Two academic recommendation letters are required for the MEXT scholarship and for almost every Japanese graduate-school direct application in 2027. They are not a formality. Together with your research plan and the professor's pre-application interest, the two letters are what convinces the admissions committee — and the funding committee, in the MEXT case — that you can actually do graduate-level research. This guide walks through who to ask, how to ask, what a strong letter looks like, and includes a complete annotated sample letter you can use as a model when working with your recommenders.
Why recommendation letters matter more in Japan than elsewhere
Japanese graduate admissions are professor-driven. The committee is checking whether the lab that wants you is making a defensible choice, and the recommendation letter is one of the few documents that gives an external academic voice on your ability. Your transcript shows grades; your research plan shows ambition; your test scores show one narrow slice of competence. The recommendation letter is the only document where another professor — someone who has worked with you, watched you struggle through a research problem, and seen the result — vouches for you in detail.
For MEXT applicants in particular, the embassy and university screening committees both read the letters carefully. A weak letter, or two letters that sound like they were written by the same person about a different applicant, can sink an otherwise strong file. A pair of strong, specific letters can pull a marginal academic record across the line. If you are still mapping out the broader application, read the MEXT scholarship 2027 complete guide first, and the application timeline for Japanese graduate schools so that the recommender request fits into the right month.
Who to ask
The right recommender is someone who has seen you do research-level work, can describe specific examples, and is willing to spend an evening writing a tailored letter. In practice, that almost always means a professor in one of these categories.
Your research advisor (mandatory if you have one)
If you have done a Bachelor's thesis, a senior research project, or any sustained independent research under a faculty supervisor, that person should be your first recommender. They have seen you propose a question, struggle with a method, recover from a setback, and produce a result. No other recommender can match that level of specific evidence. If you do not ask your research advisor, the committee will notice the absence and assume the relationship ended badly. The only acceptable reason to skip your research advisor is if they have left academia, become unreachable, or genuinely cannot write a strong letter — and in that case, you should explain the situation in your application's optional notes section.
An undergraduate professor who taught a research-heavy class
A second strong choice is a professor of an upper-division course that required a substantial research paper, a term project, or a significant analytical assignment. Lecture-only classes where you sat in the back row are not enough; the professor needs to be able to describe a specific piece of work you produced. A capstone seminar, a graduate-level course you took as an advanced undergraduate, or a research-methods course are all good sources. The work they describe in the letter will be more narrowly scoped than the research-advisor letter, but if it is concrete, it complements rather than duplicates.
A lab supervisor (PI) from research experience
If you worked in a research lab as an undergraduate research assistant or lab member — even unpaid, even part-time — the principal investigator (or a postdoctoral researcher who directly supervised you) is an excellent recommender. They have seen you in a research environment outside the classroom. If you spent a summer in an external lab, that supervisor is often the strongest second letter you can get because they observed you when you were not being graded.
A summer internship mentor (academic, not industry)
Research-focused summer programs — REU programs in the United States, JSPS Summer Program participants in Japan, DAAD summer research, RIKEN summer interns — produce excellent recommenders. The mentor saw you work for six to ten weeks on a defined project and can speak to your independence, technical skill, and collaboration in a way classroom professors usually cannot. If you participated in such a program, the mentor is often a stronger second-letter source than a professor whose course you took.
Who NOT to ask
A few categories of recommender consistently produce weak letters that hurt rather than help your application. Avoid:
- Family friends, even prestigious ones. A letter from a family friend who happens to be a professor — but who has never taught you or supervised your research — reads as exactly what it is: a favor. Japanese committees see this immediately. The letter says nothing useful about your ability and signals that you could not find a real academic recommender.
- Non-academic supervisors only. A letter from your software engineering manager, your tutoring center director, or your part-time-job supervisor is not an academic recommendation. For research-focused programs and for MEXT, this category does not substitute for an academic letter under any circumstances. If you have been working full-time for several years, see the FAQ on this question above.
- Professors you barely interacted with. A famous professor whose large lecture you attended but never spoke to in office hours will produce a generic letter saying you got an A in their course, nothing else. The committee already has your transcript; they do not need a letter to learn the grade. The prestige of the recommender does not compensate for the absence of specific content.
- Teaching assistants or graduate students. A letter from a graduate student or postdoc who supervised you can be useful as supporting material, but it does not replace a letter from a tenure-track professor. If a postdoc was your effective day-to-day supervisor, ask them to write the letter and have the principal investigator co-sign or write a parallel short letter referring to the postdoc's evaluation.
- Anyone you suspect will write a lukewarm letter. A lukewarm letter is worse than no letter. If a professor hesitates, looks uncomfortable, or asks "are you sure I am the best person?" — they are telling you the answer. Thank them for their time and find someone else.
How to ask politely
The request itself is part of the recommendation process, and a polite, well-prepared request makes a measurable difference in the quality of the letter you get back. Japanese professors, in particular, expect the request to come with all the materials they need and a realistic deadline. The pattern that works:
- Ask in person if possible, by email if not. Make a fifteen-minute appointment in the professor's office hours, explain your plans, and ask explicitly: "Would you be willing to write a strong recommendation letter for my graduate-school applications?" The word "strong" matters. It gives the professor an honest exit if they cannot write a strong one.
- Give them four to six weeks of notice. Eight is better. Recommenders with less than three weeks of lead time produce weaker letters because they cannot tailor or revise. If your deadline is December 15, ask by late October at the latest.
- Provide a complete information packet. Send the recommender, in one email, all of: your CV, your research plan or statement of purpose, your transcript, the list of programs you are applying to with deadlines, the submission method for each (email, online portal, sealed envelope), and a brief document reminding them of the specific work you did with them. This last document is the most important — it gives the recommender concrete material to cite. See the sample MEXT field-of-study statement for the research-plan format the recommender will need to read.
- Offer to draft if they prefer. Some professors will say "send me a draft and I will revise." Treat this as a normal request, not a sign of disinterest. Write a draft in their voice, with specific examples they would actually remember, and send as an editable Word document.
- Send a reminder two weeks before the deadline. One short, polite email. Confirm the submission method and deadline. Do not send three reminders.
- Send a thank-you note after submission. A handwritten card if you can, an email if you cannot. Include where you ended up applying so the professor sees the outcome of their work. This matters for future recommendations — most graduate-school applicants ask the same recommender for PhD applications, post-docs, and jobs over the next decade.
The four things Japanese committees look for in a recommendation letter
Different cultures emphasize different things in academic recommendations. US committees often look for stories about character and grit; UK committees lean analytical and precise. Japanese committees look for four specific signals, and a letter that addresses all four is dramatically stronger than a letter that addresses only one or two. For a more granular view, see what Japanese professors look for in international applicants .
1. Research ability
The single most important signal. Can the candidate define a research question, design an approach, execute the work, interpret results, and revise based on feedback? The letter should describe a specific project, not in vague terms but with enough detail that a reader in the same field can picture the work. "She conducted original research on protein folding" is weak. "She extended the equivariant message- passing framework to handle disordered protein regions, ran experiments on a curated benchmark of 1,200 IDR sequences, and identified a failure mode in the standard rotation-invariance assumption that we are now writing up as a workshop paper" is strong.
2. Work ethic and independence
Japanese labs operate on a long-hours, high-commitment culture. Committees want to know that the applicant will not collapse under pressure, will work through the tedious parts of research without supervision, and will recover from setbacks. A letter that describes a moment when the candidate hit a wall and worked through it is far more valuable than a letter that describes a candidate who never had a problem.
3. English (or Japanese) communication ability
The committee needs to know whether the applicant can function in the language of the lab. If you will work in English, the letter should mention that you communicate professionally in English; if you will work in Japanese, the letter should mention your level explicitly. Vague "good communication skills" is unhelpful. "Her written English is publication-ready and she presented at the lab seminar in English without difficulty" is useful. If you have an active JLPT result, make sure your recommender knows about it — see the JLPT N3 study hub for context on the level Japanese committees consider basic for academic survival.
4. Fit for graduate study
Finally, the committee wants to know that the applicant is not just a strong undergraduate but is genuinely suited to a research career. The letter should explicitly address graduate-level potential: capacity for sustained independent work, intellectual curiosity that goes beyond coursework, and a research direction that the recommender finds credible. The closing paragraph of the letter is the natural place for this.
Full sample recommendation letter
Below is a complete sample recommendation letter, approximately six hundred words, from a fictional Computer Science / Electrical Engineering professor recommending a strong Master's applicant. The letter is realistic, substantive, and addresses all four of the criteria above. Do not copy the wording verbatim — committees compare letters across applicants, and a letter that matches a published template gets flagged. Use it as a structural model for what your recommender's letter should look and feel like.
[University Letterhead]
Department of Computer Science
[University Name]
[Address]
[Date]
To the Graduate Admissions Committee:
I am writing to recommend Ms. Aanya Sharma for admission to the Master's program in your
department, with my strongest endorsement. I have known Aanya since September 2024, when she
joined my research group as an undergraduate research assistant during her third year. Over
the eighteen months since, I have supervised her work on equivariant graph neural networks for
molecular property prediction, and she has become one of the strongest undergraduate
researchers I have worked with in the past decade.
Aanya's principal contribution to my lab has been her undergraduate thesis project, in which
she extended our group's 2024 equivariant message-passing framework to handle small molecules
with conformational flexibility — a problem that defeats the standard rotation-invariance
assumption used in most current architectures. The starting framework was non-trivial; the
mathematical machinery is dense and the implementation crosses two large open-source
libraries. Aanya read herself into the problem in approximately six weeks, identified two
limitations of the original formulation that my postdoc and I had not flagged, and proposed a
modified parameterization that handled flexible molecules without sacrificing the equivariance
guarantee. Her implementation reproduced our published baseline within numerical tolerance and
beat it by roughly four percent on the QM9-flex benchmark she constructed. We are currently
preparing this work as a workshop submission to NeurIPS 2026 with Aanya as first author.
What distinguishes Aanya from other strong undergraduates I have supervised is the
combination of mathematical rigor and engineering discipline. Many students at her stage
either understand the theory but cannot execute the code, or write working code but cannot
explain why it should work. Aanya can do both, and more importantly, she catches her own
errors. In November 2025 she identified a subtle bug in our rotation-invariant pooling layer
that the rest of the lab — myself included — had missed for several weeks. Rather than
flagging it and waiting for instructions, she wrote a minimal reproduction, proposed two
candidate fixes, ran the experiments to compare them, and presented the result at our
weekly lab meeting. This is graduate-level research behavior in an undergraduate, and it is
the most reliable predictor I know of success in a Master's program.
Aanya's written and spoken English are excellent. She presented her thesis work at our
department's undergraduate research symposium in March 2026 and fielded questions from a
mixed audience of theorists and applied researchers without difficulty. Her thesis manuscript
required minimal editing for clarity. She has also begun studying Japanese on her own
initiative in preparation for graduate study in Japan and currently holds JLPT N3.
I will note one mild caveat: Aanya tends to under-discuss her own work in informal settings,
and occasionally needs encouragement to defend her contributions in larger lab discussions.
This is a common pattern in strong students who are genuinely modest, and I expect it to
diminish as she gains experience presenting in the larger Japanese lab seminars. It is not
an indication of weakness in the work itself.
In summary, Aanya is precisely the kind of student a research-active Japanese lab should
want: technically strong, intellectually independent, capable of sustained work on hard
problems, and genuinely committed to a research career. I recommend her without
reservation for admission to your Master's program, and I would be happy to discuss her
candidacy further by email or video call.
Sincerely,
[Recommender Name]
Professor of Computer Science
[University Name]
[Email] / [Phone]
Per-paragraph annotations
Opening paragraph. Names the applicant and the program clearly, states the strength of the endorsement, and immediately establishes the relationship — when it began, what context, how long. The reader knows in twenty seconds whether the recommender knows the applicant well enough to write a useful letter. "I have known the applicant for several years through coursework" would be a weak opener; "supervised her undergraduate research for eighteen months" is the strongest possible context.
Research paragraph. The most important section. Names a specific project, describes the technical content with enough detail that a reader in the field can evaluate it, gives concrete numerical results (four percent improvement, six weeks of background reading, a co-authored workshop submission). A reader can verify this is real work by checking the publication record. This is the research-ability signal Japanese committees look for. For applicants in CS / EE specifically, see Computer Science Master's in Japan and studying AI and ML in Japan for the research areas committees expect to see addressed.
Work-ethic paragraph. Tells a specific, dated story (the November 2025 bug). The story illustrates two things at once: technical depth (the student caught what the supervisor missed) and independence (she did not wait for instructions, she ran experiments and presented the result). Stories beat adjectives in every recommendation letter. "She is hardworking" is forgettable; "she identified the bug, proposed two fixes, ran the comparison, and presented the result" is concrete evidence.
Communication paragraph. Addresses English directly, with a verifiable instance (the March 2026 symposium presentation) rather than a generic "good English." Closes with the JLPT N3 note, which signals to a Japanese committee that the applicant has invested in language preparation. This matters enormously for fit assessments.
Caveat paragraph. The single most important non-obvious feature of a credible recommendation letter. A letter with no caveat reads as either written by the applicant themselves or written carelessly by a recommender who does not know the applicant well. A mild, fair caveat — phrased to indicate growth rather than weakness — adds credibility to every other claim in the letter. Western applicants in particular often resist asking for letters with caveats; the right framing is that a letter with one fair caveat is substantially stronger than a letter with none.
Closing paragraph. Reaffirms the endorsement, ties the recommendation back to the specific context (a research-active Japanese lab), and offers further contact. The offer to discuss by email or video call signals that the recommender stands behind the letter and would be willing to defend it. It is rare for committees to follow up, but the offer matters.
Common mistakes recommenders make
Even well-intentioned recommenders produce weak letters. If you have any influence on what your recommender writes — by drafting, by suggesting bullet points, or simply by sharing this guide with them — try to head off the four most common problems.
Too generic
The most common failure mode. The letter says the applicant is intelligent, hardworking, and a good communicator, but never describes a specific project or result. A reader cannot distinguish this letter from one written about a different student. The fix is concrete examples: name the project, the method, the result, the dates.
Overly long
A four-page recommendation letter is not stronger than a one-page one. Committees read dozens of letters per cycle; a long, meandering letter signals that the recommender did not bother to edit. Six hundred to eight hundred words on a single page is the sweet spot for a strong academic recommendation. Anything beyond a page and a half should be trimmed.
Focusing on undergraduate coursework only
A letter that lists the courses you took with the recommender, your grades, and generic praise about "her dedication to her studies" is essentially a duplicate of your transcript. Japanese committees want to know about research, independence, and graduate-level potential. If the recommender only saw you in a lecture course, you are asking the wrong recommender.
No specific research details
Closely related to the "too generic" problem, but worth calling out separately: even when the recommender does describe a research project, they often describe it at the wrong level of abstraction. "She did a thesis on machine learning" is not specific. "She extended an existing equivariant architecture to handle conformational flexibility, validated it on a custom benchmark, and improved on the prior baseline by four percent" is specific. The fix is for the recommender to write about the work the way they would describe it to a colleague at a conference.
International-applicant tips: language and formatting
Japanese graduate programs accept recommendation letters in either English or Japanese. The decision should be driven by the recommender's own language ability, not by which language you think the committee prefers.
If your recommender is non-Japanese, the letter should be in English. A letter translated into Japanese by a non-native speaker — or worse, by a translation service — almost always reads worse than the original English would have, and committees can tell. MEXT explicitly accepts English-language letters, and so do almost all Japanese national and private universities. See the MEXT embassy recommendation 2027 and MEXT university recommendation 2027 for the two MEXT tracks and the language conventions for each.
If your recommender is Japanese and writes academic Japanese fluently, a Japanese-language letter is equally welcome and at some institutions slightly preferred. Do not insist on an English translation if the recommender is comfortable in Japanese; the original-language letter will be stronger. If a program specifically requires English, the recommender can write the original in Japanese and have a faculty colleague translate, with both versions submitted.
Some applicants worry that an English letter will disadvantage them at universities with weaker English-language administrative support. In practice, in 2027, every Japanese national university and every major private university has English- capable graduate admissions staff. The language of the letter is rarely an operational problem. For the broader picture of how Japanese graduate admissions handle English applicants, read English-taught Master's in Japan 2027 .
Sealed envelope vs digital submission
Submission methods vary by program and have shifted dramatically since 2020. You will encounter four patterns in 2027:
- Direct upload by the recommender via online portal. The modern default. The applicant enters the recommender's email in the application portal, the system emails the recommender a unique upload link, and the recommender uploads a PDF directly. The applicant never sees the letter. This is the standard for almost all major Japanese national universities (Tokyo, Kyoto, Osaka, Tohoku, Tokyo Tech, Nagoya, Kyushu, Hokkaido) for international applicants in 2027.
- Email submission directly from the recommender. The recommender emails a PDF to the admissions office, with the applicant copied or not depending on program rules. Common at smaller national universities and some research-focused private universities.
- Sealed envelope submitted by the applicant. The traditional Japanese method, still required by a minority of programs. The recommender prints, signs, and seals the letter in an envelope, signs across the seal, and returns it to the applicant. The applicant submits the sealed envelope as part of a paper application package. Programs that still require this in 2027 are mostly older national-university medical and pharmacy departments and a small number of humanities programs.
- Hybrid: PDF email plus sealed envelope. The recommender submits both an email PDF for the initial review and a sealed paper letter that arrives later for the official file. This is increasingly rare and usually reserved for highly traditional programs.
Always check each program's specific requirements on the admissions page. If the program does not specify, default to email submission directly from the recommender, with the applicant copied. Never open a sealed envelope to inspect the letter — committees treat that as evidence of tampering and may discard the letter. For the broader application context across Japanese universities, see our professors directory and the guide on emailing Japanese professors , which explains the parallel process of securing professor interest before the formal application opens.
Following up if your recommender hasn't submitted
A small but real fraction of recommendation letters arrive late or not at all. Recommenders are busy, and a letter for an applicant who did not stay in close contact can slip past the deadline without being noticed. Manage this with a structured follow-up sequence rather than panicked emails.
- Two weeks before the deadline. A short, polite reminder confirming the submission method and deadline. Re-attach all the materials you originally sent. End with a sentence offering to clarify anything.
- One week before. A second short reminder, only if the first produced no response. Confirm the deadline and ask the recommender to let you know if any complication has come up.
- Three days before. A direct but respectful message. If the recommender has gone silent, offer to make the process easier — "would it help if I sent you a draft to revise?" or "I am happy to help with any formatting issues." Many letters that have stalled at this point do so because the recommender feels guilty about being late and is now avoiding the email; offering a way out usually unblocks them.
- Forty-eight hours before. If you still have no confirmation, contact the admissions office directly. Explain the situation calmly and ask whether the program can accept a recommendation letter submitted up to a week late. Most Japanese universities will accommodate small delays for international applicants where the rest of the application was on time.
- Backup recommender. Always have a third potential recommender identified before applications open. If a primary recommender genuinely cannot deliver, you can pivot quickly without panic.
Throughout this sequence, do not vent frustration to the recommender. They remain a long-term professional relationship that will outlast this single application cycle. Even if a letter arrives a week late and slightly dampens your application, a recommender who feels respected will write a better letter for your PhD applications five years later.
Bottom line
Two academic recommendation letters are required, and they matter more in Japanese graduate admissions than applicants from Western systems usually expect. Pick a research advisor and a research-engaged second recommender. Give them six to eight weeks of notice. Provide a complete materials packet. Make sure the letter addresses research ability, work ethic, language ability, and graduate-school fit, with concrete examples and at least one mild caveat for credibility. Use the sample letter above as a structural model, not a template to copy. Confirm the submission method for each program — most are now digital — and follow up politely if the deadline approaches without confirmation. Done well, two strong letters can pull a borderline application across the line; done badly, two generic letters can sink a strong one. For the next steps in your application, see the sample email to a Japanese professor to secure professor interest before the formal application opens, the sample MEXT field-of-study statement for the research-plan document, and the professors directory for finding the recommenders' future colleagues at your target labs.