Where does what you read actually go?
That question catches most people out, and the honest answers are recognisable:
- The infinite pile. New papers arrive faster than anyone can read them. “Caught up” is not a real state — strategy beats stamina.
- Read-and-forget. You read it, highlighted it, and six months later remember nothing. Reading without capture is entertainment.
- The notes graveyard. Highlights and PDFs scattered across five apps. Notes you can’t find again are notes you never took.
- No stop criterion. “I’ll start writing when I’ve read enough.” Without a definition of enough, the reading never ends.
The recension pipeline
Four steps, in order. Most literature panic comes from starting at step 3.
1. Scope the object before searching
Before you type anything into a database: ask your supervisors and your subject librarian who the must-read authors are. Start from handbooks, review articles and meta-analyses rather than primary papers — they hand you the map.
Build two artefacts as you go: a personal lexicon (the terms your field uses, including the ones you keep mixing up) and a first concept map (boxes and arrows, on paper, ten minutes).
2. Set limits — and write them down
Explicit inclusion and exclusion criteria, decided early and recorded: years (post-2010, say), geography, population, methodology, language.
Write down your keywords and your exclusions as you go. Two reasons: your methods section will need exactly this, and without a record you will re-run the same fruitless search in three months.
3. Search smart, then snowball
Databases beat Google Scholar for precision — they have controlled vocabularies, thesauri, and proper boolean operators. Learn the syntax for the one database that matters in your field; it takes an afternoon and saves weeks.
Then snowball: found a great source? Mine its reference list, and look at what has cited it since. The French call this la boule de neige, and it finds things keyword searches miss.
4. Filter in three waves, and capture as you go
Titles → abstracts → full text. Most papers should die at the title stage; that’s the system working, not you being hasty.
Use a reference manager (Zotero or Mendeley) from day one, not from the day it gets bad. Take reading notes with full references and page numbers attached. Impressions are impossible to reconstruct later — “that paper with the good figure about mice” is not a citation.
Cataloguing versus arguing
This is the difference between a literature review that reads as competent and one that reads as necessary.
| A review that catalogues | A review that argues |
|---|---|
| “Smith (2020) found X. Jones (2021) found Y. Lee (2022)…” | “Two camps disagree about X; the evidence splits by method…” |
| Organised by paper, in the order you read them | Organised by idea, tension, or gap |
| Shows that you did the reading | Shows why your project needs to exist |
| Reads like an annotated bibliography | Every cited paper is doing a job in your argument |
When to stop
Two defensible criteria — pick whichever fits your project, and say which one you used:
- Your argument holds. You can state the gap your work fills, and every claim in that statement is supported.
- Saturation. New papers stop adding new categories. You keep finding the same four positions restated.
Neither of these is “I’ve read everything.” Nobody has read everything.
Do this now — 20 minutes
- Write your inclusion and exclusion criteria for your current review, in five bullet points. Years, population, methods, languages, anything else. Save it in the document you’ll eventually write the methods section in.
- Open your reference manager — install Zotero if you don’t have one — and file the last five papers you read, with a two-line note each: what it claims and what job it does in my argument.
- Pick your stop criterion and write it down as a sentence. You will want to have it in writing at 11pm in month eleven.
Tools worth knowing
- Zotero — free, open source, syncs, and the browser connector saves a paper with one click. The safe default.
- Connected Papers — builds a visual graph of papers related to a seed paper. Excellent for finding what you’ve missed. Five free graphs a month.
- SciSpace — AI-assisted reading and literature exploration. Useful for orientation; verify everything it tells you against the actual paper.
- Litmaps and Research Rabbit — similar citation-graph tools, both with free tiers.
Where to go next
If the notes are under control but the files aren’t, Reproducible research is the hands-on one: an hour at a laptop and your analysis can be re-run — by a reviewer, by a collaborator, or by you in eighteen months.