A research question becomes searchable only after you translate it into a documented set of concepts, synonyms, and source-specific syntax. The reliable way to do this is to start from your eligibility criteria, split the question into concept blocks, populate each block with free-text terms and subject headings, test the string in a database, and record every version so a coauthor or reviewer can rerun it. Cochrane's guidance on searching for studies frames the whole task this way: plan searches from eligibility criteria, combine free-text terms with subject headings, use multiple sources, document the process, and rerun searches before publication.
This article walks through that sequence with checkpoints, a worked example, and completion criteria you can apply to a manuscript, a coauthored review, or a lab handoff.
What does a search-ready research question look like?
A question is search-ready when you can name the population, the intervention or exposure, the comparison, and the outcome you intend to study — and when each of those elements has been written down as an eligibility criterion. Cochrane's method begins precisely there: the search is planned from eligibility criteria rather than from a vague topic statement.
Before touching a database, write one sentence per criterion. For a review of a clinical intervention, that might be a population sentence, an intervention sentence, a comparator sentence, and an outcome sentence. Each sentence becomes a concept block later. If a criterion is still fuzzy, the search terms will be fuzzy too, and no amount of database syntax will fix that.
Checkpoint: a coauthor who has never seen your project should be able to read your criteria and predict roughly which concepts will appear in the search. If they cannot, revise the criteria first.
How do you split a question into concept blocks?
Take each eligibility criterion and turn it into one concept block. A concept block is a container for every way the literature might express that idea: the canonical term, spelling variants, abbreviations, older terminology, and — where the database supports it — a controlled subject heading.
Cochrane's guidance is to combine free-text terms with subject headings, so treat these as two layers inside each block rather than alternatives. Free-text terms catch new or unindexed phrasing; subject headings catch records indexed under a standardized vocabulary.
A practical block-building routine:
- Write the canonical term for the concept.
- List synonyms, abbreviations, and spelling variants.
- Identify the subject heading a database uses for that concept, if one exists.
- Note terms you deliberately exclude and why.
Checkpoint: every eligibility criterion maps to at least one block, and every block has more than a single term.
What belongs in a concept block? A reusable table
| Block element | What to record | Why it matters downstream |
|---|---|---|
| Canonical term | The standard name for the concept | Anchors the block and the manuscript description |
| Synonyms and variants | Alternate phrasing, abbreviations, spelling differences | Captures records that use different wording |
| Subject heading | The database's controlled term, where available | Adds indexed records the free text may miss |
| Exclusions | Terms deliberately left out, with a reason | Prevents later confusion during revision |
| Source tested | Which database or interface you ran it in | Makes the string reproducible |
Keep this table with the project files. It is the artifact that turns a private search into an auditable method trail.
How do you adapt one question to different databases?
Syntax is database-specific even when the concepts are identical. PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, and a Search Details view that reveals how terms were translated and combined. That Search Details panel is the fastest way to confirm that your phrase was parsed the way you intended rather than split into separate words.
Google Scholar behaves differently: it documents searching by author and title, date sorting and limits, Cited by and Related articles, citation export, library access links, email alerts, and ways to locate alternative versions of a work. Those features support citation chasing and version-finding, and they complement a structured concept-block search in a bibliographic database.
Cochrane's recommendation to use multiple sources is the reason to adapt rather than copy. Run the same concept blocks across the sources your review requires, and record the source-specific version of each string.
Checkpoint: for each source, you can point to the exact string you ran and the date you ran it.
Worked example: from question to runnable string
Suppose your eligibility criteria describe a population, an intervention, a comparison, and an outcome. Here is how the translation proceeds without inventing results:
- Population block. Canonical term plus synonyms, abbreviations, and the database subject heading where one exists.
- Intervention block. Same structure; note any older terminology the field has used.
- Comparator block. Include the generic comparison terms your criteria allow.
- Outcome block. Include the outcome terms named in your criteria.
Combine the blocks with the Boolean operators your database supports, then run the string and open the Search Details view in PubMed to confirm how each phrase and heading was translated. If a phrase was split, adjust with the phrase-searching syntax the interface documents. If a heading pulled in an unexpected scope, decide whether to keep it and record the decision.
When you need to widen the net beyond a single structured string, run the same concepts through a scholarly search interface such as WisPaper search and compare what surfaces against your database results. For readers who want a public entry point before committing to a workflow, the WisPaper Scholar Search page explains the paper-finding starting point.
Checkpoint: the string runs without syntax errors, Search Details shows the translation you expected, and you have noted any deliberate deviations.
How do you keep the search auditable across a manuscript or handoff?
Documentation is not an afterthought; Cochrane lists documenting the process alongside planning and combining terms. An auditable trail typically contains:
- The eligibility criteria the search was built from.
- The concept-block table with synonyms, headings, and exclusions.
- The exact string run in each source, with the run date.
- Notes on any syntax adjustments and why they were made.
- The screening decisions that followed, including deduplication and staged screening.
Cochrane also describes deduplicating records, screening in stages, and linking multiple reports of one study. Those steps belong in the same trail, because a reviewer rerunning your search needs to know how records were handled after retrieval.
For project-level organization, the Research Projects area gives you a place to create and keep a project together, which helps when the search trail must survive a coauthor change or a lab handoff. Papers you decide to keep can be saved and organized in My Library, so the retrieved set and the search documentation stay associated with the same project.
Checkpoint: a colleague can reconstruct your search from the files alone, without asking you questions.
When should you rerun or revise the search?
Cochrane recommends rerunning searches before publication. That makes the rerun a scheduled step, not a rescue measure. Plan for it at the point you draft the methods section, and record the rerun date alongside the original.
Revision triggers worth watching:
- A coauthor proposes a new synonym or an older term you missed.
- Peer review questions why a concept was expressed a particular way.
- A source's subject heading scope turns out to differ from what you assumed.
- The eligibility criteria themselves change during screening.
Each trigger produces a new documented version of the string rather than an untracked edit. Keeping versions side by side is what makes the method trail auditable.
Checkpoint: the manuscript's search description matches the final documented string, including the rerun date.
Where do citation checks and reference screening fit?
Once the search has produced a candidate set, two further checks protect the trail. First, verify that the references you plan to cite are real and correctly recorded. TrueCite screens BibTeX citations against real academic databases to flag hallucinated references — a useful gate before a reference list goes to a coauthor or reviewer.
Second, use the citation tools your sources provide. Google Scholar's Cited by and Related articles features, and its ability to locate alternative versions of a work, support forward and backward citation chasing that complements the concept-block search. PubMed's Similar Articles and citation lookup serve a comparable purpose inside that database.
For questions that arise while reading the retrieved papers, Library QA lets you ask questions based on papers uploaded to My Library or added from search results, which keeps clarification close to the evidence set you built.
Checkpoint: every citation in the draft traces back to a record in your documented set or to a documented citation-chasing step.
How do you know the search terms are ready?
Use explicit completion criteria rather than a feeling that the string looks right:
- Every eligibility criterion maps to a concept block.
- Every block contains free-text terms and, where available, a subject heading.
- The string has been run in each required source, with the exact version and date recorded.
- Search Details or the equivalent translation view confirms the string parsed as intended.
- Deliberate exclusions and syntax adjustments are written down with reasons.
- Deduplication and staged screening are documented in the same trail.
- A rerun is scheduled before publication, per Cochrane's guidance.
- A colleague can reproduce the search from the files alone.
When all eight hold, the search terms are ready to describe in a manuscript and defend in peer review.




