Cards for a literature review

A literature review is the piece of academic writing most often done wrong by people who know exactly what they are doing. The reading is careful, the notes are thorough, and the chapter that results is a procession: Smith found this. Jones, building on Smith, found that. Lee disagreed. Sixty paragraphs later the reader knows what everyone said and has no idea what the field thinks, where it is stuck, or why the present study needs to exist.

That chapter is an annotated bibliography in paragraph form, and it happens for the same reason the undergraduate paper stalls after the reading: the notes are organized by source, so the writing is organized by source. A review is not organized by source. It is organized by what the field claims, where it disagrees with itself, and what it has not asked — and none of those is a property of any one paper. They are properties of the set, and you cannot see them until the pieces come out of the papers they came from.

This guide is the card method applied to a body of literature. It scales: the moves are the same at sixty sources and at two hundred, and the second half of the guide is about what changes when the count gets large.

What goes on a card, at this scale

At the undergraduate scale a source yields quotations and facts. At the graduate scale it yields something more specific, and the cards have to be built to hold it.

The source card carries the citation, filled from the DOI or ISBN so that the typing is not the bottleneck, and three more things: the question the paper asked, in one line; its method, in one line; and its population or corpus, in one line. These three lines are what you will sort by later, and a source card without them is a citation and nothing else.

The finding card is the graduate equivalent of the fact card: one result, in your words, with the source, the page, and — this is the part that matters — the conditions: the sample, the setting, the measure. Effect X was found is not a finding; effect X was found in undergraduates on a self-report measure is, and the difference is where the review's argument will live.

The claim card is what the paper argues, as distinct from what it found. Papers argue beyond their findings; that is their job. A claim card marked as the source's claim, not yours, lets you hold the argument separately from the evidence for it — which is exactly the separation the review has to make.

The disagreement card is new at this scale and is the most valuable thing in the deck. When two sources say incompatible things — different findings, different interpretations of the same finding, different definitions of the term — write a card that names the disagreement: Smith and Lee disagree about whether X holds outside laboratory settings. Link it to both. The review's argument is almost always built out of these.

Your own cards — your claims about the field, your questions, your doubts — in a different color, as always.

Step 1: Bring the sources in as a batch

Sixty sources should not take sixty evenings to become source cards. Start from what you already have — a reference manager's library, a folder of PDFs, a reading list — and make the source cards as a batch: a list of DOIs and ISBNs looked up at once, or a bibliography file brought in whole. Then, and only then, fill the three lines (question, method, population) by hand, one source at a time, from the abstract. Ten minutes a source. This is the reading pass that most reviews skip, and it is the one that makes the rest possible.

Step 2: Read for findings and claims, not for summary

Read each source with cards beside it, and make finding cards and claim cards — not notes. The discipline of one thing per card (student guide 3) is harder at this scale, because a good paper has eight findings and you will want to write them as a block. Do not. Eight cards. Each one will need to be sorted independently of the others, because finding two supports one side of a disagreement and finding six the other, and a block cannot be in two piles.

Stop when a source has given you what it has. A paper that yields one card yielded one card; write it and move on. The review does not need every source to be equally present in it, and the ones that contribute nothing are still part of the record of what was read.

Step 3: Sort by question, and find the disagreements

Now the move that makes a review. Lay out the finding and claim cards and sort them not by source, and not by topic, but by the question each one answers. Does X hold outside the laboratory? Is X caused by Y or by Z? How should X be measured? Each question is a pile with a card on top naming it.

Within each pile, the sources that never met are now sitting together, and their disagreements are visible. Write a disagreement card for each one you find, and link it to the sources on either side. Sort the disagreements themselves: is it a disagreement about findings (two studies got different results), about interpretation (the same result read two ways), or about definition (they are not measuring the same thing)? These are three different kinds of card and three different kinds of paragraph, and a review that knows which is which reads as a review.

Then find the empty piles. A question the field has asked and not answered is a pile with a question card and nothing under it. A question the field has not asked is a card you write yourself, in your color, when you notice that no pile exists for it. These two kinds of empty pile are the gaps a review exists to name, and they are what justifies the study that follows.

Step 4: The argument about the field

A review has a thesis, though it is rarely called one: a claim about the state of the field. The evidence for X is strong in laboratory settings and absent outside them, and the disagreement about whether X holds in practice is a disagreement about method, not about X. That is a claim, with a verb, that someone could dispute, and every pile you have sorted is evidence for or against it.

Find it the way student guide 4 says: from the row of claims — here, the question piles and their disagreements — not before them. Then order the piles the way student guide 6 says: not chronologically, not by source, but so that each question prepares the reader for the next and the gap arrives where the reader is ready to see it. Try two orders. Keep the better.

Step 5: Write by question, cite by finding

Each pile is a section. Each section opens with the question, states what the field has found in answer, names the disagreement and its kind, and says what is unresolved. The sources appear as evidence for the findings, not as the subjects of the paragraphs. X holds in laboratory settings (Smith 2019; Lee 2021) and has not been tested outside them is a review sentence. Smith (2019) tested X in a laboratory setting and found that it held is a bibliography sentence.

The last section is the gap, and it is written from your own cards.

When the count gets large

At two hundred sources, three things change.

You will not remember what you have. The deck has to be searchable — by source, by question, by whether a finding is linked to a disagreement — and the searches you run repeatedly are worth keeping as named views: findings with no question pile, sources with no finding cards yet, disagreements with only one side linked. Each of those is a to-do list that the deck writes for you.

Sources will be cited zero times. You read them, made source cards, and they contributed nothing to any pile. This is normal and it is information; the review can say we found no studies that… only because you looked. Keep them. A view of sources cited zero times is the list of what you read and did not use, which is the cut pile at this scale, and it goes in the record (student guide 7).

The review will need to be updated. A literature review is finished when the thesis is; the field keeps publishing. A new source is a new source card and its findings sorted into the existing piles — an afternoon, not a rewrite, because the piles are the structure and the structure holds. This is the payoff of having built the review from cards rather than written it: it can be revised by adding cards, and the chapter recompiled from the same piles in the same order.

Doing this in software

Software that works from cards makes the batch step and the scale step tractable: source cards filled from a list of identifiers or a bibliography file, kept in sync with a reference manager, with the notice of a retraction arriving on the card that cites the work; finding cards linked to their sources and to the disagreements they bear on; the piles as tags or as an order; the searches as saved views that stay live as the deck grows; sources cited zero times as a question the software already knows how to ask. It also checks, when asked, whether every claim about the field has a finding under it — which is the review's version of the undergraduate's unsupported claim. Nothing about the method changes. The piles are still sorted by you, and the argument about the field is still yours to find.