How to build a literature review matrix
Forty papers you have read and cannot compare is not a literature review. It is a reading list with opinions.
25 August 2026
A literature matrix is a table. One row per paper, one column per thing you want to compare. That is the whole idea, and it sounds too simple to matter until you have tried to write a review chapter without one.
The reason it works is that a chapter is linear and your reading is not. You read paper seventeen four weeks after paper three, and by then you no longer remember exactly how paper three measured its outcome. So you write "several studies have examined this", which is the sentence that tells an examiner you did not compare them. A matrix forces the comparison at the point of reading, when you still have the paper open.
Which columns
The instinct is to record everything. Resist it, because a matrix with twenty columns does not get filled in. Start with these six, which are enough to write a review from.
| Column | What goes in it | Why |
|---|---|---|
| Citation | Authors, year, and a DOI | So the row can be traced. The DOI, not just the title, or you will spend an evening reassembling your bibliography. |
| Question | What the paper set out to find out, in your words | Two papers on the same subject often answer different questions, and this is where you notice. |
| Method and sample | Design, size, population, setting | The column that does the most work. Most disagreements in a literature turn out to be methodological. |
| Key finding | One sentence, with the direction and size of the effect where there is one | "Found a positive relationship" is not a finding. "Reduced it by about a third over two years" is. |
| Limitations | What the authors admit, and what you noticed that they did not | This column is where your gaps come from. Keep the two kinds distinguishable. |
| Relevance to me | One line on why this paper is in your review at all | If you cannot write it, the paper does not belong in the table. |
Add columns that are specific to your question, and only those. If your review is about an intervention, add how long the follow-up was. If it is about a material, add the conditions it was tested under. The point of a bespoke column is that it lets you sort by something nobody else has sorted by.
Filling it in without lying to yourself
Three habits separate a matrix you can defend from one that quietly accumulates errors.
Record the page, and quote where it matters
For any cell you might end up citing directly, paste the sentence you took it from and note the page. It takes seconds while the paper is open and it is the difference between checking a claim in ten seconds and rereading a paper you skimmed in March.
It also protects you from the failure where a summary drifts a little further from the source each time you rewrite it, until your chapter attributes something to a paper that the paper never said.
Keep their words and your words apart
Use a mark, any mark, to separate what the paper claims from what you think about it. Your judgements are the beginning of your argument and you want them, but a year later you must be able to tell which is which. This is also where accidental plagiarism comes from, when a paraphrase you meant to rewrite goes into the chapter as it stands.
Leave cells empty when they are empty
If a paper does not report its sample size, the cell says so. Do not infer it, and do not leave it blank in a way that looks like you have not filled it in yet. Missing information is a finding. A column with six empty cells is often the clearest gap in the whole table, because it means nobody in your field is reporting something they should be.
Reading the finished table
This is the part people skip. They build the matrix, feel organised, and then write the chapter paper by paper anyway, which produces the "Smith found. Jones found. Patel found." structure that every supervisor complains about.
The matrix is for reading down the columns, not across the rows. Sort by one column at a time and look for four things.
- Clusters. Sort by method. If eleven of your fifteen studies used the same design, then what your field knows is inseparable from that design's weaknesses, and that is a paragraph nobody else has written.
- Contradictions. Sort by finding. Where two papers disagree, look along their rows for what else differs. Usually it is the sample or the setting, and the explanation is your contribution.
- Holes. Sort by population, or setting, or year. An empty region in a sorted column is a gap you can point at, which is much stronger than asserting one.
- Age. Sort by year. If the important findings are all a decade old, either the question was settled or the field moved on. Both are worth knowing before you build on them.
Every one of those four is a sentence you can only write because you can see the papers side by side.
A note on tools
A spreadsheet is completely adequate, and for a review of twenty or thirty papers it is probably the right answer. Freeze the top row, widen the cells, and get on with it. Reference managers like Zotero handle the citation column well and the thinking columns badly, so most people end up with both.
The real cost is not the table. It is reading forty papers closely enough to fill six columns honestly, which is where the weeks go, and which is the part that is genuinely yours.
Where Gapnostic fits
Gapnostic builds the matrix for you and then shows its working. Each cell carries the sentence from the paper it was drawn from, so you can see the original wording without leaving the table, and it says plainly whether it had the full paper or only the abstract to work from.
The comparing and the judging stay with you, because that is the part that becomes your argument. What it removes is the four weeks of copying things into cells.