The scale of it

We should stop talking about these students as edge cases. Anatomy and physiology, the gateway to nursing, enrolls around 450,000 students a year in the US, and 30 to 40 percent don't pass on the first attempt. Gateway algebra and statistics run D/F/withdraw rates above 20 percent. Introductory courses push students out of STEM pathways disproportionately by race and income, which tells you something important: the washout is not a neutral measurement of talent. Some departments still call these courses "weed-outs," with a straight face, as if attrition were rigor. It is mostly just attrition.

What fear does to a test score

Test anxiety is not a mood. It is a measurable performance tax. In physics courses full of bioscience majors, students with higher test anxiety score lower on high-stakes exams specifically, while performing fine on lower-stakes work, and women report more of it than men in the same rooms. The mechanism is well studied: worry occupies working memory, and working memory is exactly what a physics problem needs. The student who blanks on the exam and then solves the same problem in my office an hour later is not lazy or dishonest. Her exam measured fear and knowledge blended together, and we wrote one number on it and called the number her.

The second engine of failure is the spiral. Intro STEM is cumulative, so missing two weeks (a sick kid, extra shifts, one bad breakup) makes every following lecture more expensive. Shame arrives, help-seeking stops, and by November the syllabus says "momentum" while the actual gap says "algebra from two years ago." Two thirds of undergraduates work for pay, so the two missed weeks are not a character flaw either. They are a schedule.

What the evidence says helps

Retrieval practice, with the stakes removed. Frequent low-stakes quizzing beats re-reading by a wide margin in controlled studies, and it carries a quieter benefit: exams stop being exotic. A student who has retrieved under mild pressure forty times meets the midterm as the forty-first, familiar thing.

Fix the actual gap, at the actual level. Telling a struggling student to "review the fundamentals" is a shrug dressed as advice. The productive move is diagnostic: find where understanding stops, often two courses below, and rebuild from there. This is old mastery-learning logic, and adaptive redesigns built on it have lifted gateway success rates by around ten points in institutional pilots.

Ten minutes of writing. In one of my favorite results in the literature, students who spent ten minutes before an exam writing about their worries significantly outperformed controls, with the largest gains among the most anxious. The interpretation: unloading the worry frees the working memory it was squatting in. Cheap, fast, replicated.

Teach the studying, not just the subject. Training students in planning, monitoring, and self-testing measurably improves achievement, with some of the strongest effects for students who arrived behind. The students who most need these skills are the least likely to have been taught them, anywhere, ever.

A human who notices. Structured coaching raised persistence by 4 to 7 percentage points in a randomized trial, and the mechanism was mundane: someone checked in, connected daily work to the student's own goal, and made asking for help feel normal. The honest problem is cost. A human coach for every anxious student does not fit any budget I have seen.

What I'm doing about it

In my own courses this year: more low-stakes retrieval, an explicit "if you missed weeks, come get a bridge, not a lecture" policy, and the writing exercise before every exam. It helps. It also runs into the ceiling every instructor knows: I am one person, office hours are four a week, and the student who most needs the bridge is the least likely to come ask for it.

That ceiling is why Pillar Three of ThinkAthena exists. We are designing, in the open, a course companion that does the diagnostic and the rebuild individually, treats anxiety as a design constraint, and behaves like the evidence above instead of like a louder lecture. It is at the design stage, and the design is public precisely so the students and professors it's for can tell us where it's wrong.

"I'm not a physics person" is a story students are handed, usually early, usually by a grade that was measuring something else. Stories can be rewritten. The physics was never the unclimbable part.