Here is the failure, and almost every doctoral researcher writes it once. You have read for two years. You open a document and start working through what you read: a paragraph on Ahmed, a paragraph on Brennan, a paragraph on Chowdhury, each accurately summarising a study. Forty pages later you have something that is thorough, defensible, laborious and not a literature review.
It is an annotated bibliography with the headings removed. The diagnostic takes five seconds: look at how your paragraphs begin. If most of them open with a researcher’s name, you have organised by source. A literature review is organised by claim.
This page is about closing that gap, and it assumes the reading is done. If the problem is that you have notes rather than sources, the earlier stage is covered in our workflow for turning three years of notes into chapters.
This chapter is assessed before your thesis is
Worth knowing early, because it changes how seriously candidates take it. The literature review is not only a thesis chapter; at many institutions it is a formal progression requirement in its own right.
UCL’s criteria for upgrading from MPhil to PhD status include “Demonstration of sufficient awareness of the context of the work and completion of such tasks as a review of relevant literature and a bibliography”, alongside the ability “to formulate a viable hypothesis or research question that could be completed within the normal time frame”. Those two sit next to each other for a reason: the panel is checking that your question comes out of the literature rather than being asserted alongside it. The wider mechanics of that gate are in our guide to the MPhil to PhD upgrade and confirmation review.
So this chapter gets read closely twice by different people with different jobs: once by a panel deciding whether you should continue, and once by examiners deciding whether you have contributed anything. Both are reading for the same thing, which is the subject of the next section.
What the chapter is actually for
Not to prove you have read widely. Your bibliography proves that, and nobody is impressed by volume.
The chapter exists to build the case that your contribution is necessary. Everything in it should be doing one of three jobs: establishing what is known, establishing how it is known, or establishing what is not known. The third is the one your thesis exists to change, and a review that never arrives there has failed regardless of how much it covers.
This is why the review and the contribution claim are the same argument told at different lengths. Your one-sentence statement of your original contribution to knowledge is the compressed version; the literature review is the expanded version that makes it credible. If they do not match, one of them is wrong, and it is usually not the review.
The organising decision that fixes most of it
There are four ways to order a literature review, and choosing wrongly causes more damage than any amount of poor prose.
By source. One paper per paragraph. Never correct in a doctorate. This is the failure described at the top of this page, and no amount of editing rescues it, because the problem is the skeleton rather than the surface.
Chronologically. Occasionally right, and much more rarely than it is used. Chronology earns its place only when the development of the idea is itself your point — when you are showing how a field changed its mind and why that matters to your question. Used by default, it produces a narrative with no argument, because “and then someone else did a study” is not a claim.
By theme or construct. The default for most doctorates, and correct most of the time. Each section handles one idea and reports what the field collectively knows about it, with the individual studies cited as evidence within your sentence rather than as the subject of it.
By debate. The strongest option where your field genuinely disagrees. Each section states a contested question, sets out the positions and the evidence for each, and explains what would settle it. This structure is naturally hospitable to a gap, because unresolved debates are gaps.
The practical difference is visible at sentence level. “Brennan (2021) found that supervisory contact predicted completion” is organised by source. “Supervisory contact predicts completion in several settings (Brennan, 2021; Chowdhury, 2023), though both studies aggregate across disciplines that differ sharply in independence norms” is organised by claim — it states what is known, cites who established it, and begins the move toward what is not known.
The six moves
A sequence that works whether you have 60 sources or 400.
- Write the gap sentence first. Before drafting a word of the chapter, write down what your thesis will establish that is not currently established. It will change; write it anyway. It is the destination, and a review written without one wanders.
- Sort your sources into claims, not piles. Go through the library and, for each source, record the one thing it establishes that matters to your question — not what it is about, what it shows. Sources that resist this are usually sources you do not need.
- Cluster the claims. Group them until you have somewhere between four and eight clusters. Fewer than four usually means the clusters are too coarse to say anything; more than eight usually means you are still organised by topic rather than by argument. These clusters are your sections.
- Order the clusters so each one earns the next. Section two should be a question raised by section one. If the sections could be shuffled without loss, you have a list rather than an argument, and the fix is at this stage rather than in the writing.
- Draft each section to a fixed shape. What the field holds, on what evidence, with what limitation, and what that leaves open. Four moves per section, in that order, every time. It feels mechanical while drafting and reads as authoritative afterwards.
- Write the closing synthesis last. Two or three pages that collect the “left open” sentences from every section and converge them on your research question. This is the passage your upgrade panel and your examiners will read most carefully, and it should make your question feel inevitable rather than merely reasonable.
Five failures examiners name
The review that never comes back. The commonest structural criticism of a thesis: a competent literature chapter that is then never used again, so the findings are discussed against nothing. Every cluster you build should reappear in the discussion chapter. If a section has no counterpart there, ask what it is doing in the thesis.
The gap that is merely an absence. “No one has studied this in Wales” is an absence, not a gap. A gap requires a reason the missing knowledge matters: something about the Welsh context that should make the established relationship behave differently. Absences are infinite and uninteresting; gaps are arguments.
Uncritical coverage. Reporting what studies found without ever assessing how well they found it. You are being examined on judgement, so a review in which no study has a weakness reads as one where none was read closely.
Straw-manning the field. Overstating what is missing in order to make your contribution look larger. Examiners know this literature better than you do, and an overstated gap is the easiest thing in a thesis to puncture. Claim precisely what you can defend.
Staleness. A review completed in year two and never revisited. Re-run your searches before submission, and check that nothing load-bearing has been withdrawn in the meantime — a step covered in our guide to auditing a bibliography against the retraction record.
How long, and how many sources?
There is no rule, and the honest answer is that both follow from the clusters. In a monograph-shaped thesis a single literature chapter commonly runs somewhere around 8,000 to 12,000 words, in line with other substantive chapters; where two genuinely distinct bodies of work are in play, two chapters are normal rather than indulgent. What decides it is your argument rather than a target, and the reasoning behind splitting or merging chapters at all is set out in our answer to how many chapters a PhD thesis has.
On source count, the useful test is not a number but coverage of the clusters: every claim you make should rest on enough evidence that an examiner in the field would not immediately name something you have missed. That is a far more demanding standard than “300 references”, and a far more achievable one than “everything”.
The mechanical layer, which is not optional
Two infrastructure decisions determine whether this chapter is difficult or merely long.
Your references must be in a manager, with DOIs, before you draft. A literature chapter is reordered more than any other part of a thesis — clusters merge, sections swap, whole passages move to the discussion — and every one of those moves is a citation operation. Doing it by hand is how bibliographies drift out of step with the text, which is the single most avoidable examiner criticism there is. Which manager to use, and what it costs at doctoral scale, is covered in our comparison of Zotero, Mendeley and EndNote.
Finding is a different job from synthesising. If the clusters keep revealing holes, that is a discovery problem rather than a writing problem, and the tools for it are compared in our roundup of AI research assistants for doctoral work. Do not try to solve it inside the drafting session; you will produce neither a search nor a chapter.
What remains after that is the actual work: holding four to eight clusters, several hundred sources and one argument in view at once across a chapter you will draft over weeks. Tesify holds that structure section by section and formats your references as you cite them, so the chapter stays coherent while it is being reordered and the bibliography never falls behind the text. The reading, the clustering and every sentence of the argument stay 100% written by you — which is the only version you can defend when a panel asks why your question follows from this literature. It is free to start.
When to write it
Draft it early for the upgrade, then rewrite it late. These are genuinely two different chapters wearing the same title. The upgrade version argues that a viable question exists; the final version argues that your completed work answers it, and it is written with the knowledge of what you actually found.
Candidates who treat the upgrade draft as finished spend the final year fighting a chapter that no longer matches the thesis. Expect to rewrite the closing synthesis completely, and expect at least one cluster to turn out to matter far more than you thought. Rewrite it before the introduction, since the introduction has to promise what this chapter sets up.
Frequently asked questions
What is the difference between a literature review and a summary of the literature?
Organisation. A summary is arranged by source and reports what each study said. A review is arranged by claim and reports what the field collectively knows, citing studies as evidence inside the sentence rather than as its subject. If your paragraphs begin with authors’ names, you have written the first.
How should I structure a PhD literature review?
By theme or construct in most cases, by debate where your field genuinely disagrees, and chronologically only when the development of the idea is itself your argument. Never by source. Aim for four to eight sections, ordered so that each raises the question the next answers.
How long should a doctoral literature review chapter be?
Commonly around 8,000 to 12,000 words in a monograph-shaped thesis, in line with other substantive chapters. Two chapters are normal where two distinct bodies of literature are in play. The clusters should decide the length rather than the reverse.
How many sources does a PhD literature review need?
Enough that each claim you make is properly supported and an examiner in your field would not immediately name something obvious you have missed. That is a coverage test, not a count, and it is both stricter and more achievable than any number.
Is the literature review required for my upgrade?
Frequently, yes. UCL’s upgrade criteria explicitly include “completion of such tasks as a review of relevant literature and a bibliography” alongside the ability to formulate a viable research question. Check your own criteria, since the requirement and its form are institutional.
What counts as a research gap?
A missing piece of knowledge plus a reason it matters. “Nobody has studied this population” is an absence; “this population differs in a way that should change the established relationship, and nobody has tested it” is a gap. Examiners can tell the two apart instantly.
Should I criticise the studies I review?
Yes, proportionately and specifically. You are examined on judgement, and a review where no study has a limitation reads as one where nothing was read closely. Criticise the design and the inference, not the researcher.
Do I have to include everything I have read?
No, and trying to is the main cause of overlong, shapeless chapters. Include what your argument needs. The rest of your reading is not wasted; it is what allows you to judge which sources matter.
Should the literature review come before or after the methodology?
Before, in the conventional monograph shape, because the method is justified by the gap the review establishes. In some by-publication and humanities theses the literature is woven through instead, which is a disciplinary convention worth checking against recent theses from your own department.
When should I update my literature review?
At minimum once between the upgrade draft and submission, and again shortly before you submit. Re-run your searches, add anything significant, and check that nothing you rely on has been retracted while you were writing.
Can I reuse my upgrade literature review in the final thesis?
As a foundation, yes; as a finished chapter, rarely. The upgrade version argues that a good question exists, while the final version argues that your completed work answers it, so the closing synthesis in particular usually needs rewriting from scratch.
What is the most common examiner criticism of a literature review?
That it is never used again. A strong review whose themes do not reappear in the discussion leaves the findings interpreted against nothing, and it is the structural fault examiners raise most often.
