When you have a topic idea that you think might work as a dissertation topic, it is time to start thinking about it in terms of existing research. Often your topic will be broad and may be broken down into component topics to investigate as part of the literature review. For example, a topic area of Native American student graduation rates could be looked at in terms of characteristics of Native American students, factors that contribute to academic program drop-out, factors that contribute to persistence in academic programs, and a combination of characteristics of Native American students and drop-out or persistence in academic programs. There may also be other terms that relate to your topic that could be added to your search strategy. For our example, "student success" is another term related to academic persistence that might lead to additional relevant materials.
When you have a component topic to look for, the next step is to think about keywords to use for locating information on your topic. Try to look at your topic from as many angles as you can noting what keywords might be useful in searches. Sticking with the same example topic you can easily see the importance of thinking about keywords to search with. "Native American" is not universal in it's usage. "American Indian," "Indian," "Indians of North America," or the name of a specific tribe of Native Americans could all be used as keywords leading to relevant materials. So which one is best to use? The answer is, "it depends." If you are looking for a book in a university library's catalog, the current Library of Congress Subject Heading is "Indians of North America." Doing a subject (rather than keyword) search for Indians of North America should produce the most complete results list. Doing a keyword search with any of the terms should produce a smaller number of results because different terms would have been used to catalog different sources based on the actual topics, based on changes in standard cataloging practices over time, and based on changes in the people that did the cataloging over time.
Databases are a different story. They usually use their own proprietary classification scheme rather than a standard one like the Library of Congress Classifications. What that means is each database typically has it's own list of terms used to catalog/classify items in the database into categories. Think of an index at the back of a book. In fact, most of what libraries now call databases are actually indexes that index the contents of journals or other sources that particular database has access to. Full-text databases are indexes that also contain full-text articles. Thinking back to our different keyword possibilities you can rely on trial and error to see which one works better, or if the database has a thesaurus function, that function actually checks to see if a term is on the list/index used by that database, and may recommend the correct term if not.