Literature Search Strategy: Concepts, Databases, and Records

Translate the question into searchable concepts

A literature search strategy is a reproducible way to find relevant evidence, not simply a list of words typed into one search box. Start with the research question and identify its central concepts, population, setting and outcome where applicable. A question about remote feedback and student revision can be searched through terms for feedback, online learning and revision, but a narrow phrase may miss studies that use different labels.

List synonyms, spelling variants and related subject terms for each concept. Check a few relevant papers to learn how authors and databases describe the topic. Do not add every possible word without considering precision; a broad term can flood the results with unrelated records. Use quotation marks for an exact phrase where the database supports them, and check how it handles stemming, punctuation and field restrictions.

Combine alternatives within one concept with OR and distinct concepts with AND. Parentheses make the logic visible. A search such as (feedback OR commentary) AND (revision OR rewriting) targets the overlap while allowing variation in wording. Test a query against a known relevant paper. If it fails to retrieve that paper, investigate whether the terms, indexing or database coverage explain the miss.

Choose sources for the discipline

Select databases because they cover the subject, document types and period needed, not because they are familiar. A clinical question may require biomedical databases; an education question may need education and social science indexes. Search engines can complement specialist databases but differ in ranking and reproducibility. Explain the role of each source and check whether its coverage overlaps heavily with another.

Use controlled vocabulary where a database offers it, alongside free-text terms. Subject headings can group articles using different words, but newly indexed or emerging work may not yet have them. Search titles and abstracts for emerging terms and inspect how a database maps the query. Syntax differs across platforms; adapt and record each version instead of copying a string and assuming it behaves identically.

Consider additional routes when the question warrants them. Citation tracking can find studies that build on a key paper; reference lists can reveal older foundations. Trial registries, policy repositories or dissertations may reduce publication bias for some questions. Grey literature needs its own evaluation of provenance and method. Do not add a source merely to inflate the number of records.

Record and refine the process

Pilot the search, inspect a sample of relevant and irrelevant hits, and adjust terms for a reason. If most results concern a different population, add an appropriate boundary or field restriction. If a known relevant study is missing, broaden a term or add its synonym. Keep a dated record of changes, including the search string, database, interface, filters and result count. A final strategy should reflect what was run, not what was intended.

Apply date, language and publication-type limits only when justified by the question and resources. A five-year window may be appropriate for a fast-changing technology but exclude foundational theory or a long-term outcome. An English-only limit can omit relevant work and introduce bias; acknowledge it. A filter labeled “peer reviewed” may be imperfect, so check records rather than assuming it guarantees quality.

Export records and remove duplicates using a documented method. Similar titles, preprints and later journal versions require care; an automated match can merge different works or retain duplicate versions. Record how many records were obtained and removed. For a formal review, document screening criteria and decisions so another researcher can follow the path from search results to included evidence.

Screen for relevance before judging quality

Define inclusion and exclusion criteria from the question, not from whether a study agrees with a preferred conclusion. Screen titles and abstracts, then read full texts for uncertain or promising records. Keep reasons for full-text exclusions where the review method requires them. Separate relevance from credibility: a methodologically weak study can still be relevant to the topic, though it may receive little weight in the synthesis.

Use a second reviewer or a reasoned check for consequential screening decisions when resources permit. Disagreement can reveal ambiguous criteria. If only one reviewer is available, record borderline decisions and revisit them after the initial pass. Do not claim a search is exhaustive when database access, language or time constraints prevented a full sweep.

Update the search before submission if the field changes quickly or the project took a long time. Record the final date. A study published after the search date is not evidence that the recorded search was defective, but the reader should know the boundary. For a scoping or systematic review, follow the reporting standard relevant to that design rather than treating a casual bibliography as a systematic search.

Connect search to synthesis

Preserve the records of excluded full texts and the reasons for exclusion in a formal review. A reason such as “wrong outcome” should correspond to a criterion defined before the results were read. If a record cannot be obtained, note the access problem separately from a decision that it is irrelevant. This allows readers to distinguish limitations in retrieval from the study selection logic.

The value of a search is the evidence it enables the writer to evaluate. Once sources are selected, chart their questions, populations, methods, results and limitations. Compare studies that use different outcome definitions instead of counting them as votes. If the evidence is sparse, state whether the gap reflects the field, the search boundary or both.

Present the search in enough detail that another researcher could repeat its logic, while keeping the main paper readable. A table or appendix can hold exact strings and dates. The final account should make clear where evidence was sought, what was found, what was excluded and how those choices constrain the conclusion.