Finding sources for a literature review means more than opening Google Scholar and typing in your topic. It means knowing which databases cover your field, how to build a search string that actually works, and how to expand your search when keyword results start to thin out.
This guide covers the practical side: which databases to use, how to search them effectively, and how to make sure you're not missing important papers that keyword search can't find.
Start with your research question, not your search terms
Before opening any database, you need a research question specific enough to generate useful search terms. A question like "what are the effects of sleep deprivation?" is too broad. "How does chronic sleep deprivation affect cognitive performance in adolescents?" gives you four searchable concepts: sleep deprivation, cognitive performance, adolescents, and the relationship between them.
A structured framework helps you break your question into searchable concepts. The most widely used is PICO:
- P: Population (who are you studying?)
- I: Intervention or exposure (what is being done or experienced?)
- C: Comparison (what is it being compared to?)
- O: Outcome (what are you measuring?)
For non-clinical research, PEO (Population, Exposure, Outcome) or PCC (Population, Concept, Context) are more appropriate. The framework isn't mandatory, but it helps ensure you've identified all the key concepts before you start searching.
Write out your research question and identify the key concepts. Each concept becomes a search term, and each search term needs a list of synonyms and related words, because different papers use different language for the same idea.
For example, if your concept is "cognitive performance," your synonym list might include: cognitive function, executive function, attention, memory, academic performance, neurocognitive outcomes.
Build this synonym list before you start searching. It's the foundation of a comprehensive search strategy.
Best databases for finding sources for a literature review
No single database covers all published research. The right starting point depends on your field.
- Health sciences and medicine Start with PubMed (free, covers MEDLINE, 35+ million citations). Add Embase for clinical and pharmaceutical research. Cochrane Library for systematic reviews and clinical trials. PsycINFO for psychology and psychiatry. CINAHL for nursing and allied health.
- Social sciences Scopus and Web of Science for broad multidisciplinary coverage. PsycINFO for psychology. ERIC for education. Sociological Abstracts for sociology. EconLit for economics.
- Natural sciences and engineering Web of Science for physical sciences, chemistry, and engineering. IEEE Xplore for electrical engineering and computer science. Compendex for engineering broadly. SciFinder for chemistry.
- Humanities JSTOR for journals across the humanities and social sciences. MLA International Bibliography for literature and linguistics. Historical Abstracts for history. RILM for music.
- Multidisciplinary starting points (any field) Scopus and Web of Science cover the broadest range of peer-reviewed journals. Google Scholar is the broadest of all, useful for a quick orientation and for finding preprints and grey literature, but less precise than structured databases for systematic searching.
Practical rule: Search more than one database. No single database indexes every relevant journal. For most literature reviews, a subject-specific database for your field combined with one multidisciplinary database (Scopus or Web of Science) gives solid coverage. Systematic reviews typically require broader coverage, consult your institution's librarian for guidance on your specific field.
A practical search workflow
Before diving into the details, here's the full process at a glance:
.webp)
Step 1: Build your term list
Write your research question, identify the key concepts, and generate synonyms for each. Include both current terminology and older terms that foundational papers might use.
Step 2: Search subject-specific databases first
Use your field's primary database with a structured Boolean search string and controlled vocabulary terms where available. Apply date and language filters if appropriate.
Step 3: Add a multidisciplinary database
Run the same search in Scopus or Web of Science to catch papers your subject database doesn't index.
Step 4: Run Google Scholar
Broad search to catch preprints, grey literature, and anything the structured databases missed. Use the "Cited by" link on key papers for quick forward citation searching.
Step 5: Switch to citation network search
Add your most relevant papers to ResearchRabbit. Follow references backward to find foundational work. Follow citations forward to find recent developments. Use Similar to surface cross-disciplinary connections.
Step 6: Search grey literature separately
Theses, preprints, government reports. Field-specific sources based on your topic.
Step 7: Document everything
Record search strings, databases, dates, and result counts as you go.
The rest of this guide explains each step in detail.
How to build a literature review search string
Most researchers type a phrase into a search bar and work from whatever comes back. A structured search string finds more relevant papers and fewer irrelevant ones.
Boolean operators
Boolean operators tell the database how to combine your search terms.
AND narrows your search, both terms must appear: sleep deprivation AND cognitive performance
OR expands your search, either term can appear. Use it to capture synonyms: (sleep deprivation OR sleep loss OR sleep restriction)
NOT excludes a term: cognitive performance NOT Alzheimer's
Combining them: (sleep deprivation OR sleep loss OR sleep restriction) AND (cognitive performance OR executive function OR attention) AND adolescent*
Wildcards and truncation
An asterisk (*) at the end of a word root catches all variations:
adolescent*finds adolescent, adolescents, adolescencecognitiv*finds cognitive, cognition, cognitively
A question mark (?) substitutes a single character:
wom?nfinds woman and women
Phrase searching
Quotation marks search for an exact phrase:
"cognitive performance"finds that exact phrase, not just pages containing both words separately
Putting it together
A well-built search string looks like this:
("sleep deprivation" OR "sleep restriction" OR "sleep loss") AND (cognitiv* OR "executive function" OR attention OR memory) AND (adolescent* OR teenager* OR "young people")
This captures more relevant papers than a simple keyword search, with fewer irrelevant results.
Controlled vocabulary: the search technique most researchers skip
Most major databases have a controlled vocabulary, a standardized set of terms that indexers use to tag papers regardless of what language the authors used. Searching controlled vocabulary terms alongside your keywords significantly improves recall.
Which databases have controlled vocabulary:
- PubMed: MeSH (Medical Subject Headings), the most comprehensive. MEDLINE-indexed records in PubMed are assigned MeSH terms. Search "sleep deprivation"[MeSH Terms] to find every record tagged with that concept, regardless of the words used in the text.
- Embase: Emtree, similar to MeSH but with broader pharmaceutical and clinical coverage
- CINAHL: CINAHL Subject Headings, tailored to nursing and allied health
- PsycINFO: APA Thesaurus of Psychological Index Terms
- ERIC: ERIC Thesaurus for education research
Scopus and Web of Science do not have their own controlled vocabulary. Scopus records may include indexed keywords from other thesauri (MeSH, Emtree), but inconsistently, you can't browse or search a Scopus thesaurus directly. For these databases, comprehensive keyword searching is the primary approach.
Practical shortcut: Find one paper you know is highly relevant. Look at its indexed keywords or MeSH terms in PubMed. Add those terms to your search. This is often faster than building the term list from scratch, and catches papers that use different vocabulary for the same concept.
Field-specific search tips
PubMed: Use the Advanced Search builder to combine MeSH terms with free-text keywords. Apply filters for publication type (review, clinical trial), date range, and language. The "Similar articles" feature surfaces related papers based on MeSH indexing, often more useful than keyword suggestions.
Scopus: Search in Title-Abstract-Keywords (TITLE-ABS-KEY) for most searches. The Subject Area filter limits results to your discipline. The "Analyze search results" feature shows you co-authorship patterns and most-cited papers at a glance.
Web of Science: Use the Topic search field to search across titles, abstracts, keywords, and author keywords simultaneously. The "Cited reference search" lets you find papers citing a specific study, the same as forward citation searching.
Google Scholar: Useful for broad orientation and for finding preprints and grey literature. Less reliable for systematic searching because it's harder to reproduce exactly. Use it to find PDFs of papers you've already identified, and to run quick forward citation searches via the "Cited by" link. For graduate-level literature reviews, use Google Scholar to complement structured database searches, not replace them. For a detailed comparison of Google Scholar and citation network-based discovery, see beyond Google Scholar: how researchers find relevant papers.
Grey literature: the sources most searches miss
Grey literature, theses, dissertations, conference papers, government reports, preprints, often contains relevant research that never appears in database searches. Including it gives a more complete picture, particularly in fields where policy or practice outpaces publication. It's one of the most common reasons literature reviews miss important papers, for a fuller picture of what keyword search leaves behind, see how to make sure your literature review doesn't miss important papers.
Theses and dissertations: ProQuest Dissertations and Theses (global coverage), EThOS (UK), DART-Europe (European). University institutional repositories are also searchable via Google Scholar.
Preprints: arXiv (physics, mathematics, computer science), bioRxiv and medRxiv (life sciences and medicine), SSRN (social sciences and economics). Preprints haven't been peer-reviewed, flag them as such when you cite them.
Government reports and policy documents: Search directly on agency websites (NIH, CDC, WHO, NICE, equivalent bodies in your country). Google Scholar with a site: restriction (site:cdc.gov your search terms) can also surface reports that generic searches miss.
Conference proceedings: Often published in IEEE Xplore (engineering and computer science), ACM Digital Library (computing), or accessible via Google Scholar. Conference papers in some fields represent significant original research that may not appear in journal form for years.
When to switch from keyword search to citation network search
Database searching finds papers that use your exact search terms. It won't find papers that address the same question using different vocabulary, foundational papers that predate your search terminology, or cross-disciplinary work from fields that name your concept differently.
Once you have a core set of relevant papers from your database search, citation network search, sometimes called snowball searching or citation chasing, finds what keyword search missed.
Snowball searching works in two directions. Backward snowballing follows the reference lists of your included papers to find older foundational work. Forward snowballing finds papers that have cited your included papers since publication, surfacing more recent developments.
In ResearchRabbit, add your most relevant papers as seeds. Select References to follow connections backward to the foundational work your papers were built on. Select Citations to follow connections forward to recent papers building on your seeds. Select Similar to find papers the field treats as intellectually related, even if they use completely different terminology.
The combination of structured database search and citation network search is more comprehensive than either alone. Database search gives you precision within a vocabulary. Citation network search crosses terminology boundaries and disciplinary silos.
For a deeper explanation of how citation network search works and when to use it, see citation networks: how to find papers you'd never discover with keyword search.
Document your search as you go
A search you can't reproduce is a search you'll have to redo. For every database you search, record:
- The database name and version or date searched
- The exact search string you used
- The number of results returned
- Any filters applied (date range, language, publication type)
This documentation serves three purposes. It lets you reproduce the search if you need to update it. It lets you report your methods clearly if your literature review is for a thesis or systematic review. And it prevents you from running the same search twice.
A simple spreadsheet with one row per database search is enough for most purposes.
FAQ
How many databases should I search for a literature review?
There isn't one number that works for every literature review. The right databases depend on your research question, discipline, and type of review. For many literature reviews, combining a subject-specific database with a multidisciplinary database such as Scopus or Web of Science is a useful starting point. Systematic reviews generally require a broader search across multiple relevant databases and other sources.
Is Google Scholar enough for a literature review?
It depends on the scope and requirements of your review. Google Scholar is useful for broad orientation, citation searching, and finding preprints, theses, and grey literature. But when comprehensive or reproducible searching matters, particularly for postgraduate research and systematic reviews, it should complement rather than replace structured database searches.
What's the difference between a keyword search and a controlled vocabulary search?
Keyword search looks for your exact terms in the text of papers. Controlled vocabulary search uses standardized indexing terms, MeSH in PubMed, Emtree in Embase, CINAHL headings in CINAHL, to find all papers tagged with a concept, regardless of how individual authors expressed it. Not all databases support controlled vocabulary: Scopus and Web of Science don't have their own thesauri, so keyword searching is the primary approach there. Using controlled vocabulary alongside keywords in databases that support it gives more comprehensive results.
How do I know when I've searched enough databases?
When adding a new database consistently returns papers you've already found in other databases, you've covered the core literature. Diminishing returns are a practical signal to stop adding databases and shift to citation network search and grey literature.
Can I use AI tools to find sources?
AI tools can help generate search terms, suggest databases, and summarize papers you've already found. They're not reliable for comprehensive literature searching, they can miss papers, misrepresent findings, and can't reproduce a search strategy. Use them as a supplement to structured database search, not a substitute.
What's the difference between Scopus and Web of Science?
Both are large multidisciplinary databases, but they index different journals and use different methodologies. Scopus has broader coverage overall; Web of Science applies stricter quality criteria and has more complete citation data for some purposes. For most literature reviews, either works well. For formal bibliometric analysis, Web of Science citation data is often preferred.
%20(800%20x%201036%20px).webp)
This is a big test comment on your article.