How to Write a Literature Review

A literature review that summarises one study per paragraph is a reading list with paragraph breaks.

A literature review establishes what is known about a question, how confidently, and what remains unresolved. It is not a catalogue of what you read. The distinction sounds obvious and is the single most common reason literature reviews score poorly: the writer works through sources in the order they were found, and the reader finishes with no sense of the field’s shape.

Know which kind you are writing

Two forms dominate at university level and they have different purposes. A standalone review is an assignment in its own right, assessed on how well it maps a field. An embedded review forms a chapter of a dissertation and must additionally justify the study that follows it — every section should build toward the gap your research fills.

The embedded form is more constrained, because material that does not bear on your research question does not belong however interesting it is. Deciding which you are writing determines what to cut.

Search deliberately rather than opportunistically

Random searching produces a bibliography of whatever surfaced first, and it shows. Start with a small set of core terms and their synonyms, search two or three databases appropriate to your discipline, and record what you searched and when.

Then work backwards from the two or three most relevant recent papers: their reference lists identify the foundational work far faster than keyword searching, and their introductions tell you which debates the field considers live.

Read for position, not just content

As you read, record what each source argues rather than what it covers, along with its method, its sample and its limitations. The question to answer for each is: what does this establish, and how securely?

• What is the central claim, in one sentence?
• What evidence supports it, and of what kind?
• What does the author acknowledge they cannot show?
• Which other sources does it agree or conflict with?

That last question is what makes synthesis possible later. Notes taken source by source without it produce a review that reads source by source.

Organise by theme, method or chronology

Thematic organisation is the default and usually the strongest: group sources by the question or concept they address, and treat each theme as a section that reaches a conclusion. Methodological organisation suits fields where the design determines what can be claimed. Chronological organisation suits genuine developmental narratives, where later work responds to earlier — but it is frequently chosen by default and produces a history rather than an analysis.

Whatever you choose, the section headings should name ideas rather than authors.

Synthesise, and name the relationships

Synthesis means bringing several sources into a single paragraph to make a point none makes alone. The analytical work is naming how they relate: convergence from different methods, extension into a new population, direct contradiction, or apparent disagreement that dissolves once designs are compared.

Contradiction is the most valuable and the most avoided. Explaining why two credible studies disagree — different measures, different settings, different definitions of the outcome — is exactly what distinguishes a review that analyses from one that reports.

Evaluate rather than accept

A review should convey how much weight each finding deserves. That means noticing sample sizes, design limitations, funding interests, and whether a conclusion exceeds what the data supports. Not every source deserves the same standing, and a review that treats a large randomised trial and a small convenience-sample survey as equivalent has failed to appraise.

Do this proportionately. Extended methodological critique of every source produces a critique rather than a review; the appraisal should be enough to justify the weight you give each finding.

Identify the gap precisely

Every review should arrive somewhere. In an embedded review that destination is the gap your study addresses, and vague formulations undermine it. “Little research exists” is rarely true and easy to dismiss; “existing studies have examined this in acute settings, where staffing ratios and continuity differ substantially from community provision” identifies exactly where the boundary of knowledge sits.

The gap should follow visibly from the preceding sections rather than appearing in the final paragraph.

Keep your own voice in control

The argument is yours; sources are evidence. A paragraph opening with a citation has handed the floor to someone else. Open with your claim, then bring in the sources that support, complicate or qualify it.

Paraphrase should dominate heavily over quotation. Reviews built from strings of quotations demonstrate reading rather than understanding, and they read as assembled.

Manage the volume

Literature reviews grow, and the practical machinery matters. Use reference software once you pass roughly twenty sources, record citations at the moment of reading, and keep a running synthesis document organised by theme alongside your per-source notes. That second document becomes the skeleton of the review.

Know when to stop searching. When new sources keep citing the same foundational works and stop changing your understanding, you have covered the core, and further reading is better done in response to gaps you hit while drafting.

Revise for structure before style

On a revision pass, read only your topic sentences in order. They should form a coherent account of what the field knows. If they read as a list of subjects, the review is describing rather than analysing, and no amount of sentence editing will correct it.