How to search for research papers like a pro

Google searches everything at once: shops, news, opinions and research, mixed together. Academic databases search scholarly work only, and most of them let you search it far more precisely. Some cover every subject. Others go deep into one.

You do not need all of them. Start with one broad database and one for your own subject. For Sneha that meant Google Scholar to get a feel for the topic, and PubMed and CINAHL for the nursing studies that mattered.

Databases do not read sentences the way people do. They match words. A full question makes the search engine hunt for every small word in it, including "does", "to" and "before".

So the first job is to pull out the main ideas, which librarians call concepts. Sneha's question had three: music, anxiety and surgery. For each concept, list the other words authors might use for the same thing.

Two more kinds of synonym are easy to forget: abbreviations, which authors define once and then use everywhere, and older terms a field has since replaced. Including both catches papers that a quick search misses.

PubMed goes one step further. Indexers tag most of its papers with standard subject terms called MeSH, short for Medical Subject Headings, so studies that use different words for the same idea sit under one heading. Many subject databases keep a list like this. Using it is like asking the librarian for the right shelf instead of walking every aisle.

Once you have concepts and synonyms, join them with three words, usually typed in capitals. They are called Boolean operators, after the mathematician George Boole, and they need no maths at all.

Put together, Sneha's search looked like this:

Each database handles these a little differently. Google Scholar, for example, does not support the * trick and works best with short phrases, while PubMed and CINAHL handle long bracketed searches well. A minute on the help page of each database saves an hour of confusion.

This single line took Sneha from over a million web pages to a few hundred research papers, all of them about music, anxiety and surgery.

A few hundred papers is still too many to read. The filters beside the results list are how you narrow it further.

• Years. Sneha kept the last ten years, since recent studies reflect current hospital practice.
• Document type. Articles, reviews, clinical trials or theses. One good review article can save weeks of searching.
• Subject area. Removes papers from unrelated fields that happen to use the same words.
• Language and full text. Useful, but "free full text only" can hide important studies your library can still get for you.
• Sort order. Relevance is the default. Sorting by date shows the newest work; sorting by citations shows the papers others rely on most.

Every filter is a decision. Write it down, so that later you can explain why a paper is missing.

A filter you forget about can quietly remove the one study your examiner expects to see. Keep filters few, deliberate and recorded.

One good paper is a map to others. It points backwards, through its reference list, to the older work it builds on, and forwards, through the papers that cite it, to newer work.

1. Look backwards.Read the reference list of a good paper and pick the studies that match your question. This finds the older, foundational work.
2. Look forwards.Click "Cited by" in Google Scholar, or the citation count in Scopus or Web of Science, to see newer papers that built on it.
3. Repeat with the best new finds.Each good paper leads to a few more. This is called snowballing, and it often finds papers that no keyword search did.
4. Stop when the same papers keep returning.When every trail leads back to studies you already have, you have probably found the core of the field.

Sneha started from a review article that had pooled several trials on music before surgery. Its reference list gave her eleven studies in ten minutes, and its "Cited by" page gave her six more from later years.

Free tools such as Connected Papers draw these trails as a map, which makes the most connected papers in a field easy to spot. The map is a guide, not a verdict: check each paper yourself.

Trails also show where the discussion stops, which is often where a research gap begins.

Every time you run a search, write down the database, the date, the exact words you typed, the filters you used and how many results came back.

A spreadsheet works well, and so does the first page of a notebook. The format matters less than writing every search down on the day you run it.

A log lets you run the same search a month later to catch new papers. It answers your guide when they ask how you found your sources. And if you ever write a systematic literature review, it is required: every search is reported in the methods.

Deciding what to search is thinking work. Saving, sorting and formatting references is not, and free tools do it well.

• A reference manager such as Zotero or Mendeley saves a paper with one click, keeps its details, and formats your citations later. Zotero is free and open source.
• Alerts. Google Scholar and PubMed can email you when new papers match a saved search.
• Clear file names for your PDFs, such as Author_Year_ShortTitle, so you can find them months later.

Fiberarticle's Researcher does the first round of searching for you. Give it a topic and it searches arXiv, OpenAlex, Semantic Scholar and Crossref in parallel, removes duplicates by DOI and title, applies filters such as years, Scimago quartile, open access and minimum citations, and reads the open-access PDFs it can find without ever getting around a paywall. For nursing and medicine, still search PubMed and CINAHL yourself, because specialist databases know their field best.

Sneha's final log listed four databases and two search strings. From more than a million web pages she reached 23 studies worth reading closely, and she could explain exactly how she found each one. Reading them well is the next step, and how to read a research paper picks up from there.