When a keyword search is performed, the search engine returns results that match that term or those terms in their title, abstract, author, publication, and other similar information. It does not look through the full text. We can use words that we might say in everyday speech, like "Civil War" and we will see results that match those two terms.
When a subject search is performed, the search engine returns results that match that term or those terms in their subject field. Subjects are very often classified by an expert using a controlled vocabulary, where very words are applied and used very specifically. You might see a subject heading that looks like "United States -- History -- Civil War, 1861-1865" -- there is no confusion about what we mean, but it is not exactly the way we normally speak.
What this means is that keyword searches look a lot like natural language, but we still avoid speaking in natural language or asking questions. If we have a thesis about how the US Civil War changed the American landscape, we might just pull out the key words "Civil war" and see what kinds of resources exist in the libraries. If we find too much, we might narrow down by adding another key word, like "slavery," "Lincoln," or "change."
So when do subject searches come into play? If we know the exact subject heading, we can use them to see ALL of the information on a specific topic. If we do not know the exact subject heading, we ask a subject specialist, like a librarian, or we can find books with a keyword search and use their subject headings.
Keyword searches are best if we're just starting out and want a broad overview. Subject searches are best when we want to browse through a lot of information on that particular topic.
