How to Improve Your Dissertation's Structure and Quality
To improve your dissertation, revise from the top down. First make your research questions, methods, results and conclusions line up. Then fix the job of each chapter, the order of paragraphs, the strength of your evidence, and finally your sentences and references. This order matters because structural changes may delete pages you would otherwise have spent hours polishing. The UNC Writing Center gives the same advice for any draft: finish reorganizing before you proofread.
First, read the rules that apply to you
Quality starts with meeting the standard you will be judged against. Before you revise, gather:
- Your program's handbook, including its word limit, structure and formatting rules
- The marking criteria or examiner guidance, if your program shares them
- Your institution's rules on AI use and on proofreading help
- Two or three recent dissertations your department rates highly
Structures vary. Humanities dissertations may be organized by argument rather than by methods and results. Some programs allow a thesis by publication. If your rules differ from the model below, follow your rules.
To understand the examiner's side, read Mullins and Kiley's paper "It's a PhD, not a Nobel Prize". It reports on interviews with 30 experienced examiners about the criteria they use and how they reach their judgements.
Pass 1: Run an alignment check
The most serious structural fault is a mismatch between what you asked and what you did. Build one table that traces each research question through the whole dissertation. Here is an invented study of peer tutoring.
| Research question | Method | Result reported | Discussed | Conclusion |
|---|---|---|---|---|
| RQ1: Does weekly tutoring affect pass rates? | Compare exam records for two cohorts | Section 4.2 | Section 5.1 | Yes, stated with limits |
| RQ2: Does the effect differ by prior mathematics? | Subgroup analysis | Section 4.3 | Section 5.2 | Yes, stated with limits |
| RQ3: How do students experience tutoring? | Interviews with 12 students | Section 4.4 | Not discussed | Missing |
The gap is now obvious: RQ3 has results but no discussion or conclusion. You have three honest choices. Add the missing sections, drop the question with your supervisor's agreement, or rewrite it to match what you actually studied.
Also check for the reverse problem. Look for results that answer no question, and conclusions that do not trace back to any result. Neither should survive this pass unchanged.
Pass 2: Give each chapter one job
A chapter fails when it does another chapter's job. This table describes a common empirical structure. Use the test question to check your own draft.
| Chapter | Its job | Test question |
|---|---|---|
| Introduction | State the problem, questions and scope | Could a reader state your questions after this chapter? |
| Literature review | Show what is known and build the case for your study | Does it end by making your questions feel necessary? |
| Methodology | Explain and justify what you did | Could another researcher repeat it? |
| Results or findings | Report what you found | Have you avoided interpreting here, if your field separates the two? |
| Discussion | Interpret findings against the literature | Does it answer each question and return to the sources from your review? |
| Conclusion | Say what was achieved and what follows | Does it state your contribution and limits without adding new results? |
Three faults appear often. Background that belongs in the introduction ends up in the methods. Interpretation creeps into the results. New findings first appear in the conclusion. Move each one to its proper place.
Pass 3: Reverse outline each chapter
A reverse outline shows the structure you actually wrote, rather than the one you planned. UNC describes it this way: write a brief note on what each paragraph does, copy the notes to a clean page in their current order, then study the result.
An invented section from the same study might look like this:
- Peer tutoring has grown in universities
- Tutoring is defined in several ways
- Study A: tutoring and grades
- Peer tutoring has grown in schools too
- Study B: tutoring and confidence
- Definition used in this study
Now the problems are visible. Paragraphs 1 and 4 cover similar ground. Paragraphs 2 and 6 are split. Paragraphs 3 and 5 report studies without comparing them. A better order is: definition, then what tutoring has been shown to change, with Studies A and B compared under each outcome.
Look for four things as you study the outline: repetition, gaps, paragraphs that do two jobs, and any point not connected to your questions.
Pass 4: Test the evidence and methods
Now check whether your claims can carry their weight.
Match your claims to your design. Words such as "proves" and "causes" need the right design behind them. In the invented cohort study above, "was associated with higher pass rates" is safer than "caused higher pass rates", because students were not randomly assigned.
Report your methods in full. APA's Journal Article Reporting Standards set out what to report for quantitative, qualitative and mixed methods work. They are written for journal articles, and APA also describes them as teaching tools, so they make a useful checklist for your methods and results chapters. In health research, the EQUATOR Network lists reporting guidelines by study type.
Check your numbers. Compare each figure across the abstract, text, tables and conclusion. Sample sizes, percentages and test results must match everywhere.
State your limits plainly. Do not leave a limit for your examiner to find. Naming it yourself, with its effect on your conclusions, shows judgement.
Keep terms consistent. If "peer tutoring" becomes "peer mentoring" in Chapter 4, readers may assume a new concept.
Pass 5: Add signposts
A long document needs clear signals. A reader should never have to guess why a section exists.
- Open each chapter by saying what it does and how it connects to your questions.
- Close each chapter by saying what has been established and what comes next.
- Write informative headings. "4.3 Effects by prior mathematics" helps more than "4.3 Further analysis".
- Start paragraphs with their point. A reader skimming topic sentences should be able to follow the argument.
Example of a weak chapter opening: "This chapter presents the literature."
A stronger version: "This chapter reviews evidence on peer tutoring outcomes and shows that few studies test whether effects differ by prior mathematics, which is the focus of RQ2."
Pass 6: Finish cleanly
Save the fine work for the end.
- Write or rewrite your abstract last, using our guide on how to write an abstract.
- Check that each in-text citation has a reference, and the reverse. If you use a reference manager, our Zotero guide explains how to avoid breaking linked citations.
- Reread every source you cite for a key claim.
- Check figures, tables, captions and page numbers.
- Proofread on paper or read aloud, since a change of format can reveal errors you missed on screen.
Getting useful feedback
Tell readers what kind of help you need. "Please read this" often brings back comments on commas. Try: "Read only the chapter openings and closings. Can you say what my argument is?" Your supervisor remains the main judge of direction.
For a second check, the thesis tools in CitedEvidence include outlines, chapters with citations, thesis-specific review and defence preparation. Check your institution's AI rules first, since some restrict how such tools can be used in assessed work. Disclose use as required. Your name goes on the work, so every claim and every source must be one you can defend.
If you only have a few days
| Time left | Do this | Leave out |
|---|---|---|
| One day | Alignment check, number check, reference check | Reordering whole chapters |
| Three days | Add chapter openings and closings, reverse outline your weakest chapter | Rewriting strong sections |
| One week or more | Run all six passes, then ask for feedback | Nothing, but protect time for proofreading |
Common mistakes
- Polishing too early. Fine editing can make you reluctant to cut.
- Writing the literature review as a catalogue. It should lead to your questions.
- Ignoring results that do not fit. Report them and discuss why.
- Leaving the discussion disconnected from the review. Return to the sources you introduced earlier.
- Using your conclusion as a summary only. Say what the work adds and what it does not settle.
- Treating feedback as a list of edits. Look for the underlying problem each comment points to.
FAQ
What should I revise first?
Start with the alignment check, because it can change what you need to write. If a research question has no matching result or conclusion, polishing that chapter is wasted effort.
How do I cut words without losing quality?
Use a reverse outline to find repeated points and paragraphs that do not serve a research question. Cut those before trimming sentences. Your methods and key findings are usually the last places to cut.
Can I use AI to edit my dissertation?
It depends on your institution's rules, which may treat language editing differently from drafting. Ask your supervisor, keep a record of what you used, and disclose it as required.