How to write a literature review

The single most common failure in a literature review is a structure that goes: Smith found this. Jones found that. Okonkwo found the other. That is an annotated bibliography with the annotations run together, and it is marked down every time.

A literature review is an argument about a field. Its job is to show what is known, where the field disagrees, and — crucially — what has not been settled, because that gap is the justification for your own work. It is organised by ideas, and sources appear inside the ideas.

Reviewing versus listing

The test is what your paragraph is about. If the paragraph is about a study, you are listing. If the paragraph is about a question, a method, a disagreement or a theme, and studies appear as evidence, you are reviewing.

Listing: Adeyemi (2019) surveyed 200 teachers and found X. Following this, Kaur (2021) surveyed 350 and found Y.

Reviewing: Survey evidence consistently reports X in urban schools (Adeyemi 2019; Kaur 2021; Mensah 2023), but every one of these studies drew its sample from state capitals, and the two rural studies that exist find the opposite (Bello 2020; Tran 2022). The disagreement may therefore be about sampling rather than about teaching.

The second paragraph has a claim. That is the whole difference.

The process

Define the boundary first. Which years, which languages, which disciplines, which kinds of study. Write it down. Without a boundary a literature review has no end, and this is the main reason people stall.

Search systematically and record as you go. Databases, keywords, dates. Keep a note of what you searched and when — for a formal review you may have to report it, and for any review you will otherwise repeat yourself.

Read for the conversation, not the content. As you read, note who agrees with whom, who is responding to whom, and where the field changed direction. This is what you are actually reviewing.

Build a matrix. A simple table: source, question, method, sample, finding, limitation. Fill it in as you read. When the themes emerge, they will emerge from the columns — usually from the method or limitation column.

Group into themes. Three to six. These become your sections. If a theme has one source in it, it is not a theme.

Write from the matrix, not from the papers. Have the matrix open and the papers closed. This is the single most effective trick for avoiding paragraph-per-source drift.

Structuring it

Thematically is the default and usually the best: each section is a question or a debate.

Chronologically only if the field genuinely changed over time and that change is your point.

Methodologically when the disagreement is about how the work is done — common in health and psychology.

Theoretically when competing frameworks are the story.

Within each theme: state the claim, present the agreement, present the dissent, evaluate the quality of the evidence, and say what remains open.

Ending on the gap

The final section must state what is missing, and your research question must be visibly the answer to it. A review that ends with a summary rather than a gap has not done its structural job — it may be an excellent essay and it is not yet a literature review.

Common mistakes

One paragraph per source. Describing studies without evaluating them. Citing only work that agrees with you. Including a source because you read it, not because it is needed. Uncritical use of a source's own abstract. A gap that is merely "nobody has studied X in my country" without saying why that matters. Losing track of which idea belongs to which source — a genuine plagiarism risk, not just an untidiness.

Practical tips

• Keep the matrix in a spreadsheet from day one; retrofitting it is miserable.
• Use a reference manager and let it do the formatting.
• Read the introductions of recent reviews in your field to see how the conversation is currently framed.
• If you cannot write the sentence "the field disagrees about ___", you have not read enough yet.
• Check whether your discipline expects a systematic review with a reported protocol — the standards are much stricter, and finding out late is expensive.