FIELD GUIDE / 8 MIN READ
What a literature review is actually for.
Not a summary of everything published. An argument about where the gap is.
The most common literature review looks like this: a paragraph on one study, then a paragraph on the next, then a sentence saying more research is needed. Every source gets its turn. Nothing is compared. The reader finishes knowing what has been published and not why your project should exist.
A review has one job. It has to make your question look necessary. Everything else is in service of that.
Read for disagreement, not coverage
When you read the ninth paper on your topic, you are not looking for its findings. You are looking for where it sits relative to the other eight. Three things to hunt for:
- Findings that conflict. Two studies, opposite results. Why? Different populations, different measures, different years. That "why" is often a whole project.
- A method everyone reuses without questioning it. If eleven studies all use the same self-report scale, "does this hold with observed behavior" is a legitimate question.
- A population nobody has looked at. The weakest kind of gap, and still valid / but only if you can say why that population might behave differently, not just that it is absent.
The table that does the work
Stop writing prose until this exists. One row per source, five columns:
- Author and year
- Who was studied (with the sample size)
- Design in one clause
- The main finding in your words
- What it does not tell us
Fill in twelve rows and read down column five. Patterns show up that you cannot see reading one paper at a time. Here is what a real column five looks like after eight papers on feedback timing: all undergraduate; all self-reported effort; none followed the same students into a second assignment. That is a gap, and it wrote itself.
Then write in clusters
Structure the review by argument, not by source. Each paragraph starts with a claim about the literature and cites two or three papers as evidence for it.
Studies of feedback timing consistently report stronger revision when feedback arrives within a few days (Smith 2019; Okoro 2021; Lin 2022). All three, however, measure revision as word count change, which cannot distinguish adding paragraphs from rethinking the argument.
Two sentences, three sources, and the reader now knows both what is established and where the soft ground is. Compare that with three separate paragraphs summarizing the same three studies. Same reading, a quarter of the words, and an argument.
How many sources is enough
Enough that you stop finding new arguments. You will notice it happening: around the fifteenth paper the citations start pointing back to things you have already read. That saturation point is the honest answer, and it arrives sooner than people expect for a narrow question / which is another reason to narrow the question first.
Two failure modes to watch
The defensive review. Only citing work that agrees with you. A reader who knows the field will spot the missing paper immediately, and it costs you more credibility than the disagreement would have.
The throat-clearing review. Six pages establishing that the topic is important before any actual literature appears. Importance can be one paragraph. Get to the disagreement.
A practical test before you submit: delete every citation from your review and read what is left. If the remaining sentences still make an argument, the review is doing its job. If you are left with connective tissue / "several authors have examined", "this has been widely studied" / the sources were carrying the paragraph and you have written a reading list.