Writing a Thesis or Dissertation: Structure, Workflow, and Momentum

How a thesis differs from a journal paper

A journal article makes one argument in roughly 6,000 to 10,000 words. A thesis defends a whole program of original work, typically 40,000 to 60,000 words for a master's and 60,000 to 100,000 for a PhD, and it has to prove something larger than a single result: that you can conceive, execute, situate, and criticize research on your own. That difference reshapes the writing. A thesis must be self-contained, so a reader should never need your earlier papers to follow it. It foregrounds method and justification rather than compressing them into a paragraph. It reports the dead ends and negative results where they are instructive, and it carries a full literature review instead of a two-sentence related-work note.

Many science students now write a "stapler" or three-paper thesis: two or three publishable studies bracketed by a substantial introduction and a synthesizing conclusion. It looks like a shortcut, but the connective tissue (a framing that makes three studies one contribution) is the hardest part and the part examiners probe. Whichever model your department uses, treat the thesis as the capstone of a single connected process rather than a document you bolt on at the end. The earlier you set up that pipeline, described in our guide to the modern research workflow, the less you rebuild under deadline pressure.

The standard structure and the IMRaD pattern

Most theses assemble the same canonical parts, even when disciplines reorder or rename them. Knowing what each part is for keeps you from padding one and starving another.

1. 1Abstract: 150 to 350 words covering problem, method, key findings, and contribution. Write it last, revise it most.
2. 2Introduction: the problem, why it matters, your research questions or hypotheses, and a roadmap of the chapters.
3. 3Literature review: a critical map of the field that locates your gap, not an annotated list. See our guide to the systematic literature review.
4. 4Methodology: what you did and why, in enough detail that a peer could reproduce it.
5. 5Results: what you found, reported neutrally, with figures and tables doing the heavy lifting.
6. 6Discussion: what the results mean, how they answer your questions, limitations, and how they sit against the literature.
7. 7Conclusion: the contribution restated, implications, and future work.
8. 8References and appendices: every cited source, plus instruments, code, extended tables, and proofs.

Underneath sits IMRaD: Introduction, Methods, Results, and Discussion, the backbone of empirical writing. Picture an hourglass. You open wide (the broad problem), narrow to your specific question and method, report results at the narrow waist, then widen again in the discussion back out to the field. In many STEM theses each study chapter is its own small IMRaD unit, while the thesis as a whole is a larger one. Humanities and some social-science theses instead run thematic or chronological chapters, but the hourglass logic still holds: general to specific and back to general.

Planning chapters and word budgets

Set a word budget before you draft, working backward from the total your regulations allow. A budget converts an intimidating 80,000-word mountain into a set of chapters you can actually schedule, and it exposes imbalance early, when a bloated literature review is still cheap to fix.

Give each chapter a one-line thesis statement of its own before you write a word: the single claim that chapter exists to establish. If you cannot write that sentence, the chapter is not ready to draft, and no amount of prose will hide the missing spine. Keep these statements in one place so you can read the whole argument as a paragraph and check that the chapters actually build on each other.

Write early, write often, and lower the stakes

The single most reliable predictor of finishing is whether you write while you research rather than after. Waiting until the work is "done" guarantees a brutal final year and a mountain of forgotten context. Instead, keep a running research journal and turn notes into prose continuously, so a study chapter is 60 percent drafted by the time the experiment ends.

Momentum comes from separating drafting from editing. Write a fast, ugly zero draft that no one else will see, whose only job is to get the argument out of your head and onto the page where you can fix it. Judging sentences while you generate them is what produces the blank-page freeze most students mistake for lack of ideas.

Almost all good writing begins with terrible first efforts. You need to start somewhere.Anne Lamott, Bird by Bird

Beating writer's block and perfectionism

• Set a tiny non-negotiable minimum, for example 250 words or one focused Pomodoro of 25 minutes, and protect it daily. Consistency beats occasional marathons.
• Never end a session at a natural stopping point. Stop mid-paragraph, mid-sentence even, so tomorrow you restart with momentum instead of a blank page.
• Draft out of order. Methods and results are the easiest to write because they are factual; leave the introduction and abstract until you know what you are introducing.
• Mark gaps with a literal placeholder token such as TKTK or [CITE] and keep moving. Chasing a reference mid-flow is how a writing hour becomes a browsing hour.
• Separate the writer from the editor by time of day. Generate in the morning, revise in the afternoon, and never let the editor into the drafting session.

Turning literature notes and evidence into chapter drafts

The gap between a full reference library and a written chapter is where most theses stall. The bridge is a disciplined pipeline from highlight to note to claim. When you read, capture the exact passage and where it sits (source, page, and a quotation), then write your own note about what it means for your argument. Those notes, grouped by theme, become a chapter outline almost mechanically, because each becomes a paragraph and each paragraph carries its own supporting evidence.

Write one claim per paragraph and attach the evidence to the claim, not to a vague mental memory of "a paper I read." This is the habit that survives revision: when a reviewer challenges a sentence eight months from now, you can walk from the claim back to the evidence, to the source, to the exact highlighted passage, without re-reading forty PDFs. Our guide on building an evidence base you can defend covers how to anchor each claim so it holds up under scrutiny.

Cross references, figures, and tables

Never hard-code a number. If you type "as shown in Figure 3" and later insert a figure earlier, every subsequent number is wrong and you will not catch them all. Use automatic numbering and symbolic references: in LaTeX that is \label and \ref, and Word offers cross-reference fields that do the same. Then a renumber is automatic and a broken reference is a build error rather than a silent mistake an examiner finds first. Keep figures in a consistent format, give every one a self-contained caption (a reader skimming only figures should still follow the story), and maintain a list of figures and tables that regenerates itself. If your thesis is math-heavy, our practical guide to LaTeX for researchers explains equation numbering and referencing in depth.

Keeping citations resolvable as chapters move

During a long write-up you will cut a section from Chapter 2 and paste it into Chapter 5, and its citations must travel with it intact. This is why stable citation keys matter more than they seem. Cite by a durable key (a BibTeX cite key backed by a DOI), never by a formatted string you typed by hand, so the reference resolves wherever the text lands and reformats itself if you switch styles. Run a check that flags any citation whose key has no matching entry in your bibliography before every supervisor submission, so a broken reference never reaches a reader. The mechanics of keys, styles, and BibTeX are covered in our explainer on citation styles and reference management.

Version control for prose

"thesis_final_v3_ACTUALLY_final_supervisor_edits.docx" is a system that fails at the worst moment. Treat your prose the way software teams treat code. Keep the manuscript in plain-text-friendly formats (LaTeX or Markdown) under version control with Git, commit at the end of every writing session with a message describing what changed, and you get a complete, dated history you can diff and roll back. Lost a paragraph you deleted three weeks ago? It is one command away. Working on two chapters in parallel? Use branches. If you must use Word, at minimum keep tracked changes on and store dated copies in a synced, backed-up folder, but know that Git gives you far finer-grained recovery.

The supervisor relationship and feedback loops

Your supervisor is a limited, high-value resource, and the students who finish well use them deliberately. Do not send the whole thesis at once and wait a month; send one chapter at a time with a short note saying what kind of feedback you want (structure, argument, or line edits) at that stage. Sending a rough chapter for a big-picture read while asking for grammar fixes wastes both of you. Agree an explicit cadence, for example a chapter draft every four to six weeks, so neither of you drifts.

When feedback arrives, resist the urge to defend. Log every comment, decide on a response, and keep a short revision letter that records what you changed and why. That record is not busywork: it becomes the backbone of your response to examiners later, and it stops the same comment coming back twice. Treat contradictory feedback from two supervisors as a decision to surface and resolve, not a trap; ask them directly which way to go.

A realistic chapter-by-chapter timeline

Timelines vary by field and by whether your data is already collected, but here is a workable shape for the final 18 months of a full-time PhD, assuming the bulk of the research is done. The lesson is that writing is not a single final phase; it overlaps the research and the analysis.

Preparing for the viva or defense

The viva (called a defense in North America) is an examination of your thesis and of you as an independent researcher, usually months after you submit, so re-read your own thesis cover to cover before it. Examiners are not trying to fail you; they are testing whether you understand your own contribution, its limits, and the field around it.

1. 1Re-read the whole thesis and mark, in a separate list, the weaknesses you already know about. Naming a limitation before the examiner does is a sign of maturity, not defeat.
2. 2Prepare crisp answers to the predictable questions: What is your original contribution? Why this method and not an alternative? What are the limitations? What would you do differently?
3. 3Know the papers you cite most and be ready to defend how you used them. Reread the two or three works your examiners have written that touch your topic.
4. 4Do a mock viva with peers or your supervisor. Practising saying the answers out loud is very different from thinking you know them.
5. 5Expect corrections. Minor corrections are the normal outcome, not a bad result. Keep your revision letters and provenance trail handy so you can act on them fast.

Keep the whole chain intact with Research Woven

Everything above rests on one idea: a thesis is not a document, it is a maintained chain from source to highlight to note to evidence to claim to chapter to citation, and the writing goes wrong at the joints. Research Woven is a research operations platform built to hold that chain together while you spend months rearranging chapters. Because provenance is computed rather than remembered, you can stand on any claim in your manuscript and trace it straight back to the evidence, the source, the page, and the highlighted passage behind it, then forward to every chapter that cites it.

That is what stops the slow decay that ruins the final year: a citation whose key still resolves after the paragraph moves, a cross reference that fails loudly instead of silently, an outline generated from the notes you actually wrote. The human writes the argument; the platform operates everything around it. If you want your dissertation to survive its own revision history, start by keeping every claim connected to its evidence and sources from the first note, and let the connected research workflow carry the rest. It pairs naturally with sound data management and archiving so your results stay reproducible long after you graduate.