LibGuides: Database Search Tips: Keywords and Subject Headings

Overview

To find subject headings for your topic:

Look to see if the database has an online thesaurus to browse for subjects that match your topic. PubMed does offer this ability in its database.

Another way to find subject headings:

Start with a keyword search, using words/phrases that describe your topic.
Browse the results; choose 2 or 3 that are relevant.
Look at the Subject or Descriptor field and note the terms used (write them down).
Redo your search using those terms.
Your results will be more precise than your initial keyword search.

What are keywords and subject headings?

Subject headings describe the content of each item in a database. Use these headings to find relevant items on the same topic. Searching by subject headings (a.k.a. descriptors) is the most precise way to search article databases.

It is not easy to guess which subject headings are used in a given database. For example, the phone book's Yellow Pages use subject headings. If you look for "Movie Theatres" you will find nothing, as they are listed under the subject heading "Theatres - Movies."

Keyword searching is how you typically search web search engines. Think of important words or phrases and type them in to get results.

Here are some key points about each type of search:

Keywords	vs.	Subjects
natural language words describing your topic - good to start with		pre-defined "controlled vocabulary" words used to describe the content of each item (book, journal article) in a database
more flexible to search by - can combine together in many ways		less flexible to search by - need to know the exact controlled vocabulary term
database looks for keywords anywhere in the record - not necessarily connected together		database looks for subjects only in the subject heading or descriptor field, where the most relevant words appear
may yield too many or too few results		if too many results - also uses subheadings to focus on one aspect of the broader subject
may yield many irrelevant results		results usually very relevant to the topic