How to Build a Literature Review Matrix: Columns, a Template, and How to Write From It

A literature review matrix is a table in which every paper you have read is a row, and every thing you want to compare across papers is a column: what the paper asked, how it went about it, on what data, what it found, what the authors admit it cannot show, and which of your themes it speaks to. It exists so that you write the review by theme, reading down a column, instead of paper by paper, reading across a row. Everything below follows from that one purpose.

The columns, in one table

There is no official set of columns. Library guides such as Adler University's suggest lists like "citation, keywords, location, purpose, research design, main findings, notes", and tell you to add or remove columns to fit the project. This is the set that has worked for us, in the order the information appears in a paper:

Column	What goes in it	Where in the paper
Citation	Author, year, short title. The key your reference manager uses	—
Research question	What the paper asks, in one sentence	Abstract, end of the introduction
Method / design	Experiment, survey, model, case study, corpus, review; the main technique	Methods
Data / sample	What was measured or who was studied, and how much: N, years, place	Methods, Table 1
Key findings	The two or three results the paper rests on, with the numbers	Results, abstract
Stated limitations	What the authors themselves say the study cannot show	Discussion, limitations section
Theme A, Theme B, …	One column per theme of your review. A tick, or one line on what the paper says about it	Anywhere
Relevance	How the paper bears on your own question, in your words	—
Page refs / notes	The page of every entry above, and anything you doubt	—

Two rules make the table usable months later. Every cell that states something a paper says carries its page number, so you can go back and check. And Stated limitations holds only what the authors wrote; your own objections go in Relevance or Notes, where they are clearly yours.

How to build it, step by step

Decide the scope before the first row. Write the question your review answers and three to five candidate themes. They will change; writing them down first is what turns reading into filling in a table rather than wandering. (The UNC Writing Center coach who wrote up her own matrix found that setting it up before reading gave her a checklist and made note-taking less daunting.)
Make the row before you read closely. A first pass of five to ten minutes, title to conclusion, is enough to fill Citation, Research question and Method, and to decide whether the paper deserves a full row at all.
Fill the paper's columns from the paper's own text. Research question, Method, Data, Findings and Stated limitations are facts about the paper, so paraphrase them from the sections where they live, and write the page next to each. Paraphrase rather than quote: it is shorter, and it forces you to understand the sentence before you move on.
Add themes as they appear. When a third paper says something about a topic you did not plan a column for, add the column and go back to the earlier rows. A column that stays empty after twenty papers can be deleted; it was not a theme.
Keep the row honest. A field the paper does not state stays empty. A finding you half remember is checked against the page before it goes in. A limitation you thought of goes in Notes, not in Stated limitations.
Sort and filter. Once there are more than a dozen rows, the matrix earns its keep: sort by method to see what has been done experimentally, filter by a theme to see who disagrees with whom, sort by year to see where the argument moved.
Write down the columns. Each theme column becomes a paragraph or a section: what the papers in it agree on, where they differ, what none of them did. The row never becomes a paragraph; that is an annotated bibliography, not a review.

Rows for papers, or rows for themes?

Both exist, and the names are used loosely. The table above, one row per source, is what most library guides call a literature review matrix or synthesis matrix; it is the natural shape when you are still reading, because a new paper is a new row. Some writing-center handouts flip it: one row per theme or claim, one column per source, so that a row reads straight into a paragraph of the review. The second shape is what the first one turns into when you sort by a theme column. Start with papers as rows while reading, and pivot to themes as rows when you write, in a second sheet if that helps.

From the matrix to the written review

The matrix is not the review, and the mistake every guide warns about is the same one: turning each row into a paragraph. A review organized by source ("Smith found… Jones found…") is a list. A review organized by theme ("On the question of X, three experimental studies find…, while the two surveys…") is an argument, and the matrix is what makes it possible to write it without rereading everything.

A way of working that holds up: pick a theme column, sort or filter by it, and read down. Note the agreements, the disagreements and the methodological reasons for them, and the gap, which is often the empty cells. Write that paragraph with the page numbers from the Notes column beside you. Then the next column. The sections of the review come out in the order of the columns, and the reference list comes out of the Citation column.

Common mistakes

Copying the abstract into the findings cell. The abstract is the authors' pitch. The findings cell should hold the result and its number, from the results section.
No page numbers. The cell that cannot be checked is the one you will misquote.
Mixing the authors' limitations with your criticisms. One is a fact about the paper; the other is your argument. They go in different columns.
Too many columns too early. Seven or eight is plenty. A column for every idea in every paper is a second copy of the papers.
Treating the matrix as finished. It is a working document; the themes you end up writing about are rarely the ones you started with.

A template to copy

The columns above, as a spreadsheet and as a Markdown table, with a line in each cell saying what goes there. Delete the guidance as you fill it in.

literature-review-matrix.csv opens in Excel, Numbers and Google Sheets.
literature-review-matrix.md is the same table in Markdown, for Obsidian, a Zotero note or a plain text file.

Where a reader built for papers helps

Filling the paper's own columns is the slow, mechanical part, and it is the part that can be drafted from the text. Rectoly's paper cards write each paper's research question, method, data, findings and stated limitations from the paper's paragraphs, with a page on every sentence, and line the cards up as a table with one row per paper. Each sentence is checked against the paragraphs it cites and dropped if the text does not support it; a field the paper does not state stays empty. The table exports as review-table.csv and review-table.md, so it drops into the template above with the theme and relevance columns left for you, which is the right division of labour: the facts about the paper from the paper, the argument from you. The same model runs on your own computer or tablet, and nothing is uploaded. Your own notes across the papers can be grouped by topic the same way, which is a quick way to find the theme columns you had not noticed.

Frequently asked questions

What is the difference between a literature review matrix and a synthesis matrix?

Mostly the name. Both are tables that compare sources across the same set of categories so that you can write by theme. "Synthesis matrix" is the writing-center term and sometimes means the flipped layout, with themes as rows and sources as columns; "literature review matrix" usually means one row per source. Either can be sorted into the other.

How is it different from an annotated bibliography?

An annotated bibliography is a list of sources, each with a paragraph about it: one row at a time, read across. A matrix is read down its columns, across all sources at once, which is what lets you say how the literature as a whole treats a question. The matrix can generate an annotated bibliography; the reverse does not work.

How many papers should be in it?

As many as the review covers, which for a thesis chapter is usually somewhere between twenty and a hundred. The matrix starts paying off at about a dozen rows, when you can no longer hold the differences in your head, and it is the only way to keep a hundred straight.

Which columns are essential?

Citation, research question, method, data, findings and stated limitations describe the paper; at least one theme column and a relevance column connect it to your question; and a page column makes everything checkable. Everything else is optional.

Should every cell have a page number?

Every cell that states something the paper says, yes. It costs a few seconds per cell and saves the hour of rereading when a supervisor or reviewer asks where a claim came from.

Can AI fill in the matrix for me?

It can draft the columns that describe the paper, and it should not be trusted to without a check: the known failure is a plausible sentence, or a number, that is not in the text. Use a tool that cites the page for every sentence and checks each one against the passage, as Rectoly's paper cards do, and then open the page for anything that matters. The theme and relevance columns are your argument and are better written by you.

Spreadsheet, Word table or Markdown?

A spreadsheet, as soon as you want to sort or filter, which is the point of the matrix. Word tables do not sort. Markdown is fine for a dozen rows kept next to your notes, and it converts to a spreadsheet when the table grows.

When should I start the matrix?

Before the second paper. Every row added later is a paper you have to reread.