Sciwrite Proposals

---
name: sciwrite-proposals
description: >
  Write the documents an academic career runs on besides papers: narrative review
  articles, grant and fellowship applications (specific aims, research plan),
  letters of recommendation, and personal statements. Based on Stanford's Writing
  in the Sciences. Use this skill whenever someone mentions writing or reviewing a
  grant, a fellowship application, a research proposal, a specific aims page, a
  research plan or strategy, a review article or literature review, a letter of
  recommendation or reference letter, a personal statement, a motivation letter,
  or an application essay for a PhD programme, postdoc, scholarship or residency.
  Also use it when someone asks how to ask for a recommendation letter, how to
  narrow a review topic, what reviewers of a fellowship are actually scored
  against, or how to respond to reviewers of a grant or fellowship application.
  Do NOT use it for research papers (sciwrite-manuscript) or for
  sentence-level editing alone (sciwrite-prose).
---

# Proposals, reviews, letters and statements

Four genres that share a property: the reader is deciding something about you or
your project, quickly, alongside many others. All the prose rules still apply —
use **sciwrite-prose** — but each has its own structure, and structure is where
these are won or lost.

Answer in whatever language the user writes to you in.

---

## Narrative review articles

A good project for a graduate student or early investigator: you read the current
literature properly, practise synthesis and organisation, and a well-made review
earns citations because it is useful.

**The single decision that determines whether this works: a narrow, clear
theme.** "I will review microarrays" or "I will review breast cancer" is not a
review, it is a career. "Microarrays in microbiology", "statistical techniques for
microarray data", "the history of microarray technology", "breast cancer and
exercise" — those are tractable.

**Three kinds, and this skill covers the first.** *Narrative* (non-systematic) —
your synthesis of the recent primary literature. *Systematic* — a rigorous,
prespecified search strategy aiming to find every relevant study, including
unpublished ones. *Meta-analysis* — a systematic review that statistically pools
data across studies. The latter two carry methodological requirements beyond this
skill.

**Pre-writing dominates.** Reading, note-taking and organising will take far more
time than writing. Decide the sections and subsections **early**, then file notes
from each paper directly into the section they belong to. A reference manager
(Zotero, Mendeley, EndNote) is not optional at this scale.

**Structure:**

- Abstract — written last
- Introduction — states the theme and the aim of the review explicitly
- Body — sections with headings, subsections with subheadings. In each: what is
  known, what remains unknown, and how future studies could do better
- Conclusion — what you found, plus **specific recommendations** for what the
  field should do next
- References — a narrative review can easily run to 150+

**Write for a broad audience.** People read reviews to enter a field. Do not
abbreviate heavily and do not assume prior knowledge. Numeric data across studies
goes in a summary table; tangential but interesting material goes in a sidebar or
box rather than being forced into the flow.

---

## Grants and fellowships

Full detail in `references/specific-aims.md` and `references/research-plan.md`.
Start there when someone is actually writing one. The framing that matters most:

**Get the information before you write.** Compile every opportunity you are
eligible for; ask peers and mentors what they applied to; check whether your
institution keeps a list. Then, for the one you are targeting, read the funding
announcement in its entirety — including linked attachments — and, critically,
**identify the review criteria**, because those are the questions your reviewers
will literally be scored against. For an NIH NRSA fellowship, for example, they
are: the applicant, the mentor and collaborators, the research training plan, the
training potential, and the institutional environment. Also find out your own
institution's internal deadlines, which fall before the sponsor's.

**Write regularly, not in bursts.** Break the application into tasks with a dated
timeline. Write daily or every other day, and make the time non-negotiable.
Writing becomes a habit and the anxiety drops.

**Find the gap, not just the topic.** A compelling proposal names a critical
knowledge gap or bottleneck — something that, if solved, moves the field more than
incrementally. Keep a running list of the problems inherent to your field, updated
after every paper and seminar, and narrow it through conversation with colleagues
and mentors. Compelling projects often join two threads that nobody had connected.

**Get feedback early, often, and widely.** Your reviewers will come from mixed
backgrounds, so your readers should too — including non-scientists, who are
excellent judges of clarity. Tell each reader what kind of feedback you want:
concept and feasibility, or grammar and readability. The specific aims page exists
partly to make this cheap: it is one page, so people will actually read it.

**Tell one consistent story across every document.** An application is many files
— aims, research plan, biosketch personal statement, training plan. The long-term
goal stated in one must be the long-term goal stated in the others. Leave one to
two weeks after drafting everything to read the whole package for coherence.

**Follow the formatting rules exactly.** Page limits, margins, fonts. Applications
that do not conform can be administratively rejected before review. And proofread:
an application full of errors invites reviewers to wonder how carefully you will
run the experiments.

**Plan to resubmit.** Announcements often have similar requirements, so an
application can be recycled across opportunities — check whether the sponsor
restricts concurrent submissions. And a rejected proposal improved with the
reviewers' comments is a stronger proposal. At roughly 30 % odds, persistence is
the strategy.

---

## Letters of recommendation

Full guidance, including the calibrated language, in `references/letters.md`.

**Decline if you cannot write a strong letter.** Turning someone down is doing
them a favour — a lukewarm letter sinks an application more efficiently than no
letter. (Letters exist in which the writer got the candidate's gender wrong. That
should reflect on the writer; it reflects on the candidate.) Calibrate to what
they are applying for: a small internal grant needs you to vouch that they are a
student in good standing; a competitive fellowship needs you to know them.

**Never ask a candidate to draft their own letter.** It is common practice in some
places and it defeats the purpose of an independent assessment. If you are only
willing to edit their draft, decline instead. If you are the *candidate* and
someone asks you to draft your own, that is a signal to go elsewhere.

**Show, do not tell.** Specific stories are what readers remember. Not *she meets
deadlines*, but *I sent her requests for edits at 10 a.m. and she returned the
revision by 3 p.m. the same day*.

**Length is signal.** A strong candidate gets a second page. Several short
one-page letters read, to an admissions committee, as collective lack of
enthusiasm.

**Quantify.** *She is among the top 10 % of Master's students I have taught at
Stanford* — the percentile, and the comparison group.

---

## Personal statements

Full guidance in `references/letters.md`.

**It should be personal, and it should feel slightly risky to write.** A stilted
list of accomplishments earns the word *blah* in a reviewer's notes. Reveal
something true about yourself and connect it to why you want this.

**Show, do not tell** — the same rule as letters. Readers remember stories, not
abstractions.

**Do not recite your CV.** They have it. This is the Results section problem in
another register: your job is to select and add, not to read the table aloud.

**Address weaknesses head on.** A poor transcript or a gap will be noticed
whether or not you mention it. Silence gets you discarded; a candid explanation
showing self-awareness and growth gets you noticed. This is the limitations
section of a manuscript, doing exactly the same work.

**Do your homework on the institution.** Say specifically why this programme.
Name people whose work you know. Everyone likes a compliment, and it demonstrates
you did not send the same essay to twelve places.

**Structure:** a strong, memorable opening (a story or scene beats a thesis
statement, and a longer first paragraph is fine if it is compelling); where you
want to go; what experiences led you here; what makes you a strong candidate, in
stories; why this specific programme; and a conclusion that circles back to the
opening image.

---

## Reference files

| File | Read it when |
|---|---|
| `references/specific-aims.md` | writing or reviewing a specific aims page |
| `references/research-plan.md` | writing the full research strategy |
| `references/letters.md` | letters of recommendation, personal statements, asking for letters |

Apply **sciwrite-prose** to the actual sentences in all of these. Reviewers of
grants read dozens of applications; clutter costs you more here than anywhere.