Every RSI applicant has strong grades and strong scores. That is table stakes. What the committee does not know from the transcript is what you are like to teach, to mentor, or to put in a room with a hard problem. The letters fill that in, or they do not.
A good RSI letter is specific, honest, and comparative. It shows the reader something they could not have guessed from your file. That is the entire job.
The three letters
- STEM teacher. Math, science, or computer science. Preferably someone who taught you in a hard course and has seen you hit a wall and keep going.
- Second teacher. Any subject. This letter should add a dimension the STEM letter cannot: breadth of intellectual interest, writing, leadership, character.
- Research mentor. Optional, but often the most powerful letter in the file. If you have done real research, this is the voice that can speak to whether you are ready to do more.
What a strong letter does
Uses specifics
Generic praise is invisible. "Outstanding student, top of the class" is noise. A strong letter names a moment: the question you asked, the approach you tried after a problem set stumped you, the paper you brought to office hours because you disagreed with it.
Compares you to others
The reader needs a calibration. "Among the strongest students I have taught in a decade of AP Physics" tells them something. "Very smart" does not. A teacher who cannot compare you to a clear peer set cannot help you much.
Speaks to research, not just grades
RSI is not admitting a student for a class. The letter should address how you handle open problems, ambiguity, and the kind of failure that shows up in real research.
Reads honestly
Letters that describe a flawless student are less persuasive than letters that acknowledge a real person. A brief mention of something you worked through reads as credible. A flat hymn does not.
How to ask
Pick for knowledge of you, not title.
A teacher who knows you beats a famous teacher who does not. A community lab PI who saw you work beats a university name on a letter written by someone who barely supervised you.
Ask early, in person.
At least six weeks before the deadline. In person, not by email. The question is simply "Do you feel you know me well enough to write a strong, detailed letter?" The answer tells you a lot.
Give them a packet.
Resume, essay drafts, the RSI description, the deadline, specific moments you hope they might mention, and a short note on what the other letters are likely to cover so the three do not overlap.
Follow up, lightly.
A polite reminder a week before the deadline is fine and often welcome. Thank them afterward with a handwritten note regardless of outcome.
If the letter will be weak, do not submit it.
A teacher who hesitates is telling you something. It is better to find a recommender who will invest the time than to fill a slot with a soft letter.
Why the research mentor letter carries so much weight
A classroom teacher can describe how you handle assigned problems. A research mentor can describe how you handle problems with no back-of-the-book answer. That is closer to what RSI is actually asking about.
The strongest mentor letters compare you directly to undergraduates or graduate students the mentor has worked with. "Performing at the level of a solid first-year graduate student" is the kind of calibration a high school teacher cannot provide and an RSI reader takes seriously.
If you are not yet in a lab, that is the project for the fall, not a gap to paper over with a flashier teacher letter. If you are, brief the mentor carefully on what to emphasize and give them time.
Mistakes that weaken a file
- Prestige over fit. A name-brand recommender who does not know you is a worse letter than a careful one from a teacher who does.
- Asking in December. Good letters take weeks. Short-notice letters read like what they are.
- Three letters that say the same thing. If all three cover curiosity and work ethic, two slots are wasted. Brief recommenders on angle.
- No context for the recommender. A letter written without your resume, essays, and the program description will be generic by default.
- Not verifying tone. You will not read the letter, but you can ask a teacher whether they feel positive enough to write a "strong" letter. Take "I can write something" as a no.
The short version
Pick recommenders who know you, ask early, give them what they need to write specifically, and make sure the three letters cover different ground. The file is doing its job when each letter tells the committee something the other materials did not.
Helpful resources
Related topics