PubMed is not like Google or other search engines. There is no underlying algorithm to predict what you're actually looking for in conjunction with your previous browsing habits and location. It's not searching the full text. PubMed is simply matching your entered terms against the bibliographic fields for each record.
To do that, PubMed tries to match the terms you entered with its own subject list (MeSH), and translates the terms into common variants (eg. child to children). This is referred to as Automatic Term Mapping (ATM). Most of the times this works as you'd expect:
It can be useful to see how your simple search was translated. To do so, click "Advanced" under the search bar:
Under "History and Search Details," click the arrow under "Details" to see what *actually* got searched.
