Searching a research database is not the same as searching the open web. A web search engine guesses what you want and ranks popular pages first. A research database searches structured records — titles, abstracts, authors, keywords and subject terms — so the words you type, and how you combine them, decide what you see. A vague query returns thousands of loosely related records, while a precise query returns a smaller set you can actually read and judge.
This guide explains a repeatable method: start from a clear question, choose the right database, build a query from keywords and synonyms, filter the results, and save what is useful. The examples use Google Scholar, PubMed and Scopus, but the method works in almost any scholarly database.
Every good search begins before you type anything: write your question in one sentence, then break it into its main concepts. A question such as "does vaccination reduce disease in dairy cattle" contains three concepts: vaccination, disease reduction, and dairy cattle. Each concept then needs its own list of keywords and synonyms, which you will combine later.
Decide what kind of answer you need: original research studies, a review summarising a field, a method to follow, or background for an introduction? An original study is what you need when you are checking a specific finding.
Google Scholar is a free search tool that covers scholarly material across many fields, including journal articles, theses, books and conference papers. Its main strength is breadth and simplicity: one search box and links to full text where a copy has been made available. For a first look at a topic, or for tracking down a known paper, it is often the fastest place to start.
Its limits matter too. Google Scholar does not publish a full list of what it covers, its ranking is not fully transparent, and the same search can return different results on different days. It also offers fewer precise filters than a specialist database, and it mixes peer-reviewed articles with theses, reports and other material. Use it as a broad starting point and a citation-tracing tool, then confirm important finds in a more structured database.
PubMed is a free database for the biomedical and life sciences, maintained by the United States National Library of Medicine. It is the natural first choice for questions about human and animal health, medicine, and related biology, because its records are indexed with a controlled vocabulary called Medical Subject Headings, or MeSH. A MeSH term finds articles on a concept even when authors used different words for it.
To use PubMed well, combine keyword searching with MeSH terms and filters. Start with your keywords, look at the MeSH terms attached to the most relevant records, and add the best ones to your query. Then filter by article type, such as review or clinical trial, and by publication date to keep the set current. PubMed records also link to freely available full text and to related articles, which is a quick way to widen a narrow search.
Scopus is a large abstract and citation database covering journals, books and conference proceedings across the sciences, medicine, social sciences and humanities. It is normally available through an institutional subscription, so check whether your library provides access. Its value for a searcher lies in structure: every record carries detailed fields for author, affiliation, source, document type and references, and every search can be built, saved and repeated exactly.
Two Scopus features are especially useful. First, advanced search lets you combine field codes, such as title, abstract and keywords, with Boolean operators, so you can refine a precise query step by step. Second, its citation tools show which later papers have cited a given article, and which references that article itself cited, letting you move forwards and backwards through a topic. Use Scopus when you need a controlled, repeatable search.
A strong query is built, not typed in one attempt. Take each concept from your question and list its synonyms: for "dairy cattle", you might add "dairy cows" and "bovine". Also consider spelling variants and abbreviations used in your field. A concept with only one word attached to it will miss every paper that expressed the same idea differently.
Then combine concepts with Boolean operators. AND narrows a search: every term joined by AND must appear, so it joins different concepts. OR widens a search: any term joined by OR may appear, so it joins synonyms. NOT excludes a term, but use it sparingly, because it can remove relevant papers. Brackets control the order, as in (dairy OR bovine) AND (vaccination OR immunisation), and quotation marks search an exact phrase. Build the query in stages, testing each concept on its own first.
Filters turn an unmanageable result list into a readable one; apply them deliberately rather than all at once. A date range keeps a fast-moving topic current, but setting it too narrowly can hide the original studies a field still builds on. A document-type filter separates reviews from original research and conference papers from journal articles, so you read the evidence your question needs. A field filter, such as title-only searching, is a powerful narrowing step.
Filter in stages and watch the result count at each step: apply one filter, check relevance, and only then apply the next. If a filter removes a paper you know should be there, undo it and ask why: a title-only search misses papers that mention your term only in the abstract. Keep a note of the filters you used, because a search you cannot describe cannot be repeated or reported honestly.
Do not open every result. Scan titles first, set aside anything clearly off-topic, then read the remaining abstracts and judge each against your question: does it study the same population or phenomenon, and report the outcome you need? An abstract that is vague about methods or results tells you the full text will need close reading before you rely on the paper. Shortlist a manageable number of papers for full reading, rather than collecting everything vaguely relevant.
Then follow citation trails, which often find the best papers faster than new keyword searches. Backward searching means reading a strong paper's reference list for earlier work; forward searching means finding later papers that cited it, showing how the finding was tested or challenged. Both trails are available through the "Cited by" link in Google Scholar and the citation tools in Scopus. A trail that keeps returning to the same few papers shows you have reached the core literature.
A search is only as good as the record you keep of it. Save the full query, the database used, the filters applied, and the date of the search, so you can repeat it later or explain it in your own writing. Most databases let you save searches and set up alerts, which send you new matching papers as they appear. Export the records you keep to a reference manager, where you can add notes and tags.
Organise by question, not by database: A single folder per research question, with sub-groups for background, key evidence and methods papers, is far easier to use than separate piles from each database. As you read, write one or two sentences in your own words about what each paper contributes and what its limits are. That short note turns a saved list of links into material you can actually write from weeks later.
Not everything a database returns deserves equal trust, and a search tool does not vouch for the quality of every record it lists. Check each source before you rely on it: is the editorial board named, is the peer review policy stated clearly, and are publication charges transparent? Be cautious of journals that promise unusually fast publication, hide charges until after submission, or make claims you cannot verify against an official source.
Also check whether a paper has been corrected or retracted. A retraction notice means the paper should not be relied on, and citing one unknowingly is a real risk. Many databases flag retractions on the record itself, and a quick check of the journal's site takes only a minute. The same caution applies to preprints and reports, which have not been peer-reviewed: label them as such and never present them as peer-reviewed evidence.
No single database covers everything, and searching only one is the easiest way to miss relevant work. Different databases cover different journals and publication types, and each has its own subject vocabulary. A biomedical question may need PubMed plus a broader tool, while a social science or interdisciplinary question may need a general database plus a specialist one for the field. Searching a second database with the same concepts, adapted to its vocabulary, often turns up papers the first search missed.
Beyond the major databases, consider the free tools that complement them. Semantic Scholar offers a broad, freely searchable index with useful features for finding related papers and checking citation contexts. Your library's subject guides often point to specialist databases for your field. For grey literature — reports, theses and conference papers — a check of institutional repositories can fill gaps journal databases leave.
Suppose your question is whether vaccination reduces disease in dairy cattle. Its concepts are vaccination, disease reduction, and dairy cattle. Synonyms for each: vaccination, immunisation, vaccine; disease, morbidity, infection; dairy cattle, dairy cows, bovine. Your core query then becomes (vaccination OR immunisation OR vaccine) AND (disease OR morbidity OR infection) AND (dairy cattle OR dairy cows OR bovine), adapted for each database rather than pasted unchanged.
In Google Scholar, run the query broadly first, scan the top results, and follow the "Cited by" links from the strongest papers to find later work. In PubMed, add MeSH terms, filter by article type and date, and use related-articles links to widen a narrow set. In Scopus, build the same query with field codes for title, abstract and keywords, and use citation tools to move forwards and backwards. Comparing the three result sets shows what each database adds, and their overlap confirms you have found the core literature.
Before you finish a search, run through a short checklist. Have you written the question down, broken it into concepts with synonyms for each, and searched more than one database? Have you applied filters in stages, checked that a known relevant paper still appears, and recorded the query and filters? Have you followed citation trails and saved useful records with your own notes? Any gap here is where your next improvement will come from.
Finally, judge the search by its results, not its effort. A good search returns a set you can read and defend: the key papers are present, irrelevant material is mostly filtered out, and you can explain how you found them. If the set is too large, narrow a concept or add a filter; if too small, widen the synonyms or drop a filter. Searching is iterative: revise the query several times rather than accepting the first result list you see.
| Database | Best for | Access note |
|---|---|---|
| Google Scholar | Broad first searches across fields | Free to search |
| PubMed | Biomedical and life sciences literature | Free to search |
| Scopus | Structured, repeatable searches and citation tracking | Usually via an institutional subscription |
| Web of Science | Citation searching across disciplines | Usually via an institutional subscription |
| Semantic Scholar | Finding related papers and citation context | Free to search |
These tools are the databases and search services mentioned in this guide:
Which database should I search first?
Start with the database that fits your question: PubMed for biomedical topics, a broad tool such as Google Scholar for a first look across fields, and a structured database such as Scopus when you need a controlled, repeatable search. For most questions, searching more than one database gives a more complete result than any single search.
Why does my search return too many results?
Usually because a concept is too broad, a synonym list is too long, or no filters have been applied. Narrow one concept, add a date or document-type filter, or search within titles only, and check the result count after each change rather than applying every filter at once.
Why does my search return almost nothing?
Check for spelling errors, overly narrow filters, and missing synonyms. Try searching each concept on its own to find which one is empty, then widen that concept with related terms or a subject heading from a relevant record before combining the concepts again.
What are Boolean operators, in simple terms?
AND requires every joined term to appear, so it narrows a search between different concepts. OR allows any joined term to appear, so it widens a search between synonyms. NOT excludes a term, but it should be used sparingly because it can remove relevant papers that mention the term in passing.
How do I know when I have searched enough?
You have searched enough when new searches and citation trails keep returning papers you already have, the key studies on your question are all present, and you can describe your query, databases and filters clearly. At that point, further searching usually adds duplicates rather than new evidence.
Can I rely on a paper just because a database lists it?
No. A database lists records; it does not endorse every paper it returns. Check the journal's peer review policy and editorial board, look for corrections or retractions, and read the methods before relying on any paper in your own work.