You can know the procedure for a literature review perfectly and still not be able to write one. The step-by-step version — search, screen, chart, synthesise — tells you what to do with sources. It does not tell you what the resulting sentences look like. This page shows the sentences. Each passage below is a real piece of review prose, followed by an annotation naming the move it performs and the fault it avoids. The procedural side is covered separately in the step-by-step literature review guide; start there if you have not yet done the searching.
The passages are shapes, not claims. They contain no real study and no finding attributed to a real author. Author names are placeholders in the form Author A, Author B; [cite] marks a reference you supply and [figure] a number you must find. Nothing here should be copied into a thesis as content — only as form.
What makes review prose different from summary prose
Almost every weak literature review has the same shape: one paragraph per source, each beginning with an author’s name, each ending where the study ended. It reads like a filing cabinet because it is one. Strong review prose inverts the relationship. The idea leads the sentence and the sources arrive as support, usually several at a time, and the paragraph ends somewhere the individual studies did not reach.
Compare these two openings of the same paragraph.
| Summary prose (weak) | Review prose (strong) |
|---|---|
| Author A (2021) studied peer feedback in first-year writing courses and found improvements in revision quality. Author B (2022) also studied peer feedback and found similar results. Author C (2023) studied it in a different context. | Peer feedback improves revision quality in first-year writing, a result that has now replicated across three course contexts [cite]. The agreement, however, is confined to revision behaviour: none of these studies followed writing quality beyond the revised draft. |
The right-hand version says the same thing in half the words, names what the agreement covers, and then does the one thing the left-hand version never does — it states the limit of the agreement. That limit is where your thesis will live.
Passage 1: establishing the territory
Interest in the organisation of community-based services has grown alongside the shift of care out of institutional settings, and the research literature now spans three fairly separate traditions: health services research concerned with utilisation, organisational studies concerned with coordination, and a smaller policy literature concerned with funding mechanisms
[cite]. These traditions rarely cite one another. The review that follows is organised by the question each tradition answers rather than by discipline, because the boundaries between them turn out to matter less for the present study than the measurement choices they share.
The move: territory, plus a declared organising principle. The first sentence establishes that the field exists and is large. The second is the useful one, because “these traditions rarely cite one another” is an observation about the literature, not a summary of it — and only a reader who has read widely can make it. The third sentence tells your examiner how to read the chapter, which is a courtesy that also protects you: a reader who knows your organising principle cannot accuse you of a disorganised chapter.
Passage 2: synthesis across several studies
Where studies have measured coordination directly, they have converged on two findings. First, coordination scores predict service continuity more strongly than either staffing levels or funding intensity
[cite], a pattern that holds across both urban and rural samples[cite]. Second, the effect attenuates sharply once client complexity is controlled[cite], suggesting that part of what coordination measures is the case mix rather than the organisation. Taken together, these results describe a measure that is predictive and partly confounded, which is a more specific problem than the literature’s usual framing of coordination as “difficult to measure”.
The move: synthesis, then a sharpened claim. Three sources, two findings, one conclusion that belongs to you and not to any of them. Note the final clause: it names the field’s own vague framing and replaces it with a precise one. That sentence is the difference between a chapter that reports the literature and a chapter that has read it.
Passage 3: handling disagreement without hedging
The evidence on caseload size is genuinely divided. Three studies report an inverse association with service continuity
[cite]; two report no association once setting is accounted for[cite]. The division does not appear to be methodological: both groups use administrative data and comparable outcome definitions. It does track the unit of analysis. Studies finding an association measure caseload at the level of the individual worker, while those finding none measure it at the level of the team. If caseload operates through individual workload rather than team capacity, the two sets of results are compatible, and the apparent conflict is an artefact of aggregation.
The move: adjudicating a disagreement. This is the highest-value paragraph type in any review chapter, and most theses avoid it. The structure is fixed: state the split, rule out the obvious explanation, identify the real difference, then propose the reconciliation. Notice the modal verbs — “does not appear”, “if… then” — which claim exactly as much as the evidence supports. Getting that calibration right is a matter of verb choice, which is why reporting verbs carry so much of your stance.
Passage 4: the methodological critique
A further limitation runs through the quantitative literature. Of the
[figure]studies reviewed here that model continuity as an outcome, all but two use cross-sectional designs[cite], and the two longitudinal exceptions follow clients for a single fiscal year[cite]. Because the organisational changes of interest unfold over multiple years, the existing designs are structurally unable to distinguish a coordination effect from a selection effect — organisations that coordinate well may simply retain clients who were easier to serve. The literature acknowledges this in discussion sections but has not addressed it in design.
The move: a critique that creates room for your study. The critique is specific (design, not quality), counted rather than asserted, and it ends by pointing out that the field knows and has not acted. That last sentence is what makes a supervisor write “good” in the margin, because it establishes that your contribution is not merely another study.
Passage 5: positioning the present study
This review supports three conclusions. Coordination is a meaningful predictor of continuity but is partly confounded with case mix; the caseload literature’s apparent conflict is plausibly an artefact of aggregation; and the design of existing studies cannot separate coordination from selection. The present study addresses the third of these directly, by following a single cohort across a three-year organisational change and modelling continuity with client complexity measured at intake rather than inferred. It does not resolve the first two, which would require measurement work beyond the scope of a master’s thesis.
The move: closing the chapter. Three numbered conclusions, one of which your study takes on, and an explicit statement of the two it does not. That last sentence looks like a weakness and functions as a strength: it shows you know the size of your own contribution, and it removes two obvious defence questions. The boundaries you name here should match the ones in your introduction chapter word for word.
The matrix that produces the prose
Review paragraphs like passage 2 are not written from a pile of PDFs. They are written from a table. The minimum useful synthesis matrix has one row per source and these columns:
| Column | What goes in it | Which passage it feeds |
|---|---|---|
| Source | Author, year, full citation | All |
| Setting and sample | Country, sector, N, population | Territory, synthesis |
| Design | Cross-sectional, longitudinal, qualitative, mixed | Methodological critique |
| Construct and how measured | The exact instrument or operational definition | Disagreement, critique |
| Unit of analysis | Individual, team, organisation, region | Disagreement |
| Key finding | One sentence, in your own words | Synthesis |
| Limitation the authors state | Their words, not yours | Critique |
| Your note | What this source does for your argument | Positioning |
The two columns people leave out are the two that do the work. “Unit of analysis” is what let passage 3 reconcile a contradiction. “Construct and how measured” is what let passage 2 distinguish prediction from confounding. Without them the matrix is a bibliography with extra steps.
Sort the matrix by a column, and the paragraph writes itself: everything sharing a value in that column becomes one synthesised sentence. That is the whole trick. Keeping the underlying references tidy while you build it is what a reference manager is for, and AI literature search tools can help populate the first three columns — never the last three, which require you to have read the paper.
Five faults supervisors send back
- One paragraph per source. The diagnostic is mechanical: if more than a third of your paragraphs begin with an author’s name, rewrite them idea-first.
- No disagreement anywhere. A field with no conflicting findings is a field you have not read widely enough in. Find the split and adjudicate it.
- Critique by adjective. “Limited”, “insufficient” and “lacking” are not critiques. Say what design, what measure, what population — and count.
- The chapter does not end anywhere. A review that stops rather than concludes leaves the reader to construct your gap for you, and they will construct a different one.
- It does not match the introduction. The gap in your introduction and the conclusions of your review must be the same gap. Read them side by side before you submit, together with the problem statement.
Length, structure and when it is finished
In a Canadian master’s thesis the review chapter is commonly 20 to 35 pages and in a doctoral thesis considerably more; a major research paper compresses it to 10 to 15. Field norms vary enough that the honest answer is to look at three recently completed theses from your own department in your university’s repository and count. If your programme uses a systematic method rather than a narrative review, the reporting requirements are different and stricter — see the systematic review guide.
The chapter is finished when you can state your gap in one sentence that does not appear in any source you cited. Not before.
Tesify drafts review sections from the sources you have already read, so the matrix turns into prose you can argue with rather than a blank page. Start your thesis with Tesify — 100% written by you.
Frequently asked questions
How long should a literature review chapter be?
Commonly 20 to 35 pages in a Canadian master’s thesis, more in a doctoral thesis, and 10 to 15 in a major research paper. Field norms differ enough that counting pages in three recent theses from your own department is a better guide than any general figure.
How many sources do I need?
There is no number. The workable test is coverage: you should be able to name the main positions in your area and say who holds each. If a supervisor can name a position you did not cover, you are not finished.
Should I organise by theme or chronologically?
By theme or by question, almost always. Chronological organisation only earns its place when the development over time is itself part of your argument.
Can I include a study I disagree with?
You should. A review with no disagreement in it reads as either incurious or incomplete, and adjudicating a split is the strongest paragraph type available to you.
What is a synthesis matrix?
A table with one row per source and columns for setting, design, construct and measurement, unit of analysis, finding, stated limitation and your own note. Sorting it by a column gives you a synthesised paragraph.
How do I avoid writing one paragraph per study?
Start every paragraph with an idea rather than a name, and require each paragraph to cite at least two sources. If a paragraph can only cite one, it is a summary, not a synthesis.
Do I cite the same source in several places?
Yes, and that is a sign the chapter is working. A source that appears once is being reported; a source that appears in the design critique and again in the synthesis is being used.
Where does the theoretical framework go?
Usually at the end of the review chapter or as its own section within it, after the review has established why that framework rather than another. Some programmes place it in the introduction — check your template.
Can AI write my literature review?
It can help you search and chart, and it cannot read the papers for you. The three matrix columns that matter — measurement, unit of analysis, stated limitation — require you to have opened the source. Check your university’s rules before using any tool.
When is the chapter finished?
When you can state your gap in a single sentence that appears in none of your sources, and your introduction states the same gap in the same terms.