# Scientific Writing Checklist: Step-by-Step Guide By the end of this guide you will be able to take a finished body of research and turn it into a manuscript, proposal, or lab report using a repeatable checklist. You will know what order to write in, what to check at each stage, how to keep a [lab notebook](/knowledge/molecular-biology/lab-notebook) that feeds directly into your writing, and which mistakes cost the most time during peer review. Here is what you need on hand before you start: - Your lab notebook or electronic lab notebook records, raw data files, and figures - The author guidelines for your target journal, or the funding agency's call for proposals - A reference manager such as Zotero, Mendeley, or EndNote - A plain text editor for the templates in this guide (Notepad, TextEdit in plain text mode, VS Code, or any code editor) - A method for version control, even a simple dated file naming system The checklist below works for research articles, grant proposals, theses, and class lab reports. The wording of individual checks changes slightly by format, but the sequence does not. ## Why a Checklist Beats Motivation Scientific writing fails most often at the planning and revision stages, not at the keyboard. A checklist converts writing from a talent question into a process question. Evidence from structured writing support backs this up. In a national trial of [grant writing](/knowledge/molecular-biology/grant-writing) coaching groups, 367 faculty and postdoctoral fellows were organized into 76 coaching groups, and 69 percent of participants submitted at least one national grant application within the study window, with 32 percent receiving at least one award [1]. The intervention that produced those numbers was not extra talent. It was structure: scheduled group sessions, deadlines, and feedback loops. Structure matters at the mentorship level too. In that same trial, faculty whose scientific mentors engaged in a structured way with the coaching process had significantly higher odds of receiving any award and of receiving an NIH mentored career development (K) award compared with faculty whose mentors were unstructured [1]. Translate that to your own desk: a checklist is a way of giving yourself the structured engagement that improves outcomes, even when no coach is in the room. Classroom research points the same direction. A course-based research experience built around the Paleobiology Database teaches students to write a scientific paper alongside observation, reading primary literature, and [statistical analysis](/blog/guides/statistical-analysis), and it uses an alternative grading scheme built on spiral feedback, where work is revisited and improved across multiple passes rather than graded once [2]. The checklist in this guide is built for that same spiral model. You will not pass through it once. You will pass through it several times, each pass narrower than the last. ## The Scientific Writing Checklist, Step by Step This walkthrough assumes you have data in hand and a target outlet chosen. If you are writing a grant proposal instead of a paper, the same steps apply with "specific aims" in place of "abstract" and "preliminary data" in place of "results." ### Step 1: Write Your One-Sentence Message Before any drafting, complete this sentence in one line: "This paper shows that [finding], which matters because [consequence]." Example: ``` This paper shows that longitudinal connections integrate distant levels of the hippocampus, which matters because the widely used transverse slice preparation assumes connectivity is largely transverse. ``` That example is real. Witter and Amaral wrote a review arguing that longitudinal connections in the hippocampal formation are as important as transverse connections, revising a long-standing model in which the structure was treated as a set of largely independent transverse "chips" [3]. The paper succeeded because the message fit in a sentence, and every section served it. If you cannot write your sentence, you are not ready to draft. If your sentence needs three clauses and two hedges, your project may have more than one paper in it, and you should choose one. **Check:** A colleague can repeat your one-sentence message back to you after hearing it once. ### Step 2: Read the Target's Rules Before You Write a Word Open the author guidelines for your journal or the funding announcement. Record, in your notebook, the answers to these questions: - Word or page limit for the main text - Figure and table limits and format requirements - Reference style and any citation manager style file offered - Required sections (some journals mandate a "Limitations" statement) - Whether the abstract is structured (labeled headings such as Background, Methods, Results, Conclusions) or unstructured - Data and code availability requirements Writing 7,000 words against a 4,000-word limit is the most expensive drafting error in science, because cutting a finished draft takes longer than writing to length. **Check:** You have a written list of five to eight format rules from the actual guidelines, not from memory. ### Step 3: Build an Evidence Inventory From Your Lab Notebook A lab notebook that was kept well makes this step fast. A notebook kept badly makes it slow, which is the practical argument for good notebook habits. The key habits are dated entries, enough methodological detail that someone else could repeat the work, and immediate recording of observations and failures. Course-based research programs reinforce this deliberately, for example by having students make and record weekly observations so that the data behind a paper accumulate in usable form rather than in memory [2]. If your notebook lives on paper and you want a searchable digital copy, a simple approach is a daily markdown file with a fixed header. Save the following block as a template file called `notebook_template.md`: ```markdown # 2026-11-11 ## Today's question ## What I did - [ ] - [ ] ## Conditions and reagents | Item | Lot/ID | Amount | Notes | |------|--------|--------|-------| ## Results (raw numbers, sketches described in words) ## What went wrong ## Tomorrow ``` Then create each day's entry with a one-line command: ```bash cp notebook_template.md notebook_2026-11-11.md ``` Expected result: a new file appears with today's date in the name and the full template inside, ready to fill. To convert a folder of these entries into a single PDF notebook, many people use pandoc, a free command-line document converter: ```bash pandoc notebook_*.md -o lab_notebook_2026.pdf ``` Expected output: a single PDF file containing every dated entry in order. If pandoc reports `pandoc: notebook_*.md: openBinaryFile: does not exist`, your shell did not expand the wildcard, which happens on some Windows setups. List the files explicitly instead: ```bash pandoc notebook_2026-11-01.md notebook_2026-11-02.md -o lab_notebook_2026.pdf ``` **Check:** Every figure and key number in your planned paper can be traced to a dated notebook entry. ### Step 4: Make a Skeleton Outline in IMRaD IMRaD stands for Introduction, Methods, Results, and Discussion, and it is the standard structure for research articles in the life sciences. Draft the skeleton as headings with one bullet under each, before writing prose. A minimal research template looks like this. Save it as `paper_template.md`: ```markdown # Title (working) ## 1. Abstract (150 to 250 words, written LAST) ## 2. Introduction - Broad problem - Specific gap - One-sentence message preview ## 3. Methods - Model system / sample - Key procedures with parameters - Analysis approach - Replication and statistics ## 4. Results - Figure 1: - Figure 2: - Figure 3: - One paragraph per figure, in figure order ## 5. Discussion - Main finding restated plainly - Consistency with prior work - Limitations - Next experiment ## 6. References ``` This template is deliberately spare. Add subheadings as your methods and results actually require them, and delete sections the target journal does not want. **Check:** Every planned figure has a caption stub, and every claim in the outline bullet list maps to a specific figure, table, or citation. ### Step 5: Write Results First, Around Your Figures ![Five-step flow showing Results written first, then Methods, Introduction, Discussion, and title and abstract last.](/article-images/illustrations/scientific-writing-checklist-writing-order.webp) Write the Results section before anything else. Make or finalize your figures first, then write one paragraph per figure in the order the figures will appear. Each paragraph states what was measured, the direction and size of the effect in plain numbers, and what the reader should see in the figure. State statistics in the format your target journal requires, for example "treatment increased output relative to control (P = 0.003, n = 6 per group)" with the exact test named in Methods. A useful self-check is to cover the figure and read your paragraph aloud. If the text does not let a listener sketch the figure, the paragraph is too vague. **Check:** No Results paragraph describes a figure that does not exist, and no figure appears that Results never discusses. ### Step 6: Write Methods So Someone Else Could Repeat the Work The Methods section is a recipe. Include the model or sample, key materials with sources, critical parameters (temperatures, durations, doses where relevant, software and versions), the analysis approach, and how you handled exclusions or missing data. Write Methods in past tense. If a procedure is standard, cite the original or canonical description rather than re-describing it, but still note any deviations. The hippocampus review discussed above shows why this discipline pays off over decades. The widely used transverse hippocampal slice preparation, a 300 to 400 micrometer slice maintained for hours in vitro, grew out of the lamellar model of hippocampal organization [3]. That preparation is a methods decision, and it rested on an anatomical claim. When anatomical tracing later showed widespread longitudinal projections, including a substantial longitudinal component in the mossy fibers, the field had to revise the model [3]. Methods written with enough precision to be checked are what allow that kind of productive correction. **Check:** A labmate from a different subgroup can read your Methods and list every item they would need to buy or build. ### Step 7: Write the Introduction as a Funnel The Introduction moves from broad importance, to what is known, to the specific gap, to your question or hypothesis, in that order, typically in four to six paragraphs for a research article. Cite the primary literature, not reviews alone, for claims you rely on. End the Introduction by stating the question or hypothesis and, in many journals, a one-sentence preview of the approach. A caution from real editorial history: do not overstate certainty about the current model you plan to challenge. The lamellar hypothesis of hippocampal organization shaped decades of experimental design before evidence accumulated against it [3]. Framing the existing model honestly, as a strong but testable framework, positions your work as a contribution rather than a demolition. **Check:** Every sentence in the Introduction either narrows the topic, supports a claim with a citation, or states your question. Delete anything else. ### Step 8: Write the Discussion in Three Moves The Discussion does three things in order: restates the main finding in plain language, places it against prior work, and states limitations and next steps. Resist adding new results here. If a result belongs in Discussion, it belongs in Results. Limitations should be specific and honest, for example "our sample covers only two time points, so we cannot infer the rate of change." Generic limitations ("more research is needed") say nothing and reviewers notice. **Check:** A reader who skips Results entirely can still learn your main finding and its boundary conditions from the Discussion alone. ### Step 9: Write the Title, Abstract, and Keywords Last Write the title and abstract after the full draft exists, because they must describe the paper you actually wrote. The title should name the variables or system and, where honest, the direction of the finding. The abstract mirrors the paper's structure: one or two sentences of context, the question, the approach in one sentence, the main result with a number where possible, and the interpretation in one sentence. Choose keywords by asking what terms a stranger would search to find this work. **Check:** Someone who reads only the title and abstract can state your finding. If they cannot, revise until they can. ### Step 10: Fix References and Formatting Import all citations into your reference manager, generate the bibliography in the journal's style, and then check every in-text citation against your bibliography, because reference managers produce duplicates and phantom entries. Confirm that every reference is cited in the text and every in-text citation appears in the list. Match the journal's figure resolution and file format requirements, and place tables according to the guidelines (many journals want tables at the end or as separate files). **Check:** Your reference manager reports zero uncited items and zero missing items after a final sync. ### Step 11: Run a Feedback Loop, Not a Single Review Hand the draft to at least two readers with different jobs. One should be a content expert who attacks the logic. One should be a non-specialist who flags jargon and unclear transitions. This mirrors the spiral feedback model used in effective research training, where the same document improves across repeated, focused passes instead of one overwhelming critique [2]. If you are an early-career scientist writing a proposal, structured mentor engagement has measurable value. In the coaching group trial, structured engagement of a scientific mentor was associated with higher odds of any award among faculty, with an adjusted odds ratio of 1.87, and with substantially higher odds of a mentored career development award, adjusted odds ratio 3.32, along with higher proposal success rates [1]. Ask your mentor for a specific kind of review at each pass: logic on pass one, clarity on pass two, mechanics on pass three. **Check:** You received written comments from at least two readers, and you addressed each comment in writing, even if the answer was a polite refusal with a reason. ### Step 12: Do a Cold-Eyed Final Pass Wait at least one day, then do a mechanical pass with fresh eyes. Useful low-tech and free tools: ```bash wc -w manuscript.md ``` Expected output is a line like `6412 manuscript.md`, giving your current word count against the journal's limit. For spelling, hunspell checks American English directly: ```bash hunspell -d en_US manuscript.md ``` Expected output: a numbered list of questionable words with suggestions, or nothing if the file is clean. Ignore false positives from technical terms, and add repeated terms to a personal dictionary so the list shrinks each pass. Finally, read the paper backward, last paragraph first. Backward reading defeats the brain's habit of skimming familiar text, which is why typos survive fifteen normal readings. **Check:** Word count is within the limit, spelling check is clean, and the abstract, figures, tables, and reference list all match the guidelines from Step 2. ## Summary Table of the Checklist | Stage | Main action | Pass-fail check | |--|--|--| | 1. Message | Write one sentence stating finding and importance | A colleague can repeat it back | | 2. Rules | Record target limits and formats | Five to eight rules written down | | 3. Evidence | Trace every claim to dated notebook entries | No orphan figures or numbers | | 4. Outline | Build IMRaD skeleton with figure stubs | Every claim maps to a figure or citation | | 5. Results | One paragraph per figure, figures first | No text about missing figures | | 6. Methods | Repeatable recipe with parameters | Outsider can list needed materials | | 7. Introduction | Funnel from broad to your question | Every sentence narrows or cites | | 8. Discussion | Finding, context, specific limitations | Skippable Results still leaves reader informed | | 9. Abstract | Title and abstract written last | Title plus abstract conveys the finding | | 10. References | Sync manager, check duplicates | Zero uncited, zero missing | | 11. Feedback | Two readers, three focused passes | Every comment answered in writing | | 12. Final pass | Word count, spellcheck, backward read | Within limits, clean check | ## A Reusable Research Template for Any Project The `paper_template.md` from Step 4 works as a general research template for papers, and with small edits it serves proposals and reports. For a grant proposal, replace the Results section with "Preliminary Data" and the Discussion with "Approach and Expected Outcomes." For a class lab report, drop the Introduction's literature funnel to two paragraphs and keep everything else. Keep your templates in one folder and copy, never overwrite, the master. A simple versioning habit prevents the classic disaster of writing over your only clean draft: ```bash mkdir mypaper_v1 cp paper_template.md mypaper_v1/draft_2026-11-11.md cd mypaper_v1 git init git add draft_2026-11-11.md git commit -m "first draft from template" ``` Expected output from the commit is a line like `[main (root-commit) 3f2a9c1] first draft from template` with a file change summary. The hash value will differ on your machine. After each writing session, `git add .` and `git commit -m "date + what changed"` gives you an exact history of every draft, which is also useful when a coauthor asks what changed since last month. ## Common Mistakes and Limitations These are the failures reviewers and editors see most often, roughly in order of cost. **Writing before the message is fixed.** Drafting without the one-sentence message produces papers that reviewers describe as "unclear what question is being asked." Two hours on Step 1 saves weeks. **Ignoring the author guidelines.** Wrong abstract length, unstructured abstract where a structured one is required, figures below required resolution. Desk rejection, meaning rejection before review, often traces to format violations. **Notebooks written after the fact.** A notebook reconstructed from memory weeks later is not a record, and it cannot support your Methods or your data provenance. Entries belong on the day they describe, including negative results, because "what went wrong" entries frequently supply the Discussion's limitations section honestly. **Results paragraphs that narrate rather than report.** "We then examined X" is narration. "X increased under treatment (P = 0.003)" is a report. Reviewers flag the first, approve the second. **Overclaiming.** A finding in one preparation does not establish a general principle. The hippocampus literature again supplies the cautionary example, since the lamellar model shaped experimental design for years before more sensitive tracing methods revealed widespread longitudinal connectivity that the model could not accommodate [3]. Describe what your data show in your system, and mark extrapolation as extrapolation. **One-shot feedback.** Asking a mentor "any comments?" invites vague comments. Asking "does the logic of Figure 2 support the claim in the last paragraph of Results" invites usable answers. The trial evidence on coaching shows that structured engagement outperforms unstructured engagement in writing outcomes [1], and the same principle scales down to a single lab. **Template worship.** A research template is a starting skeleton, not a cage. If your journal requires a different structure, follow the journal. If your field expects combined Results and Discussion, combine them. **A limit of checklists generally.** A checklist ensures completeness, not correctness. It will not catch a flawed experimental design, and it cannot substitute for honest data. It also cannot tell you how your specific manuscript, with its specific coauthors and deadline, should be scheduled. Use the checklist as the floor, not the ceiling. ## How to Know Your Document Is Actually Ready ![Decision path with five yes or no checks on abstract, figures, paragraph flow, hedges, and coauthor sign-off.](/article-images/illustrations/scientific-writing-checklist-ready-to-submit.webp) Before submission, run this final verification in order: 1. Read only the abstract. Does it state a finding? 2. Read only the figures and captions. Do they tell the story? 3. Read only the first and last sentences of each paragraph. Does the argument flow? 4. Search your draft for hedges. If "may," "might," and "possibly" cluster around your central claim, the message from Step 1 has drifted. 5. Confirm every coauthor has read the final version and agrees with the interpretation. Author disagreements discovered by reviewers are far more damaging than disagreements discovered at your desk. If all five pass, submit. Waiting for a perfect draft is the most common way good papers never leave the drawer. ## Frequently Asked Questions ### What is a scientific writing checklist? A scientific writing checklist is an ordered list of concrete checks that a manuscript, proposal, or report must pass at each stage of planning, drafting, and revision. It turns writing into a repeatable process so that format rules, evidence traceability, and feedback loops are verified rather than remembered. ### What order should I write a scientific paper in? Write Results first, built around finalized figures, then Methods, then Introduction, then Discussion, and finally the title, abstract, and keywords. This order works because each section depends on the ones before it, and the abstract must describe a paper that already exists. ### What is IMRaD? IMRaD stands for Introduction, Methods, Results, and Discussion, the standard structure for research articles in the life sciences. It mirrors how science is done, which makes it easy for reviewers to find what they need and easy for you to audit your own logic. ### How do I keep a lab notebook that makes writing easier? Date every entry, record methods in enough detail for someone else to repeat them, capture raw numbers and observations the day they occur, and include a running "what went wrong" section. A daily markdown template with fixed headings keeps entries searchable and convertible to a single PDF with a tool like pandoc. ### Where can I get a lab notebook template PDF? You can make one in minutes from a text template rather than hunting for a download. Save a dated markdown template, create one file per day, and run `pandoc notebook_*.md -o lab_notebook_2026.pdf` to compile the whole series into a single PDF. This gives you a notebook format you control and can adapt as your project changes. ### How do I get useful feedback on my manuscript? Ask two readers for different, specific jobs: a content expert to attack the logic and a non-specialist to flag jargon and structure. Request one focus per pass, such as logic, then clarity, then mechanics, and answer every comment in writing so the revision is deliberate. ### How many revisions does a manuscript need? Plan for at least three focused passes before submission: logic, clarity, and mechanics, in that order. More passes help if each has a distinct focus, but endless unfocused polishing usually signals that the one-sentence message from the first step never got fixed. ### Do checklists work for grant proposals too? Yes, with section names swapped. The same sequence of message, rules, evidence inventory, outline, results as preliminary data, methods, abstract as specific aims, references, structured mentor feedback, and final mechanical pass applies. Structured engagement with a scientific mentor during proposal writing is associated with better award outcomes in a large randomized trial, so build that feedback loop in from the start. ## Sources 1. [Grant writing coaching groups: A randomized trial to discern the influences of coaching duration and engagement of scientific mentors on participants' proposal submission and success rates.](https://pubmed.ncbi.nlm.nih.gov/43008041/) 2. [Using the Paleobiology Database to Teach Scientific Practices: an Approach Aligned to Relational Culture Theory.](https://pubmed.ncbi.nlm.nih.gov/43011488/) 3. [Establishing the Three-Dimensional Organization of the Hippocampus-A mini Scientific Revolution?](https://pubmed.ncbi.nlm.nih.gov/43004154/)