# Research Statement

> Write, review, and revise academic Research Statements / Research Visions / Research Plans / 5-Year Plans for faculty, postdoc, fellowship, and promotion applications across disciplines and regions. Covers Mission/Strategy/Evidence/Story, US-school guidance from Cornell, CMU, MIT, Caltech, Penn, Yale, Harvard, UConn, Notre Dame, Delaware, Berkeley, length tiers, openings, past-work theming, fundable future lines, department fit, funding vectors, rubric scoring, red flags, and examples. Use when asked to write, draft, outline, edit, critique, score, red-team, or improve any research statement, research vision, research plan, future research statement, or research pitch.

- Skill: `drunkcoding/research-statement` (Agent Skill, multi-file: 15 files)
- Install (CLI): `npx skillmds@latest add drunkcoding/research-statement`
- Raw SKILL.md: https://api.skillmd.com/api/skills/drunkcoding/research-statement/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: drunkcoding (https://skillmd.com/u/drunkcoding)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/drunkcoding/research-statement

---


# Research Statement

## Core Purpose

Write a research statement that **gets the candidate an interview**. The document's only job is to convince a non-specialist search-committee member, in 5 minutes of skimming, that:

1. The candidate has a **clear research vision** (one sentence they could repeat at lunch).
2. The candidate has the **track record** to execute it.
3. The candidate's **next 3–5 years** will produce fundable, citeable, hireable work.
4. The candidate **fits this department** specifically — not a generic department.

Everything else — eloquence, autobiography, exhaustive coverage — is secondary and usually counterproductive.

## Final delivery: humanize the output

**REQUIRED SUB-SKILL (run before returning any draft to the user).** Load the `humanizer` skill (via the skill tool) and run the finished text through it as a final editing pass. It removes AI-writing tells (inflated significance, copula avoidance like "serves as", rule-of-three, overused AI vocabulary such as "delve"/"testament"/"landscape", filler phrases, sycophancy, and em/en-dash overuse) without inventing any fact, name, number, date, or citation. Preserve the format and conventions this skill requires (length tier, section order, department-fit specifics), and if the user supplied a writing sample, pass it so humanizer matches their voice.

## The Mission/Strategy/Evidence/Story Frame

Every effective research statement, regardless of discipline or region, can be checked against four questions. If any answer is missing, the draft is incomplete.

| Element | Question the Reader Asks | Where It Lives |
|---|---|---|
| **Mission** | What single problem/question drives this person's research? | Opening paragraph, one sentence |
| **Strategy** | How do they propose to attack it? Is it credible? | Future-research section |
| **Evidence** | Have they done it before? Can they do it again? | Past-work section |
| **Story** | Why this person, why now, why here? | Voice + closing + fit paragraph |

This frame is adapted from Arie van Deursen's TU Delft hiring rubric and matches what NSF CAREER, NWO Veni, and ERC StG reviewers score on. **Use it as the master checklist for every draft.**

## Workflow

### Stage 1: Diagnose the Brief (15 min, mandatory)

Before drafting anything, extract from the job ad:

1. **Document name** — "research statement", "research vision", "research plan", "future research statement", "5-year plan". These are not interchangeable. See `references/terminology.md`.
2. **Length cap** — 1, 2, 3, 4-5, or 7-10 pages? This dictates content allocation. See `references/length_tiers.md`.
3. **Region/venue** — US R1, UK, EU/Dutch, Canadian, Australian, Asian. Voice register and fit-paragraph weighting differ. See `references/regional_voice.md`.
4. **Discipline** — CS / engineering / life sciences / math / social sciences / humanities have different expectations for past-work density, future-work specificity, and the role of grants/equipment/archives. See `references/discipline_variants.md`.
5. **Special instructions** — words like "from the heart", "intrinsic motivation", "vision", "fit with our research lines", or "potential for funding" each carry **specific obligations** the brief is testing for. Treat each as a checkbox.

For any **US faculty or postdoc application**, also load `references/us_school_guidance.md`. US university guidance is not a single R1 template: Cornell / CMU / Penn / Yale often describe 1–4 page committee-readable statements, MIT / Harvard / Caltech often expect STEM statements with 3–5 pages and multi-horizon planning, and Berkeley Math expects a nontechnical executive summary before technical depth. The job ad and discipline still win.

If the brief uses "from the heart" / "personal motivation" / "what drives you" language, **also load `references/personal_voice.md`** — these phrases are common in Dutch and Northern European applications and require a calibrated register that is neither US-emotional nor sterile-CV.

### Stage 2: Draft Inside-Out (5 short sessions)

Drafting in the order the document is read produces bad statements. Drafting **inside-out** produces good ones:

1. **North Star** — one sentence stating the animating question. Iterate 5+ versions before moving on.
2. **Future research lines** — 2–3 (rarely 4) lines, each shaped to be NSF-CAREER / Veni / ERC-StG / NIH-K fundable. See `references/future_research.md` for the 6-element shape.
3. **Past-work themes** — 2–3 themes, each told as Problem → Approach → Result → Impact (PARI). Organize **by theme, not chronology**. See `references/past_work.md`.
4. **Fit paragraph + closing credo** — name 2–3 specific groups, 1–2 specific faculty (with one specific paper or project each), 1 cross-cutting initiative, and one realistic funding pathway. See `references/department_fit.md`.
5. **Opening paragraph** — written LAST. Must do four jobs in 4–6 sentences: name the problem domain, name the gap, state the vision, hint at method. See `references/openings.md`.

For US statements, run the **brand/story overlay** from `references/us_school_guidance.md` after the North Star: state the candidate's research identity in one sentence, then test each past theme and future line against CMU's Sequence / Novelty / Problem moves and Caltech's ABT (And–But–Therefore) story logic.

### Stage 3: Red-Team & Score (mandatory before delivery)

A draft is not done until it has passed two checks:

1. **Score against the rubric** in `references/rubric.md`. Any criterion below threshold → revise that section before continuing.
2. **Run the red-flag scan** in `references/red_flags.md`. Every flagged phrase must be removed or replaced.

If revising for someone else, **always perform both checks and report scores explicitly** — do not just say "looks good".

## Universal Structural Skeleton

This is the load-bearing structure every effective research statement implements, scaled to length tier.

```
[1] Opening / Vision         — North Star + framing  (≈10–15% of length)
[2] Past Work                — 2–3 themes, PARI shape (≈30–40%)
[3] Future Research          — 2–3 fundable lines    (≈35–45%)
[4] Department Fit           — specific groups/people (≈8–12%)
[5] Closing / Credo          — personal commitment   (≈3–5%)
```

For 1-page UK statements, [4] often collapses into the cover letter. For 7-10 page humanities statements, [2] expands into a "book project" arc. For most CS / engineering positions, this allocation holds. See `references/length_tiers.md` for tier-specific allocations.

## Region & Discipline Selection

Read these references **only** when they apply:

| Trigger | Reference to load |
|---|---|
| Dutch / TU Delft / TU/e / UvA / VU / Utrecht / Leiden / Groningen / Wageningen | `references/regional_voice.md` (Dutch section) + `references/personal_voice.md` |
| US R1 / US teaching-research institutions (MIT / Stanford / CMU / Berkeley / Cornell / Penn / Yale / Caltech etc.) | `references/regional_voice.md` (US section) + `references/us_school_guidance.md` |
| UK (Cambridge / Oxford / Imperial / UCL / Edinburgh / Russell Group) | `references/regional_voice.md` (UK section) |
| Other EU (ETH / EPFL / KTH / TU Munich / Saarland / MPI) | `references/regional_voice.md` (Continental section) |
| Canada / Australia / Asia | `references/regional_voice.md` (final section) |
| CS / engineering / ML systems / SE / PL / security / HCI | `references/discipline_variants.md` (CS/Eng section) |
| Life sciences / biology / chemistry / medicine | `references/discipline_variants.md` (life-sci section) |
| Math / theoretical CS | `references/discipline_variants.md` (math section) |
| Social sciences / economics / political science / psychology | `references/discipline_variants.md` (social-sci section) |
| Humanities (history / literature / philosophy / classics) | `references/discipline_variants.md` (humanities section) |

If the user does not specify, **ask once** which discipline + region apply. Do not guess past CS/US default.

## Critical Anti-Patterns (always block)

Never let a draft ship with any of these. Full list with replacement language in `references/red_flags.md`.

1. ❌ Childhood / "ever since I was young" opening
2. ❌ Definition of the field as opening sentence
3. ❌ "Passion", "passionate", "love", "fascinated" as standalone declarations without immediate evidence
4. ❌ "Novel", "groundbreaking", "revolutionary", "paradigm-shifting"
5. ❌ "Extending" or "adding to" advisor's work — claim independence
6. ❌ Negative positioning — "Previous work fails to..." — stay in what YOU did
7. ❌ Generic department praise ("X University is world-renowned in...")
8. ❌ Future research lines under 80 words each — too vague to fund
9. ❌ Zero named faculty / groups / initiatives in the fit paragraph
10. ❌ No funding-vector mention where the brief asks for "potential for funding"
11. ❌ Closing with thanks / hopes / "I look forward to..."
12. ❌ Over-length — more disqualifying than under-length

## What "From the Heart" Actually Means

In Dutch, German, and some UK briefs, the instruction "written from the heart" / "intrinsic motivation" / "what drives you" is **not a request for emotional disclosure**. It is a request for a **personal scholarly credo**: the conviction that a specific question MUST be answered, evidenced by a sustained pattern of work directed at it. The model is closer to Hamming's *You and Your Research* or Dijkstra's EWD essays than to a US diversity statement.

Concretely, this means **show, don't tell**:

- ❌ "I am passionate about reliable software."
- ✅ "I have spent the last six years on a single question: why does software fail in ways its authors did not imagine?"

Full guidance and a phrasebook of authentic motivation patterns is in `references/personal_voice.md`. Load it whenever the brief uses heart/passion/motivation/vision language explicitly.

## Self-Editing Protocol

Run a 5-pass revision in this order. Each pass takes 15–30 minutes. Earlier passes catch structural issues; later passes catch surface issues. Do not invert.

1. **Vision pass** — Can you state the North Star in one sentence? Does the opening paragraph deliver it? If a non-specialist reads only paragraph 1, can they summarize?
2. **Claim-evidence pass** — Every assertion of impact backed by specific evidence (paper, citation count, deployment, award)? Every future-research line has a method, an outcome, a horizon, and a funding vector?
3. **Fit pass** — At least 2 specific groups + 1–2 named faculty + 1 cross-cutting initiative + 1 realistic funding pathway?
4. **Voice pass** — First-person singular consistent? Hedging level matches region? No anti-pattern phrases (`references/red_flags.md`)?
5. **Mechanics pass** — Length within cap? Tense consistent (past for done, present for ongoing, future for planned)? No typos? Could not be pasted into a different application without major editing?

If any pass fails, fix and **re-run all subsequent passes** — fixes propagate.

## References (load on demand)

Core references — load whenever drafting or reviewing:

- `references/rubric.md` — Self-grading scorecard with 8 criteria, 0–5 scale, total /40. Use to score every draft before delivery.
- `references/red_flags.md` — Annotated list of phrases and patterns that mark weak statements, with concrete replacements.
- `references/openings.md` — 10 successful opening paragraphs verbatim, decomposed; opening anti-patterns; the four-job opening checklist.
- `references/past_work.md` — PARI structure, theme selection, advisor-independence rules, quantification guidance.
- `references/future_research.md` — The 6-element fundable-line shape; horizon planning; high-risk vs. safer line balance; funding-vector vocabulary.
- `references/department_fit.md` — How to research a target department; the 2-groups-2-faculty-1-initiative formula; the gap-not-praise rule.

Conditional references — load only when applicable:

- `references/personal_voice.md` — When brief contains "heart", "passion", "intrinsic motivation", "drives you", or "vision". Heavy emphasis on Dutch / Northern European register.
- `references/us_school_guidance.md` — For any US faculty/postdoc statement or when the user cites Cornell, CMU, MIT, Caltech, Penn, Yale, Harvard, UConn, Notre Dame, Delaware, Berkeley, or similar US guidance. Source-synthesized overlay for audience, length, story, future planning, institutional fit, funding, design, and feedback.
- `references/regional_voice.md` — When the target region or register matters. Sections for US, UK, Dutch/EU, Continental EU, Canada/Australia/Asia.
- `references/discipline_variants.md` — Per-discipline rubric variations and example structures. Sections for CS/Eng, life sciences, math, social sciences, humanities.
- `references/length_tiers.md` — Content allocation tables for 1, 2, 3, 4–5, and 7–10 page caps; what gets cut at each tier.
- `references/terminology.md` — Disambiguates "research statement" vs. "research vision" vs. "research plan" vs. "5-year plan" vs. "research pitch" vs. "future research statement".
- `references/examples.md` — 10 dissected real opening paragraphs from successful CS faculty applications, with structural breakdowns. Patterns extend to other disciplines.
- `references/bad_examples.md` — 5 annotated bad-example excerpts with diagnosis-and-fix.