Who is Literature-review map for?
Students and researchers with a pile of downloaded papers and no structure to write from yet.
GuidesPlaybook
Literature-review map
A one-page map: 4–6 clusters, two anchor papers each, three disagreements with the design choice behind them, and the gaps your next search has to close.
IntermediateSteps5
Time~155 min
Tools5 tools
Before you start
- Ten to twenty papers you have collected, or two you already trust
- A reference manager you actually open, with PDFs attached
Tools in this guide
0%0/5 steps
Review builder · 2–3 hours
Sort papers by the words in their titles and you end up with five paragraphs that each start with “several studies have examined”.
This runbook groups papers by what they claim and measure, so the disagreements are visible before you draft a single section.
Review map
Review map
One page: 4–6 clusters named by claim, two anchor papers per cluster with a methods note and a limitation note, three disagreement lines with paper pairs, and the questions your next search has to answer.
Cluster 3: automated feedback in first-year writing. Anchors: Lee 2023 (RCT, n=210, graded drafts) and Ortiz 2024 (three-course case study). Tension: Lee finds more revision cycles, Ortiz finds the same suggested edits copied verbatim. Gap: nobody follows the same students into the next term.
Write four lines: population, intervention or concept, context, and years. Add one neighbouring topic you will exclude even though it is tempting. Run the boundary through Elicit and watch what it drags in that you never asked for.
A four-line boundary and one written exclusion, at the top of the map.
My review boundary: [population, concept, context, years]. First list papers that fit all four lines. Then list papers that look relevant but fail one line, and say which line. Do not widen the years or the population.
Pitfall
Widening the boundary because one good paper sits just outside it. Cite it as an exception instead, or the review keeps growing until the deadline decides for you.Verification
You can reject a genuinely good paper and point at the boundary line it fails.Seed Connected Papers with two DOIs you have read end to end, not two you found this morning. Work down the neighbours: each one gets a seed relation and a reason to open it, or it goes. Check the metadata in Semantic Scholar before anything enters your library. Free accounts get five graphs a month, so choose the seeds deliberately.
A candidate list where every entry names its seed and a reason to open it.
Pitfall
Reading the graph as a quality ranking. Proximity means two papers share citations, which is equally true of the weakest work in the field.Verification
No candidate survives on “the graph put it near the middle” alone.Open the papers you kept and write three lines for each: what it claims, what it actually measured, and what it cannot answer. Group by the claim line. Two papers that share a keyword but measure different outcomes belong in different clusters.
4–6 clusters with two anchor papers each, plus one methods note and one limitation note per cluster.
Pitfall
Writing the measure line from the abstract. Abstracts report the finding the authors liked, rarely the outcome they preregistered.Verification
Every cluster has a methods note and a limitation note written from the full text.Ask Consensus the yes-or-no version of your question and collect the paper pairs pointing in opposite directions. Open both sides. Write which design choice explains the split: the sample, the outcome measure, the follow-up length, or who paid for the study.
Three disagreement lines, each with a paper pair and the design difference behind it.
Where do findings on [question] disagree? Return the paper pairs and the claimed split for each pair. Do not average them into one conclusion; I will open both papers myself.
Pitfall
Writing “findings are mixed”. It hides that one study measured grades at eight weeks and the other measured confidence at one.Verification
For each tension you can name the design choice that produces it.Write headings that answer three questions: what is established, what is contested, and what your next search has to resolve. Paste the cluster notes into Claude to compress them into candidate headings, then delete every heading you cannot attach papers to.
A review outline where every heading names its papers and the gap the section closes.
Here are my cluster notes with paper names: [notes]. Propose section headings. Every heading must point to papers I named and say whether the section covers established or contested evidence. Do not write the prose and do not add papers.
Pitfall
Keeping a heading that sounds right but has no papers under it. It survives the outline and collapses the day you try to write the section.Verification
Each heading lists its papers and the gap that justifies the section.You should now have: Review map
- Every cluster contains papers you opened, with a methods note taken from the full text.
- Each disagreement line names the sample, measure, or follow-up difference that creates it.
- Every heading in the outline points at named papers and at a gap.
PrivacyLicensed course packs and unpublished manuscripts stay out of consumer tools unless your institution allows it. Bibliographic metadata is safe to paste; your supervisor's comments and your own unpublished results are not. When you are unsure, share the DOI and keep the PDF local.
Questions this playbook answers
Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.
What do you finish Literature-review map with?
One page: 4–6 clusters named by claim, two anchor papers per cluster with a methods note and a limitation note, three disagreement lines with paper pairs, and the questions your next search has to answer.
How long does Literature-review map take?
2–3 hours
What goes wrong in Literature-review map?
Each step names the mistake you are about to make:
- Widening the boundary because one good paper sits just outside it. Cite it as an exception instead, or the review keeps growing until the deadline decides for you.
- Reading the graph as a quality ranking. Proximity means two papers share citations, which is equally true of the weakest work in the field.
- Writing the measure line from the abstract. Abstracts report the finding the authors liked, rarely the outcome they preregistered.
- Writing “findings are mixed”. It hides that one study measured grades at eight weeks and the other measured confidence at one.
- Keeping a heading that sounds right but has no papers under it. It survives the outline and collapses the day you try to write the section.