Literature Review Method: Search Log, Synthesis Matrix, Draft

Students who struggle with literature reviews almost never struggle because they cannot write. They struggle because they start writing with a pile of PDFs and no system for comparing them. The fix is a process that turns reading into a table before it turns into paragraphs. This article lays out that process in the order you would actually do it, with a filled-in matrix on an example topic so you can see what each step produces.

The example topic is the use of short quizzes at the start of class (sometimes called retrieval practice) in high school science courses. The sources are illustrative placeholders, marked Author A through Author H, so treat the content as a demonstration of structure, not of findings. For what a review is and why it is organized around themes rather than authors, How to Write a Literature Review That Maps the Conversation covers the concept; this page is about the work.

Step 1: Write inclusion rules before you open a database

A review without rules grows until it collapses. Before searching, write four lines that say what is in and what is out.

Topic:        start-of-class quizzes / retrieval practice
Population:   high school (ages 14–18), science subjects
Date window:  2014 onward; earlier only if foundational
Source types: peer-reviewed articles, dissertations; no blog posts, no news

These rules do two jobs. They stop you from reading a fascinating study on medical students that will never fit, and they give you a sentence for the introduction of the review: “This review covers peer-reviewed studies of retrieval practice in secondary science classrooms published since 2014.” Readers and markers like to see that the boundaries were chosen rather than stumbled into.

If your topic is still too wide to write rules for, narrow it first; How to Choose a Research Paper Topic You Can Actually Finish has a funnel that works for reviews as well as papers.

Step 2: Keep a search log

Every search you run goes in a log with the database, the exact terms, the number of results and how many you kept. It takes ten seconds per search and saves hours when your supervisor asks how you found your sources, or when you need to rerun a search three months later.

The last row shows a habit worth building: once you find one central study, look at who has cited it since. Forward citation searching usually finds the most recent work faster than keyword searching does.

Step 3: Read every source for the same five things

Reading for a review is not reading for pleasure or even for full understanding. You are extracting the same five items from each source so that the sources become comparable.

1. Question or purpose. What did they set out to find?
2. Sample and setting. Who, how many, where, what subject?
3. Method. Experiment, survey, observation, interviews, meta-analysis?
4. Main finding. One sentence, in your own words.
5. Limit. Where does the study stop? What can it not say?

Item 5 is the one students skip, and it is the one that produces the gap at the end of the review. A study of 60 students in one school cannot speak for a district; a study that measured recall after one week cannot speak about retention at exam time. Write the limit down while it is fresh.

Read in this order: abstract, conclusion, then method, then the results if you still need them. Most of item 4 sits in the conclusion, and reading the method before the results lets you judge whether to trust them.

Step 4: Build the synthesis matrix

This is the center of the method. A synthesis matrix has one row per source and one column per theme. The themes are not decided in advance; they emerge after you have read five or six sources and noticed what they keep talking about. Start with the five items above as columns, then add theme columns as they appear.

Here is the matrix for the example topic after eight sources. Findings are illustrative.

Look at what the matrix already tells you before a word of prose exists. Read down the “Effect on unit test” column and the answer is nearly unanimous. Read down “Effect on retention” and half the cells say “not measured,” which is a gap. Read across Author D and Author F together and a tension appears: the schedule that helps test scores most may cost something in anxiety. Read the “Limit” column and you notice that most studies are single-school or single-teacher, which is a second gap. None of that would be visible from a stack of separate summaries.

Step 5: Turn columns into paragraphs

Each theme column becomes one section or paragraph. The sources in the column are the evidence; the pattern down the column is your point. Write the point first, then the evidence, then the limit.

A paragraph template that works for most columns:

[Claim about what the sources in this column show, in your words.]
[Two to four sources as evidence, grouped by what they agree on.]
[The disagreement or difference, with the likely reason for it.]
[What this column cannot tell us — the limit that points to the gap.]

Applied to the “Effect on retention” column, it produces something like this:

Evidence that start-of-class quizzes improve longer-term retention is thinner than evidence for immediate test gains. Only three of the eight studies reviewed measured performance four or more weeks after the unit. Where it was measured, the effect was positive but smaller than the unit-test effect (Author B, 2016; Author E, 2019), and one study found that daily and weekly schedules produced similar retention despite different unit-test results (Author D, 2018). Because the studies that did measure retention used different delays and different test formats, the size of the long-term benefit in secondary science remains unclear.

The paragraph names its claim in the first sentence, cites by grouping rather than one at a time, explains a difference, and ends on a limit. Compare that with “Author B found that… Author D found that… Author E found that…” and the difference is the difference between a review and a reading list.

Signal phrases that keep the synthesis visible without sounding mechanical:

• Across the studies that measured X, …
• Two of the larger studies agree that …; a smaller one diverges, likely because …
• This finding holds in every sample except …
• None of the studies located for this review measured …
• Author G’s design isolates a variable the earlier studies left mixed: …

Citation format for the paragraphs should follow whatever style your course uses; Citations in Research Papers: A Practical Guide to Crediting Sources covers the mechanics.

Step 6: Order the sections

Once each column is a paragraph, arrange them. Three orderings cover nearly every review.

Thematic is the default. Chronological suits a topic where a key study reset the field. Methodological suits a topic where the survey studies and the experimental studies disagree and you want to explain why. Whichever you choose, put the section closest to your own research question last, so the gap it exposes leads straight into your study.

Step 7: Write the ends after the middle

The introduction and conclusion are written last because they describe what the middle contains.

Introduction (one or two paragraphs): the topic and why it is worth reviewing; the inclusion rules from Step 1 in prose; a one-sentence map of the sections.

Conclusion (one or two paragraphs): what the sources agree on; what they disagree on and why; what none of them answer; and, if the review sits inside a larger project, how that project answers it. The conclusion of the example review would say, roughly, that immediate gains are well established, long-term retention is under-measured, the anxiety cost of daily quizzing is reported but unquantified, and almost all studies are single-site. Any one of those is a research question for a dissertation. If that is where you are heading, How to Plan and Write a Dissertation Without Losing Your Way picks up from here.

A two-week schedule

This assumes a stand-alone review of 2,000 to 3,000 words on 12 to 20 sources, alongside other coursework.

Day 11 is not a joke. A gap before revision is when you notice that two paragraphs make the same point or that the gap you named in the conclusion never appeared in the body.

Mistakes and the fix for each

One source per paragraph. The review reads as a series of summaries. Fix: return to the matrix and write from columns, not rows. If a paragraph cites only one study, either merge it into a theme or cut it.

Every finding presented as equally reliable. A 45-student study without a control group is quoted next to a meta-analysis as if they carried the same weight. Fix: use the “Method” and “Limit” columns when you write; say “a small uncontrolled study suggests” and “a meta-analysis of 22 studies confirms.”

No gap, or a gap that appears only in the conclusion. Fix: each body paragraph ends with its own limit. The conclusion collects them. A reader should be able to predict the gap before reaching it.

The matrix was built but the prose ignores it. Usually because the writer drafted from memory. Fix: keep the matrix open while writing and check that every cell you rely on is cited.

Themes chosen before reading. The columns were set from the assignment brief and the sources were forced into them. Fix: let columns emerge after five or six sources; rename or merge them freely as the picture changes.

Quoting instead of paraphrasing. Long quotations hand your analysis to the source. Fix: paraphrase the finding in the matrix in your own words the first time, and write from the matrix, so quotation becomes the exception.

Losing track of where a claim came from. Fix: the search log and the matrix are your audit trail. A claim that cannot be traced back to a row is deleted.

The one-page version

If you remember nothing else from this article: rules, log, five items per source, matrix, columns become paragraphs, ends last, rest a day, revise for synthesis. The matrix is what makes the writing easy. Students who skip it write the review three times; students who build it write it once.

Frequently asked questions

How many sources does a student literature review need?

It depends on the assignment, not on a magic number. A stand-alone 2,000-word review usually rests on 12 to 20 sources; a dissertation chapter may use 40 or more. A better test than counting is saturation: when new searches keep returning studies you already have, and the last three articles added nothing new to your matrix, you have read enough for this stage.

What is a synthesis matrix and why use one?

A synthesis matrix is a table with one row per source and one column per theme or question. Filling it forces you to read every source for the same things, and once it is full, each column is a ready-made paragraph: the sources in that column are the evidence, and the pattern across them is your point. It is the fastest way to stop writing summaries and start writing a review.

How do I handle sources that contradict each other?

Put them in the same paragraph and explain the disagreement rather than hiding it. Usually the conflict comes from a difference in sample, setting, measure or date, and naming that difference is real analysis. A sentence like “the two studies diverge, most likely because one measured self-reported use and the other logged actual use” is worth more to a reader than either finding alone.

Should the literature review include very old sources?

Include an older source when it founded the concept you are using or when later work still argues with it. Exclude older sources that have simply been superseded by better data. Set a default date window in your inclusion rules, say the last ten years, and record any exception with a one-line reason so a reader can see the choice was deliberate.