01What it is matching, and what it is not.
A web search engine reads your phrase, infers intent, corrects spelling, expands synonyms and ranks by a mixture of relevance and popularity. It is doing a great deal of interpretive work on your behalf and it is doing it invisibly.
A database does almost none of that. In the ordinary case it matches your terms against specific fields of a record: title, abstract, author, subject headings, sometimes full text. It matches them literally. If your word does not appear in the field being searched, the record does not come back, no matter how relevant it would have been.
That single difference explains nearly every frustrating result. Nothing found does not mean nothing exists; it usually means the field you searched does not contain the word you used. Which is a solvable problem rather than a dead end.
02The four operators worth knowing.
Almost every database supports these and almost no student uses them, which is a shame because the whole return on learning them arrives inside ten minutes.
- AND narrows. Two concepts joined by AND returns only records containing both, which is what you want when a search is returning thousands of things.
- OR widens, and it is the one that rescues empty searches, because it is how you catch the several words a field uses for one idea.
- NOT excludes, and it is the most dangerous of the four. Excluding a term also excludes anything that merely mentions it, so it removes useful records surprisingly often.
- Quotation marks hold a phrase together, so a two-word concept is searched as one thing rather than as two words appearing anywhere in the record.
- And a truncation symbol, usually an asterisk, catches a word's variants at once. One search instead of four, which matters more than it sounds.
03Subject headings are the actual shortcut.
This is the part almost nobody is shown. Many databases have somebody assign each record a set of controlled subject terms from a fixed vocabulary. Those terms are the same regardless of what words the authors happened to use, which means searching by subject heading catches papers your keyword search would miss entirely.
The practical route in is simple. Run any ordinary keyword search, open the single most relevant thing that comes back, and look at the subject terms listed on its record. Those are the field's own words for your topic. Search those instead and the results change character completely.
That trick also solves the problem of not knowing a field's vocabulary yet, which is its own issue at searching before you know the words. You do not need to know the terminology in advance. You need one relevant record, and the record will tell you.
04Why your search returned nothing.
In rough order of frequency. Too many concepts joined with AND, so the search is asking for a paper that does everything at once. A phrase in quotation marks that nobody actually writes that way. A word the field does not use, which is the commonest of all and which subject headings fix.
Then the mechanical ones: searching full text when the database only indexes abstracts, or the reverse. A date limit narrower than the literature. A filter left on from a previous search, which is responsible for more baffling empty screens than any conceptual error.
The habit that fixes most of it is to start deliberately too wide and narrow afterwards. Two concepts, no filters, see the volume, then add. Starting narrow and widening is much harder, because an empty result tells you nothing about which of your five constraints was the problem.
05The person whose job this is.
Every institution with a library has somebody employed to do exactly this, and online students use them at a fraction of the rate campus students do, usually because they assume the service is physical.
It is not. Most run appointments by video or email, most will take a topic and demonstrate a search strategy on it, and most consider this the most interesting part of their week rather than an imposition. A thirty-minute conversation with one of them is worth more than any article about searching, this one included.
Where the difficulty is reaching a specific paper rather than finding it, that is a different problem with its own routes, at reaching a paywalled paper.
Questions at this desk.
Which database should I be using?
Whichever your field uses, and your library's subject guides will name them. Searching a general database for a specialised topic returns thin results and convinces people the literature is thin, when what is thin is the coverage. If in doubt, ask a librarian which two are standard in your area.
Is a general web search ever appropriate?
For orientation, yes, and it is genuinely useful for learning the vocabulary before a proper search. What it is not is a substitute, because it does not index most of the literature and it cannot filter by the things that matter, like study design or peer-reviewed status.
Why do I keep getting thousands of results?
Usually because the concepts are too broad, or because a two-word idea is being searched as two separate words. Put phrases in quotation marks, join distinct concepts with AND, and add one filter at a time so you can see which one actually did the narrowing.
How many searches should a paper take?
More than one, and that is the useful expectation to carry. A single search is a first attempt at guessing a field's vocabulary. Three or four, each informed by the subject headings on what came back, is a search strategy, and the difference in what surfaces is substantial.
Notes in the margin, from the people at the desks.
READ THE LETTER