How to write a literature review that argues

Ask ten students what a literature review does and nine will say 'it summarises what has been written on the topic'. That definition is why so many literature reviews read like a queue: Smith found this, then Jones found that, then Patel found something else. Each paragraph is competent; the chapter as a whole says nothing.

A literature review is an argument about a field. Its job is to map what is known, show how the knowledge was produced, expose where it breaks down, and make your research question feel like the obvious next move. This guide takes you from defining a boundary through to the sentence patterns that make synthesis visible on the page.

The purpose: an argument, not an inventory

Every literature review is building towards one destination: the gap. Everything before it exists to make that gap credible. If a reader finishes your chapter and cannot say in one sentence why your study needs to exist, the review has failed regardless of how many sources it contains.

That reframing changes what belongs in the chapter. You are no longer obliged to mention every study you read. You are obliged to establish a small number of claims about the field — that a consensus exists here, that findings diverge there, that a method dominates and carries a particular blind spot — and to evidence each claim with the studies that best demonstrate it.

It also changes the unit of analysis. The paragraph is not about a study; it is about a claim, and studies are the evidence for it. If your paragraph opens with an author's name, you are almost certainly writing an inventory. If it opens with a statement about the field, you are writing a review.

Define the boundary before you search

Unbounded reading is the most common way to lose a month. Before you open a database, write down four decisions: the population or setting you care about, the concept or intervention, the outcome or phenomenon of interest, and the time window. Then write down what you are deliberately excluding and why.

For example: peer-reviewed empirical studies of remote-working arrangements in UK knowledge-sector organisations, published 2019–2026, examining self-reported productivity. Excluded: pre-pandemic studies, because the baseline changed; non-UK studies, because employment law differs materially; grey literature from consultancies, because methods are rarely reported.

Those exclusions are not weaknesses to hide. Written into your methodology, they read as scholarly judgement, and they protect you from the examiner's question 'why did you not include…?'. The answer is already on the page.

Search systematically, and record what you did

Use two or three subject databases rather than a general search engine — Scopus, Web of Science, PsycINFO, Business Source Complete or whichever your discipline uses — and build your search string from synonyms joined with OR, then combined with AND. 'remote work' OR telework OR 'hybrid working' AND productivity AND UK is a starting point, not a finished query.

Keep a simple search log: database, exact string, date, number of hits, number retained. It takes two minutes per search and it is the difference between a methodology section you can write in an hour and one you have to reconstruct from memory. In some disciplines a PRISMA-style flow diagram is expected; even where it is not, the log makes your process defensible.

Then work outwards from your best sources. Backward chaining means mining the reference lists of the most relevant papers; forward chaining means finding everything that has since cited them. Two or three rounds of this will surface the papers your keywords missed, which are frequently the important ones.

Build a synthesis matrix

The reason most reviews end up as inventories is that the notes were taken one paper at a time, so the writing follows the same order. A synthesis matrix breaks that habit by forcing you to read across sources instead of down them.

Make a spreadsheet with one row per study and columns for: research question, sample and setting, method, key findings, stated limitations, and — the important one — your own note on how this relates to the other studies. Fill it in as you read. After fifteen or twenty rows, patterns become visible that no amount of re-reading individual papers would reveal.

Now sort and re-sort the sheet: by method, by finding, by year, by sample size. Each sort suggests a possible theme. When sorting by method shows that eleven of your fifteen studies used cross-sectional surveys, you have not just found a pattern — you have found a limitation of the field, and quite possibly your gap.

• Study, author and year — plus where you found it, so you can retrace your steps.
• Research question, and whether it was answered.
• Sample, setting and time period.
• Method and analytical approach.
• Headline findings, in your own words, with page numbers.
• Limitations the authors admit, and limitations they do not.
• Your note: who does this agree with, contradict or extend?

Organise thematically, never author by author

Thematic organisation is the default expectation at every level from undergraduate dissertation upwards. Each section addresses one theme, debate or methodological cluster, and within it you group studies by what they show rather than who wrote them.

Two alternatives are sometimes appropriate. A chronological structure works when the field has genuinely evolved through identifiable phases and that evolution is part of your argument. A methodological structure works when the central problem is how the phenomenon has been measured. Both are still arguments — they are simply arguing about time or method rather than theme.

Whichever you choose, open each section with the claim and close it with the consequence. A section on measurement might open 'Self-report instruments dominate the field, and their limitations are well documented but rarely addressed' and close by explaining what that means for what can currently be known.

Name the gap and make it lead to your question

Gaps come in recognisable types, and naming the type makes your justification much stronger than a vague 'little research exists'. An empirical gap means a population, setting or period has not been studied. A methodological gap means the phenomenon has only been examined one way. A theoretical gap means a framework used elsewhere has not been applied here. A contradiction gap means credible studies disagree and nobody has explained why.

'Little research exists' is also the riskiest sentence in a dissertation, because an examiner who knows of three studies you missed will discount the whole chapter. Prefer precise formulations: 'No study identified in this review has examined X in a Y setting using Z', which is a claim about your search rather than about the entire world.

Then close the loop explicitly. The final paragraph of the review should restate the gap and present your research question as the direct response to it, using shared vocabulary so the reader can see the join. If your gap is about measurement and your question is about attitudes, something has come loose.

Sentence patterns that make synthesis visible

Synthesis is a way of writing as much as a way of thinking, and a handful of sentence patterns will drag your prose out of summary mode. Each of the following puts the field, not the author, in the subject position — which is exactly the shift you are trying to make.

• 'Three broadly consistent findings emerge from this literature. First…'
• 'While Okafor (2023) and Lindqvist (2024) both report X, they differ sharply on whether…'
• 'This consensus rests on a shared methodological choice: all four studies used…'
• 'The evidence is stronger for A than for B, largely because B has been measured only in…'
• 'Barnes (2026) is the exception, and the exception is instructive because…'
• 'What none of these studies addresses is…'